# Drugnews｜藥時事 — full text for AI search and agents Each block: title, canonical URL, date, language, summary, key points, then the article body without images. Per-article Markdown and JSON twins sit next to every article URL (.md / .json). Index: https://drugnews.com.tw/llms.txt ## Celltrion's Record Profit Starts in Its Own Factories URL: https://drugnews.com.tw/articles/2026-10-15-sandoz-celltrion-biosimilar-profit-capacity-en.html Date: 2026-10-15 Language: en Summary: Celltrion's third-quarter operating profit exceeded KRW 500 billion for the first time, with a margin of about 34%, powered by its own factories. Sandoz is also turning its European plants into a network capable of accommodating more than 100 medicines. The next round in biosimilars is about whose new products sell fast enough to fill the new capacity. KRW 510 billion. That is the preliminary third-quarter operating profit Celltrion announced on October 12, the first time it has crossed KRW 500 billion since the company was founded. Revenue reached KRW 1.5 trillion, up 45.8% year on year. Operating profit rose 69.2%, putting the operating margin at about 34%. Profit grew faster than revenue. The company gave two reasons: it sold more high-margin products, and its factories made those medicines more cheaply. Our view: after a decade of price competition in biosimilars, the money is flowing toward companies that make the medicines themselves. Switzerland's Sandoz has reached the same conclusion and is turning its European factories into a network capable of accommodating more than 100 medicines. But factories require money upfront. The next contest is about whose new products sell fast enough to fill the new capacity. ## 01｜From Making Medicines for Others to Choosing Its Own Celltrion started out making medicines for other companies. In 2005, it signed a contract with Bristol Myers Squibb to manufacture the rheumatoid arthritis medicine Orencia at its factory in Songdo, Incheon. Contract manufacturing taught the Korean company a difficult first lesson: how to build a large-scale antibody plant and pass a U.S. FDA inspection. Later, it began using its own factories to make its own biosimilars. Today, Songdo Plants 1 through 3 have a combined drug-substance capacity of 250,000 liters. On September 22, all three completed EU GMP renewal. Plant 3, which began commercial production only in late 2024, is expected to produce 128 batches this year, about 142% more than last year. Output across Songdo is expected to rise by about 15%. Those two figures belong together. The new plant is ramping up from a low base, which explains its steep growth rate. The increase in medicines available to sell across the site is on the scale of that 15% figure. Most factory costs are fixed. When the same buildings and quality-control teams produce more batches that meet quality standards, the cost allocated to each batch falls. Some products are also moving to higher-yield processes, allowing the same-sized bioreactor to produce more antibody. That is the arithmetic behind 34%. *Figure 1｜The route from capacity to profit: output that meets quality standards, drug-product release, and actual sales. If the medicines do not sell, the same factory becomes a cost burden.* Next, Celltrion plans to bring the final drug-product manufacturing stage in-house as well, reducing its reliance on outsourcing. When it announced the update in September, the new drug-product plant was still undergoing equipment qualification. ## 02｜Sandoz's Challenge: 100 Medicines Cannot All Fit the Same Factory Sandoz's story starts at the other end. In October 2023, it separated from Novartis and became an independently listed company. Generics still account for about two-thirds of its revenue. In the first half of this year, biosimilar sales reached USD 1.875 billion, roughly one-third of the total. At its September 8 Capital Markets Day, Sandoz unveiled “Bio100”: expanding its biosimilar portfolio from 13 at the time to more than 100 by 2040. From 2035 onward, it aims to cover about 80% of the value of originator medicines losing patent protection. The difficulty is that sales volumes vary widely across 100 medicines. High-volume products need large tanks running for long periods. For low-volume products, using a large tank can be wasteful. Sandoz's answer is to run three manufacturing approaches side by side: - Lendava, Slovenia: a high-volume drug-substance center opened on September 29. - Toulouse, France: continuous manufacturing. - Ljubljana, Slovenia: an approximately USD 300 million, 8,000-liter single-use manufacturing facility, expected to become operational in 2029, serving clinical-stage and low- to medium-volume products. The difference in one sentence: Celltrion is filling its existing factories, while Sandoz is building a network that can accommodate a wider range of medicines. | | Celltrion | Sandoz | |---|---|---| | Latest results | Third-quarter operating profit of KRW 510 billion and an operating margin of about 34% (preliminary) | First-half biosimilar revenue of USD 1.875 billion, about 33% of group revenue | | Manufacturing strategy | Raise yields at Songdo's 250,000-liter site and bring drug-product manufacturing in-house | Three manufacturing approaches, with another 8,000 liters coming in 2029 | | Product goal | Scale up high-margin new products | Expand from 13 biosimilars to more than 100 by 2040 | | Next test | Can Omlyclo scale up in the United States? | Can new products fill the new network? | *Figure 2｜Both companies are bringing manufacturing under their own control: Sandoz is expanding across different production approaches, while Celltrion is raising output at existing factories and bringing drug-product manufacturing in-house.* ## 03｜The Factories Are Built. Who Will Pay? Capacity is only half the story. The other half is whether hospitals and insurers will use the products. Celltrion launched Omlyclo in the United States in early October. It is a biosimilar to Xolair, a medicine for allergies and asthma, and currently the only version with FDA interchangeability status. Priced about 15% below the originator's wholesale price, it entered the market with preferred commercial-insurance formulary placement at Optum, one of the three largest U.S. pharmacy benefit managers. Europe has already provided an answer. Omlyclo is sold in 26 countries and had captured more than 90% of the Spanish market by July. After winning Denmark's national supply contract, the company expects its market share there to exceed 98% by year-end. Sandoz has a model of its own. Its North American biosimilar sales grew 47% at constant exchange rates in the first half, driven mainly by Wyost and Jubbonti, two denosumab biosimilars launched in June 2025. The sequence in this contest is clear: build the factories, secure coverage, then bring in prescriptions. Fourth-quarter U.S. prescription volume for Omlyclo will provide the first answer to whether there is demand for Celltrion's new capacity. *Figure 3｜Factory expansion, product launches, and coverage happen at different times. Sandoz's additional 8,000-liter facility will not become operational until 2029.* ## 04｜34% and 20.9% Cannot Be Used to Rank the Companies It is tempting to compare Celltrion's 34% with Sandoz's first-half core EBITDA margin of 20.9%. The two figures use different definitions: one is a preliminary operating margin for a single quarter; the other covers six months, includes company-defined adjustments, and includes the generics business. What matters is each company's own trend. Will Celltrion's finalized financial results confirm its preliminary figure? As biosimilars account for a larger share of Sandoz's business, can its margin move toward the company's target of more than 30% in 2035? *Figure 4｜Celltrion's 34% is its preliminary consolidated operating margin for the third quarter of 2026. Sandoz's 20.9% is its group core EBITDA margin for the first half of 2026.* ## Conclusion｜The Next Round Is About How Quickly Factories Fill Up Ten years ago, biosimilar competition was about who secured approval first and who cut prices the most. Celltrion's record this quarter shows that the rules have changed: companies that can make medicines more cheaply and sell them are the ones that keep the profit. Watch two figures next: Omlyclo's fourth-quarter U.S. prescription volume, and whether Sandoz's new products can sustain the 47% North American growth achieved in the first half. Whichever company fills its additional capacity first will be the first to pocket the money from this round. *This article is an industry and company analysis and does not constitute investment or medical advice.* --- ## Celltrion 獲利破紀錄，贏在自己的工廠 URL: https://drugnews.com.tw/articles/2026-10-15-sandoz-celltrion-biosimilar-profit-capacity.html Date: 2026-10-15 Language: zh-Hant Summary: Celltrion 第三季營業利益首破 5,000 億韓元，利益率約 34%，靠的是自己的工廠。Sandoz 也在把歐洲工廠改成能裝下上百種藥的網路。生物相似藥的下一局，比的是誰的新藥賣得夠快、填得滿新產能。 5,100 億韓元。這是 Celltrion 10 月 12 日公布的第三季暫定營業利益，公司成立以來第一次突破 5,000 億韓元。 營收 1.5 兆韓元，年增 45.8%；營業利益年增 69.2%，利益率約 34%。 利益長得比營收快。公司給的理由只有兩個：高毛利的產品賣得更多，工廠把藥做得更便宜。 我們的判斷是：生物相似藥打了十年降價戰，錢正在流向「自己把藥做出來」的公司。瑞士的 Sandoz 也看懂了，正把歐洲工廠改造成能裝下上百種藥的網路。但工廠是先付錢的生意，接下來比的是誰的新藥賣得夠快，填得滿新產能。 ## 01｜從替人代工，到自己決定做什麼藥 Celltrion 起家時，是替別人做藥。 2005 年，它和必治妥施貴寶簽約，在仁川松島的工廠替對方生產類風濕關節炎藥 Orencia。代工讓這家韓國公司先學會一件難事：把大型抗體工廠蓋起來，並通過美國 FDA 的檢查。後來，它開始用自己的工廠，做自己的生物相似藥。 今天，松島一到三廠合計有 25 萬公升的原料藥產能。9 月 22 日，三座廠一起完成歐盟 GMP 更新。2024 年底才開始商業生產的三廠，今年預計做 128 批，比去年多約 142%；整個松島的產量則預計增加約 15%。 這兩個數字要一起看。新廠從低點爬升，成長率才會這麼高；真正多出來、可以拿去賣的藥，是 15% 這個量級。 工廠的成本大多是固定的。同一批廠房和品管團隊，做出更多批合格的藥，每一批分攤的成本就下降。再加上部分產品換成產率更高的製程，同樣大小的培養槽能收成更多抗體。34% 背後，就是這套算術。 *圖1｜產能變成利潤的路：合格產出、製劑放行、實際銷售。藥賣不動時，同一座工廠就從資產變成成本。* 下一步，Celltrion 要把最後的「製劑」這一段也收回來自己做，少靠外包。9 月公告時，新的製劑廠還在做設備驗證。 ## 02｜Sandoz 的難題：100 種藥，塞不進同一種工廠 Sandoz 的故事從另一頭開始。 2023 年 10 月，它從諾華分家，獨立上市。到今天，學名藥仍占它營收約三分之二；今年上半年，生物相似藥賣了 18.75 億美元，約占三分之一。 9 月 8 日的投資人日，Sandoz 喊出「Bio100」：生物相似藥要從當時的 13 種，到 2040 年擴大到超過 100 種；從 2035 年起，布局要涵蓋專利到期原廠藥約八成的價值。 難處在於，100 種藥的銷量差很多。賣得多的，需要大槽長時間開；賣得少的，開大槽反而浪費。Sandoz 的答案是三種工廠並行： - 斯洛維尼亞 Lendava：9 月 29 日啟用的高產量原料藥中心。 - 法國 Toulouse：連續製造。 - 斯洛維尼亞 Ljubljana：投資約 3 億美元、8,000 公升的一次性生產設施，預計 2029 年運作，接臨床和中低銷量的產品。 一句話講兩家的差別：Celltrion 在把現有工廠做滿，Sandoz 在織一張能裝下更多種藥的網。 | | Celltrion | Sandoz | |---|---|---| | 最新成績 | 第三季營業利益 5,100 億韓元，利益率約 34%（暫定） | 上半年生物相似藥營收 18.75 億美元，占集團約 33% | | 製造策略 | 松島 25 萬公升，提高產率、把製劑收回自製 | 三種工廠並行，2029 年再加 8,000 公升 | | 產品目標 | 讓高毛利新品放量 | 13 種到 2040 年超過 100 種 | | 下一個證明點 | 美國 Omlyclo 能不能放量 | 新品能不能填滿新網路 | *圖2｜兩家都在把製造收回自己手上：Sandoz 擴充多種製造方式，Celltrion 提高既有工廠的產出與製劑自製。* ## 03｜工廠蓋好了，誰來買單？ 產能只是一半。另一半，是醫院和保險願不願意用。 Celltrion 10 月初在美國推出 Omlyclo，這是過敏與氣喘藥 Xolair 的生物相似藥，也是目前唯一拿到 FDA「可互換」資格的版本。定價比原廠批發價低約 15%，上市時就被美國三大藥品福利管理公司之一的 Optum，列為商業保險的優先用藥。 歐洲已經先給了答案。Omlyclo 在 26 國銷售，7 月時拿下西班牙九成以上的市占；在丹麥贏得全國供貨合約後，公司預計年底市占超過 98%。 Sandoz 也有自己的樣板。上半年北美生物相似藥按固定匯率成長 47%，主要靠 2025 年 6 月上市的兩款 denosumab 骨鬆藥 Wyost 與 Jubbonti。 所以這場比賽的順序是：工廠先蓋好，給付名單打開，處方才會進來。Omlyclo 在美國第四季的處方量，就是 Celltrion 新產能有沒有人接的第一個答案。 *圖3｜擴廠、上市、給付各在不同時間點發生；Sandoz 新增的 8,000 公升設施要到 2029 年才運作。* ## 04｜34% 和 20.9%，不能拿來排名 看到 Celltrion 的 34%，很容易拿去和 Sandoz 上半年 20.9% 的核心 EBITDA 利潤率比高低。這兩個數字口徑不同：一個是單季、暫定的營業利益率；一個是半年、經公司調整、而且包含學名藥的指標。 真正該看的是各自的趨勢。Celltrion 的正式財報會不會確認這個暫定數字；Sandoz 生物相似藥比重上升時，利潤率能不能往它設定的 2035 年「30% 以上」靠近。 *圖4｜Celltrion 的 34% 是 2026 年第三季暫定合併營業利益率；Sandoz 的 20.9% 是 2026 年上半年集團核心 EBITDA 利潤率。* ## 結語｜下一局，比的是填滿工廠的速度 十年前，生物相似藥比的是誰先拿到核准、誰降價降得多。Celltrion 這一季的紀錄說明，規則已經換了：能把藥做得更便宜、又賣得出去的公司，才留得住利潤。 接下來盯兩個數字：Omlyclo 在美國的第四季處方量，以及 Sandoz 新品能不能延續上半年北美 47% 的成長。哪一家先把新增產能填滿，哪一家就先把這一局的錢放進口袋。 *本文為產業與公司分析，不構成投資或醫療建議。* --- ## The Two-Drug Bet in Weight-Loss Injections URL: https://drugnews.com.tw/articles/2026-10-14-eloratzp-weight-loss-tolerability-persistence-en.html Date: 2026-10-14 Language: en Summary: In a Phase 2 trial, Lilly added eloralintide to tirzepatide, gaining additional weight-loss potential but also facing higher rates of treatment discontinuation due to adverse events. Whether Phase 3 dose escalation and co-formulation can turn that efficacy into a product upgrade patients are willing to continue remains to be demonstrated. The comparator Lilly chose this time was its own tirzepatide. In Phase 2 results announced on September 30, it added investigational eloralintide to tirzepatide to create EloraTZP. Among adults with type 2 diabetes and obesity or overweight, the highest-dose combination produced an estimated mean weight reduction of 23.3% at 48 weeks, versus 14.8% with tirzepatide alone in the same trial: a difference of 8.5 percentage points. These figures assume continued treatment. That increment gives Lilly a reason to keep investing. A challenge emerged alongside it: a higher proportion of participants in the combination groups discontinued treatment because of adverse events than with monotherapy. Adding a second molecule pushed the weight-loss curve down, but the treatment burden up. The company still plans to advance into Phase 3 before the end of 2026, improve dose escalation, and develop a co-formulation of the two drugs. What determines the outcome of this two-drug bet next is how much additional efficacy it can retain, and how many people it can keep on treatment. ## 01 | A New Partner Must First Beat the Existing In-House Option Tirzepatide is already the active ingredient in Mounjaro and Zepbound. Lilly has an existing option, so the new combination must answer a harder question: why would patients need an additional molecule? Eloralintide offers another route. Tirzepatide acts on GIP and GLP-1 receptors, which participate in regulating appetite and blood glucose; eloralintide targets amylin receptors, helping generate satiety and reduce food intake through another signal involved in regulating eating. This involves two molecules and three receptor pathways, rather than simply increasing the tirzepatide dose further. Lilly is also studying eloralintide alone in Phase 3. The company is testing both whether the new molecule can become a product in its own right and how much it can add to an existing drug. From a portfolio perspective, the combination's appeal is an additional option for people who need a greater improvement. Human data must show whether it can deliver that. This 367-participant, 48-week study provided the first strong reason to pursue that route. Eloralintide 9 mg plus tirzepatide 15 mg produced an estimated mean weight reduction 8.5 percentage points greater than tirzepatide 15 mg alone. Including a monotherapy group in the same trial lets readers see what adding the new molecule actually contributes. The primary comparison was the combination versus placebo; comparisons with treatments such as tirzepatide were secondary endpoints. The difference therefore supports further development, but does not establish the combination as the best of all weight-loss drugs. *Figure 1 | Mean weight reduction at week 48 in the same Phase 2 trial, in participants with type 2 diabetes and obesity or overweight. The 9 + 15 mg labels refer to the separate doses of the two drugs, which were injected separately in this trial; the pen shapes are illustrative only. The 23.3% and 14.8% figures are estimates under the assumption of continued treatment, a difference of 8.5 percentage points.* ## 02 | Another Column of Numbers Adds a Cost to the Upgrade The efficacy story is easy to tell. The discontinuation story is more complicated. Treatment discontinuation due to adverse events ranged from 10.8% to 27.0% across the combination groups, versus 2.9% with tirzepatide and 16.7% with placebo. These are discontinuations due to adverse events, not all-cause dropout rates. The announcement did not map that range to individual groups, so 27.0% cannot be assigned directly to the group achieving 23.3% weight reduction. Groups were limited in size and the placebo rate was also substantial; interpretation still requires the specific reasons. This gap directly affects product choices. Lilly needs to find a pairing that retains efficacy worth pursuing without making the process of obtaining it too difficult. The highest weight reduction and a range of discontinuation rates across groups are not enough to identify that balance. The company said gastrointestinal adverse events were most common, generally mild to moderate, concentrated mainly during dose escalation, and more frequent with the combination than with monotherapy. Mild to moderate describes severity. Whether those events interrupt an ongoing treatment is a separate question that must be answered. That is also why the method behind the 23.3% estimate deserves attention. It asks: assuming all randomized participants continued receiving study treatment, what effect could be achieved at 48 weeks? The framework allows possible interruptions or dose adjustments; it does not simply remove those who stopped treatment and count only completers. It reveals the drug's weight-loss potential, which is still a different question from the average result in everyday use for everyone who starts treatment. *Figure 2 | The 23.3% figure is estimated under the assumption that participants continued receiving the study treatment. The dashed line illustrates a statistical assumption, not an observed patient trajectory; the efficacy estimand is also not equivalent to calculating results only for completers.* The earlier SURMOUNT-4 study provides context for why continued treatment matters. Adults with obesity or overweight but without diabetes received tirzepatide for 36 weeks, then were assigned to continue treatment or switch to placebo. Over the following 52 weeks, those switched to placebo regained weight on average, while those continuing treatment lost further weight. These are not EloraTZP withdrawal results, but they clarify the long-term problem facing developers: after weight is lost, treatment and its results still need to be sustained. ## 03 | Another Pathway Adds Another Product-Development Test Receptor pathways may complement one another, but patients experience the entire treatment. EloraTZP has shown additional weight-loss potential. The next step is to put that potential into a regimen people can more readily continue. *Figure 3 | Eloralintide acts on amylin receptors, while tirzepatide acts on GIP and GLP-1 receptors. This is an illustration of the pathways of action, not a complete diagram of signaling across organs. The value of using the two molecules together must still be assessed in human trials.* Lilly's proposed optimization of dose escalation responds to the observation that gastrointestinal problems occurred mainly while doses were increasing. Starting intensity, escalation speed, and target dose are all adjustments that can be tested. The next question is not just which combination yields the greatest weight reduction, but how much additional efficacy remains after adjustment. Two possible directions are worth tracking. If the burden is concentrated in a particular combination or escalation stage, the next phase can test whether targeted changes improve continued treatment. If different pairings carry similar burdens, developers must weigh more carefully how much additional weight reduction justifies that discomfort. Group-specific reasons for discontinuation, its timing, and outcomes captured after interruptions will gradually distinguish these paths. More complete data therefore do more than add tables to the announcement. They will influence which formulation Lilly advances, and whom the new product has the strongest reason to serve. *Figure 4 | Dose escalation, tolerability, and continued follow-up need to be designed together. The optimized regimen remains to be validated. The staircase does not represent actual doses or an adjustment schedule and is not personal medication guidance.* ## 04 | A Co-Formulation Still Has to Clear the Tolerability Hurdle The two drugs were injected separately in this Phase 2 trial. Lilly's planned co-formulation may simplify administration. The two development efforts have their own questions to answer: can formulation integration make injections more convenient, and can dose-escalation changes improve tolerability while preserving additional efficacy? Considering them together shows how much overall improvement patients gain. This also determines how the new combination could coexist with existing tirzepatide. Monotherapy already works. The two-drug approach must offer people needing a greater improvement a trade-off worth choosing. Pricing and payment arrangements for the new combination have not been announced. The most relevant commercial question now is which patients Lilly can offer a more compelling treatment option. ## Conclusion | Worth Advancing, with Trade-Offs the Next Test These Phase 2 results warrant Lilly's continued development. There is evidence of additional efficacy, and the cost that must be addressed is also visible. The most meaningful next development will be how much benefit remains after adjusting dose escalation, and how treatment discontinuation due to adverse events changes. If Lilly can do both well, the second molecule will have a better chance of turning additional efficacy into a new option that physicians and patients have reason to choose. *This article is an analysis of the industry and company. It does not constitute investment or medical advice.* --- ## 瘦瘦針的雙藥賭注 URL: https://drugnews.com.tw/articles/2026-10-14-eloratzp-weight-loss-tolerability-persistence.html Date: 2026-10-14 Language: zh-Hant Summary: 禮來在第二期試驗把eloralintide加入tirzepatide，換到額外減重潛力，也遇到較高的因不良事件停藥比例。第三期的加量節奏與共製劑，能否讓這段藥效成為病人願意持續接受的產品升級，仍待驗證。 禮來這次拿來比較的對手，正是自家的tirzepatide。 9月30日公布的第二期試驗裡，它把研發中的eloralintide加到tirzepatide上，組成EloraTZP。在有第二型糖尿病、肥胖或過重的成人中，最高劑量組合的48週平均減重估計達23.3%，同場tirzepatide單藥為14.8%，相差8.5個百分點。這組數字採持續治療假設。 這個增量值得禮來繼續投入。難題也跟著出現：組合組因不良事件停藥的比例，比單藥高。加進第二個分子，減重曲線往下走，治療的負擔卻往上走。 公司仍計畫在2026年底前推進第三期，改良加量節奏，並開發兩藥共製劑。這場雙藥賭注接下來的勝負，就看它能保住多少額外藥效，又能讓多少人繼續治療。 ## 01｜新搭檔先得勝過自家的現成選項 tirzepatide已是Mounjaro與Zepbound的有效成分。禮來手上有現成選項，新組合要成立，就得回答一個更難的問題：病人為什麼需要多加一個分子？ eloralintide提供了另一條路。tirzepatide作用於GIP與GLP-1受體，參與食慾與血糖調節；eloralintide則瞄準胰澱素受體，透過另一條參與進食調節的訊號，幫助產生飽足感、減少攝食。這是兩個分子、三條受體路徑，並非單純把tirzepatide的劑量再往上加。 禮來也讓eloralintide以單藥進行第三期試驗。新分子能否獨立成為產品、能替既有藥增加多少效果，公司兩條路一起試。對產品布局而言，組合的吸引力，是為需要更大改善的人增加一個選項；能否做到，要由人體資料回答。 這次367人、48週的研究，給了這條路第一個有力理由。eloralintide 9毫克加tirzepatide 15毫克，相較tirzepatide 15毫克單藥，多出8.5個百分點的平均減重估計。同一試驗設有單藥組，讓讀者能看到加上新分子究竟增加了什麼。 研究的主要比較是組合對安慰劑，與tirzepatide等治療的比較屬次要終點。因此，這個差距支持繼續開發，還不是把新組合排成所有減重藥第一名的依據。 *圖1｜同一第二期試驗、第48週的平均體重減幅；對象有第二型糖尿病及肥胖或過重。9＋15毫克是兩藥各自的劑量，本次分開注射，圖中的筆形僅為示意。23.3%與14.8%是持續治療假設下的估計，差距為8.5個百分點。* ## 02｜另一欄數字，讓升級多了一筆代價 效果的故事很好講，停藥的故事複雜得多。 不同組合組因不良事件停藥的比例為10.8%至27.0%，tirzepatide組為2.9%，安慰劑組為16.7%。這是因不良事件停止治療，不是所有原因的退出率。公告尚未逐組對應這個範圍，不能把27.0%直接配給減重23.3%的那一組；各組人數有限、安慰劑組比例也不低，解讀仍須看具體原因。 這項缺口直接影響產品選擇。禮來需要找出的，是哪一種搭配能留下值得追的藥效，又不讓取得藥效的過程變得太難。只握有最高減幅與跨組停藥範圍，還選不出這個平衡點。 公司指出，最常見的是腸胃道不良事件，多為輕至中度，主要集中在加量期間，組合組比單藥組常見。輕至中度是嚴重程度的描述；對一項需要持續接受的治療，會不會因此中斷，是另一個必須回答的問題。 也因此，23.3%的估計方式值得看清楚。它回答的是：假設所有隨機分派的參與者持續接受研究治療，48週能達到多少效果？這個框架容許可能的暫停或劑量調整，並非直接刪掉停藥者、只算完成者。它揭示藥物的減重潛力，與所有開始治療的人在日常使用中平均能得到什麼，仍是不同問題。 *圖2｜23.3%是在持續接受研究治療的假設下估計。圖中虛線是統計假設的示意，不是觀察到的患者軌跡；效力估計也不等於只計算完成者。* 持續治療為何重要，較早的SURMOUNT-4提供了背景。沒有糖尿病的肥胖或過重成人先接受36週tirzepatide，再分組繼續用藥或改用安慰劑；其後52週，改用安慰劑者平均體重回升，持續用藥者則進一步下降。這不是EloraTZP的停藥結果，但它讓開發者面對的長期問題更清楚：減下來之後，治療與成果還要維持得住。 ## 03｜多一條作用路徑，也多一道產品考題 受體路徑可以互補，病人承受的卻是整套治療。EloraTZP已顯示額外減重潛力，接下來要把這份潛力放進更能持續接受的方案。 *圖3｜eloralintide作用於胰澱素受體，tirzepatide作用於GIP與GLP-1受體。圖為作用路徑示意，並非完整的器官訊號圖；兩分子共同使用的價值，仍由人體試驗判斷。* 禮來提出優化劑量遞增方案，與這次腸胃道問題多出現在加量期間的觀察相呼應。起始強度、增加速度、目標劑量，都是可檢驗的調整環節。下一步不是只問減得最多的配方是哪一個，而是調整之後，額外藥效還能留下多少。 這裡有兩種值得追的走向。如果負擔集中在某個組合或某段加量過程，下一階段就能檢驗針對性的調整能否改善持續治療；如果各種搭配都承受相似負擔，就得更仔細衡量，多少額外減重值得交換這些不適。逐組停藥原因、發生時間，以及中斷後追蹤到的結果，會讓這兩條路逐漸分清。 所以，更完整的資料不是替公告補幾個表格而已。它會影響禮來把哪一種配方往前推，也會影響這個新產品最有理由提供給誰。 *圖4｜加量、耐受與持續追蹤需要一起設計。優化方案尚待驗證；圖中的階梯不是實際劑量或調整時程，也不是個人用藥指引。* ## 04｜做成共製劑，還要跨過耐受這一關 本次第二期採兩藥分開注射，禮來準備開發共製劑，有機會簡化給藥。接下來兩項進度各有自己的問題要回答：製劑整合能否讓施打更方便；加量調整能否改善耐受，同時保住額外藥效。把它們放在一起看，才能知道病人獲得的整體改善有多少。 這也決定了新組合如何與既有tirzepatide並存。單藥已有療效，雙藥得讓需要更大改善的人看見值得選擇的取捨。新組合的定價與支付安排尚未公布；目前最值得追的商業問題，是禮來能為哪些人提供更有說服力的治療選項。 ## 結語｜值得往前押，下一關看取捨 這份第二期結果，值得禮來繼續開發。它已有額外藥效的證據，也已看見必須處理的代價。 接下來最有分量的進展，會是調整加量之後，增益還剩多少、因不良事件停藥的情況如何改變。禮來若能把這兩件事一起做好，第二個分子就更有機會把額外藥效，變成醫師與病人有理由選擇的新選項。 *本文為產業與公司分析，不構成投資或醫療建議。* --- ## 滴血驗癌真的來了？ URL: https://drugnews.com.tw/articles/2026-10-13-galleri-positive-signal-diagnostic-journey.html Date: 2026-10-13 Language: zh-Hant Summary: Theranos曾以指尖少量血液做多種臨床檢驗為訴求；Galleri採靜脈抽血兩管、每管10毫升。產品已有訂單，但主要晚期癌症終點未達標，「早」仍須更多證據。 從指尖取幾滴血，就能完成多種臨床檢驗。矽谷新創 Theranos 曾用這個承諾吸引投資人，最後卻走進詐欺案。2022年，創辦人 Elizabeth Holmes 因詐騙投資人相關罪名，被判刑11年3個月。 如今，另一家公司 GRAIL 正把抽血檢測帶進癌症篩檢。它的 Galleri 實際使用靜脈抽血，標準採集兩管、每管10毫升，再從血液中的 DNA 尋找多種癌症訊號。產品已經賣起來，一季檢測量超過六萬次；現在要回答的，是癌症找到了，究竟夠不夠早？ 2026年9月23日，FDA 諮詢委員以6比4支持 Galleri 的有效性，對效益大於風險也投下多數贊成票。但約14.2萬人參與的英國 NHS-Galleri 試驗，主要的第三、第四期癌症合併終點並未達標。有檢測、有訂單，與證明它能讓族群少一點晚期癌症，仍是不同的成績。 依 FDA 會議摘要，委員普遍對「early」用語有疑慮，部分贊成票附上移除這個字、追加研究等條件。這場投票是諮詢意見；GRAIL 的10月5日公告仍明列，Galleri 尚未獲 FDA 許可或批准。 公司正在走的路，是先服務願意付費的客戶，再爭取讓更大的醫療支付體系接手。卡在中間的，正是這個「早」字能替付款人換到什麼。 *圖1｜FDA 諮詢委員會三項投票：多數支持，但「early」宣稱仍被質疑；諮詢投票不等於正式批准* ## 01｜癌症還小，血裡的線索也可能太少 Galleri 想解決一個真實缺口：不少癌症沒有常規篩檢工具，等症狀出現，往往已經錯過更早處理的機會。一次抽血若能同時找到多種癌症，就不必先猜該檢查哪個器官。 它讀的是血液中的游離 DNA。正常細胞與癌細胞都可能釋出 DNA 碎片，Galleri 從混在一起的碎片裡，找出與癌症相關的甲基化模式。 甲基化是 DNA 上的化學標記。不同組織、不同細胞狀態，標記的分布不同；這些模式提供了兩種線索：有沒有可疑的癌症訊號，以及可能來自哪個組織或器官。 難處在於，有些腫瘤釋放的 DNA 很少，早期的小病灶也未必送出足夠訊號。FDA 委員看到的資料顯示，Galleri 對第一期癌症的偵測，整體不如第二至第四期。能比症狀出現更早發現癌症，仍不等於多數癌症都能在第一期被找到。 *圖2｜從血中游離 DNA 到甲基化模式：為何早期癌可能漏掉；未偵測到訊號不能排除癌症* 這讓 NHS-Galleri 的隨機試驗格外重要。研究把約14.2萬人分成篩檢與對照兩組，進行三輪年度篩檢並追蹤，要看預先指定12種癌症的第三、第四期診斷發生率，能否因加入 Galleri 而下降。 9月刊出的論文確認，主要終點未達標：發生率比1.03，95%信賴區間0.92至1.14，P值0.63。這次試驗沒有顯示，Galleri 讓這個合併晚期終點減少。 同一組癌症的第四期關鍵次要終點，發生率比為0.86；ASCO 資料表保留的95%信賴區間是0.744至0.998，達名目統計顯著。這個方向值得追蹤，但不能改寫主要終點失利，也沒有直接證明癌症死亡減少。 對 GRAIL 而言，困難不只是找到更多癌症。若希望每年替大批沒有症狀的人抽血，付款人還會問：多找出來的診斷，究竟讓多少人避開晚期疾病？這份答案目前仍有缺口，擴大客群便不能只靠檢測銷量說服人。 *圖3｜NHS-Galleri 終點：主要第三、四期合併終點未達標，第四期次要終點僅名目顯著，未證明死亡減少* ## 02｜還沒拿到 FDA 核准，為何已經能賣？ Galleri 目前在美國以實驗室開發檢測，也就是 LDT，提供服務。GRAIL 在第二季申報文件中說明，現行銷售不以取得 FDA 核准為前提；檢體由其符合 CLIA 要求、取得 CAP 認證的實驗室處理。這些實驗室資格，與 FDA 對產品作出的市場許可是不同層次的要求。 因此，公司可以先接自費訂單，也與自保雇主及部分健康保險計畫合作。官網列價949美元，符合指定開單方式的病人自費價為799美元；部分雇主、壽險業者提供全額或部分福利支付。Galleri 尚未取得廣泛的 Medicare 與大型商業健康保險給付。 這條路已經帶來收入。第二季 Galleri 檢測量超過61,000次、年增35%，檢測收入4,260萬美元、年增24%。不過，量增沒有等幅變成收入：同一季申報文件列出的平均售價，年減8%。 公司也揭露，大客戶、國際經銷商的議價及折扣等安排，會影響每次檢測的收入。爭取更多訂單，可以把實驗室設備與營運成本分攤到更多檢體；接到大客戶，卻不代表每一單都能按官網價格收錢。 這就是現階段市場選擇的代價：GRAIL 能先累積訂單與使用經驗，成長卻仍受買方願付多少、哪些人有福利補助所限制。要把服務推向更廣的受檢族群，公司仍在爭取 FDA 核准，讓大型付款人有機會正式評估納入給付。 2026年2月3日成法的 Public Law 119-75，第6221節，替這件事開了一個時間窗口：自2029年1月1日起，符合條件的多癌症早期偵測可循 Medicare 路徑爭取給付。 門檻包括 FDA 市場許可，以及 CMS 透過全國給付決定，認定對適當受益人屬合理且必要，另有年齡與檢測間隔條件。法律有了，Galleri 的資格還沒拿到；即使取得 FDA 核准，也不保證後續給付。下一批可能的大買方，仍要自己判斷證據值不值得付錢。 ## 03｜一次抽血的報價，還不是篩檢的總帳單 當支付從個人自費走向整個族群，檢測之外的費用就更難略過。 NHS-Galleri 第一輪有722人驗出癌症訊號，後續419人確認原發癌症，陽性預測值約58%。陽性報告只能提示往哪裡找；醫師仍須依可能來源與個人情況，安排影像、內視鏡，或必要的取樣與病理檢查。 這意味著，買下檢測的預算之外，還需要能完成診斷的醫療資源。多少陽性者最後沒有找到癌症、用了多少檢查、等待多久，都會改變擴大篩檢的負擔。陰性也不能排除癌症，原本建議的癌症篩檢仍須繼續。 競爭者 Guardant Health 的 Shield，選的是較窄的入口：2024年7月獲 FDA 核准，聚焦45歲以上、平均風險成人的大腸直腸癌篩檢，陽性後接大腸鏡確認。 Galleri 選擇一次涵蓋多種癌症，能碰到更多篩檢空白，也把後續問題帶進不同器官與專科。兩項產品的用途不同，不能跨研究直接比誰比較準；它們讓買方承接的診斷安排，本來就不同。 所以，GRAIL 除了等待更長期的疾病結果，還得把現有陽性者的就醫過程量清楚。這類資料能幫醫院與保險計畫估算導入所需的資源，讓原本只買一次檢測的討論，走到是否能年年提供服務的決策。 *圖4｜陽性訊號之後的診斷流程：醫師評估、依情況選檢查、必要時取樣病理；非固定檢查套餐* ## 04｜三星的資金，讓另一條市場路線繼續走 等待證據與支付決定的同時，GRAIL 仍有一家公司要經營。第二季全公司淨損1.102億美元，季末現金及短期投資8.616億美元；淨損包括研發、商業化與會計攤銷等項目，並非每次檢測的虧損或當季現金流出。 6月25日，GRAIL 完成三星關係企業合計1.1億美元的股權投資。公告把這筆資金連到美國核准、給付進程與國際市場拓展，合作則規劃在南韓商業化，並探索日本、新加坡的潛在機會，仍須滿足監管及其他條件。 安排中有一個值得注意的細節：初期檢測會交給 GRAIL 在美國北卡羅來納州的實驗室處理。 依這個安排，亞洲合作初期便可沿用既有的美國實驗室。從商業安排看，公司可以一邊爭取美國的大型付款人，一邊準備海外入口；但實驗室能出報告，並不會自動解決當地准入、付款與陽性後就醫的問題。合作能把訂單帶近哪一步，仍取決於各市場接下來完成什麼。 10月5日，GRAIL 預告將在10月25、26日的 ESMO 場次，發表癌症訊號引導的後續診斷流程，以及臨床實務中的診斷檢查資料。截至10月12日，這些仍待發表。 這批資料值得看的是：醫院要花多少時間與檢查，才能把 Galleri 的訊號追成診斷。流程若改善，可能降低醫療機構導入的負擔；它仍不能補上主要晚期終點未達標留下的缺口。 GRAIL 可以用診斷資料，協助合作方決定如何接更多檢測；要讓「早」成為大規模、持續支付的理由，還得用晚期疾病與更長期的健康結果說服付款人。眼前的訂單能繼續做，下一步能走多大，取決於公司交出的是哪一種證據。 *本文為產業與公司分析，不構成投資或醫療建議。* --- ## Has Cancer Detection from a Blood Sample Really Arrived? URL: https://drugnews.com.tw/articles/2026-10-13-galleri-positive-signal-diagnostic-journey-en.html Date: 2026-10-13 Language: en Summary: Theranos promised multiple clinical tests from a small finger-prick sample. Galleri uses two 10 mL tubes of venous blood. Orders are growing, but the primary late-stage cancer endpoint was not met, leaving “early” in need of more evidence. A few drops of blood from a fingertip, enough to perform a wide range of clinical tests. Silicon Valley startup Theranos once used that promise to attract investors, only to end up at the center of a fraud case. In 2022, founder Elizabeth Holmes was sentenced to 11 years and three months in prison on charges related to defrauding investors. Now another company, GRAIL, is bringing blood testing into cancer screening. Its Galleri test actually uses a venous blood draw, with a standard collection of two tubes containing 10 milliliters each, and looks for signals of multiple cancers in DNA circulating in the blood. The product is already selling, with more than 60,000 tests in a quarter. The question now is whether the cancers it finds are being found early enough. On September 23, 2026, FDA advisory committee members voted 6–4 in favor of Galleri's effectiveness, and a majority also voted that its benefits outweighed its risks. But the UK NHS-Galleri trial, involving approximately 142,000 people, did not meet its primary endpoint combining stage III and stage IV cancers. Having a test and receiving orders are different achievements from demonstrating fewer late-stage cancers across a population. According to the FDA meeting summary, committee members broadly expressed concerns about the word “early.” Some favorable votes came with conditions, including removing that word and conducting additional studies. The vote was advisory. GRAIL's October 5 announcement still explicitly stated that Galleri had not been cleared or approved by the FDA. The company is pursuing a path that starts with customers willing to pay, then seeks to bring larger healthcare payment systems on board. The question in between is what the word “early” can deliver for payers. *Figure 1 | Three FDA advisory committee votes: majority support, but the “early” claim remains in question; an advisory vote is not formal approval* ## 01 | When a cancer is small, clues in the blood may also be scarce Galleri aims to address a real gap. Many cancers lack routine screening tools, and by the time symptoms appear, opportunities for earlier intervention have often been missed. If one blood draw could detect multiple cancers, there would be less need to decide in advance which organ to investigate. The test reads cell-free DNA in the blood. Both normal cells and cancer cells can release DNA fragments. Galleri looks through that mixture for methylation patterns associated with cancer. Methylation consists of chemical marks on DNA. Their distribution differs across tissues and cell states. These patterns offer two kinds of clues: whether a suspicious cancer signal is present, and which tissue or organ it may come from. The difficulty is that some tumors release very little DNA, and small, early-stage lesions may not produce enough signal. The data reviewed by FDA committee members showed that Galleri's overall detection of stage I cancers was lower than its detection of stage II through IV cancers. Finding a cancer before symptoms emerge still does not mean that most cancers can be found at stage I. *Figure 2 | From cell-free DNA to methylation patterns: why early-stage cancers may be missed; no detected signal does not rule out cancer* This makes the randomized NHS-Galleri trial particularly important. The study divided approximately 142,000 people into screening and control groups, with three rounds of annual screening and follow-up. It examined whether adding Galleri could reduce the incidence of stage III and IV diagnoses across 12 prespecified cancers. The paper published in September confirmed that the primary endpoint was not met: the incidence rate ratio was 1.03, with a 95% confidence interval of 0.92 to 1.14 and a P value of 0.63. The trial did not show that Galleri reduced this combined late-stage endpoint. For the key secondary endpoint of stage IV disease in the same group of cancers, the incidence rate ratio was 0.86. The ASCO fact sheet reported the 95% confidence interval at its original precision of 0.744 to 0.998, reaching nominal statistical significance. That direction warrants follow-up, but it does not reverse the primary endpoint failure or directly demonstrate a reduction in cancer deaths. For GRAIL, the challenge goes beyond finding more cancers. If the aim is to draw blood annually from large numbers of people without symptoms, payers will also ask how many people avoid late-stage disease because of those additional diagnoses. That answer remains incomplete, so test sales alone cannot make the case for a broader customer base. *Figure 3 | NHS-Galleri endpoints: the primary combined stage III and IV endpoint was not met; the stage IV secondary endpoint was only nominally significant, with no demonstrated mortality reduction* ## 02 | Why can Galleri be sold before FDA approval? Galleri is currently offered in the United States as a laboratory-developed test, or LDT. In its second-quarter filing, GRAIL explained that current sales do not depend on obtaining FDA approval. Samples are processed in its CLIA-compliant, CAP-accredited laboratory. Those laboratory qualifications and FDA marketing authorization for a product are different levels of requirements. The company can therefore accept self-pay orders while also working with self-insured employers and some health plans. The website lists a price of $949, with a patient self-pay price of $799 for qualifying ordering routes. Some employers and life insurers provide full or partial benefit coverage. Galleri has not secured broad coverage from Medicare or major commercial health insurers. This route is already generating revenue. Second-quarter Galleri test volume exceeded 61,000, up 35% year over year, while testing revenue reached $42.6 million, up 24%. Volume growth did not translate into an equivalent rise in revenue: the filing for the same quarter reported an 8% year-over-year decline in average selling price. The company also disclosed that negotiated pricing and discounts for large customers and international distributors, among other arrangements, affect revenue per test. Winning more orders can spread laboratory equipment and operating costs across more samples. Landing a large customer, however, does not mean that every order will command the website's listed price. That is the trade-off in GRAIL's current market approach. The company can build orders and experience with the test, but growth remains constrained by what buyers are willing to pay and who has access to benefits that help cover the cost. To reach a broader screening population, GRAIL is still seeking FDA approval, giving large payers an opportunity to formally evaluate coverage. Section 6221 of Public Law 119-75, enacted on February 3, 2026, opened a window for that process: beginning January 1, 2029, qualifying multi-cancer early detection tests can pursue coverage through a Medicare pathway. The requirements include FDA marketing authorization and a CMS national coverage determination finding the test reasonable and necessary for appropriate beneficiaries, along with age and testing-interval conditions. The law is in place, but Galleri has not yet qualified. Even FDA approval would not guarantee subsequent coverage. The next potential large buyers still have to judge whether the evidence is worth paying for. ## 03 | The price of a blood test is not the full screening bill As payment shifts from individual self-pay customers to entire populations, costs beyond the test become harder to leave out. In the first round of NHS-Galleri, 722 participants had a cancer signal detected, and 419 were subsequently confirmed to have a primary cancer, giving a positive predictive value of approximately 58%. A positive report can only suggest where to look. Physicians must still arrange imaging, endoscopy, or, where necessary, tissue sampling and pathology, based on the predicted origin and the individual's circumstances. This means that a testing budget also needs healthcare resources capable of completing the diagnosis. How many people with positive results ultimately have no cancer found, how many investigations they undergo, and how long they wait all affect the burden of expanding screening. A negative result cannot rule out cancer either, and previously recommended cancer screening must continue. Competitor Guardant Health's Shield takes a narrower route. Approved by the FDA in July 2024, it focuses on colorectal cancer screening in average-risk adults aged 45 and older, with a positive result followed by colonoscopy for confirmation. Galleri's choice to cover multiple cancers at once addresses more screening gaps, while also extending follow-up questions across different organs and specialties. The products serve different purposes, so their accuracy cannot be directly ranked across studies. The diagnostic arrangements they require buyers to support are different from the outset. Alongside waiting for longer-term disease outcomes, GRAIL therefore needs to measure the care journeys of people who already have positive results. Such data can help hospitals and health plans estimate the resources required for adoption, moving the discussion from buying a single test to deciding whether the service can be offered year after year. *Figure 4 | The diagnostic pathway after a positive signal: physician assessment, investigations selected for the circumstances, and tissue sampling and pathology when needed; not a fixed package of tests* ## 04 | Samsung's funding keeps another market route moving While evidence and coverage decisions are pending, GRAIL still has a business to run. The company reported a second-quarter net loss of $110.2 million and ended the quarter with $861.6 million in cash and short-term investments. Net loss includes research and development, commercialization, accounting amortization, and other items; it is neither a loss per test nor the quarter's cash outflow. On June 25, GRAIL completed a total of $110 million in equity investment from Samsung-affiliated entities. The announcement linked the funding to the US approval and reimbursement process and international expansion. The collaboration envisages commercialization in South Korea and exploration of potential opportunities in Japan and Singapore, subject to regulatory and other conditions. One detail in the arrangement deserves attention: initial testing will be processed at GRAIL's laboratory in North Carolina, in the United States. Under this arrangement, the Asian collaboration can initially use the existing US laboratory. From a commercial perspective, GRAIL can pursue large US payers while preparing an overseas entry route. But a laboratory's ability to issue reports does not automatically resolve local market access, payment, or care after a positive result. How far the partnership brings the company toward orders still depends on what each market completes next. On October 5, GRAIL announced plans to present data at ESMO sessions on October 25 and 26 on cancer-signal-directed diagnostic pathways and diagnostic investigations in clinical practice. As of October 12, those presentations were still pending. The question to watch in these data is how much time and how many investigations hospitals need to turn a Galleri signal into a diagnosis. Better pathways may ease the burden of adoption for healthcare providers. They still cannot fill the gap left by the failure to meet the primary late-stage endpoint. GRAIL can use diagnostic data to help partners decide how to handle more tests. To make “early” a reason for large-scale, sustained payment, it still needs to persuade payers with evidence on late-stage disease and longer-term health outcomes. Current orders can continue. How far the next step takes the company depends on which kind of evidence it delivers. *This article is an analysis of the industry and company and does not constitute investment or medical advice.* *本文為產業與公司分析，不構成投資或醫療建議。* --- ## 乾燥症三期喊停，argenx 押錯的是機制不是藥 URL: https://drugnews.com.tw/articles/2026-10-13-argenx-sjogrens-unity-futility-mechanism.html Date: 2026-10-13 Language: zh-Hant Summary: argenx 10 月 8 日因無效性中止乾燥症三期 UNITY；同一個月諾華與 Amgen 的 B 細胞與 CD40L 路線相繼達標。差別不在藥，在機制動到免疫系統的哪一層，這也重畫了 efgartigimod 擴張清單的分類。 乾燥症三期喊停，argenx 押錯的是機制不是藥 2026 年 10 月 8 日，argenx 宣布中止 efgartigimod 皮下劑型在乾燥症（Sjögren's disease）的第三期 UNITY 試驗。獨立資料監測委員會在期中分析後建議「因無效性停止」，結論是這項試驗無法達到主要終點。公司原本預計 2027 年下半年才會有結果，現在提前一年出局。[1][2] 同一天，argenx 也公布了 FB102 在乳糜瀉第二期的正向結果。市場看到的是「一勝一敗」，但對投資人真正重要的是另一個問題：efgartigimod 在重症肌無力、CIDP、ITP 都成功，為什麼到了乾燥症就失靈？答案指向 argenx 整個擴張清單的篩選邏輯，而不只是一個適應症。[3] *圖1｜三種機制攻乾燥症，只有直接動 B 細胞與 T–B 共刺激的達標* ## 01｜三週內，三家公司給了乾燥症三個答案 乾燥症是第二常見的風濕免疫病，美國估計有 100 萬到 400 萬名患者，九成是女性，三到四成血清抗體陰性。幾十年來沒有一款針對病因的藥獲准：rituximab 在 TEARS 與 TRACTISS 兩項試驗都沒達到主要終點，belimumab 有訊號但沒走到核准。[4] 2026 年秋天局面突然改變。9 月 22 日，Amgen 宣布 CD40 配體拮抗劑 dazodalibep 的第三期 OASIZ 301（約 621 人，ESSDAI 至少 5 分）達到主要終點，第 48 週 ESSDAI 改善具統計與臨床意義，第 4 週就看得到變化。[5] 更早的 1 月，諾華的 ianalumab 拿到 FDA 突破性療法認定。它的兩項第三期 NEPTUNUS-1 與 NEPTUNUS-2 都達標，第 48 週 ESSDAI 下降 6.4 分對安慰劑 5.1 分、6.5 分對 5.5 分，病人整體評估也首次在乾燥症三期出現正向結果。諾華計畫 2026 年起向各國送件。[4][6] 然後是 10 月 8 日的 argenx。三個結果擺在一起，差別不在公司規模或試驗設計，而在三種藥各自動了免疫系統的哪一層。 ## 02｜efgartigimod 動的那一層，在乾燥症不是病因 efgartigimod 的機制是阻斷 FcRn，讓血中 IgG 加速分解。它在重症肌無力有效，因為那個病的病因就是一群攻擊乙醯膽鹼受體的 IgG 抗體；把 IgG 拉低，症狀就改善。CIDP 與 ITP 的邏輯相同：抗體就是兇手，清掉抗體病就退。[7] 乾燥症不一樣。病人體內確實有 anti-Ro/SSA 等自體抗體，UNITY 也只收抗體陽性的病人；但腺體裡真正造成破壞的，是持續活化的 B 細胞與 T 細胞在淚腺、唾液腺裡聚集、增生、分泌發炎物質。抗體更像這個過程留下的指紋，不是扣扳機的手。[4] ianalumab 直接清除 B 細胞，同時封住 BAFF 受體，讓 B 細胞清掉之後長不回來。dazodalibep 擋住 CD40L，切斷 T 細胞給 B 細胞的活化訊號。兩者都動到「製造發炎的機器」。efgartigimod 只處理機器的產品，機器還在運轉。這就是為什麼同樣鎖定抗體陽性病人，一個達標、一個達標、一個中途喊停。[4][5][7] *圖2｜2026 年乾燥症三張成績單的時間線* argenx 的醫學長 Luc Truyen 在聲明裡說，乾燥症是「免疫學裡最異質、最複雜的疾病之一」。這句話說對了一半；另一半是，異質的疾病更需要在三期之前先證明機制抓對了層。UNITY 沒有公開的第二期乾燥症數據可以回頭對照，這一點在期中就被無效性分析放大。[1] ## 03｜對 argenx 的影響：少一個市場，多一個問題 先看帳面。argenx 2026 年第二季產品淨銷售 15.16 億美元，年增 60%；上半年 28.13 億美元；6 月底現金與短期金融資產 52 億美元。乾燥症中止不會動到任何一條現有收入，也不會改變 Vision 2030「治療 5 萬名病人」的路徑。[8] 真正被動到的，是市場給 argenx 的「一個產品、多個適應症」估值邏輯。efgartigimod 的擴張清單包括眼肌型重症肌無力（已正向）、肌炎（ALKIVIA，原訂 2026 年第三季公布）、原發性 ITP（2027 上半年）、Graves 病（登記試驗進行中）。UNITY 的結果等於告訴投資人：這張清單要重新分類，哪些是抗體驅動、哪些只是抗體相關。[8] 肌炎是最靠近的下一題。皮肌炎與多發性肌炎有一群特異性自體抗體，但肌肉破壞同樣牽涉 T 細胞浸潤與補體。如果 ALKIVIA 也達標，乾燥症就是一次個案；如果不達標，市場會把 efgartigimod 的天花板畫在「純抗體病」。這是接下來三個月 argenx 最重要的一個數字。[8] *圖3｜三款藥在乾燥症的 ESSDAI 數字：有數字的達標，沒數字的中止* ## 04｜乾燥症市場誰拿走：諾華先到，Amgen 跟上，J&J 還在路上 對乾燥症這個市場，10 月 8 日之後的格局是兩家領先、一家觀望。諾華 ianalumab 有兩項三期與突破性認定，2026 年送件，可能成為第一個獲准的病因治療。Amgen dazodalibep 三期達標，低系統活性族群的 OASIZ 303 預計 2026 年第四季完成，若也達標，適用人群會比諾華寬。[5][6] 還有一家要盯：嬌生的 nipocalimab 與 efgartigimod 同屬 FcRn 阻斷劑，乾燥症三期結果預計 2026 年底到 2027 年初。如果嬌生也失敗，FcRn 在乾燥症的故事就結束；如果嬌生成功，argenx 要解釋的就不只是機制，還有自己的試驗設計。[4] 對台灣讀者，乾燥症（修格蘭氏症）是風濕科常見門診病，長期只有人工淚液、唾液刺激劑與免疫抑制劑可用。第一個病因治療的定價與給付，會是 2027 年健保與原廠談判的新題目；台灣臨床試驗中心若在 OASIZ 303 或 nipocalimab 三期有收案，也會在數據公布時被點名。 *圖4｜efgartigimod 擴張清單重新分類：抗體驅動的已成，抗體相關的待驗* ## 05｜接下來看什麼 三個事件會決定這篇文章的判斷是否成立。第一，argenx 10 月 22 日第三季財報是否同時交出 ALKIVIA 肌炎數據，以及管理層如何重新描述擴張清單。第二，諾華 ianalumab 的送件時間與 FDA 審查時程，決定第一個乾燥症病因藥何時上市。第三，嬌生 nipocalimab 的乾燥症三期，決定 FcRn 這一層在乾燥症是「argenx 做錯」還是「機制本身不對」。 UNITY 的中止不是 argenx 的財務事件，是一次機制分類的修正。藥沒有變差，是這個病不照藥的邏輯生病。 *本文為產業與公司分析，不構成投資或醫療建議。* --- ## Defending its blockbuster, Merck hits a patent wall URL: https://drugnews.com.tw/articles/2026-10-12-keytruda-qlex-subcutaneous-delivery-value-en.html Date: 2026-10-12 Language: en Summary: One drug accounts for half a company's revenue. That does not give its patents any extra years. One drug accounts for half a company's revenue. That does not give its patents any extra years. Merck chose South Korea's Alteogen to turn Keytruda into a more convenient subcutaneous formulation. Work on the next blockbuster must continue. For now, the company is trying to secure another chapter in the commercial life of its current mainstay. The product was developed, and the FDA approved it. But another US company, Halozyme, now stands in the way of its launch. On October 7, Halozyme announced that a Dutch court had issued an injunction concerning the MDASE patent EP622, restricting Merck subsidiary MSD BV from manufacturing and selling subcutaneous Keytruda in eight markets, including the Netherlands, France and Italy. The intravenous version is not affected by this injunction. The scope described above comes from Halozyme's disclosure; the full court judgment has not been obtained. Merck has already paid for access to its partner's technology. Why can a third company still stand in its way? And what additional costs might this injunction impose on a deal already agreed? # 01 | Half the company's revenue cannot simply wait for the next drug In the second quarter of 2026, Keytruda and its subcutaneous version, QLEX, generated a combined $8.366 billion in revenue, about half of Merck's revenue for the quarter. Behind that figure is a problem a large pharmaceutical company can hardly avoid: its biggest revenue generator is approaching patent expiry, while the success of a successor drug cannot be scheduled to suit the company's needs. Merck's 2025 annual report states that Keytruda's principal US compound patent will expire in 2028. Two other patents in the same family extend to 2029, and the company expects that they may face litigation. The listed expiry in major European markets is 2031. The US and Europe run on different clocks. The year 2028 does not mark the simultaneous loss of all protection worldwide, nor does it establish that alternatives will appear in Europe that same year. Finding new targets and developing new antibodies opens up future possibilities. But clinical trials, approvals and hospital adoption all take time. Competition for Keytruda will not automatically wait because a successor is not ready. The subcutaneous version therefore became one way to defend the existing business. Doctors already know pembrolizumab. What Merck is offering is an administration option that can make some suitable treatment regimens more convenient. The original antibody's patent clock has not restarted, but the way hospitals compare products may change. If both options require IV infusion, price can easily become the focus. If one can save administration time in suitable regimens, purchasing decisions also need to account for nursing work, treatment space and patients' time. Whether a new formulation can defend the business ultimately depends on what that convenience is worth inside a hospital. Benefits listed in a presentation first have to become a service the hospital wants to keep. *Figure 1 | About one or two minutes, depending on the regimen, refers to QLEX injection time, not the whole visit. Both products contain pembrolizumab as the anticancer component.* # 02 | One enzyme earns a share in a blockbuster business Alteogen entered this business by opening up a path for the drug solution beneath the skin. Pembrolizumab in Keytruda blocks PD-1 signalling on T cells, relieving some of the immune suppression caused by that pathway and giving antitumour immune responses an opportunity to function. In the subcutaneous version, the antibody remains the anticancer component. Subcutaneous tissue, however, is not an empty pouch waiting to be filled with medicine. The network between cells contains hyaluronan, which resists the dispersion of a drug solution. An IV infusion enters the bloodstream directly. A subcutaneous injection first has to disperse through tissue and then be absorbed. Alteogen's hyaluronidase, berahyaluronidase alfa, temporarily breaks down hyaluronan locally, increasing tissue permeability and helping a larger volume of drug solution disperse. Its role is to improve delivery; it does not change the antibody's action in mobilising an immune response. *Figure 2 | The antibody blocks the PD-1 pathway; hyaluronidase temporarily increases local subcutaneous tissue permeability, supporting dispersion and absorption. Mechanistic illustration, not to scale.* Being able to inject the drug is only part of the work. Developers also have to show that the body still receives appropriate drug exposure through the different route. The 377-patient randomised trial supporting FDA approval met its predefined drug-exposure comparability criteria. Efficacy was assessed descriptively; the trial did not show that the subcutaneous version treats cancer more powerfully. The convenience gained from that development is concrete: a healthcare professional administers the injection in about one or two minutes, depending on the regimen, compared with an approximately 30-minute IV infusion. If the patient still needs IV chemotherapy that day, only one part of the treatment is shortened. The whole visit does not become a two-minute appointment. In February 2024, Merck converted the previously non-exclusive collaboration into a global exclusive licence for subcutaneous pembrolizumab. Alteogen is also responsible for commercial enzyme supply. The enzyme's compensation became tied to the blockbuster's progress. Merck's 2025 annual report lists regulatory milestones and up to $1 billion in total annual and cumulative sales milestones under the collaboration. Royalties in the low-single-digit percentage range of net sales begin only after all sales milestones have been achieved. Following US and EU approvals in 2025, Merck paid $40 million in regulatory milestones. The company also considered it probable that future sales would trigger $890 million in milestones and recognised the related liabilities. That $890 million had not been paid at the time; payment remained contingent on the corresponding sales conditions being met. Alteogen does not itself own pembrolizumab, yet it can receive compensation as the product wins approval, sells and crosses milestones. A technology company does not necessarily have to create another blockbuster. Solving a blockbuster's immediate problem can also build a substantial business. How much do a few percentage points matter to Merck? Consider a scenario constructed here solely to examine royalty-rate sensitivity: if annual net sales of the subcutaneous product were $10 billion, each one-percentage-point difference in royalties would amount to $100 million a year. This is an illustration of scale, not a QLEX sales forecast or savings already realised in this case. But total costs also include earlier milestone payments, enzyme-supply charges and payment timing. Low-single-digit royalties begin only after all sales milestones have been achieved. A range alone cannot reveal the exact rate or reconstruct the complete, comparable terms for the two companies. The importance of a royalty rate can be estimated. Public information does not give a complete answer to how Merck originally weighed price, supply and technology. # 03 | A signed agreement does not remove a third party's patents Merck and Alteogen can agree on how the technology will be used, how the enzyme will be supplied and how returns will be shared. Halozyme's asserted rights do not disappear when those two companies sign a contract. Halozyme licenses subcutaneous delivery technology through ENHANZE. But it explicitly states that the MDASE patents at issue in this case are different from the ENHANZE patents and are not included in the ENHANZE licensing programme. This litigation therefore has not established a case of unpaid ENHANZE fees being pursued as a debt. The dispute concerns whether the formulation Merck uses falls within the scope of a third party's patent protection. Alteogen can license its own technology, and the FDA can approve the product for treatment under specified conditions. Neither development also resolves a third-party patent dispute. Nor have courts in different jurisdictions given the same answer. Merck's second-quarter 2026 SEC filing states that Halozyme had agreed in the UK to revoke the UK portion of EP622. The Dutch outcome in October, by contrast, was unfavourable to Merck. Developments in one jurisdiction cannot establish a worldwide winner. But the commercial consequences cannot wait until every lawsuit has ended before they are considered. MSD BV, affected by the Dutch injunction, is the European marketing authorisation holder and a packaging and distribution hub. Launch arrangements for the new formulation are consequently disrupted in the restricted markets described in the announcement. The drug is ready, but its routes to supply and sale are restricted. That problem will travel from legal documents to hospital purchasing decisions. # 04 | Market access can be contested again. How can lost time be recovered? The IV version remains available. This injunction has not closed off Merck's existing treatment route. Nor can its commercial consequences simply be estimated using all of Keytruda's revenue. But continuing to sell an existing product and establishing a new formulation in time are different things. A hospital cannot change an administration workflow as quickly as a pharmaceutical company can change a product poster. For suitable patients who do not need IV treatment on the same day, subcutaneous injection may offer clear convenience. To make it a routine service, hospitals still have to assess reimbursement, purchasing, nursing responsibilities and scheduling, gradually building experience with its use. If supply is restricted during this period, that work cannot proceed at the original pace. Once a launch can resume, the product will have to compete again for a place in hospital workflows. Price competition in the market may not have stood still. *Figure 3 | Whether faster administration produces hospital value depends on the whole regimen and workflow. Other IV treatment that day may still require a stay. The illustrative situations are not measured cost savings.* QLEX generated $463 million in sales in the second quarter of 2026, about 5.5% of revenue from the Keytruda product portfolio. That is a revenue share. It does not mean that 5.5% of patients have switched. *Figure 4 | Combined second-quarter 2026 sales of $8.366 billion already include QLEX's $463 million. Internal product switching cannot be counted again as a net-new company market. Source: Merck, August 4, 2026.* When a patient previously receiving IV Keytruda switches to QLEX, subcutaneous revenue can rise without the company gaining a second patient. That means this defensive effort cannot be assessed only by how quickly the new formulation grows. Another question matters: after paying for the technology, how much demand has it retained that might otherwise have been lost? If hospitals have merely changed a product name without gaining workflow improvements worth keeping, price may still come first in future decisions. Only when convenience becomes part of routine service does the subcutaneous version have grounds to compete on a different basis. This also defines the roles of two strategies. Better delivery makes an existing business more competitive; next-generation medicines still have to create new therapeutic value. Litigation over the subcutaneous version will not bring forward the completion of clinical trials for molecules still in development. Merck has to keep fighting for room to launch while also confronting the cost of time lost in building adoption. When the product can reach these hospitals without obstruction, how much more will they be willing to pay for the convenience of this injection? This article provides industry information and commercial analysis. It is not individualized medical or investment advice. # Sources 1. [Halozyme, October 7, 2026: announcement of the Dutch injunction](https://ir.halozyme.com/news/news-details/2026/Halozyme-Wins-Injunction-Stopping-Manufacture-and-Sale-of-Mercks-Keytruda-SC-in-Multiple-European-Markets/default.aspx). Disclosure by a party to the litigation; also the source for the distinction between MDASE and ENHANZE. 2. [Merck, 2025 Form 10-K](https://www.merck.com/wp-content/uploads/sites/124/2026/05/MRK-12.31.2025-10K-FINAL.pdf): principal patent table and footnote 10 on printed page 14. The Alteogen collaboration, payments made and future milestones are covered in [Note 3](https://www.sec.gov/Archives/edgar/data/310158/000031015826000063/R13.htm). 3. [Merck, August 4, 2026: second-quarter results and product sales table](https://www.merck.com/news/merck-highlights-key-regulatory-and-clinical-milestones-across-broad-diverse-pipeline/). 4. [FDA, September 19, 2025: QLEX approval and the supporting 377-patient trial](https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-pembrolizumab-and-berahyaluronidase-alfa-pmph-subcutaneous-injection). 5. [DailyMed: QLEX prescribing information, Section 12.1](https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=097d166f-b73b-41d3-9b37-7653cd2a0c41), covering the PD-1 and hyaluronidase mechanisms. 6. [Merck, September 19, 2025: QLEX approval announcement, administration by healthcare professionals and injection times](https://www.merck.com/news/fda-approves-mercks-keytruda-qlex-pembrolizumab-and-berahyaluronidase-alfa-pmph-injection-for-subcutaneous-use-in-adults-across-most-solid-tumor-indications-for-keytruda-pem/). 7. [Alteogen, February 22, 2024: announcement of the amended exclusive licence with MSD and commercial enzyme supply](https://alteogen.com/en/sub/ir/news.php?bid=13&idx=163&mode=view). 8. [Merck, second-quarter 2026 SEC disclosure of legal matters](https://www.sec.gov/Archives/edgar/data/310158/000031015826000212/R15.htm). This second-quarter disclosure predates the Dutch announcement in October. --- ## 藥王保衛戰，默沙東撞上專利牆 URL: https://drugnews.com.tw/articles/2026-10-12-keytruda-qlex-subcutaneous-delivery-value.html Date: 2026-10-12 Language: zh-Hant Summary: 一支藥撐起半家公司的營收，專利卻不會因此多給幾年。 一支藥撐起半家公司的營收，專利卻不會因此多給幾年。 默沙東選了韓國Alteogen，把Keytruda做成更方便的皮下劑型。下一支重磅新藥還得繼續開發，眼前這支藥王，先爭取下一段商業生命。 產品做出來了，FDA也核准了。擋在上市路上的，卻是另一家美國公司Halozyme。 10月7日，Halozyme宣布，荷蘭法院就MDASE專利EP622下達禁令，限制默沙東旗下MSD BV在荷蘭、法國、義大利等八個市場製造、銷售皮下版Keytruda。 靜脈版不受這道禁令影響。上述範圍來自Halozyme披露，法院判決全文尚未取得。 默沙東已經付錢買進合作夥伴的技術，為什麼還會被第三家公司攔下？這張禁令，又會讓原本談好的生意多付出什麼代價？ # 01｜半家公司的營收，不能只等下一支藥來接 2026年第二季，Keytruda與皮下版Qlex合計帶來83.66億美元營收，約占默沙東當季營收的一半。 這個數字背後，是一家大藥廠很難迴避的問題：最能帶來營收的產品，正在接近專利期限；接班的新藥，卻不能按公司的需要指定哪一天成功。 默沙東2025年報列明，Keytruda在美國的主要化合物專利將於2028年到期。同專利族另有兩件延至2029年，公司預期可能面臨訴訟；主要歐洲市場的期限則列在2031年。 美國與歐洲各有時鐘。2028不是全球同時失去所有保護的日期，也不能拿來推定歐洲同年就會出現替代品。 找新靶點、做新抗體，是為未來開疆闢土。只是臨床、核准、醫院採用，每一步都要時間。Keytruda面對的競爭，不會因為接班藥物還沒準備好，就自動延後。 皮下版於是成了守住既有生意的一種做法。 醫師已經熟悉pembrolizumab，默沙東要提供的，是讓部分合適療程更方便的給藥選擇。原抗體專利沒有重新計時，但醫院比較產品的方式，可能因此改變。 若兩邊都要靜脈輸注，價格容易成為焦點；若其中一邊能省下合適療程的給藥時間，採購便得把護理操作、治療空間與病人的時間算進去。 一支新劑型能否守住生意，最後要看這些便利在醫院裡值不值錢。寫在簡報上的優點，得先變成院方願意留下的服務。 *圖1｜依療程約1或2分鐘是QLEX的注射時間，不是完整看診時間；兩種產品的抗癌主成分皆為pembrolizumab。* # 02｜一支酶，拿到了大藥生意的分潤門票 Alteogen切進這場生意，靠的是替藥液打開皮下的通道。 Keytruda的pembrolizumab會阻斷T細胞上的PD-1訊號，解除這條路徑造成的部分免疫抑制，讓抗腫瘤免疫反應有機會發揮。換成皮下版，抗癌主角仍是這支抗體。 然而，皮下並不是一個等著裝藥的空袋子。 細胞之間的網絡含有玻尿酸，會對藥液擴散形成阻力。靜脈輸注直接進血液，皮下注射則得先在組織裡散開，再被吸收。 Alteogen的玻尿酸酶berahyaluronidase alfa，能暫時、局部拆解玻尿酸，增加組織通透性，幫較大體積的藥液散開。它負責改善遞送，沒有改變抗體調動免疫反應的作用。 *圖2｜抗體阻斷PD-1路徑；玻尿酸酶暫時、局部增加皮下組織通透性，協助藥液擴散與吸收。機制示意，不按比例。* 能注射進去，還得證明換了路徑，身體仍能獲得合適的藥物暴露。 FDA核准依據中的377人隨機試驗，達到預設的藥物暴露量可比門檻；療效以描述性方式觀察，沒有顯示皮下版治癌更強。 開發換來的便利則很具體：由醫療人員操作，依療程約1或2分鐘完成注射，對照靜脈版約30分鐘的輸注。若同日仍需靜脈化療，省下的是其中一個環節，整次就醫不會縮成兩分鐘。 2024年2月，默沙東把原本非獨家的合作，改為皮下pembrolizumab的全球獨家授權，Alteogen並負責商業供酶。 這支酶的報酬，從此與重磅藥物的進展綁在一起。 默沙東2025年報列出，合作除了監管里程碑，還有合計最高10億美元的年度與累計銷售里程碑；全部銷售里程碑達成後，才開始支付淨銷售額低個位數百分比的權利金。 2025年美國與歐盟核准後，默沙東已實付4,000萬美元監管里程碑。 公司也判斷，未來銷售很可能觸發8.9億美元里程碑，因而認列相關負債。這筆8.9億美元當時尚未支付，仍待對應銷售條件達成。 Alteogen沒有自己持有pembrolizumab，卻能隨著它獲准、賣出、跨過門檻，取得相應報酬。技術公司不一定要再造一支藥王；解決藥王眼前的難題，也能形成可觀的生意。 對默沙東，幾個百分點又有多重要？ 這裡自建一個只看費率敏感度的情境：皮下產品年淨銷售額若為100億美元，權利金每差1個百分點，一年就是1億美元。這是量級演算，並非Qlex銷售預測或本案已實現的節費。 但完整成本還有前段里程碑、供酶費與付款時序。低個位數權利金是在全部銷售里程碑之後才開始；精確費率及兩家可比的完整條款，不能由一個區間還原。 費率有多重要，可以估算；默沙東當年如何權衡價格、供應與技術，公開資料沒有給出完整答案。 # 03｜合約談好了，第三方的專利仍在門外 默沙東與Alteogen可以談妥怎麼用技術、怎麼供酶、怎麼分潤。Halozyme提出的權利主張，卻不會隨著兩家公司簽約而消失。 Halozyme透過ENHANZE經營皮下遞送授權，但它明確表示，這次涉案的MDASE專利與ENHANZE的專利不同，也不在ENHANZE授權計畫內。 因此，這場官司並非已證實的「沒付ENHANZE使用費，被追討欠款」。要爭的是，默沙東使用的配方是否落入第三方專利的保護範圍。 Alteogen能授予自家技術的使用權，FDA能核准產品按條件用於治療；這些進展，沒有一併替第三方專利爭議結案。 而且，各地法庭並未給出同一個答案。 默沙東2026年第二季SEC文件披露指出，Halozyme曾在英國同意撤銷EP622的英國部分。10月荷蘭的結果，則對默沙東不利。 不能拿一地的進展宣布全球勝負，卻也不能等到所有官司結束，才開始計算商業影響。 荷蘭禁令碰到的MSD BV，是歐洲上市許可持有人，也是包裝配送樞紐。在公告所述受限市場，新劑型的推出安排因而受阻。 藥已經準備好，供應與銷售的路卻被限制。這個問題，會從法務文件一路走到醫院採購桌前。 # 04｜上市空間可以再爭，錯過的時間怎麼追回？ 靜脈版仍可供應，默沙東眼前的治療路徑沒有被這道禁令關掉。禁令的商業後果，也不能直接拿整個Keytruda營收來估。 但原有產品繼續賣，和新劑型及時站穩，是兩回事。 醫院換一個給藥流程，沒有藥廠換一張產品海報那麼快。 對不需要同日靜脈治療的合適病人，皮下注射可能帶來明顯便利；院方要把它做成日常服務，還得評估給付、採購、護理分工與排程，逐步累積使用經驗。 若供應在這段期間受限，這些工作便無法照原來的節奏展開。日後恢復上市，還得重新爭取進入醫院流程的機會；市場上的價格競爭，未必仍停在原地。 *圖3｜給藥更快是否轉為院端價值，取決於整次療程與工作流程；同日其他靜脈治療可能仍需停留。圖中情境不代表已實測的成本節省。* Qlex在2026年第二季已有4.63億美元銷售，約占Keytruda產品組合營收5.5%。這是收入占比，不是已有5.5%的病人完成轉換。 *圖4｜2026年第二季組合銷售83.66億美元已包含QLEX的4.63億美元。產品內部轉換不是可以重複加總的公司新增市場。來源：Merck，2026年8月4日。* 一位原本使用靜脈版的病人改用Qlex，可以讓皮下版收入增加，公司卻沒有因此多出第二位病人。 所以，這場保衛戰不能只看新劑型成長多快。還要問：付出技術成本之後，它保留了多少本來可能流失的需求？ 如果醫院只是換了產品名稱，流程沒有得到值得保留的改善，未來仍可能把價格放在最前面。只有當便利真正進入日常服務，皮下版才有條件爭取不同的選擇理由。 這也劃出了兩種策略的分工。改良遞送，是替既有生意增加競爭力；下一代新藥，則仍須創造新的治療價值。皮下版遇上官司，不會讓還在開發的新分子提早完成臨床。 默沙東得繼續爭取上市空間，也得面對採用時間正在流失的代價。 等產品能順利送進這些醫院時，院方會為這一針的便利，願意多付多少？ 本文提供產業資訊與商業分析，不構成個別化醫療或投資建議。 # 來源 1. Halozyme，2026-10-07，荷蘭禁令公告（訴訟當事人披露）；MDASE與ENHANZE的區別亦出自此文。 https://ir.halozyme.com/news/news-details/2026/Halozyme-Wins-Injunction-Stopping-Manufacture-and-Sale-of-Mercks-Keytruda-SC-in-Multiple-European-Markets/default.aspx 2. Merck，2025 Form 10-K，主要專利表及註10（印刷第14頁）；Alteogen合作、實付款與未來里程碑見Note 3。 https://www.merck.com/wp-content/uploads/sites/124/2026/05/MRK-12.31.2025-10K-FINAL.pdf https://www.sec.gov/Archives/edgar/data/310158/000031015826000063/R13.htm 3. Merck，2026-08-04，第二季財報與產品銷售表。 https://www.merck.com/news/merck-highlights-key-regulatory-and-clinical-milestones-across-broad-diverse-pipeline/ 4. FDA，2025-09-19，Qlex核准及377人試驗依據。 https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-pembrolizumab-and-berahyaluronidase-alfa-pmph-subcutaneous-injection 5. DailyMed，Qlex仿單第12.1節，PD-1及玻尿酸酶機制。 https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=097d166f-b73b-41d3-9b37-7653cd2a0c41 6. Merck，2025-09-19，Qlex核准公告、醫療人員給藥與注射時間。 https://www.merck.com/news/fda-approves-mercks-keytruda-qlex-pembrolizumab-and-berahyaluronidase-alfa-pmph-injection-for-subcutaneous-use-in-adults-across-most-solid-tumor-indications-for-keytruda-pem/ 7. Alteogen，2024-02-22，MSD獨家授權修約與商業供酶公告。 https://alteogen.com/en/sub/ir/news.php?bid=13&idx=163&mode=view 8. Merck，2026年第二季SEC法律事項披露，早於荷蘭10月公告。 https://www.sec.gov/Archives/edgar/data/310158/000031015826000212/R15.htm --- ## 兒少肥胖用藥：Novo 的先發優勢，為何還要靠 Wegovy 接棒？ URL: https://drugnews.com.tw/articles/2026-10-11-pediatric-obesity-novo-wegovy-lilly-product-strategy.html Date: 2026-10-11 Language: zh-Hant Summary: Novo 已有低齡證據與歐盟標籤，Saxenda 原品牌卻準備退出美國；Wegovy 的 STEP Young 與 Lilly 的共病、口服開發路徑，如何改變兒少肥胖用藥的產品競爭？ 圖1｜劑型不同，不一定換了靶點 Liraglutide、semaglutide與orforglipron活化GLP-1受體，tirzepatide則同時活化GLP-1與GIP受體；這是給藥方式與受體作用的比較，雙受體不代表已證兒科療效較佳。Semaglutide的每週注射標示僅指本文兒科注射劑，該分子另有成人口服產品。圖為作用機制示意，不是受體結構或人體神經迴路。Tirzepatide與orforglipron的兒少肥胖用途仍在研究。〔3、6、7、12〕 # 01　同一個孩子，指引與藥品標籤可能給出不同答案 一款藥已在歐盟核准用於符合條件的六歲兒童，全球照護指引卻不建議九歲以下用藥。兒少肥胖的競爭，就從這個落差開始。〔2、15〕 WHO 在2026年10月7日發布兒少肥胖照護指引。依公告，0–9歲不建議藥物治療；10–19歲則要在受監督的多模式生活介入未達預期後，才考慮已核准藥物。另一方面，歐盟 Saxenda（liraglutide）標籤已涵蓋6至未滿12歲、體重至少45公斤且符合肥胖條件的兒童。WHO 提供照護建議，EMA 評估產品核准，兩者作用不同，前者沒有撤銷後者的標籤。〔2、15〕 這讓 Novo Nordisk 的領先多了一道考題：它已有低齡藥物證據與歐盟標籤，但每週注射、已有青少年標籤的 semaglutide，還要取得自己的低齡核准。對手 Eli Lilly 則正在研究另一種切入方式：從合併代謝問題的青少年，以及每日口服選項進場。兩家的差距，要沿著產品、年齡與治療情境來看。〔1、2、11、13〕 # 02　Saxenda 開了路，原品牌卻準備離開美國 Saxenda 是每天注射的 GLP-1 受體促效劑，歐盟已把適用年齡向下延伸。但在美國，它的兒科肥胖標籤仍是12歲以上、體重超過60公斤；FDA 又在2025年8月核准 Teva 首款以 Saxenda 為參照的學名藥。Novo 的美國官網如今預告，原品牌最後一批批發商出貨將在2027年1月下旬。公司給出的理由是使用人數下降、已有更多核准替代治療，並明確排除安全或品質原因；公告沒有將此事歸因於兒科策略。〔2、3、4、5〕 把這些事件放在一起，值得追的就不只是「兒童能不能用 GLP-1」，還包括「哪一個產品能承接新增的臨床價值」。Liraglutide 的低齡證據可以繼續存在，Saxenda 在美國的品牌經營卻走向結束。對 Novo 而言，若要延續兒少肥胖的原廠產品優勢，每週注射的 semaglutide 就成為關鍵觀察點。這是從產品組合推得的商業課題，並非公司已宣布讓兒童轉藥。〔1、2、4、5、6〕 Wegovy 注射劑在美國已有12歲以上肥胖適應症；Novo 9月公布的 STEP Young，則把 semaglutide 的第三期研究推到6至未滿12歲。165名兒童參與，68週 BMI 變化的主要終點達標。這比拿成人減重數字推測兒童效果前進了一步，但截至10月11日，這個低齡用途仍屬研發中。〔1、6〕 公告中的40.4%，指的是在假設依從治療的分析下，semaglutide 組有這個比例在68週時降到依年齡與性別判定的肥胖門檻以下，安慰劑組為0%；兩組都有生活介入。它不是體重減少40.4%。完整平均 BMI 變化與停藥情形仍要看後續數據，才能判斷這個亮眼的分類改善有多容易在實際照護中重現。〔1〕 圖2｜40.4%到底量到了什麼 STEP Young共有165名6至未滿12歲兒童參與；兩組均接受生活介入。第68週低於肥胖分類門檻的估計比例為40.4%與0%，採假設所有兒童遵從治療的分析，並非體重減少幅度。低於門檻仍可能屬過重；各組人數與此比例的分析分母未公布，不能用165直接換算反應人數。這是2026年9月7日公司初步結果中的分類指標，不是主要BMI變化終點的數值。〔1〕 如果完整資料與後續核准支持使用，Novo 有機會把既有青少年產品經驗延伸到更低年齡，減少重新建立產品認識的成本。這項時間優勢面臨兩面壓力：Teva 的學名藥在美國提供另一個競爭位置，Lilly 的新方案則可能改變選藥理由。前者能否形成價格或可及性優勢，取決於實際價格、供應與給付；目前尚無足夠資料判定這股壓力有多大。〔1、4、6、11、13〕 圖3｜兒少肥胖：兩家藥廠，四條產品路徑 本圖選取EU／US代表性標籤，不是完整核准清單；灰格表示未展開低齡研究，並非沒有計畫。STEP Young為6至未滿12歲第三期初步結果，低齡用途仍在研發。ADOLESCENTS-2要求肥胖且至少兩項指定共病：高血壓、血脂異常或糖尿病前期，並排除第二型糖尿病；另有SURMOUNT-ADOLESCENTS第三期研究。圖中ADOLESCENTS-2與orforglipron青少年試驗皆要求已進入青春期。Mounjaro美國10歲以上第二型糖尿病核准不等於兒科肥胖核准，未放入本矩陣。這是開發定位比較，不是療效或安全性排名。資料截至2026年10月11日。〔1、2、3、6、7、8、9、11、12、13〕 # 03　Lilly 的兩張牌：共病與口服 Lilly 並非沒有兒科資產：Mounjaro（tirzepatide）在美國已核准用於10歲以上第二型糖尿病。不過，糖尿病標籤不能代替肥胖標籤；Zepbound 的美國兒科安全性與有效性仍未確立。用成人競爭位置直接判定誰會贏得兒少肥胖，會漏掉這道適應症邊界。〔7、8〕 Tirzepatide 的 SURMOUNT-ADOLESCENTS 已完成收案；另一項第三期 SURMOUNT-ADOLESCENTS-2，則鎖定12–17歲、肥胖並至少合併兩項體重相關共病的族群，包括高血壓、血脂異常或糖尿病前期，並排除已罹患第二型糖尿病者。它要觀察的不只體重，還有心血管風險因子。這種設計若得到有臨床意義的結果，Lilly 就能圍繞「多個健康問題一起處理」建立選藥理由；它目前還不是已證實降低心血管事件的兒科療法。〔9、11〕 另一張牌是每日口服 GLP-1 藥物 orforglipron。它在美國以 Foundayo 用於成人；12–17歲的第三期研究另行進行，排除尚未進入青春期者。若兒科結果與核准成立，家庭可能在每週打針與每天服藥之間多一個選擇。給藥負擔會改變，但口服是否更能持續使用，需要實際資料；而且 Qsymia 早在2022年就取得美國12歲以上肥胖適應症，兒少口服治療並非從 Lilly 才開始。〔12、13、14〕 這兩條路，讓 Lilly 有機會避開只追趕年齡標籤的競爭：前一條要證明在多共病青少年中，BMI 以外能增加哪些治療價值；後一條要證明另一種給藥方式值得選擇。如果只有 BMI 改善，卻沒有足以影響臨床選擇的共病結果、耐受性或使用體驗，差異化就會縮小。Novo 因而可能有更多時間累積兒科產品經驗；反過來，真正有用的兒科差異也可能讓先進場者重新面對選藥競爭。〔11、13〕 圖4｜證據還在擴張，原品牌卻走向退場 歐盟liraglutide低齡標籤、美國學名藥核准與原品牌退場，以及semaglutide低齡新證據，是不同地區與階段的事件；連線只表時間先後。Novo所述美國Saxenda退場原因是使用人數下降與更多核准替代治療。2027年1月下旬為預定最後批發出貨，不代表全美藥局庫存當月歸零或全球liraglutide退出。新產品承接臨床價值是商業推論，並非已公布的兒童轉藥計畫。STEP Young詳細結果預定於2026年11月14–17日ObesityWeek發表。資料截至2026年10月11日。〔1、2、4、5、6〕 # 04　下一個催化劑，要看能否完成接棒 最近的檢驗點是2026年11月14–17日 ObesityWeek，Novo 預告將公布 STEP Young 詳細結果。最值得看的是平均 BMI 變化、把停藥納入後的效果、停藥原因，以及生長與青春期資料。若理想依從情境與納入停藥的分析差距很大，先發優勢可能停留在試驗新聞，難以轉成穩定的治療選擇。〔1〕 另一個時點是2027年1月下旬的美國 Saxenda 最後批發出貨。原品牌淡出之後，學名藥能否承接供應、新產品能否建立兒科位置，會分別影響市場留下哪些選擇。Lilly 的2026年第二季簡報則把 ADOLESCENTS-2 與 orforglipron 青少年試驗的主要完成時間，列在2027年10月與12月；這是開發時程，並非承諾在當月公布結果或上市。〔5、10〕 兒少肥胖這一局，Novo 已拿到比較具體的低齡證據位置，但維持原廠產品優勢仍要靠 semaglutide 把接棒做成。Lilly 的機會則在於，用共病或口服資料改變醫師與家庭選擇治療的理由。成人市場的強弱不會自動搬過來：在兒少，優勢要逐個年齡、逐個適應症、逐個產品重新建立。〔1、2、4、5、6、11、13〕 資料截至2026年10月11日。 --- ## From a Daily Tablet to Factory Orders: Who Can Profit from Oral Cyclic Peptides? URL: https://drugnews.com.tw/articles/2026-10-05-cyclic-peptide-upstream-2026-en.html Date: 2026-10-05 Language: en Summary: On September 17, 2026, Novo Nordisk and Orbis Medicines announced a collaboration to develop oral macrocycles for cardiometabolic diseases. Upfront payments and potential development and commercial milestones total up to US$1.4 billion, with additional sales royalties; Novo will also make a strategic investment in Orbis. A large drugmaker is buying opportunities to discover new medicines. That headline ceiling is not yet an order held by a raw-material factory, and the announcement does not disclose commercial purchase volumes or suppliers. On September 17, 2026, Novo Nordisk and Orbis Medicines announced a collaboration to develop oral macrocycles for cardiometabolic diseases. Upfront payments and potential development and commercial milestones total up to US$1.4 billion, with additional sales royalties; Novo will also make a strategic investment in Orbis. A large drugmaker is buying opportunities to discover new medicines. That headline ceiling is not yet an order held by a raw-material factory, and the announcement does not disclose commercial purchase volumes or suppliers. At the product end, Merck’s oral cyclic peptide enlicitide, marketed as Lipfendra, received U.S. FDA approval on July 15 to lower LDL cholesterol in adults with hypercholesterolemia. It provides a defined product and dose for analysis; commercial uptake and cardiovascular-event benefits remain separate questions. Making a medicine oral may appear to create more raw-material business, but the manufacturing process can change that calculation first. Enlicitide research has already demonstrated fewer assembly steps, potentially changing both materials and working time per kilogram of acceptable product. The question worth following is who can evolve with the new route, not simply who owns a raw-material plant. # One Tablet a Day: First Calculate What the Finished Product Contains Enlicitide binds PCSK9, preventing it from promoting degradation of LDL receptors in the liver and allowing more receptors to participate in cholesterol clearance. This illustrates how macrocycles can target interactions between proteins. For upstream analysis, the approved product’s dose provides a concrete starting point for quantification. The U.S. label specifies one tablet containing 20 mg of enlicitide daily. Assuming 100,000 people use it continuously for 30 days, the calculation is 20 mg × 30 days × 100,000 people = 60 kg of active ingredient. The population and duration form a transparent demand scenario, not a company sales forecast. The label contains another conversion: 20 mg of enlicitide is equivalent to 22.21 mg of its decanoate salt. The same scenario therefore corresponds to 66.63 kg on a salt-form basis. These are two mass expressions of the same dose and must not be added together. The approved oral dose already specifies the amount actually administered; it does not require another bioavailability adjustment. Moving from that quantity to factory purchasing requires the actual route, individual-step yields, purification recovery and formulation losses. The active pharmaceutical ingredient (API) provides the pharmacological activity, while a finished tablet also contains excipients and involves processing. They are different cost categories. The demand scenario completes only the first layer of calculation. The number of users and persistence determine finished-product requirements; the manufacturing method determines material inputs and equipment occupancy. The finished medicine’s price also includes R&D, commercialization and other costs. Taking a fixed percentage of it does not establish upstream orders. *Figure 1 | Assuming 100,000 people treated for 30 days, 20 mg of enlicitide daily equals 22.21 mg of its decanoate salt: 60 kg of active ingredient equals 66.63 kg on a salt basis. These express the same dose. * # Closing a Chain into a Ring Is Not Just a Final-Step Problem The enlicitide synthesis paper published in the Journal of the American Chemical Society in 2025 revealed a structure considerably more complex than joining a few amino acids: six of its eight amino acids are noncanonical, and it also contains substantial nonpeptidic structural elements. These specialized building blocks bring supply-chain questions into development early. Process designers must decide which materials to make internally and which to outsource. Suppliers must control stereochemical configuration and impurities, not merely deliver a compound. Even with the same atom connectivity, a different spatial arrangement can produce stereoisomers that must be controlled. If batch quality varies, a deviation may become apparent only after assembly, when earlier materials and processing time have already been consumed. The paper’s early route used protecting groups to temporarily mask reactive sites, prevent unwanted connections and then expose those sites when needed. Every additional operation creates another opportunity for transfer, separation or material loss. Individual steps may look acceptable while cumulative recovery across a long route becomes the factory’s real burden. Cyclization poses another challenge. The aim is to connect positions within the same molecule, but they may instead react with another molecule and form unwanted larger products. The study’s early Northern-fragment ring closure used high dilution to reduce such oligomerization—a choice specific to the precursor and route. If obtaining the same amount of product requires more solvent per batch, reactor volume, solvent recovery and batch time are all consumed. However large the plant, capacity must ultimately be translated into kilograms of acceptable product deliverable each month—not simply liters of installed equipment. Manufacturing competition is therefore not just a race to expand capacity. A shorter route to the same product could save batch time and free space in existing equipment. A supplier participating in those improvements offers more than kilograms of a material: it can help make the customer's production schedule more reliable. Whether the supplier retains that value still depends on service scope and contractual allocation. *Figure 2 | An assembled precursor can undergo intramolecular cyclization or form intermolecular oligomers. High dilution in the early route reduced the latter at the cost of solvent and equipment occupancy. * # An Enzymatic Route May Change Material Demand and Bargaining Positions Enlicitide process research published in Science in May 2026 used engineered enzymes for fragment formation, coupling and macrocyclization. These assembly steps did not rely on protecting groups, and crystallization replaced chromatography. The number of steps was reduced by more than half relative to previous methods. Chromatography separates components of a mixture through a separation system; crystallization uses suitable conditions to form an isolable solid product. Whether one can replace the other depends on the molecule and its impurities. When it is feasible, production scheduling, solvent handling and recovery can all change. For upstream investors, the purchasing mix behind each kilogram of acceptable API may be rewritten. If a route reduces the use of a particular protected intermediate, end-market growth will not flow proportionally into that material. Fewer steps may also allow more product to be delivered from existing equipment, reducing the need for expansion. A supplier limited to building blocks for an older route therefore faces different bargaining conditions from one participating in enzyme, reaction and purification optimization. The former depends on continued use of its material; the latter can create value by improving deliverable output and managing batch failure and switching costs. Halving the steps still does not mean halving costs. Enzyme development, production and downstream processing carry costs, and the remaining steps may concentrate much of the expense. The material cited here is insufficient to estimate commercial cost per kilogram. What can be established is progress in process methods and the need to revisit earlier assumptions about material consumption. *Figure 3 | Engineered enzymes support protecting-group-free fragment formation, coupling and macrocyclization, alongside crystallization. Steps are reduced by more than half relative to prior methods; route changes may affect material demand. * # Once Capacity Exists, Why Would a Customer Choose It? An initial hurdle for an order is whether the material meets the product’s specifications. The FDA-published ICH Q7 API guidance recommends systems for supplier evaluation, material specifications, quality release and process validation. The document explicitly states that its recommendations are not independently legally binding; actual obligations depend on applicable law and filing commitments. The practical objective of quality management is consistent delivery of acceptable product. For an API manufacturer, producing the correct main ingredient is only the starting point. Which impurities arise, whether analytical methods can identify them and whether scale-up remains consistent all influence the customer’s assessment. Process reproducibility lies between an attractive laboratory yield and sustained delivery on a contractual schedule. A drug developer also may not outsource everything. Early work could involve purchasing specialized building blocks, commissioning process development or ordering small clinical batches, followed later by decisions about internal production, outsourcing or multiple sources. Development fees, batch-manufacturing revenue and long-term purchase contracts represent different degrees of demand commitment; they should not be collapsed into a single confirmed commercial order. Even after a customer adopts a supplier, the contractual commitment matters. Forecast demand, reserved capacity and committed purchase volume are not identical. If equipment is installed without corresponding commitments and end-market sales fall short, the upstream company may bear the idle-capacity cost first. Once a customer has validated a route and source, switching requires reassessment and can create some stickiness. But more qualified suppliers or an improved process at the originator can put pressure on prices. What manufacturers need to preserve is bargaining power supported by delivery capability, not simply more kilograms passing through the revenue line. # In Taiwan, Look First at Deliverable Capabilities, Then at Product-Specific Adoption ScinoPharm Taiwan (1789) lists solid-phase, solution-phase and hybrid synthesis on its official peptide-services page, alongside process optimization, analytical-method validation, clinical material manufactured under current GMP and commercial production. These are concrete capabilities adjacent to cyclic-peptide manufacturing needs. That service page does not disclose supply of enlicitide, participation in the Orbis–Novo collaboration or commercial delivery using the cyclization route discussed here. If it participates in development services, value initially lies in solving route and analytical problems. If it moves into supply, product qualification, acceptable batches and contractual terms are then needed to establish recurring revenue. *Figure 4 | Product demand, delivery and quality, supplier qualification by customers and supply commitments are distinct commercial conditions. Development services and material supply require separate consideration of revenue terms. * As drugmakers try to put complex molecules into a daily tablet, factories could move from selling materials toward solving route and delivery problems. For an upstream company, the next announcement with real weight is not merely another piece of equipment. It is which process, which specification and which batches of acceptable product a customer is willing to entrust to it on an ongoing basis. This article provides industry information and business analysis and does not constitute individualized medical or investment advice. # Sources [Orbis Medicines: Multi-target drug-discovery partnership with Novo Nordisk — 2026-09-17](https://www.orbismedicines.com/news/orbis-medicines-enters-multi-target-drug-discovery-partnership-with-novo-nordisk-worth-up-to-usd-14-billion) [FDA: Lipfendra approval letter, NDA 220848 — 2026-07-15](https://www.accessdata.fda.gov/drugsatfda_docs/appletter/2026/220848Orig1s000ltr.pdf) [FDA: Lipfendra prescribing information — Revised July 2026](https://www.accessdata.fda.gov/drugsatfda_docs/label/2026/220848Orig1s000lbl.pdf) [Journal of the American Chemical Society: Total Synthesis of Enlicitide Decanoate — 2025-03-24](https://doi.org/10.1021/jacs.4c15966) [Science: Biocatalytic cascades enable manufacture of the macrocyclic peptide enlicitide — 2026-05-07](https://pubmed.ncbi.nlm.nih.gov/42096573/) [FDA/ICH Q7: Good Manufacturing Practice Guidance for Active Pharmaceutical Ingredients — September 2016, Revision 1](https://www.fda.gov/media/71518/download) [ScinoPharm Taiwan: Peptide synthesis and process-development services — Undated service page](https://www.scinopharm.com/tw/products-detail/54/) [Merck: FDA approval of Lipfendra to reduce LDL cholesterol — 2026-07-16](https://www.merck.com/news/mercks-lipfendra-enlicitide-is-the-first-and-only-once-daily-oral-pcsk9-inhibitor-approved-by-the-u-s-fda-to-reduce-ldl-c-in-adults-with-hypercholesterolemia/) [S4 full-paper public mirror on the Aarhus University server](https://maillist.au.dk/hyperkitty/list/hauser.chem%40maillist.au.dk/message/RWSJBSY53CCDATJCHCKPUXW4QBJVKMR3/attachment/6/total-synthesis-of-enlicitide-decanoate.pdf) --- ## 從每天一顆藥到工廠訂單：口服環肽的錢，誰賺得到？ URL: https://drugnews.com.tw/articles/2026-10-05-cyclic-peptide-upstream-2026.html Date: 2026-10-05 Language: zh-Hant Summary: 2026年9月17日，諾和諾德（Novo Nordisk）與Orbis Medicines宣布合作，開發心血管代謝疾病的口服大環肽。前期款與可能達成的開發、商業里程碑合計最高14億美元，另計銷售權利金，Novo亦將策略投資Orbis。大藥廠正在買新的藥物發現機會；這筆上限金額，還不是原料工廠手上的訂單，公告也未披露商業採購量或供應商。 2026年9月17日，諾和諾德（Novo Nordisk）與Orbis Medicines宣布合作，開發心血管代謝疾病的口服大環肽。前期款與可能達成的開發、商業里程碑合計最高14億美元，另計銷售權利金，Novo亦將策略投資Orbis。 大藥廠正在買新的藥物發現機會；這筆上限金額，還不是原料工廠手上的訂單，公告也未披露商業採購量或供應商。 另一端，Merck的口服環肽enlicitide，商品名Lipfendra，已在7月15日獲美國FDA核准，用於降低成人高膽固醇血症患者的低密度脂蛋白膽固醇。它提供了可核的產品與劑量；實際放量及心血管事件效益則是另外的問題。 一顆藥走向口服，看起來會帶動更多原料生意，製程卻可能先把這個算式改掉。Enlicitide的研究已展示如何減少組裝步驟；同樣一公斤合格產品，材料與工時都可能不同。真正值得追的，是誰能跟著新路線一起進化，而不只是誰有一座原料工廠。 # 每天一顆，先算成品需要多少 Enlicitide藉由結合PCSK9，阻止它促使肝臟的低密度脂蛋白受體降解，讓更多受體參與清除膽固醇。這說明大環肽可以瞄準蛋白質之間的作用；對上游而言，真正能開始計量的是正式產品的劑量。 依美國標籤，每日一顆含20毫克enlicitide的錠劑。假設有10萬人連續使用30天，用量就是20毫克×30天×10萬人＝60公斤有效成分。10萬人與30天只是透明的需求情境，不是公司銷售預測。 標籤還有一個換算：20毫克enlicitide相當於22.21毫克的癸酸鹽形式，因此同一情境的鹽形式含量相當於66.63公斤。這是同一劑量的不同質量口徑，兩者不能相加。 正式口服劑量已是產品實際使用的量，不必再按生物利用度放大。從這裡推到工廠採購，才需要加入實際路線、各步收率、純化回收與製劑損耗。原料藥（API）是發揮藥理作用的成分；最終錠劑還包含賦形劑與加工，兩者也不是同一成本項目。 需求情境只完成第一層計算。用藥人數與持續使用時間決定成品需要多少，製造方法才決定要投入多少材料、占用多少設備。終端藥價還包含研發、商業化與其他成本，不能直接抽一個固定百分比當作上游訂單。 圖1｜假設10萬人服藥30天：每日20毫克enlicitide＝22.21毫克癸酸鹽，對應60公斤有效成分＝66.63公斤鹽形式；兩者是同一劑量的質量換算。 # 把一條鏈關成環，難處不只在最後那一步 2025年《美國化學會誌》的enlicitide合成論文，揭露了比「幾個胺基酸串起來」複雜得多的結構：八個胺基酸中有六個屬非標準胺基酸，另含重要的非胜肽結構。 這些特殊組件讓供應鏈問題提早出現。製程設計者得決定哪些原料自行製備、哪些委外；供應商除了交出化合物，還得控制立體構型與雜質。即使原子連接方式相同，空間排列不同，也可能形成必須控制的立體異構物。若原料批次品質不穩，下游可能直到組裝後才發現偏差，前面的材料與時間已經消耗。 論文的早期路線使用保護基，暫時遮住某些反應位置，避免接錯地方，再於需要時移除。每多一道處理，就多一次轉移、分離或損耗的機會；單步看似可接受，串成長路線後，累積回收才是工廠要承擔的結果。 環化還有另一個難題：希望同一分子的兩端接起來，卻可能接到別的分子，形成不需要的較大產物。該研究早期北端片段的關環，採高稀釋條件降低這類寡聚化；這是特定前驅物及路線的選擇。 若每批需要更多溶劑才能取得同樣產物，反應槽容積、溶劑回收與批次時間都會被占用。工廠再大，也要換算成每月能交付多少合格公斤，而不只是設備有幾公升。 也因此，製造競爭不只是一場擴產競賽。能把同樣的產品做在較短路線裡，可能讓客戶省下批次時間，也可能讓既有設備釋出空間。供應商若參與這種改進，賣的就不只是某公斤材料，而是客戶更可靠的生產安排；價值能否留下來，還要看服務範圍與合約如何分配。 圖2｜已連接的前驅物可以進行分子內環化，也可能形成分子間寡聚物；早期路線以高稀釋降低後者，代價是溶劑與設備占用。 # 酵素路線改變的，可能是材料需求與議價位置 2026年5月《Science》的enlicitide製程研究，用工程化酵素進行片段形成、連接與大環化。這些組裝步驟不依賴保護基，再以結晶取代層析；相較先前方法，步驟數減少超過一半。 層析像是把混合物沿分離系統拆開；結晶則利用條件讓目標物形成可分離的固體。能不能替換，取決於分子與雜質的性質。一旦可行，生產排程、溶劑處理及回收方式都可能改變。 對上游投資人，重點是每公斤合格API背後的採購組合可能被重寫。路線若減少某種受保護中間體的使用，終端成長不會等比例流向那項材料；節省的步驟也可能讓同一設備交付更多產物，降低擴建需求。 這使單純供應舊路線組件，與能參與酵素、反應及純化優化的廠商，面臨不同的議價條件。前者依賴原材料持續被使用，後者的價值則在於替客戶提高可交付產量、控制批次失敗與轉換成本。 步驟減半仍不等於成本減半。酵素開發、製備與後處理都有成本，剩餘步驟也可能集中主要費用。目前引用的資料不足以推算商業每公斤成本；能確認的是製程方法進步，以及它要求投資人重新檢查原來的耗料假設。 圖3｜工程化酵素支援無保護基的片段形成、連接與大環化，搭配結晶分離，步驟較先前方法減少逾半；路線改變可能影響材料需求。 # 有產能之後，客戶為什麼選你？ 訂單的第一個門檻，是材料能否符合特定產品的規格。FDA公布的ICH Q7提供原料藥GMP指引，建議建立供應商評估、材料規格、品質放行與製程驗證制度。文件明示建議本身不具獨立法律約束力，實際義務仍依適用法規與申報承諾；品質管理的核心，是能否持續交出合格產品。 對原料藥廠，合成出正確主成分只是起點。哪些雜質會一起形成、分析方法能否辨認、放大後是否仍穩定，都會影響客戶的評估。一次漂亮的實驗收率，與可依合約排程持續交貨之間，還有製程重現性要建立。 藥廠也未必把全部工作外包。早期可能購買特殊組件、委託製程開發或小量臨床用料，後期再評估自製、委外或多來源配置。開發服務費、批次製造收入與長期採購合約，代表不同程度的需求承諾，不能合成一筆確定的商業訂單。 即使獲得客戶採用，也要看合約承諾了什麼。預估需求、產能預留與確定採購量並不相同；若先投入設備卻沒有相應承諾，終端銷售落後預期時，閒置成本可能先留在上游。 客戶驗證過一條路線與來源後，切換需要重新評估，會形成一定黏著度。但更多供應商合格、或原廠改善路線，都可能帶來價格壓力。製造商要保住的是交付能力所帶來的議價空間，而不只是營收帳上增加的公斤數。 # 台灣先看可交付能力，再看特定產品採用 台灣神隆（1789）的官方胜肽服務頁列有固相、液相及混合式合成，以及製程優化、分析方法確效、符合現行優良製造規範的臨床用原料與商業量產。 這是與環肽上游需求相鄰的具體能力；該服務頁未披露供應enlicitide或參與Orbis／Novo合作，也未列出本文所述環化路線的商業交付紀錄。 若參與開發服務，價值先落在解決路線與分析問題；若進一步承接供應，才需要用產品資格、合格批次與合約條件確認持續收入。 圖4｜產品需求、交付與品質、客戶對供應商的資格認證及供貨承諾，是不同商業條件；研發服務與材料供應的收費安排需分開看。 當藥廠想把複雜分子裝進每天一顆的錠劑，工廠也有機會從材料賣方，走向路線與交付問題的解決者。對上游企業，下一份真正有分量的消息，不只是又多了一項設備，而是客戶願意把哪一段製程、哪一種規格與哪一批合格產品，交給它持續完成。 本文提供產業資訊與商業分析，不構成個別化醫療或投資建議。 --- ## Migraine Around Your Period? How Much Can a Seven-Day Prevention Strategy Help? URL: https://drugnews.com.tw/articles/2026-10-04-luna-atogepant-menstrual-migraine-2026-en.html Date: 2026-10-04 Language: en Summary: The days around menstruation can already be uncomfortable. Add migraine, sensitivity to light or nausea, and work, school pickups and carefully planned activities can be disrupted. For people whose attacks repeatedly cluster in this window, the question is straightforward: can the burden be reduced before the headache arrives? The days around menstruation can already be uncomfortable. Add migraine, sensitivity to light or nausea, and work, school pickups and carefully planned activities can be disrupted. For people whose attacks repeatedly cluster in this window, the question is straightforward: can the burden be reduced before the headache arrives? On September 10, 2026, AbbVie announced the initial main results, or topline findings, from its Phase 3 LUNA study. The study enrolled 468 adult women and compared seven-day short-term prevention with placebo across three menstrual cycles. Perimenstrual migraine days decreased from baseline by an average of 1.20 days versus 0.40 days, a between-group difference of 0.80 fewer days, with P<0.0001. The company also reported that secondary endpoints including functional disability were met. LUNA's appeal is to put prevention into a few days that patients can plan for: intervening before a high-risk window rather than waiting for headache to disrupt their plans. For AbbVie, that is also a route for an existing medicine to enter a new use case. # Predictable Timing Creates Another Opportunity for Prevention The LUNA study drug, atogepant, is the active ingredient in QULIPTA in the United States and an oral small-molecule CGRP receptor antagonist. CGRP stands for calcitonin gene-related peptide, a signaling molecule involved in migraine. Receptor antagonism can be understood as occupying the signal’s receiving site and reducing its ability to activate that pathway. Human research offers a clue to the relationship between the menstrual cycle and this signaling pathway. In 2023, a Charité research team compared women in different hormonal states. Among those with natural menstrual cycles, women with migraine had higher CGRP concentrations during menstruation than controls without migraine. This is evidence of an association between hormonal status and CGRP, not an established single cause of every headache around menstruation. LUNA targets a period when attack risk is more concentrated by intervening in migraine-related signaling. Its research strategy does not alter menstruation itself or use hormone replacement. Short-term prevention also has a history. A randomized study published in 2004 evaluated frovatriptan for intermittent prevention of menstrually associated migraine. LUNA adds Phase 3 controlled evidence for atogepant’s CGRP-receptor approach; it did not directly compare different short-term preventive medicines. This is neither a new molecule nor the first way to address menstrual migraine. What changes is the evidence and product positioning: the same medicine moves from general migraine prevention toward a predictable period of often-overlooked need. The patient's question moves beyond whether to use prevention to whether it can be concentrated in the period when it is most needed. *Figure 1 | LUNA schedules prevention around a predictable menstrual window and separately evaluates migraine days and functional burden. * # Seven Days Describes Dosing; the Perimenstrual Period Defines the Efficacy Window The randomized, double-blind, placebo-controlled study enrolled adult women with pure menstrual migraine or menstrually related migraine across Europe and Asia. Before enrollment, menstrual regularity and attack patterns were assessed using medical history and electronic diaries. Chronic migraine and certain other conditions were excluded. The Japanese public registry also lists Taiwan among the study countries. This is a direct research connection to the trial; the page does not provide Taiwan’s actual enrollment or regional efficacy, nor establish local approval or NHI reimbursement. If cycles fluctuate substantially or headaches occur on many days throughout the month, a strategy targeting a few days may not transfer directly. The trial used diaries to identify regularity, and actual use also requires an understanding of attack patterns. When to start matters alongside which medicine is selected in determining whether this regimen can be carried out. The study regimen was atogepant 60 mg or placebo once daily for seven consecutive days, beginning three days before anticipated menstruation and repeated across three cycles during the double-blind period. The primary endpoint counts migraine days during the perimenstrual period (PMP), defined in the registry as Day −2 through Day +3 relative to menstrual onset. Dosing is scheduled in advance using anticipated timing, while the PMP defines the assessment period. They are not the same window. For people who also have attacks at other times, whole-month burden remains another part of the treatment decision. *Figure 2 | Dosing starts three days before anticipated menstruation and continues for seven days. The perimenstrual period (PMP) assessment is separately defined as Day −2 through Day +3 relative to menstrual onset. The windows are shown separately and not to scale. * # An Additional 0.80-Day Reduction: Function Has a Positive Signal, but Magnitude Matters The primary result compares the change from baseline in PMP migraine days averaged across three menstrual cycles during the double-blind period. Accordingly, 1.20 and 0.40 days are average reductions. The 0.80-day difference is between the groups, not a three-cycle total, a count of remaining days after treatment or an identical benefit for each participant. A difference of 0.80 days may sound small, but it belongs within a limited high-risk period, not a direct comparison with headache days across a whole month. Different levels of improvement can contribute to that average: some patients may benefit more, others less. Complete baseline data, confidence intervals and responder distributions would give the average a patient-level context. The P value supports a statistical difference; it does not answer whether treatment is worthwhile for a patient. AbbVie reported that all eight ranked secondary endpoints were also met, covering headache burden, acute medication use, functional disability, cognitive function and responder outcomes. This already represents a company-reported signal of functional benefit. What the announcement does not provide is the complete effect size for each endpoint; the results cite company data on file. The registered Functional Disability Scale (FDS) endpoint helps explain what was measured. Using PMP data across three menstrual cycles, it assesses the proportion of participants meeting the criterion of no disability or only mild impairment on every PMP day. The question is whether an entire high-risk period remains at a lower level of functional burden, not merely how many migraine days are avoided. That measure is close to daily life, but it does not directly count days off work avoided or days of caregiving restored. Reporting the proportions meeting the criterion in each group and their difference would show the size of the company’s positive signal. The acute-medication endpoint separately measures days of use during the PMP. *Figure 3 | Migraine days during the perimenstrual periods of three cycles decreased from baseline by an average of 1.20 versus 0.40 days, a between-group difference of 0.80 days, P<0.0001. PMP names the assessment period. * # Repeated Short-Term Use Still Requires Efficacy and Tolerability to Be Considered Together AbbVie reported no new safety signal in LUNA and described the findings as consistent with atogepant’s established profile. Assessing repeated short-term use also requires the type, severity and duration of adverse events and any treatment discontinuations. These determine how much benefit remains worthwhile to patients after the burden of treatment is considered. The U.S. QULIPTA label lists nausea, constipation and fatigue or somnolence among common adverse reactions, with warnings concerning hypersensitivity reactions, hypertension and Raynaud’s phenomenon. These are background data for the existing product, not LUNA’s event rates. Fewer dosing days do not establish a proportional reduction in risk. The U.S. label cited here covers migraine prevention in adults and does not include LUNA’s seven-day regimen or as-needed rescue use of atogepant. AbbVie’s September 10 announcement described plans to submit data, not a completed filing or a new approval. The dosing described above is the trial protocol; individual treatment requires clinical assessment. Short-term does not mean one-off. If the regimen becomes part of daily life, the patient will face the same choice in the next cycle: was the previous improvement worth taking the medicine again? The full report's handling of missing diary entries and withdrawals also addresses that question; it cannot retain only people who completed dosing successfully. An adoptable regimen must account for benefit, physical burden and difficulty carrying it out together. *Figure 4 | The company has reported a positive functional-endpoint result. Complete effect sizes, adverse events and tolerability, continued use and regulatory information will further define the value of the short-term regimen. * # AbbVie Has an Evidence-Supported Use Case, Not Yet a Whole New Mass Market LUNA gives AbbVie a product entry point supported by positive Phase 3 results: patients with clustered attacks who can plan ahead but have not received adequate prevention. That is more meaningful than counting all women in a market estimate. It places the need in a use case that physicians can identify and patients can describe. That positioning involves a trade-off. People with regular cycles, clustered attacks and previously inadequate prevention could represent new use. Those already on a preventive regimen may instead be switching. Physicians need to consider prior response and headache burden over the entire month; fewer dosing days alone do not make a seven-day regimen a natural replacement for existing treatment. For AbbVie, an established molecule and brand offer clinical experience and product familiarity that may support an additional use case. Yet changing an existing patient to short-term use and treating an additional previously untreated patient have different revenue implications. There is another commercial possibility: for patients previously treated only after an attack, a choice aimed at a predictable window could make prevention easier to discuss in the clinic. This is a route to expand addressable demand rather than simply increase dosing days. Education, identifying suitable patients and their willingness to return may influence adoption alongside the medicine. An established brand can provide an entry point, but new treatment experience must still retain those patients. Seven treatment days per cycle, the number of cycles treated and the clinical settings from which new users come will jointly determine medication volume. Even with more patients, annual use per person may differ. Until the user mix and reimbursement terms are clearer, directly raising annual revenue assumptions would be premature. For payers, the proportion meeting the FDS criterion and days of acute-medication use move the discussion from the mean reduction in migraine days toward whether the high-risk period can remain at a lower burden. The difference between groups in functional responders would be more persuasive than repeating the P value. If that improvement can persist across cycles with acceptable tolerability, AbbVie could turn a positive result into a regimen patients choose again in the following cycle. This article provides industry information and business analysis and does not constitute individualized medical or investment advice. # Sources [AbbVie: Positive Phase 3 atogepant results in menstrual migraine — 2026-09-10](https://news.abbvie.com/2026-09-10-AbbVie-Extends-Migraine-Leadership-with-Positive-Phase-3-Atogepant-Results-in-Menstrual-Migraine) [QULIPTA U.S. Prescribing Information — Revised September 2025](https://www.rxabbvie.com/pdf/qulipta_pi.pdf) [ClinicalTrials.gov: NCT06806293 LUNA study record — Study registration](https://clinicaltrials.gov/study/NCT06806293) [Japan jRCT: jRCT2051240268 study record — Study registration](https://jrct.mhlw.go.jp/latest-detail/jRCT2051240268) [Japan Ministry of Health, Labour and Welfare: LUNA clinical-trial information portal — Record updated 2026-09-27](https://rctportal.mhlw.go.jp/s/detail/um?trial_id=jRCT2051240268) [Charité: Why migraine frequently occurs during menstruation — 2023-02-23](https://mentalhealth.charite.de/en/metas/en_meldung/artikel/detail/why_migraine_frequently_occurs_during_menstruation/) [A randomized trial of frovatriptan for intermittent prevention of menstrual migraine — 2004-07-27](https://pubmed.ncbi.nlm.nih.gov/15277618/) [AbbVie: M24-859 LUNA study overview — Undated study overview](https://www.abbvieclinicaltrials.com/study/M24-859) --- ## 偏頭痛總挑月經前後來？7天預防研究，能少痛多少？ URL: https://drugnews.com.tw/articles/2026-10-04-luna-atogepant-menstrual-migraine-2026.html Date: 2026-10-04 Language: zh-Hant Summary: 月經前後已經不好受，再碰上偏頭痛、怕光或噁心，工作、接送與原本排好的行程都可能被打亂。對發作時間反覆落在這個窗口的人，問題很直接：能不能在頭痛來之前，先降低那幾天的負擔？ 月經前後已經不好受，再碰上偏頭痛、怕光或噁心，工作、接送與原本排好的行程都可能被打亂。對發作時間反覆落在這個窗口的人，問題很直接：能不能在頭痛來之前，先降低那幾天的負擔？ 2026年9月10日，AbbVie公布LUNA第三期研究的初步主要結果（topline）。研究納入468名成年女性，在三個月經週期比較7天短程預防與安慰劑；圍月經期偏頭痛日數相對基線平均減少1.20天對0.40天，組間多減少0.80天，P<0.0001。 公司也報告功能障礙等次要終點達標。LUNA的吸引力，在於把預防放進患者能預先安排的幾天：不是等頭痛打亂行程才處理，而是試著在高風險窗口之前介入。對AbbVie，這同時是一條讓既有藥物走進新使用情境的路。 # 時間可預期，讓預防多了一個切入點 LUNA研究藥物atogepant，是美國QULIPTA的有效成分，也是口服小分子CGRP受體拮抗劑。CGRP中文為降鈣素基因相關胜肽，是參與偏頭痛的訊號分子；「受體拮抗」可理解成占住訊號的接收位置，減少CGRP啟動這條路徑。 月經週期與這條訊號的關係，有人體研究提供線索。2023年Charité研究團隊比較不同荷爾蒙狀態的女性，觀察到在自然月經週期中，有偏頭痛者於月經期間的CGRP濃度高於無偏頭痛對照。這是荷爾蒙狀態與CGRP的關聯線索，尚未確立每次經期頭痛的單一原因。 LUNA利用的是發作風險較集中的時間，介入偏頭痛相關訊號。它不調整月經本身，也不以荷爾蒙補充作為研究策略。 短程預防本身也有歷史。2004年已有frovatriptan用於月經相關偏頭痛間歇預防的隨機研究。LUNA增加的是atogepant這條CGRP受體路徑的第三期對照證據；本試驗沒有直接比較不同短程預防藥物。 這裡沒有一個全新的分子，也不是月經偏頭痛突然有了第一種處理方法。改變的是證據與產品定位：同一顆藥，從一般的偏頭痛預防，走向一個時間可預期、需求卻常被忽略的情境。患者的問題因此從「要不要預防」，進一步變成「能不能只把預防放在最需要的那段時間」。 圖1｜LUNA把預防安排在可預期的月經周邊時段，並分別評估偏頭痛日數與功能負擔。 # 7天是給藥安排，圍月經期是療效評估範圍 研究在歐洲與亞洲納入純月經性或月經相關性偏頭痛成年女性，採隨機、雙盲、安慰劑對照設計。入組前先以病史與電子日誌確認月經規律及發作模式，並排除慢性偏頭痛等情況。 日本公部門登錄的研究地區也列有台灣。這是本試驗的直接研究連結，該頁未提供台灣實際收案人數或地區療效，亦非台灣核准或健保給付證據。 如果月份間波動很大，或整個月都有大量頭痛日，預先對準幾天的策略就不一定適合直接套用。研究以日誌辨認規律，實際使用時也需要能掌握自己的發作模式；「何時開始」和「選哪種藥」同樣影響這種方案能否被執行。 試驗方案為atogepant 60 mg或安慰劑每日一次，從預估月經開始前3天起連續7天，在雙盲期間重複三個週期。主要終點則計算圍月經期（perimenstrual period，PMP）的偏頭痛日數，登錄以月經開始為基準標示Day −2至Day +3。 給藥依預估時間提前安排，PMP則用來界定評估範圍，兩者不能視為同一窗口。對其他時段仍會發作的人，整月負擔仍是決定照護安排的另一部分。 圖2｜給藥自預估月經前3天起連續7天；圍月經期（PMP）評估另以月經開始為基準，定義為Day −2至Day +3。兩個窗口分開、非按比例示意。 # 平均少0.80天，功能結果已報喜，幅度還要看清楚 主要結果比較雙盲期三個月經週期的PMP偏頭痛日數相對基線變化之平均。因此，1.20天與0.40天都是平均減少量；0.80天是組間差，並非三週期合計或治療後剩餘日數，也不是每位參與者都得到同樣改善。 0.80天看起來不大，卻要放回一段有限的高風險時間來看，而不是拿去和整月頭痛日直接相比。它也可能由不同程度的改善組成：有人更受益，有人改變不多。完整基線、信賴區間與反應者分布，才會讓平均值長出患者的輪廓；P值回答差異是否有統計支持，不替患者回答值不值得用。 AbbVie表示，八個依序檢定的次要終點也達標，涵蓋頭痛負擔、急性藥物使用、功能障礙、認知功能與反應者指標。這已是公司報告的功能獲益訊號；公告尚未提供的是各項完整效應量，相關結果引用公司留存資料。 登錄中的功能障礙量表（FDS）終點尤其有助理解：它以三個月經週期的PMP資料，評估符合「PMP每一天都沒有功能障礙或僅輕度受限」的受試者比例。它關心一段高風險時間是否都維持在較低負擔，而不只計算少了幾天偏頭痛。 這種量測與患者生活很接近，但不直接等於少請幾天假或恢復幾天照顧家人的能力。若公布兩組達到這個條件的比例與差值，讀者就能判斷公司報告的正向訊號究竟有多大；急性用藥終點則另計PMP中使用藥物的日數。 圖3｜三個月經週期的圍月經期偏頭痛日數，相對基線平均減少1.20對0.40天，組間差0.80天，P<0.0001；PMP是評估期間的名稱。 # 反覆短程使用，仍要把效果與耐受性放在一起 AbbVie報告LUNA未見新安全性訊號，與atogepant既有輪廓一致。判斷反覆短程使用的負擔，還要看不良事件的種類、程度、持續時間及停藥情況；這些資料決定有多少改善是患者願意承受代價後仍能保有的。 QULIPTA美國標籤列有噁心、便祕與疲倦／嗜睡等常見不良反應，以及過敏反應、高血壓、雷諾氏現象等警示。這些是既有產品背景，不是LUNA本次發生率；少用幾天也不足以推定風險等比例降低。 目前引用的美國標籤為成人偏頭痛預防，未列入LUNA的7天短程安排或atogepant按需急救用途。AbbVie在9月10日公布的是計畫提交資料，尚不能視為已送件或取得新核准。上述服法均為試驗描述，實際治療須由醫療團隊評估。 短程不等於一次性。這個方案若進入日常，患者會在下一個週期再次面對同一個選擇：上次的改善，是否值得重複服藥？完整報告對缺失日誌與退出者的處理也在回答這件事，不能只留下順利完成服藥的人。真正可採用的方案，得把效果、身體負擔與執行難度一起交代。 圖4｜功能終點已有公司報告的正向結果；完整效應量、不良事件與耐受性、持續使用及監管資訊，將進一步界定短程方案的使用價值。 # AbbVie多了一個有證據的使用情境，還不是全新的大市場 LUNA讓AbbVie多了一個有陽性三期支撐的產品入口：找到發作集中、可以提前安排，卻還沒有獲得充分預防的患者。這比把所有女性都放進市場規模有意義；它把需求落在一個醫師能辨認、患者也能說清楚的使用情境。 這種定位也有取捨。週期規律、發作集中而原先預防不足的人，可能帶來新增使用；本來已有預防方案的人則可能是轉換。醫師需要把患者過去的反應與整月頭痛負擔一起考慮，七天方案不會只因服藥較少就自然取代既有治療。 對AbbVie，沿用既有分子與品牌，可以延伸已累積的臨床經驗與產品認知；但同一個病人改成短程使用，和多一名此前未治療的患者，對收入不是相同的變化。 商業機會還有一層：如果患者過去只有發作後才處理，針對可預期時窗的選項，可能讓預防更容易進入診間討論。這是擴大有效需求的路，而不是單純增加服藥天數；疾病教育、辨認適合人群與患者願不願意回來，可能和藥物本身同樣影響採用。現有品牌可以提供入口，卻仍要靠新的使用經驗留住人。 每個週期用7天、持續多少週期，以及新增患者來自哪一種治療情境，會共同決定用藥量。患者數增加時，每人年用藥量仍可能不同；在使用者組成與支付條件明朗前，不宜直接上調全年收入假設。 對支付方，FDS達標者比例及急性用藥日數，讓討論從「平均少痛多少」走向「那幾天能否維持較低負擔」。兩組功能達標比例相差多少，會比再次強調P值更有說服力。若這份改善能在可接受的耐受性下反覆維持，AbbVie才有機會把一份陽性結果，變成患者下個週期仍願意選擇的方案。 本文提供產業資訊與商業分析，不構成個別化醫療或投資建議。 --- ## PharmaEssentia Acquires FormuRx. Could Making Drugs Inside Cells Be Next? URL: https://drugnews.com.tw/articles/2026-10-03-pharmaessentia-formurx-rna-lnp-acquisition-2026-en.html Date: 2026-10-03 Language: en Summary: PharmaEssentia (6446) has acquired FormuRx for approximately NT$168 million. According to company-provided news published by Global Bio & Investment, FormuRx develops delivery technologies for getting RNA into cells. The acquisition brings lipid nanoparticle (LNP) formulations, microfluidic processes, analytical capabilities, equipment and a team. PharmaEssentia (6446) has acquired FormuRx for approximately NT$168 million. According to company-provided news published by Global Bio & Investment, FormuRx develops delivery technologies for getting RNA into cells. The acquisition brings lipid nanoparticle (LNP) formulations, microfluidic processes, analytical capabilities, equipment and a team. The company's formal English name is FormuRx Pharmaceuticals Co., Ltd. (FormuRx). FINDIT's public company profile lists its establishment in 2019. National Taiwan University's Chinese- and English-language incubator directories record its residency from September 2021 to August 2026, with a focus on lipid nanocarrier-based drugs, delivery systems and dosage-form design. This establishes an incubation relationship, not status as an NTU spin-off. FINDIT's technology overview also brings together lipid carriers, microfluidic processes and quality by design (QbD). This describes the team's approach to formulation development: incorporating quality and manufacturing considerations from the design stage, rather than pursuing only expression signals in experiments. It is a technology description, not government certification. The company-provided report also identifies two core development projects, FR-202 and FR-201. These give the acquisition concrete product programs to carry forward: PharmaEssentia gains technology and a team that can continue to adapt, while bringing development priorities into its own decision-making. For a company with experience in protein drugs, rare diseases and TCR-T, this may be better suited to expanding its RNA toolkit than licensing a single molecule. That is a strategic interpretation of the scope acquired, not a complete rationale for selecting the acquisition target confirmed by the announcement. PharmaEssentia has already turned a long-acting protein into a medicine. RNA offers a way to explore having cells manufacture a specified protein instead. The aim is still to put proteins to work, but the approach moves one layer deeper. The ambition behind this acquisition may be to extend existing protein and disease research into more drug designs. Formulations, processes and the team are the people and tools needed to turn those ideas into something tangible. The board had resolved on May 13, 2026, to acquire all issued shares. The company-provided report dated September 28 states that the acquisition of the entire shareholding had recently closed. The exact closing date and payment structure have not been disclosed. To understand which product this investment might advance next, the starting point is how the scientific work on the two sides can connect. Figure 01 | The acquisition brings LNP formulation, microfluidic processing, analytical methods and a team. Product value still requires development and validation. # From Giving Proteins to Having Cells Express Them Both approaches study how proteins can change disease, but the questions they need to control differ: Give the protein directly, and study how long its effects last. BESREMi (ropeginterferon alfa-2b) is a long-acting interferon conjugated to polyethylene glycol (PEG). According to its US prescribing information, it binds to the interferon receptor IFNAR on the cell membrane, activates intracellular signaling and thereby affects gene expression. The label also describes the relationship between drug concentrations and hematologic responses. Protein activity, exposure and duration of action become meaningful only when connected to the disease response. Have cells express the protein, and study where and how much they make. Once messenger RNA (mRNA) enters the cytoplasm, it can be translated into a specified protein. The research can therefore extend to intracellular functional proteins or cell-surface receptors. Beyond the amount of RNA administered, developers need to examine which cells express the protein, its location, quantity, duration of expression and actual function. The potential connection is experience in studying protein activity and disease responses. This is a hypothesis about research synergies, not a company-announced plan for RNA-based interferon or combination use with BESREMi. Nor can PEG processes be applied directly to LNPs. RNA requires its own delivery tools. LNPs first encapsulate and protect RNA and help cells take it up. After the particles enter the endosome, a membrane-bound compartment, the mRNA still has to be released into the cytoplasm before it can be translated. A cell signing for a parcel does not mean it has unpacked it and put it to work. These steps are among the delivery and manufacturing barriers discussed in the LNP research review. Figure 02 | After cellular uptake, mRNA must escape the endosome into the cytoplasm for translation. Function and quality require product-specific validation. FormuRx's three areas of expertise address different questions about the same material: how the formulation packages it, whether a microfluidic process can consistently produce the particles, and whether analytical methods can confirm quality after a change. Combining those results with protein-function tests can help distinguish more systematically between RNA that did not reach its destination, insufficient expression, and a protein that appeared but failed to perform the required function. Formulation choices can then follow the product's needs. # Two Rare-Disease Projects Take RNA Beyond Fluorescence Images The company report identifies FR-202 for telomere biology disorders, with a planned investigational new drug (IND) application in 2027, and FR-201 for Wilson disease, with a planned IND application in 2028. The payloads and target genes of the two programs have not been disclosed. These timelines are company plans, not authorization to begin human studies, and neither program is listed as a CAR-T product. Why are these two types of disease worth investigating for RNA developers? The following external studies provide concrete functional questions. They are not FormuRx results, and they cannot be used to assign ATP7B or TERT as the payloads of FR-201 or FR-202. Wilson disease: Once the protein appears, is copper being handled? A 2025 study by Ma and colleagues used LNPs to deliver ATP7B mRNA. ATP7B participates in copper transport. The study tracked protein localization, copper transport, copper burden in mice and ceruloplasmin activity. Development has to connect delivery, a protein working in the right location, and a change in disease-related function—not simply RNA expression levels. Telomere biology disorders: Can transient expression produce a change in telomeres? Telomeres are protective structures at the ends of chromosomes. A 2015 study by Ramunas and colleagues delivered modified mRNA encoding TERT into cultured human cells and observed transient telomerase activity and telomere extension. This is functional evidence at the cellular level, not evidence of anti-aging or regenerative efficacy in humans. The two diseases require different measures of progress. They also provide concrete ways for PharmaEssentia's rare-disease development experience to contribute: selecting identifiable patient populations, considering how long an effect needs to last, and deciding which functional changes can move toward clinical endpoints. If those questions are incorporated into formulation screening, the team's selection standard can move from “Which batch has the brightest expression signal?” to “Which batch has functional results worth taking toward the clinic?” That is a product decision the acquisition, connected to named development programs, could help improve. # With TCR-T Already in the Pipeline, What Door Could RNA Open? PharmaEssentia's public pipeline lists NY-ESO-1 TCR-T and Novel TCR-T. TCR-T is an engineered therapy in which T cells express a specified T-cell receptor. Receptor design is already part of the company's research and development portfolio, giving RNA delivery an existing direction with which it might connect. A 2020 study by Parayath and colleagues used anti-CD8-targeted polymeric nanocarriers to deliver mRNA encoding a CAR (chimeric antigen receptor) or TCR, investigating transient receptor expression in human T cells and mouse models. The carrier in this study was a polymer, not an LNP. A 2022 study by Rurik and colleagues used CD5-targeted LNPs to generate transient CAR-T cells in a mouse model of cardiac injury. Both are external preclinical studies. The acquisition report did not provide validation data for FormuRx's T-cell-targeted LNPs or a named in vivo CAR-T program, and transient expression is not, by itself, a guarantee of immune safety. Figure 03 | Separate external preclinical studies: CD8-targeted polymer carriers in 2020 and CD5-targeted LNPs in 2022. Their systems differ; neither validates FormuRx products. Putting these studies back into the context of the company's research direction reveals two areas of work that would need to connect: Receptor design answers “What should be recognized?” Antigen selection, receptor design and functional testing need to establish that T cells can produce the required effect after recognizing their target. This is a question that the existing TCR-T direction and research on RNA-driven receptor expression could share. Delivery design answers “Who receives the instructions?” After receptor mRNA is delivered into the appropriate T cells, the level and duration of expression still need to be adjusted so that delivery results connect with cellular function. A different delivery method changes more than the material; it may also change how cells are engineered and delivered. The near-term rare-disease RNA projects can first build experience in formulation, analysis and manufacturing processes. Over the longer term, the team can evaluate which of that experience might connect with receptor design. That broadens future development choices from “Which additional antigen should we work on?” to “How should engineered cells be generated and used?” # What Level of Control Does NT$168 Million Buy? Licensing a single molecule obtains product rights within the scope of a contract. A wholly owned acquisition can bring the team and resource allocation into the company's decision-making. According to the report, this payment is for all of FormuRx's issued shares, with its platform, equipment and personnel to be integrated into PharmaEssentia. It is not simply a capital injection into FormuRx. The approximately NT$168 million can be understood at three levels: Change the product and formulation together: The same organization can decide which project to pursue first and bring RNA design, formulation and quality questions into the same technical decision, rather than waiting until a candidate is finalized to address scale-up. Gain a starting point for subsequent products: Analytical methods, mixing processes and records of failed formulations left by the first program may help the next program rule out unsuitable paths earlier. The acquisition brings a team that can keep working, not a formulation that is used once and put away. Bring more substance to asset-partnership negotiations: When a candidate can be discussed together with its pharmacology, manufacturing and quality data, the company is better positioned to decide which stages to advance itself and where to seek partners. Collaboration can then be built around specific divisions of work. The price relates to early research and development and future development options, not the near-term cash flow of a mature product. Asset details, payment arrangements and subsequent costs have not been disclosed, so the price tag alone cannot establish whether the acquisition is cheap or expensive. Beyond the acquisition, integration, toxicology and clinical development still require investment. What matters is whether that additional investment can generate repeatable product-development results. Figure 04 | Full ownership can align product selection, formulation, processing, quality analysis and further development. This schematic does not show validated pharmacology or toxicology results. # From One Global Product to a Company That Can Keep Developing Drugs The promising path toward a “Taiwanese BeiGene” is for PharmaEssentia to use the research, clinical and overseas experience accumulated through BESREMi to integrate suitable Taiwanese startups and connect their technologies with international product development. For Taiwan to build its own international biopharmaceutical giant, the key is whether new assets can reach the market one after another. BESREMi moved from global research to US commercialization and gained approval for an additional indication, essential thrombocythemia (ET), on August 31, 2026. The FormuRx acquisition was resolved in May; these are parallel initiatives in the same year. The June 11 FORUS announcement explicitly states that its Canadian regulatory, medical, commercial and market-access teams will serve BESREMi and future products. For Taiwanese startups with candidates but without international execution resources, collaboration can go a step beyond technology licensing. If PharmaEssentia can connect process and quality development, multinational clinical planning and regulatory communication, startup teams can carry their strengths in mechanisms and product design toward more mature assets. On the market side, the company can choose between developing markets itself and partnering, according to disease and geography. This would give new technology a chance to connect with an organization capable of taking on subsequent development. Concrete figures offer a reference for scale. Public Nasdaq data obtained on October 2, 2026, showed a market capitalization of approximately US$41 billion for BeOne Medicines, formerly BeiGene. Its revenue for the second quarter of 2026 was US$1.705 billion. Market capitalization is an equity valuation; revenue is the operating result of a single quarter. Alongside the company's description of its integrated research, clinical and market footprint, this comparison gives the growth ambition a concrete direction: bringing internally developed products to market while continuing to develop the next group of products. If that path gradually takes shape at PharmaEssentia, the questions the market asks in valuing the company will also broaden. Beyond BESREMi and the existing pipeline, other questions will merit attention: How much additional investment is needed to bring acquired technologies into clinical development? How much of the product rights will the company retain? Which markets can it operate itself? The maturity of the evidence, probability of success and development costs of each program provide a basis for assessing early-stage options. As products advance, those options gradually acquire a more concrete foundation for value. If FormuRx can move its first programs forward and leave the methods behind for the next one, PharmaEssentia has a chance to become an integrator that takes Taiwanese technology overseas. When research, clinical execution and overseas support no longer have to be built from scratch for every program, how will the market value a PharmaEssentia that can keep bringing new products forward? That is the growth ambition behind the prospect of a “Taiwanese BeiGene.” This article provides industry information and commercial analysis. It does not constitute individualized medical or investment advice. # Research and Data Sources Global Bio & Investment, September 28, 2026; company-provided news Wholly owned acquisition, context of the approximately NT$168 million transaction, scope acquired and the two planned IND applications Original report: https://news.gbimonthly.com/tw/invest/show.php?num=90341&range=news BESREMi US prescribing information and the company announcement dated August 31, 2026 PEGylated interferon, IFNAR signaling, exposure and response, and ET approval DailyMed label: https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=9583405d-53a0-49dc-88eb-5e6384ebabcb Company announcement: https://us.pharmaessentia.com/fda-approves-pharmaessentias-besremi-ropeginterferon-alfa-2b-njft-for-adults-with-essential-thrombocythemia-a-rare-blood-cancer/ Hou et al., Lipid nanoparticles for mRNA delivery (2021) RNA protection, cellular uptake, endosomal release and translation to manufacturing Nature Reviews Materials: https://www.nature.com/articles/s41578-021-00358-0 Ma et al., Hepatocyte-targeted lipid nanoparticle for full-length ATP7B mRNA delivery in Wilson disease (2025) Protein localization, copper metabolism and animal functional measures in external ATP7B mRNA-LNP research Original research abstract on PubMed: https://pubmed.ncbi.nlm.nih.gov/41083009/ Ramunas et al., Transient delivery of modified mRNA encoding TERT rapidly extends telomeres in human cells (2015) Transient telomerase activity and telomere extension in cultured human cells FASEB Journal DOI: https://doi.org/10.1096/fj.14-259531 PharmaEssentia's public pipeline NY-ESO-1 TCR-T and Novel TCR-T; the page has no update date, so it is not used to assign a new development stage Company pipeline page: https://www.pharmaessentia.com/pipeline/pipeline Parayath et al. (2020); Rurik et al. (2022) CD8-targeted polymer delivery of CAR/TCR mRNA and CD5-targeted LNP research on CAR-T in mice, respectively Nature Communications study: https://pubmed.ncbi.nlm.nih.gov/33247092/ Science study: https://pubmed.ncbi.nlm.nih.gov/34990237/ PharmaEssentia's FORUS acquisition announcement, June 11, 2026 Canadian regulatory, medical, commercial and market-access capabilities for BESREMi and future products Official company announcement: https://us.pharmaessentia.com/pharmaessentia-expands-north-american-commercial-footprint-with-acquisition-of-canadian-partner-forus-therapeutics/ BeOne's public research and market-access information A reference model for integrated research, clinical and international operations Official description: https://beonemedicines.com/our-commitment-to-access/ Nasdaq ONC quote and market capitalization, obtained October 2, 2026; quote as of October 1, 2026 Market capitalization of approximately US$41 billion; the quote page is labeled October 1, 2026, and Closed, while the market-capitalization field has no separate timestamp Nasdaq quote: https://api.nasdaq.com/api/quote/ONC/info?assetclass=stocks Nasdaq market capitalization: https://api.nasdaq.com/api/quote/ONC/summary?assetclass=stocks BeOne's second-quarter 2026 Form 10-Q Number of ordinary shares and the conversion basis of one ADS representing 13 ordinary shares, for checking the company's overall equity scale Company Form 10-Q: https://ir.beonemedicines.com/media/document/e8111d44-9cdd-4326-80a1-f3158a471bd2/assets/beone2q2026_10Q.pdf?disposition=inline BeOne's second-quarter earnings announcement, August 5, 2026 Revenue of US$1.705 billion for the second quarter of 2026, as a reference for realized operating scale in a single quarter Company earnings announcement: https://ir.beonemedicines.com/media/document/06b047db-4795-4815-9a4c-633c486de43e/assets/Q2_2026_Earnings_Release_FINAL.pdf?disposition=inline National Taiwan University Innovation and Incubation Center's Chinese- and English-language directories of graduated companies; entry dated September 2, 2026 Correspondence between FormuRx's Chinese and English names, incubator residency from September 2021 to August 2026, and its focus on lipid nanocarriers, delivery and dosage forms; an incubation relationship Chinese directory: https://ntuiic.ntu.edu.tw/graduated.html English directory: https://ntuiic.ntu.edu.tw/ntuiic-en/graduated-e.html FINDIT's FormuRx company profile, updated October 2, 2026 Establishment in 2019; the technology description covers lipid carriers, microfluidic processes and QbD, and is not interpreted as government technology certification Public profile: https://findit.sme.gov.tw/tw/Startup/K2U5T25UN0QwWkRMTjhSc3IzalNlZz09/Basic --- ## 藥華藥買下予宇，下一步讓細胞造藥？ URL: https://drugnews.com.tw/articles/2026-10-03-pharmaessentia-formurx-rna-lnp-acquisition-2026.html Date: 2026-10-03 Language: zh-Hant Summary: 藥華藥（6446）買下予宇生技，交易金額約新台幣1.68億元。依環球生技刊載的公司消息，予宇做的是把RNA送進細胞的遞送研發，帶進來的包括脂質奈米顆粒（LNP）配方、微流體製程、分析、設備與團隊。 藥華藥（6446）買下予宇生技，交易金額約新台幣1.68億元。依環球生技刊載的公司消息，予宇做的是把RNA送進細胞的遞送研發，帶進來的包括脂質奈米顆粒（LNP）配方、微流體製程、分析、設備與團隊。 予宇的正式英文名是FormuRx Pharmaceuticals Co., Ltd.（FormuRx）。FINDIT公開基本資料列示公司2019年成立；臺大中文育成名錄與英文名錄則記錄它在2021年9月至2026年8月進駐，專注脂質奈米載體新藥、遞送系統與劑型設計。這段經歷是育成關係，不等於臺大衍生公司身分。 FINDIT的技術簡介進一步把脂質載體、微流體製程與「品質源於設計」（QbD）放在一起。這描述的是團隊的製劑開發方法：從設計時就把品質與製造問題納入，而非只追求實驗裡的表現訊號。它是技術介紹，不是政府認證。 公司供稿也說明，予宇已有FR-202、FR-201兩項核心研發專案。這讓收購有具體的產品承接點：藥華藥取得的是可繼續調整的技術與團隊，並可把研發排序納入自身決策。對已有蛋白藥、罕病與TCR-T經驗的公司，這可能比只取得單一分子的授權，更適合拓展RNA工具；它是從取得範圍推想的策略，未被公告確認為完整選標理由。 藥華藥已把長效蛋白做成藥；RNA則能探索讓細胞製造指定蛋白。同樣要讓蛋白質發揮作用，做法卻換了一層。這筆收購的野心，可能是把既有的蛋白與疾病研究，延伸到更多新藥設計，而配方、製程和團隊正是要把想法做出來的人與工具。 董事會早在2026年5月13日決議收購全部已發行股份；9月28日的公司供稿則說明，已於日前完成全數股權交割。確切交割日與付款結構未披露。這筆錢接下來能帶動哪一款產品，要從兩邊的科學工作如何接合看起。 圖01｜全資收購承接LNP配方、微流體製程、分析方法與團隊；產品價值仍待後續開發驗證。 # 從給蛋白，到讓細胞表現蛋白 同樣研究蛋白質如何改變疾病，兩種做法要控制的問題不同： 直接給蛋白，研究作用能維持多久。BESREMi（ropeginterferon alfa-2b）是接上聚乙二醇（PEG）的長效干擾素。依美國處方資訊，它結合細胞膜上的干擾素受體IFNAR，啟動細胞內傳訊，進而影響基因表現；標籤也描述藥物濃度與血液學反應的關係。蛋白的活性、暴露與作用時間，必須連到疾病反應才有意義。 讓細胞表現蛋白，研究在哪裡、做出多少。信使RNA（mRNA）進入細胞質後，可被翻譯成指定蛋白，研究範圍因此可延伸到細胞內功能蛋白或細胞膜受體。除了給了多少RNA，還要看哪些細胞表現、蛋白位置、數量、維持時間與實際功能。 兩邊可相接的是蛋白活性與疾病反應的研究經驗；這是研究工作的協同推想，並非公司已公布的RNA干擾素或BESREMi併用計畫，PEG製程也不能直接套到LNP。RNA還需要自己的遞送工具。 LNP先包覆與保護RNA，協助細胞攝入；顆粒進入內體這個膜包住的區室後，mRNA還得釋放到細胞質，才能翻譯。細胞簽收了包裹，不等於已經拆箱上工。LNP研究綜述討論的遞送與製造障礙，就包括這些環節。 圖02｜RNA被細胞攝入後，仍須從內體釋放到細胞質才能翻譯；功能與品質須按產品驗證。 予宇的三項專長，處理的是同一份材料的不同問題：配方怎麼包、微流體製程能否穩定做出顆粒、分析方法能否確認改動後的品質。把這些結果與蛋白功能測試放在一起，就能更有方向地分辨：RNA沒送到、表現量不夠，還是蛋白出現了卻未發揮需要的作用。配方取捨也就能跟著產品需要走。 # 兩個罕病專案，讓RNA不只停在螢光照片 公司消息列出FR-202針對端粒生物學疾病，規劃2027年提出臨床試驗申請（IND）；FR-201針對威爾森氏症，規劃2028年提出IND。兩案的載荷與靶基因未公開，時程屬公司計畫，並非已獲准進入人體，也未列為CAR-T產品。 這兩類疾病為什麼值得RNA開發者研究？以下外部研究提供了具體的功能問題，但不是予宇成果，也不能據此把ATP7B或TERT指定為FR-201、FR-202的載荷。 威爾森氏症：蛋白出現了，銅有沒有被處理？2025年Ma等人的研究以LNP遞送ATP7B mRNA。ATP7B參與銅運輸，研究追蹤蛋白定位、銅運輸，以及小鼠的銅負荷與銅藍蛋白活性。開發要接起來的是「送達、蛋白在正確位置工作、疾病相關功能改變」，不只看RNA表現量。 端粒生物學疾病：短暫表現，能否帶來端粒變化？端粒是染色體末端的保護結構。2015年Ramunas等人的研究在人類培養細胞中遞送編碼TERT的修飾mRNA，觀察到短暫端粒酶活性與端粒延長。這是細胞層級的功能證據，不等於人體抗老或再生療效。 兩種疾病要讀的成績不同，也讓藥華藥的罕病開發經驗有具體加入的位置：挑選可辨認的患者，思考作用需要維持多久，以及哪些功能變化能往臨床終點推進。 如果這些問題在配方篩選時就加入，團隊挑材料的標準，就能從「哪批表現訊號最亮」，推進到「哪批功能結果值得帶往臨床」。這是收購與具名管線接合後，可能改善的產品決策。 # 已有TCR-T，RNA可再打開哪道門？ 藥華藥官網管線列有NY-ESO-1 TCR-T與Novel TCR-T。TCR-T是讓T細胞表現指定T細胞受體的工程療法，受體設計本就在公司研發版圖中，RNA遞送因此有一個可以相接的既有方向。 2020年Parayath等人的研究使用抗CD8靶向的聚合物奈米載體，遞送CAR（嵌合抗原受體）或TCR的mRNA，在人類T細胞與小鼠模型研究短暫受體表現。這裡用的是聚合物，不是LNP。 2022年Rurik等人的研究則用CD5靶向LNP，在心臟損傷小鼠模型中產生暫時CAR-T。兩項都是外部前臨床研究；此次收購報導未提供予宇T細胞靶向LNP或具名體內CAR-T的驗證資料，短暫表現本身也不是免疫安全保證。 圖03｜兩項外部前臨床研究：2020年CD8靶向聚合物載體與2022年CD5靶向LNP，研究系統不同；不是予宇產品驗證。 把這些研究放回公司的研發方向，可看見兩項需要接合的工作： 受體設計，回答「辨識什麼」。抗原選擇、受體設計與功能檢測，要確認T細胞辨識目標後能產生需要的作用。這是既有TCR-T方向與RNA受體表達研究可能共用的問題。 遞送設計，回答「指令交給誰」。把受體mRNA送進適當T細胞後，還要調整表現量與持續時間，讓遞送結果接上細胞功能。換一種送法，改變的不只是材料，也可能是細胞改造與交付方式。 近期的罕病RNA專案可以先累積配方、分析與製程經驗；更長期，再評估哪些經驗能與受體設計接合。這讓後續開發選擇從「多做哪個抗原」，延伸到「用什麼方式生成和使用工程化細胞」。 # 1.68億元買的是哪一層控制權？ 授權單一分子，取得的是合約約定範圍的產品權利；全資收購則能把團隊與資源配置納入公司決策。依報導，這筆錢對應予宇全部已發行股份的收購，平台、設備與人員將整合進藥華藥，並非單純投入予宇的增資款。 約1.68億元可從三個層次理解： 把產品和配方一起改：同一組織能決定先做哪個專案，也能把RNA設計、配方與品質問題放在同一次技術決策裡，不必等候選物定案才處理放大。 取得後續產品的開發起點：第一案留下的分析方法、混合製程與失敗配方紀錄，可能讓下一案更早排除不適合的路。買進的是能繼續工作的團隊，不是一張用完就收起來的配方。 增加資產合作的談判內容：當候選物能連同藥理、製程與品質資料一起討論，公司才更有條件決定自行推進哪些階段、在哪些地區找夥伴，合作也能圍繞具體分工展開。 這個價格對應早期研發與後續開發選項，不是成熟產品的近期現金流。資產明細、支付方式與後續成本未公開，無從只憑價碼判斷便宜或昂貴；收購之外，仍有整合、毒理與臨床投入。真正值得估量的，是這些新增投入能否形成可重複的產品開發成果。 圖04｜全資收購可把產品、配方、製程、品質分析與後續開發放在同一決策中；圖示不是已驗證的藥效或毒理結果。 # 從一款全球產品，走向能接續出藥的公司 藥華藥值得期待的「台灣版百濟」路徑，是以BESREMi累積的研發、臨床與海外經驗，整合適合的台灣新創，把技術接上國際產品開發。台灣要長出自己的國際Biopharma巨獸，關鍵在於能否讓新資產接連走向市場。 BESREMi從全球研究走到美國商業化，並在2026年8月31日新增原發性血小板增多症（ET）適應症核准；予宇收購早在5月決議，兩者屬同年並行布局。6月11日FORUS公告則明說，加拿大法規、醫學、商業與市場准入團隊將服務BESREMi及未來產品。 對已有候選物、卻缺國際執行資源的台灣新創，合作可以往技術授權之後多走一段。藥華藥若能接上製程與品質開發、跨國臨床規劃及法規溝通，新創團隊就能把機制與產品設計的長處，往更成熟的資產推進；市場端再依疾病與地區選擇自行發展或合作。這讓新技術有機會接上一套能承擔後續開發的組織。 規模可以用具體數字參照：2026年10月2日取得的Nasdaq公開資料顯示，BeOne Medicines（原百濟神州）市值約410億美元；其2026年第二季營收為17.05億美元。市值是股權定價，營收則是單季營運成果。配合公司描述的整合研發、臨床與市場布局，這組參照讓成長想像有了具體方向：既能把自主產品帶向市場，也能接著開發下一批產品。 若這條路逐步在藥華藥成形，市場評價公司的問題也會拓寬。除了BESREMi及既有管線，還值得看：接進來的技術要追加多少投入才能推至臨床、公司保留多少產品權利，以及哪些市場能自行經營。各案的證據成熟度、成功機率與開發成本，是評估早期選項的依據；隨著產品往前走，這些選項才逐步取得更具體的價值基礎。 予宇若能讓首批專案往前走，再把方法留給下一案，藥華藥就有機會逐步成為台灣技術出海的整合者。當研發、臨床執行與海外承接不必每案從頭搭建，市場會如何評價這種能接續推出產品的藥華藥？這正是「台灣版百濟」值得期待的成長想像。 本文提供產業資訊與商業分析，不構成個別化醫療或投資建議。 # 研究與資料來源 環球生技，2026年9月28日，公司提供消息 全資收購、約1.68億元背景、取得範圍與兩項IND規劃 報導原頁：https://news.gbimonthly.com/tw/invest/show.php?num=90341&range=news BESREMi美國處方資訊及2026年8月31日公司公告 PEG化干擾素、IFNAR訊號、暴露與反應，以及ET核准 DailyMed標籤：https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=9583405d-53a0-49dc-88eb-5e6384ebabcb 公司公告：https://us.pharmaessentia.com/fda-approves-pharmaessentias-besremi-ropeginterferon-alfa-2b-njft-for-adults-with-essential-thrombocythemia-a-rare-blood-cancer/ Hou等，Lipid nanoparticles for mRNA delivery（2021） RNA保護、細胞攝入、內體釋放及製造轉譯 Nature Reviews Materials：https://www.nature.com/articles/s41578-021-00358-0 Ma等，Hepatocyte-targeted lipid nanoparticle for full-length ATP7B mRNA delivery in Wilson disease（2025） 外部ATP7B mRNA-LNP研究的蛋白定位、銅代謝及動物功能指標 PubMed原研究摘要：https://pubmed.ncbi.nlm.nih.gov/41083009/ Ramunas等，Transient delivery of modified mRNA encoding TERT rapidly extends telomeres in human cells（2015） 人類培養細胞中的暫時端粒酶活性與端粒延長 FASEB Journal DOI：https://doi.org/10.1096/fj.14-259531 藥華藥公開管線 NY-ESO-1 TCR-T與Novel TCR-T；頁面未標更新日，不據此補新階段 公司管線頁：https://www.pharmaessentia.com/pipeline/pipeline Parayath等（2020）；Rurik等（2022） 分別為CD8靶向聚合物CAR／TCR mRNA與CD5靶向LNP的小鼠CAR-T研究 Nature Communications研究：https://pubmed.ncbi.nlm.nih.gov/33247092/ Science研究：https://pubmed.ncbi.nlm.nih.gov/34990237/ 藥華藥FORUS收購公告，2026年6月11日 加拿大法規、醫學、商業及市場准入，供BESREMi與未來產品使用 公司官方公告：https://us.pharmaessentia.com/pharmaessentia-expands-north-american-commercial-footprint-with-acquisition-of-canadian-partner-forus-therapeutics/ BeOne公開研發與市場可近性資料 整合研發、臨床與國際布局的模式參照 官方說明：https://beonemedicines.com/our-commitment-to-access/ Nasdaq ONC報價與市值（2026-10-02取得；報價截至2026-10-01） 市值約410億美元；報價頁標示2026-10-01、Closed，市值欄未另標獨立時戳 Nasdaq報價：https://api.nasdaq.com/api/quote/ONC/info?assetclass=stocks Nasdaq市值：https://api.nasdaq.com/api/quote/ONC/summary?assetclass=stocks BeOne 2026年第二季10-Q 普通股數及1 ADS對應13股普通股的換算口徑，供公司整體股權規模核對 公司10-Q：https://ir.beonemedicines.com/media/document/e8111d44-9cdd-4326-80a1-f3158a471bd2/assets/beone2q2026_10Q.pdf?disposition=inline BeOne第二季財報公告，2026年8月5日 2026年第二季營收17.05億美元，作為已實現單季營運規模的參照 公司財報公告：https://ir.beonemedicines.com/media/document/06b047db-4795-4815-9a4c-633c486de43e/assets/Q2_2026_Earnings_Release_FINAL.pdf?disposition=inline 臺大創新育成中心中英文離駐企業名錄，項目日期2026年9月2日 予宇中英文名稱對應、2021年9月至2026年8月進駐及脂質奈米載體／遞送／劑型方向；屬育成關係 中文名錄：https://ntuiic.ntu.edu.tw/graduated.html 英文名錄：https://ntuiic.ntu.edu.tw/ntuiic-en/graduated-e.html FINDIT予宇基本資料，2026年10月2日更新 2019年成立；技術簡介描述脂質載體、微流體製程與QbD，不作政府技術認證解讀 公開資料頁：https://findit.sme.gov.tw/tw/Startup/K2U5T25UN0QwWkRMTjhSc3IzalNlZz09/Basic --- ## Novo Nordisk Already Has Weight-Loss Injections. Why Pursue Another Path? URL: https://drugnews.com.tw/articles/2026-09-30-novo-kallyope-non-incretin-obesity-programs-en.html Date: 2026-09-30 Language: en Summary: Why pursue non-incretin programs alongside an established GLP-1 franchise? K-554’s neural-circuit hypothesis, human evidence and usable dose range frame the relationship between efficacy, tolerability, persistence and product value. ────────── In September 2026, Novo Nordisk executive Jacob Petersen publicly stated that the company had signed an asset purchase agreement to acquire three non-incretin obesity development programs originating at Kallyope. The lead peptide candidate, K-554, was described as IND-ready, while the other two are earlier-stage small-molecule programs. The public post disclosed neither financial terms nor confirmation that the transaction had closed. [1] Novo Nordisk already has weight-management medicines such as Wegovy, or semaglutide. Why look beyond GLP-1? The question behind this move is more specific than whether patients can lose a few additional kilograms: could a different entry point into appetite regulation produce a treatment that balances efficacy with tolerability, potentially alongside existing medicines? The company identified both monotherapy and combination use as development possibilities. [1][6] For investors, the three programs expand the range of potential future products. Their value depends on whether the new biology can provide patients with something useful beyond existing therapy—not on treating an undisclosed new target as the next blockbuster before it has been tested. *Figure 1 | Established incretin biology is shown alongside new non-incretin research options. Some targets remain undisclosed, and the clinical value of the new programs in humans remains to be established. [1][2]* ## 【01｜Eating Less and Eating Less Because of Discomfort Are Different Research Questions】 Appetite is not controlled by a single switch. The gut, brain and other organs exchange neural and circulating signals that influence when eating starts, how much is consumed and when it stops. A neural circuit can be understood as a network of cells that receives information, relays it and changes behavior. [3] GLP-1, or glucagon-like peptide-1, is one of the incretin signals. Wegovy activates the GLP-1 receptor to influence appetite and caloric intake; these receptors are also present in brain regions involved in appetite regulation. Its effect cannot be reduced to “making people too nauseated to eat.” [6] In its January 7 update, Kallyope said K-554 targets an undisclosed non-incretin peptide receptor involved in a feeding-regulation neural circuit, with biology distinct from GLP-1 and amylin pathways. The company’s non-aversive satiety hypothesis aims to produce satiety signals while reducing their coupling with aversion and discomfort. [2] 🧠 That distinction matters scientifically. Reduced food intake may reflect greater satiety, but it can also include avoidance driven by discomfort. A mouse study published in *Nature* in 2024 separately examined hindbrain GLP-1-receptor circuits involved in satiety and aversion, showing that the two are not identical. [7] That was a separate animal study, however, and it did not identify K-554’s receptor. It supports testing the two responses separately, not a conclusion that K-554 already reduces nausea in humans. The product opportunity is whether appetite control can eventually be maintained at an effective dose while the burden of discomfort remains sufficiently low. ## 【02｜Three Programs Preserve Different Product-Format Options】 K-554 is designed as a once-weekly injection. Of the other two programs, one is a follow-on small molecule directed at a novel target; the other is a small-molecule receptor agonist in lead optimization, meaning that the properties of candidate compounds are still being refined. This remains some distance from a commercially usable product. [1][2] From a product-planning perspective, the package preserves peptide and small-molecule options. Peptide potency, duration of action and injectable formulation need to be developed together. Small molecules must also address absorption, exposure and selectivity. Different routes of administration or use settings could serve different needs if subsequent data support them. Molecular class alone cannot determine which option will be cheaper or necessarily suitable for oral administration. Three programs also do not represent three independent chances of success. Shared biological risks could affect several programs together. Whether different targets and modalities genuinely diversify that risk depends on their individual data. The public disclosures do not fully establish the relationships among the targets, so the two small molecules cannot be assumed to act at the same receptor as K-554. [1][2] *Figure 2 | K-554 is a peptide designed for weekly injection. The other two programs are early-stage small molecules, one of which is in lead optimization. The three cannot collectively be described as products that have entered human development. [1][2]* ## 【03｜January’s Timeline Needs to Be Checked Against Later Evidence】 On January 7, Kallyope originally planned to initiate a first-in-human Phase 1 study around midyear and obtain early safety, pharmacokinetic and pharmacodynamic data from a single-ascending-dose study in the second half of 2026. Single-ascending-dose studies assess different single doses across successive cohorts. Pharmacokinetics, or PK, concerns concentrations and persistence in the body; pharmacodynamics, or PD, concerns the associated biological effects. [2] The September post still described K-554 as IND-ready. This is a company description of submission preparedness, not a standardized regulatory approval category, and it does not independently establish that trial dosing has occurred. As of September 27, verifiable public information remains insufficient to confirm whether January’s plan was realized. Equally, failure to obtain an identifiable registry record does not establish that no filing, enrollment or dosing has occurred, or that the program has been delayed. [1][2][5] The commercial significance of the next update is not simply moving a development-stage label one box forward. Initial human data need to place drug exposure, adverse reactions and exploratory efficacy together. If participants eat less, at what concentration does that happen? Does discomfort increase at the same time? These questions begin to establish whether a usable dose range exists. Once-weekly administration is also a design objective. Supporting that interval requires evidence about duration of action, repeated dosing and variability between patients. Short-term reductions in food intake are insufficient to establish long-term weight maintenance, much less what happens after treatment stops. *Figure 3 | IND-ready itself does not answer questions about regulatory progress, human development, safety and PK/PD, efficacy or longer follow-up. These evidence domains may develop in parallel; unconfirmed does not mean an event has not occurred. [1][2][5]* ## 【04｜Eight Years of Collaboration Do Not Automatically Create One Asset History】 The companies signed a gut–brain-axis research collaboration on June 19, 2018. Novo Nordisk could exercise licensing options on up to six products discovered through the collaboration. That was a ceiling on the rights available, not evidence that six medicines had succeeded. [3] On September 10, 2024, Kallyope announced that Novo Nordisk had exercised an option on a ligand identified through the collaboration. A ligand is a molecule that binds to a biological target. Under that arrangement, Novo Nordisk assumed responsibility for further development, manufacturing and commercialization. [4] The history matters because the platform collaboration did produce something selected for further development. But the announcement did not identify the 2024 ligand as K-554. Its progress or contractual terms cannot be transferred to the three programs in the latest agreement. The 2026 arrangement needs to establish value through its own asset data and subsequent development decisions. ## 【05｜Better Tolerability Could First Change Who Can Stay on Treatment】 The Drugnews team’s interpretation is that Novo Nordisk’s search for non-incretin pathways should be assessed through the number of patients who can obtain and remain on effective treatment. The existing Wegovy label includes gastrointestinal adverse reactions such as nausea and vomiting, as well as discontinuations because of adverse reactions. Treatment burden is therefore a practical product issue alongside efficacy. [6] If K-554 eventually delivers meaningful efficacy while allowing some patients to remain on treatment more easily, it could have a monotherapy position of its own. A better-tolerability claim must nevertheless be established through an appropriate comparison. A low dose that feels more comfortable but provides insufficient efficacy does not automatically make a useful medicine. Combination with an incretin is another scenario. If the pathways are complementary, a study could test additional efficacy or whether different dose combinations produce a better overall treatment profile. Different mechanisms do not guarantee additive benefit. Biological signals may overlap, while additional adverse reactions, administration burden and cost could offset the gain. For Novo Nordisk, monotherapy could reach patients whose needs are not met by existing treatment, while combination use could extend the position of its GLP-1 products. Both would consume clinical and manufacturing resources. If the improvement is small, additional costs could materialize sooner than an additional useful treatment setting. 💼 Payers would need to compare the outcomes a new therapy adds with the additional expense. Physicians would need to know which patients are suitable, how to track benefit and how to manage discomfort. If an advantage appears only in a narrowly selected group of participants, the potential market would differ from a broad replacement for existing weight-management drugs. Even if a later trial reports satisfactory average weight loss, it matters whether people who leave the study are handled appropriately in the analysis and how long improvement lasts. Looking only at the curves of treatment completers can overstate what people who start therapy actually gain. This affects both clinical adoption and the durability of revenue, making it worth addressing at the study-design stage. Manufacturing and delivery must also keep pace. If the eventual product is a weekly injection, formulation stability, batch quality and continuous supply will contribute to cost. If the small-molecule programs create other administration options, their individual product data must support those uses. Undisclosed financial terms and payment structures also prevent investors from assessing how much Novo Nordisk is paying for these options. *Figure 4 | Long-term efficacy, tolerability, persistence and positioning as monotherapy or combination treatment require corresponding human evidence. A hypothesis of less discomfort does not, by itself, establish a commercial advantage.* The most discriminating next evidence will be K-554’s first readable human dose dataset: over the same observation period, which exposures produce appetite or efficacy signals, and where do nausea and other adverse effects occur? A sufficiently usable space between those responses would provide a reason to invest in longer-term studies. This article provides industry information and business analysis and does not constitute individualized medical or investment advice. --- ## 已有瘦瘦針，諾和諾德為何還要布局另一條減重路？ URL: https://drugnews.com.tw/articles/2026-09-30-novo-kallyope-non-incretin-obesity-programs.html Date: 2026-09-30 Language: zh-Hant Summary: 諾和諾德已有GLP-1減重產品，為何仍布局非腸泌素計畫？從K-554的神經迴路假說、人體證據進度與可用劑量，分析效果、耐受性及持續使用如何影響產品價值。 ────────── 2026年9月，諾和諾德（Novo Nordisk）高管Jacob Petersen公開表示，公司已簽署資產購買協議，擬取得Kallyope發起的三項非腸泌素肥胖症研發計畫。領先胜肽候選K-554被描述為IND-ready，另兩項是更早期的小分子；公開貼文未披露金額，也沒有確認交割完成。[1] 諾和諾德已擁有Wegovy（semaglutide）這類減重藥，為何仍要走出GLP-1？這筆布局對準的問題，比「能不能再少幾公斤」更細：能否從不同的食慾調節入口，找到兼顧效果與耐受性的治療，甚至與既有藥物搭配？公司把單用與併用都列為研發方向。[1][6] 對投資人來說，三項計畫增加的是未來產品的選擇空間。其價值要看新的生物學能否讓患者得到既有療法以外的實際好處，而非靠「新標的」三個字，就把尚待驗證的計畫算成下一款重磅藥。 *圖1｜既有腸泌素生物學與新增非腸泌素研究選項並列；新計畫的部分標的未披露，人體價值仍待驗證。[1][2]* 【01｜少吃，與不舒服得吃不下，是兩個研究問題】 食慾並非由單一開關控制。腸道、腦部與其他器官透過神經和循環中的訊號交換資訊，影響何時開始進食、吃多少與何時停止。所謂神經迴路，可以理解成一組接收訊號、傳遞訊息，再改變行為的細胞網路。[3] GLP-1是類升糖素胜肽-1，屬腸泌素訊號之一。Wegovy透過活化GLP-1受體影響食慾與熱量攝取；相關受體也存在於調節食慾的腦部區域。它的作用不能簡化成「讓人噁心，所以吃不下」。[6] Kallyope在1月7日說明，K-554瞄準一個未公開的非腸泌素胜肽受體，涉及攝食調節神經迴路，生物學有別於GLP-1與胰澱素（amylin）路徑。公司所說的non-aversive satiety，是希望產生飽足訊號，同時減少與厭惡、不適感綁在一起的作用。[2] 🧠 這個區別有研究意義：同樣觀察到少吃，可能是飽足增加，也可能夾雜不舒服而避開食物。2024年《Nature》一項小鼠研究，已把後腦GLP-1受體相關的飽足與厭惡迴路分開研究，顯示兩者並非完全相同。[7] 但那是另一套動物研究，沒有識別K-554的受體。它支持的是「兩種反應值得分開測」，不是K-554已能在人類減少噁心。真正的產品機會，在於未來能否在有效劑量下保留食慾控制，同時讓不舒服的負擔足夠低。 【02｜三個計畫，保留不同產品形式的空間】 K-554採每週一次注射設計。其餘兩項，一項是針對新標的的後續小分子計畫；另一項是小分子受體致效劑，仍在先導化合物優化，也就是持續調整候選分子的性質。這個階段離可商業使用的產品還有相當距離。[1][2] 從產品規劃看，這份組合留下胜肽與小分子的選項。胜肽的效力、作用時間和注射製劑需要一起設計；小分子也得處理吸收、暴露與選擇性等問題。若後續資料支持不同的給藥方式或使用情境，才有機會服務不同需求，不能只按分子類型決定誰較便宜、誰一定適合口服。 三個計畫也不等於三次互不相關的成功機會。共同的生物學風險可能讓多個項目一起受影響；不同標的與形式是否真正分散風險，要看各自資料。公開內容沒有把所有標的關係交代完整，因此不能認定兩項小分子與K-554作用於同一受體。[1][2] *圖2｜K-554為每週注射胜肽設計，另兩項為早期小分子，其中一項處於先導化合物優化；三項不能統稱已進入人體的產品。[1][2]* 【03｜1月的時程承諾，要用後來的文件核對】 Kallyope在1月7日原本規劃年中啟動首次人體第一期研究，並於下半年取得單次遞增劑量研究的早期安全性、PK與PD資料。單次遞增劑量是分批評估不同單次劑量；PK看體內濃度與停留時間，PD看伴隨的生物作用。[2] 9月公開貼文仍以IND-ready描述K-554。這是公司對申請準備程度的說法，並非統一的法規核准級別，也不能獨立證明試驗已給藥。截至9月27日，可核對的公開資料仍不足以確認1月的計畫是否實現；登錄查詢若未取得可辨識紀錄，同樣不能據此認定未申請、未收案或已延遲。[1][2][5] 對商業判斷而言，等待的重點不是把階段名稱往後挪一格。首次人體資料要把藥物暴露、不良反應與探索性藥效放在一起看：若吃得少了，究竟是在什麼濃度下發生？不舒服是否同時加重？這才開始回答可用劑量範圍是否存在。 每週一次也只是設計目標。要支持這個間隔，需要看到作用能維持多久、重複給藥後的表現，以及不同患者間的變異。短期吃得少，還不足以說明長期能維持體重，更不能回答停藥後的結果。 *圖3｜IND-ready本身未回答監管、人體進度、安全性與PK／PD、藥效及較長追蹤的問題；這些證據可並行累積，待確認不代表未發生。[1][2][5]* 【04｜八年的合作，不能直接變成同一個資產故事】 兩家公司在2018年6月19日就簽過腸腦軸研究合作。諾和諾德可對合作發現的最多六項產品行使授權選擇權；這是權利範圍上限，並非六款藥已被證明成功。[3] 2024年9月10日，Kallyope公告諾和諾德已對合作找到的一個配體行權。配體是能與生物標的結合的分子，該次安排把後續開發、製造與商業化責任交給諾和諾德。[4] 這段歷史的價值，在於平台合作確實曾產生被選中繼續開發的成果。但公告沒有把2024年配體指認為K-554，不能把前案的進度或條件接到這次三項計畫上。2026年的新協議，要靠自己的資產資料與後續安排建立價值。 【05｜耐受性若改善，先改變的可能是誰能留下來】 藥時事團隊的判讀是，諾和諾德尋找非腸泌素路徑，值得用「有多少患者能取得並持續接受有效治療」來評估。既有Wegovy標籤列有噁心、嘔吐等腸胃不良反應，也記錄因不良反應停藥；這說明療效之外，使用負擔是實際的產品問題。[6] 若K-554未來能提供有意義的效果，又讓部分患者更容易持續治療，單藥就可能有自己的位置。但耐受性較佳的主張，必須在適當比較下成立。低劑量比較舒服、卻無法產生足夠藥效，不會自動形成一款好產品。 另一個情境是與腸泌素併用。假設兩條路徑能互補，研究便可以測試是否增加效果，或以不同劑量組合達到較好的整體治療表現。兩個機制不同，並不保證效益相加；身體訊號可能重疊，額外不良反應、給藥負擔與成本也可能抵銷好處。 對諾和諾德，單藥可能接觸原本療法未能滿足的患者，併用則可能延伸既有GLP-1的產品位置。兩者都會占用臨床與製造資源；若改善幅度太小，新增成本就可能比新增用途更快出現。 💼 支付方真正會比較的，是新療法多帶來什麼結果，值不值得增加費用。醫師則需要知道適合哪一群患者、如何追蹤效果與處理不適。如果優勢只存在於挑選過的少數受試者，市場範圍就與廣泛替代現有減重藥的想像不同。 即使後續試驗的平均減重令人滿意，也還要看退出研究的人是否被妥善納入分析，以及改善能維持多久。只盯完成治療者的曲線，可能高估真正開始用藥的人能得到多少效益。這會同時影響臨床採用與收入持續性，值得在方案設計時就問清楚。 製造和交付同樣要跟上。若最後走每週注射，製劑穩定性、批次品質與持續供應會進入成本；若小分子形成其他給藥選項，則要由各自的產品驗證支持。公開金額與付款結構未明，也讓投資人暫時無法判斷諾和諾德為這些選項支付了多高代價。 *圖4｜長期效果、耐受性、持續使用與單用／併用定位，需由相應人體資料共同支持；較少不舒服的假說本身，還不能換算成商業優勢。* 接下來最有辨識力的，是K-554第一份可讀的人體劑量資料：在同樣追蹤期間，什麼暴露能帶出食慾或藥效訊號，噁心與其他不適又落在哪裡？若兩者能拉出足夠的可用空間，才值得繼續投入更長期研究。 本文提供產業資訊與商業分析，不構成個別化醫療或投資建議。 --- ## Drug Approval in Hand—But Is the Cash in the Bank? How Formosa and PharmaEngine Turn Licensing Payments Into Cash URL: https://drugnews.com.tw/articles/2026-09-29-biotech-milestone-payments-formosa-pharmaengine-en.html Date: 2026-09-29 Language: en Summary: Historical Genmab milestones, Formosa's equity settlement and gross-profit-based collaboration, and PharmaEngine's first-half 2026 accounts show how contractual achievements become—or differ from—collectable proceeds. ────────── Formosa Pharmaceuticals’ ophthalmic drug received U.S. FDA approval in March 2024. Once the marketing authorization transfer was completed, the first US$2 million development milestone from licensing partner Eyenovia was triggered. Yet the payment was half cash and half shares. The collaboration subsequently ended, with US$2.2 million of obligations released. In 2026, PharmaEngine’s half-year financial report illustrated another gap: royalties had been recognized, while part of the amount remained reflected in contract assets and tax-related entries. [1][2][7] For investors, these distinctions affect whether a company can fund its next trial. Putting an entire announced milestone into a cash budget as soon as it is achieved can produce the wrong estimate of the funding gap. Formosa’s experience provides a concrete starting point: a transaction that has already passed through approval, payment and a change of partner. ## 【01｜A US$2 Million Milestone Was Achieved. How Much Reached the Bank?】 Eyenovia’s 2024 annual report states that the milestone required both FDA approval and the effective acceptance of the transfer and assignment of that approval. Those conditions were met on March 14, 2024. Eyenovia paid US$1 million in cash on April 26 and delivered shares for the other portion on April 29. [1] The number of shares was calculated using the contractual US$1 million amount and an earlier share price. At settlement, however, the financial report valued those shares at approximately US$0.4 million. Three different measures therefore coexist: the contractual basis for calculating the shares, their value at delivery, and the cash actually paid. [1] 💵 Receiving US$1 million in cash does not mean the other US$1 million is available, at face value, to pay R&D expenses. Converting shares into cash depends on the eventual sale price and their ability to be monetized. The US$0.4 million settlement value is not evidence of Formosa’s eventual share-sale proceeds. This structure preserves cash for the payer while transferring part of the funding risk to the recipient. For Formosa’s financial planning, a liquidity question remains between achieving the milestone and receiving equity. *Figure 1 | Achievement of a trigger, notification and invoicing, and collection are distinct states. Contractual procedures vary; an announcement does not automatically establish receipt of the full amount in cash. [1][10]* ## 【02｜Ending the Collaboration: Releasing US$2.2 Million Is Not Another Cash Receipt】 On June 6, 2025, Formosa and Eyenovia entered into a termination arrangement. Eyenovia’s second-quarter filing states that the relevant conditions were met in July, when its total obligations of US$2.2 million were extinguished. This settled rights and obligations between the parties; it was not another US$2.2 million paid to Formosa. [2] The joint announcement also explained that Eyenovia had suspended sales of the ophthalmic drug following corporate setbacks involving other programs. That turn of events illustrates an important point: after a product receives approval, commercialization still depends on the partner’s funding, commitment and execution. [11] A license therefore cannot be assessed from its headline amount alone. Responsibility for further development and launch, payment calculations, sales reporting, and the return of rights when a party exits all influence what the originator can retain. After termination, future revenue assumptions must be rebuilt around the revised rights and obligations. Finding another partner also requires launch preparations, supply coordination and market development to be connected again. Even with an approval in place, the time needed for a new partner to take over may delay cash returns. A partner’s willingness to keep allocating commercial resources deserves as much attention as the maximum amount in the contract. *Figure 2 | Development responsibilities, payment conditions, sales reporting and exit arrangements jointly determine transaction value. Exit provisions change rights and amounts owed; they do not create an additional revenue stream. [2][10]* ## 【03｜Harrow Takes Over: Payment Now Depends on First Sale and Gross Profit】 Three days later, on June 9, 2025, Harrow announced that it had obtained exclusive U.S. rights and marketing authorization for BYQLOVI. The product is clobetasol propionate ophthalmic suspension 0.05%, used to treat inflammation and pain after ocular surgery. Formosa’s APNT nanoparticle formulation technology addresses drug dispersion, dissolution and bioavailability. This is a product with a specific formulation and approved use; the commercial task is to translate it into prescriptions and sales. [3][4] Harrow’s 8-K describes the payment terms clearly: US$500,000 is payable when Harrow makes its first commercial sale to a third party. Formosa is also eligible for additional one-time payments tied to commercial gross-profit milestones and royalties calculated on the product’s gross profits. [3] That US$500,000 first-sale payment is not an upfront fee received on the signing date. “Gross profit” is the more consequential term to follow. If sales volumes increase but contractually defined costs also rise, the base available for sharing may not expand at the same pace. Multiplying a market-size estimate by an assumed royalty rate would miss this transaction’s actual payment structure. Gross profit is also different from net profit. The public 8-K does not fully disclose allowable cost deductions, milestone thresholds or royalty rates. Investors need subsequent, verifiable settlement information to assess the collaboration’s recurring returns. Formosa’s first-quarter 2026 financial report provides one concrete update: no revenue had been recognized under the Harrow agreement from signing through March 31, 2026. [12] That accounting statement is confined to the reporting period; September collection status requires later documentation. As of September 27, the subsequent public materials used in this article do not provide an explicit receipt date for the US$500,000 payment. ## 【04｜The Same US$25 Million Meant Two Different Milestones for Genmab】 🔎 A short comparison from 2017 is instructive. On January 6, Genmab announced that DARZALEX, or daratumumab, had reached US$500 million in calendar-year sales, triggering a US$25 million payment. On November 14 that year, another US$25 million payment was triggered by the first commercial sale in Japan. [8][9] The first reflected sales accumulating to a specified scale; the second reflected entry into a new market. The amounts were identical, but their implications for future revenue differed. Neither announcement was evidence of cash reaching the bank on that same day. The publicly filed Genmab–Janssen agreement, signed in 2012, also treats notification, invoicing and payment separately, with some deadlines redacted. It permits Genmab to issue an invoice before Janssen’s notification if Genmab becomes aware that a milestone has genuinely been achieved. [10] The order of notification and invoicing, and the payment deadline, depend on the terms governing each payment. *Figure 3 | Two US$25 million milestones in 2017: one triggered by US$500 million in calendar-year sales, the other by the first commercial sale in Japan. Equal amounts do not imply equal commercial maturity. [8][9]* ## 【05｜PharmaEngine Recognized US$9.795 Million. Why Is That Not the Same as Cash Collection?】 PharmaEngine’s ONIVYDE is a pancreatic-cancer medicine that encapsulates irinotecan in liposomes, using formulation technology to change drug delivery and release. Once the product reaches the market, PharmaEngine receives sales royalties under its licensing arrangements for Europe and Asia, excluding Taiwan. This provides a route for market revenue to flow back to the drug developer. [6][7] In the first half of 2026, PharmaEngine recognized US$9.795 million in sales royalties. The financial notes state that US$4.9 million of that amount had not been fully collected as of June 30. Of this, US$4.41 million was recorded as current contract assets; the remaining US$0.49 million was reflected in a reduction of current income tax liabilities. [7] 📘 The US$4.9 million is already included in the US$9.795 million: it is a subset, not an additional amount. When assessing bank cash, the US$0.49 million tax treatment must also be separated out. The entire US$4.9 million should not be described as a missing remittance. Reducing a liability by US$0.49 million does not increase bank deposits, but it also differs from a contract asset awaiting settlement. The note does not explain the underlying tax mechanism in further detail, so that mechanism should not be assumed. Current contract assets are not synonymous with overdue or uncollectible amounts. The classification alone does not establish late payment or justify describing the amount as a bad debt. Funding budgets must distinguish recognized income, amounts awaiting settlement, and cash actually available. This also explains why an income statement alone can give a misleading picture of financial runway. Licensing income in one period may include collected amounts, pending settlements and different tax treatments. Subtracting R&D expenditure from the revenue total is not enough to estimate how long the company can operate. Collection periods, cash flows and existing available funds must also be considered. The same half-year report separately records US$1.7 million of sublicensing income and explicitly states that it was fully collected during the first half. This establishes one collected income item, separate from the sales royalties. The collection status of other income still needs to be checked individually. [7] *Figure 4 | H1 2026: US$9.795 million of royalties includes US$4.9 million described as not fully collected. Within that amount, US$4.41 million is recorded as current contract assets and US$0.49 million is reflected in lower income tax liabilities. A separate US$1.7 million of sublicensing income was fully collected. These figures should not be double-counted. [7]* ## 【06｜From “Will Receive” to “Received”: One Additional Record Changes the Conclusion】 PharmaEngine’s US$50 million milestone in 2025 provides a more complete comparison. Its February 5 announcement stated that ONIVYDE’s 2024 sales in Europe and Asia had reached the second threshold and that the company would receive the payment. The product milestone page separately records receipt of the US$50 million in February 2025. [5][6] The first record establishes the trigger and notification; the second adds collection. That historical payment belongs in the relevant 2025 period, separately from first-half 2026 income. PharmaEngine’s sales royalties and Formosa’s new gross-profit sharing arrangements also have different settlement bases. The Drugnews team’s interpretation is that licensing income can support the next development program only to the extent that payment form, continuity of the collaboration and actual cash conversion allow it. Formosa’s new agreement ties part of its return to product gross profit. The next substantive question is how much sustainable, usable R&D cash Harrow’s first sale and subsequent settlements can bring back to Formosa. This article provides industry information and business analysis and does not constitute individualized investment advice. --- ## 藥證到手，錢就到帳？從台新藥、智擎看懂授權金怎麼變成現金 URL: https://drugnews.com.tw/articles/2026-09-29-biotech-milestone-payments-formosa-pharmaengine.html Date: 2026-09-29 Language: zh-Hant Summary: 藥證、付款事件與實際收款之間，有哪些條件？以Genmab歷史里程碑、台新藥的股票結算與毛利合作，以及智擎2026上半年財報，解析新藥授權生意如何從進展走向可運用資源。 ────────── 台新藥的眼科新藥在2024年3月取得美國FDA核准，完成藥證移轉後，觸發授權夥伴Eyenovia第一筆200萬美元開發里程碑。但真正支付時，一半是現金，一半是股票；後來合作終止，還有220萬美元義務被免除。到了2026年，智擎半年報又展示另一種差距：權利金已認列，部分款項仍在合約資產與稅務項目裡。[1][2][7] 對投資人而言，這些金額會影響公司能不能支應下一個試驗。看到「達標」就把整筆金額放進現金預算，可能連資金缺口都算錯。先看台新藥這筆已經走過核准、支付與換約的交易。 【01｜200萬美元達標，銀行收到多少？】 Eyenovia的2024年報記載，這筆里程碑除了FDA核准，還包括藥證移轉與承接生效；條件在2024年3月14日達成。4月26日支付100萬美元現金，4月29日再以股票完成另一部分支付。[1] 股票數量依契約約定的100萬美元及先前股價計算；到了交付日，財報列出的股票公允價值約40萬美元。這裡同時存在三個量：契約計數基礎、交付日估值，以及真正進銀行的現金。[1] 💵 已支付100萬美元現金，並不代表另外100萬美元也能原額拿去付研發費。股票若要變成現金，還得看實際出售價格與可變現條件；40萬美元也不是台新藥最後賣股所得的證明。 這種安排讓付款方保留現金，也把一部分資金風險交給收款方。對台新藥的資金規劃而言，里程碑達成與股票入帳之間，仍有一道流動性問題。 *圖1｜觸發條件、通知請款與收款是不同狀態；各合約對程序的安排不同，不能把一項公告自動接成全額現金入帳。[1][10]* 【02｜合作結束，免除220萬美元不是又收到一筆錢】 2025年6月6日，台新藥與Eyenovia簽署終止安排。對方第二季財報記載，相關條件於7月滿足，Eyenovia合計220萬美元義務因而解除。這是在結清雙方權利義務，不是台新藥另收220萬美元。[2] 原共同公告也交代，Eyenovia因其他計畫遇到公司層面的挫折，停止推進這款眼藥的銷售。這個轉折提醒投資人：產品取得藥證後，商業化仍依靠夥伴的資金、投入與執行能力。[11] 一份授權因此不能只讀價碼。誰負責後續開發與上市工作、付款如何計算、銷售如何回報，以及一方退出後權利怎麼回來，都會影響原廠還能留下什麼。終止之後，未來收入模型也必須按新的權利義務重估。 重新找到夥伴，還意味著上市準備、供應協調與市場拓展需要重新銜接。即使藥證留在，等待新夥伴接手的時間仍可能拉長現金回流。評估授權案時，對方願不願意持續配置商業資源，與合約列出的最高金額同樣值得看。 *圖2｜開發責任、付款條件、銷售回報與退出安排共同決定交易價值；退出條款會改變權利與款項，不是額外一筆收入。[2][10]* 【03｜Harrow接手，付款改看首賣與毛利】 三天後，2025年6月9日，Harrow宣布取得BYQLOVI在美國的獨家權利與上市許可。它是0.05% clobetasol propionate眼用懸浮液，治療眼科手術後的發炎與疼痛；台新藥的APNT奈米製劑技術著重藥物的分散、溶解與可利用性。這是有具體製劑與核准用途的產品，後面的考題是如何轉成處方與銷售。[3][4] Harrow的8-K把付款寫得很清楚：首次向第三方商業銷售時，支付50萬美元；另有依商業毛利里程碑觸發的一次性款項，以及按產品毛利計算的權利金。[3] 首賣50萬美元並非簽約當天就到帳的簽約金。更值得追的是「毛利」：若銷量增加，符合合約定義的成本也增加，可分潤的基礎未必同步擴大。只拿終端市場規模乘上一個猜測的權利金率，會漏掉這份交易真正的計價方式。 毛利也不等於淨利。哪些成本可以扣、里程碑門檻與分潤比例如何設定，這份公開8-K沒有完整揭露；投資人需要後續可核的結算資料，才能判斷合作的持續收益。 台新藥2026年第一季財報提供一個具體進度：從簽約到2026年3月31日，Harrow合約尚未認列收入。[12] 這個會計狀態止於該報告期，9月的收款情況仍需後續文件。截至9月27日，本文所採後續公開資料仍未提供那筆50萬美元的明確實收日期。 【04｜同樣2,500萬美元，Genmab兩次達標的意思不同】 🔎 2017年有一個很短、卻足以提醒投資人的對照。Genmab在1月6日公告，DARZALEX（daratumumab）曆年銷售達到5億美元，觸發2,500萬美元款項；同年11月14日，另一筆2,500萬美元則由日本首次商業銷售觸發。[8][9] 前一筆反映銷售已累積到約定規模，後一筆反映產品進入一個新市場。金額相同，對未來收入的意義卻不同；兩則公告也不是同日銀行收款憑證。 Genmab與Janssen於2012年簽訂的公開合約，更把通知、發票和付款分開處理，部分期限被遮蔽。契約還容許Genmab在知悉條件已達成時，先於對方通知開立發票。[10] 通知與請款的先後、付款期限，都要回到各筆款項的約定。 *圖3｜2017年兩筆各2,500萬美元：一筆由曆年銷售達5億美元觸發，另一筆由日本首次商業銷售觸發。相同金額不代表相同商業成熟度。[8][9]* 【05｜智擎已認列979.5萬美元，為什麼不能直接當收現？】 智擎的ONIVYDE是脂質體包覆irinotecan的胰臟癌用藥，透過製劑改變藥物遞送與釋放方式。產品有市場後，智擎依授權安排分享歐亞地區、台灣除外的銷售權利金；這是後續市場收入回流研發公司的途徑。[6][7] 2026年上半年，智擎認列銷售權利金979.5萬美元。財報接著說明，截至6月30日，其中490萬美元尚未收款完畢；再往下拆，441萬美元列流動合約資產，其餘49萬美元涉及本期所得稅負債減少。[7] 📘 490萬美元已包含在979.5萬美元裡，這是一層子集關係。拆解銀行現金時，還要把其中49萬美元的稅務處理分開，不能把整筆490萬美元視為尚缺匯款。這49萬美元的負債減少不等於現金存款增加，卻也不同於仍待結算的合約資產；公開附註未進一步交代的稅務機制，不宜自行填入。 流動合約資產也不是「逾期收不回」的同義詞。單看這個分類，無法判定對方遲付，更不能據此寫成壞帳。對公司的資金預算，仍要把已認列收益、待結算款項與真正可動用的現金分開。 這也解釋了為什麼只看損益表容易誤判資金續航。同一季的授權收入可能包含本期已收款、仍待結算，以及不同稅務處理；用總額直接減掉研發支出，還不足以推算公司能撐多久。需要再對照收款期間、現金流與既有可用資金。 同一份半年報另列170萬美元再授權收入，並明確說已於上半年全數收款。這筆款項與前述銷售權利金分開揭露，提供了一項已收款的收入證據；其他收入的收款狀態仍須逐筆對照。[7] *圖4｜2026H1：US$9.795m權利金內含US$4.9m未收款完畢項目，其中US$4.41m列流動合約資產、US$0.49m涉及所得稅負債減少；另有US$1.7m再授權收入已全收，勿重複加總。[7]* 【06｜把「將收」核成「已收」，差一份文件就差一種判斷】 智擎2025年的5,000萬美元里程碑，提供了完整一點的對照。2月5日公告說，因ONIVYDE的2024年歐亞銷售達到第二個門檻，公司「將收到」款項；產品里程碑頁則另記載，2025年2月已收到這5,000萬美元。[5][6] 前者證明觸發與通知，後者補上收款。這筆歷史款項應留在2025年的對應期間，與2026年上半年收入分開。智擎的銷售權利金和台新藥的新毛利分潤，後續也各有不同的結算基礎。 藥時事團隊的判讀是，授權收益能否支撐下一個研發計畫，取決於付款形式、合作能否延續，以及收入如何落入現金流。台新藥的新合約已把部分回報綁在產品毛利上。接下來真正值得等的是：Harrow的首賣與後續結算，能為台新藥帶回多少可持續、可用於研發的現金？ 本文提供產業資訊與商業分析，不構成個別化投資建議。 --- ## US$9.2 Billion in Annual Sales—Why Does J&J Still Want to Separate Orthopaedics? URL: https://drugnews.com.tw/articles/2026-09-28-jnj-orthopaedics-separation-en.html Date: 2026-09-28 Language: en Summary: Why might an orthopaedics business with US$9.2 billion in annual sales lose priority for J&J’s next investment? Growth differences, the value flows in a spin-off versus a sale, and the costs of implants and surgical support explain the economic questions behind the proposed DePuy Synthes separation. ────────── In 2024, Johnson & Johnson’s orthopaedics business generated approximately US$9.2 billion in sales. Yet in October 2025, the company announced a plan to separate the business so that DePuy Synthes could operate independently. Now a potential buyer has emerged for this substantial medical-device business: on September 11, Bloomberg reported, citing unnamed people familiar with the matter, that Apollo was discussing an acquisition at a potential valuation close to US$20 billion. [1][3] The J&J and Apollo announcements and J&J second-quarter filing reviewed as of September 27 do not establish that a definitive agreement for this transaction has been signed or that it has closed. The company continues to describe multiple separation paths. The separation plan is formally documented; the US$20 billion figure comes from reporting on talks. They carry different levels of certainty. [2][3][4] Why would a parent want to let go of a business generating more than US$9 billion in annual sales while outside capital wants to take it on? That contrast reveals more about the calculation inside a large pharmaceutical group than the reported price alone. ## 【01｜Orthopaedics Is Still Growing, but May Not Get the Next Investment】 When J&J announced the separation plan, it identified oncology, immunology and neuroscience in Innovative Medicine, and cardiovascular, surgery and vision in MedTech as its future priorities. The company wants to concentrate its portfolio on higher-growth, higher-margin markets. [1] This year’s figures make the choice easier to understand. 📊 In the first half of 2026, J&J’s orthopaedics sales were US$4.801 billion, with operational growth excluding currency effects of 3.7%. Cardiovascular devices recorded operational growth of 6.6% over the same period. Orthopaedics has not stopped growing, but the two businesses are growing at different rates. [2] Removing currency effects reduces the influence of changes in dollar translation and brings the comparison closer to the underlying sales trend. Growth is still only one part of capital allocation: a faster-growing business may require more upfront investment, while mature products may already have established customers and distribution. Management must consider profitability, investment needs and payback periods together. The two growth rates help explain the context for its choice; they cannot independently quantify the value a separation would create. US$9.2 billion answers the question, “How large is this business today?” Budget decisions also ask, “Where will the next dollar produce more growth?” Stable revenue, brands and hospital relationships are valuable, while research, surgical instruments and service teams require continuing investment. Within a large group, the same pool of capital can be allocated to joint replacements, cardiovascular devices or the next generation of oncology drugs. Those uses compete with one another. The Drugnews team’s interpretation is that orthopaedics is facing a question of investment priority within the group. J&J wants to concentrate its resources; DePuy Synthes could gain the opportunity to allocate capital around its own product cycles, research needs and customer base, without always competing with the other businesses on the same budget sheet. One part of the calculation is easy to overlook. Removing a slower-growing business may improve the remaining group’s reported growth profile, but the group also gives up that business’s future earnings. Whether shareholders are better off depends on the cash or equity received in exchange and on subsequent operating results. 【Figure 1｜One Orthopaedics Business, Two Approaches to Capital Allocation】 J&J wants to concentrate investment across the group, while DePuy Synthes would set priorities around its own products and customers. Operating results following a separation would determine whether growth and margins improve. ## 【02｜Distributing Shares and Selling for Cash Deliver Different Assets】 J&J’s original completion target was 18–24 months from the October 2025 announcement, subject to final board approval, applicable regulatory approvals and consultations with employee representatives. Its second-quarter 2026 filing continues to describe an evaluation of multiple paths. Different arrangements can therefore be compared, but the final structure cannot yet be filled in for the company. [1][2] 🏢 A share-distribution spin-off Under this structure, the parent distributes shares in the new company to its existing shareholders, who then directly own shares in two companies. The orthopaedics business’s future performance and share-price movements continue to affect their holdings. The parent does not receive an equivalent sale price simply by distributing the shares. [7] 💰 A sale to an outside buyer In a cash sale, the parent receives transaction consideration while giving up the future earnings of the business sold. The proceeds could be reinvested in other businesses, used to address debt or allocated in other ways, depending on the company’s decisions. They are not automatically distributed in full to shareholders. “Orthopaedics leaving J&J” may sound like the same outcome, but value can flow to different places. In one arrangement, shareholders continue to own an independent business; in the other, the parent exchanges future earnings for consideration today. Counting the number of companies after a separation does not establish which route is more attractive. 【Figure 2｜Follow the Value Before Focusing on the Transaction’s Name】 A share distribution and a sale allocate rights differently. The diagram illustrates structures, not terms that J&J has selected. ## 【03｜What Would Earn Back a US$20 Billion Investment?】 Dividing US$20 billion by US$9.2 billion produces approximately 2.2 times. The calculation is intuitive, but it only compares the scale of a reported valuation with 2024 sales. It is not an established transaction valuation multiple, much less a claim that the investment would pay for itself in two years: revenue must also cover manufacturing, research, sales and service costs. [1][3] 🔍 A buyer needs to estimate how much cash the business could retain each year as a standalone operation. The accessible reporting does not fully establish which assets the near-US$20 billion figure covers, how debt would be treated or whether the figure represents enterprise value or equity value. Even at the same final price, a buyer assuming debt or J&J retaining certain obligations would change the economics for each side. Taxes, separation expenditure and transaction costs would also affect J&J’s net proceeds. Another key issue is the cost of leaving the group. Which shared headquarters, information-system and support functions would need to be established independently? How much capital would instruments and inventory require? Those expenditures attract less attention than the acquisition price but affect the cash that can subsequently be retained. If Apollo ultimately acquires the business, its challenge would extend beyond obtaining an inexpensive asset. Changes to the product mix, market expansion or operating efficiency all require balancing capital returns with product investment. Cutting necessary research and clinical support could also weaken future revenue. 【Figure 3｜Read the Price and the Transaction Stage Separately】 The original figure identifies a September 21 evidence cutoff. This article updates the review scope using announcements and filings accessible on September 27. The near-US$20 billion figure remains treated as anonymously sourced reporting on talks. ## 【04｜Acquiring Orthopaedics Means Acquiring a Surgical Business】 Orthopaedic devices involve much more than selling an implant to a hospital. 🦴 Consider knee replacement. The implant reconstructs damaged joint surfaces; the procedure also involves bone cuts, ligament tension and lower-limb alignment, supported by measurement and cutting instruments. A product’s adoption in the operating room depends on far more than implant materials. Associated tools, the team’s operating experience and service support are also relevant. [6] J&J’s second-quarter filing notes that growth in its ATTUNE knee portfolio was partly supported by the VELYS robotic-assisted surgical system. Adoption of surgical tools can support implant sales. The two need to be managed together rather than treated as unrelated items on separate product lists. [2] That is part of the value of a large orthopaedics platform. Products, hospital customers, sales teams, surgeon training and clinical support must connect over time. Improvements in a mature business can come from a better product mix and investment in markets, not only from lowering unit manufacturing costs. Apollo’s own plans would need to be established through its formal disclosures; general business analysis should not be presented as a strategy the buyer has committed to. Viewed this way, scale in orthopaedics involves more than selling a wider range of implants. If existing products and tools continue to be adopted within the same surgical workflow, investment in research, education and local services has a business rationale. Evaluating a prospective owner therefore means asking which capabilities it intends to strengthen and which costs can genuinely be reduced without undermining product use and support. That is closer to the source of a medical-device business’s value than simply guessing whether an acquisition would bring layoffs. ## 【05｜A Direct Taiwanese Comparison: United Orthopedic】 Taiwan’s United Orthopedic Corporation (4129) has direct product-market overlap: hip and knee implants, associated surgical instruments, clinical education and customer services. Its U2 Knee AiO integrates femoral measurement and cutting functions, illustrating how an orthopaedics company connects implants with the surgical workflow. [5][6] The relevant factors are its own product adoption, international expansion and service costs, rather than converting an international transaction directly into assumed orders for a Taiwanese stock. How margins and expenses change as revenue grows is closer to the company’s operating quality. This is a product-market comparison; no supply or partnership relationship in this transaction has been established here. 【Figure 4｜Bring the Taiwanese Comparison Back to Products and Operations】 United Orthopedic is a local example of similar products and surgical support. Its own sales and delivery costs provide operating evidence that can be followed over time. ────────── The most revealing information will be DePuy’s standalone cash-flow profile and debt allocation. After financing, headquarters, instruments and clinical-support costs, how much could it still invest in new products? J&J faces its own test. Where will the capital and management resources previously devoted to orthopaedics go? Can those new investments deliver a better return than continuing to hold this mature business? The price at the time of separation cannot answer that half of the question. Once the transaction’s form is settled, who will fund the next orthopaedics development program, how much will they commit, and will that spending support surgeon adoption and product renewal? Those are the harder questions for this mature business. --- ## 年營收 92 億美元，強生為什麼還想把骨科拆出去？ URL: https://drugnews.com.tw/articles/2026-09-28-jnj-orthopaedics-separation.html Date: 2026-09-28 Language: zh-Hant Summary: 一年銷售92億美元的骨科業務，為何仍可能排不到強生下一筆資金？從成長差異、分拆與出售的價值流向，到人工關節與手術支援的經營成本，拆解DePuy Synthes分離案真正要算的帳。 ────────── 2024 年，強生（Johnson & Johnson，J&J）的骨科業務賣了約 92 億美元。到了 2025 年 10 月，公司仍宣布分離計畫，打算讓 DePuy Synthes 獨立經營。如今，這塊大型醫材資產又傳出買家：Bloomberg 9 月 11 日引述匿名知情人士報導，Apollo 正洽談收購，可能估值接近 200 億美元。【1】【3】 截至 9 月 27 日查閱的 J&J、Apollo 公告與 J&J 第二季財報，尚未取得本案正式簽約或交割的證據；公司仍保留多種分離途徑。分離計畫有正式文件，200 億美元則是洽談消息，兩者的確定程度不同。【2】【3】【4】 一個一年能賣 90 多億美元的事業，為什麼母公司想放手，外部資金卻想接手？這個反差，比傳聞中的價碼更能說明大型藥廠正在算哪一筆帳。 【01｜骨科還在成長，卻未必排得到下一筆資金】 J&J 在宣布分離時，把未來的重心放在創新藥的腫瘤、免疫、神經科學，以及 MedTech 的心血管、外科與視力。公司希望藉由調整組合，往成長更快、利潤率更高的市場集中。【1】 今年的數字，讓這個選擇更容易理解。 📊 2026 年上半年，J&J 骨科銷售為 48.01 億美元，排除匯率影響的營運成長為 3.7%；同期心血管醫材的營運成長為 6.6%。骨科並沒有停止成長，但兩塊業務的速度不同。【2】 排除匯率，是為了少受美元換算變動干擾，更接近業務本身的銷售走勢。不過，成長速度只是資本配置的一部分：更快的業務也可能需要更多前期投入，較成熟的產品則可能已有穩定的客戶與通路。公司還得把利潤、投資需求及回收時間放在一起看；兩個成長率能解釋這項選擇的背景，不能單獨算出拆分後多了多少價值。 92 億美元回答的是「今天這門生意有多大」，管理層分配預算時，還要問「下一美元投在哪裡，能換回更多成長」。穩定的收入、品牌與醫院關係有價值，研發、手術器械與服務團隊也要持續花錢。對大型集團來說，同一筆資金放在人工關節、心血管器材或下一代腫瘤藥物，本來就會被放在一起比較。 藥時事團隊的判讀是，骨科這次面對的是集團內部的投資順位。J&J 希望把資源集中，DePuy Synthes 則有機會依自己的產品週期、研發需求與客戶結構安排資本，而不必總在同一張預算表上與其他業務競爭。 這裡仍有一筆容易忽略的帳：移出成長較慢的業務，可能讓留下來的集團成長率變漂亮，但集團也失去了那塊業務未來的收益。要判斷股東是否更划算，得看換回的資金或股權，以及之後的經營成果。 【圖1｜同一塊骨科業務，兩種資本配置】 J&J 想集中集團投資，DePuy Synthes 則以自身產品與客戶需求決定優先順序。能否改善成長與利潤率，要由分離後的經營結果驗證。 【02｜分拆給股東，與出售換現金，拿到的東西不同】 J&J 原公告的完成目標是 18～24 個月，從 2025 年 10 月起算；仍需董事會最終批准、適用的監管批准及員工代表諮詢。2026 年第二季財報也保留評估多條路徑的說法。因此，現在可以比較不同安排，還不能替公司把交易形式填好。【1】【2】 🏢 股權分派式分拆（spin-off） 若採這種方式，母公司把新公司股份分給原股東，股東改為直接持有兩家公司。之後骨科的經營好壞、股價漲跌，仍會反映在他們持有的股份上；母公司不會只因分派股份，就收到一筆等額售價。【7】 💰 對外出售 若以現金出售，母公司取得交易對價，同時讓出所售業務的未來收益。這筆資金可以再投入其他業務、處理負債或作其他資本配置，具體用途取決於公司決定，並不是自動全部發給股東。 所以「骨科離開 J&J」看似是同一個結果，價值卻可能走向不同地方。一邊是股東繼續持有一門獨立生意，另一邊是母公司拿未來收益交換現在的對價。哪條路更划算，不能只看拆分後有幾家公司。 【圖2｜先看價值流向，再看交易名稱】 股份分派與出售的權利安排不同；圖中是結構示意，不是 J&J 已選定的條款。 【03｜200 億美元買下來，要靠什麼收回投資？】 把 200 億美元除以 92 億美元，約是 2.2 倍。這個算式很直覺，卻只是在比較傳聞估值與 2024 年銷售的量級，不是正式估值倍數，更不是「兩年就能回本」：營收還得支付製造、研發、銷售與服務的成本。【1】【3】 🔍 買方真正需要估的是：獨立經營之後，每年能留下多少現金？ 目前可讀的報導，沒有完整釐清近 200 億美元究竟涵蓋哪些資產、如何處理負債，以及是企業價值還是股權價值。即使最終價格相同，買方若需承接債務，或 J&J 仍須保留部分責任，雙方得到的經濟結果也會不同。對 J&J 而言，稅務、分離費用與交易成本還會影響可留下的淨款。 另一個關鍵是離開集團後的成本。原本共用的總部、資訊系統及支援功能，哪些要獨立建立？器械與庫存需要多少資金？這些支出不如收購價醒目，卻會影響往後能留下的現金。 若 Apollo 最終接手，它面對的也不只是「買到便宜資產」的問題。調整產品組合、擴展市場或提高營運效率，都需要把資金回報與產品投資放在同一張表上；削掉必要的研發和臨床支援，可能同時削弱未來收入。 【圖3｜價碼與交易階段，分開判讀】 原圖標示 9 月 21 日的查核截點；本文已按 9 月 27 日可讀公告與財報更新查閱範圍。近 200 億美元仍以匿名媒體洽談消息處理。 【04｜接手骨科，接手的是整套手術生意】 骨科器材的生意，比「把植入物賣給醫院」複雜得多。 🦴 以人工膝關節為例，植入物用來重建受損的關節面；手術還要處理骨切面、韌帶張力與下肢力線，搭配量測及截骨器械。產品能不能進入手術室，牽涉的遠不只是植入物材料，還包括配套工具、團隊的操作經驗與服務支援。【6】 J&J 第二季財報提到，ATTUNE 膝關節產品的成長，部分受到 VELYS 機器人輔助手術系統帶動。對公司而言，手術工具的採用可以支持植入物銷售；兩者需要一起經營，而不是各自放在互不相干的產品清單裡。【2】 這也是大型骨科平台的價值所在。產品、醫院客戶、銷售團隊、醫師訓練與臨床支援，要能長期銜接。成熟業務的經營改善，可以來自更好的產品組合和市場投入，不只剩下壓低單件製造成本。至於 Apollo 如何規劃這些工作，仍要看其正式披露，不能把一般商業分析寫成買方已承諾的策略。 從這個角度看，一家骨科公司的規模，不只是多賣幾種植入物。既有產品與工具若能在同一套手術流程裡被持續採用，投入研發、教育及在地服務就有經營上的意義。評估接手方時，也就值得問：它準備把哪些能力做得更強，哪些成本確實可以降低，而不傷到產品使用與服務？這比只猜收購後會不會裁員，更接近一門醫材生意的價值來源。 【05｜台灣的直接對照：聯合骨科】 台灣的聯合骨科（4129）有直接的產品市場重疊：人工髖、膝關節與配套手術器械，也提供臨床教育和客戶服務。它的 U2 Knee AiO 整合股骨量測及截骨功能，正好說明骨科公司如何把植入物與手術流程接在一起。【5】【6】 這個例子值得看的，是自己的產品採用、海外市場拓展與服務成本，而非把國際交易直接換算成台股訂單。收入增加後，毛利與費用怎麼變，才更接近一間骨科公司的經營品質。本文採用的是產品市場的比較，沒有確認本案的供應或合作關係。 【圖4｜把台股比較放回產品與經營】 聯合骨科提供的是同類產品與手術支援的本地案例；它自己的銷售與交付成本，才是可持續追蹤的經營證據。 ────────── 接下來最有辨識力的資料，是 DePuy 獨立經營後的現金流與負債配置：付完融資、總部、器械及臨床支援成本，還能投入多少新產品？ J&J 也有自己的考題。原本投入骨科的資本與管理資源，最後流向哪裡？新的投資能否做出比繼續持有這塊成熟業務更好的回報？只看分離當下的價碼，還回答不了這一半。 交易形式確定之後，下一筆骨科研發費由誰支付、願意付多少，以及能否換來醫師採用與產品更新，才是這門成熟生意更難回答的問題。 【資料來源】 【1】J&J，骨科業務分離意向公告，2025-10-14。 https://www.jnj.com/media-center/press-releases/johnson-johnson-announces-intent-to-separate-its-orthopaedics-business 【2】J&J，2026 年第二季 Form 10-Q；報告期截至 2026-06-28，正文簽署日 2026-07-23。 https://www.sec.gov/Archives/edgar/data/200406/000020040626000153/jnj-20260628.htm 【3】Bloomberg，Apollo 洽購 J&J 骨科業務報導，2026-09-11 22:26 UTC（台北 9/12）；僅引用公開可讀段落。 https://news.bloomberglaw.com/pharma-and-life-sciences/apollo-global-is-said-in-talks-to-acquire-j-js-orthopedics-unit 【4】截至 2026-09-27 查閱的 J&J、Apollo 公告列表。當次可讀範圍分別為 9/10–9/25、8/11–9/17；不是對未索引頁面或未公開資訊的排除證明。 https://www.jnj.com/media-center/press-releases https://ir.apollo.com/news-events/press-releases 【5】聯合骨科，公司簡介與股務資訊；2026-09-27 查閱。 https://tw.unitedorthopedic.com/about-us https://tw.unitedorthopedic.com/investor/shareholder5/ 【6】聯合骨科，人工膝關節置換及 U2 Knee AiO 產品頁；2026-09-27 查閱。 https://tw.unitedorthopedic.com/total-knee-replacement/ https://tw.unitedorthopedic.com/innovation/u2-knee-aio 【7】SEC Investor.gov，Spin-Offs；2026-09-27 查閱。 https://www.investor.gov/introduction-investing/investing-basics/glossary/spin-offs 本文為產業資訊與商業分析，不構成個別投資建議。 #JohnsonAndJohnson #JNJ #DePuySynthes #Apollo #骨科 #醫療器材 #MedTech #聯合骨科 #4129 #藥時事 --- ## Ivonescimab Moves Beyond Lung Cancer: Phase 3 Survival Win Against the Biliary Tract Cancer Standard URL: https://drugnews.com.tw/articles/2026-09-04-ivonescimab-biliary-tract-cancer-harmoni-gi1-en.html Date: 2026-09-04 Language: en Summary: Ivonescimab plus chemotherapy beat durvalumab plus chemotherapy on overall survival in the Phase 3 HARMONi-GI1 interim analysis, but the hazard ratio, median survival, curves, safety and subgroup results remain undisclosed. For the past several years, the most recognizable part of the ivonescimab story has been its willingness to challenge Keytruda directly in lung cancer. Now, for the first time, that story has moved beyond lung cancer. On August 26, 2026, Akeso announced that ivonescimab plus chemotherapy met the primary endpoint of overall survival in the prespecified interim analysis of HARMONi-GI1, a Phase 3 study in first-line advanced biliary tract cancer. The comparator was not placebo or chemotherapy alone. It was durvalumab, marketed as Imfinzi, plus chemotherapy, one of the current first-line standards of care. In other words, ivonescimab was tested directly against an established global treatment standard. The bar was high from the outset. Akeso also reported that the study met key secondary endpoints of progression-free survival, or PFS, and objective response rate, or ORR. PFS measures how long patients remain alive without their disease worsening, while ORR measures the proportion of patients whose tumors shrink by a predefined amount. The result gives ivonescimab its first positive Phase 3 readout in a gastrointestinal tumor and offers its first clinical evidence that the drug's development story may extend from lung cancer into another cancer type. The numbers the market most wants to see, however, have not yet been disclosed. The company has not reported the magnitude of the reduction in the risk of death, the difference in median overall survival between the two groups, or the point at which the survival curves separate. Complete safety results and outcomes across the different biliary tract cancer subtypes have also not been released. What can be said today is that the trial succeeded at its interim analysis. What cannot yet be said is how large, durable, or broadly applicable that success will prove to be. This is an important win. The contest that will determine how far ivonescimab can ultimately go has only just begun. ## 1 | Why Biliary Tract Cancer Is Difficult for More Than Its Small Patient Population Biliary tract cancer is a group of malignancies arising from the bile ducts, gallbladder, and related anatomical sites. It receives less public attention than lung or breast cancer, but it is especially difficult to treat in clinical practice. Early symptoms are often subtle, so many patients are diagnosed only after surgery is no longer possible. The biological differences among tumors arising in different locations also make it difficult for one treatment approach to benefit every patient to the same degree. For a long time, the backbone of first-line therapy was chemotherapy with gemcitabine and cisplatin. The treatment landscape began to shift only after immunotherapy was added to that backbone. In the Phase 3 TOPAZ-1 study, durvalumab plus chemotherapy reduced the risk of death by 20% compared with chemotherapy alone, with a hazard ratio, or HR, of 0.80. A hazard ratio compares the rate at which an event occurs between two treatment groups. Longer-term follow-up continued to show a survival benefit, and the regimen became a first-line standard adopted by guidelines in multiple countries. The Phase 3 KEYNOTE-966 study of pembrolizumab, marketed as Keytruda, plus chemotherapy also showed a survival benefit. The regimen reduced the risk of death by 17%, corresponding to a hazard ratio of 0.83. Median overall survival was 12.7 months in the pembrolizumab group and 10.9 months in the control group. These studies changed treatment, but the size of their gains also left substantial room for improvement. First-line immunotherapy in biliary tract cancer now has established answers, yet clinicians and patients are still waiting for an approach that can push the survival curve meaningfully higher. That is precisely the position ivonescimab chose to challenge. ## 2 | What Did This Head-to-Head Trial Actually Compare? HARMONi-GI1, also known as AK112-309, is a single-region, multicenter, randomized, double-blind, registrational Phase 3 trial conducted in China. Participants received either ivonescimab plus chemotherapy or durvalumab plus chemotherapy as first-line treatment for advanced biliary tract cancer. The design therefore compares two complete treatment strategies directly within the same study rather than asking readers to infer a difference from separate trials conducted in different populations and at different times. The primary endpoint is overall survival, or OS: the time from a patient's entry into the study until death. In oncology, overall survival is one of the most clinically consequential endpoints and one of the hardest to make look better through presentation alone. Akeso reported results from a prespecified interim analysis assessed by an independent data monitoring committee. According to the company's announcement, ivonescimab produced a statistically significant and clinically meaningful overall survival advantage. The study also met its key secondary endpoints of progression-free survival and objective response rate. Clearing all three endpoints makes the result more persuasive than a report that tumors simply shrank more often. At least directionally, the findings indicate that patients lived longer, remained free from disease progression for longer, and achieved tumor responses more frequently. Those three measures point in a consistent direction. But a topline announcement supplies only the outline. Five pieces of information remain essential. First, the overall survival hazard ratio has not been disclosed. A result of 0.80, 0.70, or lower would imply materially different clinical and commercial value. Second, median overall survival has not been reported. Statistical significance does not mean that every patient receives the same extension of life, and the absolute difference must still be expressed in months. Third, the survival curves have not been shown. Clinicians will want to know whether the advantage appears early and continues to widen or emerges only later in follow-up, because the shape and timing of separation affect how durable the benefit appears. Fourth, complete safety data have not been disclosed. Because ivonescimab acts on both immune and angiogenic pathways, rates of bleeding, hypertension, proteinuria, immune-related adverse events, and treatment discontinuation need to be evaluated alongside those in the control group. Fifth, subgroup results remain unavailable. Whether patients with intrahepatic cholangiocarcinoma, extrahepatic cholangiocarcinoma, and gallbladder cancer benefit consistently will help determine how much of the broader biliary tract cancer population this regimen could realistically cover. The most accurate conclusion at this stage is therefore narrow but meaningful: ivonescimab won the prespecified interim analysis against a strong comparator. The size and completeness of that win must await presentation at a scientific meeting and publication in a peer-reviewed paper. ## 3 | Why Might One Bispecific Antibody Differ From Combining Two Separate Drugs? Ivonescimab is a tetravalent bispecific antibody designed to target both PD-1 and VEGF. PD-1 functions as an immune brake on the surface of T cells. Cancer cells exploit related signaling to weaken T-cell activity. Blocking PD-1 can help the immune system recognize and attack tumors again. VEGF, or vascular endothelial growth factor, promotes the formation of tumor blood vessels. It can also contribute to a hypoxic and disorganized tumor microenvironment that makes it harder for immune cells to enter. Inhibiting VEGF may therefore do more than reduce angiogenesis; it may also improve the conditions under which immune cells reach the tumor. Ivonescimab places both functions within a single molecule. Its four binding sites are intended to increase binding in tumor microenvironments where PD-1 and VEGF are both abundant. The design objective is straightforward: concentrate immune release and anti-angiogenic activity in the same location, potentially increasing tumor-localized activity while reducing unnecessary exposure in normal tissues. This architecture creates a pharmacological possibility that may differ from simply administering an immune checkpoint inhibitor and an anti-angiogenic agent as two separate medicines. It does not, by itself, prove a clinical advantage. The contribution of each mechanism, and whether the design produces a better therapeutic window, must be established by patient data rather than inferred from a diagram. The significance of HARMONi-GI1 is that this mechanism has now produced a positive Phase 3 signal outside lung cancer while facing a strong active comparator. It is an important human efficacy signal, not a license to treat the mechanism as proven across all tumors. ## 4 | Why Does Akeso Keep Choosing the Hardest Comparators? Many new cancer medicines first demonstrate activity against placebo or chemotherapy and only later seek a place beside an established standard. Akeso has chosen a different development path for ivonescimab: compare it directly with widely used immunotherapy regimens. In first-line PD-L1-positive non-small cell lung cancer, HARMONi-2 tested ivonescimab monotherapy directly against Keytruda. In first-line squamous non-small cell lung cancer, HARMONi-6 compared ivonescimab plus chemotherapy with another PD-1 antibody plus chemotherapy. In biliary tract cancer, the comparator became durvalumab plus chemotherapy. This strategy magnifies both risk and reward. If the trial succeeds, the product does not have to rely on a narrative assembled through cross-trial comparisons. Clinicians, regulators, and payers can examine the difference between the new regimen and the existing standard in patients enrolled under the same protocol, during the same period, and under the same trial rules. If the difference is small, however, the market can immediately focus on every hazard ratio, every subsequent data update, and every feature of the survival curves. A direct comparison leaves less room to explain away an ambiguous result. A head-to-head trial is therefore an expensive form of self-discipline. The sponsor voluntarily removes many of the easiest arguments it could otherwise use and subjects the product's value to a common patient population and a common set of study conditions. Ivonescimab has already built visibility in lung cancer. The importance of the biliary tract cancer result is that it is the first indication that this strategy of challenging a leading comparator may also work in another tumor category. ## 5 | Can a Lung Cancer Drug Become a Pan-Tumor Platform? Akeso has stated that five Phase 3 studies of ivonescimab have now produced positive results. Four are in non-small cell lung cancer, while HARMONi-GI1 is the first positive Phase 3 readout in a gastrointestinal tumor. That count changes the way the market frames the question. When a medicine succeeds in only one lung cancer population, investors mainly estimate the commercial opportunity in that indication. When the same molecule repeatedly produces positive signals across treatment lines, comparators, and cancer types, the market begins to ask whether it could become a broader platform and whether the probability assigned to success in the next tumor type should be revised upward. Akeso is also advancing studies in triple-negative breast cancer, head and neck cancer, small cell lung cancer, colorectal cancer, and pancreatic cancer. Summit is pursuing multiple global Phase 3 programs in its licensed territories, including studies in lung and colorectal cancer. There are still firm limits to extrapolation across tumors and regions. HARMONi-GI1 was conducted entirely in China, and its data were generated, managed, and analyzed by Akeso. Positive results in Chinese patients cannot automatically replace the multiregional clinical evidence needed in the United States, Europe, or other jurisdictions. Tumor immune environments, treatment backbones, and patient characteristics also differ across cancers. One successful trial cannot issue an advance pass to every program in the pipeline. Even with those limits, the result moves the question of whether ivonescimab can extend beyond lung cancer substantially closer to a clinical answer. ## 6 | What Does the Result Mean for Akeso, Summit, and AstraZeneca? The global rights to ivonescimab are divided into two broad territories. Summit holds rights in the United States, Canada, Europe, Japan, Latin America, the Middle East, and Africa. Akeso retains rights in the remaining markets. Ivonescimab is currently approved only in China for certain lung cancer indications and remains an investigational therapy in Summit's licensed territories. For Akeso, a positive biliary tract cancer study expands the potential breadth of the product in China and increases the credibility of its broader bispecific-antibody platform. If the complete data are sufficiently strong, the next tests will involve regulatory submission, reimbursement access, physician adoption, and commercial execution. A positive trial is the beginning of that sequence rather than its conclusion. For Summit, the Chinese data provide meaningful clinical de-risking. The same molecule has now shown an overall survival advantage outside lung cancer, giving the platform valuation a new point of support. Summit cannot, however, convert a single-region Chinese result directly into approval in the United States or Europe. The overseas development pathway, regional regulatory requirements, and any decision to initiate a global biliary tract cancer study remain unfinished work. For AstraZeneca, the announcement does not mean that Imfinzi has already lost its position as a standard of care. Durvalumab plus chemotherapy is supported by a mature Phase 3 dataset, longer-term follow-up, and approvals in multiple countries. Ivonescimab currently has only a topline announcement in this setting. Only after the complete data are available will it be possible to compare the magnitude of efficacy, safety, administration, and price in a way that could determine whether the first-line standard in biliary tract cancer is genuinely changing. ## 7 | If the Full Dataset Is Strong, What Could Change in Clinical Practice? What patients with biliary tract cancer need is not an impressive press release. They need more survival time of acceptable quality. First-line treatment is usually the starting point for the entire treatment journey. When patients begin systemic therapy, their physical condition, organ function, and ability to tolerate adverse effects are often relatively intact. If disease can be controlled longer at this stage, the consequences may extend to whether patients can receive second-line treatment and whether they can return to parts of daily life. An overall survival advantage is therefore only the first gate. Clinicians will also ask practical questions. How frequently must patients come to the hospital? Can the adverse effects of the regimen be managed when it is combined with chemotherapy? Are bleeding or thrombotic risks concentrated in particular groups? Can patients with jaundice, biliary infection, or impaired liver function receive it safely? Does the efficacy gain come with a higher rate of treatment discontinuation? Payers will approach the decision differently. Durvalumab and Keytruda already have established procurement, reimbursement, and treatment experience. A new regimen seeking to displace them must offer more than longer survival. It must also present a price, administration schedule, and management burden that healthcare systems can support. This is the distance between statistical significance and a genuine change in clinical standards, and it is often underestimated by the market. A drug first wins within a trial. It must then persuade physicians to switch, fit within hospital workflows, and remain affordable for patients and health systems. Only after those steps can trial success translate into real-world market share. For Akeso, the second half of the test will shift from development courage to commercial execution. The company will need to communicate the complete dataset clearly, identify the patients most likely to benefit, and demonstrate that manufacturing capacity and supply can support an additional indication. A head-to-head victory can open the door, but it does not automatically move the market through it. ## 8 | The Next Destination Is Not Applause but Five Complete Scorecards Ivonescimab has met a consequential endpoint in biliary tract cancer. The study also presents a notable image for China's innovative drug industry: a new therapy is no longer being tested only against chemotherapy but is challenging an immunotherapy standard that has already been validated globally. Investors and clinicians will now focus on five scorecards: the overall survival hazard ratio, median overall survival, the survival curves, complete safety results, and outcomes across the three principal biliary tract cancer subtypes. If those five scorecards tell a consistent story, the opportunity for ivonescimab could extend beyond adding one more indication. PD-1/VEGF bispecific therapy would begin to move from being a lung cancer headline toward the definition of a pan-tumor platform. It would also force the next generation of immuno-oncology medicines to answer a sharper question: in a market where a PD-1 or PD-L1 standard already exists, can the new therapy defeat it directly? If the margin of benefit is small, or if the safety cost is too high, the glow surrounding the head-to-head result could fade quickly. The most important conclusion today is therefore not that ivonescimab has already rewritten the standard of care in biliary tract cancer. The more accurate conclusion is that, for the first time, it has moved beyond lung cancer, stepped into the ring against an established biliary tract cancer standard, and won the overall survival point at a prespecified interim analysis. The remaining points belong to the complete dataset. ## Disclaimer This article is provided for information about the pharmaceutical and biotechnology industry. It does not constitute medical advice or investment advice. Patients should consult qualified healthcare professionals when making treatment decisions. --- ## Ivonescimab 跨出肺癌！正面挑戰膽道癌金標準，總生存期告捷 URL: https://drugnews.com.tw/articles/2026-09-04-ivonescimab-biliary-tract-cancer-harmoni-gi1.html Date: 2026-09-04 Language: zh-Hant Summary: Ivonescimab 合併化療在一線晚期膽道癌三期 HARMONi-GI1 達到總生存期主要終點，正面勝過 durvalumab 合併化療；完整風險比、生存曲線、安全性與亞組數據仍待公布。 ━━━━━━━━━━━━ 過去幾年，ivonescimab（依沃西）最響亮的標籤，是敢在肺癌正面挑戰 Keytruda。 現在，它第一次把戰線推到了肺癌之外。 2026 年 8 月 26 日，康方生物宣布，ivonescimab 合併化療，在一線晚期膽道癌三期 HARMONi-GI1 研究的預設期中分析中，達到總生存期主要終點。更關鍵的是，對照組並非安慰劑或單純化療，而是目前一線標準之一：durvalumab（度伐利尤單抗，商品名 Imfinzi）合併化療。 換句話說，新藥被直接推上世界級標準治療的擂台，眼前的欄杆一點也不低。 康方同時表示，無惡化生存期（PFS，疾病沒有惡化的時間）與客觀緩解率（ORR，腫瘤縮小達到標準的患者比例）也達到關鍵次要終點。ivonescimab 因此拿下第一個消化道腫瘤三期陽性結果，也第一次證明它的故事有機會從肺癌延伸到另一個癌種。 市場最想看的數字，卻還沒有出現。 公司目前沒有公布死亡風險降低多少、兩組中位總生存期相差多久、生存曲線何時分開，也沒有交代完整安全性與不同膽道癌亞型的結果。現階段可以確認的是「試驗成功」，還不能回答「成功得有多漂亮」。 這場勝利很重要；真正決定它能走多遠的比賽，才剛開始。 ──────────────────── ## 🩺 01｜膽道癌的難，不只在於病人少 膽道癌是一組從膽管、膽囊與相關部位長出的惡性腫瘤。它不像肺癌、乳癌那樣常被大眾談起，臨床上卻格外棘手：早期症狀不明顯，很多患者被發現時已無法手術；不同發生位置又帶來明顯的生物學差異，使同一套療法很難平均照顧所有患者。 過去很長一段時間，一線治療的骨架都是 gemcitabine（吉西他濱）加 cisplatin（順鉑）化療。免疫治療加入後，格局才開始鬆動。 阿斯特捷利康的 TOPAZ-1 三期研究顯示，durvalumab 合併化療，相較單純化療將死亡風險降低 20%，風險比（HR，用來比較兩組事件發生速度的指標）為 0.80。更長期追蹤仍維持生存優勢，這套療法也成為多國指南採用的一線標準。 默沙東的 KEYNOTE-966 三期研究則顯示，pembrolizumab（帕博利珠單抗，商品名 Keytruda）合併化療，死亡風險降低 17%，風險比為 0.83，中位總生存期為 12.7 個月，對照組為 10.9 個月。 這些結果改變了治療，但改善幅度仍留下很大空間。膽道癌的一線免疫治療已經有標準答案，臨床依舊在等一個能把生存曲線再往上推的方案。 ivonescimab 這次挑的，正是那個已經站穩位置的答案。 ──────────────────── ## ⚔️ 02｜這場「頭對頭」，到底比了什麼？ HARMONi-GI1，又稱 AK112-309，是一項在中國進行的單區、多中心、隨機、雙盲、註冊性三期試驗。受試者接受 ivonescimab 合併化療，或 durvalumab 合併化療，研究直接比較兩套方案作為一線晚期膽道癌治療的效果。 它的主要終點是總生存期（OS），也就是患者從進入研究到死亡的時間。對癌症治療而言，這是最難被包裝、也最有臨床分量的終點。 康方公布的是預設期中分析，資料由獨立資料監測委員會評估。結果顯示，ivonescimab 組在總生存期上取得具統計顯著性且有臨床意義的優勢；無惡化生存期與客觀緩解率也同步達標。 三個終點一起過關，讓這項結果比「腫瘤縮小得比較多」更有說服力。患者活得更久、疾病惡化得更慢、腫瘤縮小比例更高，三條線至少在方向上彼此呼應。 不過，頂線新聞稿只畫出了輪廓。 ① 第一，總生存期的風險比尚未公布。0.80、0.70 或更低，代表的臨床與商業價值會明顯不同。 ② 第二，中位總生存期尚未公布。統計顯著不等於每位患者都得到同樣幅度的延長，絕對差距仍要看實際月份。 ③ 第三，生存曲線尚未公布。優勢是早早出現、持續拉開，還是在後段才形成，會影響醫師對療效穩定性的判斷。 ④ 第四，完整安全性尚未公布。ivonescimab 同時作用於免疫與血管新生路徑，出血、高血壓、蛋白尿、免疫相關副作用及停藥比例，都要與對照組放在一起看。 ⑤ 第五，亞組結果尚未公布。肝內膽管癌、肝外膽管癌與膽囊癌是否一致受益，將決定這套療法真正能覆蓋多大的患者範圍。 因此，今天最準確的說法是：ivonescimab 贏下了預設期中分析，而且對手很強；完整勝差仍待學術會議與同儕審查論文揭曉。 ──────────────────── ## 🧬 03｜為什麼一支雙抗，可能和兩支藥聯用不同？ ivonescimab 是一種同時鎖定 PD-1 與 VEGF 的四價雙特異性抗體。 PD-1 是 T 細胞表面的免疫煞車。癌細胞藉由相關訊號讓 T 細胞失去攻擊力；阻斷 PD-1，可以讓免疫系統重新辨認並攻擊腫瘤。 VEGF 是血管內皮生長因子。它會促進腫瘤血管形成，也會讓腫瘤微環境更缺氧、更混亂，進一步阻礙免疫細胞進入。抑制 VEGF，除了減少新生血管，也可能改善免疫細胞進入腫瘤的條件。 ivonescimab 把兩種功能放進同一個分子，並利用四個結合位點提高在同時富含 PD-1 與 VEGF 的腫瘤微環境中的結合力。設計目標很清楚：讓免疫解鎖與抗血管新生在同一處發生，提高腫瘤局部作用，同時降低正常組織不必要的暴露。 這套結構給了它與一般「免疫藥加抗血管藥」不同的藥理可能性。真正的臨床優勢仍須由數據決定。哪一種機制貢獻最多、是否換來更好的安全治療窗，都不能只靠示意圖下結論。 HARMONi-GI1 的價值，正是在最強對照面前讓這套機制第一次跨出肺癌，留下人體三期的陽性訊號。 ──────────────────── ## 🥊 04｜康方為什麼總愛挑最難的對手？ 很多新藥會先對安慰劑或化療證明有效，再尋找與標準療法並排的位置。康方替 ivonescimab 選擇的是另一條路：直接和已被廣泛採用的免疫治療正面比較。 在 PD-L1 陽性的一線非小細胞肺癌，HARMONi-2 讓 ivonescimab 單藥對上 Keytruda；在一線鱗狀非小細胞肺癌，HARMONi-6 讓 ivonescimab 合併化療對上另一款 PD-1 抗體合併化療；到了膽道癌，對手換成 durvalumab 合併化療。 這種策略把風險與報酬同時放大。 一旦成功，產品不必再靠跨試驗比較拼湊故事，醫師、監管機構與支付方可以直接看到它和現行標準的差距。若結果只有些微波動，市場也會立刻放大每一個風險比、每一次資料更新與每一條生存曲線。 頭對頭研究因此是一種昂貴的自我約束：公司主動把最容易辯解的空間拿掉，讓產品價值接受同一批患者、同一個時間與同一套試驗規則的檢驗。 ivonescimab 已經在肺癌建立聲量。膽道癌的意義，是它第一次證明這種「挑最強對手」的打法，有機會在另一個腫瘤領域延續。 ──────────────────── ## 🧭 05｜從一款肺癌藥，走向一個泛癌種平台？ 康方表示，ivonescimab 至今已有五項三期研究讀出陽性結果：四項來自非小細胞肺癌，HARMONi-GI1 則是第一項消化道腫瘤三期陽性。 這個數字改變了市場提問的方式。 當一款藥只在單一肺癌族群成功，投資人計算的是那個適應症能賣多少；當同一個分子在不同治療線別、不同對照與不同癌種中反覆出現陽性訊號，市場開始評估它能否形成平台，下一個癌種的成功機率是否應該被上調。 康方目前還在三陰性乳癌、頭頸癌、小細胞肺癌、結直腸癌與胰臟癌等領域推進研究。Summit 也在其授權區推進多項全球三期，包括肺癌與結直腸癌。 跨癌種外推仍有清楚邊界。膽道癌這項研究完全在中國進行，所有資料由康方產生、管理與分析；中國患者的陽性結果，不能自動替代美國、歐洲或其他地區所需的多區域臨床證據。不同癌種的免疫環境、治療骨架與患者組成也各不相同，一場成功無法替所有管線預先簽發通行證。 但它至少讓「ivonescimab 能否跨出肺癌」從假說往前走了一大步。 ──────────────────── ## 💰 06｜對康方、Summit 與阿斯特捷利康，分別意味著什麼？ ivonescimab 的全球權利被分成兩塊。Summit 擁有美國、加拿大、歐洲、日本、拉丁美洲、中東與非洲等地的權利；康方保留其餘市場。ivonescimab 目前只在中國獲批部分肺癌適應症，在 Summit 的授權區仍屬研究中療法。 對康方而言，膽道癌陽性增加的是產品在中國的適應症寬度，也提高整套雙抗平台的可信度。若完整數據夠強，下一步才會進入監管申報、醫保准入、醫師採用與商業放量的考驗。 對 Summit 而言，這份中國資料具有重要的「臨床去風險」價值：同一個分子在肺癌之外仍能打出總生存期優勢，平台估值有了新支點。然而 Summit 不能直接把中國單區結果變成歐美核准。海外開發路徑、監管要求與是否啟動膽道癌全球研究，仍是尚未完成的作業。 對阿斯特捷利康而言，這不是 Imfinzi 已經失去標準地位。durvalumab 合併化療擁有成熟三期、長期追蹤與多國核准；ivonescimab 目前只有頂線結果。待完整資料公布後，市場才有條件比較療效幅度、安全性、給藥與價格，判斷膽道癌的一線標準會不會真正改寫。 ──────────────────── ## 🏥 07｜如果完整數據夠強，臨床現場會怎麼變？ 膽道癌患者真正需要的，不是一張漂亮的新聞稿，而是多一段有品質的生存時間。 一線治療通常是整條治療路徑的起點。患者第一次接受系統性治療時，體力、器官功能與承受副作用的能力往往相對完整；若能在這一段把疾病控制得更久，後續能否接受第二線治療、能否回到日常生活，都可能跟著改變。 因此，總生存期優勢只是第一道門。醫師還會問：患者需要多久來一次醫院？合併化療後的副作用能否管理？出血或血栓風險是否集中在特定族群？黃疸、膽道感染或肝功能異常的患者能否安全使用？療效提升會不會伴隨更多停藥？ 支付方關心的問題又不同。durvalumab 與 Keytruda 已有成熟的採購、給付與治療經驗。新療法若要取代它們，除了延長生存，也要提出可被醫療體系承擔的價格、給藥與管理方案。 這正是「有統計差異」到「改變臨床標準」之間最容易被市場忽略的距離。藥物先在試驗裡贏，接著要讓醫師願意換、醫院能夠用、患者負擔得起，最後才會反映在真實世界市占率。 對康方來說，後半場考驗會從研發勇氣轉向商業執行。它需要把完整數據說清楚，把適合的患者找出來，也要證明產能與供應能跟上新增適應症。頭對頭勝利替產品打開門，卻不會自動把市場搬進來。 ──────────────────── ## 📊 08｜下一站，不是掌聲，而是五張成績單 ivonescimab 在膽道癌贏下一個有分量的終點，也替中國創新藥留下難得畫面：新療法不再只和化療比，而是直接挑戰全球已被驗證的免疫治療標準。 接下來，投資人與臨床醫師會共同盯著五張成績單：總生存期風險比、中位生存期、生存曲線、完整安全性，以及三種膽道癌亞型的結果。 若五張成績單彼此一致，ivonescimab 獲得的將超越新增一個適應症的機會。它會讓 PD-1／VEGF 雙抗從肺癌明星，開始接近泛癌種平台的定義；也會迫使下一代免疫腫瘤藥物回答一個更尖銳的問題：在已有 PD-1 或 PD-L1 標準療法的市場裡，新藥究竟能不能正面贏過它？ 如果勝差不大，或安全性代價過高，這場頭對頭的光環也會迅速降溫。 所以此刻最值得記住的，並非「ivonescimab 已經改寫膽道癌」。 更準確的答案是：它第一次跨出肺癌，站上膽道癌金標準的對面，而且先拿到了總生存期這一分。 剩下的分數，等完整數據說話。 資料來源：康方生物 HARMONi-GI1 官方公告、Summit Therapeutics 官方公告與管線資料、TOPAZ-1、KEYNOTE-966、FDA 公開資料。 本文為醫藥產業資訊整理，不構成醫療建議或投資建議。治療選擇請諮詢專業醫療人員。 ## 資料來源 1. [康方生物｜HARMONi-GI1 達到總生存期主要終點](https://akesobio.com/en/media/akeso-news/260826/) 2. [Summit Therapeutics｜HARMONi-GI1 頂線結果與單區研究邊界](https://smmttx.com/news/press-releases/news-details/2026/Ivonescimab-Plus-Chemotherapy-Demonstrates-Statistically-Significant-Overall-Survival-Benefit-Compared-to-Durvalumab-Plus-Chemotherapy-in-First-Line-Advanced-Biliary-Tract-Cancer-in-Single-Region-Trial-Conducted-by-Akeso-in-China/default.aspx) 3. [TOPAZ-1｜Durvalumab 合併化療主要分析](https://pubmed.ncbi.nlm.nih.gov/38319896/) 4. [TOPAZ-1｜更新整體生存分析](https://pubmed.ncbi.nlm.nih.gov/38823398/) 5. [FDA｜Pembrolizumab 合併化療用於膽道癌](https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-pembrolizumab-chemotherapy-biliary-tract-cancer) --- ## Claude Starts Running Experiments: AI Enters the Physical Lab URL: https://drugnews.com.tw/articles/2026-09-03-anthropic-mhs-ai-lab-hardware-standard-en.html Date: 2026-09-03 Language: en Summary: Anthropic's Model Hardware Standard research preview gives AI agents a shared interface for microscopes, liquid handlers, robots and laser-control systems. The proof-of-concept results are important, but they do not mean autonomous laboratories are already mature. At 2 a.m., a laser in a quantum computer loses lock. At 4 a.m., a graduate student must return to the lab to change a plate in a PCR instrument. AI could already read papers, propose hypotheses and write analysis code. Faced with a droplet, a pipette or a real instrument, however, it stopped at the screen. On August 27, 2026, Anthropic moved that boundary forward. The company released a research preview of the Model Hardware Standard (MHS), a shared interface through which AI agents can read and operate microscopes, liquid-handling workstations, robotic arms, plate readers and laser-control systems. An agent can observe instrument state, sequence procedures, adjust parameters and, in some cases, recover from failure. This does not mean Claude has taken over the laboratory, or that scientists are about to disappear. MHS is available only to an initial group of scientific and advanced-manufacturing partners, and most published examples are proofs of concept. Complex experiments, physical judgment and safety accountability remain human responsibilities. The important shift is that AI for Science is no longer only a contest over who can produce more answers. It is becoming a contest over who can send an answer into physical equipment, collect new data and begin the next experimental cycle. ## 1 | Scientific AI has never lacked ideas Generative AI can propose thousands of protein designs, but somebody still has to make and test them. A model may predict which molecule could work, yet it cannot by itself aspirate a reagent, control temperature, read a reaction and adapt after failure. Digital candidates are becoming cheaper; physical validation is becoming the new bottleneck. Zihao Song, a doctoral researcher in the Baker and Pinglay laboratories at the University of Washington, offered a vivid example in Anthropic's case study. Designing a protein on a computer can cost as little as about $0.01, while testing one candidate at the bench can cost about $100 and require a week of labor. Once a project faces thousands of candidates, the budget and time disappear not only into deciding what to try, but into verifying it. Academic workflows change constantly. Expensive, rigid automation may repeat one procedure ten thousand times but remain ill suited to a group that changes experiments dozens of times a year. Instruments also speak incompatible vendor languages. Connecting a liquid handler, robot and microscope can require custom software for each pair; Anthropic says such integration often takes weeks or months. MHS addresses that long-neglected last mile. ## 2 | MHS is a common socket for laboratory equipment MHS is not a new machine and is not limited to one brand of robot. It is a shared language between agents and instruments. A standardized driver converts actions such as reading a temperature, setting a temperature or moving to a position into simple read-and-write commands. A reference file describes the instrument's capabilities, adjustable values, load limits and safety boundaries in a form an AI system can interpret. Once connected, an agent can orchestrate equipment through the Model Context Protocol (MCP), a command line or an application programming interface. Researchers describe a high-level objective in natural language; the agent decomposes the workflow, checks instrument state, waits for one step to finish and asks the next machine to take over. Anthropic says the standard is model-agnostic. Other agent frameworks could use the same protocols. MHS remains early, and Anthropic plans to build safety evaluations and operating practices with partners before moving toward open source. If it works, MHS could become a standard layer connecting models, instruments and laboratories: not a system that chooses the science, but one that lets AI reliably discover, understand and control programmable hardware. ## 3 | Genentech: Claude can tune parameters, but it does not understand a bubble The proof of concept closest to drug development came from Genentech. The team chose the BCA protein assay and coordinated a liquid handler, robotic arm and plate reader. Claude served as the central coordinator through MHS and searched for suitable pipetting speeds for water and viscous bovine serum albumin solution. It ran trials, read results, compared them with expert baselines and adjusted the next round. For that setup, it identified approximately 140 microliters per second for water and 10 microliters per second for the viscous protein solution; Genentech automation specialists considered the values reasonable. The failure was more instructive. When bubbles caused a detection error, Claude treated the event as a parameter problem and repeatedly retried the same well, creating more foam. A researcher had to explain that this was a physical problem: move to a clean well and reduce mixing. Claude retained the guidance in later runs, and the team converted it into a reusable liquid-handling skill. But reading large amounts of chemistry and physics does not give a language model the embodied intuition of a laboratory worker. Genentech therefore described the work as a promising proof of concept, not a fully autonomous laboratory. Scientists still set biological goals, interpret critical results and handle unknown failures. ## 4 | Baker, Pinglay and Janelia: reducing night shifts and integration work In the Baker and Pinglay laboratories, Song connected six instruments in less than a week, including the time needed to write drivers. A remote dashboard let an agent monitor qPCR amplification curves and stop a run at the appropriate point. It also coordinated a robot and liquid handler during plate transfers. Repeated tests produced no collision between the two machines. Song could monitor the procedure from his office rather than return to the bench every hour or two. This remained a demonstration: complex workflows need substantial optimization, and long-running agents add compute cost. The immediate value is reducing instrument watching, waiting and delayed detection of spoiled samples. One early MHS project began at HHMI's Janelia Research Campus, where neuroscience rigs combine lasers, lenses, focusers, sensors and software from multiple vendors. Researchers exposed equipment state through MHS so Claude could align beams, adjust optical systems and check results with sensors. One microscope setup reportedly compressed about half a day of manual configuration into one step. MHS does not remove physical trade-offs in imaging. Scan speed, coverage and signal quality still compete. A standards layer can reduce integration friction; it cannot repeal physics. ## 5 | QuEra's 99.3% result is impressive—and narrow Quantum-computing company QuEra used MHS for laser-lock recovery. Its previous specialist-built script succeeded about 58% of the time and took roughly 150 seconds per attempt. After Claude interacted with instruments and iterated, it produced a standalone script that recovered the correct lock in 695 of 700 blind tests, or 99.3%. The hardest perturbations took about 10–14 seconds; simpler cases took roughly 0.9–5.4 seconds. That result shows what AI may accomplish in a closed problem with a clear objective, readable state and adjustable controls. It does not establish that Claude understands the physical world. It still could not diagnose physical hardware failures, and it sometimes stopped for human approval when an action appeared risky. Safety cuts both ways. An overconfident agent may damage costly equipment, waste scarce samples or endanger people. An overly cautious one can defeat the purpose of unattended experiments. Adoption will depend on clear permissions, device-level limits, human approval points and accountability. ## 6 | The platform contest spans models, workbenches and hardware standards MHS makes more sense alongside Anthropic's other science initiatives. In June 2026, the company introduced the Claude Science beta, combining literature, databases, code environments, charts, compute and auditable records in one workbench. It included more than 60 scientific skills and connectors across genomics, single-cell analysis, proteomics, structural biology and cheminformatics. On August 27, Anthropic also announced 10,000 team seats for scientists for one year. Verified principal investigators at academic or nonprofit institutions can add free standard seats; premium seats with five times the usage cost $15 a month. Researchers can separately apply for up to $50,000 in AI for Science credits per project. Together, the strategy has three layers: subsidized model access for research teams; Claude Science for digital research workflows; and MHS for programmable physical equipment. Experimental data can return to the model, creating a hypothesis–design–execution–measurement–revision loop. Anthropic may be competing less to own each discovery than to own the interface through which scientific work, device data and experimental records pass. ## 7 | Who benefits, and which businesses may be rewritten? Instrument and robotics companies that support the standard could benefit first. Anthropic's early partner or testing list includes Danaher, QIAGEN, Tecan, MBF Bioscience, Automata, Doosan Robotics and Universal Robots; AWS plans support through Strands Robots. Easier discovery and control may put compatible equipment into more workflows and shift service from reactive repair toward live diagnostics. Cloud, data and model-tool providers form a second layer. Long experiments consume inference, storage, audit and access-control resources. Providers that can connect sensitive local data, cloud compute and device control securely may earn recurring revenue. Pressure may fall on firms that rely only on one-off custom connectors. Standard drivers could commoditize low-value translation work. Integrators that understand experimental validation, regulation and safety may instead move up the value chain into system qualification, risk controls and multi-site deployment. Investors should not translate ecosystem participation directly into revenue. MHS is not yet open source and remains a research preview. Partners range from exploration and testing to driver development and planned support. Supporting a standard is not the same as receiving a large order, much less proving market leadership. ## 8 | Taiwan's opportunity: turn legacy instruments into auditable data nodes Taiwan combines precision manufacturing, robotics, semiconductors, servers, medicine and biotechnology. Its most useful opportunity may not be another chatbot, but three practical layers. First, reliable drivers for existing equipment. Many hospitals, universities and factories still use instruments without modern APIs. Sensing, edge control, device digitization and safety isolation could make installed equipment accessible. Second, auditable experimental data chains. Life science depends on batch records, calibration, reagents, permissions and sample provenance. Every instruction needs an approver and version; every result must be reproducible before it can support drug development or regulation. Third, shared human–machine safety systems. Device limits, emergency stops, anomaly detection, approval gates and accountability logs may create more value than selling another robotic arm. The meaningful metrics are how many devices are connected, how much integration time falls, whether downtime and sample loss decline, and whether revenue is recurring rather than one-off. ## 9 | Do not declare scientists replaced MHS has four clear limits. It works only with equipment that has a programmable interface; language models have weak spatial and physical reasoning; safety practices are still under construction; and all published outcomes come from a small number of partners and tightly bounded tasks. Protein assays, plate transfers, microscope setup and laser locking do not validate complex cell culture, unknown biological reactions or clinical development. A more accurate statement is that Claude is beginning to acquire hands for programmable hardware. Scientists still decide where to go, what may be done, whether results are trustworthy and who is accountable when something fails. ## Conclusion | The next AI-for-Science contest is about closing the loop AI for Science used to resemble an exam: the system that read more papers, predicted more accurately and generated more candidates appeared to lead. MHS moves the contest into the physical world. When an agent can operate equipment, it can help obtain the next dataset for training and validation. A hypothesis becomes a physical action, and an experimental result becomes the starting point for another decision. Anthropic has not proved that Claude is a scientist. It has placed the model, a scientific workbench and a hardware standard on the same platform map. Today, scientists still cannot hand over the keys. MHS is a research preview, proofs of concept must scale, and physical failure and safety accountability remain unresolved. A mature autonomous laboratory will not be AI working alone behind a closed door. It will combine model speed, instrument precision and expert judgment while leaving an auditable trail among all three. AI for Science has changed because Claude has begun to touch the real world—and the real world will use every bubble, shutdown and irreproducible result to teach AI what science actually requires. --- ## Claude 開始動手做實驗！AI 搶進真實實驗室，遊戲規則變了 URL: https://drugnews.com.tw/articles/2026-09-03-anthropic-mhs-ai-lab-hardware-standard.html Date: 2026-09-03 Language: zh-Hant Summary: Anthropic 公開模型硬體標準 MHS 研究預覽，讓 AI 代理透過共通介面讀取與操作顯微鏡、移液工作站、機械手臂與雷射控制系統；這是概念驗證，不是無人實驗室已成熟。 ──────────────────── 半夜兩點，量子電腦裡的雷射突然失去鎖定；凌晨四點，博士生又得回實驗室替聚合酶連鎖反應儀換板。過去，人工智慧可以讀完論文、提出假說、寫出分析程式，碰到真實世界裡的一滴液體、一支移液器，卻只能停在螢幕前。 2026 年 8 月 27 日，Anthropic 把這條界線往前推了一步。 公司公開「模型硬體標準」（Model Hardware Standard，MHS）的研究預覽版，讓人工智慧代理透過共通介面，讀取並操作顯微鏡、移液工作站、機械手臂、酵素標定儀，甚至量子電腦裡的雷射控制系統。它可以觀察設備狀態、安排實驗順序、依結果調整參數，某些情況下還能從故障中自行恢復。 這不是「Claude 已經接管實驗室」，也不是科學家準備失業。MHS 目前只開放給首批研究機構與先進製造夥伴，案例多是概念驗證；複雜實驗、物理直覺與安全責任仍牢牢握在人手上。 但一個更重要的變化已經出現：AI for Science（人工智慧科學研究）不再只比誰能提出更多答案，開始比誰能把答案送進真實設備、拿回新數據，再展開下一輪實驗。 ──────────────────── 🧪 一、科學 AI 最缺的，從來不是點子 生成式人工智慧進入科學界後，最常遇到的質疑很合理：模型可以一口氣提出一萬個蛋白質設計，但誰來把它們做出來？它可以預測哪個分子可能有效，卻不能自己吸取試劑、控制溫度、讀取反應，再根據失敗調整下一輪。 數位世界的候選愈來愈便宜，實體驗證反而成了新的塞車點。 華盛頓大學 Baker 與 Pinglay 實驗室的博士生 Zihao Song 提供了一個很直觀的例子。依 Anthropic 官方案例，現在用電腦設計一個新蛋白質，成本最低可到約 0.01 美元；把它送上實驗台測試，卻可能花約 100 美元，還要投入一週人力。當研究一次面對上千個候選，真正吞掉預算與時間的，不再只是「想出什麼」，而是「如何把它驗證完」。 學術實驗室和固定產線不同，流程經常改變。昂貴而僵硬的自動化設備，可以把同一件事重複一萬次，卻未必適合一年更換幾十種實驗的團隊。 設備也像一座語言混亂的城市：A 廠牌的移液器、B 廠牌的機械手臂、C 廠牌的顯微鏡，各有自己的程式介面。要讓它們合作，工程師得替每一對設備打造專用連接程式。官方說明指出，這類整合常耗費數週到數月。 MHS 要處理的，正是這個被忽略很久的「最後一公尺」。 ──────────────────── 🔌 二、MHS 是什麼？可以把它想成實驗設備的共通插座 模型硬體標準不是一台新機器，也不是只讓 Claude 控制某個品牌的機械手臂。它更像一層共通語言，放在 AI 代理與各式設備之間。 每台設備都有一個標準化驅動程式，把「讀取溫度」、「設定溫度」、「移動到某個位置」等操作，轉成簡單的讀取與寫入指令。設備能做什麼、哪些數值可以調整、最大負重與安全界線，也會整理成 AI 能讀懂的參考檔。 設備接上之後，代理可以透過模型情境協定（Model Context Protocol，MCP）、命令列或應用程式介面（Application Programming Interface，API）調度。研究者用自然語言描述高階目標，AI 負責拆解流程、確認設備狀態、等待前一步完成，再叫下一台機器接手。 更關鍵的是，MHS 並未把自己綁死在 Claude。Anthropic 表示，這套標準與模型無關，其他代理框架也能透過標準協定存取。它目前仍是早期版本，Anthropic 希望先和夥伴建立安全評估與操作慣例，再走向開源。 如果這條路成功，MHS 想成為不同模型、儀器與實驗室之間的標準層。它像實驗設備的 USB：不替研究者決定要做什麼，先讓 AI 能可靠地找到、理解與控制硬體。 ──────────────────── 🧫 三、基因泰克的蛋白質實驗：AI 會調參，卻不懂一顆泡泡 最接近藥物研發的概念驗證，來自基因泰克（Genentech）。 研究團隊選擇二辛可寧酸蛋白質測定法，也就是常見的 BCA 蛋白質濃度檢測。這套流程要協調移液工作站、機械手臂與酵素標定儀：液體先被精準吸取，微孔盤由機械手臂搬運，最後再讀取吸光值。 團隊讓 Claude 透過 MHS 擔任三台設備的中央協調者，並請它自行尋找適合清水與黏稠牛血清白蛋白溶液的移液速度。Claude 執行試驗、讀取結果、和專家操作的基準比較，再調整下一輪參數。最後，它為水找到每秒約 140 微升、為黏稠蛋白質溶液找到每秒約 10 微升的流速；基因泰克自動化專家認為，這些參數對該套設備而言合理。 這段成果很吸睛，失敗的部分其實更重要。 蛋白質溶液黏稠又容易起泡。Claude 第一次遇到氣泡造成的偵測錯誤時，把它當成程式參數問題，在同一個孔位反覆重試，結果攪出更多泡沫。研究人員必須明確告訴它：這是物理問題，要換到乾淨孔位，並減少混合次數。 模型收到指導後，確實能在後續流程保留這段經驗，團隊也把它整理成可重複使用的液體操作技能。可是，這個插曲戳破了一個容易被宣傳遮住的真相：語言模型讀過大量物理與化學文字，不等於它具有實驗員面對液體時的直覺。 因此，基因泰克把成果定位為「很有希望的概念驗證」，並沒有宣布實驗室已經全自動。科學家仍設定生物學目的、判讀關鍵結果、處理未知錯誤，AI 先接走的是大量重複、機械化、可描述的工作。 ──────────────────── 🤖 四、Baker 實驗室與 Janelia：被解放的是深夜值班與設備整合 在 Baker 與 Pinglay 實驗室，Zihao Song 用 MHS 接上六台儀器，花費不到一週，時間還包含為設備撰寫驅動程式。他建了一個遠端儀表板，讓代理監看定量聚合酶連鎖反應（quantitative Polymerase Chain Reaction，qPCR）的擴增曲線，在適當時機停止流程；也讓機械手臂與移液工作站互相等待、完成微孔盤交接。 重複測試中，兩台設備沒有碰撞。他可以在辦公室監看流程，不必每隔一、兩小時回實驗台換板。這仍只是示範；複雜流程需要更多最佳化，長時間監控也會增加運算成本。它先處理的，是研究者巡看儀器、等待流程與太晚發現樣本報廢的真實痛點。 MHS 最早的起點之一，是霍華德・休斯醫學研究所 Janelia 研究園區。那裡的神經科學儀器把雷射、鏡頭、對焦器、感測器與多家廠商軟體綁在一起。研究人員用 MHS 暴露設備狀態，讓 Claude 能對準光束、調整光學系統，再由感測器檢查結果；其中一套顯微鏡，把原本約半天的手動設定壓成單一步驟。 官方並未說 MHS 消滅顯微成像的物理限制。掃描速度、涵蓋範圍與訊號維持仍須取捨；標準層能降低整合摩擦，不能取消物理定律。 ──────────────────── ⚛️ 五、QuEra 的 99.3%，為何精彩又不能亂外推？ 量子電腦公司 QuEra 把 MHS 用在雷射鎖定恢復。量子系統需要把雷射頻率維持在極高精度；溫度、震動或氣壓變化，都可能讓雷射偏離目標。傳統人工恢復約需 5 到 10 分鐘，一套由多種專家花數月打造的專用腳本，成功率約 58%，每次嘗試約 150 秒。 Claude 透過 MHS 讀取儀器、調整控制器，經過反覆實驗後寫出獨立腳本。官方案例顯示，後續盲測 700 次中成功 695 次，恢復率達 99.3%；最困難的擾動約 10 到 14 秒，簡單情況約 0.9 到 5.4 秒。 這證明在目標清楚、狀態可讀、控制項可調的封閉問題裡，AI 有機會找到比固定流程更好的策略；不能外推成「Claude 已懂物理世界」。真實硬體故障時，它仍不知怎麼排查；遇到風險操作，也可能停下等待人工批准。 安全在這裡是一個雙面問題。代理太大膽，可能損壞昂貴設備、浪費珍貴樣本，甚至帶來人身風險；代理太保守，又會讓無人值守實驗失去意義。誰能定義清楚的權限、設備級限制、人工核准點與事故責任，誰才有機會把漂亮示範變成可採用的系統。 ──────────────────── 🧩 六、真正的平台戰：模型、工作台、硬體標準一起來 若只看 MHS，會低估 Anthropic 的企圖。 2026 年 6 月，公司先推出測試版 Claude Science，把文獻、資料庫、程式環境、圖表、計算資源與可稽核紀錄放進同一個科學工作台。它預先整合超過 60 種科學技能與連接器，涵蓋基因體、單細胞、蛋白質體、結構生物與化學資訊；產出的圖表同時保留程式碼、環境與對話歷程，方便研究者追溯與重現。 8 月 27 日，Anthropic 又宣布為全球科學家開放 10,000 個團隊方案名額，期限一年。通過驗證的學術或非營利研究機構主持人，可以替團隊加入免費標準席位；五倍用量的進階席位每月 15 美元。若專案需要更大用量，研究者還可申請每案最高 5 萬美元的 AI for Science 點數。 三塊拼圖放在一起，平台輪廓就清楚了： ① 第一層，用低價或免費方案把模型送進研究團隊。 ② 第二層，用 Claude Science 承接讀文獻、跑分析、做圖與管理運算的數位工作流。 ③ 第三層，用 MHS 連接可程式化設備，讓數位推理跨進物理執行。 實驗數據再回到模型，形成「假說—設計—執行—測量—修正」的閉環。Anthropic 爭的未必是每項發現，而是科學工作、設備資料與實驗紀錄共同經過的介面。 ──────────────────── 💰 七、誰受益，誰的生意可能被重寫？ ① 第一批明顯受益者，是願意接入標準的儀器與機器人公司。 Anthropic 公布的早期合作或測試名單包括丹納赫、凱杰、Tecan、MBF Bioscience、Automata、斗山機器人與 Universal Robots；Amazon Web Services（AWS）則計畫透過 Strands Robots 支援 MHS。對硬體商而言，設備愈容易被代理發現與調用，愈可能進入更多自動化工作流程，售後服務也有機會從被動維修走向即時診斷。 ② 第二層是雲端、資料平台與模型工具。實驗若長時間運行，推理、儲存、稽核與權限管理都會消耗資源。誰能在本地敏感資料、雲端算力與設備控制間建立可信連線，誰就可能收取長期費用。 ③ 壓力最大的一群，可能是只靠客製連接程式與一次性設備整合收費的服務商。當共通驅動程式降低「翻譯」門檻，部分低價值整合工作會被標準化。不過，真正懂實驗流程、驗證、法規與安全的整合商不一定消失；他們可能升級成負責系統驗證、風險分級與跨廠區部署的高價值角色。 投資人也要避免把「加入生態」直接翻譯成營收。MHS 尚未開源，仍在研究預覽；夥伴的狀態從探索、測試、開發驅動程式到計畫支援不等。支援標準不代表已產生大額訂單，更不代表 Anthropic 一定能把標準變成市場主導權。 ──────────────────── 🇹🇼 八、台灣的機會，藏在「把舊設備變成可驗證的資料節點」 台灣同時擁有精密製造、機器人、半導體、伺服器、醫療與生技實驗能量，理論上適合切入自主實驗室。但最有價值的機會，未必是再做一個聊天機器人，而是處理更難也更務實的三件事。 ① 第一，替既有設備建立可靠驅動程式。許多醫院、學校與工廠仍在使用沒有現代應用程式介面的老儀器，MHS 目前無法直接相容。能提供感測、邊緣控制、設備數位化與安全隔離的廠商，可能成為存量改造的重要一環。 ② 第二，建立可稽核的實驗資料鏈。生命科學看重批次、校正、試劑、權限與樣本來源。設備能被 AI 操作只是起點；每一次指令由誰批准、用哪個版本執行、結果能否重現，才決定資料能不能進入藥物研發與監管流程。 ③ 第三，設計人機共同負責的安全層。台灣擅長硬體與製造管理，如果能把設備級上限、緊急停止、異常偵測、人工核准與責任紀錄做成模組，價值可能高於單純販售一支機械手臂。 🔎 真正該追的不是誰在簡報寫了「AI 實驗室」，而是接上多少設備、整合縮短多久、能否降低停機與樣本報廢，收入來自一次性專案還是可重複授權。 ──────────────────── 🛡️ 九、別急著說「科學家被取代」 MHS 目前有四道清楚的限制。 ① 它只支援具有可程式化介面的設備；大量老舊儀器仍需廠商或工程團隊改造。 ② 語言模型的空間與物理推理有限，泡沫、碰撞、液體黏度與機械磨損，不會因為讀過說明書就自動變成身體直覺。 ③ 安全規則仍在建立。Anthropic 正利用研究預覽發展額外評估、最佳實務與實體安全路線圖，這本身就說明產品尚未成熟到可以無監督普及。 ④ 最後，所有公開成果都來自少數合作夥伴與特定任務。蛋白質濃度測定、微孔盤交接、顯微鏡設定與雷射鎖定，都是邊界清楚、可以量化的工作；它們不能替複雜細胞培養、未知生物反應或臨床開發背書。 所以，更精確的說法是：Claude 開始取得「可程式化硬體的手腳」，科學家仍然負責決定往哪裡走、什麼可以做、結果是否可信，以及失敗時誰來承擔。 ──────────────────── 🎯 結語：AI4S 下一戰，不在誰最會回答，而在誰能完成閉環 過去，人工智慧科學研究的競爭像一場考試：誰能讀更多論文、預測更準、生成更多候選，誰就領先。 MHS 把競賽帶進另一個場地。 當代理能控制設備，模型就有機會自己取得下一批訓練與驗證資料。假說不再停在文字裡，而會變成一次真實操作；實驗結果也不再只是報告終點，而是下一輪決策的起點。 這才是 Anthropic 此次發布最值得注意的地方。它沒有證明 Claude 已經是一位科學家，卻第一次把模型、科學工作台與硬體標準放進同一張平台藍圖。 誰能讓這個閉環安全、可重現、跨品牌運作，誰就可能掌握未來實驗室的入口。 只是今天，科學家仍不能把鑰匙交出去。 MHS 還是研究預覽，概念驗證仍要擴大，物理故障與安全責任仍未解決。真正成熟的自主實驗室，不會是 AI 一個人關起門來做實驗，而是讓模型負責速度、設備負責精準、專家負責判斷，三者互相留下可以追溯的證據。 AI for Science 的遊戲規則確實變了。 不是因為 Claude 已經接管實驗室，而是它終於開始碰到真實世界；而真實世界，會用每一顆氣泡、每一次停機與每一份無法重現的數據，逼 AI 學會什麼叫科學。 — 資料來源：[Anthropic｜模型硬體標準研究預覽](https://www.anthropic.com/news/model-hardware-standard-research-preview)、[Anthropic｜擴大科學家支援方案](https://www.anthropic.com/news/expanding-support-for-scientists)、[Anthropic｜Claude Science 科學工作台](https://www.anthropic.com/news/claude-science-ai-workbench)。 本文為科技與生技產業資訊整理，不構成醫療建議或投資建議。MHS 與 Claude Science 目前分別處於研究預覽與測試階段；文中案例、效能與成本數據均來自 Anthropic 及其合作夥伴公開資料，仍需更多獨立、跨設備與長期驗證。 --- ## A New Drug Launched. Seven Months Later, the Company Sought Liquidation URL: https://drugnews.com.tw/articles/2026-09-02-post-approval-biotech-commercialization-death-valley-en.html Date: 2026-09-02 Language: en Summary: FDA approval does not erase business risk. Iterum, Karyopharm and Theravance show how revenue, cost, pipeline and capital clocks can begin running at the same time after a drug reaches the market. Iterum's ORLYNVAH generated just $0.4 million in net product revenue in its first quarter on the market, while selling, general and administrative expense was about $6.5 million. From commercial launch in August 2025 to its application for liquidation in March 2026 took roughly seven months. This was not a clinical failure. It was a case in which post-approval commercialization cash flow did not have time to grow. One case does not represent every biotech company, but it is a reminder that launch replaces development risk with revenue, cost and financing risk. This article is not investment advice. ## 01 | Iterum: $0.4 million in revenue collided with $6.5 million in expense Iterum's story resembles a post-launch free fall. ORLYNVAH is an oral penem antibiotic. In October 2024, the FDA approved it for uncomplicated urinary tract infections in certain adult women: infections caused by specified bacteria where oral antibacterial options are limited or unavailable. Every qualifier in that label narrows the market. Antibiotics are not ordinary chronic-disease medicines. Physicians must consider antimicrobial resistance, culture results, stewardship and reimbursement rules. A new antibiotic can have clinical value without prescriptions flowing as naturally as they might for a blood-pressure medicine. Health systems often want to reserve a new antibiotic for patients who truly need it. That is good public-health practice, but it is brutal for a small company that depends on a single product. Iterum had hoped ORLYNVAH would be its final chance to turn the company around. It spent nearly a year preparing for the market, building a commercial team and working with outside partners. The product launched commercially in August 2025. The first-quarter result arrived quickly: net product revenue was only $0.4 million, a figure that also included initial channel stocking. In the same quarter, selling, general and administrative expense reached $6.5 million, with the increase driven primarily by commercialization activity. This was not simply a case of “poor sales.” It meant that every step the company took toward the market was burning cash much faster than current revenue was arriving. Physician education, distribution, payer access, patient support and a sales team do not stop charging simply because prescriptions have not yet scaled. Before launch, R&D spending can look like the highest mountain. After launch, what can crush a company may be the market bill that must be paid every day. At the end of September 2025, Iterum had only $11 million in cash remaining. The company discussed ORLYNVAH commercial rights with two potential counterparties, but no transaction was completed. Its share price and listing status continued to deteriorate. On March 27, 2026, the company applied to the High Court of Ireland for liquidation. On April 13, the court formally ordered it to be wound up. Roughly seven months passed between ORLYNVAH's commercial launch and the liquidation application; the court issued its formal order the following month. A drug had completed clinical development, passed review and actually reached pharmacy shelves, yet it did not have time to turn prescriptions into cash flow. The lesson from Iterum is not that antibiotics lack value. It is that clinical need, patients who can be prescribed a drug, the reimbursement route and commercial cost are never the same thing. A market can look large, but if the patients who can actually be found, diagnosed, reimbursed and prescribed treatment are too few, the market size on paper is a mirage. ## 02 | Karyopharm: a first-in-class target could not rescue the balance sheet Karyopharm has not collapsed. That must be stated first, because it is not currently a replay of Iterum. It is an ongoing stress test. The company's core product, XPOVIO (selinexor), is the world's first approved exportin 1 (XPO1) inhibitor. It entered the multiple-myeloma market in 2019, and “first in class” was once Karyopharm's brightest label. But arriving first does not mean being the first to build a large market. XPOVIO entered a hematologic-oncology field with intense competition and rapidly shifting treatment options. A new mechanism faces more than a regulatory certificate: it must contend with therapies physicians already know, the condition of later-line patients, adverse-event management and pressure from other new therapies moving earlier in treatment. In the second quarter of 2026, XPOVIO generated $30.8 million in net product revenue, up about 3.7% from $29.7 million a year earlier. That is not no business, but neither does it show the kind of explosive growth that can plainly carry the cost of an entire company. In the same quarter, Karyopharm recorded $29.0 million in R&D expense and $25.9 million in selling, general and administrative expense; its operating loss was $22.5 million and its net loss was $67.0 million. The latter includes interest and substantial non-cash fair-value changes in financial instruments, so it cannot simply be treated as cash flowing out in full. It nevertheless reflects a capital structure already under strain. The more dangerous numbers are on the balance sheet. At quarter end, cash, cash equivalents, restricted cash and investments totaled $65.4 million. The company expected existing liquidity to last only into September 2026. On September 10, $15.8 million of loan principal was due. Karyopharm explicitly stated that without additional financing or a lender waiver, payment would be expected to take it below a $10 million minimum-liquidity covenant and trigger default. For a biotech company, the science clock may be measured in “next year's data.” The bank clock can be precise to a single date. Karyopharm's clinical hope is also splitting into different paths. In July 2026, the Phase 3 XPORT-EC-042 trial of XPOVIO in endometrial cancer missed its primary endpoint. Another Phase 3 study, SENTRY in myelofibrosis, delivered one positive and one negative result across its co-primary endpoints. At week 24, XPOVIO plus ruxolitinib produced at least a 35% reduction in spleen volume in 49.8% of patients, versus 28.0% in the control group, a highly significant difference. But the co-primary symptom-score endpoint did not meet its target, with a p-value of 0.825. In subsequent written feedback, the FDA said spleen-volume reduction appeared capable of serving as a surrogate endpoint reasonably likely to predict overall survival. Karyopharm therefore plans to submit a supplemental new drug application for myelofibrosis under an accelerated-approval pathway, with long-term overall-survival data to confirm clinical benefit. That is a real regulatory pathway, not a check that has already cleared. Karyopharm now faces a sharp contradiction. It does not lack a product, revenue or clinical opportunities. What it lacks is time to wait for those opportunities to become approvals and sales. A first-in-class halo can attract attention, but it cannot substitute for cash or ask creditors to postpone the calendar. ## 03 | Theravance: not unable to continue, but unwilling to wager its cash on the next bet Theravance is the easiest of the three companies to describe incorrectly. It does not face an immediate cash shortage. At the end of the second quarter of 2026, it had $387.7 million in cash and marketable securities and no debt. Its COPD medicine YUPELRI continued to grow, and the company retained revenue and milestone opportunities linked to partnered products. What disappeared was the R&D story. Theravance had placed ampreloxetine in neurogenic orthostatic hypotension. After an earlier Phase 3 study failed, a subgroup signal emerged in patients with multiple system atrophy, and the company designed the Phase 3 CYPRESS study for that population. In March 2026, CYPRESS again missed its primary endpoint. After that failure, Theravance did not use nearly $400 million of cash to search for another high-risk asset. It decided to discontinue ampreloxetine, phase out its R&D function and sharply reduce general and administrative costs. After restructuring, operating expenses were expected to be about 60% lower than in 2025. Theravance then put itself on the transaction table. On June 29, 2026, it entered into a definitive acquisition agreement with Zymeworks: $17 in cash per share plus a contingent value right that would allow former shareholders to retain some future proceeds if ampreloxetine is licensed, sold or commercialized. As of August 10, the transaction was still expected to close in the second half of 2026, subject to shareholder approval and other customary conditions. The correct description is therefore “signed and awaiting closing,” not “acquired.” Theravance was not pushed into liquidation by cash, as Iterum was, and it was not guarding an imminent loan date like Karyopharm. It made a different kind of harsh choice: when commercial assets still had value and cash remained on the balance sheet, but there was no R&D direction worth continuing to fund, it stopped acting as an independent R&D biotech company. That is not necessarily failure. For employees and scientists who had invested years in the work, it is certainly the end of an era. For shareholders, however, it may be a rational exit that avoids burning remaining value in another high-risk trial. The same headline—“a biotech company disappears”—can conceal completely different capital outcomes: liquidation, default pressure and acquisition. ## 04 | Real commercialization capability is not simply selling a drug The biotech industry easily treats “independent commercialization” as a coming-of-age ritual. Hiring a company's own sales force, negotiating with payers and bringing a medicine to patients can indeed retain more of a product's economics than licensing it to a large pharmaceutical company. It can also build organizational capability. But retaining all upside also means taking on all fixed cost, inventory, access and scale-up risk. Real commercialization capability is not proving that a company can issue an invoice. It is making each unit of commercial investment generate enough cash to fund the next round of innovation. If a company must keep raising capital after launch simply to maintain its sales force; if each additional unit of revenue first requires even more fixed cost; if the growth of the only product cannot keep up with the R&D burn of the next pipeline, then the company has merely changed from “using investors to fund R&D” to “using investors to fund R&D and sales.” That is why the post-approval death valley is often harder for the market to see early than clinical failure. Clinical results are binary events: success or failure is usually revealed in a day. Commercialization slippage looks more like chronic blood loss. Prescriptions edge up each quarter, payer coverage gradually improves, management keeps saying the second half will be better, and the company always appears to have hope. Cash does not wait. ## 05 | Before judging a commercial biotech, look at four clocks ### 1. The revenue clock Do not look only at how many patients a market contains. Ask how many will be diagnosed, fit the label, obtain coverage and ultimately start treatment. First-quarter channel stocking is not the same as patient demand, and shipments into distribution are not the same as continuing prescriptions. ### 2. The cost clock What share of revenue is absorbed by sales expense? How much physician education and patient support are needed for every additional patient? Is the company building economies of scale that improve with revenue, or must it add another layer of cost every time it sells a box of medicine? ### 3. The pipeline clock A single product can support a company, but only if it is truly large enough and durable enough. If the first drug produces only limited cash, the next indication and next pipeline cannot be delayed repeatedly. Karyopharm's problem is precisely that revenue exists, while the successor has not had time to fully deliver. ### 4. The capital clock How long can cash last? When is the next debt due? Are there minimum-liquidity covenants? If the company raises capital at a low share price, how severe will dilution be? Drug R&D is about scientific probability; the capital market ultimately asks a different question: can you survive until the day the answer is announced? The four clocks do not run separately. Slow uptake makes fixed costs look heavier. Heavy costs compress pipeline investment. Pipeline failure then reduces financing capacity. Eventually, what first appears to be a local commercial problem spreads along the balance sheet into a corporate survival problem. ## 06 | The real warning for biotech companies in Taiwan and China Chinese-speaking biotech markets have placed increasing emphasis on “independent commercialization” in recent years. The direction itself is not wrong. If every asset is licensed out too early, a company will never build its own product and market capability. But independent commercialization should not become an exercise in saving face. Whether a company should sell its own drug ought to depend on whether patients are concentrated, physicians can be covered efficiently, the reimbursement path is clear, product differentiation can support the price and cash can survive the slowest plausible uptake scenario. If the answers are not attractive, regional licensing, co-promotion, a phased team build-out or even a company sale at a good price may be more rational than insisting on selling alone. Theravance reminds the market that the best capital allocation can sometimes be to stop betting while value is still visible. Iterum reminds it that misjudging the speed of commercialization can turn an approved drug from a lifeline into an accelerant. Karyopharm tells us that the hardest state is not having no hope at all; it is having hope that needs a year when cash may have only a month left. ## Conclusion | The FDA grants a ticket, not a life-support machine The launch of a new drug is absolutely worth celebrating. It means science, clinical development, manufacturing and regulation have crossed countless gates, and patients finally have one more option. But for a company, FDA approval never means that risk has been reset to zero. It moves the questions from the laboratory to the market. Can the product be found, prescribed and reimbursed? Can revenue catch up with commercial cost? Can the next pipeline take over in time? Can the company secure its next breath before capital markets close? Those questions determine whether a biotech company can truly become a biopharma company: moving from a drug-development organization dependent on external financing to a pharmaceutical enterprise with continuing product cash flow. We used to think of Phase 3 as biotech's death valley. More and more cases now suggest that the truly lethal cliff may lie after approval: the lights are on, the store is open and the first prescription has been written, but money disappears from the balance sheet faster than patients arrive. A launched drug proves only that it is eligible to be sold. The company's survival proves that it is a business. --- ## 新藥上市才 7 個月，公司就聲請清盤：真正的死亡谷在核准之後 URL: https://drugnews.com.tw/articles/2026-09-02-post-approval-biotech-commercialization-death-valley.html Date: 2026-09-02 Language: zh-Hant Summary: ORLYNVAH 上市第一季只有 40 萬美元淨產品收入，同季銷售、一般與管理費約 650 萬美元；約七個月後，Iterum 聲請清盤。Karyopharm 與 Theravance 的情況不同，卻共同顯示新藥核准後，收入、成本、管線與資本四座時鐘會同時啟動。 Iterum 的 ORLYNVAH 上市第一季僅有 40 萬美元淨產品收入，銷售、一般與管理費卻約達 650 萬美元；從 2025 年 8 月上市到 2026 年 3 月聲請清盤，約七個月。這不是新藥臨床失敗，而是核准後的商業化現金流來不及長大。單一案例不代表所有 Biotech，卻提醒投資人：上市只是把研發風險換成收入、成本與融資風險；本文不構成投資建議。 ## 01｜Iterum：40 萬美元收入，撞上 650 萬美元費用 Iterum 的故事最像一場上市後的自由落體。 ORLYNVAH 是一款口服青黴烯類抗生素，2024 年 10 月獲 FDA 核准，用於治療特定成年女性的單純性尿路感染；適用對象必須是由指定細菌引起，而且可用的口服抗菌選擇有限或沒有替代方案。 這段標籤裡，每一個限定詞都會縮小市場。 抗生素不是一般慢性病藥。醫師要考慮抗藥性、培養結果、用藥管理與支付規則；一款新抗生素即使具有臨床價值，也不代表處方會像降血壓藥一樣自然流進來。醫療體系往往希望把新抗生素留到真正需要時才用，這對公共衛生是好事，對一家只靠單一產品活命的小公司卻很殘酷。 Iterum 原本期待 ORLYNVAH 成為公司翻身的最後籌碼。它花了將近一年準備市場，搭建商業團隊，也與外部公司合作。2025 年 8 月，產品正式上市。 第一季成績很快揭曉：淨產品收入只有 40 萬美元，而且這個數字還包含通路最初備貨；同一季的銷售、一般與管理費用卻高達 650 萬美元，增加主要來自商業化活動。 這不是單純的「賣得不好」。 它代表公司每往市場前進一步，都在用遠高於當期收入的速度燒錢。醫師教育、通路、保險准入、病患支持與銷售團隊不會因處方還沒長起來就停止收費。新藥上市前，研發支出看起來像最大的山；上市後，真正壓垮公司的，可能是一張每天都要支付的市場帳單。 截至 2025 年 9 月底，Iterum 現金只剩 1,100 萬美元。公司曾與兩個潛在交易對手討論 ORLYNVAH 的商業權益，最後沒有簽成交易；股價與上市資格也一路惡化。 2026 年 3 月 27 日，公司向愛爾蘭高等法院聲請清盤；4 月 13 日，法院正式下令清盤。ORLYNVAH 從商業上市到聲請清盤約七個月，翌月法院正式下令。 一款藥完成臨床、通過審查、真的擺上貨架，卻來不及把處方變成現金流。 Iterum 的教訓不是「抗生素沒有價值」，而是臨床需求、可開立患者、支付路徑與商業成本，從來不是同一件事。市場再大，如果真正能被找到、被診斷、被核銷、被處方的患者太少，紙上的市場規模只是海市蜃樓。 ## 02｜Karyopharm：第一名的靶點，救不了最後一名的現金表 Karyopharm 還沒有倒。 這句話必須先寫在前面，因為它目前不是 Iterum 的翻版，而是一場仍在進行的壓力測試。 公司的核心產品 XPOVIO，通用名為 selinexor，是全球第一款獲准的輸出蛋白 1（XPO1）抑制劑。它在 2019 年先進入多發性骨髓瘤市場，「全球首創」曾經是 Karyopharm 最明亮的標籤。 可第一個抵達，不等於第一個建立大市場。 XPOVIO 進入的是高度競爭、治療選擇快速變動的血液腫瘤領域。新機轉要面對的，不只是一張核准證書，還有醫師已熟悉的療法、後線患者的體力、藥物副作用管理，以及其他新療法不斷前移的壓力。 2026 年第二季，XPOVIO 淨產品收入為 3,080 萬美元，較去年同期的 2,970 萬美元成長約 3.7%。這不是沒有生意，卻也看不出足以扛起整家公司成本的爆發力。 同一季，Karyopharm 的研發費用為 2,900 萬美元，銷售、一般與管理費用為 2,590 萬美元；營業虧損 2,250 萬美元，淨損 6,700 萬美元。後者包含利息與不少非現金的金融工具公允價值變動，不能簡單當成現金全部流出，但它仍反映資本結構已經很緊。 真正危險的數字，在資產負債表上。 季底現金、約當現金、受限現金與投資合計 6,540 萬美元。公司預計現有流動性只能支應到 2026 年 9 月；9 月 10 日，一筆 1,580 萬美元的貸款本金到期。公司明確表示，如果沒有新增融資或貸款人豁免，付款後預計會跌破 1,000 萬美元的最低流動性條款，構成違約。 對生技公司來說，科學時鐘可以用「明年讀數」計算，銀行時鐘卻會精確到某一天。 偏偏 Karyopharm 的臨床希望也在分岔。2026 年 7 月，XPOVIO 用於子宮內膜癌的三期 XPORT-EC-042 未達主要終點；另一項骨髓纖維化三期 SENTRY，則在共同主要終點上交出一好一壞的成績。 第 24 週時，XPOVIO 聯合 ruxolitinib（蘆可替尼）讓 49.8% 患者達到脾臟體積縮小至少 35%，對照組為 28.0%，差異高度顯著；可是症狀分數的共同主要終點沒有達標，p 值為 0.825。 FDA 後續以書面意見表示，脾臟體積縮小看來可以作為「合理可能預測總生存期」的替代終點。Karyopharm 因此計畫走加速核准路徑，提交骨髓纖維化的補充新藥申請，並用長期總生存期資料確認臨床效益。 這是一條真實的監管路徑，不是一張已經兌現的支票。 現在 Karyopharm 面對的矛盾非常尖銳：它不是沒有產品、沒有收入或沒有臨床機會；它缺的是等待這些機會變成核准與銷售的時間。全球首創的光環可以吸引注意，卻不能代替現金，也不能要求債權人延後日曆。 ## 03｜Theravance：不是撐不下去，而是不想再拿現金賭下一次 Theravance 是三個案例裡最容易被寫錯的一家。 它短期不缺錢。2026 年第二季末，公司有 3.877 億美元現金與有價證券，而且沒有債務；慢性阻塞性肺病藥物 YUPELRI 仍持續成長，公司也享有與合作產品相關的收入與里程碑機會。 真正消失的，是研發故事。 Theravance 曾把 ampreloxetine 押在神經性直立性低血壓上。早期三期研究失敗後，多系統萎縮患者的次族群出現訊號，公司再為這群患者設計 CYPRESS 三期。2026 年 3 月，CYPRESS 仍未達主要終點。 這次失敗後，公司沒有再用近 4 億美元現金去尋找下一顆高風險資產，而是決定終止 ampreloxetine、逐步收掉研發功能，大幅縮減一般管理支出。重整後，營運費用預計較 2025 年減少約 60%。 接著，Theravance 把自己放上交易桌。 2026 年 6 月 29 日，公司與 Zymeworks 簽署確定收購協議：每股 17 美元現金，另加一項或有價值權，讓原股東保留 ampreloxetine 未來若被授權、出售或商業化時的部分收益。 截至 8 月 10 日，這筆交易仍預計在 2026 年下半年完成，尚待股東批准與其他慣常條件。因此，正確說法是「已簽約、待交割」，不是已經完成收購。 Theravance 沒有像 Iterum 一樣被現金逼進清算，也沒有像 Karyopharm 一樣守著迫近的貸款日。它做的是另一種殘酷選擇：當商業資產仍有價值、現金仍在帳上，卻看不到值得繼續下注的研發方向時，停止扮演一家獨立的研發型生技公司。 這不一定是失敗。 對員工與曾經投入多年的科學家而言，它當然是一個時代的結束；對股東而言，卻可能是避免把剩餘價值再度燒進高風險試驗的理性退出。 同樣是「一家 Biotech 消失」，清算、違約與被收購，背後代表的資本結局完全不同。 ## 04｜真正的商業化能力，不是把藥賣出去 生技產業很容易把「自主商業化」當成成人禮。 公司自己招業務、自己談支付、自己把藥送到患者手中，確實比授權給大藥廠更能保留產品利潤，也能建立組織能力。可是保留全部上檔，同時也等於承擔全部固定成本、庫存、准入與放量風險。 真正的商業化能力，不是證明公司有辦法開一張發票，而是讓每一塊商業投入，能夠帶回足以支付下一輪創新的現金。 如果一款藥上市後仍要靠持續增資維持銷售團隊；如果新增一塊收入，就要先增加更多固定費用；如果唯一產品的成長追不上下一條管線的研發消耗，那家公司只是從「靠投資人養研發」，變成「靠投資人養研發和銷售」而已。 這也是為什麼上市後的死亡谷，往往比臨床失敗更難被市場及早看見。 臨床結果是二元事件，成功或失敗通常在一天內揭曉；商業化失速卻像慢性失血。處方每季多一點、保險覆蓋慢慢改善、管理層持續說下半年會更好，公司看起來一直有希望，現金卻不等人。 ## 05｜看商業化 Biotech，先看四個時鐘 ① 第一只，是**收入時鐘**。 不要只看總市場有多少患者，要看多少人會被診斷、符合標籤、能取得給付，最後真的開始治療。首季備貨不等於患者需求，通路進貨也不等於持續處方。 ② 第二只，是**成本時鐘**。 銷售費用占收入多少？每增加一位病患，需要多少醫師教育與病患支持？公司是建立會隨收入改善的規模經濟，還是每賣一盒藥，都必須再加一層成本？ ③ 第三只，是**管線時鐘**。 單一產品可以養一家公司，前提是它真的夠大、夠久；若第一款藥只能提供有限現金，第二個適應症與下一條管線就不能一再延遲。Karyopharm 的困境，正是收入還在，接棒者卻來不及完全兌現。 ④ 第四只，是**資本時鐘**。 現金能撐多久？下一筆債何時到期？有沒有最低流動性條款？公司若在股價低點增資，稀釋會多重？醫藥研發談的是科學機率，資本市場最後問的卻是：你能不能活到答案公布那一天？ 四個時鐘不是分開走的。 放量慢，會讓固定成本顯得更重；成本太重，會壓縮管線投資；管線失敗，又會降低融資能力。最後，一次看似局部的商業問題，會沿著資產負債表擴散成公司的生存問題。 ## 06｜給台灣與中國生技公司的真正警告 華語生技市場近年越來越重視「自主商業化」，這個方向本身沒有錯。把所有資產都太早授權出去，公司永遠難以建立自己的產品與市場能力。 但自主商業化不該成為面子工程。 一家公司是否該自己賣藥，應該取決於患者是否集中、醫師是否容易覆蓋、支付路徑是否清楚、產品差異是否足以支持價格，以及現金能否撐過最慢的放量情境。若答案不漂亮，區域授權、共同銷售、分階段建隊伍，甚至在好價格出現時出售公司，都可能比「一定要自己賣」更理性。 Theravance 提醒市場，有時候最好的資本配置不是再賭一把，而是在價值還能被看見時退出；Iterum 則提醒市場，錯估商業化速度，會讓一款獲准新藥從救命繩變成加速燃燒的導火線；Karyopharm 告訴我們，最難受的狀態不是完全沒有希望，而是希望需要一年，現金只剩一個月。 ## 結語｜FDA 給的是門票，不是續命丹 新藥上市當然值得慶祝。 那代表科學、臨床、製造與監管跨過了無數道門；也代表患者終於多了一個選擇。但對公司而言，FDA 核准從來不是「風險清零」，只是把問題從實驗室搬到市場。 產品能不能被找到、被處方、被給付？收入能不能追上商業成本？下一條管線能不能及時接棒？公司能不能在資本市場關門之前拿到下一口氧氣？ 這些問題，才決定一家 Biotech 能不能真正長成 Biopharma——從仰賴外部融資的生技研發公司，變成有持續產品現金流的製藥企業。 過去，我們總把臨床三期當作生技公司的死亡谷。 現在越來越多案例證明，真正致命的懸崖可能藏在藥物獲准之後：燈都亮了、店也開了、第一張處方也出了，帳上的錢卻比患者來得更快消失。 藥上市，只證明它有資格被賣。 公司活下來，才證明這是一門生意。 — 主要資料來源：[Iterum｜2025 年第三季財報](https://ir.iterumtx.com/press-releases/detail/159/iterum-therapeutics-reports-third-quarter-2025-financial)、[Iterum｜清盤申請與臨時清算](https://www.sec.gov/Archives/edgar/data/1659323/000119312526127785/itrm-20260327.htm)、[Iterum｜法院下令清盤後續公告](https://content.bsx.com/CompanyDocuments/1099939718/Iterum%20-%20BSX%20Announcement%20-%201%20June%202026.pdf)、[Karyopharm｜2026 年第二季財報](https://investors.karyopharm.com/2026-08-13-Karyopharm-Reports-Second-Quarter-2026-Financial-Results-and-Highlights-Continued-Progress-Toward-Myelofibrosis-sNDA-Submission)、[Karyopharm｜SENTRY 三期結果](https://investors.karyopharm.com/2026-06-02-Karyopharm-to-Present-Results-from-Phase-3-SENTRY-Trial-of-Selinexor-Plus-Ruxolitinib-in-Myelofibrosis-in-Late-Breaking-Oral-Presentation-at-ASCO-2026-with-Simultaneous-Publication-in-the-Journal-of-Clinical-Oncology)、[Karyopharm｜骨髓纖維化申請路徑](https://investors.karyopharm.com/2026-07-30-Karyopharm-Plans-to-Submit-sNDA-for-Selinexor-Plus-Ruxolitinib-in-Myelofibrosis-Under-Accelerated-Approval-Pathway)、[Theravance｜CYPRESS 三期與重整](https://investor.theravance.com/news-releases/news-release-details/theravance-biopharma-reports-phase-3-cypress-study-did-not-meet)、[Theravance｜2026 年第二季財報與待完成收購](https://investor.theravance.com/news-releases/news-release-details/theravance-biopharma-inc-reports-second-quarter-2026-financial)。 本文為醫藥產業與投資教育，不構成醫療或投資建議。公司狀態與監管進度截至 2026 年 8 月 26 日；Karyopharm 尚未破產，Theravance 與 Zymeworks 的交易截至查核日仍待完成。 --- ## Can New Drugs Be Tested in a Computer First? The Virtual Cell Is Here URL: https://drugnews.com.tw/articles/2026-09-01-aido-cell-virtual-cell-drug-discovery-en.html Date: 2026-09-01 Language: en Summary: GenBio AI's AIDO Cell attempts to connect DNA, RNA, proteins and cell phenotypes through a shared state that supports sequential perturbations. The company says it matched or exceeded specialist models on 24 of 31 internal metrics, but the preview covers only K-562 and Hep-G2 cells, and novel predictions still require physical experiments, external replication and blinded testing. What if a candidate drug could be “fed” to a cell inside a computer before it ever reached a person? Could drug developers avoid more dead ends? In a 31-metric benchmark built by GenBio AI, the company says its AIDO Cell matched or outperformed the specialist models used for comparison on 24 metrics. Yet the current preview focuses on just two cancer cell lines, K-562 and Hep-G2. This is a noteworthy engineering prototype, not a substitute for cell experiments. New predictions still require physical experiments, external replication and blinded testing. ## 01 | Earlier Models Could Answer Questions. A Virtual Cell Tries to Accumulate Experience Over the past several years, artificial intelligence has become useful across many biological tasks. Give a model an amino-acid sequence and it may predict how the protein folds. Feed it single-cell data and it may identify cell types. Tell it that a gene has been switched off and it may predict part of the resulting change in gene expression. These tools are valuable, but most resemble specialists from different departments. One understands protein structure, another gene expression and another cellular images. Researchers pass the output of one model to the next and assemble the answers into a workflow. The problem is that a real cell is not a stack of independent reports. Turning off one gene can change RNA transcription and splicing. Protein abundance, location and interactions may shift. Signaling pathways respond. The cell's appearance, growth rate and drug response may then change as well. Every layer feeds back into the others. AIDO Cell's ambition is to make these layers readable from one shared “cell state.” A researcher could first knock out a gene, then add a drug and then simulate a resistance mutation. At the second step, the model would not reset to an untouched cell; it would continue from the state produced by the first intervention. That is what a stateful model is meant to do. Put more simply, an ordinary prediction tool is like taking a photograph. AIDO Cell is trying to make a film whose script can be revised while it is playing. Earlier interventions are not erased; later changes accumulate on top of them. Real experiments are sequential. Researchers rarely ask only whether a drug works. They ask what happens when one treatment is followed by another, whether a cell recovers after resistance emerges and treatment is changed, or how two experimental paths diverge from the same starting point. If a model can preserve a coherent cell state across those steps, the computer becomes more than a place to retrieve answers. It begins to resemble a digital laboratory in which experiments can be repeated and branched. ## 02 | What Has AIDO Cell Actually Achieved So Far? The easiest claim to overstate should be addressed first: AIDO Cell is not yet a complete human cell, and it has not been shown to replace physical experiments. The publicly described 1.0 preview focuses on two cancer cell lines that have been studied for decades: K-562, derived from leukemia, and Hep-G2, derived from liver cancer. Both have extensive literature and multi-omics data, which makes them practical test beds. They are still standardized laboratory cell lines, far removed from primary patient cells, immune microenvironments and organ-level interactions inside a person. According to GenBio, the team created an internal benchmark comprising five task categories and 31 metrics. It covers small-molecule perturbation, gene knockout, protein structure, RNA splicing and genome regulation. The company reports that AIDO Cell matched or exceeded the specialist models used for comparison on 24 of the 31 metrics. That is an interesting engineering signal. It is not the same as proof that the model “understands the cell.” The benchmark was developed by the company building the product, and the complete technical report currently has to be requested from GenBio. The field still needs more public evidence, external reproduction and blinded evaluation. The most valuable test is not the score on a familiar exam. It is whether the system can make a useful prediction about a genuinely unseen problem and whether a laboratory can then confirm it. AIDO Cell has not yet cleared that decisive threshold. GenBio's own explanation acknowledges the limitation. Current examples are mainly checked against established biological knowledge, while new predictions still require experimental validation. Multi-scale cell simulation also lacks a field-wide blinded standard comparable to CASP, the long-running benchmark for protein-structure prediction. The most precise description today is therefore not “the cell has been solved.” Researchers have been shown a prototype architecture that can place several biological scales in a shared state and support sequential interventions. It demonstrates that such an architecture can be built. It does not demonstrate that the architecture can replace a real cell. ## 03 | The Imatinib Example Is Not a “Miracle Prediction” Early coverage can make the imatinib example sound as if the model independently rediscovered a known mechanism. That interpretation is misleading. Imatinib is an important targeted therapy for chronic myeloid leukemia. It inhibits the abnormal ABL1 kinase. In its K-562 virtual-cell example, GenBio first used the simulated state produced by imatinib as a reference target. The system then generated structurally different candidate molecules and simulated their binding to ABL1, their gene-expression effects and indicators that might reflect cellular stress. The company reports that some candidates reproduced 65% to 88% of the imatinib-associated differentially expressed genes in simulation while retaining distinct chemical structures. The case illustrates an appealing workflow. Instead of asking only whether a molecule binds to a target, researchers specify the cellular state they want to reach and work backward to design molecules that might push the cell toward that state. But the design and ranking still take place inside the model. Whether those molecules can be synthesized, reproduce the predicted response in cell assays or create unexpected toxicity must be answered by chemistry and physical experiments. GenBio has also stated that new predictions are undergoing validation. The sensible interpretation of the imatinib example is therefore “workflow demonstration.” It is not evidence that a new medicine has been discovered, and it is certainly not evidence that a virtual cell has passed a clinical test. ## 04 | Why Would Drug Companies Still Want It? Because the most expensive part of drug development is often not generating ideas. It is eliminating the wrong ones. A team can design thousands of molecules at once, but it cannot synthesize every molecule, expose every relevant cell to each one and measure every resulting change in genes and proteins. Laboratory capacity is finite. The recurring questions are which ten candidates deserve to be made first, which signal might be a data artifact, and which combination appears promising but may hide toxicity. If virtual cells become useful, their first job will not be to make pipettes disappear. It will be to identify the experiments for which the pipettes are most worth picking up. In early discovery, a virtual cell could help narrow the candidate set. In combination therapy, it could simulate different treatment sequences. In resistance research, it could branch into alternative mutation paths. In a rare disease with very few physical samples, patterns learned from other cells might help researchers form a better starting hypothesis. There is also a paradox. The more completely a model tries to simulate a cell, the more it depends on high-quality, multimodal and traceable experimental data. AI will not remove the laboratory from drug development. It may instead increase demand for standardized perturbation datasets, single-cell analysis, cellular imaging, protein quantification and automated experimentation. A credible virtual world has to return repeatedly to the physical world for calibration. ## 05 | This Is Not a One-Company Race GenBio is not the only company pursuing a virtual cell. In March 2026, Xaira Therapeutics introduced X-Cell, a model focused on predicting how cells respond to genetic or drug perturbations. Its technical report describes a model scaled to 4.9 billion parameters, trained and evaluated with more than 20 million cells and a large collection of perturbation combinations. The two companies emphasize different approaches. X-Cell's public work concentrates more heavily on large perturbation datasets and prediction across cellular contexts. AIDO Cell emphasizes a shared state, multi-scale readouts and sequential operation. Company announcements do not yet establish which approach is ahead, and parameter count is not a direct measure of biological understanding. Any fair comparison should ask at least four questions: 1. Can the model predict outcomes in cells and perturbations it has not seen? 2. Can independent laboratories reproduce the results? 3. Can the model tell researchers where its predictions are unreliable? 4. Does using it improve experimental success rates, time or cost in practice? The winner will not be the company that says “world model” most convincingly. It will be the one that can turn an answer inside a computer into a stable result in the laboratory. ## 06 | The FDA's Direction Is Encouraging, but It Is Not a Free Pass The regulatory environment is changing. The U.S. Food and Drug Administration includes computer simulation, organ-on-chip systems and organoids in its discussion of New Approach Methodologies, or NAMs. A draft released in March 2026 says that use of a NAM should address its context of use, human biological relevance, technical characterization and fitness for purpose. Final frequently asked questions for cell and gene therapy products published in August also state that the agency may consider scientifically justified in vitro or computational approaches capable of producing valid data in appropriate circumstances. That is a meaningful change in direction. It should not be reported as “the FDA has endorsed virtual cells.” The current documents do not endorse AIDO Cell, nor do they say that one virtual-cell system can directly replace necessary animal, cell or human studies. Regulators accept validated methods for defined uses; they do not approve a product simply because its name sounds advanced. For virtual-cell companies, one of the most important commercial assets may eventually be not the model alone but an evidence package that establishes the settings in which its output can be trusted. ## 07 | Taiwan's Opportunity Is Broader Than Building a Chinese-Language Model Should Taiwan pursue virtual-cell technology? Yes, but the opportunity should not be reduced to training a larger model or creating a Chinese-language interface. The required value chain is long: high-quality clinical samples, multi-omics measurements, single-cell sequencing, pathology and cell imaging, high-throughput perturbation experiments, data standards, computing infrastructure and drug-development teams able to send predictions back to the laboratory for validation. Taiwan's healthcare system, hospital biobanks, semiconductor and server capabilities, and the experimental and manufacturing experience of parts of its biotech sector create plausible entry points. But “could benefit” and “has won orders” are different claims. For the industry and investors, the important questions are not whether a company has written AI on its pitch deck. Does it have data that can be used legally and updated continuously? Can it standardize data from different hospitals? How many novel model predictions have been confirmed experimentally? Are customers willing to pay to place the system inside a real research workflow? Those figures reveal more about a commercial moat than the number of billions of parameters. ## 08 | When the Next Breakthrough Is Announced, Check These Four Validation Points Virtual cells can produce dazzling demonstrations. Click a gene and a pathway changes color. Add a drug and the cell's appearance shifts. Ask the model to generate a new set of molecules. The more complete the screen looks, the easier it is to forget that the display may still represent one model remaining internally consistent with its own assumptions. The next time a company announces a breakthrough in whole-cell simulation, look for four answers. 1. **Was there a genuinely blinded test whose answer remained hidden until the end?** If the model has already seen closely related data, a strong score may reflect familiarity with the question type. 2. **Has the system moved beyond data-rich standard cell lines such as K-562 and Hep-G2?** Performance in primary cells, patient-derived cells or complex co-culture systems would be a more demanding test. Answering questions about textbook cell lines does not guarantee performance in the clinical world's disorder. 3. **Who validated the new prediction?** Experiments performed by the developer are an important start. Reproduction by an independent team using different equipment and batches would raise confidence further. 4. **What changed for the customer?** How many experiments were avoided, how much time was saved, and how much did the success rate improve? If those numbers remain absent, the virtual cell may be a strong research instrument without yet being a product that customers will buy at scale. These questions are simple, but they separate “the model looks impressive” from “the model changes drug development.” That distinction matters to scientists and investors alike. ## Conclusion | Biology Is Becoming Computable, but It Has Not Been Computed Away AlphaFold showed that computers can provide answers to an exceptionally difficult biological problem in a form that changes research workflows. Virtual cells want to go a step further. Instead of looking only at the shape of one protein, they aim to model how many molecules pull on one another inside a cell and how successive interventions accumulate into a new state. If that vision succeeds, drug development could shift from less directed searching toward more deliberate design. But AIDO Cell remains a preview. It has two prototype cell lines. Its public examples largely revisit known biology. New predictions still await physical experimental validation, and the field lacks a shared blinded benchmark. The right response is neither to announce the end of wet-lab research nor to dismiss the system because it is incomplete. The important transition is that researchers now have a digital-cell architecture that can be questioned sequentially. The next task is to expose every compelling simulation to the unforgiving test of the physical world. Only after passing that test can biology move from something that can be described toward something that can reliably be computed. --- ## 新藥能先在電腦裡試？「虛擬細胞」來了！ URL: https://drugnews.com.tw/articles/2026-09-01-aido-cell-virtual-cell-drug-discovery.html Date: 2026-09-01 Language: zh-Hant Summary: GenBio AI 的 AIDO Cell 嘗試用共享狀態串起 DNA、RNA、蛋白質與細胞表型，並連續模擬基因、藥物與抗藥性擾動。公司基準顯示 31 項指標中有 24 項追平或超越專用模型，但目前只有 K-562、Hep-G2 兩種原型細胞，新預測仍待實體實驗、外部重現與盲測。 如果新藥進入人體前，能先在電腦裡「餵」給細胞，研發是否能少走冤枉路？GenBio AI 的 AIDO Cell 在公司自建 31 項指標中，有 24 項追平或超越專用模型；但目前只聚焦 K-562 與 Hep-G2 兩種癌細胞株。這是值得注意的工程原型，不是細胞實驗的替代品；新預測仍需實體實驗、外部重現與盲測。 ## 01｜以前的模型會回答，虛擬細胞想學會「經歷」 過去幾年，人工智慧已經能處理不少生物學問題。 給它一段胺基酸序列，它可以預測蛋白質可能折成什麼形狀；輸入一批單細胞資料，它可以辨認細胞類型；告訴它某個基因被關閉，它也可能預測部分基因表現會怎麼改變。 這些工具很有用，卻多半像不同科別的專家：有人懂蛋白質，有人懂基因表現，有人懂細胞影像。研究者把上一個模型的答案交給下一個模型，最後拼成一條工作流程。問題是，真實細胞並不是幾張彼此獨立的報表。 一個基因被關閉，可能改變 RNA 的轉錄與剪接；蛋白質數量、位置和交互作用跟著變；訊號路徑受到牽動；最後細胞的外觀、生長速度與藥物反應也可能改寫。每一層都在互相回饋。 AIDO Cell 的野心，就是讓這些層級從同一個「細胞狀態」讀出。研究者可以先關閉一個基因，再加入一種藥，接著模擬抗藥性突變；第二步不是重新面對一顆乾乾淨淨的細胞，而是接在第一步造成的狀態上。 這就是所謂「有狀態」的模型。 更直白地說，普通預測工具像拍一張照片；AIDO Cell 想拍的是一部可以中途改劇本的電影。你做過的事不會被歸零，後面的變化會累積在前面的變化之上。 真實實驗本來就是連續的。研究者很少只問「這個藥有沒有作用」，更常問的是：先用藥、再加藥會怎樣？產生抗藥性之後換另一種治療，細胞會不會恢復？同一個起點分成兩條實驗路徑，結果差在哪裡？ 如果模型真的能保留前後一致的細胞狀態，電腦就不只是查答案的地方，而開始像一間可以反覆操作的數位實驗室。 ## 02｜它現在到底做到了什麼？ 先把最容易被誇大的地方說清楚：AIDO Cell 還不是一顆完整的人類細胞，更沒有證明能取代實體實驗。 目前公開的 1.0 預覽版，聚焦兩種被研究數十年的癌細胞株：白血病來源的 K-562，以及肝癌來源的 Hep-G2。這兩種細胞累積了大量文獻與多體學資料，適合拿來測試模型；但它們仍是實驗室裡的標準細胞株，和患者體內的原發細胞、免疫微環境、器官互動，相隔很遠。 依 GenBio 官方資料，團隊建立了一套包含五類任務、31 項指標的內部基準，涵蓋小分子擾動、基因剔除、蛋白質結構、RNA 剪接與基因體調控。公司宣稱 AIDO Cell 在 31 項指標中，有 24 項達到或超越所比較的專用模型。 這是一個值得注意的工程訊號，卻不能直接等同於「模型已經理解細胞」。原因很簡單：這套基準由開發公司建立，完整技術報告目前需向公司申請，產業界仍需要更多公開資料、外部重現與盲測。 真正令人期待的，不是它在熟悉考題上拿了幾分，而是它能不能對一個沒看過的新問題做出預測，然後被實驗室證實。 這個最關鍵的門檻，目前還沒有跨過。 GenBio 自己也在官方說明中承認：現有案例主要對照既有生物學知識，新的預測仍要接受實體實驗驗證；多尺度細胞模擬也還缺少一套像蛋白質結構預測競賽 CASP 那樣，由整個領域共同認可的盲測標準。 所以，今天較精確的說法不是「細胞已被算透」，而是：研究者第一次看到一套能把多種生物層級放進共享狀態、又能連續操作的原型系統。 它證明了一種架構可以被做出來，還沒有證明這套架構足以替代真實細胞。 ## 03｜伊馬替尼案例，不是「神預測」 原始消息常把伊馬替尼案例寫成「模型成功重現已知藥物機制」。這句話容易讓人誤會。 伊馬替尼是慢性骨髓性白血病的重要標靶藥物，會抑制異常的 ABL1 激酶。GenBio 在 K-562 虛擬細胞裡，先把伊馬替尼造成的模擬狀態設為參考目標，再生成結構不同的新候選分子，模擬它們與 ABL1 的結合、基因表現變化，以及可能反映細胞壓力的指標。 公司報告稱，部分候選分子在模擬中可重現伊馬替尼 65% 至 88% 的差異表現基因，同時呈現不同的化學結構。 這個案例展示的是一條很有吸引力的工作流程：不只問「某個分子能不能黏住靶點」，而是先指定希望細胞抵達的狀態，再反向設計可能把細胞推向那個狀態的分子。 可是，它仍然是模型裡的設計與排序。 這些候選分子是否真的能合成、是否在細胞實驗中產生相同反應、是否有意料之外的毒性，還要交給化學合成與實體實驗回答。官方也明確表示，新的預測正在驗證中。 因此，伊馬替尼案例最合理的意義是「工作流程展示」，不是一款新藥已被發現，更不是虛擬細胞已經通過臨床考驗。 ## 04｜為什麼藥廠仍會想要它？ 因為新藥開發最昂貴的，往往不是產生點子，而是排除錯誤。 一個團隊可以一次設計成千上萬個分子，卻不可能把每個分子都做出來、餵給所有細胞、測完每一種基因與蛋白質變化。實驗資源有限，真正的問題始終是：哪十個最值得先做？哪一個訊號可能只是資料偏差？什麼組合最有機會，也最可能藏著毒性？ 虛擬細胞如果有用，第一份工作不會是讓移液槍消失，而是決定哪幾次最值得拿起移液槍。 在早期藥物發現，它可以協助縮小候選範圍；在合併療法，它可以先模擬給藥順序；在抗藥性研究，它可以分岔出不同突變路徑；在罕見疾病，如果真實樣本非常少，它或許能從其他細胞學到的規律，幫研究者建立更好的起始假說。 但這裡也藏著一個悖論：模型越想模擬完整細胞，就越需要高品質、多模態、可追溯的實驗資料。人工智慧不會讓實驗室退場，反而可能提高市場對標準化擾動資料、單細胞分析、細胞影像、蛋白質定量與自動化實驗的需求。 虛擬世界要可信，必須不斷回到真實世界校正。 ## 05｜這不是一家公司獨跑的比賽 GenBio 並不是唯一追逐虛擬細胞的公司。 2026 年 3 月，Xaira Therapeutics 發表 X-Cell，聚焦預測細胞受到基因或藥物擾動後的反應。其技術報告把模型擴大到 49 億參數，訓練與評估使用超過 2,000 萬個細胞與大量擾動組合。 兩家公司切入角度不同。X-Cell 的公開成果較集中在大規模擾動資料與跨細胞情境預測；AIDO Cell 則強調共享狀態、多尺度讀出與連續操作。現在還不能憑公司新聞稿判定誰領先，更不能把參數多寡直接當成生物學理解力。 **真正公平的比較，至少要問四件事：** ① 第一，模型是否能在沒看過的細胞與擾動上預測？ ② 第二，結果能否被獨立實驗室重現？ ③ 第三，模型能否告訴研究者自己在哪些情境不可靠？ ④ 第四，使用模型後，實驗成功率、時間與成本是否真的改善？ 這場競賽最後比的不是誰最會說「世界模型」，而是誰能把電腦裡的答案，穩定變成實驗室裡的結果。 ## 06｜FDA 的風向是利多，卻不是通行證 監管環境確實正在改變。 美國食品藥物管理局（FDA）把電腦模擬、器官晶片、類器官等工具納入「新方法學」（New Approach Methodologies，NAMs）的討論。2026 年 3 月發布的草案，提出使用新方法學時應說明使用情境、人體生物相關性、技術特性與是否符合目的；8 月公布的細胞與基因治療常見問題最終版，也表示在適當情境下，願意考慮有科學依據且能產生有效資料的體外或電腦模擬方法。 這是風向轉變，但不能寫成「FDA 已認可虛擬細胞」。 現行文件沒有替 AIDO Cell 背書，也沒有說一套虛擬細胞可以直接取代必要的動物、細胞或人體研究。監管接受的是經過驗證、符合特定用途的方法；不是一個聽起來很先進的產品名稱。 對虛擬細胞公司來說，未來最大的商業資產之一，可能不是模型本身，而是「這個模型在什麼情境下值得相信」的證據包。 ## 07｜台灣真正能切入的，不只是做一個中文版模型 台灣要不要追虛擬細胞？答案當然是要，但不必把競爭想成再訓練一個更大的模型。 虛擬細胞需要的產業鏈很長：高品質臨床樣本、多體學量測、單細胞定序、病理與細胞影像、高通量擾動實驗、資料標準、運算資源，以及能把預測送回實驗驗證的藥物開發團隊。 台灣的醫療體系、醫院生物資料庫、半導體與伺服器能力，以及部分生技公司的實驗與製造經驗，確實提供了切入點。但「理論上能受益」和「已經拿到訂單」是兩回事。 對產業與投資人而言，真正該追的不是哪家公司在簡報上寫了 AI，而是它有沒有取得可合法使用、可持續更新的資料；是否能把不同醫院的資料整理成相容格式；模型提出的新預測，有多少被實驗證實；客戶是否願意付費，把它放進真實研發流程。 這些數字，比「模型有幾十億參數」更接近商業護城河。 ## 08｜下一則新聞出現時，先看這四個驗證點 虛擬細胞很容易製造令人目眩的展示：點一下基因，螢幕上的路徑變色；加入藥物，細胞外形跟著改變；再讓模型生成一批新分子。畫面愈完整，讀者愈容易忘記，那仍可能只是同一套模型對自己的推論保持一致。 **因此，下一次看到「全細胞模擬再突破」，不妨先找四個答案。** ① 第一，有沒有真正保留到最後才揭曉答案的盲測？如果模型早已看過相近資料，漂亮成績可能只是熟悉題型。 ② 第二，有沒有跨出 K-562、Hep-G2 這類資料豐富的標準細胞株，進到原發細胞、病人來源細胞或更複雜的共培養系統？能在教科書細胞上作答，不代表能處理臨床世界的混亂。 ③ 第三，新預測由誰驗證？開發公司自己的實驗是重要起點，若能由獨立團隊、不同設備與不同批次重現，可信度才會再上一階。 ④ 第四，客戶用了模型之後，究竟少做多少實驗、少花多少時間，又提高多少成功率？這些數字若長期缺席，虛擬細胞就可能是一項很好的研究工具，卻還不是能被大規模採購的產品。 四個問題都很樸素，卻能把「模型看起來很厲害」與「模型真的改變研發」分開。對科學界如此，對投資人更是如此。 ## 結語｜生物學開始可計算，但還沒有被算完 AlphaFold 讓世界看到，電腦可以在一個極難的生物問題上，給出足以改變研究流程的答案。虛擬細胞想再往前一步：不只看一顆蛋白質的形狀，而是看許多分子在細胞裡如何互相拉扯，並在一次又一次干預後，累積成新的狀態。 這個願景一旦實現，藥物研發可能從漫無目的的搜尋，走向更有方向的設計。 可是 AIDO Cell 今天仍是預覽版：只有兩種原型細胞，公開案例以已知生物學為主，新的預測尚待實體實驗驗證，整個領域也缺乏共同的盲測標準。 所以，最好的態度不是急著宣布「濕實驗將被淘汰」，也不是因為模型還不完美就忽視它。 應該記住的是這個轉折：我們開始擁有一套可以被連續追問的數位細胞；接下來，科學界要做的是讓每一個漂亮的模擬，都接受真實世界毫不留情的檢驗。 如果那一關跨得過去，生物學才會真正從「可以被描述」，走向「可以被計算」。 ## 主要資料來源 - [GenBio AI｜AIDO Cell 官方技術說明](https://genbio.ai/aido-cell-simulator/) - [GenBio AI｜官方團隊頁](https://genbio.ai/our-team/) - [Xaira Therapeutics｜X-Cell 官方發布](https://www.xaira.com/news/xaira-therapeutics-launches-x-cell-its-first-virtual-cell-model-trained-on-the-largest-ever-genome-wide-perturbation-dataset-x-atlas-pisces) - [FDA｜新方法學草案](https://www.fda.gov/regulatory-information/search-fda-guidance-documents/general-considerations-use-new-approach-methodologies-drug-development) - [FDA｜細胞與基因治療常見問題最終指引](https://www.fda.gov/regulatory-information/search-fda-guidance-documents/frequently-asked-questions-developing-potential-cellular-and-gene-therapy-products) - [Nobel Prize｜David Baker](https://www.nobelprize.org/prizes/chemistry/2024/baker/facts/) 本文為產業與科學資訊整理，不構成醫療建議或投資建議。AIDO Cell 目前為預覽版，文中公司性能數據屬開發方公布，仍待更多獨立驗證。 --- ## GLP-1 Drugs Are Making Luxury Guests Eat Less. How Can Five-Star Hotels Protect Revenue? URL: https://drugnews.com.tw/articles/2026-08-31-glp1-luxury-hospitality-spending-shift-en.html Date: 2026-08-31 Language: en Summary: A U.S. household transaction study found that grocery spending fell 5.3% on average six months after GLP-1 adoption and 8.2% among higher-income households. Some luxury hotels are already testing smaller portions, protein-forward menus, zero-proof drinks and wellness services. The strategic question is whether hospitality operators can translate lower food volume into better service, pricing and repeat demand. The champagne is cold. The dessert tower is full. The guest paying several thousand dollars for a luxury stay takes one look and says, “I really cannot eat any more.” That scene captures a new operating question for hospitality. GLP-1 medicines are best known for transforming obesity and diabetes treatment. Their commercial spillover, however, is reaching restaurants, grocery baskets, alcohol occasions, wellness services and the economics of a hotel stay. A U.S. household transaction study found that grocery spending fell 5.3% on average six months after GLP-1 adoption. Among households earning more than $125,000 a year, the decline was 8.2%. Spending at fast-food outlets, coffee shops and other limited-service restaurants fell 8.0%. Some luxury hotels are already testing smaller portions, protein-forward menus, flexible meal formats and more ambitious zero-proof programs. Those are meaningful early cases. They are not evidence that GLP-1 use has already reduced revenue across the global five-star hotel sector, and they do not prove that every change in guest behavior is caused by the drugs. The strategic question is therefore more precise: **when appetite shrinks, can a hotel preserve value by selling better design, personalization and experience rather than more food?** ## 01 | Luxury Hospitality Was Built Around the Visible Language of “More” Traditional luxury hotels know how to make value visible. The breakfast buffet resembles a market. Afternoon tea arrives on three tiers. The minibar holds chocolate, nuts, spirits and several kinds of sparkling water. The room rate buys not only accommodation, but also the feeling that someone has anticipated every possible desire. The underlying formula is abundance. GLP-1 medicines can change how some guests experience that abundance. These drugs affect appetite and satiety signaling. The U.S. Food and Drug Administration prescribing information for Wegovy, the semaglutide product approved for chronic weight management, states that it delays gastric emptying. Common gastrointestinal adverse reactions in adult weight-management trials included nausea, vomiting, diarrhea, constipation and gastroesophageal reflux disease. Not every user experiences the same effect. Not every guest taking a GLP-1 medicine suddenly rejects dessert or alcohol. Yet for some people, a large high-fat meal can stop looking generous and start feeling uncomfortable. That distinction matters for service. A portion that once communicated hospitality may now communicate waste, pressure or lack of attention. ## 02 | The Spending Signal Has Moved Beyond Appetite Surveys The most important evidence is not a social-media anecdote. It is the appearance of the signal in transaction data. The 2026 *Journal of Marketing Research* study linked U.S. household medication surveys with observed purchases. It identified 2,602 households that adopted GLP-1 therapy and matched them with similar households that did not. Six months after adoption, grocery spending was 5.3% lower on average. The reduction reached 8.2% among households with annual income above $125,000. Spending at limited-service restaurants, including fast food and coffee shops, was 8.0% lower. Higher-income households are important to luxury hospitality, which makes the result difficult to ignore. But the limits are equally important. The study examines U.S. household consumption. It does not directly measure five-star hotel revenue. It is observational rather than randomized, even though the researchers used matching and a difference-in-differences design. Income may capture access to medicine, baseline consumption, lifestyle and other differences. The study also found that grocery spending moved back toward baseline for some households after treatment discontinuation. The correct interpretation is therefore an early warning, not a financial forecast. Luxury-hotel operators should pay attention to changing food demand among affluent consumers. They should not assume that every suite guest has “switched off” appetite, or that the spending effect will be identical across markets, hotels and treatment durations. ## 03 | If the Guest Eats Less, the Hotel Cannot Look as If It Gives Less The hard part is not cutting a steak into a smaller portion. The hard part is making the smaller portion feel worth the price. Examples reported by *Forbes Travel Guide* show how that transition may work. Food options at St. Regis Washington, D.C. have moved toward lighter, smaller, protein-forward, plant-forward and nutrient-dense choices. Malibu Kitchen at The Ned Doha offers greater portion flexibility. The Broadmoor in Colorado allows guests to divide meals into smaller servings without abandoning presentation or visual impact. This is not an attempt to turn five-star cuisine into boiled chicken and plain vegetables. It asks hotels to use ingredient quality, plate design, tableware and more attentive service to make “less” feel intentional. Yesterday, value was communicated by how full the plate looked. Tomorrow, value may come from three bites that are exactly right. That is also why removing dessert is too crude a response. The point raised by The Broadmoor's pastry team was not to eliminate pleasure, but to preserve it while changing sugar, fat or other ingredients without making the guest feel that the experience had been sacrificed. Luxury service also needs to avoid public labeling. A large “weight-loss injection menu” could make guests feel classified by their medication status. A better design places smaller portions, protein-forward dishes, plant-led choices and zero-proof drinks on the regular menu. A guest using a GLP-1 medicine can choose them without disclosing a diagnosis, while every other guest retains the same option. Luxury moves from “we filled the table for you” to “we calibrated the experience for you.” The financial pressure can accumulate across a stay. A restaurant on the street may offset a lower check with faster table turnover. A luxury hotel carries banquet facilities, breakfast operations, room service, bars, pastry kitchens and labor-intensive service. One skipped drink or dessert looks small. Repeated across breakfast, tea, dinner and the minibar, the effect can reduce food-and-beverage revenue throughout the guest journey. The hotel cannot simply halve the portion and preserve the old price without changing anything else. Luxury guests pay to feel understood, not manipulated. If a smaller plate looks like hidden cost-cutting, the damage to trust can exceed the ingredient savings. GLP-1 therefore becomes a service-design test. Chefs need to improve nutrition and flavor per bite. Food-and-beverage managers need a portion and pricing architecture that feels flexible rather than cheap. Servers need to ask about preferences without asking for medical history. Bars need to keep non-drinkers inside the social center of the hotel. The difficult task is not removing one piece of cake. It is redefining abundance when abundance no longer means food weight. ## 04 | The Bar Is Not Disappearing. It Is Learning to Sell Ritual Without Alcohol Alcohol is one of the revenues luxury hotels are least willing to lose. A cocktail does not sell ethanol alone. It sells an evening ritual, an entry point to social interaction and a photogenic object on the table. If a guest drinks less, replacing the whiskey with a glass of tap water destroys most of that value. Miami Beach EDITION has expanded zero-proof options, while Four Seasons Fort Lauderdale and Omni La Costa are among the hotels offering more complete non-alcoholic drink experiences. These beverages retain aroma, glassware, color and bartending craft. Removing alcohol does not require removing price or ceremony. But the claim “GLP-1 makes people stop drinking” goes beyond current evidence. A small Phase 2 randomized trial published in *JAMA Psychiatry* in 2025 enrolled 48 adults with symptoms of alcohol use disorder. Low-dose semaglutide reduced alcohol craving, drinks per drinking day and some heavy-drinking outcomes. Not every alcohol endpoint was significant, and treatment lasted only nine weeks. Later reviews also described the signal as worthy of study while emphasizing the need for larger trials. For hotels, a stronger zero-proof program is a demand test. It is not proof that semaglutide or the broader GLP-1 class has become an established treatment for alcohol use disorder. The commercial lesson is narrower and more useful. A beverage occasion that might otherwise disappear can be redesigned as a premium non-alcoholic experience. The hotel is not surrendering to lower alcohol consumption. It is finding a way to charge for “not drinking, without leaving the occasion.” ## 05 | Smaller Meals Can Expand Into a GLP-1 Companion Economy Once the guest's needs extend from dessert choice to nutrition, muscle maintenance, gastrointestinal comfort and medication routines during travel, a hotel sees a service chain rather than one menu item. Four U.S. nutrition and obesity-medicine organizations issued joint guidance in 2025 that emphasized nutrient density, adequate protein and resistance training during GLP-1 treatment for obesity. The combination matters. A protein drink is not a substitute for resistance exercise, clinical assessment or an adequate overall diet. This helps explain why resorts and wellness properties are linking kitchens, gyms, dietitians and recovery services. Hilton Head Health's GLP-1 Support Experience combines chef-prepared meals, protein guidance, strength and recovery classes, consultations with a dietitian and cooking workshops. At the more expensive edge, hospitality begins to approach medical territory. The Four Seasons Hotel Los Angeles at Beverly Hills lists IV and “body maintenance” services provided by the third-party operator Immortelle Integrative Health. Its seasonal-treatment pages list a $750 body-maintenance IV and a $1,000 pre- and post-flight program that includes IV therapy. Those pages establish that luxury hospitality is testing new paid services. They do not establish through high-quality clinical trials that these treatments preserve muscle or skin in GLP-1 users. Four Seasons also states that the provider is a third party and that the hotel does not guarantee the services. The boundary is essential. Integrating qualified nutrition advice, reasonable exercise and thoughtful accommodation can be service innovation. Describing every IV bag as a medical breakthrough can turn luxury into expensive pseudoscience. Privacy creates another operating risk. A hotel may need to know that a guest wants refrigeration, a smaller meal or a different exercise schedule. That does not mean the guest wants to disclose a medical history. “I prefer a small portion” is not permission to ask which medicine the guest uses. Personalization that feels like interrogation can drive away the customer it is meant to retain. A mature design therefore avoids forcing guests to identify themselves as GLP-1 users. Smaller portions, zero-proof cocktails, protein-forward dishes, in-room refrigeration and strength classes can be available to everyone. When a service becomes medical, responsibility should transfer clearly to a properly qualified professional. Hotels can sell a wellness experience. They cannot blur medical accountability. ## 06 | GLP-1 Is Not Eliminating Desire. It Is Redirecting the Wallet It is tempting to tell this story as the end of snacks, desserts and hotel bars. Consumers may not stop spending. They may move the spending. PwC surveyed more than 3,000 current, former and prospective GLP-1 users in 2026. Drawing on Numerator data, it estimated that about 21% of U.S. households included a current user as of May. Among current users, 61% said they bought fewer sweets and 56% bought fewer salty snacks. The survey also reported increases across fresh produce, protein foods, nutrition, clothing, fitness and self-care categories. These are U.S. survey and consumer-association data, not a universal map for every person or country. Yet they indicate a practical direction: when companies can no longer sell “eat more” as easily, they begin selling “eat with more precision, look better and feel better cared for.” The shift can be viewed in three circles. 1. **Food itself.** Large packages, calorie density and impulse add-ons face pressure, while smaller portions and protein- or fiber-forward products gain a new position. Nestlé launched Vital Pursuit in 2024 for GLP-1 users and weight-management consumers, emphasizing portion control, protein and nutrient density. The positioning does not guarantee success, but it shows that a large food company treats lower appetite as a product-design problem rather than a passing headline. 2. **Needs created by physical change.** Clothing size, muscle maintenance, exercise, sleep, skin and social habits can change. Some money previously spent on snacks or alcohol may move toward apparel, fitness, nutrition and self-care. 3. **Experience, where hotels have an advantage.** A food company can sell one small meal. A hotel can build that meal into a complete stay: a butler records preferences, the kitchen adjusts portions, the bar creates a zero-proof pairing, a trainer builds a strength session and the guest leaves with something that can continue at home. In other words, a hotel may translate lower food volume into more service. This is an advantage over packaged food. When a cookie package shrinks, consumers often see disguised inflation. When a dish becomes smaller but more personal, better sourced and more carefully served, a guest may accept a higher unit price. That only works when the experience genuinely improves. Ordinary juice in a beautiful glass and plain chicken renamed as a “metabolic menu” are not innovation. They are old products using a popular medicine as decoration. ## 07 | Four Metrics Matter More Than a Branded “GLP-1 Hotel” Package For Taiwan's hotel operators, the right first move is not to declare a GLP-1 hospitality revolution. It is to measure behavior. Four metrics provide a useful starting point: 1. **Small-portion take rate.** Are guests actively choosing flexibility, or are they simply leaving more food uneaten? 2. **Zero-proof check value.** Do non-alcoholic drinks preserve beverage spending, replace alcohol spending or bring in a different occasion? 3. **Wellness repeat purchase.** Do nutrition, strength and recovery services bring guests back, or generate only one social-media visit? 4. **Food waste per occupied room.** Can smaller formats reduce waste while protecting price and satisfaction? If guests already leave food on the plate, a large portion does not create value; it creates waste. A smaller format becomes good business when it lowers ingredient waste, protects the check and improves satisfaction. A smaller plate with an expensive “GLP-1 friendly” label is much easier to see through. The economics will also differ by hotel type. Business hotels often include breakfast in the room rate. Lower consumption may not reduce revenue immediately, although it may reduce preparation and waste. Resorts depend more heavily on restaurants, bars and paid activities, making demand changes more direct. Banquet-led hotels face another question: if guests eat less, do weddings and large gatherings still need a full table to communicate generosity? The numbers are more valuable than the trend label. A trend reaches the income statement only when a company translates it into product design, price, cost or repeat behavior. ## Conclusion | The Next Version of Luxury May Be Knowing What Not to Put on the Table The real challenge for a five-star hotel is not removing two cans of soda from the minibar. It is making a small meal feel valuable, turning a non-alcoholic drink into a social ritual and building nutrition, strength and recovery services that guests choose again—without crossing the line between hospitality and unproven medical claims. The previous version of luxury meant that everything was waiting when the guest opened the door. The next version may mean that the hotel already knows what does not need to be there. What remains fits the guest better and feels more worth paying for. GLP-1 medicines have not removed desire. They may be changing “bring me another one” into “make this one exactly right.” --- ## 減重針逼瘋五星級飯店？客人吃少了，飯店靠什麼賺？ URL: https://drugnews.com.tw/articles/2026-08-31-glp1-luxury-hospitality-spending-shift.html Date: 2026-08-31 Language: zh-Hant Summary: 美國家戶交易研究顯示，採用 GLP-1 六個月後，雜貨支出平均下降 5.3%，高收入家庭下降 8.2%；部分高端飯店也開始測試小份餐、高蛋白、無酒精飲品與健康陪伴服務。真正的商業題目，不是客人吃少了多少，而是飯店能否把食物的量重新翻譯成服務、價格與回購。 香檳冰好了、甜點塔疊滿了，住一晚幾萬元的客人卻只說：「我真的吃不下。」一項美國家戶研究顯示，使用 GLP-1 六個月後，雜貨支出平均下降 5.3%，高收入家庭下降 8.2%。高端飯店也開始測試小份餐、高蛋白與無酒精選項；但這只是第一波案例，不能直接推論全球五星飯店營收，更不能把所有改變都歸因於減重藥。 ## 01｜五星級飯店最熟悉的語言，原本叫做「多」 傳統奢華飯店很懂得把價值做成看得見的東西。 自助早餐要像一座小型市集，下午茶要有三層甜點塔，客房冰箱裡從巧克力、堅果到烈酒一格不缺。房費裡還包含一種「任何時刻都有人替我準備好」的感覺。 這套公式的核心，是豐盛。 可是一旦客人開始使用 GLP-1（類升糖素胜肽－1）類藥物，身體對豐盛的回應可能改變。 這類藥物會影響食慾與飽足訊號；以美國食品藥物管理局核准的 semaglutide（司美格魯肽）減重產品 Wegovy 為例，仿單明確記載它會延緩胃排空，成人減重試驗中常見噁心、嘔吐、腹瀉、便祕與胃食道逆流等腸胃道反應。 不是每個人都會出現相同反應，也不是每位使用者都突然不碰甜食。但對部分客人來說，一盤過大的高脂餐點，不再代表慷慨，反而可能變成壓力。 更值得飯店緊張的，是這件事已經開始出現在消費資料裡。 2026 年刊登於《Journal of Marketing Research》的研究，把美國家戶用藥調查與實際交易資料連在一起，追蹤 2,602 個採用 GLP-1 的家戶，並配對相近的未採用家戶。結果顯示，開始使用後六個月，雜貨支出平均下降 5.3%；年收入超過 12.5 萬美元的家戶，降幅達 8.2%；速食、咖啡店與其他有限服務餐飲的支出則下降 8.0%。 這組數字特別刺眼，因為高收入族群恰好也是高端旅宿最想留住的客人。 不過，研究談的是美國家戶消費，不能直接拿來推算全球五星飯店營收；它也顯示，部分人停藥後，雜貨支出會往原本水準靠近。這比較像一張提早亮起的警示燈，還不是飯店業的財報判決書。 而且，高收入家戶降得更多，也不代表「有錢人一定比較不想吃」。研究作者指出，收入可能同時反映藥物取得能力、原本的消費水準與生活型態。採用者雖與相近的未採用家戶完成配對，現實世界研究仍無法像隨機試驗那樣排除全部干擾因素。 因此，我們可以說高端消費者的食品支出變化值得飯店警戒；不能反過來宣稱，每個住進五星套房的人都已把胃口關掉。數字提供方向，個別客人的需求仍要靠服務現場辨認。 ## 02｜客人吃少了，不能讓他覺得「飯店也少給了」 高端飯店真正難的，不是把牛排切小。 難的是牛排切小之後，客人仍然覺得這一餐值得。 《Forbes Travel Guide》列出的案例很能說明這場轉向。華盛頓特區瑞吉酒店的餐飲選項，開始朝較輕盈、小份、高蛋白、植物比例較高與營養密度較高的方向調整；杜哈 The Ned 的 Malibu Kitchen 提供更有彈性的份量；科羅拉多州 The Broadmoor 允許客人把餐點拆成較小份，卻不犧牲擺盤與視覺效果。 這不是把五星料理變成水煮餐。 恰恰相反，飯店得用更好的食材、更細的份量設計、更漂亮的器皿與更精準的服務，把「少」重新包裝成一種選擇權。 過去，價值感來自盤子有多滿；未來，價值感可能來自「我只吃三口，但每一口都剛好」。 這也是為什麼「撤掉所有甜點」不是好策略。原報導裡，The Broadmoor 的甜點師談的並非取消享受，而是讓替換糖、油或其他成分後的甜點，仍吃不出被犧牲的感覺。 高端服務最怕貼標籤。飯店若直接推出一張斗大的「減重針菜單」，可能讓客人覺得自己被公開分類；更聰明的做法，是把小份、高蛋白、植物導向與無酒精選項自然放進所有人的菜單，讓使用藥物的人可以選，沒有使用的人也不必被排除。 奢華於是從「替你堆滿」變成「替你拿捏」。 而高端飯店之所以可能比一般餐廳更早感受到這股變化，原因也不難理解。 一間路邊餐館可以靠翻桌率消化客單價的變動，五星級飯店卻同時背著宴會廳、早餐檯、客房餐飲、酒吧、甜點房與大量服務人力。客人只少點一杯酒、一份甜點，看起來不大；若同一種選擇在早餐、下午茶、晚餐與客房服務反覆發生，整趟住宿的餐飲收入就會被一層層削薄。 偏偏飯店又不能粗暴地把份量砍半、價格照舊。 因為高端客人買的是被理解，不是被算計。若小盤子看起來像偷工減料，省下的食材成本可能換來更大的品牌傷害。飯店必須讓客人感受到：份量變小，是因為更懂他的狀態；不是因為廚房想省錢。 這也讓 GLP-1 成為一場服務設計考試。廚師要重新安排每口的營養與味道，餐飲經理要設計小份但不廉價的價格階梯，服務人員要學會詢問偏好又不刺探病史，酒吧則要讓不喝酒的人仍能留在社交中心。 真正困難的從來不是「少放一塊蛋糕」，而是整家飯店得重新理解：客人要的豐盛，可能已經不再等於食物的重量。 ## 03｜酒吧沒有消失，只是「微醺」開始換一種賣法 飯店最難捨棄的收入之一，是酒。 一杯調酒賣的從來不是乙醇本身。它賣的是夜晚的儀式、社交的入場券，以及一張適合拍照的桌面。若客人減少飲酒，飯店不能只把威士忌拿走，再遞上一杯白開水。 邁阿密海灘 EDITION 因此增加了無酒精調飲的版面，Four Seasons Fort Lauderdale 與 Omni La Costa 等飯店，也出現更完整的無酒精飲品選項。這些飲料並沒有放棄香氣、杯器、色彩與調飲技術；少掉酒精，價格與儀式感未必跟著消失。 但「GLP-1 讓人不想喝酒」目前還不能寫成定論。 2025 年《JAMA Psychiatry》刊登的一項小型第二期隨機試驗，在 48 名有酒精使用障礙症狀的成人中，觀察到低劑量司美格魯肽降低酒精渴望、每個飲酒日的飲用量，以及部分重度飲酒結果；可是不是所有酒精指標都顯著，研究時間也只有九週。後續系統性回顧同樣認為訊號值得追蹤，但仍需要更大型試驗。 所以，飯店增加無酒精飲品可以視為需求測試，不能被寫成藥物已經「治好酒癮」。 商業上真正有意思的是：一個原本可能消失的酒水消費，被重新設計成高單價的無酒精體驗。飯店不是向少喝投降，而是在替「不喝也要有排場」找新的收費方式。 ## 04｜從小份餐，到一整套「用藥陪伴經濟」 當客人的問題從吃不吃甜點，延伸到如何補充營養、維持肌力、管理腸胃不適與旅行中的用藥習慣，飯店看到的會是一整條新的服務鏈，而非單張菜單。 美國四個營養與肥胖醫學專業組織在 2025 年發布聯合建議，強調使用 GLP-1 治療肥胖時，應重視營養密度、足夠蛋白質與阻力訓練，以降低營養缺口並協助維持瘦體組織。重點是「蛋白質加上肌力訓練」，不是喝一杯高蛋白飲就萬事大吉。 這也解釋了為什麼旅宿與度假村開始把廚房、健身房、營養師和恢復服務串在一起。Hilton Head Health 的官方方案，就把主廚餐點、蛋白質指導、肌力與恢復課程、營養師諮詢及料理工作坊放進住宿體驗。 更昂貴的版本甚至走向醫療邊緣。洛杉磯比佛利山四季酒店官網列出由第三方 Immortelle Integrative Health 提供的多種靜脈注射與「身體維持」服務；季節療程頁把身體維持點滴列為 750 美元，另一項包含靜脈療法的飛行前後健康方案則標價 1,000 美元。這證明高端旅宿正在測試新的付費項目，卻不等於這些療程已被高品質臨床試驗證實能替 GLP-1 使用者保住肌肉或皮膚。四季官網也明載，服務由第三方提供，飯店不為其療效背書。 這條界線很重要。 把專業營養、合理運動與貼心住宿整合起來，是服務創新；把每一瓶點滴都說成醫療奇蹟，就可能從奢華跨進智商稅。 這裡還有另一個容易被忽略的風險：隱私。 飯店為了提供冷藏空間、適合的餐點或運動安排，可能需要知道客人的部分需求；但「想吃小份」不等於客人願意公開病史，「需要冰箱」也不代表櫃檯可以追問用了哪一種藥。高端服務若把個人化做成盤問，反而會把最想留住的客人推走。 因此，真正成熟的設計不應要求住客先自我貼上「GLP-1 使用者」標籤。它應該讓任何人都能選擇小份餐、無酒精調飲、蛋白質較高的菜色、房內冷藏與肌力課程；需要醫療處置時，再清楚切換到有資格、有責任歸屬的專業人員。 飯店可以賣健康體驗，卻不能模糊醫療責任。這條線守得好，個人化會變成信任；守不好，再漂亮的健康套餐也可能變成法規與聲譽風險。 ## 05｜GLP-1 沒有消滅慾望，只是把慾望搬家 很多人把這個故事講成食品、甜點和酒水的末日。 但消費者未必真的少花錢，他們更可能把錢移到另一個地方。 PwC 在 2026 年針對 3,000 多名現任、停用與考慮使用 GLP-1 的受訪者調查顯示，截至同年 5 月，美國約 21% 家戶有現任使用者；61% 現任使用者表示較少購買甜食，56% 較少購買鹹零食，但新鮮蔬果、蛋白質食品、營養補充，以及服裝、健身和自我照護等類別，出現不同程度的增加。 這是美國市場的調查與消費關聯，不是每個國家、每個人都會照表演出。可它指出一個很清楚的方向：當「多吃一點」不再那麼容易被販售，企業會改賣「吃得更精準、看起來更好、感覺更受照顧」。 資金並沒有憑空蒸發，它比較像沿著三個圓圈向外移動。 1. **最裡面的一圈，是食物本身。** 大包裝、高熱量、靠衝動加購的品項先承受壓力；份量較小、蛋白質與纖維較高的產品得到新位置。雀巢早在 2024 年就推出 Vital Pursuit，明確把產品定位在 GLP-1 使用者與體重管理族群，主打小份、高蛋白與營養密度。這不保證產品一定成功，卻說明大型食品公司已經把「胃口變小」當成產品設計題，而不是短期新聞。 2. **第二圈，是身體改變後的新需求。** 衣服尺寸需要重買，肌力與體能需要維持，皮膚、睡眠、運動與社交習慣也可能跟著調整。原本花在甜食與酒精上的部分預算，便可能流向服飾、健身、營養與自我照護。 3. **最外面的一圈，才是旅宿最擅長的體驗。** 食品公司賣的是一盒小份餐，飯店卻能把同樣的小份餐放進一段完整旅程：管家先問偏好，廚房調整份量，酒吧準備無酒精搭餐，教練安排肌力課程，客人離開時再帶走一份可延續的飲食建議。 也就是說，飯店有機會把「少吃掉的量」重新翻譯成「多出來的服務」。 這是高端旅宿比一般零食公司更有利的地方。一包餅乾縮小，消費者很容易認為它變相漲價；一道料理縮小，如果背後多了客製、故事、食材與服務，客人可能願意支付更高的單位價格。 當然，這套轉換只有在體驗真的變好時才成立。若飯店只是把普通果汁換進漂亮杯子、把水煮雞胸肉改名為代謝套餐，再收三倍價格，那不是創新，只是借熱門藥物替舊商品貼金。 對五星級飯店而言，真正的考題不是冰箱裡少放兩罐可樂。 而是能不能把小份餐做出大價值，把不喝酒做成一場體面的社交，把肌力、營養與恢復變成客人願意續購的服務，同時守住醫療證據與行銷話術的邊界。 ## 對台灣業者來說，先量四件事，不要急著宣布革命 現在還不到宣布「GLP-1 飯店革命」的時候。更實際的是先盯三件事：小份餐的點購率是否上升、無酒精飲品能不能維持客單價、健康與恢復服務是否真的帶來回購，而不是只吸引一次打卡。 還可以再多看一個細節：浪費率。 若客人原本就吃不完，大盤餐點留下的不是價值，而是廚餘。小份化若能同時降低食材浪費、維持價格與提高滿意度，它才可能是一門好生意；若只是把盤子縮小、再掛上一張昂貴的「GLP-1 友善」標籤，熱潮退去後，消費者很快就會看穿。 不同飯店，答案也不會一樣。 商務飯店的早餐常已包在房價裡，食量下降不一定立刻減少收入，卻可能降低備餐與廚餘；度假飯店仰賴餐飲、酒吧與活動加購，需求改變就更直接；以宴會為主的飯店還得面對另一種難題——賓客吃得少了，婚宴與大型聚會是否仍要用滿桌菜色證明氣派？ 因此，台灣市場第一步不該急著做一張「減重針套餐」，而是把資料收好。哪些餐點最常被剩下？小份選項會不會提高點餐意願？無酒精飲品是取代原本的酒，還是帶來新的客群？健身與營養服務有多少人願意第二次購買？ 這些答案比一句「GLP-1 是下一個大趨勢」更值錢。因為真正能進財報的，從來不是趨勢本身，而是企業把趨勢翻譯成產品、價格與回購的能力。 以前的奢華，是客人一開門，桌上什麼都有。 下一個版本的奢華，可能是客人還沒開口，飯店就知道哪些東西不必端上來；剩下的每一樣，都更適合他，也更值得付錢。 GLP-1 沒有讓慾望消失。 它只是把「再來一份」，改成了「替我做得更剛好」。 ## 主要資料來源 1. [Forbes Travel Guide｜How GLP-1s Are Reshaping Luxury Hospitality](https://www.forbes.com/sites/forbestravelguide/2026/07/15/how-glp-1s-are-reshaping-luxury-hospitality/) 2. [Journal of Marketing Research｜The No-Hunger Games](https://journals.sagepub.com/doi/10.1177/00222437251412834) 3. [美國 FDA｜Wegovy 2025 年核准仿單](https://www.accessdata.fda.gov/drugsatfda_docs/label/2025/215256s023lbl.pdf) 4. [JAMA Psychiatry｜Once-Weekly Semaglutide in Adults With Alcohol Use Disorder](https://jamanetwork.com/journals/jamapsychiatry/fullarticle/2829811) 5. [The Obesity Society 等四組織｜GLP-1 治療期間的營養優先事項](https://www.obesity.org/wp-content/uploads/2025/06/Joint_Nutritional-Priorities-to-Support-GLP-1-Therapy-for-Obesity.pdf) 6. [PwC｜2026 GLP-1 Consumer Trends](https://www.pwc.com/us/en/industries/consumer-markets/library/glp-1-consumer-trends.html) 7. [Hilton Head Health｜GLP-1 Support Experience](https://www.hhhealth.com/glp-1-support-experience/) 8. [Four Seasons Los Angeles｜Immortelle](https://www.fourseasons.com/losangeles/landing-pages/property/immortelle/) 9. [Nestlé USA｜Vital Pursuit](https://www.nestleusa.com/media/pressreleases/vital-pursuit-glp-1) 事實查核截止：2026 年 8 月 31 日。 > 免責聲明：本文為產業與科普資訊整理，不構成醫療、用藥或投資建議。GLP-1 類藥物的適應症、副作用與使用方式，應由合格醫療專業人員依個別狀況評估。 --- ## Avexitide Reduced the Composite Rate of Level 2/3 Hypoglycemia Events by 55%—and Repriced Amylyx Overnight URL: https://drugnews.com.tw/articles/2026-08-30-amylyx-avexitide-pbh-phase3-en.html Date: 2026-08-30 Language: en Summary: Avexitide reduced the combined rate of Level 2 and Level 3 hypoglycemic events by 55% versus placebo in Amylyx's Phase 3 LUCIDITY trial, sending the shares up about 64% in one session. This analysis examines the GLP-1 mechanism in post-bariatric hypoglycemia, the evidence still missing from the topline release, Amylyx's $35.1 million asset acquisition, its $500.2 million financing and the regulatory and commercial gates ahead. Bariatric surgery can produce durable weight loss, yet some patients later face a disturbing paradox: eating a meal can send their blood glucose low enough to cause confusion, loss of consciousness or a seizure. Amylyx Pharmaceuticals is trying to treat that condition with avexitide, a drug that moves in the opposite direction from the GLP-1 medicines now reshaping obesity care. Rather than activating the GLP-1 receptor, avexitide blocks it. In the Phase 3 LUCIDITY trial, the combined rate of Level 2 and Level 3 hypoglycemic events was 55% lower with avexitide than with placebo. Amylyx shares rose about 64% on the day the topline result was announced. The result is important, but the boundaries are equally important. Avexitide is not approved by the U.S. Food and Drug Administration. Amylyx has released topline data, not a full peer-reviewed dataset. Absolute event rates, confidence intervals, long-term safety, manufacturing, regulatory review and commercial uptake still need to be established. ## 01 | Why Can a Successful Bariatric Procedure Make a Meal Dangerous? Understanding the disease explains why this drug moves against the direction of the mainstream GLP-1 market. Bariatric surgery can help people with severe obesity improve weight and metabolic disease. A subset of patients, however, develops post-bariatric hypoglycemia, or PBH, months or years after surgery. After a meal, nutrients can reach the small intestine unusually quickly. The gut releases a large amount of glucagon-like peptide-1, or GLP-1. Under normal conditions, GLP-1 signals the pancreas to release insulin and helps bring post-meal glucose back down. In PBH, the response is excessive. Insulin acts like a brake pushed too hard, and glucose falls rapidly. Milder episodes can cause palpitations, sweating, tremor, dizziness and weakness. In severe episodes, the brain does not receive enough glucose. A patient may become confused, lose cognitive function, lose consciousness or have a seizure. The unpredictable nature of these events can make people afraid to drive alone, care for children or even eat without another person nearby. The best-known medicines in the GLP-1 class—including semaglutide and tirzepatide—use GLP-1 signaling to reduce appetite and improve glycemic control. Avexitide is a GLP-1 receptor antagonist. Its objective is to temporarily block excessive GLP-1 signaling, reduce inappropriate insulin secretion and prevent post-meal glucose from collapsing. The same biological pathway can support opposite therapeutic strategies in different diseases. One program steps on the signal; the other releases it. Avexitide is therefore a useful counterpoint to the conventional GLP-1 investment narrative. ## 02 | Why Did a 78-Patient Phase 3 Trial Move the Market So Much? LUCIDITY was a multicenter, randomized, double-blind, placebo-controlled Phase 3 trial conducted at 21 U.S. centers. It enrolled 78 adults who had developed PBH after Roux-en-Y gastric bypass surgery. Participants were randomized three to two to receive a once-daily 90 mg subcutaneous injection of avexitide or placebo for 16 weeks. The primary endpoint was the combined rate of Level 2 and Level 3 hypoglycemic events. Level 2 hypoglycemia generally refers to glucose below 54 mg/dL, a threshold that requires prompt action. Level 3 is defined by severity rather than one glucose value: the patient is impaired enough to need help from another person. For someone living with PBH, preventing one such event can mean avoiding a fall, a traffic accident, an emergency-department visit or loss of consciousness. Amylyx reported that avexitide reduced the combined event rate by 55% versus placebo, with a p value of 0.000003. The company also said that every secondary endpoint was met, including Level 2 events documented through self-monitored blood glucose, Level 2 events detected by continuous glucose monitoring and independently adjudicated Level 3 events. Most adverse events were described as mild to moderate. Amylyx reported no serious adverse events related to avexitide treatment. The most common events were diarrhea, injection-site erythema and bruising. Neither arm showed a change in body weight over the 16-week treatment period. That last point matters because blocking GLP-1 could intuitively raise concern about weight regain. The short trial did not show a weight difference, but sixteen weeks does not answer the long-term question. An open-label extension is ongoing, and the complete dataset is expected to be presented at a medical meeting. This is a strong topline result. It is not yet a complete paper. The release did not provide each arm's absolute event rate, the distribution of treatment effects or confidence intervals. It therefore cannot show whether a small subgroup captured most of the benefit or whether another subgroup did not respond. The p value is compelling, but clinical interpretation still requires the full data. ## 03 | The More Dramatic Turnaround Belongs to Amylyx Itself Viewed in isolation, avexitide is a rare endocrine-disease asset with a positive Phase 3 trial. Viewed from 2024, it looks like a distressed-asset decision made when Amylyx had very few comfortable options. Amylyx entered the amyotrophic lateral sclerosis market with Relyvrio. In March 2024, however, the global Phase 3 PHOENIX trial failed to meet its prespecified primary and secondary endpoints. In April, the company announced that it had begun the process of voluntarily removing the U.S. and Canadian marketing authorizations. New patients could no longer start treatment, and Amylyx restructured, reducing its workforce by about 70%. This was not an ordinary pipeline setback. The company's commercial revenue source disappeared. Three months later, Amylyx acquired avexitide. Company disclosures and U.S. securities filings describe $35.1 million in cash consideration, plus specified cure costs and assumed liabilities. The acquired assets included patents, regulatory materials, contracts and certain drug materials. Amylyx also assumed obligations tied to future sales. The transaction resembles a distressed purchase not because the asset was necessarily cheap, but because both sides were under pressure. Seller Eiger BioPharmaceuticals was in bankruptcy proceedings, while Amylyx had just suffered its own commercial collapse. A Phase 3-ready asset with several completed Phase 2 studies and FDA Breakthrough Therapy designation changed hands when neither company had room for leisurely capital allocation. In retrospect, paying $35.1 million for an asset that later produced a positive Phase 3 result looks dramatic. It would still be wrong to say that Amylyx bought a commercial medicine for $35.1 million. The company subsequently funded Phase 3 development, manufacturing, regulatory work and launch preparation. If review is delayed, the label is narrow, payer access is weak or uptake disappoints, the acquisition price will have been only a small part of the total investment. ## 04 | Beyond the 64% Share-Price Move, Amylyx Immediately Raised Capital On the day the data were announced, Amylyx closed at $35.11, up 63.84% from $21.43 in the prior session. Trading volume rose from roughly 1.25 million shares to approximately 23.95 million. The company did not simply celebrate the revaluation. It went directly to the capital market. On August 18, Amylyx initially announced a proposed $350 million public offering of common stock. When the offering was priced on August 19, the company increased the size to 14.09 million shares at $35.50 per share, for expected gross proceeds of approximately $500.2 million before underwriting discounts, expenses and any exercise of the underwriters' option. Both “$350 million” and “$500 million” therefore appeared in official company materials, but they describe different points in the financing process. The former was the initial proposed size; the latter was the expected gross amount after pricing the next day. Amylyx expects to use the proceeds for avexitide prelaunch activities and manufacturing capacity, research and development, working capital and general corporate purposes. For shareholders, the financing means two things at once: Amylyx has more capital to turn a positive Phase 3 result into a regulatory filing and launch, and issuing new shares creates dilution. Biotechnology companies commonly finance after a major data catalyst. The relevant question is whether this $500.2 million creates approval, supply and an effective commercial launch—not merely a longer cash runway. ## 05 | Three Gates Still Separate the Trial From a Launch Amylyx plans to submit a New Drug Application to the FDA by the end of 2026. If approved, the company expects a commercial launch in 2027. The operative words are **plans** and **if approved**. 1. **Data completeness.** The FDA aligned with Amylyx on the primary endpoint, and avexitide has Breakthrough Therapy and Orphan Drug designations. Those mechanisms can improve the efficiency of communication; they do not guarantee approval. Regulators will still examine absolute event counts, missing data, consistency across measurement methods, longer-term safety, manufacturing and quality. 2. **The label.** LUCIDITY enrolled patients with PBH after Roux-en-Y gastric bypass. Real-world PBH also occurs after sleeve gastrectomy and other procedures. The breadth of the initial label will directly shape the treatable population. 3. **Commercial execution.** No therapy is currently FDA approved for PBH, and many patients may remain undiagnosed. That creates market opportunity and education cost. Physicians must identify the condition, patients must accept daily injections and payers must decide that preventing hypoglycemic events warrants reimbursement. Amylyx estimates that about 160,000 people in the United States have symptomatic PBH. This is a company estimate, not a confirmed pool of treated patients. Prevalence, diagnosis, price, reimbursement and persistence will jointly determine the real revenue opportunity. ## 06 | Once-Weekly Imapextide Validated a Concept, but MBX Is Not Funding Phase 2b Avexitide's advantage is development lead: it has already produced a positive Phase 3 result. Its practical weakness is once-daily subcutaneous administration. Amylyx is developing a potential successor, AMX0318. It is also a GLP-1 receptor antagonist, but the design objective is extended dosing. The program remains in preclinical work intended to support a future clinical application, which the company targets for 2027. That is lifecycle planning, not clinical efficacy evidence. MBX Biosciences designed imapextide, formerly MBX 1416, for once-weekly injection. In May 2026, the company disclosed preliminary proof-of-concept results from a small, open-label Phase 2 study. Across dose levels, mean post-meal glucose nadir increased and mean peak insulin decreased. In the same announcement, however, MBX explicitly said it would not commit further investment toward a Phase 2b trial of imapextide in PBH as it prioritized other programs. Imapextide therefore should not be presented as an active later-stage challenger. A more precise interpretation is that it provided early evidence for a less-frequent dosing concept but did not become a continuing development program. The two datasets should not be compared as equals. Avexitide's evidence comes from a randomized, double-blind, placebo-controlled, 78-patient Phase 3 trial using actual hypoglycemic events as the endpoint. Imapextide's evidence comes from a small, open-label Phase 2 exploration focused mainly on metabolic measures during a mixed-meal test. Once-weekly dosing supplied a convenience hypothesis; it did not establish better adherence or an active competitive program. Amylyx's commercial question is whether its development lead can build physician familiarity, patient experience and payer access while overcoming the adoption burden of daily injections. ## 07 | Investors Now Need to Model More Than Probability of Success The positive Phase 3 result substantially reduced avexitide's clinical risk. It did not make the valuation model simple. At least five items now matter: 1. Absolute event rates and consistency of effect in the complete Phase 3 dataset. 2. The FDA's response to the filing package and proposed label. 3. Manufacturing capacity, acceptance of daily injections and payer coverage. 4. Diagnostic education and first-year uptake if the drug is approved in 2027. 5. Whether dilution from the $500.2 million financing produces a higher risk-adjusted asset value rather than simply a higher fixed-cost base. This analysis does not use peak-sales forecasts that are unsupported by the company or another reliable source. PBH is a market without an approved therapy, which means both the opportunity and the uncertainty are large. Before price, label and diagnosis rates are known, a highly precise revenue estimate usually packages assumptions as an answer. ## Conclusion | The Real Turnaround Will Not Be Measured by One Trading Day Amylyx captures both the harshness and the possibility of biotechnology. An approved medicine can fail its confirmatory trial. A company can lose its product, revenue and most of its workforce within weeks. The same company can acquire an overlooked asset from a bankrupt seller and, two years later, regain the market's attention. Avexitide's Phase 3 result deserves recognition. A 55% reduction in the combined event rate, a very small p value, success across every secondary endpoint and a currently manageable safety profile give it a credible opportunity to become the first approved treatment for PBH. But a 64% one-day share-price move only reprices the probability of success. A durable turnaround requires an accepted NDA, a successful review, reliable manufacturing, patients willing to inject every day and payers willing to cover the medicine. In 2024, Amylyx retrieved a reverse-direction GLP-1 asset from the edge of a corporate cliff. In 2026, it showed that the asset may work. The next step is proving that it can become a business. --- ## 減重手術後，吃飯竟可能昏倒？一款新藥讓 Amylyx 暴漲 64%！ URL: https://drugnews.com.tw/articles/2026-08-30-amylyx-avexitide-pbh-phase3.html Date: 2026-08-30 Language: zh-Hant Summary: Amylyx 的 avexitide 在 LUCIDITY 三期試驗中，讓第二級與第三級低血糖事件合併發生率較安慰劑降低 55%，帶動股價單日上漲約 64%。本文拆解 PBH 的 GLP-1 機制、三期證據、3,510 萬美元資產收購、5 億美元募資，以及核准與商業化前仍須跨越的關卡。 減重手術能讓體重下降，部分患者吃完飯後卻可能因血糖急降而昏倒。Amylyx 的 avexitide 反向阻斷 GLP-1 訊號；三期試驗中，第二級與第三級低血糖事件的合併發生率較安慰劑降低 55%，消息公布當天股價上漲約 64%。但這仍是頂線資料，藥物尚未獲 FDA 核准，完整數據、長期安全性與商業化仍待驗證。 ## 01｜為什麼有人減重手術成功，吃完飯卻可能昏倒？ 先理解疾病，才知道這顆藥為何走反方向。 減重手術能幫助許多重度肥胖患者改善體重與代謝疾病，但有一部分人在手術數月或數年後，會出現「減重手術後低血糖」（post-bariatric hypoglycemia，PBH）。 患者吃完飯後，營養更快進入小腸，腸道釋放大量類升糖素胜肽-1（glucagon-like peptide-1，GLP-1）。GLP-1 原本會提醒胰臟分泌胰島素，協助餐後血糖回落；問題是 PBH 患者的反應太強，胰島素像煞車踩過頭，血糖快速墜落。 輕一點，可能是心悸、冒汗、顫抖、頭暈；嚴重時，大腦得不到足夠葡萄糖，會出現混亂、認知受損、失去意識，甚至癲癇。這些事件難以預測，患者可能不敢獨自開車、不敢照顧孩子，也不敢在沒有他人的情況下吃飯。 市場最熟悉的 semaglutide、tirzepatide 等藥物，是利用 GLP-1 訊號降低食慾、改善血糖；avexitide 則是一種 GLP-1 受體拮抗劑，目標是暫時擋住過度的 GLP-1 訊號，減少不合時宜的胰島素大量分泌，讓餐後血糖不要一路跌穿地板。 一邊踩油門，一邊鬆油門。 同一條生物路徑，在不同疾病裡，可以成為完全相反的治療方向。這就是標題裡所說的 GLP-1「對跖點」。 ## 02｜78 人的三期，為什麼能讓市場如此興奮？ LUCIDITY 是一項多中心、隨機、雙盲、安慰劑對照三期試驗，在美國 21 個中心納入 78 名接受過 Roux-en-Y 胃繞道手術、後來出現 PBH 的成年人。 受試者以 3：2 隨機分組，每天皮下注射一次 90 毫克 avexitide 或安慰劑，雙盲治療 16 週。主要終點是第二級與第三級低血糖事件的合併發生率。 第二級低血糖一般指血糖低於每分升 54 毫克，是需要立即處理的重要警訊；第三級則不以單一血糖數字定義，而是嚴重到患者需要他人協助。對 PBH 患者來說，少一次這類事件，可能就是少一次昏厥、跌倒、車禍或急診風險。 Amylyx 公布的頂線結果顯示，avexitide 組的合併事件率較安慰劑低 55%，p=0.000003。公司同時表示，所有次要終點也都達標，包括自行量血糖所記錄的第二級事件、連續血糖監測所記錄的第二級事件，以及經獨立判定的第三級事件。 安全性方面，大多數不良事件為輕度至中度，沒有與 avexitide 治療相關的嚴重不良事件；最常見的是腹瀉、注射部位紅斑與瘀青。16 週內，治療組與安慰劑組都沒有觀察到體重變化。 這最後一點很值得注意。阻斷 GLP-1，直覺上可能讓人擔心體重反彈；至少在這項 16 週試驗裡，公司沒有看到明顯體重差異。但 16 週不等於長期，開放標籤延伸研究仍在進行，完整數據也要等醫學會議發表。 所以這是一份非常強的頂線結果，還不是完整論文。 我們目前不知道兩組各自的絕對事件率、效果分布、信賴區間，也還不能判斷少數患者是否特別受益或沒有反應。p 值很漂亮，臨床判讀仍要等完整資料。 ## 03｜真正反轉的，是 Amylyx 自己的命運 如果只看 avexitide，這是一款三期成功的罕病內分泌藥；把時間拉回 2024 年，它更像一筆絕境中的資產配置。 Amylyx 曾靠 Relyvrio 進入 ALS 市場。可是 2024 年 3 月，全球三期 PHOENIX 試驗沒有達到預先設定的主要與次要終點。4 月，公司宣布啟動撤銷美國與加拿大上市許可的程序，新患者不再取得藥物；同時進行重整，員工人數減少約 70%。 那不是普通的管線失敗，而是公司收入來源被拔掉。 三個月後，Amylyx 買下 avexitide。依公司與美國證券文件，交易現金價為 3,510 萬美元，另加確定的補正成本與承接負債；資產包括專利、監管資料、合約與部分藥物材料。Amylyx 也承接未來銷售相關的權利金義務。 這筆交易之所以像「抄底」，不是因為價格一定便宜，而是賣方 Eiger 正在破產程序裡，買方 Amylyx 也剛經歷自己的商業崩塌。兩家都沒有從容揮霍的空間，卻把一個已完成多項二期試驗、獲美國食品藥物管理局（FDA）突破性治療認定、準備進入三期的資產，送進了新主人手裡。 現在回頭看，3,510 萬美元買到一個三期成功資產，帳面上極具戲劇性。 但這不能直接寫成「花 3,510 萬美元買到一款上市藥」。Amylyx 後續負擔了三期試驗、製造、監管與商業化成本；如果審查受阻、標籤過窄、支付不順或上市放量不及預期，早期買價只是總投資的一小部分。 ## 04｜64% 的股價之外，公司立刻做了另一件事 數據公布當天，Amylyx 股價收在 35.11 美元，較前一日 21.43 美元上漲 63.84%，成交量也由約 125 萬股放大至約 2,395 萬股。 公司沒有只享受掌聲，而是立刻進場募資。 8 月 18 日，Amylyx 初步宣布擬公開發行 3.5 億美元普通股；8 月 19 日正式定價時，規模上調為 1,409 萬股、每股 35.50 美元，預計總募資約 5.002 億美元，尚未計入承銷商額外認購選擇權，也還沒扣除費用。 因此，「募資 3.5 億」和「募資 5 億」都曾出現在官方文件裡，描述的卻不是同一個時點。前者是初步方案，後者是次日定價後的最終預計總額。 這筆錢預計用於 avexitide 上市前準備與增加製造產能、研發、營運資金及一般公司用途。對股東來說，它同時代表兩件事：公司有資本把三期成功往上市推，也代表新股發行帶來稀釋。 生技公司在好消息後趁高募資並不罕見。真正要看的，是這 5 億美元能否換來核准、產能與有效的商業啟動，而不是只讓現金消耗期延長。 ## 05｜距離上市，還有三道門 Amylyx 計畫在 2026 年底前，向 FDA 提交新藥申請（New Drug Application，NDA）；若獲准，公司預期 2027 年展開商業上市。 請注意這兩個動詞：計畫、若獲准。 1. **資料完整性。** FDA 已與公司就主要終點達成共識，avexitide 也有突破性治療與孤兒藥認定，這些都能提高溝通效率，卻不保證核准。監管機構仍會審查絕對事件數、遺漏資料、不同量測方式的一致性、長期安全性、製造與品質。 2. **標籤。** LUCIDITY 只納入 Roux-en-Y 胃繞道手術後的 PBH 患者；現實世界還包括袖狀胃切除等其他手術族群。最初核准範圍會多寬，將直接影響可治療人數。 3. **商業化。** PBH 目前沒有 FDA 核准療法，未被診斷的患者可能不少，這既是市場機會，也是教育成本。醫師要辨認疾病、患者要接受每天注射、支付方要相信減少低血糖事件值得給付，缺一不可。 Amylyx 估計，美國約有 16 萬名症狀性 PBH 患者。這是公司估算，不是已確認會接受治療的市場。患者人數、診斷率、價格、給付與持續用藥率，會共同決定真實營收。 ## 06｜每週一針曾驗證概念，卻沒有形成持續競爭 Avexitide 的優勢，是三期已經成功，開發進度明顯領先；弱點則是每天皮下注射一次。 Amylyx 自己已在準備下一棒 AMX0318。它同樣是 GLP-1 受體拮抗劑，但目標是長效給藥，目前仍在進行支持臨床試驗申請的前期研究，公司目標 2027 年提出試驗申請。這是生命週期布局，離證明療效還很遠。 MBX Biosciences 的 imapextide（MBX 1416）曾以每週一次注射取得初步概念驗證。2026 年 5 月公布的小型、開放標籤二期資料顯示，在不同劑量下，餐後血糖最低點平均提高，胰島素峰值平均下降。 但 MBX 在同一份公告中也明確表示，基於資源與資本配置，公司不會再投入 imapextide 的 PBH 二期後段試驗。因此，imapextide 不能再被視為正在積極追趕的後來者；較精確的定位是：它曾驗證低頻給藥概念，卻沒有形成持續開發競爭。 兩份資料也不能硬比。Avexitide 是 78 人、隨機雙盲、安慰劑對照三期，終點是實際低血糖事件；imapextide 的二期是小型、開放標籤探索，主要看混合餐試驗中的代謝指標。每週給藥曾提供一個便利性假設，卻不能被寫成已證實的依從性優勢或仍在推進的競品。Amylyx 真正要回答的，是能否用開發進度建立醫師、患者與支付基礎，同時克服每日注射的採用門檻。 ## 07｜投資人現在要算的，不只是成功機率 三期成功後，avexitide 的臨床風險大幅下降，估值模型卻沒有因此變簡單。 接下來至少要追五件事： 1. 完整三期數據中的絕對事件率與效果一致性。 2. FDA 對申請資料與標籤範圍的回應。 3. 製造產能、每日注射接受度與保險給付。 4. 2027 年若獲准後的診斷教育與首年放量。 5. 5.002 億美元募資後的股本稀釋，是否換來更高的資產淨現值，而不是更高的固定成本。 至於未經公司或可靠來源支持的分析師峰值銷售預測，這篇不採用。PBH 是沒有核准療法的市場，潛力與不確定性都很大；在價格、標籤與診斷率尚未出爐前，把峰值寫得太精確，通常只是把假設包裝成答案。 ## 結語｜真正的翻身，不在 64% 那一天 Amylyx 的故事很像生技產業最殘酷也最迷人的一面。 一款已上市藥物可以在確認性試驗後失敗，一家公司可以在幾週內失去產品、收入與大部分團隊；同一家公司也可能從另一家破產企業手中，買回一顆被低估的資產，兩年後重新取得市場信任。 Avexitide 的三期結果值得肯定。55% 的事件降低、極低的 p 值、所有次要終點達標，以及目前看來可管理的安全性，讓它有機會成為 PBH 的第一款核准療法。 但股價一天上漲 64%，只是市場把成功機率重新標價；真正的翻身，要等 NDA 被接受、審查通過、藥物順利製造、患者願意每天注射、支付方願意買單。 2024 年，Amylyx 從懸崖邊撿回一顆 GLP-1 的「反向藥」。 2026 年，它證明這顆藥可能有效。 下一步，才是證明它能成為一門生意。 ## 主要資料來源 1. [Amylyx｜LUCIDITY 三期頂線結果](https://investors.amylyx.com/news-releases/news-release-details/amylyx-pharmaceuticals-announces-positive-topline-results-0) 2. [Amylyx｜avexitide 資產收購](https://investors.amylyx.com/news-releases/news-release-details/amylyx-pharmaceuticals-announces-acquisition-phase-3-ready-glp-1) 3. [Amylyx｜Relyvrio 撤市與重整](https://investors.amylyx.com/news-releases/news-release-details/amylyx-pharmaceuticals-announces-formal-intention-remove) 4. [Amylyx｜歷史股價](https://investors.amylyx.com/stock-information/historical-price-lookup) 5. [Amylyx｜5.002 億美元定價募資](https://investors.amylyx.com/news-releases/news-release-details/amylyx-pharmaceuticals-announces-pricing-upsized-500-million) 6. [MBX Biosciences｜imapextide 二期更新](https://investors.mbxbio.com/news-releases/news-release-details/mbx-biosciences-provides-obesity-portfolio-update-including) 事實查核截止日：2026 年 8 月 30 日。 > ＊本文為醫藥產業與投資教育，不構成醫療、用藥或投資建議。Avexitide 尚未獲 FDA 核准；完整三期數據仍待醫學會議揭露。 --- ## Pancreatic Cancer Finally Gets a Broad RAS(ON) Drug: Inside RASONQUE's FDA Approval URL: https://drugnews.com.tw/articles/2026-08-29-revolution-rasonque-pancreatic-cancer-ras-on-en.html Date: 2026-08-29 Language: en Summary: The FDA approved Revolution Medicines' RASONQUE 6.5 months ahead of schedule for a defined metastatic pancreatic adenocarcinoma population. This analysis examines the RASolute 302 survival data, the RAS(ON) mechanism, Revolution's eleven-year strategic pivot and the pricing, access, safety and global execution tests that begin after approval. Pancreatic cancer is devastating not only because it is often detected late. The deeper frustration is that researchers have known for decades that RAS sits behind most pancreatic tumors, yet the protein resisted conventional drug design for more than forty years. On August 26, 2026, the U.S. Food and Drug Administration approved Revolution Medicines' RASONQUE, or daraxonrasib, 6.5 months ahead of the agency's original action date. It is the first approved broad RAS(ON) inhibitor capable of targeting several activated RAS forms, and the first approved therapy to directly address the dominant oncogenic driver in pancreatic cancer. In the 500-patient Phase 3 RASolute 302 trial, median overall survival reached 13.2 months with RASONQUE versus 6.7 months with physician's-choice chemotherapy. The hazard ratio for death was 0.40. Median progression-free survival was 7.2 versus 3.6 months, with a hazard ratio of 0.49. Both median survival measures nearly doubled. That is a major result in a disease with few effective later-line options. It is not, however, evidence that pancreatic cancer has been conquered. The approval covers adults with metastatic pancreatic adenocarcinoma who have received at least one prior systemic therapy or who are not candidates for multi-agent systemic therapy. It is not a universal treatment for early-stage disease, and median outcomes do not predict the result for an individual patient. The company story is equally instructive. Revolution began in 2015 as a natural-products and antifungal biotechnology company. Over eleven years, it replaced its original strategy, acquired a platform without human proof of concept, used stock to acquire roughly $1.1 billion in expected net cash and financed commercialization without surrendering control of the core asset. This was not a company that knew the right answer from the beginning. It was a company willing to change the answer as evidence changed. ## 01 | Why RAS Resisted Drug Development for Four Decades RAS functions like a growth switch inside the cell. Under normal conditions it cycles between active and inactive states. Certain mutations hold the switch in an active configuration, continually transmitting proliferative signals. Pancreatic ductal adenocarcinoma is especially dependent on this pathway, most commonly through KRAS G12D, G12V and G12R variants. Traditional small molecules struggled because the RAS surface is relatively smooth and lacks the deep pockets that drug designers usually exploit. RAS also operates through tightly coordinated protein interactions inside the cell. The challenge was less like fitting a key into a lock than trying to grip a polished metal sphere with no obvious handle. KRAS G12C inhibitors proved that RAS was not completely inaccessible. But G12C represents only a subset of tumors and is not a dominant mutation in pancreatic cancer. RASONQUE takes a different approach. It is an oral, noncovalent tri-complex inhibitor. Rather than relying on RAS alone to provide a binding pocket, the molecule recruits another intracellular protein to create a new interface and trap RAS in its active, or ON, state. This architecture enables inhibition across multiple mutant and wild-type RAS proteins. That design is commercially important. The U.S. label does not require a companion diagnostic selecting one rare mutation before treatment can be considered. The label is broader than a single-mutation strategy, although it remains limited to the approved metastatic pancreatic adenocarcinoma population. ## 02 | How Much Weight Should Investors Put on 13.2 Versus 6.7 Months? RASolute 302 was a global, randomized, open-label Phase 3 trial. Five hundred patients with metastatic pancreatic adenocarcinoma whose disease had progressed after first-line therapy were assigned one-to-one to RASONQUE 300 mg once daily or one of four physician-selected cytotoxic chemotherapy regimens. The study included patients with several RAS G12 mutations as well as patients without a detected tumor RAS mutation. It met its primary and key secondary endpoints in both the RAS G12 population and the full study population. The results in all randomized patients were: - Median overall survival: **13.2 months versus 6.7 months**. - Hazard ratio for death: **0.40**, corresponding to a 60% reduction in the relative risk of death over follow-up. - Median progression-free survival: **7.2 months versus 3.6 months**. - Hazard ratio for progression or death: **0.49**, corresponding to a 51% reduction in relative risk. These figures should not be collapsed into a single slogan. The medians describe the point at which half the patients in each group had experienced the event. Hazard ratios compare risk across the full follow-up period. A 0.40 hazard ratio does not mean that mortality becomes 40%, and a 6.5-month difference in medians does not mean that every patient lives exactly 6.5 months longer. The parallel improvement in progression-free and overall survival makes the evidence more persuasive than a result supported by only one endpoint. Yet individual outcomes will still depend on performance status, disease burden, prior therapy, subsequent treatment and adverse-event management. Safety is part of the value equation. Official materials warn about skin and soft-tissue toxicity, stomatitis, diarrhea, gastrointestinal perforation, interstitial lung disease or pneumonitis, and embryo-fetal toxicity. In the pancreatic adenocarcinoma study, 86% of RASONQUE-treated patients experienced skin toxicity and 10% had Grade 3 events. Oral administration can reduce infusion burden. It does not make the treatment trivial or cost-free. ## 03 | The First Strategic Decision: Tear Up the Founding Script Third Rock Ventures launched Revolution Medicines in 2015 with a $45 million Series A financing. The company's original lead story was not cancer or RAS. It centered on natural-product chemistry and antifungal medicines, including an effort to preserve the potency of amphotericin B while reducing kidney toxicity. A company's name, founding team and first financing are usually tied to its initial thesis. That makes admitting that the original program should no longer command the most resources psychologically and organizationally difficult. Revolution shifted toward oncology and difficult-to-drug targets. It first advanced an SHP2 inhibitor upstream of RAS signaling and entered a global partnership with Sanofi in 2018 that included a $50 million upfront payment. Many companies would have protected the first asset validated by a large pharmaceutical partner at almost any cost. Revolution instead defined itself as a company able to solve difficult molecular problems, not as the owner of one irreplaceable program. That distinction made the later move into direct RAS inhibition possible. ## 04 | The Second Strategic Decision: Buy Warp Drive Before Human Proof Revolution acquired Warp Drive Bio in 2018. The transaction brought tri-complex technology designed to engage activated RAS and a set of early oncology programs. At the time, direct RAS(ON) inhibition had no clinical proof of concept and looked speculative to much of the industry. Revolution's financial statements allocated $55.8 million to acquired in-process research and development and $13.6 million to tri-complex and genomic-mining technologies. Those accounting allocations should not be simplified into the claim that Revolution paid $69 million in cash for the company. The acquisition also demonstrates what platform discipline looks like. Revolution did not retain every program merely because it had been acquired. It kept the tri-complex capabilities aligned with RAS and oncology, then sold the non-core antibiotic genome-mining platform to Ginkgo Bioworks in 2019. Sanofi decided in 2022 to terminate the SHP2 collaboration, effective in 2023. By then, Revolution's next-generation RAS(ON) pipeline had already entered the clinic. The old lead program receded only after a new engine was operating. That sequencing helped prevent a partner exit from becoming a corporate cliff. ## 05 | The Third Strategic Decision: Acquire Cash, Not Another Drug Scientific conviction solves only half of late-stage biotechnology. Running broad RAS programs across tumor types, treatment lines and global Phase 3 studies requires an unusually deep capital base. In 2023, Revolution acquired EQRx in an all-stock transaction. EQRx's low-cost medicines model had faltered, and its pipeline offered limited strategic fit. The asset Revolution wanted was the balance sheet. The completed transaction was expected to add roughly $1.1 billion in net cash, in exchange for approximately 55 million newly issued Revolution shares. Existing shareholders absorbed dilution; the company bought more time to advance several RAS programs in parallel. In 2025, Revolution entered a flexible financing agreement with Royalty Pharma worth up to $2 billion. It included up to $1.25 billion in synthetic-royalty financing and up to $750 million in secured debt. That money was not free. Royalty Pharma can receive tiered royalties on RASONQUE sales for fifteen years, while borrowed amounts carry interest and repayment obligations. Revolution preserved global development and commercialization control by exchanging part of its future cash flow and accepting financial liabilities. At the end of June 2026, the company held about $3.9 billion in cash, cash equivalents and marketable securities. This gave it the capacity to launch independently rather than sell a core asset under financing pressure immediately before approval. Media reports have described acquisition interest from large pharmaceutical companies. But as of the FDA approval, Revolution had not announced a definitive sale, and no company or regulatory primary source established that its board formally rejected a $30 billion offer. A compelling rumor is not a completed capital-allocation decision. The verified conclusion is narrower: EQRx, equity issuance and Royalty Pharma financing bought Revolution the option not to sell immediately. ## 06 | Approval Starts the Commercial Exam RASONQUE is now available for prescription in the United States. Revolution disclosed a wholesale acquisition cost of $39,800 for a 30-day supply, or nearly $478,000 on a mechanical annualized basis. Wholesale acquisition cost is not a patient's out-of-pocket expense and is not the net revenue that Revolution will recognize. Rebates, payer discounts, patient assistance, channel fees, treatment duration, dose modification and discontinuation will separate list price from realized economics. The market will now follow at least six variables: 1. How quickly oncologists place oral RASONQUE into later-line care. 2. Whether payer access and patient-support programs shorten time to treatment. 3. Whether skin toxicity, stomatitis and diarrhea can be managed in routine practice. 4. Whether real-world treatment duration approaches clinical-trial experience. 5. Whether first-line pancreatic cancer, lung cancer and other development programs succeed. 6. Whether the $3.9 billion balance sheet can support an entire RAS product family without being consumed by rapid expansion. Asia adds another strategic variable. On August 10, 2026, Revolution and BeOne Medicines announced a collaboration covering daraxonrasib and three other clinical-stage RAS(ON) inhibitors in parts of Asia. Depending on the market, BeOne received exclusive development and commercialization rights or exclusive commercialization rights. Japan and South Korea are outside the licensed territory. U.S. approval increases the strategic value of those regional rights. It does not constitute approval in Asian markets. Local filings, clinical requirements, labels, pricing, reimbursement and launch execution remain separate gates. ## 07 | Revolution Bet on More Than One Drug It is tempting to rewrite the RASONQUE story as a straight line drawn by a visionary company that saw the answer a decade in advance. The record is more useful precisely because it was not linear. In 2015, Revolution was developing natural-product antifungals. In 2018, its most visible clinical strategy centered on SHP2. That same year, it acquired Warp Drive's RAS(ON) technology. In 2023, it issued a large block of shares to acquire EQRx's cash. In 2025, it exchanged a portion of future sales and accepted debt capacity to finance independent commercialization. Only in 2026 did RASONQUE become the world's first approved broad RAS(ON) targeted medicine. The company did not protect one answer through the entire journey. It protected a more durable operating principle: release sunk costs when evidence weakens, and concentrate talent, capital and time when a stronger hypothesis emerges. For patients, the 13.2-month median overall survival result matters more than the strategy lesson. For the biotechnology industry, the approval also clarifies what a platform should mean. A platform is not a slide with ten simultaneous pipelines. It is an underlying capability that can find a better problem after an initial thesis fails—and that can raise the probability of a second product after the first one succeeds. Pancreatic cancer has not been defeated. But a door blocked by RAS for more than forty years has finally opened a meaningful distance. --- ## 胰臟癌終於被撬開一道縫！FDA 提前 6.5 個月批准首款廣譜 RAS 新藥 URL: https://drugnews.com.tw/articles/2026-08-29-revolution-rasonque-pancreatic-cancer-ras-on.html Date: 2026-08-29 Language: zh-Hant Summary: FDA 提前 6.5 個月核准 Revolution Medicines 的 RASONQUE，用於特定轉移性胰臟腺癌成人。本文拆解 RASolute 302 的生存數據、RAS(ON) 技術、Revolution 十一年轉向史，以及藥證之後的定價、給付、安全性與全球商業化考驗。 ──────────────────── 胰臟癌最令人絕望的地方，不只是發現得晚。 更殘酷的是，人類早就知道多數胰臟癌背後站著 RAS，卻用了四十多年，始終很難真正抓住它。 2026 年 8 月 26 日，美國食品藥物管理局（FDA）提前 6.5 個月批准 Revolution Medicines 的 RASONQUE，通用名 daraxonrasib。這是第一款獲准、可同時瞄準多種活化態 RAS 的廣譜標靶藥，也是首款直接針對胰臟癌主要致癌驅動因素的核准療法。 在 500 人第三期試驗裡，RASONQUE 把中位總生存期從化療組的 6.7 個月拉到 13.2 個月，死亡風險降低 60%；中位無惡化生存期則從 3.6 個月延長至 7.2 個月。 兩條曲線幾乎都翻了一倍。 但「胰臟癌被攻克」仍然不是正確答案。這次核准只適用於已接受至少一次全身治療，或不適合多藥全身治療的轉移性胰臟腺癌成人；它不是早期患者的萬用藥，也不是所有人都能長期控制疾病。 真正值得寫下來的，是另一件事：一家 2015 年還在做抗黴菌藥的新創，如何在十一年裡換掉原始劇本、買下一套當時沒有人體證據的技術、用股票買回 11 億美元現金，最後把一個被稱為「不可成藥」的靶點推到市場。 這不是一路看對答案。 而是每次證據改變時，都敢重新下注。 一、RAS 為什麼難了四十年？ RAS 像細胞裡的生長開關。 正常狀態下，它會在「開」與「關」之間切換；發生特定突變後，開關可能長時間卡在啟動狀態，不斷把增殖訊號送下去。胰臟導管腺癌尤其依賴這條路徑，最常見的是 KRAS G12D、G12V 與 G12R 等變異。 問題是，RAS 的表面平滑、缺乏傳統小分子容易卡住的口袋，而且它與細胞內重要蛋白互動緊密。過去的藥物開發像想用一把普通鑰匙，去轉動一顆沒有鎖孔的金屬球。 近年率先成功的 KRAS G12C 抑制劑，證明 RAS 不是完全不能碰；但 G12C 只涵蓋部分腫瘤，在胰臟癌裡並不是主流突變。 RASONQUE 走的是另一條路。 它是一款口服、非共價的「三元複合體抑制劑」。藥物不是只靠 RAS 自己的表面，而是借用細胞內另一個蛋白，一起形成新的結合介面，抓住處於啟動狀態的 RAS。它可以同時抑制多種突變型與野生型 RAS 的訊號，因此不需要先用伴隨式診斷，挑出某一種單一突變才能開藥。 簡單說，過去是在等鎖孔出現；這套技術是把旁邊的蛋白一起拉過來，現場造出一個可以下手的位置。 二、13.2 個月對 6.7 個月，數據到底有多重？ 支持核准的 RASolute 302，是一項全球、隨機、開放標籤第三期試驗。 500 名先前接受過一線治療、疾病仍惡化的轉移性胰臟腺癌患者，以一比一分組，接受每天一次、300 毫克 RASONQUE，或由醫師選擇四種標準細胞毒性化療之一。 試驗同時納入多種 RAS G12 突變，也納入沒有檢出腫瘤 RAS 突變的患者。結果在 RAS G12 族群與全部受試者中，都達到主要與關鍵次要終點。 全部受試者的結果如下： • 中位總生存期：13.2 個月對 6.7 個月。 • 死亡風險：相較化療降低 60%，風險比 0.40。 • 中位無惡化生存期：7.2 個月對 3.6 個月。 • 疾病惡化或死亡風險：相較化療降低 51%，風險比 0.49。 這裡要分清楚兩種讀法。 13.2 個月對 6.7 個月，是兩組患者各自走到一半事件發生時的「中位數」；風險比 0.40，則是把整段追蹤期間的死亡風險一起放進比較。兩個數字回答的問題不同，不能把它們簡化成「每個人都一定多活 6.5 個月」，也不能把死亡風險降低 60% 寫成死亡率只剩 40%。 同樣地，無惡化生存期從 3.6 個月到 7.2 個月，代表延後疾病惡化的訊號和總生存期方向一致，讓結果不只靠單一終點支撐。可真正落到一位患者身上，仍會受到身體狀況、疾病負荷、先前治療、後續用藥與副作用管理影響。大型試驗告訴我們整體趨勢，不替任何個人預先寫好結局。 在胰臟癌這個五年相對存活率極低、後線選擇有限的市場，這不是邊緣改善。 可它也不是「治癒」。中位數代表一半患者高於、一半低於這個時間；完整判讀還要看不同患者的反應持續時間、後續治療與長期存活尾端。標題可以有爆點，醫療邊界不能被爆點吃掉。 安全性也要算進來。官方資料列出的警示包括皮膚與軟組織毒性、口腔炎、腹瀉、腸胃穿孔、間質性肺病或肺炎，以及胚胎與胎兒毒性。胰臟腺癌試驗中，86% 接受 RASONQUE 的患者出現皮膚毒性，其中 10% 為第三級。 口服不等於輕鬆，更不等於沒有代價。 三、第一個關鍵決定：親手撕掉創業劇本 Revolution Medicines 在 2015 年由 Third Rock Ventures 以 4,500 萬美元 A 輪資金孵化。 當時它的主角不是癌症，也不是 RAS，而是天然物化學與抗黴菌藥。最早的候選方向，來自兩性黴素 B：保留強大抗黴菌能力，同時設法降低腎臟毒性。 一家公司的名字、團隊與第一筆錢，往往都綁在原始故事上。 真正困難的不是發現新方向，而是承認舊方向不再值得占用最多資源。Revolution 後來把重心轉向腫瘤與難成藥靶點，先推進位於 RAS 訊號上游的 SHP2 抑制劑，2018 年再與賽諾菲簽下全球合作，取得 5,000 萬美元預付款。 很多公司走到這裡，會努力保護第一個被大藥廠看上的資產。 Revolution 沒有把 SHP2 當成永遠不能換掉的招牌。 它把自己定義成「能解決複雜分子問題的公司」，而不是「只能做某一顆藥的公司」。這個差別，後來決定了整家公司能不能轉向 RAS。 四、第二個關鍵決定：在沒有人體證據時，買下 Warp Drive 2018 年，Revolution 收購 Warp Drive Bio。 Warp Drive 帶來的關鍵資產，是利用三元複合體抓住活化態 RAS 的技術與早期腫瘤計畫。當時 RAS(ON) 還沒有人體概念驗證，直接抑制活化態 RAS 對許多人而言仍像科幻。 公司的財務文件把收購取得的在研資產記為 5,580 萬美元，另有 1,360 萬美元分配至三元複合體與基因組挖掘等技術；這些是會計上的資產配置，不宜直接簡化成「花 6,900 萬美元買公司」。 更重要的是，Revolution 沒有把收購進來的所有東西都留下。 它保留與 RAS 和腫瘤相關的三元複合體能力，2019 年把抗生素相關的基因組挖掘平台出售給 Ginkgo Bioworks。平台愈多，簡報看起來愈熱鬧；可公司資源有限，真正的策略通常不是「什麼都要」，而是知道什麼不再屬於自己。 後來賽諾菲在 2022 年決定終止 SHP2 合作，2023 年生效。此時 Revolution 的下一代 RAS(ON) 管線已經進入臨床。 舊主線退場時，新主線不是才開始畫圖。 這種提前換檔，才讓合作終止沒有變成公司的斷崖。 五、第三個關鍵決定：買的不是藥，是 11 億美元現金 科學方向看對，只解決一半問題。 把一款廣譜 RAS 藥推進多癌種、不同治療線別與全球第三期，需要的是一條極深的資本護城河。 2023 年，Revolution 以換股方式併購 EQRx。EQRx 原本的平價新藥模式受挫，管線對 Revolution 沒有明顯戰略價值；真正被買下的，是資產負債表。 交易完成後，Revolution 預計取得約 11 億美元淨現金，代價是發行約 5,500 萬股新股。這不是免費午餐：舊股東承受稀釋，公司則用股權換回把 RAS 管線平行推進後期的時間。 2025 年，公司又與 Royalty Pharma 簽下最高 20 億美元彈性融資，其中包括最高 12.5 億美元的未來銷售權利金安排，以及最高 7.5 億美元的公司債務。 這筆錢也不是贈與。 若 RASONQUE 開始銷售，Royalty Pharma 可在十五年內依銷售級距取得權利金；債務部分則要付息與償還。Revolution 保留全球開發與商業化控制權，交換的是未來部分現金流與財務義務。 截至 2026 年 6 月底，公司現金、約當現金與有價證券約 39 億美元。這讓它有能力自己上市，不必在核准前因為缺錢被迫出售核心資產。 市場曾出現大型藥廠洽購 Revolution 的媒體傳聞，但截至 RASONQUE 核准，公司沒有公告一筆確定收購，也沒有一級來源能證明它正式拒絕了「300 億美元報價」。這種故事很吸睛，卻不能當作已發生的董事會決策來寫。 真正可以確認的是：Revolution 用 EQRx、增資與 Royalty Pharma，為「不必立刻賣公司」買到了選擇權。 六、藥證到手，商業考試才剛開始 RASONQUE 已在美國開放處方。公司向美國證券交易委員會申報的批發採購價，是每 30 天 39,800 美元。 這個價格若機械年化，接近 47.8 萬美元；它不是患者實付金額，也不是公司最後能認列的淨收入。保險折扣、病患援助、通路費用、停藥率與實際治療時間，都會把牌價與營收拉開。 Revolution 從臨床階段公司，一夜跨進商業化階段。接下來市場會盯六件事： 1. 醫師是否迅速把口服 RASONQUE 放進二線治療。 2. 保險給付與病患援助能否縮短啟用時間。 3. 皮膚毒性、口腔炎與腹瀉能否被妥善管理。 4. 真實世界的用藥持續時間，能否接近試驗。 5. 第一線胰臟癌與肺癌等新適應症能否繼續成功。 6. 39 億美元現金能否養出整個 RAS 產品家族，而不是被快速擴張吞掉。 還有亞洲變數。 2026 年 8 月 10 日，Revolution 與百濟神州（現英文名 BeOne Medicines）簽下合作，百濟取得 daraxonrasib 等四款 RAS(ON) 抑制劑在部分亞洲市場的獨家開發與商業化權利，或依市場取得獨家商業化權利；日本與韓國不在該授權區域內。 美國核准提高了這組區域權利的戰略價值，卻不等於亞洲各市場已同步核准。後續仍要看當地申請、試驗、標籤、定價與上市執行。 七、Revolution 真正押中的，不只是一顆藥 回頭看，RASONQUE 的故事很容易被寫成「天才公司十年前就看見答案」。 事實正好相反。 2015 年，它在做抗黴菌天然物。 2018 年，最前面的資產是 SHP2。 同一年，它才把 Warp Drive 的 RAS(ON) 技術買進來。 2023 年，它用大量新股換取 EQRx 的現金。 2025 年，它再用未來部分銷售與債務，換取自主商業化的資本。 2026 年，RASONQUE 才成為全球首款廣譜 RAS 標靶藥。 這家公司沒有一路守著同一個答案。 它守住的是另一件事：舊證據失效時，願意放下沉沒成本；新證據足夠強時，敢把人才、資本與時間集中上去。 對患者而言，最重要的當然是 13.2 個月，而不是公司策略課。 對生技產業而言，這次核准則留下另一個答案：所謂平台，不能只在簡報裡同時擺十條管線。真正的平台，是第一個假說倒下後，底層能力還能找到更值得解的問題；第一款藥成功後，又能讓第二款藥的成功機率提高。 胰臟癌還沒有被打敗。 但那扇被 RAS 堵了四十多年的門，終於被推開一道真實的縫。 而 Revolution 最值得記住的，也許不是它一次押中 RAS。 是它在答案仍不清楚的十年裡，一次又一次，願意改掉昨天的自己。 — 主要資料來源：美國 FDA RASONQUE 核准公告與新聞稿；Revolution Medicines RASONQUE 核准公告、2026 年第二季財報、SEC 8-K／10-Q；Third Rock Ventures 2015 年創立公告；Revolution Medicines 收購 Warp Drive Bio、出售基因組挖掘平台、收購 EQRx、與 Royalty Pharma 融資及與 BeOne Medicines 合作公告。 ＊本文為醫藥產業與投資教育，不構成醫療、用藥或投資建議。RASONQUE 的美國核准範圍為特定轉移性胰臟腺癌成人；其他國家、其他癌種與更早治療線別仍須各自取得監管核准。 --- ## A New Rival for PharmaEssentia? What Zealand Actually Sold for $100 Million URL: https://drugnews.com.tw/articles/2026-08-28-zealand-rusfertide-royalty-pharma-pv-en.html Date: 2026-08-28 Language: en Summary: Zealand Pharma sold rusfertide-related milestones and a 1% global net-sales royalty for up to $100 million, not control of the drug itself. The transaction clarifies how Zealand, Protagonist and Takeda hold very different economics—and what rusfertide could mean for PharmaEssentia's BESREMi. With rusfertide approaching the final stretch of FDA review, Zealand Pharma has already signed a transaction worth up to $100 million, selling selected economics tied to the drug to Royalty Pharma. The transaction closed on August 12, 2026. Zealand received $50 million at closing and is due another $50 million on the first anniversary. But Zealand did not sell the drug itself. What changed hands were regulatory and sales milestones inherited from an earlier collaboration, plus a 1% royalty on global net sales. Takeda—not Zealand—holds the exclusive worldwide rights to develop and commercialize rusfertide. The more useful questions are therefore different: Why would a buyer pay up to $100 million for a 1% stream that is still waiting for an FDA decision and has not begun generating sales? And if rusfertide is approved, how could it affect Taiwan-listed PharmaEssentia and BESREMi? ## 01 | The $100 Million Is Not a Drug Price Zealand transferred two sets of rights: regulatory and commercial milestones tied to rusfertide, and a 1% royalty on global net sales. The company retained a narrow slice of upside. If annual worldwide net sales exceed $1.5 billion, Zealand will continue to receive 0.25% on the amount above that threshold, while Royalty Pharma receives 0.75%. That $1.5 billion figure is a contractual breakpoint. It is not Zealand's peak-sales guidance, and it should not be presented as evidence that management privately expects at least $1.5 billion in sales. The transaction is fundamentally an exchange of time and risk. Zealand monetized exposure to FDA review, launch pricing, reimbursement, uptake and competitive risk before those uncertainties were resolved. Royalty Pharma, supported by a portfolio spanning more than 35 marketed products and development-stage assets, can absorb single-asset uncertainty within a much broader pool. Royalty Pharma does not need to build a rusfertide sales force. If Takeda commercializes the drug successfully, product sales can convert directly into royalty income. For Zealand, this was a non-core asset over which it had no operating control. At the end of June, the company held DKK 14.457 billion in cash, cash equivalents and marketable securities. The deal was not emergency financing; it allowed Zealand to concentrate capital on its obesity and metabolic pipeline, including petrelintide, which is approaching Phase 3, and the late-stage asset survodutide. ## 02 | One Drug, Three Very Different Economic Positions Rusfertide's economics cannot be understood by looking at Zealand alone. Protagonist Therapeutics advanced the drug into Phase 3 and entered a global collaboration with Takeda in 2024. On April 28, 2026, Protagonist formally exited the 50:50 U.S. profit-and-loss sharing arrangement, giving Takeda exclusive worldwide development and commercialization rights. The opt-out triggered an immediate $200 million payment. If the FDA approves rusfertide, Protagonist can receive another $200 million opt-out payment and a $75 million milestone. It is also eligible for royalties ranging from 14% to 29% on annual worldwide net sales; at $1.5 billion in sales, the weighted-average rate would be about 21%. The three positions are therefore distinct: - **Takeda** bears the post-approval launch, pricing and global commercialization responsibilities. - **Protagonist** exchanged U.S. profit-and-loss exposure for substantial payments and a large global royalty stream. - **Royalty Pharma** acquired Zealand's much smaller 1% tail interest and can wait for long-term cash flow within a diversified portfolio. Both Protagonist and Zealand monetized future value, but they sold very different bundles of rights. The $100 million Zealand transaction cannot be compared directly with Protagonist's much larger payments because the scope, royalty rate and original contribution are not equivalent. ## 03 | Strong Clinical Data Still Leave Regulatory and Commercial Gates Rusfertide is a once-weekly subcutaneous hepcidin mimetic. By restricting iron availability for erythropoiesis, it is designed to reduce excessive red-blood-cell production, stabilize hematocrit and lower the need for phlebotomy. The Phase 3 VERIFY trial enrolled 293 patients with polycythemia vera, or PV, who remained phlebotomy-dependent because their hematocrit was inadequately controlled despite current treatment. Between weeks 20 and 32, 76.9% of patients receiving rusfertide plus standard therapy achieved a clinical response defined by the absence of a phlebotomy indication, compared with 32.9% on placebo plus standard therapy. From baseline through week 32, the mean number of phlebotomies was 0.5 versus 1.8, while 62.6% versus 14.4% of patients maintained hematocrit below 45%. At 52 weeks, the most common adverse events were injection-site reactions in 47.4% of patients, anemia in 25.6% and fatigue in 19.6%. Most were Grade 1 or 2, while serious adverse events occurred in 8.1%. The FDA granted priority review. Protagonist's formal filings described the PDUFA timing as August 2026, while Takeda referred to the third quarter. As of the article's August 15 fact-check cutoff, neither the companies nor the FDA had publicly disclosed a precise calendar date or announced an approval. Clinical risk has been reduced substantially, but regulatory risk has not disappeared. Even after a potential approval, pricing, reimbursement, treatment sequencing and patient acceptance of weekly injections will shape commercial performance. ## 04 | What Rusfertide Could Mean for PharmaEssentia PharmaEssentia's BESREMi is FDA-approved for adults with PV and received approval for a prefilled injection pen in June 2026. It is a long-acting interferon with a broad adult label and a growing body of long-term hematologic and molecular-response data. The FDA label states that the mechanism by which interferon alfa treats PV is not fully understood. Rusfertide offers a different value proposition: rapidly stabilizing hematocrit and reducing phlebotomy by directly restricting the iron available for red-cell production. In the near term, competition is most likely among patients who need treatment escalation. If physicians consider reducing phlebotomy the most urgent objective, rusfertide could compete for new prescriptions and payer budgets. That does not mean BESREMi would be displaced directly. VERIFY allowed rusfertide to be layered on top of existing care, including phlebotomy, hydroxyurea, interferon and ruxolitinib. It was not a head-to-head trial against BESREMi. For patients already receiving interferon but still struggling with hematocrit control, rusfertide could become an add-on rather than a substitute. Four questions matter most for PharmaEssentia investors: 1. Will the final FDA label limit rusfertide to phlebotomy-dependent patients with inadequate control? 2. Will guidelines and payers position it as first-line, later-line or add-on therapy? 3. In real-world use, will weekly injections deliver enough phlebotomy reduction and persistence to sustain adoption? 4. Can BESREMi turn its adult PV label, long-term response data and newly approved pen into deeper market penetration? There is also a strategic offset. PharmaEssentia's supplemental biologics application to expand BESREMi into essential thrombocythemia, or ET, has an FDA target date of August 30, 2026. If successful, the company's growth would no longer depend on PV alone. ## Investment Takeaway | Follow the Cash-Flow Rights, Not the Headline Zealand's choice was financially rational. It converted a non-core stream with limited operating control into up to $100 million: $50 million at closing, another $50 million one year later, plus a retained 0.25% on annual sales above $1.5 billion. Royalty Pharma is buying a long-duration cash-flow claim that can sit inside a diversified portfolio. It is not simply making a binary bet on an unresolved regulatory event. For PharmaEssentia, rusfertide is a real competitive entrant, but not a one-line story in which a new launch automatically replaces BESREMi. The two products differ in mechanism, treatment objective and likely place in therapy. The commercial outcome will depend on the final FDA label, guidelines, reimbursement and prescribing sequence. Two catalysts were still ahead at the August 15 fact-check cutoff: the FDA decision on rusfertide and the potential ET label expansion for BESREMi. One will determine whether a new mechanism enters the PV market. The other could determine whether PharmaEssentia expands its growth platform beyond PV. --- ## 藥華藥迎新對手？1 億美元賣分潤 URL: https://drugnews.com.tw/articles/2026-08-28-zealand-rusfertide-royalty-pharma-pv.html Date: 2026-08-28 Language: zh-Hant Summary: Zealand Pharma 以總對價 1 億美元出售 rusfertide 的里程碑與 1% 全球淨銷售權利金，交易不是出售藥物控制權。本文拆解 Zealand、Protagonist 與武田的不同經濟位置，並分析 rusfertide 對藥華藥 BESREMi 的競爭、互補與治療分層。 FDA 審查只差最後一段，Zealand Pharma 卻先簽下總對價 1 億美元的交易，把 rusfertide 的經濟權益賣給 Royalty Pharma。 交易已在 2026 年 8 月 12 日完成交割：5,000 萬美元立即支付，另外 5,000 萬美元在交割滿一周年時支付。 但這不是「賣掉一款新藥」。 Zealand 交出的是歷史合作留下的監管與銷售里程碑，以及全球淨銷售額 1% 的權利金；武田才是 rusfertide 的全球獨家開發與商業化權利人。 投資人真正該問的，是一條等待 FDA 決定、尚未開始銷售的 1% 現金流，為什麼買方願意給出總對價 1 億美元？以及 rusfertide 若核准，會怎麼影響台股藥華藥（6446）的 BESREMi？ ## 一、1 億美元不是藥價，是未來現金流的折現 Zealand 交出的權利包括兩部分：rusfertide 的監管與商業里程碑，以及全球淨銷售額 1% 的權利金。 公司仍留下一小段上行空間：如果年度全球淨銷售額超過 15 億美元，超出部分 Zealand 可繼續拿 0.25%，Royalty Pharma 拿 0.75%。 這個 15 億美元是合約門檻，不是 Zealand 對峰值銷售的公司指引，更不能拿來推論管理層「內心至少看 15 億美元」。 交易的本質，是時間與風險交換。 Zealand 把 FDA 審查、上市定價、給付、放量速度與競品風險提前折現；Royalty Pharma 則依靠 35 款以上已上市產品與開發中資產組合，承接單一藥物的不確定性。 對 Royalty Pharma 而言，它不需要自己養銷售團隊。只要武田把藥賣出去，銷售額就能轉成權利金收入。 對 Zealand 而言，這是一項非核心、也沒有營運控制權的資產。公司截至 6 月底握有 144.57 億丹麥克朗現金、約當現金與有價證券，並不缺救命錢；它選擇把資本放回肥胖與代謝主線，包括即將進入第三期的 petrelintide 與後期資產 survodutide。 ## 二、一款藥，三家公司拿的是三種不同報酬 rusfertide 的經濟鏈不能只看 Zealand。 Protagonist 將產品推進到第三期，2024 年再與武田簽下全球合作。2026 年 4 月 28 日，Protagonist 已正式退出美國 50:50 損益分攤，武田因此取得全球獨家開發與商業化權利。 Protagonist 行使退出權後，立即觸發 2 億美元付款；若 FDA 核准，還可再拿 2 億美元退出款與 7,500 萬美元里程碑。後續全球年淨銷售額權利金為 14% 至 29%，15 億美元銷售情境的加權平均約 21%。 三方定位因此很清楚： - **武田**承擔核准後的上市、定價與全球商業化執行。 - **Protagonist** 把美國損益風險換成高額付款與厚實的全球權利金。 - **Royalty Pharma** 買下 Zealand 的 1% 尾端權益，以組合資金等待長期現金流。 同樣是「把未來變現」，Protagonist 與 Zealand 的籌碼厚度完全不同。1 億美元不能直接拿去和 Protagonist 的數億美元付款比較，因為兩者出售的權利範圍、權利金比例與原始貢獻都不一樣。 ## 三、臨床數據夠強，但 FDA 與商業化仍是兩道門 rusfertide 是每週皮下注射的鐵調素模擬胜肽，透過限制鐵供應來降低紅血球過量生成，目標是穩定血比容並減少放血。 第三期 VERIFY 納入 293 名即使接受現有治療，仍因血比容控制不佳而依賴放血的真性紅血球增多症（PV）患者。 第 20 至 32 週，rusfertide 加標準治療組有 76.9% 達到「沒有放血適應症」的臨床反應；安慰劑加標準治療組為 32.9%。0 至 32 週平均放血次數為 0.5 次對 1.8 次，維持血比容低於 45% 的比例為 62.6% 對 14.4%。 52 週最常見不良事件為注射部位反應 47.4%、貧血 25.6% 與疲倦 19.6%，多為 1 至 2 級；嚴重不良事件發生率為 8.1%。 FDA 已給予優先審查。Protagonist 正式文件把 FDA 審查目標日（PDUFA）寫在 2026 年 8 月，武田對外寫第三季；截至 8 月 15 日，沒有公司或 FDA 一級來源公開精確日曆日，也沒有核准公告。 因此，臨床風險已大幅下降，監管風險仍未歸零。即使獲准，商業化還要回答價格、給付、處方排序與每週注射接受度。 ## 四、對藥華藥的影響：競爭、互補、分層同時存在 藥華藥 BESREMi 已獲 FDA 核准治療成人 PV，2026 年 6 月再取得預充填注射筆核准。它是長效干擾素，擁有廣泛成人標籤，也累積了長期血液學與分子反應資料；FDA 標籤明載，干擾素治療 PV 的作用機轉仍未完全闡明。 rusfertide 的賣點不同：快速穩定血比容、減少放血，直接處理紅血球過量生成。 短期看，競爭最可能出現在需要升級治療的患者。若醫師認為「先減少放血」更迫切，rusfertide 可能分走新增處方與保險預算。 但這不等於 BESREMi 會被直接替代。 VERIFY 允許 rusfertide 疊加在放血、羥基脲、干擾素或 ruxolitinib 等現有標準治療上，並不是把它和 BESREMi 做頭對頭。對已使用干擾素、血比容仍控制不佳的患者，rusfertide 反而可能成為加成治療。 對 6446 的真正觀察點有四個： 1. FDA 最終標籤是否限制在放血依賴、控制不佳的族群。 2. 指引與支付方把 rusfertide 放在首線、後線，還是加成位置。 3. 真實世界中，每週注射能否換來足夠明顯的放血減量與持續用藥。 4. BESREMi 能否把成人 PV 標籤、長期血球控制資料與 2026 年新核准的注射筆便利性，轉成更高滲透率。 還有一個對沖變數：BESREMi 擴張至原發性血小板增多症（ET）的補充生物製劑申請，FDA 目標日是 2026 年 8 月 30 日。若成功，藥華藥的成長來源將不再只押 PV。 ## 投資結論｜別急著判斷誰賣飛，先看誰掌握現金流 Zealand 的選擇很理性：用總對價 1 億美元把非核心、低控制權的未來現金流提前定價，其中 5,000 萬美元在交割支付、另 5,000 萬美元一年後支付，再保留年銷售超過 15 億美元部分的 0.25% 上行權利。 Royalty Pharma 看中的，是能放進多資產組合、長期收取的權利金，不是押一場尚未落幕的核准事件。 對藥華藥而言，rusfertide 是真競爭，但不是「一上市就取代 BESREMi」的單線劇本。兩者機制、使用目標與治療位置不同；最後的商業勝負，會由 FDA 標籤、指引、支付與醫師處方順序共同決定。 最值得追的兩個催化，恰好都落在 8 月：rusfertide 的 FDA 決定，以及 BESREMi 的 ET 標籤擴張。 一個決定 PV 市場會不會加入新機制；另一個決定藥華藥能不能把成長版圖往 PV 之外推進。 --- ## DNA Is Not the Nvidia of AI Drug Discovery, At Least Not Yet URL: https://drugnews.com.tw/articles/2026-08-27-twist-dna-ai-drug-discovery-shovels-en.html Date: 2026-08-27 Language: en Summary: Twist Bioscience is seeing rapid growth in AI-enabled drug-discovery orders and DSPS revenue, but demand growth, unit value and industry moat are not the same thing. The scarce asset may be the full wet-lab system that converts massive sequence volumes into proteins, functional data and model feedback. DNA synthesis has suddenly been elevated to the throne of “the Nvidia of AI drug discovery.” Twist Bioscience’s latest results are undeniably strong. Fiscal third-quarter 2026 revenue reached $118.4 million, up 23% year over year. DNA Synthesis and Protein Solutions, or DSPS, generated $56.6 million, up 39%. Revenue from therapeutics customers, a customer-industry classification rather than a separate product segment, reached $40.4 million, up 49%. The company also said orders tied to AI-enabled drug discovery are expected to grow at a triple-digit percentage in fiscal 2026 versus fiscal 2025, with another triple-digit increase visible for fiscal 2027. That wording matters. Twist is describing order growth, not revenue already recognized in its financial statements. Likewise, the $40.4 million therapeutics figure groups customers by industry; it is neither a stand-alone product segment nor a disclosure of AI-specific revenue. Treating either number as a pure AI revenue line would overstate what the company has actually reported. The demand cycle is real. But orders are not revenue, and revenue growth does not automatically become monopoly power. The market is collapsing three different propositions into one: more demand, greater value per unit and a supplier moat that is hard to replace. Nvidia became formidable because all three were true at the same time. DNA synthesis has so far proved only the first. ## 01 | Faster AI Design Makes the Wet Lab Busier Twist’s strongest evidence lies in actual orders and workflow. AI is channeling a flood of digital sequences into physical laboratories. Models generate vast numbers of DNA or protein sequences. Selected molecules must then be synthesized, expressed as proteins and tested for binding, stability, developability and function. The resulting experimental data are fed back into the model, starting the next design cycle. That is the Design-Build-Test-Learn loop. Twist describes customers moving through three stages: building the biological data needed to train models, repeatedly turning the design-build-test crank, and then iterating the model with new data while expanding into formats such as IgG and bispecific antibodies. The company’s own description is direct: an AI-ready wet lab at scale. AI has not eliminated experiments. It is turning experimentation into a higher-throughput industrial system. ## 02 | One Million Sequences Do Not Equal One Million Equivalent Orders To understand the opportunity, it helps to set aside the catchy “picks and shovels” metaphor and look at three variables: - **N: the number of candidates generated by AI** - **P: the proportion that enters physical validation** - **V: the economic value captured by each experiment** The wet-lab opportunity can be framed roughly as `N × P × V`. This is an analytical framework, not company guidance, and the variables interact with one another. N is almost certain to expand rapidly. Yet as models improve at eliminating low-probability candidates in silico, P could fall. Automation, miniaturization and platform improvements may also reduce the price or cost of each experiment, so V does not necessarily rise with N. Twist’s own engineering data put this tension in plain view. Across a company dataset of roughly one million oligonucleotides produced at femtomole scale, from the first half of 2023 through fiscal 2026’s second quarter: - Manufacturing cost fell by about 60%. - Chemical waste declined from 51.4 liters to 14.4 liters, a reduction of roughly 70%. - Turnaround time fell from 26 hours to 7 hours, a reduction of roughly 73%. - Capacity increased by about fourfold. These figures show a platform becoming faster and less expensive to operate. They do not mean selling prices fell by the same amount. But they do underscore an important point: sequence volume, revenue and material consumption will not remain linearly linked forever. For investors, this is the central unit-economics question. Higher sequence volume can expand platform utilization while lower cost per run changes the value captured from each order. The investment case therefore depends not only on how much work enters the factory, but also on what customers buy around that work and how much of the resulting productivity the supplier can retain. When a platform completes more sequences with less time, reagent use and waste, the value may accrue first to the platform and its workflow rather than to each gram of input material. ## 03 | Twist Itself Is Already Reaching Beyond DNA Revenue If DNA synthesis alone captured the value of AI drug discovery, Twist would have the strongest incentive to say so. Instead, its third-quarter presentation maps a much broader opportunity. The company estimates that by 2030, the serviceable available market for DNA Synthesis and Protein Solutions will exceed $7 billion, within a total addressable market of roughly $18 billion. Its chart places the incremental serviceable market from AI-enabled drug discovery in antibody discovery services and protein expression, at roughly $500 million each. It does not add a separate AI increment to the DNA synthesis column. That does not mean DNA will miss the upside. It means the share of wallet created by AI is extending into protein expression, characterization, functional testing and standardized data that can be fed back into models. Twist’s product roadmap is moving downstream as well, with IgG characterization, yeast display, CHO expansion and high-throughput bispecific antibodies. The prize is a continuous stream of orders spanning genes, proteins and data. That is why asking only whether a company can synthesize DNA is no longer enough. ## 04 | If This Is Not Nvidia, Where Could the Moat Be? Nvidia’s GPUs combined scarce compute, high value per system and a software ecosystem. The life-science tool chain is longer and more fragmented, and automation can continue to push down unit costs at many points. Investors should therefore look beyond sequence counts. Four indicators are harder to fake: 1. **Delivery speed:** Can complex sequences be delivered predictably, rather than through a single record-setting order? 2. **Attach rate:** After buying DNA, how many customers also purchase protein expression, characterization and functional testing? 3. **Data standardization:** Can each experimental round be compared under consistent conditions and become a high-quality label for the model? 4. **Repeat orders:** Does the customer’s design-experiment-learning wheel keep turning, embedding the supplier in routine R&D? If all four improve together, a platform has a chance to move from contract manufacturing toward an R&D operating system. One quarter of rapid revenue growth is not enough to prove that degree of stickiness. ## Final Take DNA matters. As AI moves deeper into antibody, protein and nucleic-acid drug design, demand for complex DNA and high-throughput synthesis has room to grow. Twist’s third-quarter results have moved that thesis from narrative into orders. But DNA is only the first doorway through which a digital molecule enters the physical world. Beyond it lie protein expression, binding, stability, functional testing and model feedback. Value can be redistributed at every step. The shovel rush may continue. The crown should wait. The scarce asset in AI drug discovery may not be a single DNA sequence. It may be the entire experimental factory that can reliably turn massive sequence volumes into proteins and functional data, then send those data back into the model. --- ## DNA 不是 AI 製藥的輝達，至少現在還不是 URL: https://drugnews.com.tw/articles/2026-08-27-twist-dna-ai-drug-discovery-shovels.html Date: 2026-08-27 Language: zh-Hant Summary: Twist Bioscience 的 AI 輔助藥物發現訂單與 DSPS 營收快速成長，但需求增加、單位價值與產業壁壘不是同一件事。真正稀缺的，可能不是單一 DNA 合成步驟，而是能把海量序列轉成蛋白質、功能數據並回饋模型的整座濕實驗工廠。 DNA 合成，忽然被抬上了「AI 製藥的輝達」神壇。 Twist Bioscience 的財報確實亮眼。會計年度 2026 第三季營收 1.184 億美元，年增 23%；DNA 合成與蛋白質解決方案（DSPS）營收 5,660 萬美元，年增 39%。按客戶產業分類的治療領域營收則達 4,040 萬美元，年增 49%。 公司還表示，AI 輔助藥物發現的訂單，2026 會計年度對 2025 年預計出現三位數百分比成長；對 2027 年也看得到同等量級的訂單成長。 景氣來了，這點不必否認。但訂單不是營收，營收成長也不會自動長成壟斷。市場正把三個層次疊在一起：需求增加、單位價值上升、供應商擁有難以取代的壁壘。 輝達之所以可怕，是三件事曾在同一時間成立。DNA 合成才剛證明第一件。 ## 01｜AI 設計得越快，濕實驗室反而越忙 Twist 給市場最有力的證據，落在實際訂單與工作流：AI 正把數位世界的序列洪水，引進真實實驗室。 模型先生成數不清的 DNA 或蛋白質序列；候選分子被合成、表現成蛋白質，再接受結合力、穩定性、可開發性與功能性測試。實驗數據回灌模型，開始下一輪設計。 這就是 Design–Build–Test–Learn 的循環。Twist 把客戶分成三個階段：先建立模型所需的生物資料；然後持續轉動設計、構建、測試；最後用新數據迭代模型，並擴充 IgG、雙特異性抗體等新類型。 公司對自己的定位也很直接：大規模、為 AI 準備好的濕實驗室。 AI 沒有消滅實驗。它把實驗變成更高吞吐量的工業系統。 ## 02｜一百萬條序列，不等於一百萬份等價訂單 要看懂這波需求，可以先放下「鏟子」這個太順口的比喻，用三個變數重算一次： - **N：AI 生成的候選數量** - **P：進入物理驗證的比例** - **V：每次實驗能捕捉的經濟價值** 濕實驗的機會大致取決於 `N × P × V`。這是分析框架，不是公司財測，三個變數也會彼此影響。 N 幾乎肯定快速放大。可是模型越懂得在電腦裡淘汰低機率候選，P 就有可能下降；自動化、微量化與平台改進則會壓低單次實驗的價格或成本，V 不一定跟著 N 上升。 Twist 自己的工程數據，已經把這個張力擺到桌上。 在約 100 萬條寡核苷酸（oligo）、飛莫耳（femtomole）規模的公司數據中，2023 上半年到 2026 會計年度第二季： - 製造成本約下降 60% - 化學廢棄物從 51.4 公升降到 14.4 公升，約少 70% - 作業時間從 26 小時縮短到 7 小時，約少 73% - 產能約提高 4 倍 這些數字證明平台正在加速、降本，不能被解讀成售價也下降同樣幅度。但它們至少提醒一件事：序列數量、營收與原料消耗，不會永遠線性同步。 當一座平台用更少時間、試劑與廢棄物完成更多條序列，最先變貴的可能是平台與工作流，不是每一克原料。 ## 03｜Twist 自己，早就不只想收 DNA 的錢 若 DNA 合成足以獨佔 AI 製藥的價值，最應該大聲宣告的就是 Twist。可是它在第三季簡報裡，畫出的地圖更寬。 公司估計，2030 年 DNA 合成與蛋白質解決方案的可服務市場超過 70 億美元，總可觸及市場約 180 億美元。圖表把 AI 輔助藥物發現的新增可服務市場，分別畫在「抗體發現服務」與「蛋白質表現」，各約 5 億美元；DNA 合成欄並沒有另畫一塊 AI 增量。 這不是說 DNA 沒有受惠。它說明的是，AI 帶來的錢包份額，正往蛋白質、特性分析、功能測試與可回饋模型的標準化數據延伸。 Twist 的下一步產品地圖也往下游走：IgG 特性分析、酵母展示、CHO 擴充、高通量雙特異性抗體。它要爭取的，是從 Gene、Protein 到 Data 的連續訂單。 這也是為什麼單看「會不會做 DNA」，已經不夠了。 ## 04｜如果不是輝達，壁壘藏在哪裡？ 輝達的 GPU 同時卡住稀缺算力、高單機價值與軟體生態。生命科學工具鏈更長、節點更散，也更容易被自動化持續壓低單位成本。 所以，投資人別只追序列數量。以下四個指標更難作假： 1. **交付速度：**複雜序列能否可預測地交付，重點不在一筆最快紀錄。 2. **附加率：**客戶買 DNA 之後，有多少同時買蛋白質表現、特性分析與功能測試。 3. **數據標準化：**每輪實驗能否在一致條件下比較，直接變成模型的高品質標籤。 4. **重複訂單：**客戶的設計—實驗—學習輪子能否持續轉動，讓供應商嵌入日常研發流程。 這四項若一起改善，平台才有機會從代工商變成研發系統。單單一季營收快速成長，還算不出這層黏著度。 ## 寫在最後 DNA 當然重要。AI 越深入抗體、蛋白質與核酸藥物設計，複雜 DNA 與高通量合成的需求就越有機會成長。Twist 第三季財報，已經讓這條線索從故事走進訂單。 但 DNA 只是數位分子踏入現實世界的第一道門。門後還有蛋白質表現、結合力、穩定性、功能性與模型回饋，每一站都可能重新分配價值。 鏟子熱可以燒，皇冠先別戴。 AI 製藥最稀缺的，未必是一條 DNA；更可能是能把海量序列穩定變成蛋白質、功能數據，再送回模型的整座實驗工廠。 --- ## Merck’s Oral PCSK9 Drug Enters an Injectable Market: Who Really Loses? URL: https://drugnews.com.tw/articles/2026-08-26-merck-lipfendra-oral-pcsk9-en.html Date: 2026-08-26 Language: en Summary: Lipfendra is the first FDA-approved oral PCSK9 inhibitor, lowering LDL-C by 56% and 59% in two Phase 3 trials. Oral dosing opens a major market, but cardiovascular outcomes, fasting adherence, payer access and Taiwan approval will decide whether it expands the category or displaces injections. For the past decade, patients seeking a large LDL-cholesterol reduction through PCSK9 inhibition have generally needed an injection. Merck has now turned that therapeutic route into a once-daily pill. On July 15, 2026, the U.S. Food and Drug Administration approved Lipfendra, or enlicitide. It is the first FDA-approved oral PCSK9 inhibitor for adults with hypercholesterolemia, including heterozygous familial hypercholesterolemia, as an adjunct to diet and exercise. Across two pivotal Phase 3 trials enrolling 3,207 patients, placebo-adjusted LDL-C reductions at Week 24 were 56% and 59%. Those numbers invite an obvious commercial question: can a pill take the market away from injectable PCSK9 therapies? Not so fast. Lipfendra has crossed an important threshold, but the real contest will depend on cardiovascular outcomes, long-term adherence, physician behavior and payer access. The doorway has changed; the competition behind it is only beginning. ## 1 | What Did the First Oral PCSK9 Inhibitor Actually Change? PCSK9 acts like a disposal tag for hepatic LDL receptors. When PCSK9 binds the receptor, the complex is more likely to be internalized and degraded. Fewer receptors remain on the liver-cell surface to remove LDL particles from circulation, and blood LDL-C rises. Existing approaches intervene in different ways. Amgen’s Repatha and Regeneron/Sanofi’s Praluent are antibodies that bind circulating PCSK9. Novartis’ Leqvio uses small interfering RNA to reduce PCSK9 production in the liver. All require injection. Lipfendra is a macrocyclic peptide. Its constrained structure blocks the interaction between PCSK9 and the LDL receptor, allowing more receptors to remain on the hepatocyte surface and continue clearing LDL-C. The FDA label keeps the two pivotal populations separate: - CORALreef Lipids enrolled 2,904 adults with atherosclerotic cardiovascular disease or high cardiovascular risk. At Week 24, the placebo-adjusted mean LDL-C reduction was 56% (95% CI, 51% to 61% lower). - CORALreef HeFH enrolled 303 patients with heterozygous familial hypercholesterolemia. The placebo-adjusted mean reduction at Week 24 was 59% (95% CI, 53% to 66% lower). These results came from different populations. They should not be merged into a universal “roughly 60%” claim, and they should not be used to rank products across separate trials. Lipfendra is also not a frictionless pill. The label specifies one 20 mg tablet every morning on an empty stomach, taken with water, black coffee or plain tea, followed by at least 30 minutes before food or other beverages. That routine may be easier to accept than an injection, but patients must remember it every day. Safety requires the same nuance. Overall adverse-reaction frequency in the high-risk hypercholesterolemia trial was similar to placebo. In the HeFH study, diarrhea occurred in 7% versus 2% with placebo and dizziness in 9% versus 4%. Those findings do not eliminate the need for larger and longer post-launch safety data. Convenience still carries an adherence cost. ## 2 | Daily Oral Dosing Versus Infrequent Injection Leqvio, or inclisiran, uses a very different adherence model. A healthcare professional administers it subcutaneously at initiation, again at Month 3 and then every six months. Calling it “two injections a year” is reasonable after the loading phase, but incomplete if the initial and three-month doses disappear from the description. Lipfendra transfers control to the patient. It avoids scheduled injection visits and could extend more easily into routine outpatient practice. The trade-off is a daily fasting routine and a 30-minute waiting period. Leqvio transfers scheduling to the healthcare system. Patients do not need to remember a pill each day, and clinics can track completed administration. The trade-offs are return visits, product stocking and a provider-administered reimbursement pathway. For adherent patients who dislike injections, oral dosing may be compelling. For older, high-risk patients already managing multiple medicines, a twice-yearly clinic routine may be more reliable. This is not a simple oral-versus-injectable victory. It is a question of which adherence architecture works for a specific patient. ## 3 | Strong LDL-C Reduction Is Not Yet a Cardiovascular Outcomes Claim Lipfendra was approved on LDL-C lowering. Its label does not claim a reduction in cardiovascular death, myocardial infarction or stroke. The PCSK9 target itself is well validated. Repatha and Praluent have reduced major cardiovascular events in large outcomes trials, while statins have an even deeper hard-outcomes evidence base. That supports LDL-C lowering as a risk-reduction strategy. It does not allow a new product to inherit another drug’s outcome claim. CORALreef Outcomes is designed to answer the Lipfendra-specific question. [ClinicalTrials.gov](https://clinicaltrials.gov/study/NCT06008756) lists an estimated enrollment of 14,550 high-risk participants, while Merck has said more than 14,500 patients have been enrolled. The primary endpoint is a composite of coronary heart-disease death, myocardial infarction, ischemic stroke, acute limb ischemia or major amputation, and urgent arterial revascularization. The registry currently lists November 2029 as the primary-completion date. Leqvio is also waiting for dedicated cardiovascular-outcomes evidence. Novartis expects ORION-4 and VICTORION-2P to read out in 2027. Until then, infrequent dosing should not be described as proven to reduce cardiovascular events. The 2026 ACC/AHA dyslipidemia guideline lowered the LDL-C target below 55 mg/dL for some very-high-risk patients. It supports adding ezetimibe and/or PCSK9 monoclonal antibodies for selected patients who remain above goal despite maximally tolerated statin therapy. Inclisiran is an option in situations such as intolerance, limited access to antibodies or a strong preference for infrequent dosing. The guideline was released before Lipfendra’s approval. It could not have endorsed the product in advance, and it should not be summarized as a blanket recommendation for every PCSK9 approach. ## 4 | Merck and Novartis Are Selling Different Treatment Experiences Merck listed Lipfendra's FDA approval as a key second-quarter milestone and linked new-product launches with its long-term growth trajectory. Approval alone does not guarantee a rapid ramp: physician adoption, payer access and real-world commercialization still need to be demonstrated. That caution is realistic. A new drug must enter formularies, pharmacies and clinical workflows; payers must agree to cover it; physicians must learn when to use it; and patients must maintain the fasting routine. Novelty does not rewrite a prescribing system overnight. Leqvio is also still expanding. Novartis reported second-quarter 2026 sales of $480 million, up 59% at constant currencies. First-half sales reached $932 million, up 64% at constant currencies. The 64% figure applies to the first half, not the quarter. The company also reported U.S. monthly prescription growth of 49% and a 23.3% share of new prescriptions in the Medicare Part B channel, while adoption increased after national reimbursement in China. These are company-reported operating measures, not proof that an injectable will defeat or lose to an oral medicine. They do show that the provider-administered model is still gaining scale. The market may ultimately segment rather than collapse into one winner: patients preferring daily oral dosing, patients suited to twice-yearly clinic administration, and patients prioritizing antibodies with established cardiovascular-outcomes evidence. ## 5 | What Taiwan Should Watch Taiwan already has a National Health Insurance pathway for injectable PCSK9 therapies. Effective September 1, 2025, reimbursement criteria for evolocumab and alirocumab were broadened for selected populations. The LDL-C threshold for initiating treatment fell from 135 to 100 mg/dL, and prior-authorization review moved from every six months to every 12 months. The National Health Insurance Administration estimated roughly 5,005 beneficiaries per year and NT$443 million in additional annual drug spending. That improves access to intensive LDL lowering for high-risk patients. But as of August 19, 2026, the primary-source review used for this article had not identified a Taiwan approval or reimbursement announcement for Lipfendra. U.S. approval does not mean the product can already be prescribed in Taiwan. Four issues now matter locally: whether the Taiwan FDA accepts and approves the product, what the final label says, whether pricing fits the reimbursement system, and whether the daily fasting routine creates a new source of missed doses in real-world practice. ## Conclusion | Oral PCSK9 Opens a Door; It Does Not Sentence Injections to Death Lipfendra’s central achievement is proving that a target dominated by injectable therapies can be addressed with a once-daily oral medicine producing substantial LDL-C reductions. The cholesterol market, however, is not decided by one laboratory measure. Long-term adherence, prescriber uptake, payer coverage and product-specific cardiovascular outcomes will determine whether Lipfendra is mainly a convenience upgrade or a true market reset. Injectables will not disappear tomorrow. The market has gained another option — and a more complicated comparison. Merck has secured entry into the first round of oral PCSK9 competition. Whether Lipfendra becomes the category leader will depend on hard outcomes after 2029 and on real-world prescribing before then. --- ## 默沙東口服藥殺進「降膽固醇針」市場，誰會被搶走？ URL: https://drugnews.com.tw/articles/2026-08-26-merck-lipfendra-oral-pcsk9.html Date: 2026-08-26 Language: zh-Hant Summary: FDA 核准首款每日口服 PCSK9 抑制劑 Lipfendra，兩項第三期試驗在第 24 週降低 LDL-C 56% 與 59%。但口服不等於針劑退場：心血管硬終點、每日空腹依從性、保險准入與台灣核准，才會決定市場如何重切。 過去十年，想用 PCSK9 這個靶點把壞膽固醇大幅壓低，幾乎都得接受注射。 現在，默沙東把這條路做成一顆每天吞的藥。 2026 年 7 月 15 日，美國食品藥物管理局（FDA）正式核准 Lipfendra（學名 enlicitide）。它是第一款獲 FDA 核准的口服 PCSK9 抑制劑，用於成人高膽固醇血症，包括異合子家族性高膽固醇血症（HeFH），需搭配飲食與運動。 兩項關鍵第三期試驗共 3,207 人。第 24 週時，Lipfendra 相較安慰劑校正後，讓低密度脂蛋白膽固醇（LDL-C，俗稱壞膽固醇）下降 56% 與 59%。 這個數字，足以讓市場開始想像：一顆藥，會不會把針劑的市場整碗端走？ 先別急著替針劑辦告別式。 Lipfendra 確實跨過一道大門，但降脂藥真正的勝負，還要看心血管硬終點、長期依從性、醫師處方與保險給付。口服只是開門方式改了，門後的仗才剛開始。 ## 01｜第一顆口服 PCSK9，突破在哪裡？ PCSK9（會加速肝臟 LDL 受體被分解的蛋白）像一個「拆除標記」。它和 LDL 受體結合後，受體容易被帶進細胞內分解；肝臟能回收壞膽固醇的受體變少，血中的 LDL-C 就會堆高。 過去，臨床多靠抗體藥物在血液中抓住 PCSK9，像安進的 Repatha、再生元／賽諾菲的 Praluent；諾華 Leqvio 則用小干擾 RNA（siRNA）降低肝臟製造 PCSK9。這些路線都需要注射。 Lipfendra 是大環胜肽。它把胜肽折成較穩定的環狀結構，阻止 PCSK9 與 LDL 受體結合，讓更多受體留在肝細胞表面，繼續把 LDL-C 從血液中清掉。 FDA 仿單把兩項試驗分得很清楚： - CORALreef Lipids 納入 2,904 名已有動脈粥樣硬化性心血管疾病（ASCVD）或具高風險的高膽固醇成人。第 24 週相較安慰劑校正後，LDL-C 平均下降 56%（95% 信賴區間：下降 61% 至 51%）。 - CORALreef HeFH 納入 303 名異合子家族性高膽固醇血症患者。第 24 週相較安慰劑校正後，LDL-C 平均下降 59%（95% 信賴區間：下降 66% 至 53%）。 這兩個數字來自不同受試族群，不能揉成一個「近 60%」的萬用數字，更不能直接拿去和別家試驗排名。 但 Lipfendra 並非拿起來就吞。 仿單要求每天早晨空腹服用一顆 20 mg，搭配開水、黑咖啡或無添加的茶，至少等 30 分鐘才能進食或喝其他飲料。這比打針容易接受，卻仍是一套需要每天記住的規則。 安全性也不是一片空白。高風險高膽固醇試驗中，整體不良反應頻率與安慰劑相近；HeFH 試驗則較常見腹瀉（7% 對 2%）與頭暈（9% 對 4%）。這些數字不代表重大風險已經排除，上市後仍要看更長期、更大規模的真實世界資料。 方便，不等於完全沒有依從性成本。 ## 02｜口服對上半年一針，誰才真的方便？ 諾華 Leqvio（inclisiran）的打法完全不同。 它由醫療專業人員皮下注射，第一次給藥後，第 3 個月再打一劑，之後每 6 個月一次。完成起始療程後，確實接近「一年兩針」，但若只寫一年兩針，就把前面的起始與第 3 個月劑量藏掉了。 兩款藥競爭的，其實是兩套依從性邏輯。 Lipfendra 把用藥控制權交給患者。好處是不用安排注射、容易延伸到更多一般門診；代價是每天都要記得吃，還得守住空腹與等待 30 分鐘的規則。 Leqvio 把時程交給醫療體系。患者不必天天記藥，院所能追蹤是否完成給藥；代價是必須回診、醫療端要備藥，支付也走較特殊的院所給藥路徑。 對服藥規律、排斥針頭的人，口服很有吸引力；對藥物很多、容易漏服的高齡高風險患者，半年回院一次也可能更穩。 所以這不是「口服一定贏針劑」。真正的問題是：哪一種管理方式，更適合眼前這位患者？ ## 03｜LDL-C 降得漂亮，還不等於心肌梗塞真的變少 Lipfendra 這次核准，依據是 LDL-C 降幅。它沒有拿到「降低心血管死亡、心肌梗塞或中風」的適應症。 PCSK9 這個靶點並不陌生。Repatha、Praluent 等單株抗體，已在大型心血管結局試驗中證明可以降低重大心血管事件；他汀類藥物的硬終點證據更完整。這些證據支持「把 LDL-C 壓低能改善風險」的大方向。 但靶點成立，不等於每一款新產品都能把別人的硬終點證據直接搬來用。 Lipfendra 的 CORALreef Outcomes 正在回答這道題。[ClinicalTrials.gov](https://clinicaltrials.gov/study/NCT06008756) 目前列預估收案 14,550 名高風險受試者，默沙東則表示已完成超過 14,500 人入組。主要終點是冠心病死亡、心肌梗塞、缺血性中風、急性肢體缺血或重大截肢，以及緊急動脈血管重建的複合事件；目前登錄的主要完成日期為 2029 年 11 月。 Leqvio 也在等自己的專屬結局資料。諾華在第二季簡報中預期 ORION-4 與 VICTORION-2P 於 2027 年讀出；在資料出來前，也不能把低頻注射的便利直接寫成已證實降低心血管事件。 2026 年美國心臟病學會／美國心臟協會（ACC／AHA）血脂異常指引確實存在，也把部分高風險患者的 LDL-C 目標壓到 55 mg/dL 以下。對使用最大耐受劑量他汀後仍未達標的特定患者，指引支持加入依折麥布（ezetimibe）和／或已有證據的 PCSK9 單株抗體；inclisiran 則在無法耐受、無法取得單抗，或強烈偏好低頻給藥時成為選項。 這份指引發布在 Lipfendra 核准之前，當然不可能預先替它背書。把它寫成「所有 PCSK9 都已全面前移」，會把指引說得太滿。 ## 04｜市場預期為何分化？兩家公司賣的不是同一種體驗 默沙東自己沒有把上市初期說得太樂觀。 默沙東在第二季財報中把 Lipfendra 的 FDA 核准列為關鍵里程碑，並把新產品上市與長期成長連在一起。但核准不等於快速放量；醫師採用、支付准入與實際商業化速度，仍要逐步驗證。 這是很現實的判斷。新藥要進醫院與藥局、談保險給付、教育醫師，還要讓患者願意每天照規則服用。第一款口服 PCSK9 有新鮮感，處方系統卻不會一夜改寫。 Leqvio 也不是站著等人搶。 2026 年第二季，Leqvio 銷售額為 4.80 億美元，按固定匯率成長 59%；上半年為 9.32 億美元，按固定匯率成長 64%。這裡最容易看錯：64% 是上半年口徑，不是第二季。 諾華披露，美國第二季每月處方量成長 49%，Medicare Part B 新處方市占達 23.3%；中國納入國家醫保後，相關市場占有率也翻倍。這些都是公司自己的營運指標，不能直接拿來當口服藥一定贏或一定輸的答案，卻證明院所給藥這條路仍在擴張。 市場最後很可能不是一款藥吃掉全部，而是被重新切成幾塊：偏好每天口服的人、適合半年回院給藥的人、需要單抗且重視既有硬終點證據的人，各自走向不同方案。 ## 05｜台灣真正要看的：方便之外，還有核准與健保 台灣目前已經有 PCSK9 注射型治療的健保路徑。 健保署公告，自 2025 年 9 月 1 日起，evolocumab 與 alirocumab 的給付條件放寬：部分適用族群開始治療的 LDL-C 門檻由 135 mg/dL 下修至 100 mg/dL，事前審查由每 6 個月延長為每 12 個月一次；估計一年約 5,005 人受惠，新增藥費約新台幣 4.43 億元。 這代表台灣高風險患者對強效降脂的可近性正在提高。不過，截至 2026 年 8 月 19 日，本次查核的一級來源尚未見 Lipfendra 的台灣核准或健保收載公告。 美國核准，不能直接翻譯成「台灣已經能開」。 台灣接下來要盯四件事：台灣食品藥物管理署（TFDA）是否受理與核准、最終標籤怎麼寫、價格能否進入健保可接受範圍、每日空腹規則在真實世界會不會變成新的漏藥來源。 ## 結語｜第一顆口服 PCSK9 打開大門，沒有宣判針劑死刑 Lipfendra 最厲害的地方，是它證明原本由注射主導的 PCSK9 靶點，也能做成每天一顆、降幅明確的口服藥。 但降脂市場從來不只比一張 LDL-C 報告。 患者願不願意長期照規則吃、醫師是否容易開、保險是否願意付、心血管硬終點能不能兌現，才會決定這顆藥最後是便利性升級，還是真正改寫治療版圖。 針劑不會明天消失。市場反而多了一種選擇，也多了一套更難的比較。 口服 PCSK9 的第一戰，默沙東已經拿到入場券；要不要封王，得等 2029 年以後的硬證據與真實世界處方，一起給答案。 --- ## The Stronger the Backbone, the More Valuable the Missing Layer: Can Anbogen's ABT-301 Enter the Next Wave of Oncology BD? URL: https://drugnews.com.tw/articles/2026-08-25-anbogen-abt-301-combination-layer-tokyo-comparative-oncology-en.html Date: 2026-08-25 Language: en Summary: Big Pharma is paying for incremental efficacy around established therapies. ABT-301 has completed Phase 1; the CRC triplet's efficacy, sustainable dosing, durability and biomarker readouts will determine whether it can become a Combination Layer and open substantive partnering opportunities. From Revolution x BeOne and PD-1/VEGF to ADCs: Big Pharma is paying for incremental efficacy, and the next re-rating of Ambogen (7784) will hinge on positive readout of the CRC triplet of ABT-301, positioning as a combination foundation and opening the door to strategic partnering By Drugnews | Data cut: August 24, 2026 > Drugnews Insight | Big Pharma is paying for incremental efficacy around established therapies. Anbogen (7784) has completed a 23-patient Phase 1 study of ABT-301, with 1 PR and 8 SD, and is now advancing a BeOne-supported global Phase 1/2 triplet trial of ABT-301 + tislelizumab + bevacizumab in pMMR CRC (NCT07244705). The key is simple: better efficacy, sustainable dosing, durable benefit, and identifiable responders. If all four hold, ABT-301 could open the door to Big Pharma partnerships and a valuation re-rating. In August 2026, Revolution Medicines, which now holds one of the industry's most advanced RAS portfolios, agreed to have BeOne Medicines fund and execute a global registrational Phase 3 trial for one of its candidates. BeOne also secured development or commercialization rights to four RAS(ON) inhibitors in selected Asian markets, while the two companies opened the door to combinations with a PRMT5 inhibitor and an EGFRxMETxMET trispecific antibody. Why did the deal get attention? Start with Revolution's position. By the end of July 2026, its market capitalization was approaching $40 billion. Daraxonrasib had already delivered a positive Phase 3 readout in pancreatic cancer, and the FDA had accepted the NDA for review. As of the end of June, the company held roughly $3.9 billion in cash, cash equivalents and marketable securities, and had already begun building a U.S. commercial infrastructure. Revolution had capital, late-stage assets and multiple strategic options, yet still chose to hand a global registrational trial to BeOne. That says something important: at the late-development stage, the bottleneck can shift from molecule quality and funding to the organizational capacity required to run multiple global trials in parallel. Figure 1 | The next wave of oncology BD: the stronger the backbone, the more valuable the Combination Layer Global partnerships are increasingly aimed at the gaps left by established backbone therapies - primary non-response, resistance and limited durability. Assets that can deliver measurable incremental benefit are moving onto Big Pharma's collaboration shortlist. What BeOne brings to the table is a global development system that has already been tested repeatedly at scale. Its 2025 interim report cited more than 3,700 employees across global R&D and clinical operations, trials running on six continents, and relationships with regulators and investigators in more than 45 countries. Zanubrutinib had enrolled roughly 7,100 patients, while tislelizumab had been studied in about 14,000 participants across 70 trials in 35 countries. For Revolution, those numbers translate into immediately deployable capabilities in patient enrollment, data operations and multi-region regulatory execution. Revolution has high-value molecules; BeOne has its own oncology pipeline. What they are exchanging is global clinical execution, regional market access and entry points for combination development. Once a backbone therapy reaches late-stage development, a different scarcity emerges: how to treat more patients, extend the duration of benefit and enter tumor types that the backbone cannot reach on its own. The same pattern is visible elsewhere. Revolution and Summit are pairing three RAS(ON) inhibitors with the PD-1/VEGF bispecific ivonescimab. GSK has added the B7-H3 ADC risvutatug rezetecan alongside ivonescimab. Each partnership is aimed at a clinical break point - primary non-response, insufficient depth of response, limited durability or acquired resistance. These deals are best used as a buyer's yardstick: what type of incremental benefit is Big Pharma willing to pay for? That is why Anbogen's ABT-301 deserves to be placed in the same buyer framework now—not because Anbogen is comparable to Revolution, but because its CRC triplet is testing the same kind of value buyers increasingly seek. If the trial can demonstrate clinically meaningful benefit in difficult-to-treat pMMR/non-MSI-H advanced colorectal cancer, ABT-301 could move beyond an early-stage HDAC asset and toward a combination foundation that a larger oncology company could plug into an existing franchise. . > Anbogen's re-rating does not require guessing the buyer. The market should first focus on three questions: Can ABT-301 deliver incremental efficacy? Can the dose be maintained over time? Can the responder population be identified clearly? ## 1 | The Stronger the Backbone, the More Valuable the Incremental Layer Once a backbone therapy enters a major indication, the buyer has already invested in global trials, manufacturing, regulatory infrastructure, physician education and commercial channels. Adding an external asset that improves outcomes can leverage that installed base, making the capital efficiency of incremental innovation potentially much higher than building an entirely new standalone franchise. Revolution and Summit offer an intuitive example. RAS(ON) inhibitors suppress a core oncogenic driver, while ivonescimab targets both PD-1 and VEGF in an effort to address immune escape, abnormal vasculature and the tumor microenvironment. GSK then adds a B7-H3 ADC, layering targeted cytotoxicity onto the same therapeutic backbone. BeOne and Revolution are taking the same logic into the RAS ecosystem through PRMT5, EGFR and MET biology. Clinical outcomes from these combinations still need to be proven, but the buyer need is already clear: a valuable Combination Layer should accomplish at least one of three jobs. If Revolution only needed outsourced trial execution, it could simply hire a CRO. Instead, this transaction exchanges regional rights, global development responsibilities and combination access. BeOne is putting an internal system that can shorten the path to a registrational milestone on the table; in return, it gains rights to four RAS assets in selected Asian markets and a path for BGB-58067 and BG-T187 to enter Revolution's combination network. The fact that clinical-development capability itself is embedded in the deal structure is one of the most relevant takeaways for other biotech companies. Turn non-responders into responders. Make existing responses last longer. Open a backbone therapy to a tumor type it could not reach before. For investors, these three jobs are often more closely tied to asset value than mechanistic novelty alone. The larger the patient base of the backbone, the more quickly a reproducible incremental benefit can translate into additional patients, longer treatment duration and label expansion. Figure 2 | The stronger the backbone, the more valuable the gap around it Partnerships built around RAS + PD-1/VEGF, PD-1/VEGF + ADC, and RAS + PRMT5/EGFR-MET show that buyers are actively assembling therapeutic layers that address non-response, resistance and limited durability. ## 2 | What Big Pharma Pays For: A Defined Gap and an Executable Next Step Recent transactions with disclosed economics show how Big Pharma is beginning to price these incremental layers. Takeda's global oncology collaboration with Innovent included a $1.2 billion upfront payment and covered IBI363, a PD-1/IL-2alpha-biased bispecific fusion protein; IBI343, a Claudin18.2 ADC; and an option on the early-stage dual-target ADC IBI3001. Together, the three assets add immune activation, precision cytotoxicity and a next-generation combination option. In 2026, Pfizer and Innovent expanded the model across 12 early and discovery-stage oncology programs, with Pfizer paying $650 million upfront and up to $9.85 billion in potential development, regulatory and commercial milestones. Pfizer had already built a major ADC franchise through Seagen; this collaboration adds new targets, new payload concepts and multispecific immune-engaging designs around that installed platform. Astellas and Vir moved even earlier with VIR-5500. The PSMA-targeted dual-masked T-cell engager was still in Phase 1 when Vir became eligible for $335 million in upfront and near-term payments, plus up to $1.37 billion in additional milestones. Astellas has an established prostate-cancer franchise, and VIR-5500 can plug directly into that franchise as a T-cell engagement layer - giving the buyer a reason to take risk earlier in development. The three transactions point to the same buyer standard: clinical stage is only one variable. An early asset can still command meaningful economics if it addresses a clearly defined treatment gap, is supported by enough human or translational evidence, and comes with a clear next-trial plan - including the patient population, backbone therapy and development responsibilities. Ultimately, Big Pharma is paying for an executable next step: which gap the asset fills, whether it can fill that gap in patients, and what the buyer gains by plugging it into an existing franchise. The more complete those answers are, the less discovery work the buyer has to absorb - and the closer the asset moves toward transaction readiness. Figure 3 | Which gaps are Big Pharma paying to fill? Recent transactions by Takeda, Pfizer and Astellas add immune activation, ADCs, multispecific immune engagement and T-cell engagement to existing oncology franchises. The economics reflect clear product complementarity and an executable development path. ## 3 | ABT-301 Through the Buyer's Lens: What Has Been De-Risked, and What Still Needs to Be Proven Anbogen's lead asset, ABT-301 (imofinostat), is an oral HDAC inhibitor that primarily targets HDAC1/2/3. HDAC drugs have established commercial precedent in hematologic malignancies, but single-agent development in solid tumors has historically been limited by modest efficacy and tolerability. Anbogen's strategy is therefore combination-led: use epigenetic and tumor-microenvironment modulation to help immunotherapy reach tumors that have remained poorly responsive. ABT-301 has completed a first-in-human Phase 1 study in 23 patients with advanced solid tumors. The study established 150 mg as the potential recommended Phase 2 dose (RP2D), reported one partial response and eight cases of stable disease, and showed an increase in acetylated histone H3, supporting target engagement in humans. Some cases of stable disease lasted 16 to 53 weeks, while the partial response lasted 24 weeks. Those data accomplish several basic de-risking tasks for an early asset: the drug has been dosed in humans, dose and PK/PD behavior have begun to take shape, and the target can be engaged. But this was still a small, uncontrolled, multi-tumor Phase 1 study and cannot establish efficacy in colorectal cancer. Treatment-related thrombocytopenia, nausea and vomiting, together with dose-limiting toxicities seen at higher dose levels, also make dose intensity a critical issue for the ongoing triplet. The one partial response and eight cases of stable disease should not be treated as a response rate, but the duration data do provide evidence of biological activity: the partial response lasted 24 weeks, and some stable-disease cases lasted 16 to 53 weeks. The AcH3 increase also links clinical exposure to HDAC inhibition. Taken together, whether disease stabilized, how long it remained stable and whether the target was engaged means Anbogen no longer has to prove the drug is usable from scratch; development capital can now be focused on combination dose optimization and indication selection. Anbogen is now advancing a global Phase 1/2 trial of ABT-301 + tislelizumab + bevacizumab in patients with locally advanced or metastatic pMMR/non-MSI-H colorectal cancer, with planned enrollment of 66 patients across the full study, including 36 patients in the Phase 2 preliminary efficacy portion.. This population has historically shown limited sensitivity to immune-checkpoint inhibition, making the clinical gap well defined. Each component addresses a different part of the problem. Tislelizumab releases the PD-1 immune brake; bevacizumab targets VEGF-driven abnormal vasculature and immune exclusion; ABT-301 is intended to increase antigen presentation, enhance CD8 T-cell activity and reduce immunosuppressive myeloid cells. The central question is whether adding this third layer creates a measurable incremental benefit. CAPability-01 provides an external human benchmark. In that randomized Phase 2 study, a different HDAC inhibitor, chidamide, was combined with sintilimab and bevacizumab in MSS/pMMR metastatic colorectal cancer. The triplet delivered a 44% ORR and 7.3-month median PFS, versus 13% and 1.5 months with the doublet. The study also reported two treatment-related deaths, reinforcing that efficacy and toxicity must be evaluated together. CAPability-01 provides an external human benchmark. The randomized Phase 2 study enrolled 48 patients: 23 received chidamide + sintilimab, and 25 received chidamide + sintilimab + bevacizumab. The prespecified primary endpoint was met, with an 18-week PFS rate of 43.8% (21/48) across the full study. The triplet versus doublet results were 64.0% versus 21.7% for 18-week PFS rate, 44.0% versus 13.0% for ORR, and 7.3 versus 1.5 months for median PFS; median DOR across the study was 12.0 months. Safety is equally important: the Nature Medicine report attributed two treatment-related deaths, one to hepatic failure and one to pneumonitis, underscoring that the commercial value of a combination is discounted when toxicity compromises sustained treatment. CAPability-01 supports the class-level hypothesis that HDACi + PD-1 + VEGF can help unlock immune-cold colorectal tumors. ABT-301 still needs to validate its own asset profile across three dimensions: incremental efficacy, durability and a workable safety window. That gap between class validation and asset validation is the source of Anbogen's next potential re-rating. Figure 4 | Where does ABT-301 stand today? ABT-301 has established a Phase 1 foundation in human dose, PK/PD and target engagement and has moved into a CRC triplet. The next step is to prove durable add-on efficacy. ## 4 | How the CRC Triplet Could Reframe ABT-301's Valuation As of August 20, 2026, ClinicalTrials.gov lists NCT07244705 as Recruiting, with planned enrollment of 66 patients, an estimated Primary Completion in April 2028 and Study Completion in July 2028. Under the company trial design, Phase 1 includes 30 patients for triplet dose escalation and RP2D exploration, while Phase 2 includes 36 patients across two active dose levels for preliminary efficacy assessment. Registry dates may be updated over time. Before full data disclosure, investors and analysts can follow three public readout nodes over the next two years: Node 1 — Phase 1 enrollment, dose selection and RP2D; Node 2 — confirmed ORR/DCR and early PFS; Node 3 — DOR/PFS maturation and biomarker readouts. Because this is an early, open-label study, ORR alone can be misleading. The market will need to triangulate four data sets: "Add-on" means response depth, disease control and early PFS should look better than a reasonable historical expectation; "Tolerable" means the RP2D, dose reductions, interruptions and overall dose intensity must support long-term triplet treatment; "Durable" means DOR and PFS need to rule out transient tumor shrinkage. "Explainable" means tumor tissue, peripheral blood or other biomarkers should identify a reproducible, enrollable responder population. The asset identity changes only if all four pieces connect. Acceptable safety without incremental efficacy leaves ABT-301 as an HDAC asset that can be combined. A response signal without durability or biomarkers may support collaboration discussions, but bargaining power remains limited. If efficacy, tolerability, durability and patient selection all line up, Anbogen will have its first Human Combination PoC. CRC triplet readout path: three scenarios for investors to track | Data scenario | How the market is likely to read it | Asset position | | --- | --- | --- | | Bull case | Nodes 1–3 validate sequentially. / Incremental efficacy, DOR/PFS, manageable safety and an identifiable responder population all line up; a buyer can design the next trial directly from the package. | Partner-ready | | Base case | Node 1 validates; Node 2 shows activity; Node 3 remains incomplete. / A response signal emerges, but durability or biomarkers remain incomplete; partnership options exist, but bargaining power is limited. | Option value preserved | | Bear case | Nodes 1–2 do not validate. / No clear incremental benefit, or toxicity prevents adequate dose intensity; dose, population or backbone needs to be redesigned. | Development thesis reset | What the market will price after each readout is residual risk. The more nodes are cleared, the less dose-finding, patient-segmentation and biomarker work a buyer must redo—and the more likely a relationship can move from research collaboration toward co-development or licensing. The value of the PoC is therefore not a single ORR number. It is whether a buyer can see which additional patients become treatable, what dose is sustainable, how long the benefit lasts and how the next trial should be designed. Figure 5 | The four questions the CRC triplet must answer "Adds value, can be sustained, lasts and can be explained" map to incremental efficacy, maintainable dosing, durability and patient selection. Only when all four data sets connect does the trial become combination evidence that can support BD evaluation. ## 5 | The University of Tokyo Collaboration: Lowering the Cost of the Next Clinical Miss The CRC trial asks whether ABT-301 can fill one clinical gap. To broaden the buyer universe, the asset must answer a second question: can the same biology be reproduced in another tumor type, another patient segment or another treatment backbone? According to the collaboration materials provided by the company, Anbogen and the Laboratory of Veterinary Surgery at the University of Tokyo's Graduate School of Agricultural and Life Sciences plan to use canine and feline tumor cell lines, organoids, in vivo models and naturally occurring cancers in companion animals to evaluate ABT-301 as monotherapy and in combinations with different drugs, and to identify molecular backgrounds that may predict response. Naturally occurring tumors in dogs and cats retain intact immune systems, tumor heterogeneity and disease evolution, adding translational information that conventional cell lines and mouse models may not fully capture. The commercial value of the platform lies in selection and elimination: which tumor type deserves human testing, which patients are most likely to benefit, which partner produces the largest incremental effect and which directions should be stopped before an expensive clinical trial begins. The University of Tokyo team's published work spans canine malignant melanoma, urothelial carcinoma, bladder-cancer organoids, as well as canine CAR-T, immune-checkpoint and combination-immunotherapy research. These tumor types do not necessarily need to become new Anbogen indications. Their more useful role may be to test whether ABT-301's TME-modulating biology persists in an intact immune environment and to turn a broad mechanistic thesis into testable hypotheses for patient selection and combination design. Investors should not read this collaboration as a plan to launch multiple new pipelines at once. The ideal output is a narrowing translational map that breaks the CRC human signal into responder biology, biomarkers, the best backbone and Go/No-go decisions, while creating a clear priority order for a second validation setting. That map has direct implications for capital efficiency. If Anbogen can compress ten plausible directions into one high-probability coordinate, the next clinical dollar can be concentrated, and a prospective buyer can more easily estimate the time, cost and probability of success after taking over the program. > For a small biotech, a translational platform should eliminate nine low-probability paths before adding ten new pipeline arrows. Figure 6 | University of Tokyo Comparative Oncology: a translational de-risking engine for AnBogen’s ABT-301 By combining naturally occurring tumors with an intact immune environment, the platform is designed to screen for the right Cancer, Biomarker, Partner and Go/No-go decision - with the goal of improving the hit rate of the next human trial. ## 6 | Partner-ready Means the Buyer Knows What to Do Next The BD value of an early-stage asset is often determined by how far the buyer still is from its next major decision point. If the buyer must re-establish dose, rebuild patient segmentation, guess the right combination backbone and create a biomarker strategy from scratch, the transaction carries a large exploration discount. The more complete the data package, the more room there is for economics to move higher. ABT-301's path from clinical proof to partnership—and then to higher strategic valuation—can be viewed in three layers. Layer 1 | Human Combination PoC — opens partnership potential. If the CRC triplet establishes a maintainable RP2D, clear incremental efficacy and maturing DOR/PFS, large oncology companies can begin evaluating substantive partnership opportunities. Layer 2 | Translational Map — supports valuation uplift. If the University of Tokyo work and clinical biomarker program clarify who benefits, why they benefit and which class of backbone is most rational, they can reduce exploratory work and execution risk in the next trial. Layer 3 | Cross-backbone Replication — broadens the buyer pool. If incremental benefit is reproduced with a second backbone or in a second high-priority indication, ABT-301 can move from a regimen-specific asset toward a Combination Layer that can plug into multiple oncology franchises. Anbogen has disclosed preclinical data showing that ABT-301 can increase chemotherapy sensitivity in KRAS-mutant pancreatic cancer through the HDAC3-NRF2 axis. Those data support pancreatic cancer as a research direction, but there is currently no public evidence of a direct ABT-301 + RAS(ON) inhibitor combination. RAS can therefore serve as a future experimental entry point, but it should not yet be priced into a transaction thesis. The two agents need to be tested directly for response depth, durability and delayed resistance before the combination can justify clinical development. If all three layers are completed, a prospective partner would receive an executable asset package: a human working dose, manageable safety, an identifiable responder population, the best backbone, biomarkers, the next indication and a blueprint for the next trial. That package reduces the buyer's cost of taking over development and could broaden the buyer universe beyond a single PD-1 partner to multiple oncology companies seeking TME modulation. > A signal in one regimen usually creates only a small number of negotiating counterparts. Replication across backbones is what can turn a single partnership discussion into buyer competition. Figure 7 | How ABT-301 could move from a clinical interface to buyer competition The value-creation path runs from Clinical Interface to Human Combination PoC to Translational De-risking and ultimately Buyer Pool Expansion. Replicability across backbones will determine bargaining power. ## Conclusion | Anbogen Is Now on a Verified Re-rating Path Anbogen is still in the data-generation phase. This is not yet the point to guess which Big Pharma company might buy the asset, nor should the tislelizumab supply arrangement be treated as an endorsement from BeOne. Investors can instead track three milestones to determine whether the asset is truly moving up the value curve.成 First, can the CRC triplet deliver a Human Combination PoC that is both interpretable and sustainable? Second, can the University of Tokyo collaboration and clinical biomarker program identify reproducible responder biology and the best partner? Third, can a second backbone generate incremental benefit again, showing that ABT-301's value extends beyond a single regimen? Phase 1 has established a human foundation for ABT-301, and CAPability-01 provides a class-level benchmark. If the CRC triplet establishes a Human Combination PoC, the door to substantive Big Pharma partnering could open. A University of Tokyo translational map that narrows the responder population, biomarkers and best partner—followed by replication with a second backbone—would further reduce the buyer's development burden and support a higher strategic valuation. The starting point for Anbogen's next re-rating is the first interpretable human incremental-efficacy signal from the CRC triplet; a drug-supply agreement by itself is not enough. A backbone drug typically has one rights holder. A second layer that can expand the responder pool across multiple backbones has the potential to turn a one-on-one partnership discussion into buyer competition. --- ## 主藥越強，旁邊缺的那一層越值錢：安邦ABT-301能否進入下一輪腫瘤BD名單？ URL: https://drugnews.com.tw/articles/2026-08-25-anbogen-abt-301-combination-layer-tokyo-comparative-oncology.html Date: 2026-08-25 Language: zh-Hant Summary: 大藥廠正為主藥旁的增量療效付費。安邦ABT-301已完成Phase I，接下來由CRC三聯的療效、劑量持續性、耐久性與biomarker決定其能否走向Combination Layer與大型藥廠合作。 從Revolution × BeOne、PD-1／VEGF到ADC：大藥廠正為「增量療效」付錢，安邦下一次估值轉折將由CRC三聯人體療效決定ABT-301聯合用藥基石，打開大型藥廠合作 撰文｜藥時事 Drugnews　　資料基準日｜2026年8月24日 > 藥時事觀點Drugnews Insight | 大藥廠正為補強主藥「增量療效」付錢。安邦（7784）ABT-301一期23人已見1人腫瘤縮小、8人病情穩定，目前已與BeOne展開供藥合作，推進ABT-301＋tislelizumab＋bevacizumab全球I/II期pMMR大腸直腸癌三聯試驗（NCT07244705）。下一步就看：療效是否更好、能否持續用、效 果是否持久、哪些人最受益；四關成立，就有機會打開大藥廠合作與估值重估。 2026年8月，手握後期RAS資產的Revolution Medicines，將其中一款候選藥的全球註冊性Phase III交由BeOne Medicines出資並執行。BeOne另取得四款RAS(ON)抑制劑在部分亞洲市場的開發或商業化權益，雙方也把PRMT5抑制劑、EGFR×MET×MET三特異性抗體接進RAS聯合治療。 這筆合作受到關注，先看Revolution的份量。2026年7月底，公司市值逼近400億美元；daraxonrasib的胰臟癌Phase III已公布陽性結果，FDA也受理上市申請。公司在6月底持有約39億美元現金、現金等價物與有價證券，並已為美國上市建立商業化基礎。它有資金、有後期產品，也有多個合作選項，仍將一項全球註冊性試驗交給BeOne。後期開發的瓶頸開始由分子與現金，轉向同時推動多個全球試驗的組織容量。 圖1｜下一輪腫瘤BD的核心：主藥越強，Combination Layer越值錢 全球合作正集中處理主藥留下的非反應、抗藥與療效持久性缺口；能提供可量化增量的第二層，開始進入大藥廠合作清單。 BeOne能拿來交換的，是一套已跑過多次全球開發的內部系統。2025年半年報顯示，其全球研發與臨床營運團隊超過3,700人，在六大洲開展試驗，並與45個以上國家的監管機關及研究者合作；zanubrutinib累計入組約7,100人，tislelizumab則在35個國家進行70項試驗、累計約1.4萬名受試者。對Revolution來說，這些數字代表可以立即使用的患者入組、資料管理與跨區申報能力。 Revolution有熱門分子，BeOne也有自己的腫瘤管線。交易桌上交換的是全球臨床執行、區域市場與管線聯用入口。當一顆主藥進入後期，下一個稀缺能力開始浮現：如何讓它多打一群病人、撐得更久，或進入原本碰不到的癌種。 同一條脈絡出現在其他合作。Revolution與Summit把三款RAS(ON)抑制劑接上PD-1／VEGF雙抗ivonescimab；GSK再把B7-H3 ADC risvutatug rezetecan放到ivonescimab旁邊。每筆合作都對準臨床留下的斷點：初始非反應、反應不夠深、療效不夠久，以及後續抗藥。 這些合作更適合當成一把買方尺：大藥廠願意為哪一類增量付錢？這也正是安邦（7784）ABT-301值得提前放進同一個買方框架的原因。把這把尺放到ABT-301，投資命題落在一個清楚分岔口——安邦的CRC三聯試驗能在pMMR/non-MSI-H這群難治型末期大腸直腸癌病人證明臨床獲益，安邦就有機會跨進Combination Layer的估值框架，走向可被大型藥廠接進既有腫瘤franchise的聯合用藥基石。 > 安邦的重估不必先猜買家。市場先看三件事：ABT-301能否帶來增量療效、劑量能否長期維持、受益患者能否被清楚辨識。 ## 一、主藥越強，旁邊的增量越容易被放大 一款主藥進入大型適應症後，買方已經投入全球試驗、製造、法規、醫師教育與商業化通路。此時加入一個能改善治療結果的外部資產，新增價值可以沿用既有基礎設施，資本效率通常高於重新建立一條獨立產品線。 Revolution與Summit提供一個直觀案例。RAS(ON)抑制劑壓制腫瘤的核心驅動訊號，ivonescimab同時作用於PD-1與VEGF，試圖補上免疫逃脫、異常血管與腫瘤微環境。GSK接著加入B7-H3 ADC，讓同一套骨幹再多一層定向細胞毒殺傷。 BeOne與Revolution則把PRMT5、EGFR與MET等機轉接進RAS生態。這些組合的臨床結果仍待驗證，買方需求已經相當具體：一顆好的Combination Layer，至少要完成三項任務之一。 若只需要試驗執行，Revolution可以繼續採用CRO。這筆合作以區域權益、全球開發責任與聯用入口交換。BeOne把能縮短資產抵達註冊節點的內部系統放上談判桌，對價則是四款RAS資產在部分亞洲市場的權益，以及BGB-58067、BG-T187進入Revolution聯合治療網絡的機會。臨床能力本身被寫進交易結構，才是這筆合作對其他Biotech最值得參照的地方。 讓原本不反應的病人開始反應。 讓已經反應的病人維持更久。 讓一顆主藥進入原本打不開的癌種。 對投資人而言，這三項任務比機轉的新穎度更接近資產價格。主藥的患者基礎愈大，第二層只要帶來可重複的增量，新增患者、延長治療時間與擴張適應症便能直接轉成商業價值。 圖2｜主藥越強，旁邊的缺口越值錢 RAS＋PD-1／VEGF、PD-1／VEGF＋ADC、RAS＋PRMT5／EGFR-MET等合作，顯示買方正在配置能補足非反應、抗藥與療效持久性的作用層。 ## 二、大藥廠付錢買的是清楚缺口，以及可以立刻執行的下一步 近期幾筆有明確對價的交易，進一步說明大藥廠如何替第二層定價。 Takeda與信達生物的全球腫瘤合作包含12億美元首付款，涵蓋PD-1／IL-2α-bias雙特異性融合蛋白IBI363、Claudin18.2 ADC IBI343，以及早期雙標的ADC IBI3001選擇權。三項資產分別補進免疫活化、精準殺傷與下一代聯合治療選項。 Pfizer在2026年與信達一次納入12項早期與新源頭腫瘤計畫，支付6.5億美元首付款，後續潛在開發、法規與商業里程碑最高98.5億美元。Pfizer已藉Seagen建立大型ADC版圖，這筆合作替既有平台增加新靶點、新payload與多特異性免疫設計。 Astellas與Vir的VIR-5500合作更早。這款PSMA標靶dual-masked T-cell engager仍在Phase I，Vir即可取得3.35億美元前期與近期付款，另有最高13.7億美元里程碑。Astellas長期經營前列腺癌，VIR-5500可以直接補進既有franchise的T-cell engagement層，買方因此願意在早期承擔風險。 三筆交易傳達同一個買方標準：臨床階段只是其中一項決定因素。資產若對準明確治療斷點，已有足夠人體或轉譯證據支持，且下一場試驗的患者、骨幹與開發責任清楚，早期項目同樣能取得高額對價。 大藥廠最後會替一套可執行的下一步付錢：補哪個缺口、人體裡補不補得起來、接進既有管線後多得到什麼。答案愈完整，買方還要承擔的探索成本愈低，資產也愈接近可交易狀態。 圖3｜大藥廠正在花錢補哪一個缺口？ Takeda、Pfizer與Astellas近期合作將免疫活化、ADC、多特異性與T-cell engagement接進既有腫瘤版圖；交易對價反映清楚的產品互補與可執行開發路徑。 ## 三、把ABT-301放進買方模型：已完成的風險，和尚未完成的證明 安邦生技的核心資產ABT-301（imofinostat）是一款口服HDAC抑制劑，以HDAC1／2／3為主要作用標的。HDAC類別在血液腫瘤已有商業化經驗，實體瘤單藥開發則長期面臨療效有限與耐受性的挑戰。安邦因此把ABT-301放進聯合治療，利用表觀遺傳與腫瘤微環境調節，協助免疫治療進入原本反應不佳的腫瘤。 ABT-301已完成23位晚期實體瘤患者的first-in-human Phase I。試驗確立150 mg為潛在的二期推薦劑量（RP2D），整體出現1位PR與8位SD，並觀察到acetylated histone 3上升，支持人體target engagement；部分疾病穩定維持16至53週，PR維持24週。 這組資料完成早期資產的幾項基本工作：藥物已進入人體，劑量與PK／PD有初步輪廓，標的也能被打到。試驗仍是跨癌種、無對照的小型Phase I，無法推論大腸直腸癌療效。血小板低下、噁心與嘔吐等治療相關不良事件，以及高劑量組曾出現的DLT，也讓後續三聯的dose intensity成為關鍵。 Phase I中的1位PR與8位SD不能被當成有效率，但持續時間提供了資產仍有生物活性的線索：PR維持24週，部分SD維持16至53週。AcH3上升則把臨床暴露與HDAC抑制接起來。病人是否穩定、穩定多久、標的是否被打到，三類資料讓安邦不必從零開始證明藥物可用，後續資金可以集中在組合劑量與適應症選擇。 安邦目前推進ABT-301＋tislelizumab＋bevacizumab的全球Phase I/II，對象為pMMR／non-MSI-H局部晚期或轉移性大腸直腸癌，規劃收案66人，其中Phase II療效評估部分為36人。這群病人對免疫檢查點抑制劑的反應長期有限，臨床缺口明確。 三顆藥各自處理不同問題。Tislelizumab解除PD-1免疫煞車；bevacizumab抑制VEGF相關異常血管與免疫排斥；ABT-301嘗試提高抗原呈現、增加CD8 T細胞活性並降低免疫抑制性髓系細胞。試驗的核心，是第三層加入後能否產生可辨識的增量。 CAPability-01提供一個外部人體參照。該隨機Phase II使用另一款HDAC抑制劑chidamide搭配sintilimab與bevacizumab，共納入48人（chidamide+sintilimab雙聯23人、chidamide+sintilimab+bevacizumab三聯25人），主要終點18週PFS率為43.8%（21/48），達到預設主要終點；三聯對雙聯的18週PFS率為64.0% vs 21.7%，ORR為44.0% vs 13.0%，中位PFS為7.3 vs 1.5個月，全體中位DOR為12.0個月。安全性同樣重要。CAPability-01原文報告兩例治療相關死亡，分別為hepatic failure與pneumonitis，也提醒市場，組合療法的商業價值會被安全性折現。 CAPability-01支持HDACi＋PD-1＋VEGF打開冷腫瘤的類別假說，ABT-301仍要以自己的「增量療效」、「維持久」與「安全性窗口」完成資產驗證。這段差距，就是安邦下一次估值轉折的來源。 圖4｜ABT-301目前站在哪裡？ ABT-301已完成Phase I的人體劑量、PK／PD與target engagement地基，並進入CRC三聯；下一關是證明可維持的add-on efficacy。 ## 四、CRC三聯會把安邦帶向哪一種估值？ 投資人看這場Phase I/II，依照ClinicalTrials.gov臨床試驗NCT07244705公開資料時間來看，目前為Recruiting，預計總收案66人；Primary Completion預計2028/04，Study Completion預計2028/07。依公司試驗設計，Phase I為三聯劑量爬坡與RP2D探索試驗共30人，Phase II則為三聯兩個有效劑量初步療效評估試驗共36人；日期後續仍可能更新。 完整資料揭露前，未來兩年內的公開讀出判斷可以拆成三個節點：(節點1) 看近期Phase I的收案、劑量與RP2D；(節點2) 看confirmed ORR／DCR與早期PFS；(節點3) 再看DOR／PFS與biomarker。 這是早期、開放式試驗，單看ORR容易失真，市場最後會用四組資料交叉判讀: 加得上，看response depth、疾病控制與早期PFS是否高於合理歷史預期； 撐得住，看RP2D、減量、中斷與dose intensity能否支持長期三聯； 維持久，看DOR與PFS能否排除短暫縮瘤； 說得清，則要用腫瘤組織、周邊血液或其他biomarker辨識可重現、可收案的受益患者。 四題接起來，資產身份才會改變。安全性可接受但缺乏療效增量，ABT-301仍停留在可聯用的HDAC資產；出現反應訊號但缺少持久性與biomarker，合作討論可能存在，議價能力仍弱；療效、耐受性、持久性與患者選擇同時成立，安邦才會取得第一個Human Combination PoC。 CRC三聯資料出現後，市場可能走向三種估值情境 | 資料情境 | 市場如何解讀 | 資產位置 | | --- | --- | --- | | 多方 | 3個節點連續成立，增量療效、DOR／PFS、可管理安全性與受益族群同時成立；買方可直接設計下一場試驗。 | Partner-ready | | 中性 | 節點1成立，節點2出現反應訊號，節點3仍不完整，出現反應訊號，持久性或biomarker仍不完整；合作選項存在，議價能力有限。 | 保留選擇權 | | 空方 | 節點1-2未成立，無法辨識增量，或毒性使dose intensity不足；需調整劑量、族群或聯用骨幹。 | 開發假說重整 | 資料讀出後，市場看的是剩餘風險。三個節點愈完整，買方需要自行補做的劑量、患者分層與biomarker工作愈少，合作也愈可能從共同研究往共同開發或授權升級。 因此，這個PoC的價值不在單一ORR，而在於讓買方知道：多打開哪群病人、需要多少劑量、效果能維持多久，以及下一場試驗怎麼設計。 圖5｜CRC三聯要回答的四個問題 「加得上、撐得住、維持久、說得清」分別對應增量療效、可維持劑量、持久性與患者選擇；四組資料接起來，才構成可供BD評估的人體組合證據。 ## 五、東京大學合作的價值，在於降低下一場試驗的失敗成本 CRC試驗回答ABT-301能否補上一個臨床缺口。資產若要擴大買方範圍，還需要回答第二個問題：相同生物學能不能在另一個癌種、另一群患者或另一個治療骨幹中重現？ 依公司提供的合作研究資料，安邦與東京大學農學生命科學研究科獸醫外科學實驗室將運用犬貓腫瘤細胞株、類器官、體內模型與自然罹癌伴侶動物，評估ABT-301單藥與不同藥物聯用，並尋找可預測反應的分子背景。 自然發生的犬貓腫瘤具備完整免疫系統、腫瘤異質性與疾病演化，能補充細胞株和小鼠模型較難呈現的轉譯資訊。平台的商業價值來自篩選與淘汰：哪一個癌種值得進人體、哪一群病人最可能受益、哪一種partner能產生最大增量、哪些方向應在昂貴臨床啟動前停止。 東京大學團隊的公開研究涵蓋犬惡性黑色素瘤、泌尿上皮癌、膀胱癌類器官，以及犬CAR-T、免疫checkpoint與複合免疫療法。這些癌種未必會直接變成安邦的新適應症；它們更適合拿來測試ABT-301在完整免疫環境中的作用是否跨模型存在，並將HDAC可調節TME壓縮成可檢驗的病人選擇與聯用假說。 投資人不宜把這項合作解讀成一口氣增加多條pipeline。最理想的交付是一張縮小範圍的轉譯地圖，將CRC人體訊號拆成responder biology、biomarker、最佳骨幹與Go／No-go，並替第二個驗證場景提供清楚優先順序。 這張地圖直接影響資本效率。安邦若能把十個可能方向縮成一個高機率座標，下一筆臨床資金會更集中，潛在買方也更容易估算接手後的時間、成本與成功機率。 > 對小型Biotech而言，轉譯平台應優先刪掉九條低機率路徑；多畫十條管線只會分散有限資本。 圖6｜東京大學Comparative Oncology：安邦ABT-301的轉譯去風險引擎 平台利用自然腫瘤與完整免疫環境篩選Cancer、Biomarker、Partner與Go／No-go，目標是提高下一場人體試驗的命中率。 ## 六、Partner-ready的定義：買方拿到資料後，知道下一場試驗怎麼做 一顆早期資產的BD價格，常由買方距離下一個關鍵決策點還有多遠所決定。買方若仍要重新找劑量、重做患者分層、猜聯用骨幹並建立biomarker，交易會承擔大量探索折價。資料包愈完整，對價愈有機會往上移動。 ABT-301走向Partner-ready，需要三層證據連在一起。 第一層｜Human Combination PoC。打開合作可能。CRC三聯若交出可維持的RP2D、明確增量與DOR／PFS，大型藥廠即可開始評估實質合作。 第二層｜Translational Map。提高valuation。東京大學與臨床biomarker若能回答誰受益、為何受益、和哪一類主藥搭配最合理，就能降低下一場試驗的探索成本。 第三層｜Cross-backbone Replication：擴大買方池。若第二個backbone或高優先適應症重現增量，ABT-301就能由單一regimen資產走向多個腫瘤franchise。 安邦已公開ABT-301透過HDAC3–NRF2路徑提高KRAS突變型胰臟癌對化療敏感性的臨床前資料。這項結果支持胰臟癌研究方向，目前沒有ABT-301與RAS(ON)抑制劑直接聯用的公開證據。RAS可以成為後續實驗入口，現階段不宜提前納入交易估值。把兩顆藥直接放進模型，觀察response depth、持久性與抗藥延後，才能決定是否值得進入臨床。 當三層證據完成，潛在合作方取得的會是一套可執行資產包：人體工作劑量、可管理安全性、受益患者、最佳骨幹、biomarker、下一個癌種與後續試驗設計。這套答案能降低接手成本，也能把買方範圍由單一PD-1合作方擴大到多個需要TME調節的腫瘤公司。 > 一個regimen的訊號通常只帶來少數談判對手；跨backbone重複，才有機會把單一合作談判推進成買方競爭。 圖7｜ABT-301如何從臨床接口，走到買方競爭？ 價值遞增路徑由Clinical Interface、Human Combination PoC、Translational De-risking走向Buyer Pool Expansion；跨backbone可複製性將決定議價權。 ## 結語｜安邦值得追蹤的，是一條可以被驗證的估值路徑 安邦目前仍處於資料生成期。這個階段不適合先猜哪一家大藥廠會買，也不適合把tislelizumab供藥安排視為BeOne背書。投資人可以沿著三個里程碑判斷資產是否升級。 第一，CRC三聯能否交出可解釋、可維持的Human Combination PoC。 第二，東京大學與臨床biomarker能否找出可重複的responder biology與最佳partner。 第三，第二個backbone能否再次產生增量，證明ABT-301的價值可以跨越單一regimen。 Phase I已替ABT-301建立人體地基，CAPability-01提供類別參照，CRC三聯負責證明安邦自己的增量，若CRC Human Combination PoC成立，合作可能就被打開；東京大學平台再把患者、biomarker與最佳partner收斂，第二個backbone再驗證可複製性，則會進一步降低買方接手成本並推升valuation。。 安邦下一次重估的起點，是CRC三聯第一組可解釋的人體增量資料；供藥協議本身不夠。一顆主藥通常只有一個權利人，能替多個主藥擴大受益人群的第二層，才有機會把單一合作談判變成買方競爭。 --- ## Curium and Lantheus Agree to an Up-to-$8 Billion Merger: Radiopharma Becomes a Full-Stack Business URL: https://drugnews.com.tw/articles/2026-08-24-curium-lantheus-definitive-merger-en.html Date: 2026-08-24 Language: en Summary: The $7 billion Curium-Lantheus rumor covered in June became a definitive agreement on August 3: $102.50 in cash plus up to $12.00 per share in CVRs, for total consideration of up to $8 billion. The strategic prize is an integrated chain spanning isotopes, manufacturing, distribution, diagnostics, therapy and U.S. hospital access. ## 00 | The story has moved from a $7 billion rumor to a definitive agreement worth up to $8 billion Radiopharma is entering a different phase of competition. On August 3, 2026, Curium and Lantheus Holdings announced that they had signed a definitive merger agreement. Under the agreement, Lantheus shareholders would receive $102.50 per share in cash at closing, together with contingent value rights worth up to another $12.00 per share if specified commercial milestones are achieved. The maximum aggregate transaction value is approximately $8.0 billion. The status boundary matters. **A definitive agreement is not a completed acquisition.** The transaction still requires approval by Lantheus shareholders, regulatory clearances and other customary closing conditions. The CVR milestones may never be achieved, so the additional $12.00 per share should not be treated as certain consideration. [Drugnews' June 9 article](https://drugnews.com.tw/articles/2026-06-09-curium-lantheus-radiopharma-acquisition-en.html) analyzed the earlier report that Curium might acquire Lantheus for roughly $7 billion. This article is the next chapter after the August 3 agreement: price, structure and strategic intent are now clearer, even though the merger has not yet closed. The central issue is not the headline price. It is the way radiopharma acquisition logic is evolving. Eli Lilly's acquisition of POINT Biopharma, Bristol Myers Squibb's purchase of RayzeBio and AstraZeneca's acquisition of Fusion Pharmaceuticals each added a radiopharmaceutical platform, pipeline or isotope-linked capability to a much larger oncology portfolio. Curium-Lantheus is different. It combines two nuclear-medicine specialists and attempts to connect isotope access, manufacturing, distribution, commercial diagnostics, hospital channels and therapeutic programs inside one operating system. Radiopharma competition is therefore becoming less about owning one attractive molecule and more about controlling the entire theranostic chain. ## 01 | Curium needs the U.S. endpoint; Lantheus needs upstream depth and a stronger therapy arm Curium's advantage sits upstream. The company is focused entirely on nuclear medicine and operates a vertically integrated radiopharmaceutical manufacturing and distribution network. Its infrastructure spans isotope production, reactor- and cyclotron-based manufacturing, radiopharmacies and specialized logistics. The merger announcement says the combined company would serve oncology, neurology and cardiology patients in more than 70 countries. Curium also brings therapeutic development assets. One of the most important is 177Lu-PSMA-I&T, a PSMA-targeted radioligand therapy being studied in metastatic castration-resistant prostate cancer. ECLIPSE, ClinicalTrials.gov identifier NCT05204927, is a randomized phase 3 trial sponsored by Curium US. Curium's weakness is equally visible. It has isotope, manufacturing, distribution and therapeutic depth, but it does not possess a U.S. diagnostic franchise and hospital commercialization network comparable with Lantheus. Lantheus offers the mirror image. PYLARIFY, or piflufolastat F 18, is the leading PSMA PET imaging agent in the United States. When Lantheus announced FDA approval of the new PYLARIFY TruVu formulation in March 2026, it said that more than 760,000 people with prostate cancer had received PSMA PET scans with PYLARIFY since the original product's 2021 approval. TruVu is designed to preserve diagnostic performance while supporting larger batches and greater patient availability; Lantheus expects a phased U.S. launch beginning in the fourth quarter of 2026. The company also markets DEFINITY, an ultrasound-enhancing agent used in echocardiography, and Neuraceq, or florbetaben F 18, a beta-amyloid PET imaging agent. Lantheus' 2025 shareholder materials described Neuraceq as the second most utilized and fastest-growing commercial amyloid PET agent in the United States. The strategic complement is straightforward. Curium supplies upstream production, logistics and a therapeutic arm. Lantheus supplies established radiodiagnostics and a U.S. hospital entry point. The merger is not filling one product gap; it is attempting to join the hardest parts of the radiopharma value chain. ## 02 | Half-life does not merely constrain radiopharma; it determines the business model A conventional small molecule or antibody can often be manufactured, stored and shipped over long distances with a reasonably forgiving clock. A radiopharmaceutical cannot be managed in the same way. Radioactive isotopes decay continuously. A short half-life means that production, labeling, quality testing, release, shielded transportation and hospital scheduling must be synchronized. Inventory is perishable by design. A delayed shipment does not simply arrive late; it may arrive with insufficient activity for the planned procedure. That is why radiopharma's competitive barrier begins well before clinical efficacy. A company must secure reliable isotope supply, manufacture sterile products under GMP conditions, operate or coordinate specialized transport, work with nuclear-medicine departments and deliver the correct activity to the correct patient at the scheduled time. Therapeutic radiopharmaceuticals add another layer. Products such as Pluvicto, or lutetium Lu 177 vipivotide tetraxetan, require manufacturing and distribution to align with a named patient's treatment slot. Half-life, batch quality, hospital capacity and patient readiness all affect whether a dose can be used. The moat is therefore not a molecular structure in isolation. It is the ability to make the isotope, build the product, release it, move it and administer it as one dependable process. The Curium-Lantheus combination is an attempt to turn that operational chain into a strategic platform. ## 03 | Theranostics creates radiopharma's most valuable commercial loop Radiopharma can connect diagnosis and treatment around the same biological target. A diagnostic isotope is used with PET or SPECT to identify patients whose tumors express the target. A therapeutic isotope can then deliver radiation near those tumor cells. This is the foundation of theranostics. PSMA in prostate cancer is the most mature example. Novartis uses Locametz, or gallium Ga 68 gozetotide, for PSMA PET imaging and Pluvicto for PSMA-positive prostate cancer. The sequence is clinically intuitive: identify the right patient, administer targeted therapy and use subsequent imaging and biomarkers to follow disease. Novartis reported full-year 2025 Pluvicto sales of $1.994 billion, up 42% at constant currencies. That commercial performance shows that nuclear medicine can become a global blockbuster business rather than remain a specialist diagnostic niche. Lantheus is strongest at the diagnostic gateway. Curium has upstream capabilities and therapeutic assets including 177Lu-PSMA-I&T. If the merger closes and integration works, Curium may be able to connect isotope supply and therapeutic development to Lantheus' established U.S. PSMA imaging network. Lantheus, in turn, may gain a more credible path from imaging revenue toward therapy economics. This is the meaning of Lantheus' "Find, Fight and Follow" positioning. Find the patient through imaging. Fight the disease with a targeted therapy. Follow the patient with continued imaging and disease management. The value lies in owning more than one transaction across that cycle. Execution, however, is not automatic. Diagnostic adoption does not guarantee that a therapeutic asset will succeed clinically or win approval. Hospital relationships do not eliminate manufacturing constraints. The strategic fit is strong, but the therapeutic and integration risks remain real. ## 04 | The competitive rule is changing from pipeline assembly to full-chain integration Recent radiopharma M&A has often looked like a set of puzzle pieces being added to Big Pharma. Bristol Myers Squibb acquired RayzeBio and its actinium-based radiopharmaceutical platform. AstraZeneca acquired Fusion Pharmaceuticals to add next-generation radioconjugates and actinium-225 capabilities. Lilly acquired POINT Biopharma to enter clinical and preclinical radioligand therapy. Those transactions treated radiopharma as a new oncology modality. Curium-Lantheus looks more like consolidation within the modality itself. It brings isotope supply, radiopharmaceutical manufacturing, diagnostics, U.S. hospital access, therapeutic programs and commercialization infrastructure into one corporate structure. A company that owns only one candidate may find it increasingly difficult to build a durable advantage. The more demanding questions are operational. Where will the isotope come from? Which facilities can manufacture the dose? How wide is the practical distribution radius? Will nuclear-medicine departments adopt the workflow? Can diagnostics establish a patient funnel before therapy launches? Can the therapeutic product create recurring economics? Can follow-up imaging deepen physician and patient relationships? Traditional drugmakers do not always possess these capabilities. For a radiopharma company, they are fundamental. ## 05 | Taiwan has few direct radiopharma equities, but it does have imaging and nuclear-medicine infrastructure Taiwan's listed market contains few companies that directly resemble Curium, Lantheus or Novartis' radioligand-therapy franchise. The local ecosystem is distributed across nuclear pharmacies, radiopharmaceutical manufacturing and delivery, hospital nuclear-medicine departments, imaging software and interpretation tools. Taiwan Global Medical Solutions, or GMS, is one example of infrastructure close to the physical radiopharma chain. The company describes a nationwide network for nuclear-medicine pharmacy and logistics, and its Taichung plant has passed PIC/S GMP inspection. Taiwan's FDA lists the facility as approved to manufacture sterile radiopharmaceutical injections. This is the unglamorous but indispensable part of nuclear medicine: a product must not only be made; it must arrive on time and remain usable. Among listed companies, the more defensible observation is imaging and information infrastructure—not a claim that Taiwan already has its own Lantheus. Acer Medical, ticker 6841 in Taiwan, has a TFDA-cleared computer-aided detection platform for brain dopamine-transporter SPECT imaging. It uses AI to analyze Tc-99m dopamine-transporter scans and therefore sits at the intersection of nuclear imaging and assisted interpretation. It is not a radiopharmaceutical company. EBM Technologies, ticker 8409, provides PACS, RIS, DICOM, FHIR and AI-imaging integration. As theranostics moves from isolated scans toward longitudinal patient management, image flow, cross-institution exchange, quantitative analysis and workflow integration may become more important. There is currently no public evidence that the Curium-Lantheus transaction itself creates a specific order or revenue stream for EBM. The investment boundary is important. A thematic connection to nuclear imaging is not the same as direct transaction exposure, a commercial contract or a radiopharma pipeline. ## Conclusion | Radiopharma is no longer a one-pipeline competition The Curium-Lantheus agreement matters because it makes the next stage of nuclear medicine visible. Radiopharmaceuticals have half-life constraints, radiation-handling requirements, specialized manufacturing and distribution, and hospital workflows that must coordinate diagnosis with therapy. Future competition will not be determined only by who owns a promising drug candidate. It will also depend on who can use diagnostics to enter hospitals, secure and manufacture isotopes, connect therapy to an existing patient funnel and retain the relationship through follow-up imaging. That is the larger message behind the transaction. Radiopharma is moving from competition among individual molecules toward competition among full-chain systems. For any company seeking to enter the field, the question is no longer just whether the drug works. The company must also show that it can deliver the dose to the patient safely, reliably and on time. --- Facts checked through August 24, 2026. The transaction remains subject to shareholder approval, regulatory clearances and other closing conditions, and the CVR milestones may not be achieved. Investigational products carry clinical, regulatory and commercial uncertainty. This article is provided for industry research and education only; it does not constitute investment, medical, fundraising or securities advice. --- ## 核藥史上最大併購：Curium 與 Lantheus 合併，改寫的不只是交易金額 URL: https://drugnews.com.tw/articles/2026-08-24-curium-lantheus-definitive-merger.html Date: 2026-08-24 Language: zh-Hant Summary: 6 月官網追蹤的 Curium–Lantheus 70 億美元收購傳聞，已在 8 月 3 日推進為最終合併協議：每股 102.50 美元現金加最高 12 美元 CVR，總交易價值最高約 80 億美元。真正被整合的是核素、製造、配送、診斷、治療與美國院端網絡。 ## 00｜從 70 億美元傳聞，走到最高 80 億美元最終協議 核藥產業的競爭，正在進入新的階段。 2026 年 8 月 3 日，Curium 與 Lantheus Holdings 宣布簽署最終合併協議。依協議，Lantheus 股東在交易完成時可取得每股 102.50 美元現金，另有最高每股 12 美元、綁定特定商業里程碑的或有價值權利（CVR）；總交易價值最高約 80 億美元。 這裡先把狀態說清楚：**簽署最終協議不等於交易已完成**。交易仍須經 Lantheus 股東表決、監管核准與其他成交條件；CVR 對應的里程碑也不保證達成。 [藥時事 6 月 9 日的早期文章](https://drugnews.com.tw/articles/2026-06-09-curium-lantheus-radiopharma-acquisition.html)討論的是 Curium 擬以約 70 億美元收購 Lantheus 的傳聞與產業邏輯。本篇則是 8 月 3 日最終協議公布後的後續：價格、交易結構與雙方整合方向都已進一步明確，卻仍不是已交割併購。 這筆交易真正值得看的，不是 80 億美元本身，而是核藥產業的併購邏輯正在改變。過去幾年，Eli Lilly 收購 POINT Biopharma、Bristol Myers Squibb 收購 RayzeBio、AstraZeneca 收購 Fusion Pharmaceuticals，主要是在大型藥廠的腫瘤版圖中補進一個放射性藥物平台、一組候選管線或特定核素技術。 Curium 與 Lantheus 不太一樣。這不是大型藥廠買一條管線，而是兩家核藥公司把上游同位素、製造、配送、診斷產品、美國醫院通路與治療資產直接接起來。核藥競爭已經不再只是誰有一個好分子；真正的關鍵，開始變成誰能掌握整條診療一體化產業鏈。 ## 01｜Curium 缺的是美國終端，Lantheus 缺的是上游與治療端 Curium 的強項在上游。公司長期專注核醫藥物的開發、製造與供應，擁有大型垂直整合製造與配送網絡。交易公告指出，合併後公司將服務超過 70 個國家的腫瘤、神經與心血管患者。 Curium 也正在推進 177Lu-PSMA-I&T。這是一款 PSMA 靶向放射性配體治療候選藥，用於轉移性去勢抗性前列腺癌；ECLIPSE 是由 Curium US 贊助的隨機 Phase 3 研究，ClinicalTrials.gov 登記號為 NCT05204927。 Curium 的短板也很明顯：它有核素、製造、全球配送與治療端資產，卻缺少 Lantheus 在美國已經跑通的診斷產品與醫院端商業網絡。 Lantheus 的強項剛好相反。PYLARIFY（piflufolastat F 18）是美國市場領先的 PSMA PET 前列腺癌影像產品。Lantheus 在 2026 年 3 月公布 PYLARIFY TruVu 獲 FDA 核准時表示，自 2021 年原產品獲准以來，已有超過 76 萬名患者接受 PYLARIFY PSMA PET 掃描。TruVu 新配方的重點，則是支援更大批次與更高患者可近性；公司規劃在 2026 年第四季分階段推出。 此外，Lantheus 還有心臟超音波造影劑 DEFINITY，以及 β-amyloid PET 顯像劑 Neuraceq（florbetaben F 18）。公司 2025 年股東資料指出，Neuraceq 在美國已是使用量第二、成長最快的 amyloid PET 顯像劑。 所以這筆交易的互補性很清楚：Curium 有上游核素、製造、配送與治療端；Lantheus 有診斷產品與美國市場入口。兩家公司合在一起，補的不是單一產品，而是核藥產業最難補的一整段鏈條。 ## 02｜核藥為什麼特殊？因為半衰期決定商業模式 核藥和一般藥物最大的不同，在於它不是只看藥效，還要看時間。 許多放射性核素半衰期很短，無法像一般小分子或抗體藥一樣長期庫存、再慢慢送往遠方。核藥從核素生產、放射標記、GMP 製造、品質檢測、屏蔽運輸到醫院排程，每一步都受到時間限制。 這也是為什麼核藥的競爭壁壘從一開始就不只是研發。企業需要穩定的同位素來源、符合規範的無菌製造、放射性物質運輸經驗、核醫科合作網絡，以及把正確活度的產品在正確時間送到正確患者身上的調度能力。 治療用核藥更難。以 Pluvicto（lutetium Lu 177 vipivotide tetraxetan）這類放射性配體治療為例，產品製造與配送必須配合患者排程；半衰期、批次品質、醫院治療時間與患者狀態都會影響最終使用。核藥的核心護城河不是一個分子，而是能不能把分子、核素、製造、配送、診斷、治療與院端流程串成穩定系統。 Curium 與 Lantheus 的合併，正是朝這個系統化方向前進。 ## 03｜診療一體化，才是核藥最有價值的商業閉環 核藥最有價值的特性之一，是同一個生物靶點可以連接診斷與治療。先用診斷性核素做 PET 或 SPECT 顯像，找出表達目標靶點的患者；再用治療性核素把放射能量送到腫瘤附近。這就是 theranostics，也就是診療一體化。 前列腺癌 PSMA 是目前最成熟的例子。Novartis 以 Locametz（gallium Ga 68 gozetotide）做 PSMA PET 顯像，再以 Pluvicto 治療 PSMA-positive 前列腺癌，形成「先顯像篩選、再精準治療」的模式。Novartis 公布的 2025 年全年 Pluvicto 銷售額為 19.94 億美元，年增 42%，證明核藥可以做成全球重磅產品。 Lantheus 的主要優勢仍在診斷端；Curium 則擁有 177Lu-PSMA-I&T 等治療端資產。合併後，Curium 有機會把上游核素與治療管線接到 Lantheus 已建立的美國 PSMA 診斷網絡；Lantheus 也有機會從影像診斷進一步接近治療端收入。 這不是單純併購產品，而是把「Find, Fight, Follow」變成商業閉環：先找到病人，再治療病人，最後用影像追蹤病人。 ## 04｜競爭規則改寫：從單點管線，走向全鏈條整合 過去幾年的核藥交易，多半像大型藥廠補拼圖：BMS 買 RayzeBio，取得 actinium-based radiopharmaceutical 平台；AstraZeneca 買 Fusion Pharmaceuticals，補進 next-generation radioconjugates 與 actinium-225 能力；Lilly 買 POINT Biopharma，切入臨床與臨床前放射性配體治療管線。 這些交易的共同點，是把核藥當成腫瘤版圖中的新 modality。Curium 與 Lantheus 的交易更像產業升級：它把同位素供應、放射性藥物製造、診斷顯像、醫院准入、治療管線與商業通路放進同一體系。 未來只靠一條候選分子，可能很難形成長期優勢。真正有價值的公司，要同時回答幾個問題：核素從哪裡來？製造基地在哪裡？配送半徑多大？醫院核醫科願不願意用？診斷端能不能建立患者入口？治療端能不能形成收入？後續影像追蹤能不能留下資料與醫師黏著度？ 這些問題，傳統藥廠不一定熟悉，卻是核藥公司生死攸關的基本功。 ## 05｜台灣直接核藥個股少，但影像與核醫基礎設施值得觀察 台灣上市櫃市場裡，直接對標 Curium、Lantheus 或 Novartis 放射性配體治療的公司不多。核醫藥物產業更多分布在核醫藥局、放射性藥品製造與配送、醫院核醫部、影像資訊與判讀工具。 臺灣新吉美碩長期布局核醫藥品調製、製造與物流。公司資料顯示，其運輸網絡覆蓋台灣各地，臺中藥廠通過 PIC/S GMP；TFDA 也列出該廠核定製造放射性無菌注射液。這類能力非常貼近核藥「做得出來，也要準時送得到」的本質。 上市櫃公司中，可觀察的是核醫影像與資訊基礎設施，而不是把任何公司硬說成下一個 Lantheus。 第一個是長佳智能（6841）。TFDA 公布的 AI／機器學習醫材許可資料列有「長佳智能腦部多巴胺轉運體斷層掃描電腦輔助偵測平台」。它分析 Tc-99m 多巴胺轉運體掃描影像，站在核醫影像判讀與 AI 輔助分析的交會點；但它不是核藥本身。 第二個是商之器（8409）。公司長期提供 PACS、RIS、DICOM、FHIR 與 AI 醫療影像整合。當核藥從單點檢查走向診療一體化，影像資料流、跨院交換、定量分析與院端流程整合都會變得更重要；但目前沒有公開證據顯示本次 Curium–Lantheus 交易直接帶給商之器特定訂單。 ## 結語｜核藥競爭，已經不是一條管線的競爭 Curium 與 Lantheus 簽署最終合併協議，之所以重要，是因為它把核藥產業的下一階段講得很清楚。 核藥有半衰期限制、放射性物質管理、製造配送難題與醫院核醫科流程，診斷與治療還必須互相配合。未來的競爭不會只看誰有一款候選藥，而會看誰能先用診斷產品打進醫院、穩定供應核素與藥物、把治療產品接上患者入口，再以影像追蹤形成長期閉環。 這才是 Curium 與 Lantheus 交易背後最大的啟示：核藥產業正在從「單點分子競爭」，進入「全產業鏈系統戰」。對任何想進入核藥的公司來說，問題不只是藥有沒有用，還要問有沒有能力把它準時、穩定、安全地送到病人身上。 --- 資料查核至 2026 年 8 月 24 日。交易仍待股東表決、監管核准與其他成交條件；CVR 里程碑可能無法達成。研究中候選藥的臨床、監管與商業結果均有不確定性。 --- ## Who Could Be Big Pharma's Next Major Biotech Acquisition? URL: https://drugnews.com.tw/articles/2026-08-23-big-pharma-biotech-ma-next-targets-2026-en.html Date: 2026-08-23 Language: en Summary: Biopharma acquisition capital in 2026 is clustering around a narrow window: human risk has started to fall, but strategic value is not yet fully priced. Alkermes, Dyne, First Tracks, Legend and Sail illustrate five ways a biotech can enter that window, while Taiwan's Pharmosa Biopharm and Belite Bio show how late-stage assets can speak the same global language. The global biotech M&A market is making one preference unusually clear in 2026. Large pharmaceutical companies still have cash, and the need to refill portfolios ahead of major patent expirations has not disappeared. What has changed is the distribution of that capital. Buyers are no longer spreading it evenly across the biotech universe. They are concentrating on a much smaller group of companies positioned at a particular point on the risk curve: clinical uncertainty has begun to fall, while commercial and strategic value has not yet been fully priced. Big Pharma is not out of money. It has simply become more selective. The biotechs that enter an acquisition conversation often share several features. Their lead asset has produced an initial human signal. Their therapeutic field has already been validated by a large transaction. The product can extend into more than one indication. The platform fills a specific gap in a buyer's portfolio. Above all, the timing cannot be too early or too late. Too early, and the science remains too risky. Too late, and the asset may already be prohibitively expensive. That is why biotech M&A in 2026 increasingly looks like a competition over timing. ## 01 | Recent deals have rewritten the acquisition standard Several transactions over the past year show what large pharmaceutical companies now want. Novartis acquired Avidity Biosciences for approximately $12 billion, adding a muscle-targeted antibody oligonucleotide conjugate platform and several late-stage neuromuscular programs. Novartis framed the acquisition as a route to multiple potential launches before 2030 and access to multibillion-dollar market opportunities. Gilead Sciences paid approximately $7.8 billion for its long-term partner Arcellx, bringing the BCMA CAR-T therapy anitocabtagene autoleucel, or anito-cel, fully into the Kite cell-therapy organization. The sequence is instructive: collaborate first, observe the data and execution, and acquire the company once the strategic fit has been de-risked. Eli Lilly has also remained active. It agreed to acquire Centessa Pharmaceuticals for roughly $6.3 billion upfront plus a contingent value right of up to approximately $1.5 billion, gaining a sleep-wake pipeline centered on OX2R agonism. Lilly then moved further into neuroscience with an agreement to buy AtaiBeckley for about $2.8 billion upfront plus a CVR worth up to $1 billion, adding new approaches to treatment-resistant depression. Immunology is attracting similar premiums. argenx agreed to acquire Forte Biosciences for approximately $2.2 billion, adding the anti-CD122 antibody FB102. The transaction followed positive Phase 1b vitiligo data. It showed that a leading immunology company was willing to pay for an early asset when the mechanism was coherent and the indication expansion opportunity was broad. The common standard is straightforward. A low price is not enough, and a fashionable target is not enough. Buyers want an initial human signal, credible indication expansion, a clean fit with existing capabilities and a reason to believe that waiting will make the asset more expensive. ## 02 | As orexin heats up, Alkermes occupies a more sensitive position Lilly's acquisition of Centessa makes Alkermes strategically more visible because of one mechanism: orexin. Orexins are hypothalamic neuropeptides that play a central role in maintaining wakefulness. Traditional sleep medicines generally help patients fall or remain asleep. An OX2R agonist works in the opposite direction: it activates wake-promoting circuitry. The therapeutic opportunity includes narcolepsy type 1, narcolepsy type 2 and idiopathic hypersomnia. Alkermes' alixorexton is an oral, selective OX2R agonist in development across all three conditions. The US Food and Drug Administration has granted Breakthrough Therapy designation for alixorexton in narcolepsy type 1. In the Phase 2 Vibrance-2 study in narcolepsy type 2, the drug produced statistically significant and clinically meaningful improvements on both the Maintenance of Wakefulness Test and the Epworth Sleepiness Scale. Alkermes has since initiated the global Phase 3 Brilliance program in narcolepsy types 1 and 2. That creates two layers of value. First, Alkermes has a core OX2R asset in Phase 3. Second, the company is not confining the mechanism to rare sleep disorders. It is also exploring ALKS 7290 in adults with attention-deficit/hyperactivity disorder. If OX2R agonism can move from narcolepsy to idiopathic hypersomnia and eventually into larger indications such as ADHD, the proposition changes. It is no longer a single-product sleep-disorder story; it becomes a neuroscience platform. For a large pharmaceutical company, the pattern of beginning in a focused indication and expanding into larger markets can be especially attractive. ## 03 | After Avidity, Dyne is the most visible independent muscle-delivery platform Novartis' acquisition of Avidity naturally raises a follow-on question: which independent company now represents the next muscle-targeted oligonucleotide platform? Dyne Therapeutics is the most obvious name. Dyne's platform uses an antibody-based targeting component to deliver oligonucleotide therapies to muscle, with programs in Duchenne muscular dystrophy, myotonic dystrophy type 1 and facioscapulohumeral muscular dystrophy. Its most commercially advanced asset is zeleciment rostudirsen, also known as z-rostudirsen or DYNE-251, for patients with DMD who are amenable to exon 51 skipping. Dyne submitted a biologics license application seeking accelerated approval in May 2026. The FDA accepted the filing and granted Priority Review. This is close to the timing window favored by strategic buyers. The product is not yet fully commercial, but it is approaching a regulatory decision. Meaningful risk remains, yet the greatest uncertainties are lower than they were when the platform was purely early-stage. Dyne also extends beyond DMD. Its DM1 and FSHD programs matter because a buyer would not be purchasing only an exon-skipping drug. It would be acquiring a muscle-delivery platform. The approximately $12 billion Avidity transaction created a reference price and validated the strategic importance of that technology class. Dyne is therefore likely to be evaluated against a different benchmark than it was before that deal. ## 04 | Once CD122 became transactible, First Tracks became harder to ignore Immunology has long been one of the areas where Big Pharma is most willing to spend. In the summer of 2026, CD122 suddenly became a transaction-backed mechanism. argenx's purchase of Forte after early FB102 data in vitiligo signaled that the IL-2/IL-15 receptor beta chain could become a meaningful new immunology entry point. Following that deal, First Tracks Biotherapeutics stands out as an independent company working on the same target class. First Tracks' ANB033 is an anti-CD122 antagonist directed at the shared beta subunit of the IL-2 and IL-15 receptors. The program is in Phase 1b development in celiac disease and eosinophilic esophagitis. The company's second-quarter 2026 update guides to topline data from the first celiac cohort in the fourth quarter of 2026, the second celiac cohort in the first quarter of 2027, and eosinophilic esophagitis data in the third quarter of 2027. The company has two additional immunology programs. Rosnilimab is designed to selectively deplete pathogenic T cells and has completed a Phase 2b trial in rheumatoid arthritis. ANB101 is an earlier program designed to modulate plasmacytoid dendritic cells through BDCA2. First Tracks belongs on an M&A watch list for a simple reason. If ANB033 establishes human proof of concept in a second independent indication, CD122 will no longer be a story supported by a single acquired company. It will begin to look like a competitive immunology platform with multiple strategic entry points. ## 05 | Legend Biotech is now worth more than Carvykti alone Legend Biotech represents a different acquisition logic. This is not an early platform built entirely on clinical promise. It already has a commercially successful CAR-T product. Carvykti, or ciltacabtagene autoleucel, generated $597 million in global net sales in the first quarter of 2026, a 62% increase year over year. Growth outside the United States was even faster. The product had reached 18 markets and more than 300 treatment centers. Legend also reported a 99% manufacturing success rate and said more than 95% of orders were released on or before their final product delivery date. The asset that expands Legend's strategic imagination, however, is LB2501. LB2501 is an in vivo CD19/CD20 dual-targeting CAR-T candidate. Conventional autologous CAR-T therapy requires a patient's T cells to be collected, engineered outside the body and reinfused. In vivo CAR-T attempts to reprogram the cells inside the patient, potentially reducing the complexity of individualized manufacturing. In June 2026, Legend reported preliminary Phase 1 data. Among six patients with relapsed or refractory B-cell non-Hodgkin lymphoma treated at the higher dose level, the objective response rate was 100% and the complete response rate was 83.3%; all responses were ongoing at the data cutoff. Six patients are far too few to declare clinical success. That limitation must remain explicit. Yet the strategic signal is already visible: Legend combines an established CAR-T revenue stream with a potential next-generation platform that could move cell therapy from ex vivo manufacturing toward in vivo engineering. That pairing—mature revenue plus a next-generation platform—is precisely the type of structure that can draw large pharmaceutical interest. ## 06 | J&J has already secured an option on Sail Some transactions begin well before a full acquisition. Sail Biomedicines is a useful example. In July 2026, Johnson & Johnson entered a strategic collaboration with Sail to develop in vivo CAR-T therapies for immune-mediated diseases. J&J agreed to provide a total of $785 million in upfront consideration, including a $465 million equity investment, and received an exclusive option to acquire Sail for $2.58 billion. This structure deserves attention. Big Pharma does not always need to buy a platform outright at the start. It can invest, collaborate, gain an internal observation window and secure a future purchase right. If the platform fails, the buyer has avoided paying the full acquisition price immediately. If human data validate the approach, the buyer is already inside the relationship and rivals will find it much harder to intervene. Option-based collaborations may therefore become a recurring transaction template for high-potential platform biotechs. The buyer controls timing; the biotech secures development capital without giving up the company on day one. ## 07 | Which Taiwan companies speak the new acquisition language? Bringing the discussion back to Taiwan requires discipline. The useful question is not which local company will be acquired. It is which companies have assets that global buyers can understand using the same strategic language. One example is Pharmosa Biopharm, listed in Taiwan under ticker 6875. Its L606 program is a liposomal treprostinil inhalation suspension for pulmonary arterial hypertension and pulmonary hypertension associated with interstitial lung disease. L606 has been licensed to Liquidia for North America and is being evaluated in the global Phase 3 Re-Spire trial in PH-ILD. It combines a focused specialty market, differentiated formulation, late-stage clinical development and an international partnership—four attributes that can make an asset legible to a global portfolio owner. A second example is Lin BioScience, listed under ticker 6696, through its subsidiary Belite Bio. Belite's tinlarebant, also known as LBS-008, is being developed for Stargardt disease and geographic atrophy. Belite has completed the rolling submission of a US new drug application for tinlarebant in Stargardt disease, following positive topline results from the Phase 3 DRAGON trial. Rare ophthalmic diseases can offer concentrated clinical pathways, clearly defined patient needs and globally valuable regulatory milestones. Pharmosa Biopharm and Lin BioScience are not direct comparators to Alkermes, Dyne or Legend. Their relevance lies elsewhere. Both illustrate how Taiwanese companies can create value through specialty diseases, late-stage evidence, regulatory execution and international partnerships rather than through narrative momentum alone. Global buyers do not purchase attractive stories. They purchase assets that can be modeled, developed and integrated into a worldwide product portfolio. ## Conclusion | The next major deal may happen where risk has just started to fall The most likely acquisition candidates in the second half of 2026 are not necessarily the cheapest biotechs or the companies attached to the most fashionable platform. The more consequential candidates tend to sit where several conditions intersect: human proof of concept has emerged; the platform can expand across indications; commercialization is within sight; a large transaction has established a price reference for the field; and a specific buyer can integrate the asset into its clinical, regulatory or commercial system. Alkermes represents OX2R agonism and sleep-wake disorders. Dyne represents muscle-targeted oligonucleotide delivery. First Tracks represents CD122 immunomodulation. Legend combines mature CAR-T revenue with a next-generation in vivo CAR-T platform. Sail illustrates how a pharmaceutical company can secure a strategic position through an acquisition option before committing to full ownership. None of these companies is certain to be acquired, and none of the investigational programs is certain to succeed. Their shared feature is more precise: they occupy the point that Big Pharma now watches most closely, where risk is beginning to fall but future value has not been completely realized. For a large pharmaceutical company, the best acquisition is rarely the cheapest company. It is the opportunity to secure several years of future growth before the rest of the market has fully priced it. ## Disclaimer The acquisition candidates and transaction structures discussed here are industry observations based on public information. They do not imply that any company will be acquired, any clinical trial will succeed, or any investigational therapy has been proven safe or effective. This article is intended for industry research and knowledge sharing only. It does not constitute investment, medical, fundraising, or individual stock advice. --- ## Biotech 下一筆重磅併購，會落在誰身上？ URL: https://drugnews.com.tw/articles/2026-08-23-big-pharma-biotech-ma-next-targets-2026.html Date: 2026-08-23 Language: zh-Hant Summary: 2026 年的生技併購資金，正集中到臨床風險開始下降、商業價值尚未完全定價的公司。從 Alkermes、Dyne、First Tracks、Legend 到 Sail，再回看國邑與仁新，真正值得追的是人體概念驗證、平台外延、商業化距離與買方策略契合度。 2026 年的全球 Biotech 併購市場，出現一個很清楚的變化。 大型藥廠手上仍有現金，專利懸崖帶來的產品補洞壓力也沒有消失。但錢不再平均灑向所有 Biotech，而是集中流向一小群「剛好站在臨床風險下降、商業價值還沒完全定價」的公司。 換句話說，Big Pharma 不是沒錢，只是變得更挑。 能進入收購視野的 Biotech，通常有幾個共同特徵：核心資產已經有初步人體數據；所處領域已有大型交易驗證；產品可以擴展到多個適應症；技術平台能補足買方現有版圖；更重要的是，時間點不能太早，也不能太晚。 太早，科學風險太高；太晚，價格已經貴到難以下手。這也是為什麼 2026 年的併購交易，越來越像一場對「時點」的競賽。 ## 01｜過去一年的交易，已經把新標準寫出來 過去一年幾筆交易，幾乎把大型藥廠的新審美說得很清楚。 Novartis 以約 120 億美元收購 Avidity Biosciences，取得肌肉靶向 Antibody Oligonucleotide Conjugate（AOC）平台，以及多個後期神經肌肉疾病管線。Novartis 表示，這筆交易可望在 2030 年前帶來多個產品上市機會，並打開數十億美元級市場。 Gilead Sciences 則以約 78 億美元收購長期合作夥伴 Arcellx，把 anito-cel（anitocabtagene autoleucel）這款 BCMA CAR-T 完全收進自家 Kite 細胞治療體系。這是一種典型的「合作觀察後完整併購」：先合作、看資料、看執行，再決定把資產整合進來。 Eli Lilly 也持續出手。它以約 63 億美元前期對價、另附最高約 15 億美元 CVR 收購 Centessa Pharmaceuticals，取得以 OX2R agonist 為核心的睡眠覺醒障礙管線；之後又以約 28 億美元前期對價、最高 10 億美元 CVR 收購 AtaiBeckley，進一步加碼治療抗性憂鬱症與精神醫學新機制。 免疫領域也在升溫。argenx 宣布以約 22 億美元收購 Forte Biosciences，取得 anti-CD122 antibody FB102。這筆交易發生在 Forte 公布白斑症 Phase 1b 積極資料之後，顯示大型免疫公司願意為早期但機制清楚、平台外延大的免疫資產支付高溢價。 這些交易的共通點很直接：不是便宜就買，也不是熱門就買。要有初步人體資料，要有多適應症外延，要能接進買方既有業務，還要讓買方相信：再晚一點，價格會更貴。 ## 02｜Orexin 升溫後，Alkermes 的位置變得更敏感 Centessa 被 Lilly 收走後，Alkermes 的戰略位置明顯變得更突出。原因在於 orexin。 Orexin，也就是食慾素，是由下丘腦產生的神經胜肽，對維持清醒狀態非常關鍵。過去睡眠藥物多半是讓人睡得著，但 OX2R agonist 的邏輯反過來：直接增強清醒相關神經通路，用於 narcolepsy type 1、narcolepsy type 2、idiopathic hypersomnia 等疾病。 Alkermes 的 alixorexton 是口服選擇性 OX2R agonist，目前已經在 narcolepsy type 1、narcolepsy type 2 和 idiopathic hypersomnia 中推進開發。FDA 已授予 alixorexton 治療 narcolepsy type 1 的 Breakthrough Therapy designation。更重要的是，alixorexton 在 narcolepsy type 2 的 Phase 2 Vibrance-2 研究中，於 Maintenance of Wakefulness Test（MWT）與 Epworth Sleepiness Scale（ESS）兩個主要終點上，都展現統計顯著且具臨床意義的改善。Alkermes 也已啟動針對 narcolepsy type 1 與 type 2 的全球 Phase 3 Brilliance 計畫。 這讓 Alkermes 具備兩層價值。第一，它有一條已進入 Phase 3 的 OX2R 核心資產；第二，公司沒有把 OX2R 限定在罕見睡眠疾病，還在探索 ALKS 7290 用於成人 ADHD。 若 OX2R 能從 narcolepsy 擴展到 idiopathic hypersomnia，再往 ADHD 等更大市場延伸，這就不再是單一產品故事，而是神經科學平台故事。對大型藥廠來說，這種「小適應症起步、大適應症外延」的結構很有吸引力。 ## 03｜Avidity 之後，Dyne 成為肌肉靶向核酸藥物最醒目的獨立標的 Avidity 被 Novartis 收購後，市場自然會問：下一個肌肉靶向核酸藥物標的會是誰？最容易被看見的，是 Dyne Therapeutics。 Dyne 的核心技術，是把寡核苷酸藥物透過抗體導向肌肉細胞，處理 Duchenne muscular dystrophy（DMD）、myotonic dystrophy type 1（DM1）與 FSHD 等神經肌肉疾病。最接近商業化的是 zeleciment rostudirsen（z-rostudirsen，DYNE-251），用於適合 exon 51 skipping 的 DMD 患者。2026 年 5 月，Dyne 向 FDA 提交 BLA，尋求 accelerated approval；FDA 隨後接受申請並給予 priority review。 這種狀態非常符合併購市場喜歡的時間點。產品還沒完全商業化，但已經接近監管決策。風險還在，但最大的不確定性已經比早期平台低很多。 Dyne 也不是只有 DMD。它還有 DM1、FSHD 等管線，這正是買方在意的地方：它買到的不是一個 exon skipping 藥，而是一個肌肉遞送平台。Avidity 已經用 120 億美元交易為這條技術路線定錨，Dyne 自然會被市場重新估值。 ## 04｜CD122 變熱後，First Tracks 會被盯上 免疫學一直是 Big Pharma 最願意花錢的領域。2026 年夏天，CD122 突然成為交易熱點。 Forte 的 FB102 在白斑症早期資料後，被 argenx 以 22 億美元買走，代表市場開始相信 CD122 可能成為新的免疫調節入口。Forte 交易之後，同靶點上仍獨立的 First Tracks Biotherapeutics，就變得更值得注意。 First Tracks 的 ANB033 是 anti-CD122 antagonist，靶向 IL-2／IL-15 receptor shared beta subunit。公司管線資料顯示，ANB033 正在 celiac disease 與 eosinophilic esophagitis 中進行 Phase 1b 研究；celiac disease 的 Cohort 1 topline data 預計於 2026 年第四季讀出，Cohort 2 預計在 2027 年第一季，eosinophilic esophagitis 則預計 2027 年第三季讀出。 First Tracks 還有 rosnilimab 和 ANB101。Rosnilimab 目標是選擇性清除 pathogenic T cells，已完成 rheumatoid arthritis Phase 2b；ANB101 則透過 BDCA2 調節 plasmacytoid dendritic cells，目前處於早期開發階段。 這類公司最適合被放進「下一筆可能交易」清單。原因很簡單：如果 ANB033 在第二個獨立適應症中建立人體概念驗證，CD122 就不再只是 Forte 一家公司的故事，而會變成一條可被多家公司競逐的免疫平台路線。 ## 05｜Legend Biotech 的價值，已經不只 Carvykti Legend Biotech 是另一種併購邏輯。它已經不是只有早期平台與臨床想像，而是有一款真正商業化成功的 CAR-T 產品。 Carvykti（ciltacabtagene autoleucel）2026 年第一季全球淨銷售額達 5.97 億美元，年增 62%；美國以外市場成長更快，產品已進入 18 個市場、超過 300 個治療中心。Legend 也披露製造成功率達 99%，超過 95% 訂單能按最終產品交付日期準時釋放。 但真正讓 Legend 想像力變大的，是 LB2501。 LB2501 是 in vivo CD19／CD20 dual-targeting CAR-T。傳統自體 CAR-T 需要取出患者 T 細胞，在體外改造後再回輸；in vivo CAR-T 則試圖直接在患者體內完成 T 細胞改造，降低個體化製造的複雜度。2026 年 6 月，Legend 公布 LB2501 Phase 1 初步資料。在較高劑量組 6 名復發或難治性 B 細胞非何杰金氏淋巴瘤患者中，ORR 達 100%，CR 達 83.3%，且所有反應在資料截止時仍持續。 樣本數很小，遠遠不能宣告成功。但對併購市場來說，意義已經足夠清楚：Legend 不是只有已上市 CAR-T 收入，還有可能把細胞治療從 ex vivo 推向 in vivo 的下一代平台。 這種「成熟收入＋下一代平台」的組合，正是大型藥廠最容易動心的類型。 ## 06｜Sail 已經被 J&J 提前卡位 有些交易，不會等到完整併購才發生。Sail Biomedicines 就是例子。 2026 年 7 月，Johnson & Johnson 與 Sail 建立戰略合作，開發用於免疫介導疾病的 in vivo CAR-T。J&J 將支付總計 7.85 億美元初始款項，其中包含 4.65 億美元股權投資，並取得以 25.8 億美元收購 Sail 的獨家選擇權。 這種模式很值得注意。Big Pharma 不一定要一開始就全買。它可以先投資、先合作、先取得內部觀察窗口，並鎖住未來收購權。如果平台失敗，買方沒有立刻承擔完整收購成本；如果人體數據跑出來，買方已經站在門內，競爭對手很難中途搶走。這可能會成為未來高潛力平台型 Biotech 很常見的交易模板。 ## 07｜台灣誰符合「新併購語言」 這個題目放回台灣，不能簡單問哪家公司會被併購。更實際的問題是：哪些公司比較接近全球 Big Pharma 現在願意理解的資產語言？ 第一個可以看國邑（6875）。 國邑的 L606（Liposomal Treprostinil Inhalation Suspension）是吸入型微脂體 treprostinil，用於 pulmonary arterial hypertension（PAH）與 PH-ILD 等肺高壓相關疾病。L606 已授權給 Liquidia 推進北美市場，並啟動 PH-ILD 全球 Phase 3 Re-Spire 試驗。這是一個很典型的專科市場、差異化製劑、後期臨床與國際合作結構。 第二個是仁新（6696）。 仁新子公司 Belite Bio 的 tinlarebant（LBS-008）瞄準 Stargardt disease 與 geographic atrophy。公司公告指出，Belite 已完成向 FDA 滾動提交 tinlarebant 治療 Stargardt disease 的 NDA；先前 DRAGON 試驗也已公布 positive topline data。這類罕見眼科疾病，具備患者需求明確、臨床路徑集中、國際藥證價值高的特徵。 國邑和仁新不是 Alkermes、Dyne 或 Legend 的直接對標。但它們共同符合一個方向：不是靠題材熱度，而是靠專科疾病、後期資料、法規路徑與國際合作來建立價值。這才是台灣公司最值得學的地方。國際買方要的不是漂亮故事，而是能被放進全球產品組合裡計算的資產。 ## 結語｜下一筆重磅併購，會落在「風險剛下降」的公司身上 2026 年下半年的 Biotech 併購，最值得看的不一定是最低估值公司，也不一定是最熱門平台公司。 真正有機會被收購的，往往站在幾個條件交會處：人體概念驗證已經出現；平台能擴展到多個適應症；產品距離商業化不遠；所處領域已有大型交易定價；買方能把它接進既有臨床、法規或商業化體系。 Alkermes 代表 OX2R 與睡眠覺醒障礙；Dyne 代表肌肉靶向核酸遞送；First Tracks 代表 CD122 免疫調節；Legend 代表成熟 CAR-T 收入與 in vivo CAR-T 下一代平台；Sail 則代表大型藥廠提前卡位的選擇權模式。 這些公司不一定都會被買，也不一定都會成功，但它們都站在 Big Pharma 現在最敏感的位置：風險開始下降，價值尚未完全兌現。 對大型藥廠來說，最好的併購從來不是買最便宜的公司，而是在市場還沒完全定價前，先買下未來幾年的成長權。 ## 免責聲明 本文提及的併購候選與交易結構，僅為依公開資料所作的產業觀察，不代表任何公司必然被收購、任何試驗必然成功或任何藥物已證實安全有效。 --- ## BeOne Bets on Four RAS Drugs: The Real Asset Is Development Optionality URL: https://drugnews.com.tw/articles/2026-08-22-revolution-medicines-beone-four-ras-programs-en.html Date: 2026-08-22 Language: en Summary: BeOne is taking regional rights to four RAS(ON) programs, opening two candidate combination routes and funding one global registrational Phase 3. The strategic asset is not the sum of four drugs, but the optionality to decide where capital and development speed matter most. On August 10, BeOne Medicines and Revolution Medicines laid out a collaboration spanning four RAS(ON) programs, two BeOne combination assets, commercialization rights in selected Asian markets and one global registrational Phase 3 study. This is not an acquisition, and the global rights to all four drugs have not changed hands. BeOne is gaining a development and commercialization entry point in selected Asian markets, together with opportunities to test its own oncology assets alongside Revolution Medicines' RAS portfolio. Revolution Medicines retains rights outside the licensed territory while adding a partner that can fund and execute a global registrational Phase 3 trial. Four programs make the announcement look large. The more important question is which company is buying time for the other. ## 01 | Four RAS strategies enter at different stages RAS can be understood as a molecular switch inside a cancer cell. Oncogenic mutations can hold that switch in an active position, continuously transmitting growth signals downstream. Earlier drugs often trapped an inactive form of RAS. Revolution Medicines' RAS(ON) platform is designed to engage the active state that is driving the tumor. The four clinical programs covered by the collaboration are not equally mature. **Daraxonrasib (RMC-6236)** Daraxonrasib is a multi-selective RAS(ON) inhibitor intended to cover several common RAS genotypes. The US Food and Drug Administration accepted its new drug application on July 22, 2026, for previously treated metastatic pancreatic ductal adenocarcinoma. That filing is supported by data from a global randomized Phase 3 study. **Zoldonrasib (RMC-9805)** Zoldonrasib targets RAS G12D. Revolution Medicines has started the global Phase 3 RASolute 305 study, testing zoldonrasib with standard chemotherapy in previously untreated metastatic RAS G12D pancreatic cancer. **Elironrasib (RMC-6291)** Elironrasib targets RAS G12C and remains in early clinical development. The recruiting Phase 1b/2 study includes both monotherapy and a combination with daraxonrasib. **RMC-5127** RMC-5127 targets RAS G12V. The first patient was dosed in January 2026, and the program remains in Phase 1/1b. BeOne is therefore not taking in four equally ripe assets. One has reached FDA review, one is in Phase 3, and two are still defining dose, patient selection and the combinations most worth pursuing. ## 02 | The combinations are the most ambitious part of the deal The companies point to two BeOne assets as potential combination partners: - **BGB-58067**, an MTA-cooperative PRMT5 inhibitor designed around the dependence of MTAP-deleted tumors on PRMT5; and - **BG-T187**, an EGFR × MET × MET trispecific antibody with one EGFR-binding site and two MET-binding sites, intended to address upstream signaling and bypass activation. The wording of the announcement matters. Planned studies will combine BGB-58067 and BG-T187 with daraxonrasib “or” zoldonrasib. This is a pool of candidate combinations. The companies have not disclosed a fixed pairing for every asset, the tumor types, doses or start dates. It would be inaccurate to draw all four permutations as active programs. The scientific logic is straightforward. After RAS is suppressed, a tumor may reconnect through upstream signaling, bypass pathways or alternative survival programs. PRMT5 and EGFR/MET offer two different ways to surround that escape route. But combining two attractive targets does not automatically create a better therapy. Human data must establish whether efficacy can be added, toxicity remains manageable, and sequencing and dose produce a usable therapeutic window. At present, the companies have a mechanistic hypothesis and a collaboration framework, not evidence that the combinations work. ## 03 | The Asian rights look broad, but the disclosed boundary is narrow Revolution Medicines is granting BeOne exclusive rights to the four programs in “selected Asian markets.” Depending on the market, those rights may include development and commercialization or commercialization alone. The companies have not published the full country list. The agreement therefore cannot be described as covering all of Asia, and Taiwan cannot be assumed to be inside the territory. What is explicit is that Japan and South Korea remain outside the licensed territory, with Revolution Medicines retaining development and commercial rights in both countries. The financial terms leave an equally large blank space. Revolution Medicines is eligible for development milestones, sales milestones and tiered royalties on net sales in the collaboration territory. The upfront payment, aggregate milestone value and royalty rates were not disclosed. “Eligible to receive” is an important phrase. It describes payments that may occur if future conditions are met; it does not mean the money has already been received. ## 04 | The biggest unanswered question: which program gets the global Phase 3? The heaviest commitment in the agreement is BeOne's obligation to fund and conduct one global registrational Phase 3 study for one of Revolution Medicines' four RAS(ON) inhibitors. As of August 12, the companies had not identified the drug, tumor type, line of therapy or start date. That blank may be more revealing than the undisclosed transaction value. Daraxonrasib already has several global Phase 3 programs. Zoldonrasib has entered a first-line pancreatic-cancer Phase 3. Elironrasib and RMC-5127 are earlier. The final choice will show how the partners rank scientific risk, clinical speed and market opportunity. Revolution Medicines has not handed over global development. The announcement explicitly says the company will continue its broad global registrational program. BeOne is taking responsibility for one additional Phase 3 study and the agreed development or commercial work inside the collaboration territory. ## 05 | What did each company actually buy? For BeOne, the collaboration is an acceleration ticket in solid tumors. BeOne already has PRMT5 and EGFR/MET assets, cross-regional clinical execution and a commercial network across Asia. By connecting four RAS(ON) inhibitors, it gains regional rights, combination entry points and the opportunity to lead a global Phase 3 without building an entire RAS chemistry platform from zero. For Revolution Medicines, the exchange creates development capacity and geographic reach. The company retains rights outside the licensed territory while a partner funds and executes one global Phase 3. Milestones and royalties allow Revolution Medicines to participate in part of the value created in selected Asian markets. That structure may preserve internal capital for its other registrational studies while opening markets where its footprint is thinner. Drugnews' reading is that the central asset in this deal is optionality, not the simple sum of four drugs. Three items now deserve close attention: 1. **The monotherapy foundation.** Daraxonrasib's regulatory progress, zoldonrasib's Phase 3 trial, and whether the two earlier programs can produce durable safety and efficacy signals. 2. **The combination window.** Which pairings actually start, how patients are selected, and whether toxicity can be controlled. 3. **The global Phase 3 choice.** Which program is selected, how regional work is divided, and whether the collaboration moves faster than Revolution Medicines could on its own. Four drugs make the network look large. One selected Phase 3 will reveal where that network is actually tightening. ## Disclaimer All four drug candidates remain investigational. This collaboration does not establish their safety, efficacy or regulatory approval. This article provides biotech-industry analysis based on public information and does not constitute investment advice, medical advice or treatment guidance. --- ## 百濟神州一口氣押上 4 款 RAS 新藥 URL: https://drugnews.com.tw/articles/2026-08-22-revolution-medicines-beone-four-ras-programs.html Date: 2026-08-22 Language: zh-Hant Summary: 百濟神州一次接進四款成熟度不同的 RAS(ON) 資產、兩條候選聯用路線與一項全球註冊三期。真正值得追的不是四藥相加，而是區域權利、開發分工與選擇權如何替雙方搶時間。 8 月 10 日，百濟神州（BeOne Medicines）與 Revolution Medicines，把一張「四款 RAS(ON) 新藥、兩款聯用資產、亞洲商業化、全球三期」的合作網攤上桌。 這筆交易沒有買下一家公司，也沒有把四款藥的全球權利整批搬走。百濟神州拿到的，是部分亞洲市場的開發與商業化入口，加上把自家腫瘤資產接進 RAS 管線的試驗機會；Revolution Medicines 則保留區域外權利，同時多了一支能出資並執行全球註冊性三期的隊伍。 四款資產很搶眼；更關鍵的問題是，誰在替誰搶時間。 ## 01｜一次接進四種 RAS 戰法 RAS 可以想成癌細胞裡的分子開關。突變讓它長時間卡在啟動狀態，持續把生長訊號往下游送。早期藥物多半抓住關閉狀態；Revolution Medicines 的 RAS(ON) 平台，鎖定正在驅動腫瘤的活化狀態。 這次納入合作的四款臨床資產，成熟度並不相同： **daraxonrasib（RMC-6236）** 多重選擇性 RAS(ON) 抑制劑，可涵蓋多種常見 RAS 基因型。美國 FDA 已在 2026 年 7 月 22 日受理其用於既往治療過的轉移性胰臟導管腺癌 NDA，背後已有全球隨機三期資料。 **zoldonrasib（RMC-9805）** 鎖定 RAS G12D。公司已啟動全球三期 RASolute 305，測試它搭配標準化療，用於未治療的轉移性 RAS G12D 胰臟癌。 **elironrasib（RMC-6291）** 鎖定 RAS G12C。目前仍在早期臨床開發，登錄中的 Phase 1b/2 研究同時評估單藥，以及與 daraxonrasib 的聯用。 **RMC-5127** 鎖定 RAS G12V，2026 年 1 月才完成首位患者給藥，仍在 Phase 1/1b。 換句話說，百濟神州接進來的不是四顆同樣成熟的果實。裡面有一款已走到 FDA 審查、一款正在三期，也有兩款仍在尋找劑量、病人與最佳組合。 ## 02｜聯用才是這筆合作最有野心的地方 雙方公告把聯用方向指向百濟神州的兩款資產： - **BGB-58067**：MTA 協同型 PRMT5 抑制劑，瞄準 MTAP 缺失腫瘤對 PRMT5 的依賴。 - **BG-T187**：EGFR × MET × MET 三特異性抗體，以一個 EGFR 結合位置、兩個 MET 結合位置處理上游訊號與旁路活化。 公告寫得很精準：規劃中的研究，會讓 BGB-58067、BG-T187 搭配 daraxonrasib「或」zoldonrasib。這是一個候選聯用池，尚未公布每一項固定配對、癌別、劑量與啟動時程，不能自行畫成四種組合全部開跑。 科學邏輯不難理解。RAS 被壓住後，腫瘤可能透過上游訊號、旁路路徑或細胞存活機制重新接電。PRMT5 與 EGFR／MET 提供兩條不同的包抄路線。 但聯用從來不是把兩個好靶點放在一起就會自然變好。療效能否疊加、毒性能否承受、先後次序與劑量是否找得到窗口，都要由人體資料回答。現在有的是機制假說與合作框架，還沒有聯用成功的結論。 ## 03｜亞洲權利看起來很大，邊界其實很細 Revolution Medicines 授予百濟神州四款藥在「部分亞洲市場」的獨家權利。依不同市場，權利可能是開發加商業化，也可能只有商業化。 官方沒有公布完整國家清單，因此不能直接寫成「亞洲全拿」，也不能推定台灣就在授權範圍。已經確定的是，日本與韓國留在授權區外，仍由 Revolution Medicines 保有開發及商業權利。 財務條款也留了大片空白。公告只確認 Revolution Medicines 有資格取得開發里程碑、銷售里程碑，以及合作區域淨銷售額的分級權利金；首付款、里程碑總額與權利金級距都沒有公開。 這些字眼很重要。「有資格取得」代表未來達標才可能發生，不等於款項已經入帳。 ## 04｜全案最大的伏筆：哪一款進全球三期？ 合作裡最重的一句，是百濟神州將出資並執行一項全球註冊性 Phase 3，對象是 Revolution Medicines 四款 RAS(ON) 抑制劑其中之一。 截至 8 月 12 日，雙方沒有說明是哪一款藥、哪個癌別、哪一線治療，也沒有給出啟動時間。 這個空白比交易金額更值得追。daraxonrasib 已有多項全球三期，zoldonrasib 也已進入胰臟癌一線三期；elironrasib 與 RMC-5127 的成熟度更早。最終選擇會透露雙方如何排序科學風險、臨床速度與市場空間。 Revolution Medicines 並未把全球開發交出去。公告明寫，公司仍會推進自身廣泛的全球註冊研究；百濟神州承擔的是其中一項額外三期，以及合作區域內約定的開發或商業任務。 ## 05｜兩家公司各自買到什麼？ 對百濟神州來說，這是一張實體瘤加速券。 公司手上已有 PRMT5 與 EGFR／MET 資產，也有跨區域臨床執行與亞洲商業網路。接進四款 RAS(ON) 抑制劑後，它不必從零搭出一整套 RAS 化學平台，就能取得區域權利、聯用入口與一項全球三期的主導機會。 對 Revolution Medicines 來說，換到的是開發產能與地理覆蓋。它保留區域外權利，讓合作夥伴替一項全球三期出資執行，並用里程碑與權利金分享部分亞洲價值。這有機會把內部資源留給其他註冊研究，同時進入原本布局較薄的市場。 藥時事的判讀是：這筆合作的核心資產叫「選擇權」，不是四款藥的簡單加總。 接下來要追三件事： 1. **單藥底盤**：daraxonrasib 的監管進展、zoldonrasib 三期，以及兩款早期資產能否交出可延續的安全性與療效。 2. **聯用窗口**：最後啟動哪些配對、患者怎麼選、毒性是否壓得住。 3. **全球三期落點**：哪一款藥被選中，誰負責哪些地區與工作，時程能否比 Revolution Medicines 單打獨鬥更快。 四款藥讓合作看起來很大；一項被選中的三期，才會告訴市場這張網究竟往哪裡收。 ## 免責聲明 所有候選藥皆屬研究中產品，尚未因此次合作證明安全、有效或取得核准。本文為生技醫藥產業與公開資料整理，不構成投資、醫療或用藥建議。 --- ## Moderna Surges 176.97% After Its Personalized Cancer Vaccine Clears a Phase 3 Test URL: https://drugnews.com.tw/articles/2026-08-21-moderna-merck-intismeran-phase3-melanoma-en.html Date: 2026-08-21 Language: en Summary: Positive topline results from INTerpath-001 sent Moderna shares up 176.97% in one session. The deeper valuation question is whether patient-specific design, manufacturing, release and clinical delivery can become a scalable platform. Start with the boundary: this is not a cure for cancer, it is not evidence across every tumor type, and intismeran is not approved. What Moderna and Merck have reported is a positive Phase 3 topline result in patients with high-risk cutaneous melanoma whose tumors had already been completely removed. It is not a product that can be ordered for a patient tomorrow. On August 19, Moderna shares closed at $174.38, up $111.42, or 176.97%, in a single session. The arithmetic matters. Subtracting the $111.42 gain from the closing price gives a prior close of $62.96. Dividing $174.38 by $62.96 yields roughly 2.77. A 176.97% increase means the stock added 1.7697 times its starting value; it does not mean the stock merely reached 176.97% of that value. Those figures establish the violence of the market reaction. They do not by themselves establish an exact increase in market capitalization, which would require a consistent share-count and corporate-action basis. Drugnews therefore does not repeat the unverified claim that Moderna added $44.5 billion overnight, nor do we use an unverified volume figure to dramatize the move. The catalyst was a Phase 3 announcement from Moderna and Merck. Intismeran autogene, a personalized neoantigen mRNA therapy, met two endpoints in INTerpath-001 when combined with Keytruda, or pembrolizumab. The result applies to a specific setting: patients with completely resected, high-risk melanoma. In that setting, adding a patient-specific therapy to Keytruda delayed recurrence and distant metastasis versus Keytruda alone. A headline may ask whether this means goodbye to cancer. The evidence permits only one answer: far too soon. Why, then, did the market reprice Moderna by nearly threefold in one day? Because investors were buying more than one melanoma trial. The result raised expectations across four layers at once: the probability of success for the product, the possibility of extension to other tumors, Moderna's post-pandemic second growth engine, and the value of an operating system that can design and manufacture a different medicine for every patient. ## 01 | What did Phase 3 actually establish? INTerpath-001 enrolled 1,137 patients with completely resected stage IIB, IIC, III or IV cutaneous melanoma. Participants were randomized 2:1 to: - intismeran plus Keytruda; or - placebo plus Keytruda. The control arm was not untreated. The trial asked whether adding a personalized therapy could further reduce recurrence risk on top of established Keytruda adjuvant treatment. According to the prespecified interim topline announcement: - the primary endpoint of recurrence-free survival, or RFS, was met; - the key secondary endpoint of distant metastasis-free survival, or DMFS, was met; and - the safety profile was consistent with prior experience, with no new safety signal observed. That is enough to change the probability assigned to the asset. It is not enough to complete a valuation model. The companies have not disclosed the Phase 3 hazard ratios, confidence intervals, event counts, absolute differences or survival curves. Overall survival will continue to be followed. Regulatory discussions come next. In other words, we know that the study crossed its statistical gates. We do not yet know the size of the win. Intismeran also remains investigational and has not been approved by the US Food and Drug Administration. The market has already purchased part of the future. The full dataset will determine how much of that optimism survives. ## 02 | This is not an off-the-shelf vaccine A conventional vaccine identifies an antigen shared across a population and then produces one standardized product. Intismeran follows a different operating model. After surgery, tumor and blood samples are collected. DNA and RNA sequencing identifies mutations present in the cancer. An algorithm considers the patient's human leukocyte antigen, or HLA, type and selects neoantigens that are more likely to be recognized by T cells. Instructions for as many as 34 neoantigens are encoded in one mRNA construct, packaged in a lipid nanoparticle and returned to that individual patient. When the patient changes, the mutation set changes, and the drug sequence changes with it. That is the hard meaning of personalized. A hospital is not simply drawing a vial from inventory. It is initiating a new design, manufacturing and quality-release order for each patient. The vaccine component teaches the immune system what to recognize. Keytruda blocks the PD-1 checkpoint, giving activated T cells a better chance to continue attacking residual cancer cells. One numerical boundary is important. The 1,137 figure is total trial enrollment. It does not mean that 1,137 intismeran products were manufactured. Only the treatment arm received the patient-specific therapy in the 2:1 randomized study. ## 03 | Why might this cancer-vaccine approach be different? Cancer vaccines are not new. Earlier approaches often targeted tumor-associated antigens shared by many patients, and they repeatedly encountered three barriers. The first was immune tolerance. Some tumor-associated antigens are also present in normal tissue, so the immune system may be reluctant to attack them strongly. The second was tumor heterogeneity: cancer cells lacking a single target can escape. The third was the tumor microenvironment, which can suppress T cells even after they have been activated. Intismeran attempts to address all three pieces. It selects tumor-specific neoantigens, can encode up to 34 candidate targets and is combined with PD-1 blockade. The trial also places the intervention in the adjuvant setting after surgery, where the vaccine is confronting potential microscopic residual disease rather than a large, established tumor with a deeply suppressive microenvironment. These features make the result biologically understandable. They do not create a passport into every cancer type. Melanoma has a relatively high mutational burden and is often responsive to immune therapy, making it a comparatively favorable first proving ground. Colder and more complex tumors must still produce their own evidence. ## 04 | Why mRNA? Industrial control over designed variation Personalized therapy is not simply a traditional factory reduced to one miniature plant per patient. It changes the manufacturing problem. For most medicines, quality systems try to prevent one batch from differing from the next. With intismeran, the encoded content is supposed to differ. The operating question becomes whether many distinct designs can still be governed by common rules for purity, potency, safety, traceability and on-time release. mRNA provides an engineering handle for that controlled variation. The neoantigen combination can change while in-vitro transcription, purification, lipid-nanoparticle encapsulation and release testing remain modular and verifiable. The difficulty has not disappeared. It has shifted from rebuilding a process to reliably scheduling different orders through one validated system. The algorithm selects candidate neoantigens, manufacturing turns the design into a usable product, and the clinical trial determines whether it helps patients. No one layer can compensate for failure in another. The platform value therefore does not lie in how many sequences a company can write. It lies in how much variation the company can manage. Personalized therapy can move beyond a small number of research centers only if every design may differ while delivery time and quality promises remain stable. ## 05 | The market bought four distinct sources of value **First, the probability of success for the melanoma asset.** A large randomized Phase 3 study met both its primary and key secondary endpoints. That moves intismeran materially closer to a regulatory filing. The eventual effect size will still influence the eligible population, label, pricing and pace of adoption. **Second, the possibility that the platform can be repeated.** As of June 2026, the companies described nine Phase 2 or Phase 3 studies across melanoma, non-small-cell lung cancer, bladder cancer and renal-cell carcinoma. The first Phase 3 success raises the credibility of the broader program. But melanoma is also the most favorable first station. It tends to carry a high mutational burden and respond to immunotherapy, and this study treats lower disease burden after surgery. Lung, kidney and bladder cancer still need to submit their own scorecards. **Third, a post-pandemic second engine for Moderna.** Since the COVID-19 peak, investors have asked whether Moderna is mainly a respiratory-vaccine company or a platform capable of repeatedly producing programmable medicines. This Phase 3 outcome gives the second description its strongest clinical support so far. For Merck, it adds another potential growth engine to the Keytruda combination strategy and product life cycle. The companies have disclosed that they generally share relevant global development costs and profits equally. That does not mean every dollar of revenue will be divided mechanically in half. Manufacturing costs, expense allocation and commercial execution still shape the economics. **Fourth, the personalized manufacturing system itself.** Sampling, shipment, sequencing, algorithmic design, mRNA manufacturing, patient-level batch release, cold-chain logistics and return to the clinic form one time-sensitive chain. A delay at any point can cause a patient to miss the intended treatment window. The future moat will therefore be measured in days from sample receipt to administration, successful batch-release rates, cost per patient and the number of distinct sequences a facility can process in parallel. The molecule is the product. The system that gets a personalized medicine back to the patient on time may become the platform. ## 06 | Personalized economics require a different operating report Unit economics cannot be reduced to patient count multiplied by price. Every patient launches a workflow. Was the sample usable? Did the design succeed on the first attempt? Was the batch released on time? Did the cold chain reach the hospital? A failure can add cost without producing a treatable patient or recognized revenue. The most important commercial questions remain undisclosed. How long does sample-to-dose take? What proportion of batches pass release testing? How many patients can one facility process simultaneously? How far can automation reduce manual work? How much margin will failed batches and urgent logistics consume? Will payers reimburse the entire personalized service? Future financial reporting may therefore resemble a pharmaceutical company, a data company and a precision-manufacturing operation at the same time. The sequence matters. Speed, reliability, traceability and the cost curve may prove to be the harder moat. ## 07 | A 176.97% surge raises the risk as well as the expectations The August 19 close of $174.38 and the 176.97% increase come from Nasdaq's official market data. They demonstrate market enthusiasm; they do not guarantee that long-term value has already been secured. Three common misreadings should be removed. Asking whether cancer is over does not establish efficacy across tumor types. The 1,137-patient trial is not the same as 1,137 individualized treatment batches. And a successful Phase 3 topline announcement does not permit investors to import a Phase 2 hazard ratio into a still-empty Phase 3 field. At least five gates remain: 1. Are the complete Phase 3 hazard ratios and absolute effects large enough? 2. Will overall survival provide a harder clinical benefit signal? 3. What filing path and label will the FDA accept? 4. Can patient-specific manufacturing control cost and delivery time at commercial scale? 5. Can other cancer types reproduce the melanoma result? The Phase 2 five-year follow-up reported a hazard ratio of 0.51 for recurrence or death and 0.411 for distant metastasis or death. Those figures came from a 157-patient Phase 2 study. They cannot be substituted for the undisclosed Phase 3 results. The share price has already crossed its finish line. Clinical development and commercialization have not. ## 08 | Taiwan participated, but local beneficiaries should not be invented ClinicalTrials.gov lists four Taiwanese study locations: Kaohsiung Chang Gung Memorial Hospital, National Cheng Kung University Hospital, National Taiwan University Hospital and Linkou Chang Gung Memorial Hospital. Taiwan's clinical system therefore participated in a large multinational Phase 3 study of personalized cancer therapy. That experience can build capabilities in sample handling, patient care and cross-border logistics. A trial site is not a Taiwanese approval, and it is not evidence that a local sequencing, materials or manufacturing company has received an order. A Taiwanese company belongs in an investment map only after a public collaboration, procurement agreement, regulatory filing, clinical responsibility or revenue signal appears. ## Conclusion | Cancer has not said goodbye; the market priced the ability to manage difference The August 19 surge brought a rarely confronted pharmaceutical question into view: when the medicine must change with the patient, can a company still deliver an industrial-grade schedule and quality promise? INTerpath-001 answered one clinical portion of that question. A highly personalized workflow operated inside a large randomized study and crossed two prespecified endpoints. It has not yet answered the questions of approval, overall survival, commercial capacity, cost or reproducibility across tumor types. Cancer has not said goodbye. What has been revalued is the ability to control variation: the design may change from person to person, but delivery cannot become unpredictable; every batch may be different, but the release standard cannot loosen. The move from fixed formulations to controlled variation creates a new industrial threshold. Intismeran currently holds a positive Phase 3 topline ticket to approach that threshold. It is not yet a guarantee of approval, revenue or success across cancer types. ## Disclaimer This article provides biotech-industry analysis based on public information. It is not investment advice, medical advice or treatment guidance. Market data are stated as of the US close on August 19, 2026; a one-day share-price move does not indicate future returns. Intismeran remains investigational and is not FDA approved. Phase 3 disclosure is currently limited to topline results; the full effect size, overall-survival evidence, regulatory timing and commercialization conditions remain pending. --- ## 再見癌症？癌症疫苗來了！莫德納單日暴漲176.97% URL: https://drugnews.com.tw/articles/2026-08-21-moderna-merck-intismeran-phase3-melanoma.html Date: 2026-08-21 Language: zh-Hant Summary: INTerpath-001 第三期頂線結果讓莫德納股價單日上漲 176.97%。真正被重新估價的，不只是一款黑色素瘤療法，而是個人化設計、製造、質控與臨床交付能否成為可擴張的平台。 先說答案：這不是「癌症治癒」，不是所有癌種通用，intismeran 也尚未獲核准。現在看到的，是已切除高風險皮膚黑色素瘤的第三期頂線陽性，而非明天能在醫院施打的成品。 8 月 19 日，莫德納（Moderna，股票代號 MRNA）收在 174.38 美元，一天上漲 111.42 美元，漲幅高達 176.97%。 一家公司，股價在一個交易日被重新定到前一天的近 2.77 倍。 這筆帳可以直接驗算：174.38 美元減去當日上漲的 111.42 美元，前一日收盤是 62.96 美元；174.38 除以 62.96，約為 2.77。所謂上漲 176.97%，是原價之外又增加 1.7697 倍，不是漲到原價的 176.97%。 這組數字證明市場反應有多劇烈，不能直接代替精確市值增量。市值還需要同一時點的流通股數與公司行動口徑；本文不採用未另行核實的「一夜增加 445 億美元」，也不以未驗證交易量替行情加戲。 引爆行情的，是莫德納與默沙東公布的一則三期消息：個人化新抗原信使核糖核酸（mRNA）療法 intismeran autogene，聯用 Keytruda（帕博利珠單抗），在 INTerpath-001 試驗達到兩項終點。 它目前驗證的是：在已完成手術切除、仍有高復發風險的皮膚黑色素瘤患者中，個人化療法加上 Keytruda，比 Keytruda 單藥更能延後復發與遠端轉移。標題可以追問「再見癌症」，證據只允許我們回答：還太早。 市場為什麼仍願意一天重估近三倍？ 因為投資人買進的範圍，早已超出一個黑色素瘤試驗。這次結果同時推高了四層想像：產品成功率、平台外推、莫德納的後疫情第二曲線，以及一套「為每位患者重新設計、製造一批個人化藥物」的系統。 ## 01｜176.97% 背後，三期到底證明了什麼？ INTerpath-001 納入 1,137 名已完成手術切除的 IIB、IIC、III 或 IV 期皮膚黑色素瘤患者，按 2:1 隨機分成兩組： - intismeran＋Keytruda - 安慰劑＋Keytruda 對照組並非沒有治療。這場試驗要回答的，是在既有的 Keytruda 輔助治療上，再加一套患者專屬療法，能不能繼續降低復發風險。 公司公布的預設期中頂線結果顯示： - 無復發存活期（RFS）主要終點達標。 - 無遠端轉移存活期（DMFS）關鍵次要終點達標。 - 安全性與既往資料一致，未觀察到新的安全訊號。 這三點足以改變資產的成功機率，卻還不夠填完估值模型。 第三期的風險比、信賴區間、事件數、絕對差異與存活曲線都還沒有公布；總生存期（OS）也會繼續追蹤。兩家公司接下來才要與監管機構討論申請路徑。 換句話說，我們知道試驗「過關了」，還不知道它贏了多少。Intismeran 也尚未獲美國食品藥物管理局（FDA）核准。 市場先把未來買進去了，完整數據將決定這份樂觀能留下多少。 ## 02｜這不是貨架上的疫苗，而是一張患者專屬通緝令 一般疫苗先找到大家共有的抗原，再生產同一種成品。Intismeran 走的是另一條路。 患者手術後，研究團隊取得腫瘤與血液樣本，讀取腫瘤的去氧核糖核酸（DNA）與核糖核酸（RNA），找出只存在癌細胞的突變。演算法再依患者的人類白血球抗原（HLA）型別，挑選較可能被 T 細胞看見的新抗原。 最多 34 個新抗原的指令，會被寫進同一條 mRNA，包進脂質奈米顆粒，再送回這位患者體內。 病人換了，腫瘤突變不同，藥物序列也跟著改變。 這就是「個人化」最硬的一層意思：醫院面對的不是庫存，而是一筆筆重新設計、重新製造、重新質控的訂單。 疫苗負責教免疫系統認人；Keytruda 則阻斷程式性死亡蛋白 1（PD-1）這道免疫煞車，讓被喚醒的 T 細胞有機會繼續追殺殘存癌細胞。 要特別提醒：1,137 是全試驗人數，不代表製造了 1,137 支 intismeran。試驗採 2:1 分組，只有治療組接受患者專屬療法。 ## 03｜癌症疫苗跌倒多年，這次為何不一樣？ 癌症疫苗不是今天才出現。過去許多療法鎖定多數患者共有的腫瘤相關抗原，卻常撞上三堵牆。 第一堵，是免疫耐受。某些腫瘤相關抗原也存在於正常組織，免疫系統未必願意猛烈攻擊。第二堵，是腫瘤異質性；若只追一個標記，沒有該標記的癌細胞可能逃走。第三堵，是腫瘤微環境會重新關閉已被喚醒的 T 細胞。 Intismeran 嘗試把三塊拼圖同時放回來：選擇患者腫瘤特有的新抗原，一次最多編碼 34 個候選目標，再由 Keytruda 解除 PD-1 免疫煞車。這次又放在手術後的輔助治療，疫苗面對的是可能殘留的微小病灶，而非一個已長成、內部充滿抑制訊號的大型腫瘤。 這些差異解釋了成功的合理性，沒有替其他癌種簽發通行證。黑色素瘤突變負荷較高、對免疫治療較敏感，本來就是相對友善的試驗場；更冷、更複雜的癌種仍要自己證明。 ## 04｜為什麼偏偏是 mRNA？差異也要被工業管理 個人化療法不是把一條傳統產線縮小成「一人一座工廠」，而是把製造難題換了方向。 多數藥物的品管，最怕不同批次做出不同結果。Intismeran 的內容本來就會隨患者改變，真正的考題因此變成：不同設計能否被一組共同規則接住，仍準時跨過純度、效價、安全性與可追溯性的門檻。 mRNA 讓這種「可控的不同」有了工程抓手。每位患者要編碼的新抗原組合可以替換，製造端則把體外轉錄、純化、脂質奈米顆粒包封與放行拆成可驗證的模組。難度沒有消失，只是從反覆另建製程，轉成如何精準編排不同訂單。 演算法負責從大量突變中挑選候選新抗原，製造系統負責把設計落成可用產品，臨床試驗再回答它是否真的有效。三者少一環，都不能靠另一環補分。 所以平台價值不在「能寫出多少序列」，而在「能管理多少差異」。當每張設計單都不同，卻仍能給出穩定的交期與品質承諾，個人化才可能從少數研究中心走向真正的醫療服務。 ## 05｜市場一天買進的，其實是四筆價值 **第一筆，是黑色素瘤資產的成功率。** 大型隨機三期同時達到主要終點與關鍵次要終點，讓 intismeran 從「有希望」向監管申請跨了一大步。完整療效幅度，仍會影響適用族群、標籤強度、定價與滲透速度。 **第二筆，是平台能不能複製。** 截至 2026 年 6 月，兩家公司已布局九項第二／三期研究，涵蓋黑色素瘤、非小細胞肺癌、膀胱癌與腎細胞癌。第一個三期成功，會讓後續管線的可信度上升。 黑色素瘤同時也是相對友善的第一站：突變負荷較高、對免疫治療較敏感，這次又用在手術後腫瘤量較低的輔助治療。肺癌、腎癌與膀胱癌仍要一張張交出自己的成績單。 **第三筆，是莫德納終於摸到後疫情第二曲線。** 新冠疫情後，市場一直追問：莫德納究竟是一家呼吸道疫苗公司，還是一家能反覆產出「可編程藥物」的平台公司？ 這次三期達標，第一次讓後者不再只停留在簡報。對默沙東（Merck，股票代號 MRK）而言，則是替 Keytruda 的組合治療與後續產品週期增加一具新引擎。 依兩家公司揭露，雙方原則上均分相關全球開發成本與損益。這不等於每一美元營收都直接對半；製造成本、費用分攤與商業執行仍會改變最後留下的價值。 **第四筆，是個人化製造系統本身。** 從取樣、運送、定序、演算法設計，到 mRNA 製造、個別批次質控、放行、冷鏈與回院給藥，任何一環延誤，都可能錯過患者的治療時程。 真正的護城河，未來要看的是：從取樣到給藥需要幾天、每批成功放行的比例、單位患者成本，以及一座工廠能同時處理多少條不同序列。 分子是產品；讓個人化療法準時抵達患者的整套系統，才可能變成平台。 ## 06｜個人化療法能不能賺錢，要看一張新報表 個人化療法的單位經濟，不能只用患者人數乘上藥價。每位患者都會啟動一條工作流：樣本是否合格、設計能否一次完成、批次能否準時放行、冷鏈能否送達，任何失敗都可能增加成本，卻無法形成有效收入。 商業化後真正該盯的，是幾個目前尚未公開的答案：從收樣到給藥要多久？批次放行成功率多高？一座工廠能同時承接多少患者？自動化能把人工作業壓到什麼程度？失敗批次與急件物流會吃掉多少毛利？支付方是否願意為這套個人化服務買單？ 因此，未來的財務報表可能同時帶著藥廠、資料公司與精密製造業的影子。序列本身很重要；速度、穩定度、可追溯性與成本曲線，才是那道更難看見的護城河。 ## 07｜暴漲 176.97%，也把風險一起拉高了 8 月 19 日的 174.38 美元與 176.97% 漲幅，來自那斯達克（Nasdaq）官方收盤資料。它證明市場情緒有多強，不能保證長期價值已經落袋。 先拆掉三個最容易出現的誤讀：標題問「再見癌症」，不代表療法已跨癌種成立；1,137 是全試驗人數，不等於 1,137 批個人化藥；第三期達標，也不能拿第二期風險比補進尚未公布的欄位。 接下來至少有五道關卡： 1. 完整第三期風險比與絕對差異夠不夠大？ 2. 總生存期能否提供更硬的臨床證據？ 3. FDA 最後接受什麼申請與標籤？ 4. 個人化製造能否在商業規模下控制成本與交期？ 5. 其他癌種能不能複製黑色素瘤的成功？ 還有一個常見陷阱：2026 年公布的第二期五年追蹤顯示，復發或死亡風險比為 0.51，遠端轉移或死亡風險比為 0.411。那是 157 人的第二期資料，不能拿來補上這次第三期尚未公開的數字。 股價已經先衝過終點線，臨床與商業化還在跑。 ## 08｜台灣不是旁觀者，但概念股別先搶跑 美國臨床試驗登錄網站 ClinicalTrials.gov 列出四個台灣試驗中心：高雄長庚、成大醫院、台大醫院與林口長庚。 這代表台灣醫療體系確實參與了個人化癌症療法的大型跨國三期，也有機會累積樣本處理、臨床照護與跨境物流經驗。 試驗中心不等於台灣已核准，更不等於本地定序、原料或製造公司已經拿到訂單。若要把台灣企業放進投資地圖，至少要等到公開合作、採購、申請、臨床分工或收入證據出現。 ## 結語｜癌症還沒說再見，市場先替「管理差異」開價 8 月 19 日的暴漲，把一個製藥業過去很少正面回答的問題推上檯面：當藥物必須隨患者改變，藥廠還能不能給出工業級的時間表與品質承諾？ INTerpath-001 先回答了臨床的一部分。它告訴市場，這套高度個人化的流程可以在大型隨機試驗中運作，並讓兩項預設終點達標。它還沒有回答的，則是核准、總生存期、商業產能、成本，以及其他癌種能否重現成果。 癌症仍未說再見。這次真正被重新估價的，是藥廠駕馭差異的能力：設計可以因人而變，交期不能失控；成品各自不同，放行標準不能鬆動。 從固定配方走向可控差異，新的產業門檻已經出現。但 intismeran 手上目前只有第三期頂線陽性的入場券，還不是核准、營收或跨癌種成功的保證。 ## 免責聲明 本文為生技醫藥產業與公開資料整理，不構成投資、醫療或用藥建議。行情資料截至美東時間 2026 年 8 月 19 日收盤；單日漲幅不代表未來報酬。Intismeran 仍屬試驗性療法，尚未獲 FDA 核准；第三期目前僅公布頂線結果，完整療效幅度、總生存期、監管時程與商業化條件仍待後續資料。 --- ## NVIDIA Enters Pharma: BMS Readies 576 GPUs as Lilly Commits Up to $1 Billion URL: https://drugnews.com.tw/articles/2026-08-20-big-pharma-ai-arms-race-nvidia-bms-lilly-en.html Date: 2026-08-20 Language: en Summary: BMS is preparing to deploy what NVIDIA's published specifications imply is a 576-GPU Rubin system, while Lilly and NVIDIA plan to commit up to $1 billion over five years. The real contest is whether proprietary data, wet-lab execution and clinical evidence can become one learning loop. Bristol Myers Squibb is preparing to deploy eight NVIDIA DGX Vera Rubin NVL72 systems. Based on NVIDIA's published specifications, the installation implies a total of 576 Rubin GPUs. Eli Lilly has gone even further. The company and NVIDIA are establishing a joint AI laboratory with a planned five-year commitment of up to $1 billion. What Big Pharma fears now is falling behind the learning curve. Compute, models, laboratory robotics, clinical documentation and decades of accumulated failure data are being connected into a single research-and-development chain. This contest has moved far beyond buying a few chatbots to help employees write reports. As BMS, Lilly, Roche, Merck, Novo Nordisk and Sanofi all increase their commitments, the central question in AI-enabled drug development has changed. Running a model faster is only the beginning. The eventual winners will be the companies that make the next experiment faster and more informative, then carry a candidate through the unforgiving test of human clinical trials. ## 01 | BMS is building three layers at once BMS's 2026 moves provide a particularly useful example. In May, the company announced that it would roll out Anthropic's Claude Enterprise across its global operations for more than 30,000 employees. The scope spans research, clinical development, manufacturing, quality, commercial operations and enterprise knowledge. BMS wants the platform to help organize development documents, connect internal knowledge and allow AI agents to participate in selected workflows. In July, BMS announced the construction of an NVIDIA DGX SuperPOD. NVIDIA subsequently disclosed that the system would comprise eight DGX Vera Rubin NVL72 systems. Each carries 72 Rubin GPUs, which yields the 576-GPU figure. BMS described the installation as the most powerful single-company-owned NVIDIA infrastructure in the life-sciences industry. That claim has a narrow comparison set and comes from the two companies involved, so it should not be treated as an independent ranking. The scale of the investment nevertheless sends a clear signal: pharmaceutical companies increasingly want core compute capacity under their own control. The reason is data. Published success stories are available to everyone. Large pharmaceutical companies possess something far scarcer: unpublished failed experiments, compound-toxicity records, patient-stratification data, manufacturing deviations and clinical-operating histories accumulated over decades. These datasets involve intellectual property and patient privacy. If they can be used to train models inside a controlled environment, the competitive moat can extend beyond algorithms to the data itself. BMS has said that AI now participates in the design of all its small-molecule programs and most of its large-molecule programs. That demonstrates depth of use, not drug success. A long clinical road still separates designing a molecule from proving that it is safe and effective in humans. One week later, Schrödinger introduced an early-access version of Bunsen, an AI co-scientist that can plan, execute and interpret computational molecular-discovery workflows. The important boundary is that Schrödinger's announcement did not name BMS as a Bunsen customer. Joining the two announcements into a supposed "BMS-Bunsen" relationship would go beyond the available evidence. ## 02 | Big Pharma is buying an entire R&D factory BMS is not acting alone. The 2026 AI infrastructure ledger is already crowded. ### January | Lilly and NVIDIA The companies announced a joint five-year commitment of up to $1 billion to establish an AI co-innovation laboratory in the San Francisco Bay Area. The plan is to connect a computational "dry lab" with a robotic "wet lab," allowing AI to propose designs, machines to run experiments and the results to flow back into the model in a near-continuous cycle. ### March | Roche Roche announced the addition of 2,176 NVIDIA Blackwell GPUs. Together with cloud resources, its hybrid architecture exceeds 3,500 GPUs. The intended uses span drug discovery, pathology, diagnostics, manufacturing digital twins and digital health. ### April | Merck & Co. and Google Cloud Merck announced a multiyear collaboration of up to $1 billion to deploy Gemini Enterprise across research, manufacturing, commercial operations and corporate functions. This is the U.S. company Merck & Co., known as MSD outside the United States and Canada, not Germany's Merck KGaA. ### April | Novo Nordisk and OpenAI Their strategic collaboration reaches from candidate discovery to manufacturing, supply chain and commercial operations, with global integration planned to progress through the end of 2026. ### An earlier foundation | Sanofi and Owkin Sanofi moved earlier. In 2021, it made a $180 million equity investment in Owkin and established a separate three-year collaboration worth $90 million. The program uses federated learning to analyze multimodal data distributed across hospitals in search of cancer biomarkers and treatment-response patterns. Viewed together, these investments show that Big Pharma is not buying one uniform form of AI. Some companies are building compute first. Others are deploying enterprise-wide agents, while still others are connecting models directly to laboratory operations. Hardware expenditure is only the admission price. Whether proprietary data, scientific judgment, experimental execution and regulatory records can be connected into a closed learning loop will determine whether the spending becomes an R&D asset or merely an expensive server room. ## 03 | AI can move faster, but the drug still has to work in people The most easily misread statistic in AI drug discovery is the claim that Phase 1 success rates reach 80% to 90%. That range comes from a 2024 early review of programs developed by AI-native companies. The sample remained small, companies used different definitions of what counted as "AI-discovered," and public disclosure may have favored successful programs. More importantly, Phase 1 is primarily designed to evaluate safety and dosing. It cannot by itself establish whether a drug meaningfully treats a disease. For Phase 2, the review observed a success rate of roughly 40%, much closer to historical norms. In other words, AI may be improving the selection of molecules that look more drug-like, but it has not yet proved that it understands complex human disease better. Two of the most closely watched programs illustrate the evidence boundary. Insilico Medicine's rentosertib is a pulmonary-fibrosis candidate whose target identification and molecular design used generative AI. In 2025, *Nature Medicine* published its randomized Phase 2a study. Seventy-one participants were treated for 12 weeks, and the highest-dose group produced an encouraging signal on a secondary forced-vital-capacity endpoint. Yet the study was small and short, 16 participants discontinued early, and the paper itself called for larger and longer trials. MindRank's MDR-001 is an oral small-molecule GLP-1 candidate. In February 2026, the company announced that the first participant had been dosed in a Chinese Phase 3 study expected to enroll about 750 people. That places the program among the most clinically advanced AI-assisted drug candidates, but starting Phase 3 is not the same as succeeding in Phase 3. The company's Phase 2 summary also still requires support from complete data. The right conclusion is therefore narrower. 2026 may be a year in which AI-designed or AI-assisted drugs face increasingly demanding clinical stress tests. It is not yet the year in which AI has been clinically proven as a superior drug-development system. ## 04 | Taiwan has a role, but a supply-chain list is not a purchase order Taiwan's most direct connection to this wave of pharmaceutical investment lies in AI infrastructure. NVIDIA's published Vera Rubin ecosystem includes Taiwanese companies across chips, packaging, systems and volume manufacturing. TSMC, SPIL under ASE Technology Holding, Foxconn, Quanta's QCT, Wistron, Inventec and Pegatron are among the companies identified in those layers. NVIDIA has also said that the relevant platforms have 150 ecosystem partners in Taiwan. There is still no public evidence showing which Taiwanese supplier, if any, will fill a specific part of the BMS or Roche deployments. Appearing in NVIDIA's ecosystem indicates the ability to participate in the broader market. It does not establish that a company has won a particular pharmaceutical order, and it cannot be translated directly into revenue. For Taiwan's biotech industry, the second lesson is even sharper. Pharmaceutical companies are willing to spend heavily because they possess decades of high-quality, traceable data, including failures. When formats are inconsistent, experiments cannot be reproduced and clinical fields do not connect, even the strongest model can do little more than organize noise. The AI era will amplify the value of data assets, but it will amplify data debt as well. ## Conclusion | The final judge is still the patient's clinical outcome GPU counts are easy to compare, and $1 billion makes an irresistible headline. Drug development, however, has never been won by a leaderboard. Three questions now deserve sustained attention. Can a pharmaceutical company turn one experiment into a better next decision? Can AI-designed candidates preserve an advantage through Phase 2 and Phase 3? Can the capital being committed ultimately produce more useful medicines? BMS and Lilly have placed large bets on the table. The final judge remains the same: clinical outcomes in patients. ## Disclaimer This article provides biotech-industry information and analysis based on public sources. It is not medical advice, diagnostic guidance, treatment advice or investment advice. Many company investment figures discussed here are multiyear ceilings or plans rather than amounts already spent. Infrastructure scale, early clinical signals and supply-chain ecosystem participation do not establish drug success, approval probability or a specific supplier order. --- ## 輝達殺進製藥業！BMS 搶 576 顆 GPU，禮來砸 10 億美元 URL: https://drugnews.com.tw/articles/2026-08-20-big-pharma-ai-arms-race-nvidia-bms-lilly.html Date: 2026-08-20 Language: zh-Hant Summary: BMS 準備部署依 NVIDIA 規格推算共 576 顆 Rubin GPU，禮來與 NVIDIA 五年共同投入上限 10 億美元；真正的競爭不只在算力，而在能否把專有資料、濕實驗與臨床證據接成研發閉環。 BMS 準備部署 8 套 NVIDIA DGX Vera Rubin NVL72。按 NVIDIA 公布的規格推算，裡面合計有 576 顆 Rubin GPU。 禮來更直接：與 NVIDIA 共建 AI 實驗室，五年投入上限 10 億美元。 大藥廠現在怕的，是學習曲線落後。 算力、模型、實驗機器人、臨床文件與數十年累積的失敗資料，正被接成同一條研發鏈。這場競爭早已超過「買幾個聊天機器人、替員工寫報告」的層次。 當 BMS、禮來、Roche、默沙東、諾和諾德、Sanofi 同時加碼，AI 製藥的勝負也換了題目：模型能不能跑得快，只是起點；誰能讓下一次實驗更快、更準，最後把候選藥送過人體試驗，才有資格談贏家。 ## 01｜BMS 一口氣補齊三層 BMS 今年的動作很有代表性。 5 月，公司宣布把 Anthropic 的 Claude Enterprise 導入全球營運，服務超過 3 萬名員工，範圍橫跨研究、臨床開發、製造、品質、商務與企業知識。BMS 希望它能協助整理開發文件、串接內部知識，甚至讓 AI 代理人參與部分工作流程。 7 月，BMS 再宣布建置 NVIDIA DGX SuperPOD。NVIDIA 隨後披露，系統由 8 套 DGX Vera Rubin NVL72 組成；每套搭載 72 顆 Rubin GPU，換算就是 576 顆。 BMS 把它稱為生命科學界最強大的單一公司自有 NVIDIA 基礎設施。這句話有很窄的比較範圍，也出自合作雙方，不能當成獨立排行榜。但投資規模釋放的訊號很清楚：藥廠開始把核心算力留在自己手上。 原因就在資料。 論文裡的成功案例人人都看得到。大藥廠手上那些沒有公開的失敗實驗、化合物毒性、病人分層、製程偏差與臨床操作紀錄，反而更稀缺。這些資料牽涉智慧財產與病人隱私，若能留在受控環境中訓練模型，護城河就會從演算法延伸到資料本身。 BMS 表示，AI 已參與公司所有小分子計畫，以及多數大分子計畫的設計。這證明的是「使用深度」，還不是「藥物成功」。從設計一個分子，到在人體裡證明安全、有效，中間仍隔著漫長的臨床路。 一週後，Schrödinger 也推出分子研發代理人 Bunsen 的早期使用版本，可自行規劃、執行並解讀計算流程。值得注意的是，官方公告沒有把 BMS 列為 Bunsen 客戶。把兩件事硬接成「BMS×Bunsen」，目前證據不足。 ## 02｜大藥廠買的是一整座研發工廠 BMS 並不孤單。2026 年的軍備表，已經排得很長： ### 1 月｜禮來×NVIDIA 雙方宣布五年共同投入最高 10 億美元，在舊金山灣區建立 AI 共同創新實驗室。規劃把運算端的「乾實驗室」與機器人「濕實驗室」接起來，讓 AI 提出設計、機器執行實驗、結果再回饋模型，朝 24 小時循環運作前進。 ### 3 月｜Roche 新增 2,176 顆 NVIDIA Blackwell GPU；加上雲端資源，整體混合架構超過 3,500 顆。用途涵蓋藥物研發、病理、診斷、製造數位分身與數位醫療。 ### 4 月｜美國默沙東×Google Cloud 多年合作投入上限 10 億美元，導入 Gemini Enterprise，覆蓋研發、製造、商務與公司營運。這裡指的是 Merck & Co.，不是德國 Merck KGaA。 ### 4 月｜諾和諾德×OpenAI 從候選藥發現一路延伸到製造、供應鏈與商務，並預計在 2026 年底前逐步完成全球整合。 ### 較早布局｜Sanofi×Owkin Sanofi 早在 2021 年就投資 Owkin 1.8 億美元，另設三年 9,000 萬美元合作，利用聯邦學習分析分散在醫院的多模態資料，尋找癌症標記與治療反應。 把這些交易放在一起，會發現大藥廠要的不是同一種 AI。 有人先補算力，有人先鋪全公司的代理人，也有人把模型直接接進實驗室。硬體支出只是門票。能不能把專有資料、科學家判斷、實驗執行與法規紀錄連成閉環，才會決定這筆錢是研發資產，還是昂貴的機房。 ## 03｜AI 跑得快，藥還是得過人體這一關 AI 製藥最容易被誤讀的數字，是「一期成功率 80%到90%」。 這個區間來自 2024 年一篇針對 AI 原生公司的早期回顧。樣本數仍小，各家公司對「AI 發現」的定義不同，公開資料也可能偏向成功項目。更重要的是，一期主要看安全性與劑量，不能直接回答藥物有沒有治療效果。 到了二期，該回顧觀察到的成功率約 40%，已接近歷史水準。換句話說，AI 可能更會挑出「像藥的分子」，卻尚未證明它更懂複雜的人體疾病。 目前最受矚目的兩個案例，也把證據邊界畫得很清楚。 Insilico Medicine 的 rentosertib，是用生成式 AI 找到標靶並設計的肺纖維化候選藥。2025 年《Nature Medicine》刊登其 IIa 期隨機對照研究：共 71 人、治療 12 週，最高劑量組的肺活量次要終點出現鼓舞訊號。然而樣本小、追蹤短，16 人提前退出；論文本身也要求更大、更長的研究。 中國 MindRank 的口服小分子 GLP-1 藥物 MDR-001，已於 2026 年 2 月完成中國三期首位受試者給藥，預計約 750 人參與。它確實站上目前 AI 輔助藥物最前面的臨床位置，但三期啟動不等於三期成功，公司公布的二期摘要也仍需要完整資料支持。 所以，2026 年可以是 AI 藥物接受臨床壓力測試的一年，還不能叫作「AI 已被臨床證明的一年」。 ## 04｜台灣有位置，但別把供應鏈名單當訂單 這波大藥廠投資，台灣最直接的交集在 AI 基礎設施。 NVIDIA 公布的 Vera Rubin 生態系裡，台積電、日月光集團旗下矽品、鴻海、廣達旗下 QCT、緯創、英業達、和碩等台灣業者，都出現在晶片、封裝、系統與整機量產環節。NVIDIA 更表示，相關平台在台灣有 150 家生態系夥伴。 但目前沒有公開資料能證明，BMS 或 Roche 的這批設備由哪一家台廠承接。出現在 NVIDIA 供應鏈，代表有能力參與整體市場；它不等於拿到特定藥廠訂單，更不能直接換算營收。 對台灣生技業來說，另一個提醒更尖銳。 藥廠願意花大錢的原因，是手裡有數十年高品質、可追溯、包含失敗結果的資料。若資料格式混亂、實驗無法重現、臨床欄位彼此不通，再強的模型也只能整理雜訊。AI 時代會放大資料資產，也會放大資料債。 ## 寫在最後 GPU 數量很好比，10 億美元也很好寫進標題。可惜新藥研發從來不靠排行榜決勝。 接下來值得盯的，有三件事：大藥廠能否把一次實驗變成下一次決策；AI 設計的候選藥能否在二、三期維持優勢；投入的資本最後能不能換成更多好藥。 BMS 和禮來已經把籌碼推上桌。最後的裁判，依舊是病人的臨床結果。 ## 免責聲明 本文為生技醫藥產業與公開資料整理，不構成醫療診斷、治療建議或投資建議。文中公司投資金額多為多年投入上限或規劃，不代表已支付金額；基礎設施規模、早期臨床訊號與供應鏈生態系關係，均不等同於新藥成功、核准機率或特定訂單。 --- ## Just $74 Million Short: Can Skyrizi Succeed Humira as the Autoimmune Drug King? URL: https://drugnews.com.tw/articles/2026-08-19-skyrizi-humira-autoimmune-drug-king-en.html Date: 2026-08-19 Language: en Summary: Skyrizi generated $5.505 billion in Q2 2026, only $74 million below Humira's historic quarterly peak. The real succession test spans efficacy, treatment burden, multi-indication reach and access. Humira spent years at the summit of global prescription-drug sales. Its patent cliff has only just arrived, yet AbbVie's Skyrizi is already closing in on the quarterly record Humira itself left behind. Skyrizi generated $5.505 billion in the second quarter of 2026, up 24.4% year over year. Humira's quarterly peak was $5.579 billion in the fourth quarter of 2022. The gap is only $74 million. The jump is not a one-quarter anomaly. Skyrizi generated $4.483 billion in the first quarter and grew another 22.8% sequentially in the second. If that momentum holds, the Humira record that has stood for four years could soon be broken by another AbbVie product. One boundary needs to be clear from the start. The $5.505 billion includes Skyrizi revenue across psoriasis, psoriatic arthritis, Crohn's disease, ulcerative colitis and its other approved uses. AbbVie does not disclose psoriasis revenue separately, so the entire figure cannot be attributed to plaque psoriasis. That is precisely why the number matters. Skyrizi has grown from a psoriasis launch into AbbVie's next immunology platform. In the race to become the next autoimmune drug king, an efficacy table is only the first test. ## 01 | Humira is declining, and AbbVie has built another $5.5 billion product Humira generated $21.237 billion in full-year 2022 revenue, including $5.579 billion in the fourth quarter. Sales then declined rapidly after U.S. exclusivity ended and biosimilar competition entered the market. Every blockbuster eventually declines. The danger is when the successor has not matured in time. AbbVie's answer is now visible in the numbers. In the second quarter of 2026, Skyrizi generated $5.505 billion and Rinvoq generated $2.525 billion. Together they produced $8.030 billion, or 47.3% of AbbVie's $16.990 billion in quarterly revenue. The Humira cliff is no longer an existential threat to the company. But a new issue is forming at the same time. Skyrizi alone accounted for about 32.4% of AbbVie's second-quarter revenue. The company has reduced its dependence on Humira, only to begin concentrating revenue in a new flagship product. That is the dual nature of an autoimmune drug king. It can rescue a company's growth trajectory while also redefining the company's risk. ## 02 | Why can psoriasis support a megablockbuster? Psoriasis extends far beyond plaques and scaling on the skin. It is a chronic immune-mediated disease that can recur over a lifetime and may coexist with joint, metabolic and cardiovascular disease. A study based on the U.S. National Health and Nutrition Examination Survey estimated that psoriasis affects about 3.0% of adults aged 20 or older in the United States, equivalent to roughly 7.55 million people. A large patient population, a long disease course and the need for sustained control create an enormous long-term treatment market. That does not mean every new drug automatically captures the opportunity. Psoriasis already has mature TNF, IL-17 and IL-23 treatment pathways. As choices expand, a product cannot stop at being effective. It has to help more patients achieve deeper skin clearance, sustain that response and reduce the burden of long-term treatment. That is where Skyrizi established its position. ## 03 | Skyrizi's core advantage: deep clearance with a lower treatment burden Skyrizi is an IL-23-targeting biologic. Under the U.S. FDA label, patients with moderate-to-severe plaque psoriasis receive subcutaneous injections at Week 0 and Week 4, followed by one injection every 12 weeks. After the two initiation doses, that usually means only four maintenance injections a year. This is not a regimen that begins monthly and is later stretched to quarterly dosing once the disease is controlled. The 12-week maintenance interval is built into the approved label rather than improvised after improvement. The efficacy bar is also high. In the two adult Phase 3 plaque-psoriasis trials reported in the FDA label, 75% of patients in each study achieved PASI 90 at Week 16, meaning at least a 90% improvement in skin lesions. PASI 100, or complete skin clearance, was achieved by 36% and 51% of patients, respectively. With approvals extending to psoriatic arthritis, Crohn's disease and ulcerative colitis, Skyrizi is no longer positioned as just a psoriasis injection. One brand now spans inflammation in the skin, joints and gastrointestinal tract. That distinction is central to the commercial story. Psoriasis opened the first door with physicians and patients. Multiple indications raised the ceiling. ## 04 | An oral drug has arrived. Does that mean the injectable loses? In March 2026, the FDA approved Johnson & Johnson's ICOTYDE, or icotrokinra, a once-daily oral peptide that targets the IL-23 receptor. It addresses an intuitive patient need. Many people dislike injections, and swallowing a pill can feel easier. But oral administration is not an automatic victory. In the two adult Phase 3 trials in the ICOTYDE label, 55% and 58% of patients achieved PASI 90 at Week 16, while 31% and 32% achieved PASI 100. In the two Skyrizi label trials cited above, PASI 90 was 75% in both studies, while PASI 100 was 36% and 51%. Those figures can provide a sense of scale, but they cannot determine a winner. The products were not compared in the same head-to-head trial. Patient populations, trial designs and statistical methods differed. Clinical decisions require several ledgers to be considered at once. Remembering a pill every day is one kind of treatment burden. One injection every 12 weeks can also be a form of convenience. Depth of response, safety, comorbidities, reimbursement, patient preference and physician experience ultimately shape the prescription. Oral peptides therefore open a new part of the market, while long-acting biologics remain firmly in the race. ## 05 | The next competitive cycle is more complicated than injection versus pill Psoriasis frequently coexists with other immune-mediated conditions. Whether a product can be used across specialties and indications can materially influence physician choice. The FDA labels for some IL-17 inhibitors include warnings about new-onset or worsening inflammatory bowel disease and call for patients to be monitored. That does not mean every patient will experience the event, nor does it justify condemning an entire drug class. For patients with gastrointestinal comorbidities, the decision still depends on the individual product label and the patient's disease profile. The IL-23 pathway has established approved indications in psoriasis and selected inflammatory bowel diseases, which supports Skyrizi's brand expansion. Skyrizi also carries label requirements related to infection risk and tuberculosis evaluation. No immunology drug can be judged only by its advantages while its risks are ignored. AbbVie itself is not betting on a single formulation. The company acquired Nimble Therapeutics in 2025 to obtain an oral IL-23 receptor peptide platform. In AbbVie's June 2026 pipeline presentation, the candidate ABBV-859 was listed as expected to enter Phase 1 in the third quarter of 2026. In other words, the company now leading with a long-acting injectable is also preparing its own oral option. That is the standard Big Pharma playbook: protect the current strength while acquiring the formulation that could change the rules. ## Conclusion | The next autoimmune drug king will be decided by the entire treatment experience Skyrizi is only $74 million away from Humira's quarterly record. Reducing the story to "psoriasis created another blockbuster" misses the larger signal. The more important question is how an immunology drug becomes a long-duration platform across diseases. Skyrizi's success reflects several forces working together: deep skin clearance, maintenance dosing every 12 weeks, expansion across indications and AbbVie's commercialization capability. Chronic-disease markets do not reward a single attractive efficacy number forever. Efficacy, treatment burden, comorbidities, safety and access have to become a treatment experience that patients can remain on for the long term. Oral peptides will bring new competition, while IL-17, IL-23 and other mechanisms continue to evolve. Whether Skyrizi ultimately succeeds Humira as the autoimmune drug king will depend on sustained growth, pricing and reimbursement, and the continued expansion of revenue across indications. But in the second quarter of 2026, AbbVie had already accomplished the hardest part of the transition: After Humira's decline, the company built a successor that has come within $74 million of its predecessor's quarterly record. ## Disclaimer This article provides biotech-industry information and analysis based on public sources. It is not medical advice, a diagnosis, treatment guidance or investment advice. Results from different clinical trials should not be treated as direct comparative evidence. Treatment decisions should be made by qualified healthcare professionals based on each patient's condition and the applicable approved label. --- ## 55 億美元只差一步：Skyrizi 要接棒 Humira 當自體免疫藥王？ URL: https://drugnews.com.tw/articles/2026-08-19-skyrizi-humira-autoimmune-drug-king.html Date: 2026-08-19 Language: zh-Hant Summary: Skyrizi 2026 年第二季收入達 55.05 億美元，距 Humira 的歷史單季高點只差 7,400 萬美元；真正的接班考驗，是療效、治療負擔、多適應症與支付能力能否一起延續。 Humira（復邁）曾連續多年站在全球處方藥銷售頂峰。它的專利懸崖剛落下，AbbVie 手中的 Skyrizi（喜開悅）已逼近 Humira 自己留下的單季紀錄。 2026 年第二季，Skyrizi 單季收入衝到 55.05 億美元，年增 24.4%。Humira 在 2022 年第四季寫下 55.79 億美元的單季高點。 兩者只差 7,400 萬美元。 單季跳升背後還有連續性。Skyrizi 第一季收入 44.83 億美元，第二季再成長 22.8%。若接下來維持目前動能，Humira 封存四年的單季紀錄，真的可能很快被自家後輩改寫。 但先把一條重要界線說清楚：這 55.05 億美元，是 Skyrizi 橫跨乾癬、乾癬性關節炎、克隆氏症與潰瘍性結腸炎等適應症的總收入；AbbVie 沒有拆出乾癬單項銷售，不能把整筆收入都算到銀屑病頭上。 這組數字的分量，在於 Skyrizi 已從乾癬新藥長成 AbbVie 的新免疫平台。自體免疫藥王的接力賽，療效表只是第一關。 ## 01｜Humira 退場，AbbVie 又養出一個 55 億美元產品 Humira 在 2022 年全年收入達 212.37 億美元，第四季單季 55.79 億美元，之後隨美國專利獨占期結束、生物相似性藥品進場而快速下滑。 明星藥總會退場，危險往往出在接班人還沒長大。 艾伯維如今交出的答案很直接：2026 年第二季，Skyrizi 55.05 億美元、Rinvoq 25.25 億美元；兩者合計 80.30 億美元，占公司單季總收入 169.90 億美元的 47.3%。 Humira 的懸崖，已不再是攸關公司生死的危機。 不過，新問題也同時浮現。Skyrizi 一款藥就占 AbbVie 第二季收入約 32.4%。公司擺脫了對 Humira 的依賴，新的明星產品集中度也跟著成形。 自體免疫新藥王有兩面：它可以拯救一家公司的成長，也會重新決定整家公司的風險。 ## 02｜乾癬為什麼能養出超級重磅藥？ 乾癬的影響遠超過皮膚表面的紅斑與脫屑。它是一種慢性、免疫介導的疾病，常反覆發作，也可能合併關節、代謝與心血管問題。 一項以美國全國健康與營養調查為基礎的研究估計，美國 20 歲以上成人乾癬盛行率約 3.0%，相當於 755 萬人。 患者基數大、病程長、需要持續控制，自然形成巨大的長期治療市場。但這並不代表任何新藥都能吃到紅利。 乾癬治療已經有 TNF、IL-17、IL-23 等多條成熟路線。當選擇變多，產品不能只做到「有效」，還得讓更多患者清得更乾淨、維持得更久，同時降低長期治療的負擔。 Skyrizi 的起點正是這裡。 ## 03｜它的王牌：清得乾淨，也把負擔壓低 Skyrizi 是針對 IL-23 的生物製劑。依美國 FDA 標籤，中重度斑塊型乾癬患者在第 0 週、第 4 週各接受一次皮下注射，之後每 12 週一次。 換句話說，完成起始劑量後，一年通常只需要四次維持注射。 這與「先每月一次，病情控制後再改每季一次」不同。每 12 週一次早已寫進標籤的維持療程，與病況改善後臨時拉長間隔無關。 療效也建立了很高的門檻。FDA 標籤中的兩項成人第三期試驗顯示，第 16 週達到 PASI 90、也就是皮損改善至少 90% 的比例，兩項試驗都是 75%；達到完全清除的 PASI 100，分別是 36% 與 51%。 加上乾癬性關節炎、克隆氏症與潰瘍性結腸炎，Skyrizi 已跨出「一支乾癬針」的定位，同一個品牌串起皮膚、關節與腸道發炎。 這點非常關鍵。乾癬替它打開醫師與患者的第一扇門，多適應症才把天花板往上推。 ## 04｜口服新藥來了，針劑就一定輸嗎？ 2026 年 3 月，美國 FDA 核准 Johnson & Johnson 的 ICOTYDE（icotrokinra），這是一款每天一次、鎖定 IL-23 受體的口服胜肽藥物。 它擊中了一個很直觀的需求：不少患者不喜歡打針，吞藥看起來更輕鬆。 但「口服」不等於自動勝出。 ICOTYDE 標籤中，兩項成人第三期試驗第 16 週的 PASI 90 分別為 55% 與 58%，PASI 100 為 31% 與 32%。Skyrizi 標籤中的兩項試驗，同期 PASI 90 都是 75%，PASI 100 為 36% 與 51%。 這些數字只能用來看大致量級，不能直接判定誰贏。兩者並非在同一場頭對頭試驗中比較，受試者組成、試驗設計與統計方法都不同。 臨床決策要同時算上多本帳。每天記得服藥，是一種負擔；每 12 週注射一次，也可能是一種便利。療效深度、安全性、共病、給付、患者偏好與醫師經驗，最後會一起決定處方。 所以，口服胜肽打開的是一塊新市場，長效生物製劑仍會留在賽場。 ## 05｜下一輪競爭，會比「針劑對口服」更複雜 乾癬患者常伴隨其他免疫疾病，產品能否跨科別、跨適應症使用，會影響醫師選擇。 部分 IL-17 抑制劑的 FDA 標籤，列有新發或惡化發炎性腸道疾病的警示，要求治療時留意症狀；這不等於所有患者一定發生，也不能把整個藥物類別一概判死刑。對合併腸道疾病的患者而言，仍需依個別產品標籤與病況評估。 IL-23 路線則已在乾癬與部分發炎性腸道疾病建立核准適應症，讓 Skyrizi 的品牌延伸更有利。但它同樣有感染、結核病評估等標籤要求，沒有任何免疫藥能只談優點、不談風險。 更有意思的是，艾伯維自己也沒有押單邊。 它在 2025 年買下 Nimble Therapeutics，取得口服 IL-23 受體胜肽平台。AbbVie 2026 年 6 月的官方簡報把候選藥 ABBV-859 列為預計 2026 年第三季啟動第一期臨床。 也就是說，今天靠長效針劑領跑的艾伯維，也在準備自己的口服選項。 大藥廠的標準打法，就是守住現有強項，也提前買下可能改變規則的劑型。 ## 結語｜下一個自體免疫藥王，比的是整套治療體驗 Skyrizi 距離 Humira 的單季紀錄，只差 7,400 萬美元。 把這個故事縮成「乾癬又養出一支大藥」，太可惜。更大的訊號是，一款免疫藥如何被做成跨病種的長期平台。 Skyrizi 的成功，是高皮膚清除率、每 12 週一次維持注射、多適應症擴張與 AbbVie 商業化能力疊在一起的結果。慢性病市場不會只為一個漂亮數字買單。療效、治療負擔、共病、安全與支付，必須整合成一套患者願意長期留在裡面的治療方案。 口服胜肽會帶來新競爭，IL-17、IL-23 與其他機制也會繼續迭代。Skyrizi 能不能接下 Humira 的自體免疫藥王位置，還要看成長能否延續、價格與給付是否守得住，以及多適應症收入能否繼續擴大。 但至少在 2026 年第二季，艾伯維已經把一件最難的事做成： Humira 退場後，AbbVie 已經養出一個逼近其單季紀錄的接班人。 ## 免責聲明 本文為生技醫藥產業與公開資料整理，不構成醫療診斷、治療建議或投資建議。不同臨床試驗不可直接橫向判定療效優劣，實際用藥應由專業醫療團隊依患者病況與核准標籤評估。 --- ## Eli Lilly Sold $23 Billion in One Quarter. Why Is It Still Rushing to Buy the Next Drug King? URL: https://drugnews.com.tw/articles/2026-08-18-eli-lilly-tirzepatide-next-drug-king-en.html Date: 2026-08-18 Language: en Summary: Tirzepatide has allowed Eli Lilly to dominate both diabetes and obesity, but revenue concentration, pricing pressure and the next product relay are the investment questions that matter now. Eli Lilly has delivered another set of results that almost do not look like a traditional Big Pharma quarter. Second-quarter 2026 revenue reached $22.974 billion, up 48% year over year. Net income was $7.095 billion, up 25%. The company also raised full-year revenue guidance to $85 billion to $87 billion. Mounjaro, the diabetes brand, generated $9.943 billion. Zepbound, the obesity brand, generated $4.928 billion. Together, the two brands produced $14.871 billion, or 64.7% of Lilly's quarterly revenue. But investors should not be fooled by the two names. Mounjaro and Zepbound are both powered by the same molecule: tirzepatide. In other words, Lilly appears to have two money-printing machines, but the core engine is one molecule. That is why the better the numbers get, the more urgently Lilly has been buying companies. The most comfortable tailwind is also where concentration risk becomes most visible. ## 01 | One molecule, two large markets Tirzepatide is a dual GIP and GLP-1 incretin agonist. In type 2 diabetes it is sold as Mounjaro. In adults with obesity, or overweight adults who meet treatment criteria, it is sold in the United States as Zepbound. The molecule first established its strength clinically. In the 751-person SURMOUNT-5 Phase 3 head-to-head study, adults with obesity or overweight without diabetes treated for 72 weeks had an average weight reduction of 20.2% with tirzepatide versus 13.7% with semaglutide. That result supports tirzepatide's advantage under the conditions of that study. It does not mean every patient will experience the same effect, and it cannot be casually transferred to every indication. The second step was commercial execution. Lilly placed the same molecule into diabetes, weight management and sleep-apnea-related markets, then used LillyDirect, self-pay programs, insurance access and manufacturing expansion to turn clinical differentiation into prescription volume. The second-quarter numbers are the result of that machine. Mounjaro generated $9.943 billion, up 91% year over year. Zepbound generated $4.928 billion, up 46%. The combined $14.871 billion represented 64.7% of quarterly revenue. The first half showed the same structure. The two brands produced $27.693 billion against Lilly's total first-half revenue of $42.773 billion, also 64.7%. This is no longer a one-quarter coincidence. It is the company's revenue architecture. ## 02 | Lilly is leading, but the price war has already started Behind the high growth is another set of numbers that investors should take seriously. Lilly's second-quarter volume increased 60%, while realized price declined 13%. In the United States, excluding rebate and discount estimate adjustments, overall realized price declined about 9%. That 9% is a company-level U.S. business metric, not a single-product Zepbound price cut. But the direction is clear: the GLP-1 market is trading lower realized price for broader access and greater volume. China shows the same kind of exchange. After Mounjaro was included in China's national reimbursement drug list, overseas realized price declined while volume increased sharply, driving overseas Mounjaro revenue up 172% year over year. Novo Nordisk has not left the field either. The company still reported 7% constant-currency adjusted sales growth in the second quarter of 2026. As of July 17, weekly U.S. prescriptions for oral Wegovy had exceeded 265,000. Lilly is winning the current product and execution cycle. It has not won permanent monopoly status. The more successful a product becomes, the more payers will demand concessions. The more competitors arrive, the more options physicians and patients will have. Today's 60% volume growth must keep outrunning price erosion if revenue momentum is to hold. ## 03 | Oral and triple agonist programs are Lilly's own relay Lilly is not waiting for tirzepatide to slow before preparing the next leg. The first follow-on is Foundayo, or orforglipron, a once-daily oral small-molecule GLP-1 drug. FDA approved it on April 1, 2026 for chronic weight management in eligible adults with obesity or overweight, and Lilly began shipping it on April 6. In its first selling quarter, the product contributed $98 million. The boundary matters. Foundayo is currently approved in the United States for weight management. Its type 2 diabetes indication remains under regulatory review. Its value is not that it will replace every injectable product, but that it can bring patients who dislike injections, face injection inconvenience or prefer an oral option into treatment. The second follow-on is retatrutide. This once-weekly triple agonist targets GIP, GLP-1 and glucagon receptors and remains investigational. In the TRIUMPH-1 Phase 3 topline readout, the 12 mg dose produced an average weight reduction of 28.3% at 80 weeks using the efficacy estimand and 25.0% using the treatment-regimen estimand. Lilly plans to submit in the United States in the first quarter of 2027. One molecule protects the cash flow. One pill expands the entry point. One triple agonist pushes the efficacy ceiling. If the relay works, Lilly can turn a single-product cycle into a platform cycle. But any weak leg changes the market's view of the growth premium currently attached to tirzepatide. Foundayo has to prove long-term prescription persistence. Retatrutide still has to pass regulatory review and real-world use. Manufacturing has to scale with a larger patient pool. ## 04 | Sleep, CAR-T and vaccines: Lilly is buying a second mountain Lilly is also using external capital to move beyond metabolic disease. Acquisitions completed in the second quarter included Orna, Ajax, Centessa and Kelonia. After the quarter, Lilly also completed three infectious-disease acquisitions involving Curevo, LimmaTech and Vaccine Company, and it signed a merger agreement with AtaiBeckley. The directions look scattered at first. The common logic is that Lilly wants assets with clinical entry points or platform barriers, then intends to apply its development, manufacturing and commercialization capability at scale. Centessa gives Lilly exposure to sleep-wake disorders. The transaction has closed, with about $6.3 billion in cash consideration at closing and up to $1.5 billion in contingent value rights. Orna and Kelonia point toward in vivo CAR-T and genetic delivery. Orna's transaction value is up to $2.4 billion. Kelonia included $3.25 billion upfront and up to $7.0 billion in total consideration including milestones. Curevo, LimmaTech and Vaccine Company add shingles, drug-resistant bacterial infection and EBV vaccine exposure. The announced maximum consideration across the three deals was about $3.83 billion. AtaiBeckley gives Lilly a potential entry into treatment-resistant depression and related psychiatric disorders, with about $2.8 billion in cash consideration at closing and up to $1.0 billion in CVRs. As of the verification date, that deal should still be treated as pending. The easiest mistake is to add every maximum number together and describe it as money already spent. Milestones and CVRs require clinical, regulatory or commercial conditions to be met. Announced ceilings are not current cash outflows. The income statement already shows the cost of this expansion. Lilly recorded $2.8 billion of acquired in-process research and development expense in the quarter, plus $703 million of impairment, restructuring and other special charges, mainly related to Kelonia and Centessa closing and integration. This is an expensive succession project. What Lilly is buying is optionality, not a guarantee. ## 05 | The real moat is moving science into factories In May, Lilly committed another $4.5 billion to expand two Indiana manufacturing sites. Since 2020, its committed capital expenditure in the state has exceeded $21 billion. The new facilities cover active pharmaceutical ingredients, Foundayo, retatrutide and genetic medicine. This may matter more than the acquisition headlines. Drug development does not always fail at molecule design. Some products succeed clinically and then stumble on capacity, yield, device configuration, supply chain or launch speed. Lilly's current advantage is that the cash generated by a mega-franchise can be directed immediately into the clinical programs and factories required for the next generation of molecules. Scale creates speed. It also creates risk. The broader the pipeline becomes, the harder resource prioritization gets. The wider the platform set, the greater the integration cost. If acquired assets remain trapped in early clinical development, today's capital intensity can become tomorrow's impairment intensity. ## 06 | Taiwan read-through: Caliway is a research case, not a Lilly concept stock For Taiwan, the closest public research reference is Caliway Biopharmaceuticals, ticker 6919. The company has submitted an IND to the U.S. FDA for a multicenter Phase 2 trial of CBL-514 in combination with tirzepatide. The indication is reduction of abdominal subcutaneous fat in adults with overweight or obesity. The planned trial is expected to enroll 120 people and use MRI-measured change in abdominal subcutaneous fat volume as the primary endpoint. The study logic is that after GLP-1 therapy reduces overall body weight, persistent subcutaneous fat and body-composition issues may become a separate clinical problem. The evidence boundary must be clear. As of August 11, 2026, the company had submitted the Phase 2 IND. That is not the same as FDA allowing the trial to begin. The rebound-weight improvement data cited by the company came from animal research, and human clinical benefit has not yet been established. Caliway is not part of Lilly's supply chain, and the use of Zepbound in a trial should not be packaged as a Lilly concept-stock story. The more useful question is whether a large GLP-1 market creates clinically testable complementary needs for Taiwanese companies, not whether local names can be forced into the same main battlefield as global giants. Lilly's greatest strength today is tirzepatide. Its biggest strategic problem is also that tirzepatide has become too strong. One quarter of $22.974 billion in revenue gives Lilly the capital to pursue oral drugs, triple agonists, sleep disorders, CAR-T, vaccines and psychiatry at the same time. If those investments become products, Lilly may be buying the next decade. If they merely become a stack of expensive early-stage assets, today's strength will magnify tomorrow's gap. The next drug king is never simply bought. It has to grow through clinical trials, regulation, factories and markets, one gate at a time. ## Disclaimer This article is for industry and business analysis only. It does not constitute medical advice, investment advice, or a recommendation to buy or sell any security. Treatment decisions should be made by qualified healthcare professionals based on individual circumstances. --- ## 一季狂賣 230 億美元，禮來卻急著買下一個藥王 URL: https://drugnews.com.tw/articles/2026-08-18-eli-lilly-tirzepatide-next-drug-king.html Date: 2026-08-18 Language: zh-Hant Summary: Tirzepatide 讓禮來同時吃下糖尿病與減重兩個大市場，但收入集中、價格壓力與下一代產品接班，才是市場真正該看的投資問題。 禮來又交出一張幾乎不像大藥廠的成績單。 2026 年第二季營收 229.74 億美元，年增 48%；淨利 70.95 億美元，年增 25%。公司也把全年營收指引上調至 850 億至 870 億美元。其中，糖尿病藥 Mounjaro 賣了 99.43 億美元，減重藥 Zepbound 賣了 49.28 億美元。 兩個品牌加起來 148.71 億美元，占禮來整季營收 64.7%。 但別被兩個名字騙了：Mounjaro 與 Zepbound 背後，都是同一個分子 tirzepatide（替爾泊肽）。 換句話說，禮來目前看似有兩台印鈔機，核心引擎其實只有一具。 這也解釋了為什麼業績越好，禮來買公司的速度反而越快。最舒服的順風局，最怕只有一個引擎。 ——— 【01｜同一個分子，吃下兩個大市場】💉 Tirzepatide 是 GIP／GLP-1 雙重腸泌素促效劑。用在第二型糖尿病時，以 Mounjaro 銷售；用在肥胖或符合條件的過重成人體重管理時，美國品牌是 Zepbound。 這個分子的厲害之處，先在臨床端建立。 在 751 人參與的 SURMOUNT-5 第三期頭對頭研究中，沒有糖尿病的肥胖或過重成人治療 72 週後，tirzepatide 組平均減重 20.2%，semaglutide 組為 13.7%。這項結果支持 tirzepatide 在該研究條件下的減重優勢；它不代表所有患者都能得到相同效果，也不能用來推論不同適應症的療效。 接著是商業執行。禮來把同一個分子放進糖尿病、體重管理與睡眠呼吸中止等市場，再用 LillyDirect、自費方案、保險准入與產能擴張，把臨床差異轉成處方量。 第二季數字正是這套機器的結果： 📌 Mounjaro：99.43 億美元，年增 91%。 📌 Zepbound：49.28 億美元，年增 46%。 📌 兩者合計：148.71 億美元，占季度營收 64.7%。 上半年也幾乎一樣：兩品牌合計 276.93 億美元，占禮來 427.73 億美元營收的 64.7%。這不是單季偶然，而是公司收入結構。 ——— 【02｜禮來領先了，價格戰也已經開始】⚔️ 高速成長背後，財報藏著另一組值得警覺的數字。 禮來第二季銷量增加 60%，實現價格卻下降 13%。美國市場若排除回扣與折扣估計調整，整體實現價格約下降 9%。這個 9% 是公司美國業務口徑，不能偷換成「Zepbound 單品降價 9%」；但方向很清楚：GLP-1 市場正在用更低價格換更大覆蓋。 中國也呈現同一種交換。Mounjaro 納入中國國家醫保藥品目錄後，海外實現價格下降，銷量則大幅增加，推動海外 Mounjaro 營收年增 172%。 而且，諾和諾德遠沒有退出戰場。該公司 2026 年第二季調整後銷售仍以固定匯率成長 7%；截至 7 月 17 日，Wegovy 口服錠在美國單週處方已超過 26.5 萬張。 所以，禮來贏的是目前這一輪產品與執行，不是永久壟斷。 產品越成功，支付端越會要求降價；競爭者越多，患者與醫師的選項越多。今天的 60% 銷量成長，必須持續跑得比價格下滑更快，營收才守得住。 ——— 【03｜口服、三靶點：禮來先接好自己的下一棒】🧬 禮來沒有等 tirzepatide 放慢才準備接班。 第一棒是 Foundayo（orforglipron），一款每日一次的口服小分子 GLP-1 藥物。FDA 在 2026 年 4 月 1 日核准它用於符合條件的肥胖或過重成人，4 月 6 日開始出貨；第一個銷售季度已貢獻 9,800 萬美元。 這裡要說清楚：Foundayo 目前在美國核准的是體重管理，第二型糖尿病適應症仍在送件階段。它的價值不是取代所有針劑，而是把不想打針、注射不便或偏好口服的族群拉進治療市場。 第二棒是 retatrutide。這款每週一次的 GIP／GLP-1／升糖素三重受體促效劑仍在研究中。TRIUMPH-1 第三期頂線數據顯示，12 mg 組依療效估計策略計算，80 週平均減重 28.3%；若依治療方案估計策略計算，則為 25.0%。禮來規劃 2027 年第一季在美國提出申請。 一個分子守住現金流，一顆藥錠擴大入口，一個三重促效劑往療效上限再推。這條接力如果順利，禮來就能把單品週期拉成平台週期。 但任何一棒出問題，都會讓市場重新檢視目前給 tirzepatide 的高成長溢價。Foundayo 需要證明長期處方黏著度；retatrutide 仍要經過監管審查與真實世界使用；產能則必須跟得上更大的患者池。 ——— 【04｜買睡眠、CAR-T、疫苗：禮來在買第二座山】🧩 除了自己做，禮來也把資金撒向代謝疾病之外。 第二季完成的收購包括 Orna、Ajax、Centessa 與 Kelonia；季後又完成 Curevo、LimmaTech、Vaccine Company 三筆感染症收購，並簽下 AtaiBeckley 的併購協議。 方向看似很散，背後其實有一條共同標準：先拿到有臨床入口或平台壁壘的資產，再用禮來的開發、製造與商業化能力把它放大。 🔹 Centessa：睡眠—清醒障礙。交易已完成，交割現金對價約 63 億美元，另有最高 15 億美元或有價值權（CVR）。 🔹 Orna／Kelonia：體內 CAR-T 與基因遞送。Orna 交易最高 24 億美元；Kelonia 交割預付款 32.5 億美元，含里程碑的最高總對價 70 億美元。 🔹 Curevo／LimmaTech／Vaccine Company：帶狀疱疹、抗藥性細菌與 EBV 疫苗。三筆公告的最高總對價合計約 38.3 億美元。 🔹 AtaiBeckley：難治型憂鬱症等精神疾病。禮來已簽約，交割現金對價約 28 億美元，另有最高 10 億美元 CVR；截至查證日仍應視為待完成交易。 這裡最容易犯的錯，是把所有「最高金額」加起來，直接說禮來已經花掉多少錢。里程碑與 CVR 要達成臨床、法規或商業條件才會支付，公告上限不等於當期現金支出。 財報已經先看得到買公司的代價：第二季收購中研發資產費用達 28 億美元，另有 7.03 億美元減損、重整與其他特殊費用，主要與 Kelonia、Centessa 的交割與整合相關。 這是一場很昂貴的接班工程。買進來的是選擇權，不是保證書。 ——— 【05｜真正的護城河，是把科學送進工廠】🏭 禮來 5 月再加碼 45 億美元擴建印第安納州兩座製造基地，2020 年以來在當地承諾的資本支出已超過 210 億美元。新設施涵蓋原料藥、Foundayo、retatrutide 與基因醫學。 這一步比併購新聞更關鍵。 製藥業最常見的失敗，未必發生在分子設計。有些產品臨床成功，最後卻卡在產能、良率、裝置、供應鏈或上市速度。禮來現在的優勢，是能把超級單品產生的現金，直接灌進下一代分子的臨床與工廠。 規模帶來速度，也帶來另一種風險：管線越多，資源排序越難；平台越廣，整合成本越高。若買來的資產長期停在早期臨床，今天的資本密度，明天就可能變成減損密度。 ——— 【06｜台灣座標：康霈是研究案例，不是禮來概念股】🔍 台灣若要找直接相關的研究座標，目前可看康霈生技（6919）。 公司已向美國 FDA 遞交 CBL-514 合併 tirzepatide 的多中心二期臨床試驗 IND，適應症是「減少過重或肥胖成人的腹部皮下脂肪」，預計納入 120 人，以 MRI 量測腹部皮下脂肪體積變化。 這項研究的邏輯，是在 GLP-1 降低整體體重後，進一步處理頑固皮下脂肪與停藥後體組成問題。 但證據邊界要畫清楚：截至 2026 年 8 月 11 日，公司僅完成二期 IND 送件，尚不等於 FDA 已准予試驗啟動；公司提出的復胖改善數據來自動物研究，人體臨床效益尚未證實。康霈也不是禮來供應鏈，更不能因為研究使用 Zepbound 就包裝成「禮來概念股」。 真正值得觀察的，是大型 GLP-1 市場出現後，台灣公司能否找到可被臨床驗證的互補問題，而不是在同一條主賽道上硬碰巨頭。 ——— 禮來現在最強的地方，是 tirzepatide；最需要解決的問題，也正是 tirzepatide 太強。 一季 230 億美元營收，讓它有本錢同時押注口服藥、三重促效劑、睡眠、CAR-T、疫苗與精神醫學。這些投資若能長成產品，禮來買到的會是下一個十年；若只長成一疊昂貴的早期管線，今天的盛況反而會放大明天的落差。 真正的藥王，從來不是被買回來的。 它得在臨床、監管、工廠與市場裡，一關一關長出來。 ## 免責聲明 本文為產業資訊整理與商業分析，不構成醫療建議、投資建議或任何買賣有價證券之推薦。藥物使用與治療選擇應由合格醫療人員依個別情況評估。 --- ## Revenue Jumped 39%, Yet Iovance Still Lost $47 Million: Can Cancer Cell Therapy Become a Business? URL: https://drugnews.com.tw/articles/2026-08-17-iovance-amtagvi-til-commercialization-en.html Date: 2026-08-17 Language: en Summary: Iovance's Amtagvi rebounded sharply in the second quarter, but the central question is no longer whether TIL therapy can work. It is whether a personalized cell therapy can become a reproducible, reimbursable and scalable medical service. A year ago, the market was still asking whether a cancer cell therapy manufactured individually for every patient could ever become a real business. In the second quarter of 2026, Iovance gave investors a much stronger answer than before. Total product revenue reached $99.31 million, up 66% year over year and 39% sequentially. U.S. Amtagvi revenue was about $90.72 million, above the company's prior quarterly guidance range of $79 million to $81 million. The numbers are bright. But the important question is no longer the old one: whether TIL therapy can work. Amtagvi is already the first FDA-approved tumor-infiltrating lymphocyte therapy. The question now is whether a treatment that requires tumor resection, patient-specific cell expansion and hospital reinfusion can be turned into a stable, reproducible and reimbursable medical service. This competition has moved far beyond the laboratory. ## 01 | This rebound is about a process that is beginning to work Iovance disclosed that its authorized treatment center network now exceeds 95 sites across the United States, Canada and Australia, with roughly one-third of the network represented by community centers. It also reported that turnaround time from tumor receipt to product availability has been reduced to under 31 days. Those figures need careful reading. The 95 sites are not all in the United States, and the sub-31-day turnaround is an operational disclosure, not a guarantee that every patient can complete therapy within that period. The other key change is manufacturing. In the first quarter of 2026, Iovance centralized commercial and clinical manufacturing at its own iCTC cell therapy center. The company says the site has potential annual capacity for more than 5,000 patients. Centralized manufacturing, a broader treatment-center network and shorter turnaround time together created the conditions for the second-quarter revenue rebound. Financially, the picture also improved. Iovance reported product gross margin of 56%. But that gross margin excludes depreciation and amortization. The company still recorded a $51.91 million operating loss and a $47.32 million net loss for the quarter. It also did not raise full-year revenue guidance after beating its prior quarterly target. Instead, it said it was reviewing its 2026 revenue guidance of $350 million to $370 million and expected to update the market in the third quarter. So this was a rebound worth acknowledging. It was not proof that the profit model is already settled. Revenue could have been affected by surgery schedules, manufacturing release, bed availability or payer approvals that accumulated from earlier periods. Until the company discloses fuller conversion metrics, a 39% sequential increase should not be drawn as a straight line into every future quarter. ## 02 | TIL is not a bag of drug. It is a relay that cannot break In simplified terms, physicians first remove part of a patient's tumor, identify immune cells that have already infiltrated the tumor, send those cells to a manufacturing site for expansion, prepare the patient with lymphodepleting chemotherapy, reinfuse the expanded cells and then support activation with interleukin-2. Every step from referral, sampling, manufacturing, shipping, hospitalization and infusion can become a dropout point. The first gate is referral. A patient has to be recognized by the treating oncologist as a potential TIL candidate and referred to a center capable of both surgery and cell therapy. A listed treatment center is not the same as a mature referral network. The second gate is tumor harvest and waiting time. The patient must have resectable tumor tissue, and the disease must remain controlled enough during the manufacturing period for the patient to proceed to treatment. For rapidly progressing advanced cancer, time itself is clinical risk. The third gate is manufacturing and product release. Each batch belongs to one patient. Unlike an antibody drug, it cannot be produced in bulk, stored and then distributed. A failed quality release or logistics delay may mean a lost manufacturing batch and, more importantly, a missed treatment window. The fourth gate is hospitalization and payment. Surgery, lymphodepletion, infusion, IL-2, bed scheduling and adverse-event management all have to fit on the same timeline. If payer prior authorization is delayed, referral and tumor harvest may still fail to translate into revenue. Iovance's 10-Q makes this funnel explicit: product revenue is recognized when the patient completes infusion. A new center, a resected tumor or a manufactured batch does not immediately become revenue. The financial statement only sees the result after the patient completes the whole path. The FDA label also leaves a concrete historical reminder. In the registration dataset, 189 patients had tumor tissue resected, but 33 ultimately did not receive Amtagvi. Reasons included manufacturing failure, disease progression, death, adverse events from lymphodepletion and patient withdrawal or alternative treatment. That is why manufacturing success is only one part of the story. Whether patients can survive the waiting period, hospitals can arrange surgery and beds, and payers can approve the total episode of care will all show up in revenue. ## 03 | The overlooked cost is inside the hospital Amtagvi is not an outpatient injection. The FDA label requires administration in an inpatient hospital setting with access to intensive care support. The full treatment includes surgery, lymphodepleting chemotherapy, cell infusion, IL-2, hospitalization and adverse-event management. Payers are evaluating a full care episode, not simply the price of a cell product. Safety cannot be hidden behind revenue growth either. The FDA used different analysis sets for different safety questions. Among 160 patients who initiated the full regimen, 12 deaths, or 7.5%, were considered by FDA to be at least possibly related to the treatment regimen. In Cohort 4, 21 of 89 infused patients, or 23.6%, required intensive care for reasons not related to the infusion itself. In the main safety population of 156 patients, 137 patients, or 87.8%, had at least one Grade 4 treatment-emergent adverse event within 30 days after infusion. Those figures should not be lazily described as the adverse-event rate of Amtagvi alone. The submitted regimen includes lymphodepleting chemotherapy and IL-2, and FDA noted that attribution across the entire regimen is difficult. This is why Iovance has to standardize patient selection, bed planning, multidisciplinary coordination and adverse-event management. TIL commercialization capability and hospital execution capability are effectively the same issue. ## 04 | Ninety million dollars is not the finish line. It is the entry ticket The quarter at least demonstrated that personalized cell therapy is not destined to remain confined to a handful of academic centers. Community centers are joining, manufacturing is centralized and turnaround time is improving. That suggests the process may be capable of replication. But three pieces remain unresolved. First, the company has not disclosed an updated commercial manufacturing success rate or patient dropout rate. Investors still cannot see the full conversion rate from tumor resection to actual infusion. Second, 56% product gross margin has not produced profitability, and the reported figure excludes depreciation and amortization. As volume increases, investors still need evidence that equipment, labor, quality control and failed batches can be absorbed efficiently. Third, Amtagvi is under accelerated approval. The objective response rate supporting FDA approval was 31.5%, but continued approval depends on confirmatory evidence. The ongoing Phase 3 trial is expected to complete its primary analysis in 2028. A single quarter of better sales cannot replace long-term clinical evidence. ## 05 | Taiwan should focus on the whole chain from factory to bedside For Taiwan, the most direct public reference point is Timing Pharma, ticker 6659 on the emerging stock market. The company discloses a GTP-compliant cell preparation center, a TAF-accredited laboratory and a TIL project with Far Eastern Memorial Hospital. The disclosed status remains under application. This only shows that Taiwanese groups are accumulating cell manufacturing and clinical collaboration capabilities. It does not make Timing Pharma an Iovance supplier, a direct peer competitor or an automatic beneficiary of Iovance's commercial progress. Other local cell-therapy companies should not be inserted into the story unless they have direct public TIL evidence. The metrics worth tracking are operational rather than promotional: the success rate from tumor harvest to infusion, manufacturing turnaround days, patients treated per center, reimbursement for the full treatment episode and complete gross margin after volume increases. The meaning of Iovance's second quarter is that the market can now see a clearer possibility that personalized cancer cell therapy may become a scalable system. Whether it can become a profitable long-term business is still not answered by this quarter's revenue. The answer lies in whether the next group of patients can move through the entire path faster, more safely and more consistently. ## Disclaimer This article is for industry and business analysis only. It does not constitute medical advice, investment advice, or a recommendation to buy or sell any security. Treatment decisions should be made by qualified healthcare professionals based on individual circumstances. --- ## 營收跳增 39%，還是虧了 4,700 萬美元：癌症細胞療法翻身了？ URL: https://drugnews.com.tw/articles/2026-08-17-iovance-amtagvi-til-commercialization.html Date: 2026-08-17 Language: zh-Hant Summary: Iovance 的 Amtagvi 第二季營收明顯反彈，但真正要驗證的不是 TIL 是否有療效，而是個人化細胞療法能否被做成可複製、可支付、可放大的醫療服務。 一年前，市場還在懷疑：這種每位患者都要重新製造一次的癌症細胞療法，真的能做成生意嗎？ 2026 年第二季，Iovance 給出一張明顯反彈的成績單。總產品營收 9,931 萬美元，年增 66%、季增 39%；其中，美國 Amtagvi 營收約 9,072 萬美元，也高於公司先前提出的 7,900 萬至 8,100 萬美元季度指引。 數字很亮眼，真正重要的卻不是「TIL（腫瘤浸潤淋巴球）有沒有用」這個老問題。 Amtagvi 已經是 FDA 核准的第一款腫瘤浸潤淋巴球療法。現在市場要驗證的是：一個必須開刀取腫瘤、替每位患者個別培養細胞、再送回醫院輸注的療程，能不能被做成穩定、可複製、付得起錢的醫療服務？ 這場競爭，勝負早已走出實驗室。 ——— 【01｜這次反彈，賣的是流程開始順了】📈 Iovance 第二季揭露，Amtagvi 的授權治療中心（ATC）網路已超過 95 家，約三分之一是社區型中心；從接收腫瘤組織到產品可運送的周轉時間，已壓到 31 天以內。 但有一個口徑必須說清楚：95 家是美國、加拿大與澳洲合計，不是全在美國。31 天也是公司的營運揭露，不等於每位患者都能在 31 天內完成治療。 另一個關鍵，是 Iovance 在 2026 年第一季把商業與臨床製造集中到自有細胞治療中心（iCTC）。公司表示，這座工廠潛在年產能超過 5,000 名患者。中央製造、中心擴張與周轉縮短一起發生，才讓第二季營收有條件往上跳。 財務也確實改善：公司列示的產品毛利率升至 56%。不過這個數字排除了折舊與攤銷；當季仍有 5,191 萬美元營業損失、4,732 萬美元淨損。更值得注意的是，公司沒有趁著超越季度指引就上調全年目標，而是表示正在重新檢視 3.5 億至 3.7 億美元的 2026 年營收指引，預計第三季再更新。 所以，這是一季值得肯定的反彈，還不是獲利模式已經定型。營收可能受到前一季手術排程、工廠交付、病床釋出或支付核准集中認列的影響；在公司尚未公開完整轉換率前，不能把 39% 的季增直接畫成往後每季都會延續的直線。 ——— 【02｜TIL 不是一袋藥，而是一條不能斷的接力】🧬 簡單說，醫師先切下一部分腫瘤，找出已經進入腫瘤內部的免疫細胞，再送到工廠大量培養；患者接受清除原有淋巴球的化療後，才把這批細胞輸回體內，並搭配介白素-2（IL-2）支持細胞活化。 從取樣、製造、運送、住院到輸注，每一棒都可能讓患者中途掉隊。 這是一個比「多開幾家中心」更難的轉換漏斗。 第一關是轉介。患者要先被原本的腫瘤科醫師認出適合 TIL，再送到能執行手術與細胞治療的中心。中心掛牌存在，不等於轉介網路已經成熟。 第二關是取樣與等待。患者必須有可切除的腫瘤組織，細胞送進工廠後，病情還要在等待期間維持到能接受後續療程。對快速惡化的晚期癌症來說，時間本身就是臨床風險。 第三關是製造與放行。每一批產品只屬於一名患者，無法像抗體藥物那樣先大量製造再分裝；品質不合格或物流延誤，報廢的可能是一批貨，錯過的則是患者的治療窗口。 第四關是住院與支付。手術、清淋化療、輸注、IL-2、病床與不良事件處理必須排在同一條時間線上；保險事前核准若拖延，前面完成的轉介與取樣也可能無法轉成收入。 Iovance 的 10-Q 也把這個漏斗寫得很清楚：產品收入要到患者真正完成輸注時才認列。換句話說，新增一家中心、取出一塊腫瘤、甚至做出一批細胞，都還不能立刻算成營收；只有患者走完整條路，財報才看得到結果。 FDA 標籤留下了一組很直觀的歷史資料：送審試驗中，189 名已完成腫瘤切除的患者，有 33 人最後沒有接受 Amtagvi。原因包括產品無法製造、疾病惡化、死亡、清淋療程不良事件，以及患者改用其他治療或退出。 這也說明，細胞做不做得出來只是其中一題。患者能否撐過等待期、醫院能否排到手術與病床、療程費用能否獲得支付，全部都會反映在最後的收入上。 ——— 【03｜最容易被忽略的成本，其實在醫院裡】🏥 Amtagvi 不是在門診打一針就能回家。FDA 標籤要求在住院醫院使用，現場還必須具備加護病房（ICU）照護能力。療程涉及手術、清淋化療、細胞輸注、IL-2、住院與不良事件處理，保險公司評估的也不是一袋細胞的價格，而是整套照護成本。 安全性也不能被漂亮營收掩蓋。FDA 分別使用三個不同的分析集：160 名啟動完整療程者中，有 12 人（7.5%）死亡被 FDA 判定與研究療程至少可能相關；Cohort 4 接受輸注的 89 人中，有 21 人（23.6%）因非輸注相關照護需求轉入加護病房（ICU）；主要安全性分析集 156 人中，有 137 人（87.8%）於輸注後 30 天內至少出現一項第四級（Grade 4，危及生命等級）治療後不良事件。 但這些數字不能偷懶寫成「Amtagvi 單藥副作用率」。送審療程同時包含清淋化療與 IL-2，FDA 也指出，很難把每一項風險完全拆給單一環節。 這正是 Iovance 必須把患者篩選、床位安排、跨科團隊與不良事件處理一起標準化的原因。TIL 的商業化能力，與醫院的臨床執行能力其實是同一件事。 ——— 【04｜9,000 萬美元不是終點，只是一張入場券】⚠️ 這一季至少證明，個人化細胞療法並非注定只能停在少數學術中心。 社區中心開始加入、製造集中、周轉時間縮短，代表流程有機會被複製到更多地方。 但長期故事還缺三塊拼圖。 第一，公司沒有公開最新的商業製造成功率與患者流失率，我們還看不到從腫瘤切除到真正輸注的完整轉換率。 第二，56% 毛利率仍未帶來獲利，而且公司揭露的毛利率排除了折舊與攤銷。產量放大後，設備、人工、品質管制與失敗批次能否被有效攤薄，還要再看幾季。 第三，Amtagvi 目前是加速核准。FDA 核准時的客觀緩解率（ORR，也就是腫瘤明顯縮小的患者比例）為 31.5%，持續核准仍要等確認性第三期試驗。該研究預計 2028 年才完成主要分析，單季銷售反彈不能替代長期臨床證據。 ——— 【05｜台灣真正該看的是「工廠到病房」整條鏈】🔍 台灣市場若要找產業座標，目前較直接的公開案例是興櫃的天明製藥（6659）。公司官網揭露具備符合人體細胞組織優良操作規範（GTP）的細胞製備中心與 TAF 認證實驗室，也列出與亞東紀念醫院合作的 TIL 計畫，但目前狀態仍是「申請中」。 這只能說明台灣已有團隊在累積細胞製備與臨床協作能力，不能把天明寫成 Iovance 的供應鏈、同級競爭者或直接受惠者。其餘常見細胞治療公司，若沒有直接 TIL 公開證據，也不該為了湊台股名單硬塞進來。 真正值得追蹤的，不是哪家公司說自己也能做細胞，而是五個營運指標：從取樣到輸注的成功率、製造周轉天數、單一中心能治療多少人、整套照護能否順利獲得支付，以及放量後的完整毛利率。 Iovance 第二季的意義，是讓市場第一次較清楚地看到：個人化癌症細胞療法的流程，可能被做成一套可以擴張的系統。 至於它能不能成為一門長期獲利的生意，答案仍不在這一季的營收，而在下一批患者能否更快、更安全、更穩定地走完整條路。 ## 免責聲明 本文為產業資訊整理與商業分析，不構成醫療建議、投資建議或任何買賣有價證券之推薦。藥物使用與治療選擇應由合格醫療人員依個別情況評估。 --- ## Pfizer Paid Billions for Obesity Assets. Is It Already Cutting Its Losses? URL: https://drugnews.com.tw/articles/2026-08-16-obesity-drug-shakeout-pfizer-roche-amgen-en.html Date: 2026-08-16 Language: en Summary: Pfizer, Roche and Amgen are pruning obesity programs that lack clear differentiation. The GLP-1 market has moved from a race to enter the clinic to a contest across efficacy, safety, convenience and commercial execution. ------------------------------------------------ ## Another Wave of Obesity Drugs Falls: Ordinary Players No Longer Have a Ticket to the GLP-1 Market The obesity-drug shakeout is accelerating. During the second-quarter 2026 earnings season, several obesity programs that had once carried substantial expectations were removed from Big Pharma pipelines. The most visible example came from Pfizer. Multiple oral GLP-1 assets obtained through its expensive acquisition of Metsera were rapidly cleared out. Roche has also begun to prune parts of the obesity portfolio it acquired with Carmot Therapeutics. Amgen, meanwhile, has closed an early obesity program and concentrated its resources on the few core assets it believes can still break through. On the surface, the explanations are familiar: the data were not good enough, tolerability was not clean enough, or internal priorities changed. The deeper reason is that the competitive bar in obesity has been reset. Tirzepatide and semaglutide have already established efficacy, reimbursement experience, physician familiarity and brand trust. The market no longer needs another weight-loss drug that is merely comparable. Any late entrant now has to answer a brutal question: exactly where is it better than Eli Lilly and Novo Nordisk? - Does it produce greater weight loss? - Is it safer or easier to tolerate? - Is dosing more convenient? - Does it improve more obesity-related complications? - Can it serve needs that current incretin therapies do not adequately cover? If the answer is unclear, even a clinical-stage asset can be abandoned quickly. This round of pipeline cuts does not mean that large drugmakers have collectively lost confidence in obesity. It means they now accept that the market has moved beyond the era in which simply owning a GLP-1 program was enough to create a compelling story. ----------------- ## 01 | Pfizer: An Expensive Oral GLP-1 Portfolio, Cut Without Hesitation Pfizer provides the most dramatic case. It was among the earliest large pharmaceutical companies to pursue oral GLP-1 drugs. From lotiglipron to danuglipron, and later its acquisition of Metsera, Pfizer did not lack effort or ambition. Instead, it encountered several of the hardest problems in the field: liver toxicity, gastrointestinal tolerability, insufficient efficacy, dose design and an increasingly unforgiving comparison with powerful marketed products. Pfizer announced its acquisition of Metsera in 2025 as a way to re-enter the obesity race through external business development. One of Metsera's most closely watched assets was MET-224o, an oral, ultra-long-acting GLP-1 receptor agonist. It was once regarded as a key component of Pfizer's effort to challenge Lilly and Novo Nordisk in oral obesity therapy. By the second quarter of 2026, however, Pfizer had removed MET-224o from its pipeline update. The company also discontinued its internally developed PF-07976016. Fierce Biotech reported that Pfizer terminated four programs in the quarterly clearout, including the Metsera-derived MET-224o and the GIP receptor candidate PF-07976016. The decision left much of the oral GLP-1 combination that had justified Pfizer's renewed obesity push effectively cleared away. The irony is that Pfizer did not lack strategic imagination. It had envisioned an obesity matrix spanning oral agents, monthly and quarterly dosing, and multiple mechanisms. It had also planned to combine a GIP receptor antagonist with a GLP-1 receptor agonist in an attempt to reproduce or surpass the synergistic logic behind tirzepatide. But a coherent portfolio narrative does not guarantee a clinical or commercial pass. The removal of MET-224o shows that the threshold for oral GLP-1 development is higher than many investors assumed. An oral product must deliver more than convenience; it also has to solve exposure, dose, tolerability, manufacturability and competitive efficacy. The discontinuation of PF-07976016 makes a related point. A GIP receptor antagonist combination without a clearly demonstrated efficacy increment can lose value if administration becomes more complicated and clinical acceptance remains uncertain. Pfizer still wants to compete. The difference is that today's obesity market no longer allows a merely plausible asset to consume large amounts of capital indefinitely. ----------------- ## 02 | Roche and Amgen Are Pruning Too: No Advantage Is a Disadvantage The same logic is visible at Roche and Amgen. Roche's acquisition of Carmot Therapeutics was a major step into obesity. The transaction brought in several incretin-related assets, including enicepatide, formerly CT-388, and acmopatide, formerly CT-868. Roche nevertheless moved quickly to prioritize. Acmopatide is a once-daily dual GLP-1/GIP receptor agonist. Although an earlier study in people with type 1 diabetes showed a signal of HbA1c improvement, Roche chose not to advance the program and redirected resources toward enicepatide. BioSpace reported that the company was prioritizing enicepatide across obesity, type 2 diabetes and cardiovascular outcomes. Amgen has made a similar choice. Its second-quarter 2026 pipeline update confirmed that the company stopped developing the early obesity candidate AMG 513 while continuing to advance its lead obesity program, MariTide, or maridebart cafraglutide, also known as AMG 133. These discontinued programs were not necessarily devoid of biological activity. The problem is that activity alone is no longer an entry ticket. New drugs must compete against products that are already marketed, scaling and familiar to prescribers. Lilly's tirzepatide has pushed injectable weight loss to roughly 20% or more in major studies. Its next-generation candidate retatrutide, a triple GIP, GLP-1 and glucagon receptor agonist, produced an average 24.2% weight reduction at the 12 mg dose after 48 weeks in a Phase 2 trial. Against that standard, an early candidate showing only 10% to 15% weight loss, without a decisive tolerability or convenience advantage, may not justify a Phase 3 program enrolling thousands of participants and requiring development spending at the billion-dollar scale. The rule for the second half of the obesity race is straightforward: lacking an advantage is itself a disadvantage. ----------------- ## 03 | Why Are Large Drugmakers Cutting So Aggressively? Phase 3 Is Expensive and the Market Is Crowded Obesity is not a small indication. Once a program reaches Phase 3, the sponsor faces large sample sizes, long follow-up and development across multiple adjacent indications. Cardiovascular outcomes, obstructive sleep apnea, knee osteoarthritis, metabolic dysfunction-associated steatohepatitis and diabetes can all become part of the evidence and commercial battlefield. This is not a game that can be completed with tens of millions of dollars. It is a billion-dollar competition. The market is also moving from unconstrained expansion toward direct competition for patients, prescribers, reimbursement and supply. The practical questions are now harder: - Which product will patients choose? - Which one will physicians prescribe? - Which therapy will insurers reimburse? - Can the manufacturer's supply chain keep pace? - Will long-term adherence hold? - What happens to body weight after treatment stops? Novo Nordisk and Lilly have already built moats in brand, manufacturing capacity, payer experience and clinical evidence. A later entrant therefore needs a clear point of differentiation on at least one important dimension. That is why pipeline cuts at Pfizer, Roche and Amgen should not automatically be read as failure. Pruning can be evidence of strategic maturity. In innovative drug development, failing quickly can be more valuable than keeping an uncompetitive program alive. A company that acknowledges insufficient differentiation early can redirect capital and talent toward assets with a genuine chance of changing the market. ----------------- ## 04 | Cutting Ordinary Assets Does Not Mean Abandoning Obesity This cleanup does not signal a Big Pharma exit from obesity. It shows that companies are concentrating their resources on flagship assets. Pfizer's leading obesity asset is now berobenatide, or MET-097i, an injectable ultra-long-acting GLP-1 receptor agonist. In June 2026, Pfizer reported Phase 2b VESPER-1 extension data showing an unadjusted mean weight reduction of 15.9% at the 2.4 mg dose after 32 weeks, with no apparent plateau. The company has also described plans for more than 20 studies across obesity and related comorbidities, including multiple Phase 3 trials. Roche is placing its emphasis on enicepatide. In Phase 2 data released in 2026, the once-weekly subcutaneous drug produced a placebo-adjusted 22.5% weight reduction at 48 weeks in the highest 24 mg dose group. The effect had not reached a plateau, and the safety profile was reported as consistent with the drug class. Amgen's lead bet is MariTide. Its differentiation is not simply another GLP-1 mechanism, but the possibility of substantially less frequent dosing. If monthly or even less frequent administration can preserve efficacy and tolerability in Phase 3, MariTide could appeal to patients who do not want weekly injections or who need a more convenient option after prior incretin treatment. Amgen continues to advance the program, with Phase 3 readouts expected in 2027. The multinational companies are therefore concentrating, not retreating. - An oral asset that is not good enough gets cut. - An early mechanism without clear differentiation gets cut. - A program whose adverse effects erode its efficacy proposition gets cut. Capital is being reserved for products that could genuinely alter the competitive landscape. ----------------- ## 05 | The Taiwan Read-Through: Caliway and ANYA Approach the Market from Different Angles The Taiwan comparison requires a clear boundary. No listed Taiwanese biotech company currently offers a direct equivalent to global-scale incretin assets such as tirzepatide, semaglutide or retatrutide. But several adjacent strategies are worth tracking. The first is Caliway Biopharmaceuticals, listed under ticker 6919. Caliway's CBL-514 is neither a GLP-1 therapy nor a systemic weight-loss drug. It is a small-molecule treatment being developed for large-area local fat reduction. That makes its route distinct from the global GLP-1 battlefield, but potentially relevant to the next stage of obesity medicine. As patients lose weight on incretin therapies, body composition, regional fat, weight maintenance and post-treatment rebound can become additional treatment needs. In 2026, Caliway presented animal data at ECO and ADA involving CBL-514 in combination with GLP-1 receptor-based therapies. The company has framed the work around adipose-tissue remodeling and durability of weight reduction. The second is ANYA Biopharm, ticker 7776, an investment within the Center Laboratories group. ANYA is developing an oral peptide platform. Its ANY-002 is a GLP-1 receptor agonist, and public information from Center Laboratories says the product has been licensed to two international listed companies. This approach matters because Pfizer's experience demonstrates why oral GLP-1 development is not simply a matter of converting an injectable product into a tablet. The hard problems include absorption, stability, dosing, gastrointestinal tolerability, cost and large-scale manufacturing. An oral peptide platform will attract serious international interest only if it can demonstrate a meaningful advantage across those dimensions. The clearest Taiwan read-through is therefore: - Caliway represents the potential need for adipose-tissue remodeling and weight maintenance in the GLP-1 era. - ANYA tests whether an oral peptide platform can overcome the technical barriers that have defeated multiple oral GLP-1 candidates. Neither company sits in the front row of the global incretin contest, but both connect to changes now unfolding across obesity medicine. ----------------- ## Conclusion | Obesity Drugs Are Not Cooling Down. The Bar Is Rising Another group of obesity candidates has fallen, but that does not mean the market is ending. The opposite is true. Obesity drug development is entering a more demanding and more mature phase. Early in the cycle, an association with GLP-1 was often enough for capital markets to assign strategic value. Now investors and drugmakers demand tangible differentiation. An oral drug must prove that it is not merely convenient, but also effective, tolerable and manufacturable. A long-acting injectable must show that lower dosing frequency can preserve both efficacy and tolerability. A multi-target drug must demonstrate that adding mechanisms produces better clinical outcomes, rather than assuming that more targets are inherently better. A non-incretin therapy must prove that it can address needs current GLP-1 drugs leave unresolved. This pipeline cleanup is Big Pharma delivering an expensive message to the market: the obesity opportunity remains enormous, but there is no longer room for ordinary players. The survivors will not be the companies with the most elaborate GLP-1 narrative. They will be the ones that can establish a clear advantage in efficacy, safety, convenience, tolerability, manufacturing cost or an unmet clinical need. Every late entrant should answer one question before committing to pivotal development: Why would physicians and patients choose this drug instead of a product from Lilly or Novo Nordisk? ----------------- ### Disclaimer This article is provided solely for industry research and educational purposes. It does not constitute investment, medical, fundraising or securities advice. --- ## 百億美元搶來的減肥藥，輝瑞冤大頭又丟掉了？ URL: https://drugnews.com.tw/articles/2026-08-16-obesity-drug-shakeout-pfizer-roche-amgen.html Date: 2026-08-16 Language: zh-Hant Summary: Pfizer、Roche 與 Amgen 同步清理缺乏差異化的肥胖症管線，顯示 GLP-1 下半場的門檻已從「有藥可做」升高為療效、安全、便利性與商業化能力的全面競爭。 ------------------------------------------------ 【又一批減重藥倒下：GLP-1 的下半場，普通玩家已經沒有門票】 減重藥的淘汰賽，還在加速。 2026 年第二季財報季，一批原本被寄予厚望的 obesity 管線被大型藥廠直接砍掉。 最受矚目的，是 Pfizer 以高價收購 Metsera 後取得的多款口服 GLP-1 資產，快速被清理；Roche 從 Carmot Therapeutics 收進來的部分減重管線，也開始陸續退出；Amgen 則關閉早期減重候選藥，把火力集中到真正有機會突圍的核心資產。 表面上，這些項目被砍，是因為數據不夠好、安全性不夠漂亮、或內部優先順序改變。 但更深層的原因，是減重市場的門檻已經被徹底拉高。 當 tirzepatide 和 semaglutide 已經建立療效、支付、醫師使用經驗與品牌信任之後，市場不再需要一款「差不多」的減重藥。 後來者若想進場，必須回答一個非常殘酷的問題：你到底比 Lilly 和 Novo Nordisk 好在哪裡？ 📌 是減重更多？ 📌 安全性更好？ 📌 給藥更方便？ 📌 能改善更多併發症？ 📌 還是能吃下 GLP-1 沒有覆蓋的新需求？ 如果答案不清楚，就算已經進入臨床，也可能被快速放棄。這一輪管線調整，不是大型藥廠集體退縮，而是它們開始承認：減重藥市場已經不再是「誰有 GLP-1，誰就有故事」的時代。 ----------------- 【01｜Pfizer：百億美元買來的口服 GLP-1，說砍就砍】 最戲劇化的，是 Pfizer。 Pfizer 曾是最早布局口服 GLP-1 的大型藥廠之一。從 lotiglipron 到 danuglipron，再到後來透過併購 Metsera 補強管線，它不是沒有努力過，而是一路踩到這個市場最難的幾個坑：肝毒性、胃腸耐受性、療效不足、劑量設計，以及和已上市強勢產品的差距。 2025 年，Pfizer 宣布收購 Metsera，希望透過外部 BD 重回肥胖症賽道。當時 Metsera 最受關注的資產之一，是口服、超長效 GLP-1 receptor agonist MET-224o。這款藥曾被視為 Pfizer 重新挑戰 Lilly 與 Novo Nordisk 口服減重市場的關鍵拼圖。但到了 2026 年第二季，Pfizer 在管線更新中移除 MET-224o，同時也砍掉自研的 PF-07976016。 Fierce Biotech 報導指出，Pfizer 這次一口氣終止四款管線，其中就包含來自 Metsera 的 MET-224o，以及 GIPR 相關候選藥 PF-07976016。這代表 Pfizer 重金押注的口服 GLP-1 組合幾乎被清空。 更諷刺的是，Pfizer 不是沒有想像力。 它曾試圖建立一套涵蓋口服、月度、季度、多機制的肥胖症矩陣；也曾規劃把 GIPR 拮抗劑與 GLP-1 receptor agonist 聯用，希望複製甚至超越 tirzepatide 的協同邏輯。但臨床市場不會因為故事完整就給通行證。 📌MET-224o 被砍，說明口服 GLP-1 的門檻比想像中更高。 📌PF-07976016 被停，也說明 GIPR 拮抗聯用若沒有明確療效增益，反而可能因給藥方式複雜、臨床接受度不明而失去價值。 Pfizer 不是不想贏，而是現在減重市場已經不允許「看起來有機會」的藥繼續大量燒錢。 ----------------- 【02｜Roche、Amgen 也在清倉：無優勢，就是劣勢】 同樣的故事，也發生在 Roche 與 Amgen 身上。 📌 Roche 進入減重賽道的重要一步，是收購 Carmot Therapeutics。 這筆交易讓 Roche 拿到多款 incretin 相關資產，包括 enicepatide（CT-388）與 acmopatide（CT-868）。但 Roche 很快做出取捨。Acmopatide 是每日一次 GLP-1/GIP dual receptor agonist。即使先前在第 1 型糖尿病患者中看到 HbA1c 改善訊號，Roche 仍選擇停止後續開發，把資源轉向 enicepatide。BioSpace 報導指出，Roche 放棄 acmopatide，是為了優先推進 enicepatide 在減重、第二型糖尿病與心血管結局方向的開發。 📌Amgen 也在做同樣的事。 Reuters 報導，Amgen 在 2026 年第二季停止開發早期肥胖症候選藥 AMG 513，但仍持續推進其核心減重藥 MariTide（maridebart cafraglutide，AMG 133）。 這些管線不是完全無效。問題是，在今天的減重市場，「有效」已經不是門票。後來者要面對的是已上市、已放量、已被醫師熟悉的巨頭產品。 Lilly 的 tirzepatide 已經把注射型減重效果推到 20% 左右以上；下一代 retatrutide 作為 GIP／GLP-1／glucagon receptor triple agonist，在 Phase 2 研究中，12 mg 劑量 48 週平均減重達 24.2%。 在這種標準下，後來者若只有 10% 到 15% 的早期減重、或副作用與給藥便利性沒有明顯優勢，就很難說服公司投入數千人三期試驗與十億美元級開發成本。 ✅ 減重藥下半場的規則很簡單：沒有優勢，就是劣勢。 ----------------- 【03｜為什麼大型藥廠砍得這麼狠？因為 Phase 3 太貴，市場太擠】 減重藥不是小適應症。 一旦進入 Phase 3，就要面對巨大樣本數、長時間追蹤、多適應症布局、心血管結局、睡眠呼吸中止、膝骨關節炎、MASH、糖尿病等延伸戰場。 這不是幾千萬美元能玩完的遊戲，而是十億美元級競爭。更重要的是，減重市場已經進入存量競爭。早期大家想像的是一個無限擴張的藍海，但現在真正的問題變成： ✅患者會選誰？ ✅醫師會開誰？ ✅保險會給付誰？ ✅藥廠供應鏈能不能撐住？ ✅長期依從性好不好？ ✅停藥後體重反彈怎麼辦？ 當 Novo Nordisk 和 Lilly 已經建立品牌、產能、支付與臨床資料護城河，後進者要突圍，必須在某個維度形成清楚差異。 所以，Pfizer、Roche、Amgen 砍管線不是壞事。這反而是一種成熟。 📌 在創新藥開發裡，「快速失敗」有時比「拖著不死」更重要。 早點承認一款藥沒有足夠競爭力，把資金與團隊集中到真正有機會的資產上，才是大型藥廠能在高競爭市場活下來的方式。 ----------------- 【04｜砍掉普通管線，不代表放棄減重市場】 這一輪管線清理，並不代表 Big Pharma 放棄減重，相反，它們正在把資源集中到王牌資產。 📌 Pfizer 現在最核心的減重資產，是 berobenatide（MET-097i），一款超長效 GLP-1 receptor agonist 注射劑。Pfizer 2026 年 6 月公布 Phase 2b VESPER-1 資料，表示 berobenatide 在 32 週延長研究中，2.4 mg 劑量未經安慰劑校正平均體重下降 15.9%，且仍未看到減重平台期；公司也規劃推進超過 20 項肥胖與相關共病試驗，包括多項 Phase 3。 📌 Roche 則押注 enicepatide（CT-388）。Roche 2026 年公布 Phase 2 資料顯示，enicepatide 每週一次皮下注射，在最高 24 mg 劑量組 48 週達到 22.5% 安慰劑校正體重下降，且未達減重平台期，安全性也與同類藥物一致。 📌Amgen 的核心，是 MariTide。MariTide 的差異化不在於「又一個 GLP-1」，而在於更長給藥間隔。若每月甚至更低頻給藥能在 Phase 3 證明療效與耐受性，它有機會切入那些不想每週打針、或已使用 GLP-1 但需要更便利方案的患者。Reuters 也指出，Amgen 仍在持續推進 MariTide，Phase 3 資料預計 2027 年讀出。 也就是說，MNC 不是退出，而是集中。 📌 口服藥不夠好，就砍。 📌 早期機制沒有差異，就砍。 📌 副作用拖累療效，就砍。 資源要留給真正能改變競爭格局的產品。 ----------------- 【05｜台灣可以怎麼看？康霈、奧孟亞，各站在不同切口】 這個題目放回台灣，要先講清楚：台灣目前沒有直接對標 tirzepatide、semaglutide、retatrutide 這類全球級 incretin 藥物的上市櫃公司，但有幾個相鄰方向值得觀察。 📌 第一個是康霈生技（6919）。 康霈的 CBL-514 不是 GLP-1，也不是全身性減重藥，而是大範圍局部減脂小分子新藥。 這條路線和全球 GLP-1 戰場不同，但反而很貼近下一階段肥胖醫學的需求：當患者用 GLP-1 類藥物體重下降後，身體組成、局部脂肪、減重維持與停藥後反彈，會成為新的治療需求。康霈 2026 年曾在 ECO 與 ADA 發表 CBL-514 併用 GLP-1R 類藥物的動物數據，公司也表示該方向聚焦於脂肪組織重塑與減重持久性。 📌第二個是晟德集團轉投資的奧孟亞（7776)。 奧孟亞主打口服胜肽平台，其 ANY-002 為 GLP-1 receptor agonist，晟德公開資料也提到該產品已授權兩家國際上市公司。這條路線對台灣很重要，因為 Pfizer 的教訓剛好說明：口服 GLP-1 不是只要把注射劑變成口服就成功。 真正難的是吸收、穩定性、劑量、胃腸耐受性、成本與大規模製造。如果奧孟亞這類平台未來能證明在口服胜肽上有真正差異，才有機會被國際買方看見。 所以台灣段落最自然的連結是： 📌 康霈看 GLP-1 時代後的脂肪組織重塑與減重維持。 📌 奧孟亞看口服胜肽平台能否突破口服 GLP-1 的技術門檻。 這兩條都不是主戰場第一排，但都能接上全球肥胖醫學正在發生的變化。 ----------------- 【結語｜減重藥沒有退燒，是門檻變高了】 又一批減重藥倒下，不代表減重市場結束。 恰好相反，這代表市場進入更殘酷、也更成熟的階段。 早期，只要碰到 GLP-1，資本就願意給想像。 現在，市場要的是實打實的差異化。 📌 口服藥要證明不只方便，還要有效、安全、可製造。 📌 長效注射要證明不只打得少，還要維持療效與耐受性。 📌 多靶點藥物要證明不是機制越多越好，而是臨床結果真的更好。 📌 非 incretin 類藥物則要證明自己能補上 GLP-1 沒有處理好的問題。 這輪管線清理，是大型藥廠用最昂貴的方式告訴市場：減重藥賽道還很大，但普通玩家已經沒位置。 未來能活下來的，不是 GLP-1 故事講得最完整的公司，而是能在療效、安全性、便利性、耐受性、製造成本或新需求上，真正拿出明確優勢的公司。 對後來者來說，開發前最好先問自己一句： ✅ 這款藥，究竟有哪一點能讓醫師和患者不選 Lilly、不選 Novo Nordisk，而選我？ ----------------- ### 免責聲明 --- ## GSK's US$2 Billion Cough-Drug Bet Hit a Phase 3 Reproducibility Wall URL: https://drugnews.com.tw/articles/2026-08-15-gsk-camlipixant-calm-phase3-reproducibility-en.html Date: 2026-08-15 Language: en Summary: After GSK's roughly US$2 billion acquisition of BELLUS Health, CALM-1 met its endpoint but CALM-2 did not reproduce the result. The divergence shows why Phase 2 promise, patient experience and reproducibility must move together. GSK paid roughly US$2 billion for a chronic-cough drug. Three years later, the programme stopped in Phase 3. On 17 July 2026, GSK said it would discontinue development of camlipixant for refractory chronic cough (RCC). The uncomfortable part was not a clean sweep of negative trials. CALM-1 met its primary endpoint, while CALM-2 failed to reproduce the result. When GSK acquired BELLUS Health in 2023, the investment case looked unusually coherent: a large patient population, no approved treatment in the United States or Europe, encouraging Phase 2 data, and a P2X3 inhibitor that might avoid the taste disturbance that had complicated the class. In CALM-1, camlipixant 50 mg twice daily met the primary endpoint at Week 12. In CALM-2, the same dose did not achieve statistical significance at Week 24. The 25 mg twice-daily dose missed in both studies, and key secondary measures—including the Chronic Cough Diary (CCD)—also failed to reach their prespecified thresholds. One positive pivotal trial and one negative one can sound better than an across-the-board failure. For a medicine expected to establish a new standard of care, however, it was not enough. GSK's US$2 billion P2X3 bet ran into the most expensive problem in late-stage drug development: a signal that could not be reproduced consistently. That does not invalidate the entire cough target. It does force investors and developers to ask the question that Phase 3 exists to answer: can the early result happen again? ### 01 | Read CALM-1 and CALM-2 separately GSK reported the combined headline results from CALM-1 and CALM-2 on 17 July 2026. Both were Phase 3 trials in adults with refractory chronic cough. RCC refers to cough lasting more than eight weeks that persists despite appropriate treatment of possible underlying causes, or cannot be adequately explained by an identifiable cause. For patients, this is not an occasional need to clear the throat. It can mean hundreds of coughs a day, disrupted sleep, impaired work, social isolation and a sustained loss of quality of life. CALM-1's primary efficacy time point was Week 12; CALM-2's was Week 24. Both studies measured 24-hour cough frequency and compared camlipixant with placebo. The headline results were: - **CALM-1:** camlipixant 50 mg twice daily significantly reduced 24-hour cough frequency versus placebo at Week 12 and met the primary endpoint. - **CALM-2:** the same 50 mg twice-daily regimen did not achieve statistical significance at Week 24. - **25 mg twice daily:** neither study reached statistical significance. - **Key secondary endpoints:** both trials missed prespecified thresholds, including measures based on the Chronic Cough Diary. - **Safety:** the overall incidence and severity of treatment-related adverse events were similar between camlipixant and placebo in both trials. There was no obvious headline safety failure, and the pharmacology did not disappear completely. GSK stopped because the overall efficacy was limited and unlikely to transform patient care. The discontinuation applies to the RCC indication. One boundary matters. The Phase 2b BALANCE study in diarrhoea-predominant and mixed irritable bowel syndrome is continuing. Camlipixant has not disappeared from every disease programme, and an RCC outcome cannot be mechanically extrapolated to IBS. ### 02 | Phase 2 gave GSK a credible reason to buy In April 2023, GSK agreed to acquire BELLUS Health for US$14.75 per share in cash, representing an aggregate equity value of approximately US$2 billion. The transaction closed on 28 June. GSK bought the company, but camlipixant was clearly the central strategic asset. The SOOTHE Phase 2b study provided the supporting signal. The 50 mg and 200 mg twice-daily regimens both met the primary endpoint, with placebo-adjusted reductions in 24-hour cough frequency of approximately 34.4% and 34.2%, respectively. Taste-related adverse events occurred in no more than 6.5% of participants across doses. That tolerability profile mattered. P2X3 receptors participate in cough-reflex hypersensitivity, so inhibiting them may reduce cough. But interference with related receptors can also produce taste disturbance, making long-term treatment difficult. Camlipixant was designed as a selective P2X3 antagonist, and the Phase 2 data appeared to combine efficacy with comparatively manageable taste effects. The commercial opportunity also looked compelling. GSK's acquisition announcement estimated that about 28 million people worldwide had chronic cough, roughly 10 million had RCC lasting longer than a year, and about six million lived in the United States and European Union. At the time, neither region had an approved RCC therapy. Those figures were company estimates made in 2023. They should not be treated as a current, precise count of diagnosed or addressable patients. They do explain the acquisition logic: a large population, clear unmet need, a late-stage asset, encouraging Phase 2 efficacy, and a potential tolerability advantage within the class. GSK then expected an approval and launch in 2026, with accretion to adjusted earnings per share from 2027. In a 2024 presentation, the company also assigned camlipixant peak-year sales potential of more than £2.5 billion. The currency and status of that estimate matter. It was expressed in pounds sterling and represented a forward-looking, unrisked product potential—not realised revenue. ### 03 | Why one successful Phase 3 study could not rescue the programme If CALM-1 was positive, why could GSK not simply file on the strength of the successful study? A p-value below a statistical threshold is only the first requirement for a medicine expected to change practice. The magnitude and durability of benefit must be consistent; patients should feel a meaningful difference; and the overall benefit-risk profile must support chronic use. CALM-1 showed a statistically significant difference at Week 12, whereas CALM-2 did not reproduce that finding at Week 24. Potential explanations include placebo response, patient heterogeneity, execution, timing, effect size or other study-specific factors. GSK has so far disclosed headline conclusions rather than the full data set, so it would be premature to assign a single cause. The Week 12 and Week 24 results also came from two different trials with different participants. They do not prove that efficacy simply faded by Week 24. Assessing durability would require the full time-course data from both trials, baseline characteristics, estimated treatment differences and confidence intervals. Nor does CALM-1's positive primary endpoint mean that every RCC patient benefited. Objective cough monitors count frequency. Patients also care about intensity, night-time disruption, urinary incontinence, fatigue and social limitation. The failure of CCD and other key secondary endpoints to provide support therefore weakens the total evidence: a measurable instrument signal that does not connect consistently with patient experience carries less regulatory and commercial value. The development decision nevertheless had a clear direction. The lower dose missed in both studies, the higher dose succeeded on the primary endpoint in only one, and key secondary endpoints failed to support either study. A programme cannot readily claim to transform patient care when benefit appears at a single trial and time point but does not move several patient-relevant measures together. This is also why a relatively clean safety headline could not save the asset. Similar treatment-related adverse-event rates versus placebo suggest that the topline data did not reveal an obvious tolerability price. In a chronic condition, however, safety must accompany reliable benefit; it cannot replace it. ### 04 | Phase 3 is a new test, not a larger photocopy Drug development is often described as a straight line: Phase 1 checks safety, Phase 2 finds a signal, Phase 3 expands the sample, and approval follows. Reality is closer to sitting the exam again. Phase 2 trials are usually smaller. Patient and site selection can be more concentrated, while dose, endpoint and statistical assumptions are still being refined. Phase 3 moves the proposed effect into a larger and more heterogeneous population, more centres, longer follow-up and a stricter execution environment. A larger sample can improve statistical power, but it also exposes disease heterogeneity, measurement noise and real-world operational difficulty. CALM's central problem is that a promising Phase 2 signal did not translate into consistent Phase 3 evidence. SOOTHE reported placebo-adjusted reductions of roughly 34% with both 50 mg and 200 mg twice daily, alongside a low incidence of taste-related events. In Phase 3, GSK evaluated 25 mg and 50 mg. The higher dose met the primary endpoint in one study and missed in the other; the lower dose missed in both. The results are not a simple biological zero, but they do not support a stable benefit-risk conclusion. The public evidence available at the time supported moving into Phase 3. CALM-1 and CALM-2 were the new information that changed the decision. The more useful diligence questions are therefore not whether GSK should have predicted the future perfectly. They are how the company modelled variability in cough-frequency measurement, placebo response, dose selection and joint Phase 3 success—and whether management reallocated resources promptly once the evidence changed. ### 05 | What the market now has to reprice It is tempting to say that US$2 billion simply disappeared. The accounting conclusion requires more discipline. GSK paid approximately US$2 billion for all of BELLUS Health's equity. Any impairment charge must be read from the company's financial reporting rather than inferred from the clinical decision. What investors can already reprice are the lost peak-sales scenario, the resulting gap in GSK's 2031 product portfolio, and management's record of converting external business development into durable products. The failure of one asset has not, by itself, rewritten GSK's 2026 full-year guidance or its 2031 ambitions. The 17 July programme update did not present a withdrawal or restatement of group guidance as its central message. ### 06 | Does P2X3 still have a path forward? After camlipixant's discontinuation in RCC, it is easy to declare P2X3 another false target. That conclusion moves too quickly. The high-dose CALM-1 result shows measurable activity in at least one study at a defined time point. The broader development and regulatory history of P2X3 inhibitors also indicates that the mechanism has human evidence. The unresolved challenge is how inhibition, selectivity, patient segmentation, endpoints and follow-up duration must be combined to produce a benefit that is both meaningful and reproducible. A target can be biologically valid while a particular molecule, dose, population or study design is insufficient. Conversely, one positive trial cannot validate an entire class. Investors should therefore watch the full subgroup analyses, baseline cough frequency, placebo response, time-course curves, patient-reported outcomes and whether the taste profile still resembles the advantage suggested in Phase 2. ### 07 | Three things Taiwan-based readers should track There is not enough primary evidence to identify a Taiwan-listed company as a co-developer, manufacturer, clinical partner or commercial beneficiary of camlipixant. Forcing a local stock connection would create a claim that the evidence does not support. Three industry questions are more useful: 1. **The complete CALM data.** The topline release does not yet show whether the discrepancy was driven mainly by effect size, patient mix, placebo response, timing or another factor. 2. **Whether BALANCE continues and when it reads out.** IBS has different biology and endpoints. RCC failure cannot prove that the IBS study will fail, just as an ongoing trial cannot be treated as evidence of a second life for the molecule. 3. **How GSK accounts for the BELLUS transaction and reallocates R&D resources.** This is where the clinical stop will eventually become visible on the financial statements and in the portfolio. ### Conclusion | Phase 3 makes a compelling story prove itself again Camlipixant's path is severe but familiar. Phase 2 offered efficacy, a potentially favourable taste profile and substantial unmet need. A global pharmaceutical company paid US$2 billion, planned for a 2026 launch and assigned more than £2.5 billion in peak-year sales potential. Then one pivotal trial met its primary endpoint, another did not, and key secondary endpoints failed to support either study. GSK stopped development in RCC. The Phase 2 evidence was once sufficient to justify Phase 3. It was not sufficient to reproduce the result across two pivotal studies. The US$2 billion transaction bought the right to test a late-stage hypothesis—not certainty of approval. When the evidence no longer supported the programme, stopping became the final development decision. ### Disclaimer This article is an analysis of biopharma and industry trends. It does not constitute medical diagnosis, treatment advice or investment advice. Trial results and subsequent development status should be verified against company disclosures, trial registries and regulatory materials. --- ## GSK 砸 20 億美元買止咳新藥，三期卻翻車！ URL: https://drugnews.com.tw/articles/2026-08-15-gsk-camlipixant-calm-phase3-reproducibility.html Date: 2026-08-15 Language: zh-Hant Summary: GSK 約 20 億美元收購 BELLUS Health 後，camlipixant 的 CALM-1 達標、CALM-2 卻未能複製。這場第三期落差揭示：二期訊號、患者感受與可重現性必須一起過關。 GSK 砸 20 億美元買止咳新藥，三期卻翻車！ ────────── GSK 曾花約 20 億美元買下一款慢性咳嗽新藥。三年後，它停在第三期。 2026 年 7 月 17 日，GSK 宣布停止 camlipixant 在難治性慢性咳嗽的開發。更棘手的是：CALM-1 過關，CALM-2 卻無法複製。 2023 年收購 BELLUS Health 時，市場看到的是一個很完整的故事：患者多、歐美缺乏核准治療、二期訊號漂亮，camlipixant 還可能避開同類藥最麻煩的味覺副作用。 CALM-1 的 50 mg 每日兩次，在第 12 週達到主要終點；問題是 CALM-2 的同一劑量，在第 24 週沒有達到統計顯著。25 mg 每日兩次則在兩項研究都沒過，慢性咳嗽日記等關鍵次要終點也沒有達到預設門檻。 一個試驗過、一個試驗沒過，聽起來比「全面失敗」好一點；對一個想成為新標準的藥，卻仍然不夠。 GSK 花 20 億美元押下的 P2X3 資產，在三期撞上了「無法重現」。這沒有替整個止咳靶點判死刑，卻把 CALM 最昂貴的教訓攤在桌上：早期訊號，能不能再做一次？ ────────── 📌 01｜先把 CALM-1 與 CALM-2 分開看 GSK 在 2026 年 7 月 17 日公布 CALM-1 與 CALM-2 綜合結果。 兩項都是針對成人難治性慢性咳嗽（refractory chronic cough, RCC）的第三期試驗。RCC 指咳嗽持續超過八週，已針對可能的原因接受適當治療後仍持續，或找不到能充分解釋的原因。對患者來說，這不是偶爾清喉嚨，而可能是一天反覆數百次、影響睡眠、工作、社交與生活品質的長期疾病。 CALM-1 的主要療效時間點是第 12 週；CALM-2 則是第 24 週。兩個研究都測量 24 小時咳嗽頻率，並比較 camlipixant 與安慰劑。 正式結果如下： - CALM-1：50 mg 每日兩次，在第 12 週相較安慰劑顯著降低 24 小時咳嗽頻率，主要終點達標； - CALM-2：相同 50 mg 每日兩次，在第 24 週沒有達到統計顯著； - 25 mg 每日兩次：兩項試驗都沒有達到統計顯著； - 關鍵次要終點：包括慢性咳嗽日記（CCD）在內，兩項試驗都沒有達到目標門檻； - 安全性：兩項試驗中，camlipixant 與安慰劑的治療相關不良事件整體發生率與嚴重度相近。 安全性沒有爆雷，藥理活性也未完全消失；讓 GSK 停手的是整體療效有限，難以改變患者照護。公司停止的是 RCC 適應症後續開發。 另一個邊界也要講清楚：評估腹瀉型與混合型腸躁症的 IIb 期 BALANCE 試驗仍會繼續，camlipixant 尚未從所有疾病開發中消失。 ────────── 🧪 02｜二期看起來，真的很有理由買 回到 2023 年 4 月，GSK 宣布以每股 14.75 美元現金收購 BELLUS Health，總股權價值約 20 億美元；6 月 28 日，公司另行公告交易完成。GSK 買的是整家公司，但 camlipixant 顯然是核心理由。 支持收購的是 SOOTHE IIb 期研究。50 mg 與 200 mg 每日兩次都達到主要終點，經安慰劑校正的 24 小時咳嗽頻率降幅分別約 34.4% 與 34.2%。味覺相關不良事件在不同劑量的發生率不高於 6.5%。 這個安全性輪廓很重要。P2X3 受體參與咳嗽反射過度敏感，抑制它有機會降低咳嗽；但如果藥物同時明顯干擾相關受體，可能帶來味覺異常，讓患者難以長期服用。camlipixant 被設計成具選擇性的 P2X3 拮抗劑，二期資料看起來同時保留療效與味覺耐受性。 市場空間也很誘人。GSK 收購公告引用的估計是，全球約 2,800 萬人有慢性咳嗽，其中約 1,000 萬人的 RCC 超過一年，美國與歐盟約 600 萬人。當時美國與歐盟沒有核准的 RCC 藥物。 這些數字都是 2023 年收購時的公司估計，不能當成今天精準的診斷人數或可治療市場。不過它們解釋了 GSK 為何願意出手：大患者池、明確未滿足需求、後期資產、二期訊號，以及可能優於同類的味覺安全性。 GSK 當時預期 2026 年核准與上市，並希望從 2027 年起增厚調整後每股盈餘。2024 年的公司簡報更曾給出「峰值年銷售潛力超過 25 億英鎊」的估計。 這個口徑是英鎊，而且屬於公司前瞻、未經風險調整的產品潛力，尚未形成收入。 ────────── ⚖️ 03｜一個三期過關，為何仍救不了專案？ 讀到 CALM-1 過關，直覺會問：為什麼 GSK 不拿成功的那一個繼續申請？ 一個 p 值小於門檻，只是新藥改變臨床的第一關。效果是否穩定、病人是否真的感受到、風險效益能否支持長期使用，缺一不可。 CALM-1 在第 12 週看到顯著差異，CALM-2 在更長的第 24 週沒有複製。這可能涉及安慰劑反應、患者異質性、試驗執行、時間點、效果大小或其他因素；GSK 目前只公布頂線結論，完整數據尚待會議或論文，因此不能替公司指定單一失敗原因。 第 12 週與第 24 週來自兩項不同試驗、不同受試者，無法直接推論「藥效到第 24 週就消失」。若要判斷療效是否隨時間衰退，至少要看到兩項試驗各時間點的完整曲線、基線特徵、估計差異與信賴區間。 同理，CALM-1 的陽性也不表示所有 RCC 患者都受益。客觀咳嗽監測器算到的是頻率，患者在意的還有咳嗽強度、夜間干擾、尿失禁、疲勞與社交限制。這正是 CCD 等關鍵次要終點沒有支持，會讓整體證據變得更弱的原因：儀器上的變化若沒有穩定連到患者感受，商業與監管價值都會打折。 但開發決策已經有足夠方向：低劑量兩個研究都沒過，高劑量只過一個主要終點，兩個研究的關鍵次要終點也沒有支持。若藥物只在某個試驗、某個時間點顯示統計優勢，卻無法讓多項患者相關指標一起移動，要宣稱它能「transform patient care」就很困難。 這也是為什麼安全性沒有救回專案。camlipixant 的治療相關不良事件與安慰劑相近，表示它至少沒有在頂線資料出現明顯的安全性代價；慢性病上市還需要穩定療效，再乾淨的耐受性也補不上這一格。 ────────── 🔁 04｜三期是一場重新驗證 新藥故事常被寫成一條直線：一期看安全、二期看訊號、三期放大樣本，最後核准。 現實比較像重新考一次。 二期研究通常規模較小，受試者與試驗中心選擇也可能較集中；劑量、終點與統計假設仍在摸索。看到正面訊號後，三期必須把那個效果放進更大的患者群、更多中心、更長追蹤與更嚴格執行環境。樣本變大可以提高統計能力，也會把疾病異質性、量測雜訊與真實執行難度一起放大。 CALM 最刺眼之處，是二期訊號沒有穩定轉成三期的一致性。 SOOTHE 的 50 mg 與 200 mg 都看到約 34% 的安慰劑校正降幅，味覺事件又少，這讓資產看起來相當可推進。到了三期，GSK 改以 25 mg 與 50 mg 評估；高劑量在一項試驗達標，另一項沒過，低劑量則兩項都沒過。結果無法化成零與一；這組數據很難支撐穩定的風險效益結論。 當時能看到的公開證據支持進三期；CALM-1 與 CALM-2 的結果，才是後來新增的關鍵資訊。 更合理的問題是：GSK 在盡職調查中如何評估咳嗽頻率的量測變異、安慰劑反應、劑量選擇與兩個三期的成功機率？以及資料延後後，管理層是否及時重新分配資源？這些才是收購品質的核心。 ────────── 💰 05｜20 億美元，市場重砍四筆帳 標題很容易寫成「20 億美元打水漂」，但財務上要保守。 GSK 以約 20 億美元買下 BELLUS Health 全部股權，實際減損仍要等財報。市場現在同時重估消失的峰值銷售想像、2031 年產品組合缺口，以及 GSK 外部併購的命中率。 單一資產失敗尚未改寫 GSK 的 2026 全年財測或 2031 年目標。公司沒有正式更新指引，7 月 17 日的專案公告也沒有把全年指引寫成撤回或重申的核心事項。 ────────── 🧭 06｜P2X3 還有沒有下一條路？ camlipixant 停止後，很容易把 P2X3 寫成「又一個偽靶點」。這個結論走太快。 CALM-1 的高劑量主要終點達標，本身就顯示藥物在至少一個研究、特定時間點有可測量的作用；其他 P2X3 路線的開發與監管歷史，也說明這個生物機制已有人體證據。尚待解決的是抑制程度、選擇性、患者分層、終點與追蹤時間，怎麼組合才能得到足夠且可重現的效益。 靶點可以是對的，分子、劑量、族群或試驗設計仍可能不夠好；反過來，一個試驗陽性也不能替整個靶點背書。 這個區分很重要。投資人接下來應追完整亞組、基線咳嗽頻率、安慰劑反應、不同時間點曲線、患者報告結果，以及味覺相關事件是否仍維持二期優勢。 ────────── 🇹🇼 07｜台灣讀者應該追的三件事 目前沒有足夠一級證據，能把台灣上市櫃公司列為 camlipixant 的共同開發、製造、臨床或商業化受益者，因此不硬湊台股。 產業上更值得追的是： 第一，GSK 後續發表的 CALM 完整數據。現在只有頂線結果，尚不能判斷失敗主要來自效果大小、患者差異或時間點。 第二，BALANCE IIb 期是否繼續、何時讀出。腸躁症的疾病機制與終點不同，RCC 失敗不能直接推導 IBS 一定失敗；同樣也不能因為專案還在，就先說分子有第二次生命。 第三，GSK 財報如何處理 BELLUS 交易相關資產與研發優先順序。這才會告訴投資人，臨床止損最後如何落到帳上。 ────────── 🎯 結語｜三期的價值，是讓漂亮故事接受複製 camlipixant 的歷程很殘酷，也很典型。 二期看見療效、味覺安全性與巨大未滿足需求；大藥廠用 20 億美元買進，設定 2026 年上市與超過 25 億英鎊峰值銷售潛力；最後，一個三期過、一個三期沒過，兩個研究的關鍵次要終點又都沒有支持，GSK 選擇停止 RCC。 二期資料曾經足以支持進三期，卻沒能在兩個關鍵試驗中重現。20 億美元買到的是驗證假設的機會；當證據不夠，及時停手就是最後一項開發決策。 參考資料（查證截止：2026-07-19）： 1. GSK, CALM-1／CALM-2 phase III update, 2026-07-17 https://www.gsk.com/en-gb/media/press-releases/gsk-provides-an-update-on-calm-1-and-calm-2-phase-iii-trials/ 2. GSK, agreement to acquire BELLUS Health, 2023-04-18 https://www.gsk.com/en-gb/media/press-releases/gsk-reaches-agreement-to-acquire-late-stage-biopharmaceutical-company-bellus-health/ 3. GSK, completion of BELLUS Health acquisition, 2023-06-28 https://www.gsk.com/en-gb/media/press-releases/gsk-completes-acquisition-of-bellus-health/ 4. SOOTHE original peer-reviewed paper, phase 2b, NCT04678206 https://academic.oup.com/ajrccm/article/211/6/1038/8513749 5. ClinicalTrials.gov, SOOTHE, NCT04678206 https://clinicaltrials.gov/study/NCT04678206 6. GSK, J.P. Morgan Healthcare Conference transcript（峰值年銷售潛力） https://www.gsk.com/media/tsvlrat3/jp-morgan-healthcare-conference-transcript.pdf 7. ClinicalTrials.gov, CALM-1, NCT05599191 https://clinicaltrials.gov/study/NCT05599191 8. ClinicalTrials.gov, CALM-2, NCT05600777 https://clinicaltrials.gov/study/NCT05600777 9. ClinicalTrials.gov, BALANCE, NCT07519395 https://clinicaltrials.gov/study/NCT07519395 #GSK #Camlipixant #P2X3 #慢性咳嗽 #臨床三期 #新藥研發 #Biotech #藥時事 免責聲明：本文為生技醫藥與產業趨勢整理，不構成醫療診斷、治療建議或投資建議。試驗結果與後續開發狀態應以公司公告、試驗登錄與監管資料為準。 --- ## This Taiwanese Biotech Is Building a Nanorocket Inside the Human Body URL: https://drugnews.com.tw/articles/2026-08-13-intelligene-rna-delivery-nanorocket-en.html Date: 2026-08-13 Language: en Summary: Oligonucleotide therapeutics are moving from sequence design to extrahepatic delivery. Intelligene is using EV and LNP platforms to target the lung and airways, testing whether RNA medicines can truly reach disease sites beyond the liver. The sequence is the payload. Delivery is the rocket. The company that can move RNA beyond the liver may hold the key to the next valuation cycle in oligonucleotide therapeutics. Source date: August 3, 2026 Some technologies begin as scientific experiments. Only after major pharmaceutical companies commit meaningful capital does the market recognize that they have become durable industry themes. Oligonucleotide (Small RNA) therapeutics are approaching that inflection point. In 2026, Alnylam Pharmaceuticals entered into a collaboration with AI biotech company Inceptive Nucleics valued at up to $2 billion. At first glance, the agreement looks like an RNAi leader placing a strategic bet on artificial intelligence. The deeper message is more consequential: competition in RNA medicines is no longer limited to who can design the most elegant sequence. It is increasingly about who can move a candidate into the clinic faster, more reproducibly, and with a higher probability of success. The agreement includes a $30 million upfront payment, an equity investment, and potential preclinical, regulatory, and commercial milestones, with AI being used to accelerate RNA candidate design and selection. The fit between AI and oligonucleotide drug discovery is straightforward. Traditional small-molecule programs search for binding pockets on proteins, while antibodies generally target extracellular or cell-surface proteins. Oligonucleotides intervene further upstream in the flow of biological information: siRNA can trigger the degradation of a target mRNA, antisense oligonucleotides can alter mRNA degradation or splicing, and miRNA-based approaches may regulate broader gene networks. In other words, oligonucleotide therapeutics shift drug discovery from finding a structure to engineering a sequence. Once drug development begins to resemble sequence engineering, AI becomes a natural accelerator. It can support sequence design, off-target prediction, and chemical-modification optimization while reducing the volume of wet-lab trial and error. But that is only the first half of the value chain. Even the best-designed RNA sequence cannot reach a disease site on its own. The harder challenge comes next. An RNA medicine must survive the body’s defensive barriers, avoid rapid degradation and excessive immune interception, reach the correct organ and cell type, and generate sufficient pharmacologic activity at a tolerable dose. That is the most unforgiving—and potentially the most valuable—problem in the second half of oligonucleotide therapeutics: delivery. *Figure 1 | The first half of oligonucleotide therapeutics is sequence design; the second half is delivery beyond the liver.* > Core Investment Thesis • AI is accelerating the design and selection of RNA sequences. • The next major value inflection will come from reproducible delivery beyond the liver. • A validated extrahepatic delivery route can reframe a company from a single-asset story into a platform asset. ## 01 | Oligonucleotide Therapeutics Have Moved from Scientific Promise to Commercial Reality To distinguish a durable technology platform from a market bubble, three questions matter: Are there products? Are patients using them? Is the category generating revenue? Oligonucleotide therapeutics have begun to produce a credible commercial scorecard. Alnylam is the clearest example. Its RNAi portfolio has crossed the threshold from “can this modality become a drug?” into commercial scale-up. AMVUTTRA (vutrisiran) has become a flagship product for the category, while ONPATTRO (patisiran) remains a landmark in the commercialization of RNAi therapeutics. Alnylam has built its identity around RNA interference as a way to silence disease-causing genes and reduce the production of harmful proteins at the source. Other global pharmaceutical companies have pushed oligonucleotide drugs into increasingly large disease markets. Novartis brought siRNA into hypercholesterolemia with LEQVIO (inclisiran). Biogen’s SPINRAZA (nusinersen) became a defining antisense therapy for spinal muscular atrophy. QFITLIA (fitusiran), developed through the Sanofi–Alnylam partnership, entered prophylactic treatment for hemophilia A and B. Together, these products show that oligonucleotide therapeutics are no longer confined to technically elegant rare-disease programs; they can address chronic disease, inherited disorders, hematology, and infectious disease. An even more consequential signal has emerged in chronic hepatitis B. Bepirovirsen, developed by GSK and Ionis, achieved a 19% functional cure rate in the overall population and 26% in the low-viral-activity subgroup in the Phase 3 B-Well program, compared with 0% in the control arm. GSK has also noted that more than 240 million people worldwide are living with chronic hepatitis B, while current standards of care generally require long-term treatment and deliver low rates of functional cure. Those numbers matter. For decades, the treatment objective in chronic hepatitis B has largely been long-term viral suppression rather than functional cure after a finite course of therapy. If bepirovirsen ultimately reaches the market, it would show that oligonucleotide drugs can do more than prove themselves in rare diseases; they may reshape the treatment logic of major global indications. That is why the category is being re-rated. Oligonucleotide therapeutics are not simply a way to treat one disease. They represent a new, programmable language for intervening in disease biology. *Figure 2 | Commercial RNA medicines, hepatitis B functional-cure data, and extrahepatic delivery are reshaping the small-RNA opportunity.* ## 02 | The Liver Was the First Orbit, Not the Final Destination The liver was the dominant proving ground for the first commercial wave of oligonucleotide therapeutics. The reason is practical. The liver naturally takes up nucleic-acid medicines from the circulation, and mature hepatocyte-targeting technologies such as GalNAc have enabled many siRNA programs to generate clinical and commercial results more efficiently. But maturity also creates crowding. Once the liver route was validated, the market moved to the next question: - Can RNA be delivered beyond the liver? The lung and airways, central nervous system, muscle, adipose tissue, and immune cells are the new territories that next-generation oligonucleotide platform companies are trying to open. These tissues are linked to large markets across chronic disease, infectious disease, neurology, autoimmunity, and oncology. The path will not be easy. RNA is not a conventional small molecule. It is fragile, charged, vulnerable to nuclease degradation, and generally unable to cross cell membranes on its own. Designing the molecule is not enough. It must be protected, transported, released, and activated inside the relevant disease cell. That is why the second half of oligonucleotide therapeutics cannot be reduced to a claim that a company “has a sequence.” *Figure 3 | Organ-level delivery difficulty is beginning to reprice RNA platform companies.* > Payload vs. Rocket • The sequence is the payload. • Delivery is the rocket. • Without a reliable launch system, even the best payload never reaches its destination. ## 03 | Markets Are Repricing RNA Platforms by Organ-Level Difficulty and Degree of Validation The next phase of the market is beginning to value RNA platforms along two dimensions: the difficulty of the target organ and the degree to which the delivery approach has been validated. If a company is making incremental advances within the established liver route, investors will focus primarily on indication size, clinical stage, and commercial execution. If a company can open a new extrahepatic organ, however, the market is no longer evaluating only a single product. It is evaluating whether the company has created a new delivery gateway. That is why muscle, the central nervous system, kidney, adipose tissue, and lung are becoming the new coordinates of the oligonucleotide industry. Several major transactions over the past year illustrate the shift. Novartis acquired Avidity for approximately $12 billion, not simply for exposure to one muscle disease, but for an AOC platform capable of delivering oligonucleotides into muscle tissue. Arrowhead and Novartis entered into a collaboration worth up to $2 billion around ARO-SNCA, pointing toward the more difficult central nervous system. Lilly’s agreement with Ascidian, worth up to $1.9 billion, extended RNA medicines into kidney disease and RNA exon editing. GSK’s collaboration with SiranBio showed that adipocyte-targeted siRNA is also attracting early strategic interest from large pharmaceutical companies. Together, these transactions make one point clear: The competitive question in the second half of oligonucleotide therapeutics is no longer simply whose RNA sequence is better. It is who can be first to open a new organ frontier. Muscle delivery has already reached high-value acquisition territory. The central nervous system remains exceptionally difficult, but the potential market in neurodegenerative disease is enormous. Kidney and adipose tissue are being reopened through RNA editing, siRNA, and new targeting technologies. Lung and airway delivery remain less fully priced by the market, but a clinically validated platform in this area would be highly differentiated. The competitive landscape can be read in three ways. First, by organ. Each organ represents a different technical barrier and a different market opportunity. The liver is established; muscle is receiving high-value validation; and the central nervous system, kidney, adipose tissue, and lung are emerging as the next platform battlegrounds. Second, by position. A company or transaction further to the right on the development spectrum has more clinical or business-development validation. A company further to the left remains at the platform-concept or early-validation stage. This distinction helps investors separate platforms supported by data from platforms still waiting for a decisive milestone. Third, by transaction or valuation scale. A larger value reflects a higher market capitalization, acquisition value, or deal ceiling. The point is not merely to compare who is larger today. It is to use capital-market pricing as a proxy for the value that a validated delivery platform may unlock. *Figure 4 | Intelligene is using lung and airway programs to build an EV and LNP RNA delivery platform.* Table 1. Competitive Landscape for Extrahepatic RNA Delivery | Company / Transaction | Primary Delivery Target | Stage / Validation | Market Cap / Acquisition Value / Deal Ceiling | Technology / Platform | Key Investment Read-Through | | --- | --- | --- | --- | --- | --- | | Alnylam | Primarily liver, expanding toward multiple organs | Late-stage clinical / commercial | Approx. US$38.6B market cap | RNAi / siRNA | Commercial RNAi leader; demonstrates that oligonucleotide drugs can scale into meaningful sales. | | Ionis | Liver, CNS, rare and chronic diseases | Late-stage clinical / commercial | Approx. US$12.2B market cap | ASO / LICA | Bepirovirsen extends the ASO opportunity from rare disease into a major chronic hepatitis B market. | | Avidity | Muscle | Late-stage clinical / near-commercial | Approx. US$12.0B acquisition value | AOC (antibody–oligonucleotide conjugate) | Novartis’s high-value acquisition strategically validated muscle delivery as a platform asset. | | Arrowhead | CNS, lung, cardiometabolic tissues and other targets | Early clinical / BD | Approx. US$10.3B market cap | TRiM RNAi | The ARO-SNCA collaboration with Novartis supports the licensing value of an extrahepatic RNAi platform. | | Dyne | Muscle | Early- to mid-stage clinical | Approx. US$3.1B market cap | FORCE platform | Commonly viewed as an important benchmark for Avidity and the muscle-delivery field. | | Wave | Adipose / metabolism / CNS | Platform concept to early clinical | Approx. US$1.1B market cap | Stereopure oligonucleotides | Shows that extrahepatic delivery still requires validation through clinical endpoints and a clear commercial position. | | Alnylam × Inceptive | Primarily liver; focused on sequence-design capabilities | Early clinical / BD | Deal ceiling: US$2.0B | AI + RNA design | Signals that the sequence-design half of oligonucleotide R&D is entering an AI-enabled industrialization phase. | | Lilly × Ascidian | Kidney | Early clinical / BD | Deal ceiling: US$1.9B | RNA exon editing | Shows RNA value expanding from gene silencing toward functional repair. | | SiranBio × GSK | Adipose tissue | Early clinical / BD | Deal ceiling: US$1.0B | Adipocyte-targeted siRNA | A representative transaction in the emerging field of extrahepatic adipose targeting. | | Intelligene | Lung / respiratory tract | Platform concept to early clinical | NT$6.96B market cap (Aug. 3, 2026; update before publication) | EV + LNP RNA delivery platform | The core question is not current size, but whether IND, Phase 1, and CMC data can validate pulmonary RNA delivery. | Note: Market capitalizations, acquisition values, and deal ceilings are retained from the source draft. Intelligene’s market capitalization is stated as of August 3, 2026 and should be refreshed before publication. Breaking the landscape down by company, target organ, platform, and transaction value makes the underlying message more explicit. This is not a ranking of who has already won. It shows that the oligonucleotide industry is moving from a liver-centric model toward multi-organ delivery competition. Each newly validated organ may become a new entry point for platform re-rating. That is why the market should track not only the companies that have already attracted high-value acquisitions or licensing deals, but also earlier-stage companies attempting to open new organ routes. Within this landscape, the central question for Intelligene is whether its lung and airway strategy can become the next validated extrahepatic delivery entry point. ## 04 | Intelligene Is Building a Nanorocket Inside the Human Body Against this backdrop, Intelligene Inc. (7832) should be viewed through a different lens. What makes the company strategically interesting is not simply that it is developing RNA medicines, nor that its story can be framed around COVID-19 or influenza. It is that Intelligene has chosen the lung and respiratory tract as its initial battleground. The company’s lead program, IG-001, is being developed for COVID-19 and other coronavirus infections. IG-002 is being advanced as a broad-spectrum influenza therapy. On the surface, these are respiratory antiviral candidates. At a platform level, however, the two infectious-disease programs are intended to accelerate the build-out of a broader extrahepatic RNA therapeutic platform for the respiratory system. That distinction matters. If IG-001 is viewed only as a post-pandemic standalone asset, the market is likely to value it against short-term demand. Viewed as a platform program, the objective is more durable: design RNA medicines against relatively conserved viral sequences that are less prone to rapid mutation, thereby suppressing viral replication through a strategy that may be broader and easier to update. Respiratory viruses are not a one-time problem. COVID-19 demonstrated the threat posed by emerging viruses. Influenza provides an annual reminder that seasonal respiratory disease has never disappeared. Older adults, patients with chronic disease, and immunocompromised populations continue to need better antiviral options. Traditional antiviral drugs can face variant escape and resistance. If RNA medicines can target more conserved viral regions, they may support broader-spectrum and more rapidly updateable treatment strategies. That is the basis of Intelligene’s differentiation. The company is not entering the most mature and crowded liver-delivery segment. It is pursuing the more difficult—and potentially more distinctive—lung and airway route, with the longer-term objective of expanding into a strategically important extrahepatic targeting system. Difficulty creates risk. But if the platform works, difficulty is also what makes the value visible. ## 05 | Extracellular Vesicles: The Real Barrier Is CMC, Not Imagination If RNA is the payload, extracellular vesicles are the delivery vehicle—the nanorocket Intelligene is trying to build. Extracellular vesicles, or EVs, can be understood as small, naturally occurring vesicles used by cells to exchange information. According to the company’s disclosures, EVs are approximately 40 to 150 nanometers in size, can carry biological molecules, and are capable of crossing cell membranes and supporting intercellular communication. Intelligene is developing both lipid nanoparticles (LNPs) and EV-based delivery systems. At its GMP pilot plant and process R&D center in the Taipei Biotechnology Park, the company is building process development, scale-up validation, and chemistry, manufacturing, and controls capabilities to support preclinical and early clinical programs. EVs are not magic. The real difficulty is not the story. It is CMC. Which cell source should be used? How can culture conditions be standardized? How will the EVs be purified? How will RNA loading be measured? Is particle-size distribution consistent from batch to batch? Does biological activity remain stable? Can product quality be maintained after manufacturing is scaled? These questions—not the conceptual appeal of exosomes—will determine whether an EV-based medicine can enter the clinic. Many platform companies ultimately fail not because the science is uninteresting, but because manufacturing does not translate. Producing a promising result in the laboratory is not the same as producing clinical-grade drug. A high-quality small batch does not guarantee that the process will remain stable at scale. For that reason, Intelligene’s investment in its own GMP pilot plant and process-development center should not be treated as a simple facilities announcement. The company states that it has established an initial foundation and process stability in EV manufacturing. The pilot facility is scheduled for completion in the second half of the year, followed by equipment installation and pilot production. For an early-stage RNA company, process capability is strategic leverage when discussing clinical development, licensing, and international partnerships. A real nanorocket is not an arrow drawn on a presentation slide. It is a product that can be manufactured consistently, pass quality testing, and support clinical supply batch after batch. For Intelligene’s EV platform to be re-rated, the company will need to demonstrate that EVs can become a manufacturable, controllable, and scalable drug-delivery vehicle. ## 06 | Australia Connects a Taiwan-Built Platform to an International Clinical Pathway A second part of Intelligene’s strategy that warrants attention is its Australian footprint. The company has entered into a collaboration with Australia’s Kirby Institute. The collaboration provides access to BRIDGE program support equivalent to as much as A$5 million—approximately NT$100 million—in research resources for nucleic-acid drug development, linking Australian R&D, clinical-research, and regulatory capabilities to the advancement of RNA therapeutics. Intelligene has also established an overseas R&D center in Australia’s Gold Coast Health and Knowledge Precinct (GCHKP). For an early-stage drug developer, the value of an overseas site is not the address itself. It is whether the site can connect R&D, clinical development, regulatory strategy, and international partnerships into a functioning pathway. Any RNA therapeutic intended for global development will eventually face toxicology packages, clinical-trial design, CMC documentation, regulatory interactions, and diligence by potential licensing partners. Connecting to international development infrastructure earlier can reduce the risk of costly detours later. Viewed together, the Taiwan and Australia strategy has a clearer logic. Taiwan provides the manufacturing base and pilot facility. Australia provides access to R&D, clinical, regulatory, and international collaboration resources. One is the root; the other is the gateway. The former gives the company greater control over manufacturing and quality, while the latter gives the technology a path toward larger markets. For an early-stage biotech company, that is a practical and strategically coherent configuration. ## 07 | The Market Will Now Look for Lift-Off Data Intelligene remains an early-stage, R&D-intensive drug-development company. It is not a mature pharmaceutical business and should not be valued through the price-to-earnings logic applied to revenue-generating companies. The more relevant framework is whether the company can deliver a sequence of de-risking milestones. - Can IG-001/IG-002 complete its preclinical package and advance into Phase 1? - Can IG-001/IG-002 demonstrate broad-spectrum potential across multiple COVID-19 and influenza strains? - Can EV-based candidates advance meaningfully toward Phase 1? - Can the EV platform generate data on RNA loading, stability, delivery efficiency, and batch-to-batch consistency? - Can the Nangang GMP pilot plant complete commissioning and validation and support the manufacture of clinical-trial material? - Can the Australian R&D center translate into tangible clinical, regulatory, and international-partnership progress? These are not slogans. They are measurable, auditable indicators. A company is not truly re-rated by thematic exposure or a memorable tagline alone. It must produce data at the points where the technical risk is highest. For Intelligene, the “nanorocket inside the human body” is more than a marketing metaphor. It is a concise description of the problem the company is trying to solve: turning an RNA sequence created in the laboratory into a drug that can reach disease sites in the lung and respiratory tract. *Figure 5 | Intelligene's next valuation variables include preclinical data, IND progress, Phase 1 entry, CMC execution, and international partnerships.* ## Conclusion | In the Second Half of Oligonucleotide Therapeutics, Value Lies in Reaching the Target The oligonucleotide story is no longer written only in the future tense. Alnylam’s deal of up to $2 billion with Inceptive shows that RNA drug discovery is becoming more industrialized through AI. Lilly’s agreement with Ascidian, worth up to $1.9 billion, shows RNA value expanding from gene silencing toward functional repair and kidney delivery. GSK’s bepirovirsen has created the possibility of functional cure in chronic hepatitis B. Novartis’s approximately $12 billion acquisition of Avidity sends an even clearer message: once extrahepatic RNA delivery produces clinically persuasive evidence, it is no longer a platform narrative—it becomes a strategic asset that large pharmaceutical companies are willing to acquire at scale. Yet the most important part of the second half is only beginning. The liver has been validated. The next questions are whether RNA can reach the lung, muscle, central nervous system, and immune cells. Those questions contain the valuation logic for the next generation of oligonucleotide platform companies. That is why Intelligene’s position is worth examining now. The company is not competing in the most mature and crowded liver route. It is targeting the earlier, more difficult, and more differentiated lung and airway opportunity, with a longer-term ambition to establish broader extrahepatic targeting capabilities. It is linking RNA drug candidates, EV-based delivery, Australian R&D collaboration, a Taiwan-based pilot plant, and Phase 1 development objectives into an international platform pathway that remains uncommon among Taiwan’s biotech companies. In the first half of oligonucleotide therapeutics, the market asked who could design an effective sequence. In the second half, the market will ask who can actually deliver it to the disease site. That is the challenge Intelligene has chosen to take on: building a nanorocket inside the human body. ## Disclaimer This article provides an analysis of industry trends only and does not constitute investment advice. Biotech stocks are highly volatile; please assess the risks yourself. --- ## 這家台灣Biotech正在打造人體內奈米火箭 URL: https://drugnews.com.tw/articles/2026-08-13-intelligene-rna-delivery-nanorocket.html Date: 2026-08-13 Language: zh-Hant Summary: 小核酸藥物的競爭正從 RNA 序列設計走向肝外遞送；智新生技以 EV 與 LNP 平台切入肺部與呼吸道，挑戰小核酸下半場最值錢的問題：能否真正抵達病灶。 【序列是有效載荷，遞送才是火箭；誰能把 RNA 送到肝臟以外，誰才握住下一輪小核酸估值密碼】 有些技術，剛開始像是一場科學實驗；等到大藥廠真金白銀砸下去時，才會被市場看見它其實是一條產業主線。 小核酸藥物正在走到這個位置。 今年 Alnylam Pharmaceuticals 與 AI 生技公司 Inceptive Nucleics 簽下最高 20 億美元合作，表面上是 RNAi 龍頭押注人工智慧；更深一層看，這筆交易其實在宣告：小核酸藥物的競爭，已經不只是在比誰能寫出更好的序列，而是比誰能更快、更穩、更有效地把 RNA 藥物推向臨床。Alnylam 這筆合作包含 3,000 萬美元前期對價、股權投資，以及臨床前、監管與銷售里程碑，目標是用 AI 加速 RNA 藥物候選分子的設計與篩選。 小核酸之所以特別適合 AI，不難理解。傳統小分子藥物要尋找蛋白質上的結合口袋，抗體藥物多半要鎖定細胞外或細胞膜上的蛋白；但小核酸更像是在生命訊息的上游動刀。siRNA（小干擾核酸） 可以讓目標 mRNA 被降解，ASO（反義寡核苷酸） 可以影響 mRNA 降解或剪接，miRNA（微小核糖核酸） 則有機會調節一整串基因網路。 換句話說，小核酸把藥物開發從「尋找結構」推向「設計序列」。 當藥物變得更像序列工程，AI 自然會成為加速器。它可以協助序列設計、脫靶風險預測、化學修飾優化，也可以減少過去大量濕實驗試錯的時間。可是，這只是故事的上半場。 因為 RNA 序列設計得再漂亮，也不會自己抵達病灶。 真正困難的地方在後面：它要穿過人體裡層層防線，不被分解，不被免疫系統過度攔截，最後還要進到正確器官、正確細胞，並且在安全劑量下產生足夠藥效。 這就是小核酸下半場最殘酷、也最值錢的問題：遞送。 *圖1｜小核酸藥物的上半場是序列設計，下半場是遞送能否抵達病灶* ## 01｜小核酸已經從故事，變成賣得動的藥 判斷一個新技術是不是泡沫，不能只看論文，也不能只看想像空間，最後還是要問三件事：有沒有產品？有沒有病人使用？有沒有營收？ 小核酸現在已經開始交成績單。 Alnylam 是最具代表性的公司。它的 RNAi 藥物已經跨過「能不能成藥」的門檻，進入商業放量階段。AMVUTTRA（vutrisiran） 成為小核酸商業化的重要代表，ONPATTRO（patisiran） 則是 RNAi 藥物走向市場的里程碑。Alnylam 自身也把公司定位為 RNAi 治療領域的先行者，強調 RNAi 能透過沉默致病基因，從源頭減少有害蛋白產生。 其他國際藥廠也陸續把小核酸推進更大的疾病市場。 Novartis 的 LEQVIO（inclisiran） 把 siRNA 帶進高膽固醇血症市場；Biogen 的 SPINRAZA（nusinersen） 是 ASO 治療脊髓性肌肉萎縮症的重要代表；Sanofi 與 Alnylam 合作的 QFITLIA（fitusiran） 則切入血友病 A 與 B 的預防性治療。這些產品共同證明，小核酸不再只是罕見疾病裡的漂亮技術，而是可以進入慢病、遺傳疾病、血液疾病與病毒感染等更大戰場。 更具指標性的突破，來自慢性 B 型肝炎。 GSK 與 Ionis 合作開發的 bepirovirsen，在三期 B-Well 研究中，於整體族群達到 19% 功能性治癒率，在低病毒活性族群達到 26%，對照組則為 0%。GSK 也指出，目前全球有超過 2.4 億人受慢性 B 肝影響，而現行標準治療多半需要長期用藥，功能性治癒率仍低。 這個數據很有重量。 B 型肝炎過去很長一段時間，是「壓制病毒」而不是「有限療程後追求功能性治癒」。如果 bepirovirsen 最終能順利走向市場，代表小核酸不只是在罕見病裡證明自己，而是有機會改寫全球大病種的治療邏輯。 這正是小核酸被重新估值的原因。 它不是只解決一種病，而是提供了一套新的疾病干預語言。 *圖2｜小核酸藥物從商業化產品、慢性 B 肝功能性治癒訊號到肝外遞送新戰場* ## 02｜肝臟是第一個軌道，但不是最後一站 小核酸的前半場，肝臟是最重要的舞台。 原因很現實：肝臟本來就容易吸收血液循環中的核酸藥物，再加上 GalNAc 這類肝細胞導向技術成熟，讓許多 siRNA 藥物能比較順利跑出臨床與商業化成果。 但成熟，也意味著擁擠。 當肝臟路線被證明可行，市場自然會往下一個問題看： RNA 能不能送到肝臟之外？ 肺部、呼吸道、中樞神經、肌肉、脂肪、免疫細胞，這些才是下一輪小核酸平台公司真正想打開的新版圖。這些組織背後連著的，是龐大的慢病、感染症、神經疾病、自體免疫與腫瘤市場。 只是，這條路不會好走。 RNA 不是一般小分子。它脆弱、帶電、容易被核酸酶分解，也不容易自己穿過細胞膜。藥物不是設計出來就算完成，還要被保護、被運送、被釋放，最後在病灶細胞裡產生效果。 所以，小核酸的下半場，不能只談「我有序列」。 - 序列是有效載荷。 - 遞送才是火箭。 - 沒有火箭，再好的有效載荷也抵達不了目的地。 *圖3｜肝臟之外的器官遞送難度，正在重新定價 RNA 平台公司* ## 03｜市場正在用「器官難度」重新定價 RNA 平台 小核酸的下半場用「器官難度」與「驗證程度」重新定價平台公司。 如果一家公司只是在肝臟路線上做增量創新，市場會更重視適應症大小、臨床進度與商業化能力；但如果一家公司能打開肝臟以外的新器官，市場看的就不只是單一產品，而是它是否有機會建立新的遞送入口。 這也是為什麼肌肉、中樞神經、腎臟、脂肪與肺部，正在成為小核酸產業的新座標。 其中，近一年幾個重要交易特別值得注意。 Novartis 以約 120 億美元收購 Avidity，看中的不是單一肌肉疾病題材，而是 Avidity 能把寡核苷酸送進肌肉組織的 AOC 平台。Arrowhead 與 Novartis 針對 ARO-SNCA 簽下最高 20 億美元合作，指向的是更困難的中樞神經系統。Lilly 與 Ascidian 的最高 19 億美元合作，則把 RNA 藥物推向腎臟疾病與 RNA exon editing。GSK 與 SiranBio 的合作，則代表脂肪細胞導向 siRNA 也開始被大藥廠提前押注。 這些事件共同說明一件事： 小核酸下半場的競爭，不再只是「誰的 RNA 序列比較好」，而是「誰能率先打開新的器官戰場」。 肌肉已經被 Avidity 這類公司推到高價併購階段；中樞神經雖然難度最高，但神經退化疾病市場想像巨大；腎臟與脂肪正在被 RNA editing、siRNA 與新型靶向技術重新打開；而肺部與呼吸道，則是另一個尚未完全被市場定價、但一旦跑出來就會很有辨識度的方向。 *圖4｜智新生技以肺部與呼吸道為起點，打造 EV 與 LNP RNA 遞送平台* 下面這張競爭格局圖，不只是把公司放在一起比較，而是把「器官」、「成熟度」與「估值」三件事放在同一張地圖裡。 小核酸肝外遞送的格局地圖可以用三個方向來讀： 第一，看器官。 不同器官代表不同難度，也代表不同的市場想像。肝臟已經成熟，肌肉正在被高價驗證，中樞神經、腎臟、脂肪與肺部則是下一批平台公司想打開的新戰場。 第二，看位置。 越往右，代表臨床或 BD 驗證越成熟；越靠左，代表仍在平台概念或早期驗證階段。這能幫助市場區分，哪些公司已經被數據驗證，哪些公司還處在等待關鍵數據的階段。 第三，看圓圈大小。 圓圈越大，代表市場市值、收購估值或交易上限越高。這不是單純比誰現在比較大，而是用資本市場的定價，反映不同平台被驗證後可能打開的價值空間。 | 表1、肝外遞送競爭格局比較表 | 表1、肝外遞送競爭格局比較表 | 表1、肝外遞送競爭格局比較表 | 表1、肝外遞送競爭格局比較表 | 表1、肝外遞送競爭格局比較表 | 表1、肝外遞送競爭格局比較表 | | --- | --- | --- | --- | --- | --- | | 公司 / 事件 | 主要遞送器官 | 成熟度位置 | 市值 / 估值 / 交易上限 | 技術 / 平台 | 觀察重點 | | Alnylam | 肝臟為主，並向多器官延伸 | 後期臨床 / 商業化 | 約 US$38.6B 市值 | RNAi / siRNA | RNAi 商業化龍頭，證明小核酸已能放量銷售。 | | Ionis | 肝臟、CNS、罕病 / 慢病 | 後期臨床 / 商業化 | 約 US$12.2B 市值 | ASO / LICA | bepirovirsen 讓 ASO 從罕病延伸至慢性 B 肝大病種。 | | Avidity | 肌肉 | 後期臨床 / 商業化邊緣 | 約 US$12.0B 收購估值 | AOC（抗體-寡核苷酸偶聯物） | Novartis 高價收購，代表肌肉遞送平台被戰略性驗證。 | | Arrowhead | CNS、肺部、心代謝等 | 早期臨床 / BD | 約 US$10.3B 市值 | TRiM RNAi | 與 Novartis 的 ARO-SNCA 合作顯示肝外 RNAi 平台具授權價值。 | | Dyne | 肌肉 | 早期至中期臨床 | 約 US$3.1B 市值 | FORCE 平台 | 常被視為 Avidity 的重要對標公司。 | | Wave | 脂肪 / 代謝 / CNS | 平台概念至早期臨床 | 約 US$1.1B 市值 | stereopure oligonucleotide | 顯示肝外遞送仍需以臨床終點與商業定位驗證價值。 | | Alnylam × Inceptive | 肝臟為主，偏序列設計端 | 早期臨床 / BD | 交易上限 US$2.0B | AI + RNA 設計 | 說明小核酸上半場正進入 AI 工業化設計。 | | Lilly × Ascidian | 腎臟 | 早期臨床 / BD | 交易上限 US$1.9B | RNA exon editing | 代表 RNA 價值由沉默基因走向功能修復。 | | SiranBio × GSK | 脂肪 | 早期臨床 / BD | 交易上限 US$1.0B | 脂肪細胞導向 siRNA | 是脂肪組織肝外靶向的代表性新交易。 | | 智新生技 | 肺部 / 呼吸道 | 平台概念至早期臨床 | NT$69.6億新台幣 (2026.08.03數據) | EV + LNP RNA 遞送平台 | 核心看點不是現值大小，而是能否以 IND / Phase 1 / CMC 數據證明肺部 RNA遞送可行。 | 進一步把主要公司、標的器官、技術平台與交易估值拆開來看。 它要傳達的重點不是「誰已經贏了」，而是：小核酸產業正在從肝臟單一路線，走向多器官遞送競爭。每一個新器官，都可能成為下一個平台重估的入口。 也因此，真正值得關注的，不只是已經被高價收購或授權的公司，也包括那些還在早期、但正在挑戰新器官遞送的公司。 在這張肝外遞送競爭地圖裡，智新正在挑戰的肺部與呼吸道路線，是否有機會成為下一個被驗證的新器官入口？ ## 04｜智新打造人體內的奈米火箭 放在這個脈絡下，智新生技（7832） 值得被重新理解。 智新真正有意思的地方，不是單純說自己也做 RNA，也不是把故事停在 COVID-19 或流感題材，而是它選擇把戰場先放在肺部與呼吸道。 公司主力管線 IG-001 鎖定 COVID-19 與冠狀病毒感染相關治療，IG-002 則朝廣效型流感治療推進。表面看，這是呼吸道病毒藥物；但深一層看，它其實是在透過這兩個傳染性疾病藥物管線的推進來加速建立一個全面的肝外靶向RNA 治療平台。 這個差別很大。 如果只把 IG-001 看成疫情後的單一產品，市場很容易用短線需求去評價它；但如果從平台角度看，智新想做的是針對呼吸道病毒中相對保守、較不容易快速變異的序列區段，設計 RNA 藥物來抑制病毒複製。 呼吸道病毒從來不是一次性問題。 COVID-19 讓世界看見新興病毒的威脅，流感則每年提醒我們，季節性病毒從未離開。高齡者、慢性病患者、免疫功能較弱族群，長期都需要更好的抗病毒工具。傳統抗病毒藥物常常會面對變異株逃脫與抗藥性問題；若 RNA 藥物能鎖定病毒較保守的序列，就有機會建立更廣效、也更能快速更新的治療策略。 這正是智新的差異化所在。 它不是選擇最成熟、最擁擠的肝臟小核酸路線，而是切入更難、也更有辨識度的肺部與呼吸道，進一步拓展到更具備戰略意義與市場價值的肝外靶向系統中。 難，代表風險。但難，也代表一旦做出來，平台價值會被看見。 ## 05｜EV 外泌體：真正的門檻在製程，不在想像 如果 RNA 是有效載荷，那 EV (外泌體) 就是智新正在打造的奈米火箭。外泌體可以理解為細胞之間傳遞訊息的小型天然囊泡。智新目前揭露，外泌體約 40 至 150 奈米，能攜帶生物分子，也具備跨越細胞膜與進行細胞溝通的特性；智新同時布局 LNP（脂質奈米顆粒） 與 外泌體，並在台北生技園區的GMP先導工廠-製程研發中心推進 RNA 藥物之製程開發、放大驗證（scale-up）與 CMC 能力建立，以支援臨床前及早期臨床需求。 外泌體不是魔法。 它真正困難的地方，不在故事，而在 CMC（製程與品質管制）。 細胞來源怎麼選？培養條件怎麼固定？外泌體怎麼純化？RNA 包覆率怎麼測？每一批粒徑分布是否一致？藥物活性是否穩定？放大製造後，品質是否還能維持？ 這些問題，才是 外泌體藥物 能不能走進臨床的關鍵。 很多平台公司最後不是輸在科學，而是輸在製造。實驗室做得出來，不等於臨床用藥做得出來；小批量做得漂亮，不代表放大後仍然穩定。 所以智新自己建置先導工廠-製程研發中心，這件事不能只當成硬體新聞，公司在外泌體製程上已建立一定基礎與穩定性，且先導工廠將於下半年完工並進駐設備展開試產。對早期 RNA 公司來說，製程能力就是未來談臨床、談授權、談國際合作時的底氣。 真正的「奈米火箭」，不是簡報上畫出來的箭頭，而是能不能一批一批穩定做出來，能不能通過品質驗證，能不能支撐臨床用藥。 智新的 外泌體 平台若要被市場重估，關鍵在於它能不能把外泌體變成可製造、可品管、可放大的藥物載體。 ## 06｜澳洲布局：把台灣平台接上國際臨床航道 智新另一個值得觀察的點，是澳洲布局。 公司已與澳洲 Kirby Institute 簽署合作，獲得BRIDGE計畫贊助最高 500 萬澳幣（約新台幣 1億元）等同研究資源,用於核酸新藥開發 。連結澳洲研發、臨床研究與法規資源，以支持 RNA therapeutics之研究開發；智新目前也在澳洲黃金海岸的健康與知識創新園區(GCHKP)設置海外研發中心。 對早期新藥公司來說，海外據點的重點不只是掛招牌，而是能不能把研發、臨床、法規與國際合作真正串起來。 RNA 藥物若要走向國際，最後一定會面對毒理資料、臨床試驗設計、製程文件、監管溝通與潛在授權夥伴審查。越早接上國際場域，越能降低後面走彎路的機率。 智新的路徑若把台灣與澳洲合起來看，輪廓就更清楚。 台灣這邊建立製程與先導工廠，澳洲那邊接軌研發、臨床與國際合作。前者是根，後者是門；前者讓公司掌握製造與品質，後者讓技術有機會走向更大的市場。 這對一家早期生技公司來說，是很實際的布局。 ## 07｜接下來，市場會看的是升空數據 當然，智新仍然是一家高研發投入、早期階段的新藥公司。它不是成熟藥廠，也不能用營收型公司的本益比去評價。 真正適合它的觀察方式，是看里程碑能不能連續兌現。 - IG-001/IG-002 能不能順利完成臨床前資料，並推進到臨床一期？ - IG-001/IG-002 能不能展現對多種新冠與流感病毒株的廣效潛力？ - EV 類候選藥物能不能真正往臨床一期前進？ - 外泌體平台能不能拿出包覆率、穩定性、遞送效率與批次一致性的數據？ - 南港的GMP先導工廠能不能完成試車、驗證並支援臨床用藥製備？ - 澳洲研發中心能不能持續轉化為臨床、法規與國際合作進展？ 這些不是口號，而是可以被檢查的硬指標。 一家公司要被真正重估，不能只靠題材，也不能只靠一句漂亮標語。它必須在最困難的地方，逐步拿出資料。 對智新來說，「人體內的奈米火箭」不是為了聽起來炫，而是很貼近它正在解的問題：如何把 RNA 從實驗室裡的序列，變成能被送進肺部與呼吸道病灶的藥物。 *圖5｜智新生技的下一階段估值變數，包括臨床前資料、IND、Phase 1、CMC 與國際合作* ## 結語｜小核酸下半場，真正的價值在抵達 小核酸的故事，已經不只是未來式。 Alnylam 用 20 億美元押注 AI，說明 RNA 藥物研發正在加速工業化；Lilly 與 Ascidian 最高 19 億美元 RNA exon editing 腎病合作，說明 RNA 藥物價值正在從沉默基因，走向功能修復與肝外器官戰場；GSK 的 bepirovirsen 讓慢性 B 型肝炎看見功能性治癒希望；Avidity 被 Novartis 用約 120 億美元收購，更直接告訴市場：只要肝外 RNA 遞送能拿出臨床說服力，它就不再只是平台故事，而是大藥廠願意重金買下的戰略資產。 可是，真正精彩的下半場才剛開始。 肝臟已經被證明，接下來大家會追問：RNA 能不能送進肺？能不能送進肌肉？能不能送進中樞神經？能不能送進免疫細胞？ 這些問題背後，藏著下一輪小核酸平台公司的估值密碼。 智新生技的價值，正是在這個時候變得值得討論。 它卡位的不是最成熟、最擁擠的肝臟路線，而是更難、更早期、也更具差異化的肺部與呼吸道與進一步的全方位肝外靶向遞送。它把 RNA 核酸新藥、EV 外泌體遞送、澳洲研發合作、台灣先導工廠與臨床一期推進目標串在一起，形成台灣生技市場少見的國際化平台路徑。 小核酸的上半場，市場看的是誰能寫出有效序列。 小核酸的下半場，市場會看誰能真正抵達病灶。 而智新正在挑戰的，就是打造人體內的奈米火箭。 ## 免責聲明 本文僅為產業趨勢分析，不構成任何投資建議。生技股波動大，請自行評估風險。 --- ## Novo Nordisk’s 6,300-Patient ZEUS Failure: A Textbook Lesson in Translational Risk URL: https://drugnews.com.tw/articles/2026-08-13-ziltivekimab-zeus-mace-phase3-failure-en.html Date: 2026-08-13 Language: en Summary: ZEUS showed the central translational risk in cardiovascular drug development: ziltivekimab suppressed the IL-6 pathway, but the 6,300-patient outcomes trial did not reduce three-point MACE. # Novo Nordisk’s Phase 3 Setback: A 6,300-Patient Textbook Lesson On July 31, 2026, Novo Nordisk disclosed the Phase 3 ZEUS outcome trial for ziltivekimab. The drug had looked almost ideal in Phase 2: inflammation markers fell sharply, dose response was clear, and the same pharmacodynamic signal was reproduced in Japanese participants. The market was waiting for the next step: proof that lowering inflammation could translate into fewer heart attacks or strokes. What ZEUS delivered instead was blunt. More than 6,300 patients. Primary endpoint hazard ratio 0.99. 95% confidence interval 0.88–1.11. In this patient population, at this dose, under this follow-up structure, ziltivekimab did not prove that it could reduce major adverse cardiovascular events. The harder contrast is that pharmacology was still there. Novo Nordisk said free IL-6 and high-sensitivity C-reactive protein, or hsCRP, fell as expected. Yet the three-point MACE endpoint—cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke—did not move. This drug hit the IL-6 pathway. It did not make patients suffer fewer heart attacks or strokes. For biotech investors, that kind of failure is more useful than a simple “the drug did nothing” story. The question is not whether the biomarker moved. The question is why a nearly perfect biomarker story could not carry a hard outcomes trial. *Figure 1 | ZEUS enrolled more than 6,300 patients; inflammation markers fell, but three-point MACE did not improve.* ## 01 | A Phase 2 star, a Phase 3 miss ZEUS was a large, randomized, double-blind, placebo-controlled, event-driven cardiovascular outcomes trial. Participants had established atherosclerotic cardiovascular disease, chronic kidney disease, and hsCRP of at least 2 mg/L. All patients received standard-of-care therapy. The study compared monthly subcutaneous ziltivekimab 15 mg with placebo. The primary endpoint was three-point MACE: cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke. The final hazard ratio was 0.99. That number should not be translated into “precisely zero effect.” The 95% confidence interval of 0.88–1.11 still allows for limited benefit or limited harm. But for development and regulatory purposes, the conclusion is clear: ZEUS did not demonstrate the prespecified primary endpoint benefit. Safety also matters. Novo Nordisk said overall adverse events and serious adverse events were similar between groups, and all-cause mortality did not differ. But serious infections were higher in the ziltivekimab arm. The most important lesson is that the drug appeared to engage the IL-6 pathway, yet patient outcomes did not improve. ## 02 | RESCUE really was strong, but it answered a different question Ziltivekimab’s key Phase 2 study was RESCUE. RESCUE was a randomized, double-blind, placebo-controlled study in 264 patients with stage 3 to 5 chronic kidney disease who were not on dialysis and had hsCRP of at least 2 mg/L. Patients were assigned to placebo or ziltivekimab 7.5 mg, 15 mg, or 30 mg every four weeks for up to 24 weeks. The primary endpoint was change in hsCRP at week 12. The study was designed to ask whether the IL-6 pathway could be suppressed, how dose should be selected, and what short-term safety looked like. The result was strong. At week 12, median hsCRP reductions were 77%, 88%, and 92% for the 7.5 mg, 15 mg, and 30 mg groups, respectively. Placebo fell only 4%. Fibrinogen, serum amyloid A, haptoglobin, secretory phospholipase A2, and lipoprotein(a) also fell in a dose-related way. From a pharmacodynamic perspective, that is an impressive answer: target engagement, a large inflammation-marker response, clear dose response, and 15 mg approaching a pharmacologic plateau. But RESCUE proved that the drug could suppress the IL-6 pathway. It did not prove that patients would have fewer heart attacks, strokes, or cardiovascular deaths. A 264-patient, 12-week biomarker trial was never built to answer that question. ZEUS was. *Figure 2 | RESCUE and ZEUS answered different questions: biomarker response is not outcome proof.* ## 03 | Three layers investors must not confuse ZEUS is easy to misread because investors often compress three different kinds of evidence into one story. The first layer is a risk marker. A higher value identifies patients who are more likely to have future events. It helps describe who is at risk. The second layer is a pharmacodynamic marker. A marker changes after dosing, showing that the drug has produced a biological effect. The reductions in free IL-6 and hsCRP belong here. The third layer is a surrogate endpoint. This is a much higher bar. A surrogate must be supported by enough evidence to reliably predict clinical benefit. A marker can predict risk and respond beautifully to therapy, yet still fail to stand in for heart attack, stroke, or death. ZEUS drew that line clearly. hsCRP helped identify a high-risk inflammatory population. It also showed that the IL-6 pathway was being suppressed. But a large hsCRP reduction did not translate into fewer three-point MACE events. Could imaging have solved the gap? Not automatically. Atherosclerosis is not a straight line where smaller plaque always means fewer events. Plaque vulnerability, local inflammation, thrombosis, and blood flow all matter. A local image also cannot represent the entire arterial system. Darapladib is a warning. Early biomarker and plaque-related signals were not enough. In the large Phase 3 STABILITY trial, the primary cardiovascular endpoint was not significantly improved. Imaging can strengthen an evidence package. Unless it has been validated as a surrogate endpoint, it still cannot replace major cardiovascular event trials. *Figure 3 | Risk markers, pharmacodynamic markers, and surrogate endpoints sit on different evidence layers.* ## 04 | Two hypotheses, neither proven yet There are two intuitive explanations for ZEUS. First, the IL-6–CRP axis may be associated with risk without being the dominant driver of the next event in this ASCVD-plus-CKD population. In that case, the inflammatory marker can fall while the disease process that produces myocardial infarction or stroke continues. Second, the drug may have delivered some anti-inflammatory or antithrombotic benefit, but that benefit may have been offset by infection or by other liabilities not captured by the biomarkers that were measured. Both are plausible. Neither is proven yet. Novo Nordisk has disclosed a higher proportion of serious infections in the ziltivekimab arm, but not the absolute event counts, types, timing, or enough detail to prove that infections offset cardiovascular benefit. Small and short biomarker studies are not designed to reveal low-frequency, long-term risk with confidence. The only firm conclusion today is that ZEUS missed its primary endpoint. The mechanism of failure still needs the full dataset. ## 05 | Cardiovascular history is full of beautiful markers that failed ZEUS is not an isolated case. The first example is the HDL cholesterol hypothesis. In ILLUMINATE, torcetrapib increased HDL-C by 72.1% and lowered LDL-C by 24.9%. Cardiovascular events increased by 25%, and all-cause mortality increased by 58%. The second is triglycerides. In PROMINENT, pemafibrate lowered triglycerides by 26.2%, but the primary endpoint hazard ratio was 1.03. ApoB rose 4.8%. Lowering lipid content and lowering atherogenic particle burden are not the same thing. The third is inflammation and plaque-related biology. Varespladib moved certain biomarkers in the intended direction, but VISTA-16 did not improve the primary endpoint and showed increased myocardial infarction risk. The trial was terminated early. The fourth is homocysteine. NORVIT lowered homocysteine by about 27% with folic acid and B vitamins. The primary endpoint still did not improve. The same lesson keeps coming back: changing a number associated with risk is not the same as changing patient fate. Between the two are causality, disease modification, and net clinical benefit. ZEUS also does not kill the inflammation hypothesis. CANTOS gave positive evidence. In patients with prior myocardial infarction and persistently elevated hsCRP, canakinumab 150 mg significantly reduced the primary cardiovascular endpoint, with an HR of 0.85. Fatal infection increased, and all-cause mortality did not significantly differ. LoDoCo2 also showed that low-dose colchicine reduced the primary composite endpoint in chronic coronary disease: 6.8% versus 9.6%, HR 0.69. Those trials show that certain anti-inflammatory interventions can reduce events in specific patients. Ziltivekimab still had to answer for itself. The more precise conclusion is this: a biomarker must sit on a validated causal pathway, the drug must modify the disease process, and the clinical benefit must survive the subtraction of infection or other harm. ## 06 | Three questions before trusting a beautiful Phase 2 curve The next time a Phase 1 or Phase 2 study lowers a biomarker by 80% or 90%, investors should not automatically price in Phase 3 success. Ask three questions first. One: is this number merely a risk marker, or has it been validated as a surrogate endpoint through hard-outcome trials? Two: is the drug modifying a causal disease unit, or only changing a signal left behind by the disease process? Three: is there an unbooked liability that the biomarker does not show—serious infection, toxicity, metabolic harm, or off-target biology—that only a large Phase 3 trial can settle? If any of those answers is unclear, the Phase 2 curve can justify moving forward. It should not be discounted as if Phase 3 success were already earned. ## Conclusion | Every beautiful early chart is only a premise Ziltivekimab’s Phase 2 data were not fake. They still matter. They proved that the antibody could engage the IL-6 pathway. They did not prove that patients would have fewer myocardial infarctions, strokes, or cardiovascular deaths. In cardiovascular drug development, that final question is the only one that settles the account. Risk signal, target engagement, biomarker movement, and patient outcome are four different evidence layers. A beautiful earlier layer cannot prepay success in the next one. ZEUS has simply confirmed the old rule again: in cardiovascular medicine, the final invoice is paid in patient events. The beautiful numbers before that only earn a drug the right to enter the final. *Figure 4 | What ZEUS tells us, what remains unknown, and what investors should not extrapolate.* ## Disclaimer This article is for biotech industry and capital-market analysis only. It is not medical advice, investment advice, or a recommendation to buy or sell securities. Clinical and investment decisions require full source data and professional judgment. --- ## 諾和諾德大三期慘敗！6,300 人的教科書級教訓 URL: https://drugnews.com.tw/articles/2026-08-13-ziltivekimab-zeus-mace-phase3-failure.html Date: 2026-08-13 Language: zh-Hant Summary: ZEUS 超過 6,300 人卻交出 HR 0.99：ziltivekimab 壓低 free IL-6 與 hsCRP，卻沒有證明能降低三點 MACE。這不是抗發炎假說死亡，而是 biomarker 不能替患者硬終點預支成功。 諾和諾德大三期慘敗！6,300 人的教科書級教訓 2026 年 7 月 31 日，諾和諾德公布 ziltivekimab 的 ZEUS 第三期結果。 這款藥在第二期幾乎交出一張滿分成績單：發炎指標大幅下降、劑量反應清楚，日本受試者也重現相同訊號。市場原本等著它證明「降發炎可以少一次心肌梗塞或中風」，最後等到的卻是—— 超過 6,300 人，主要終點風險比（HR）0.99，95% 信賴區間 0.88–1.11。 ZEUS 給的答案很乾脆：在這群患者、這個劑量與追蹤條件下，ziltivekimab 沒有證明能降低重大心血管事件。 更強烈的反差是：藥效學訊號仍在。諾和諾德表示，游離 IL-6 與高敏感度 C 反應蛋白（hsCRP，常用的發炎指標）都如預期下降；但由心血管死亡、非致命心肌梗塞與非致命中風構成的三點重大心血管事件（MACE），就是沒有跟著下降。 這款藥確實打中 IL-6 通路，卻沒讓患者少一次心肌梗塞或中風。對生技產業來說，這種失敗比「藥完全沒作用」更值得記住。 為什麼第一、二期把發炎指標幾乎打穿，仍然推不出第三期硬終點成功？ZEUS 是最新的一堂課。 *圖1｜ZEUS 超過 6,300 人、HR 0.99，發炎指標下降但三點 MACE 未降低* ## 01｜二期明星，三期翻車 ZEUS 是一項大型、隨機、雙盲、安慰劑對照、事件驅動的心血管結局試驗。 受試者已確診動脈粥樣硬化心血管疾病（ASCVD），同時患有慢性腎臟病（CKD），而且 hsCRP 至少 2 mg/L。所有人都接受標準治療，再比較每月一次皮下注射 ziltivekimab 15 mg 與安慰劑。 主要終點是三點 MACE：心血管死亡、非致命心肌梗塞與非致命中風。 最後 HR 是 0.99。 這個數字不能被翻成「精確證明零效果」。95% 信賴區間 0.88–1.11，仍容納有限的效益或有限的傷害；但在開發與監管上，結論已經很清楚：ZEUS 沒有證明預設的主要終點獲益。 安全性也出現一筆不能忽視的帳。公司表示，兩組整體不良事件與嚴重不良事件比例相近，全因死亡沒有差異，但 ziltivekimab 組的嚴重感染比例較高。 這場失敗最值得研究的地方，正是藥物命中了 IL-6 通路，卻沒有換來患者結局的改善。 *圖2｜RESCUE 與 ZEUS 回答不同問題，二期 biomarker 不能直接外推三期硬終點* ## 02｜第二期有多漂亮？幾乎把發炎指標打穿 Ziltivekimab 的核心第二期研究是 RESCUE。 這項隨機、雙盲、安慰劑對照試驗納入 264 名第三至第五期慢性腎臟病、尚未透析且 hsCRP 至少 2 mg/L 的高風險患者，分成安慰劑與 7.5、15、30 mg 三個劑量組，每四週注射一次，最長治療 24 週。 主要終點是第 12 週 hsCRP 變化，回答的是能否壓住 IL-6 路徑、劑量怎麼選與短期安全性。 結果非常強。 第 12 週，7.5、15、30 mg 組的 hsCRP 中位降幅分別為 77%、88% 與 92%；安慰劑組只有 4%。纖維蛋白原、血清澱粉樣蛋白 A、觸珠蛋白、分泌型磷脂酶 A2 與脂蛋白(a)也出現劑量相關下降。 從藥效學角度看，這是一張漂亮答卷：靶點命中、發炎指標大降、劑量反應清楚，15 mg 也接近藥效平台。 RESCUE 真實證明了藥物能壓低 IL-6 通路；264 人、12 週的設計，原本就無法回答心肌梗塞、中風、死亡與低頻長期感染。ZEUS 接手的正是這張考卷。 *圖3｜風險標誌物、藥效學標誌物與替代終點是三種不同證據層級* ## 03｜三種指標，一定要分開看 這場失敗之所以容易被誤讀，是因為三種性質完全不同的指標，常被市場混成同一件事。 第一種是風險標誌物。 某個數值較高的人，未來比較容易出事；它先描述的是「誰風險較高」。 第二種是藥效學標誌物。 給藥後指標照預期改變，證明藥物產生生物學作用。游離 IL-6 與 hsCRP 下降，就屬於這一層。 第三種才是替代終點。 它的門檻最高：必須經過足夠試驗證明，改善這個指標能可靠預測患者結局也會改善。 風險預測與藥效反應都可能很準，仍不足以取代心肌梗塞、中風與死亡等臨床事件。 ZEUS 正好把這條界線劃得很清楚：hsCRP 幫研究團隊找到帶有全身性發炎的高風險患者，也忠實反映 IL-6 被壓下去；但 hsCRP 大幅下降，並沒有讓三點 MACE 跟著下降。 接著看 hsCRP 與血管斑塊。 hsCRP 能標出高風險患者，也受腎臟病、代謝與感染影響。它可能只是伴隨訊號，或只涵蓋部分致病路徑；目前完整 MACE 分項、事件率與亞組尚未公布，因此這些都只是待驗證的解釋。 影像能不能補上這一格？ 如果抽血指標不夠，直接看血管斑塊影像可以嗎？ 答案仍然是：可以增加證據，但不能自動取代臨床事件。 動脈粥樣硬化不是「斑塊越大，事件一定越多」的直線關係。斑塊脆弱度、局部發炎、血栓與血流都會影響事件；局部影像也無法代表全身動脈。 Darapladib 就是警告：早期曾出現酵素與斑塊訊號，後來大型第三期 STABILITY 的主要心血管終點仍未顯著改善。 影像仍能補強證據；只要尚未被充分驗證為替代終點，它就不能取代實際的重大心血管事件試驗。 ## 04｜ZEUS 為什麼失敗？目前只有兩個假說 回到 ZEUS，目前有兩種最直觀的解釋。 第一種，IL-6—CRP 軸雖然和風險相關，卻可能沒有主導這群 ASCVD＋CKD 患者的下一次事件；發炎指標降下來，造成心肌梗塞或中風的核心病程仍繼續往前。 第二種，藥物可能帶來部分抗發炎或抗血栓效益，但被感染或其他未被目前生物指標捕捉的代價抵銷，最後讓淨效益回到接近零。 兩種說法都合理，但現在都還不能被當成定論。 公司只說嚴重感染比例較高，尚未公布絕對人數、類型與時序，也沒有證明「感染抵銷效益」。小型、短期試驗本來就難以看清低頻風險。 現在能確定的只有主要終點失敗；失敗機制仍待完整資料回答。 ## 05｜漂亮數字翻車，心血管史上早有前例 把 ZEUS 放回心血管藥物開發史，它並不孤單。 第一類是「好膽固醇」假說。 Torcetrapib 在 ILLUMINATE 讓 HDL-C 上升 72.1%、LDL-C 下降 24.9%，心血管事件反而增加 25%，全因死亡增加 58%。 第二類是三酸甘油酯。 PROMINENT 中，pemafibrate 讓三酸甘油酯下降 26.2%，主要終點卻是 HR 1.03；ApoB 反而上升 4.8%。降低脂質含量，和減少顆粒數並不是同一件事。 第三類是發炎與斑塊相關指標。 Varespladib 曾讓相關生物指標朝有利方向變化，最後 VISTA-16 既沒改善主要終點，心肌梗塞風險還增加，試驗提前終止。 第四類是同半胱胺酸。 NORVIT 用葉酸與維生素 B 群把同半胱胺酸降低約 27%，主要終點仍沒有獲益。能預測疾病的指標，不一定位在可改變結局的因果鏈上。 這一系列歷史試驗反覆提醒同一件事：一款藥改變了與風險相關的數字，和它改變了患者命運，中間還隔著因果、疾病修飾與淨臨床效益三道關卡。 可是，ZEUS 也沒有推翻整個發炎假說。 如果因此得到「所有生物指標都不可信」或「抗發炎治心血管全部無效」的結論，又會犯另一種錯。 CANTOS 提供過正面證據。IL-1β 抗體 canakinumab 在既往心肌梗塞、hsCRP 持續升高患者中，150 mg 組顯著降低主要心血管終點，HR 0.85；但致命性感染增加，全因死亡沒有顯著差異。 LoDoCo2 也顯示，慢性冠狀動脈疾病患者使用低劑量 colchicine，主要複合終點為 6.8% 對 9.6%，HR 0.69。 這些研究證明特定抗發炎介入能在特定患者降低事件；ziltivekimab 仍必須靠自己的試驗作答。 更精準的結論是：生物指標必須代表經過驗證的因果通路，藥物也要改變疾病過程，最後再扣掉感染或其他傷害，才可能預測患者獲益。 ## 06｜看到漂亮的第二期，先問這三題 下一次看到第一、二期把某個生物指標壓低 80%、90%，先不要把第三期成功當成預設答案，而要先問三件事： 第一，這個數字是風險標誌物，還是已被硬終點試驗驗證過的替代終點？ 第二，藥物改變的是疾病的致病單元，還是病程留下的伴隨訊號？ 第三，有沒有一筆生物指標沒寫在帳上的負債——感染、毒性、代謝或其他脫靶風險——會等到大型第三期才結算？ 這三題只要有一題答不清楚，漂亮的第二期曲線可以支持「值得進第三期」，卻不應被直接折算成「第三期理應成功」。 *圖4｜ZEUS 目前已知、未知與不能外推的結論邊界* ## 結語｜前面每一張漂亮圖，都只是前提 Ziltivekimab 的第二期資料沒有造假，也沒有失去價值。 它證明這款抗體能命中 IL-6 通路，卻沒有證明患者能少一次心肌梗塞、中風或心血管死亡。 而在心血管領域，這恰恰是最後唯一能結帳的問題。 風險訊號、靶點命中、生物指標與患者結局，是四層不同證據。前一層再漂亮，也不能替下一層預支成功。 ZEUS 只是再一次確認： 心血管新藥最後能結帳的，仍是患者事件；前面的漂亮數字，只負責把它送進決賽。 --- ## Can $1.2 Billion Help AI Double the Pace of Drug R&D? URL: https://drugnews.com.tw/articles/2026-08-12-bio-genesis-mission-ai-drug-rd-en.html Date: 2026-08-12 Language: en Summary: The Bio Genesis Mission is an attempt to connect supercomputing, biomedical data, AI agents, automated laboratories, clinical networks, and regulation into one R&D loop. Its real test is whether governed data and reproducible experiments can reach patients faster. From target discovery and experimental design to clinical development, the United States is trying to reconnect the entire drug R&D process around AI. On July 22, the US National Institutes of Health launched the Bio Genesis Mission, aligning more than $1.2 billion in funding. Its official ambition is larger still: double the pace of biomedical innovation over the next five to ten years, or cut the time from scientific discovery to patients in half within a decade. Bio Genesis is meant to connect national-laboratory supercomputing, federal genomic and clinical data, AI agents, automated laboratories, hospitals, cancer centers, and FDA regulatory capabilities. What the United States wants to rebuild is the operating system for biomedical research. But the $1.2 billion is not a newly approved fund that has already been disbursed in full. It combines funding committed in fiscal 2026 with resources planned for fiscal 2027. The White House's figure of more than $5 billion also covers the broader Genesis Mission across energy, manufacturing, defense, health, and other scientific priorities; it cannot all be attributed to Bio Genesis. The hardest question may not be whether the AI is intelligent enough. It is whether sensitive data can move safely across institutions, whether AI-generated hypotheses can survive experimental replication, and whether the resulting evidence can actually reach patients. *Figure 1 | Bio Genesis aims to connect governed biomedical data, AI supercomputing, hypothesis generation, automated experiments, and clinical evidence in a repeatable R&D loop.* ## 1. DOE Builds the Scientific Foundation, Then NIH Connects Biomedicine Bio Genesis sits inside the broader Genesis Mission, which was launched by executive order in November 2025. The US Department of Energy is responsible for building the American Science and Security Platform, connecting supercomputers, national laboratories, scientific datasets, and research facilities. Health, childhood cancer, chronic disease, and translational medicine were brought into the same framework in July 2026. A more accurate picture is that DOE is building the scientific highway. The National Institutes of Health, Department of Health and Human Services, Food and Drug Administration, and Advanced Research Projects Agency for Health then connect the data they may legally provide, their clinical networks, and their regulatory capabilities. Individual models can be replaced. Once cross-agency infrastructure is built, however, it can reshape the entire research process. ## 2. Connecting Cells All the Way to Patients NIH has placed six missions in the same pipeline. At the front end, AI is expected to predict the behavior of cells, tissues, and complex biological systems. It is also expected to connect molecular, genomic, phenotypic, and real-world data to accelerate target discovery, drug repurposing, and clinical translation. The middle of the pipeline addresses two gaps: how biological discoveries become scalable manufacturing processes, and how abnormal pathogen signals can be detected and attributed earlier. One connects scientific discovery to industrialization; the other connects it to public health and national security. The final destination is the patient. Childhood cancers contain many rare subtypes, making it difficult for any single hospital to accumulate enough data. Chronic disease requires long-term population data, environmental exposure, genetics, and fundamental biology to be analyzed together. All six missions face the same wall: data are scattered across laboratories, hospitals, databases, and regulatory systems; computational findings are slow to return to experiments, and experiments often fail to connect with clinical development. An AI signal is only a starting point. It still has to be tested in cells, animals, prospective clinical studies, and regulatory review. Treating correlation as causation would merely allow researchers to take the wrong path faster. *Figure 2 | The six Bio Genesis missions span complex biological systems, manufacturing scale-up, biological threats, childhood cancer, translational medicine, and chronic disease. AI is one shared layer of the infrastructure.* ## 3. The $1.2 Billion, $5 Billion, and $800 Million Figures Describe Different Things Large policy programs are easily reduced to a headline, so the funding definitions need to be separated. The first figure is NIH's "more than $1.2 billion." That amount contains three components: funding already committed in 2026, funding planned for 2027, and resources previously directed toward related fields that are now being aligned under the same mission framework. New Bio Genesis funding opportunities are still to be announced. The second figure is the White House's "more than $5 billion." This is the federal commitment to the Genesis Mission as a whole. It spans health, energy, infrastructure, manufacturing, defense, and other scientific missions. It does not mean Bio Genesis alone has received $5 billion. The third figure is "more than $800 million" in commitments from industry partners. DOE explicitly says that this support includes compute and credits, access to foundation models, cloud infrastructure, scientific expertise, research collaboration, and direct funding. The $800 million therefore cannot be treated entirely as cash, much less converted directly into revenue for any named company. On the same day, DOE selected 278 projects: 168 led by universities, 87 led by DOE or National Nuclear Security Administration laboratories, 19 led by companies, and four led by nonprofit organizations. Those 278 projects remain under award negotiation, and DOE retains the right to withdraw selections. What can be confirmed today is that policy and resources are beginning to align. Whether research actually becomes twice as fast will have to be judged from completed scientific work. *Figure 3 | The $1.2 billion, $5 billion, and $800 million figures refer to NIH resources, the broader Genesis Mission, and partner commitments, respectively. They cannot be added together or treated as one fund.* ## 4. What Could Change in Drug Development Is the Speed of Feedback Drug development today often gets stuck in a cumbersome cycle. Researchers begin with the literature or an existing dataset, propose a hypothesis, design an experiment, wait for a result, and then decide what to do next. Data formats, access permissions, and experimental equipment differ across institutions. Simply organizing information so that it can be compared may consume a large share of the available time. The Genesis Mission imagines something closer to a closed loop. Data first enter an environment with access controls and provenance tracking. Models propose candidate targets or designs. AI agents coordinate simulations and analysis. Automated laboratories test the hypotheses. The new results return to the models to determine the next cycle. Only after sufficient evidence has accumulated does the work enter clinical and regulatory development. The most practical value of AI is to shorten each cycle of asking a question, running an experiment, receiving a result, and revising the hypothesis, while preserving an auditable record. This also explains why DOE suddenly matters so much. NIH has biomedical research and clinical networks. FDA has regulatory data and review expertise. ARPA-H supports high-risk health research. DOE contributes national laboratories, supercomputing, scientific facilities, and experience with automated experimentation. Bio Genesis is betting that connecting these capabilities creates more leverage than any one agency building AI on its own. Unfortunately, arrows on an organization chart are usually smoother than the real world. Once computing capacity is available, the difficult part begins: data. Medical records, genomic data, and clinical datasets each carry their own consent scope, ethical constraints, de-identification requirements, retention policies, and limits on reuse. Different fields, sampling methods, and patient populations cannot simply be pooled for training. Institutions must also determine who is responsible for dataset bias, provenance, intellectual-property allocation, AI-agent errors, and cross-agency cybersecurity. The executive order already requires attention to privacy, intellectual property, cybersecurity, data standards, and provenance. The NIH director has also stressed that AI will not replace scientific judgment or peer review. A supercomputer may compress a week of computation into an hour. It cannot turn a medical record that lacks consent, provenance, or reproducibility into reliable evidence. ## 5. Technology Companies Are Already Inside, but Procurement Values Are Not Public DOE's collaboration list includes OpenAI, Google, Microsoft, AWS, NVIDIA, AMD, IBM, Intel, Oracle, Palantir, xAI, and other companies. The list shows who has entered the network; it does not disclose procurement amounts. The participants should be viewed across three layers. The first is compute, cloud infrastructure, permissions, cybersecurity, and provenance, which provide a secure environment for sensitive scientific data. The second is whether scientific models and automated experiments can reproduce results across datasets and laboratories. The third is whether pharmaceutical companies, CROs, hospitals, diagnostics providers, and manufacturers can carry those results into clinical development. No matter how polished the first two layers look, the system remains an expensive research demonstration if the third layer never reaches patients. ## 6. Two Capability Coordinates to Watch in Taiwan For Taiwan, a more useful question is this: which local companies genuinely sit at the points where data become computable and research results become translatable? AetherAI Emerging Co. (7803) develops digital-pathology workflows, pathology-image annotation, and AI-assisted analysis, and integrates those systems into hospital pathology departments and pharmaceutical research workflows. Its capabilities map to a critical part of Bio Genesis: converting information that is scattered across glass slides, manual interpretation, and hospital processes into a digital workflow that models can analyze and clinicians can use. Genomics BioSci & Tech Emerging Co. (4195) provides multiple generations of sequencing, whole-genome, RNA, methylation, single-cell, microbiome, proteomics, and bioinformatics services, and has expanded into nucleic-acid and peptide CRDMO services. It sits closer to another part of the chain: generating and organizing multi-omics data, then advancing the work toward biomarkers and early translational manufacturing. The two companies represent different modules. AetherAI is oriented toward clinical imaging and pathology workflows; Genomics BioSci & Tech is oriented toward genomics, multi-omics, and early translational services. Neither is currently listed as a Genesis Mission partner, and there is no primary-source evidence that either has received a Bio Genesis order. That boundary matters. Capability alignment and direct commercial exposure are not the same thing. The larger lesson for Taiwan is that no single company can reproduce a national closed loop of this scale. The gaps to address are cross-hospital data standards, patient consent and governance, trusted computing environments, clinical validation, and collaboration that connects research findings with manufacturing and regulation. *Figure 4 | In Taiwan, digital pathology and clinical-data governance form one capability coordinate, while genomics and multi-omics form another. Capability alignment does not imply a US government contract.* ## 7. The Next Test Is Three Implementation Scorecards Bio Genesis has only just begun. It is too early to declare that it will double the pace of drug development. The first scorecard will be funding: when NIH announces new programs, how much each of the six missions receives, and how much of the total is genuinely incremental. The second will be the platform: whether the American Science and Security Platform can demonstrate a publicly verifiable data-to-model-to-experiment workflow. The most important third scorecard remains clinical and regulatory. How will sensitive data be licensed and traced? How many AI-identified targets enter prospective experiments? What standards will FDA use to evaluate AI-generated evidence? It does not matter how many times faster a slide deck claims the process has become. The final metric is how much less time patients have to wait. The grandeur of the program's name is beside the point. If Bio Genesis is valuable, it will eventually be visible in one concrete outcome: whether a reliable discovery reaches patients sooner. ## Disclaimer This article organizes information about biotechnology, public policy, and industry developments. It does not constitute medical, investment, or product-specific advice. Funding levels, project status, and partner lists may change; readers should consult the latest official disclosures. --- ## 美國砸 12 億美元，想用 AI 把新藥研發快一倍！ URL: https://drugnews.com.tw/articles/2026-08-12-bio-genesis-mission-ai-drug-rd.html Date: 2026-08-12 Language: zh-Hant Summary: 美國以 Bio Genesis Mission 串起超算、生醫資料、AI 代理、自動化實驗、臨床與監管能力；真正的成敗不在模型口號，而在資料治理、實驗重現與患者轉譯能否形成閉環。 從找靶點、做實驗到進臨床，美國想用 AI 把整條新藥研發流程重接一次。 7 月 22 日，美國國家衛生研究院（NIH）啟動「生物創世紀計畫」（Bio Genesis Mission），對齊超過 12 億美元資金。官方目標更大：未來 5 至 10 年讓生醫創新速度加倍，或在 10 年內把科學發現走到患者的時間縮短一半。 Bio Genesis 要接起國家實驗室超算、聯邦基因體與臨床資料、AI 代理、自動化實驗室、醫院、癌症中心與 FDA 監管能力。美國想重做的，是整個生醫研究作業系統。 但 12 億美元不是一筆剛核定、全數撥付的新基金；其中同時包含 2026 會計年度已承諾與 2027 會計年度規劃中的相關資金。白宮所說超過 50 億美元的聯邦承諾，也屬於涵蓋能源、製造、國防與健康任務的整體 Genesis Mission，不能全部算給 Bio Genesis。 接下來最難的未必是 AI 夠不夠聰明，而是敏感資料能不能安全互通、AI 提出的假說能不能被實驗重現，以及研究結果能不能真的走進患者身上。 *圖1｜Bio Genesis Mission 想把生醫資料、AI 超算、假說、自動化實驗與臨床證據接成可反覆驗證的研發閉環。* ## 一、能源部先搭科研底座，NIH 再把生醫接上去 Bio Genesis 的母體，是 2025 年 11 月由行政命令啟動的 Genesis Mission。美國能源部（DOE）負責建置「美國科學與安全平台」，把超算、國家實驗室、科學資料與研究設施接起來；2026 年 7 月，健康、兒童癌症、慢性病與藥物轉譯才被拉進同一框架。 更接近的畫面是：能源部先搭科研高速公路，再由美國國家衛生研究院（NIH）、衛生與公共服務部（HHS）、食品藥物管理局（FDA）、衛生高級研究計畫局（ARPA-H）等機構，把能合法提供的資料、臨床網路與監管能力接上去。單一模型可以更換，跨機構基礎設施一旦建成，影響的是整條研究流程。 ## 二、從細胞一路接到患者 NIH 把六個任務裝進同一條管線。前端要用 AI 預測細胞、組織與複雜生命系統，也要把分子、基因體、表型與真實世界資料接起來，加快找靶點、藥物再定位與臨床轉譯。 中段瞄準兩個落差：生物學發現如何走到可放大製造，以及異常病原訊號如何更早被發現與歸因。這兩件事一邊連著產業化，一邊連著公共衛生與國家安全。 最後落到患者。兒童癌症有大量罕見亞型，單一醫院很難累積足夠資料；慢性病則要把長期人群、環境暴露、遺傳與基礎生物學放在一起看。六個任務都卡在同一堵牆：資料散在實驗室、醫院、資料庫與監管系統，計算結果很難迅速回到實驗，實驗也常接不上臨床。 AI 找到訊號只是起點。它仍要回到細胞、動物、前瞻性臨床與監管判讀；把相關性誤當因果，只會更快走錯路。 *圖2｜Bio Genesis 的六項任務涵蓋複雜生物系統、製造放大、生物威脅、兒童癌症、藥物轉譯與慢性病；AI 只是共同底座的一層。* ## 三、12 億、50 億、8 億，各有自己的口徑 這類大型政策最容易被 headline 帶著走，所以資金口徑一定要拆開。 第一個數字，是 NIH 的「超過 12 億美元」。 這 12 億美元由三部分組成：2026 年已承諾資金、2027 年規劃中資金，以及原本投入相關領域、如今被納入同一任務框架的資源；新的 Bio Genesis 資助機會之後才會公布。 第二個數字，是白宮的「超過 50 億美元」。 這是整個 Genesis Mission 的聯邦承諾，涵蓋健康、能源、基礎設施、製造、國防與其他科學任務，不是 Bio Genesis 單一計畫拿到 50 億美元。 第三個數字，是產業聯盟的「超過 8 億美元」。 DOE 說得很清楚，這些支持包含算力與額度、基礎模型使用權、雲端基礎設施、科學專業、研究合作以及直接資金。也就是說，不能把 8 億美元全部當成現金，更不能直接推算成哪一家公司的訂單。 同一天 DOE 還選出 278 個項目，包括 168 個大學主導、87 個 DOE 或 NNSA 國家實驗室主導、19 個企業主導與 4 個非營利組織主導。 278 個項目目前仍在授獎談判，能源部保留撤回權。現階段能確認的是政策與資源已開始對齊；研究速度是否真的加倍，要等完成的科學成果。 *圖3｜12 億、50 億與 8 億美元分屬 NIH、整體 Genesis Mission 與產業夥伴承諾三種不同口徑，不能相加或視為同一基金。* ## 四、新藥研發可能被改變的，是回饋速度 現在的新藥研發常卡在一個很笨重的循環。 研究者先從文獻或資料中提出假說，設計實驗，等結果，再決定下一輪。不同資料格式不一樣、權限不一樣、實驗設備不一樣，光把資料整理到可以比較，就可能耗掉大量時間。 Genesis Mission 想像的流程更像閉環： 資料先進入有權限與來源追蹤的環境，模型提出候選靶點或設計；AI 代理安排模擬與分析；自動化實驗室驗證；新結果再回到模型，決定下一輪；有足夠證據後，才接進臨床與監管流程。 AI 最實際的價值，是把「提出問題—做實驗—拿結果—再修正」的每一輪縮短，並留下可稽核紀錄。 這也解釋了為什麼 DOE 的角色會突然變得很重。NIH 有生醫研究與臨床網路，FDA 有監管資料與判讀能力，ARPA-H 做高風險健康研究；DOE 則有國家實驗室、超算、科學設施與自動化實驗經驗。Bio Genesis 賭的是把這些不同能力接起來，比任何一個部門自己做 AI 都更有槓桿。 可惜，組織圖上的箭頭通常比現實世界順。 算力到位後，真正難關才出現：資料。 病歷、基因體與臨床資料各有同意範圍、倫理、去識別、保存與再利用限制；欄位、採樣與患者族群不同，也不能直接混在一起訓練。還要回答資料偏差、來源追蹤、智財分配、AI 代理錯誤與跨機構資安由誰負責。 行政命令已把隱私、智財、資安、資料標準與來源追蹤寫進要求；NIH 主任也強調 AI 不會取代科學判斷與同儕審查。 超算可以把一週的計算縮成一小時，卻不能把沒有同意、沒有來源、不能重現的病歷變成可靠證據。 ## 五、科技公司先進場，採購金額還沒公開 DOE 合作名單可見 OpenAI、Google、Microsoft、AWS、NVIDIA、AMD、IBM、Intel、Oracle、Palantir、xAI 等公司。合作名單只顯示誰已進入網路，採購金額仍未公開。 這張名單要分三層看。第一層是算力、雲端、權限、資安與來源追蹤，讓敏感科研資料有地方安全運作；第二層是科學模型與自動化實驗能否跨資料集、跨實驗室重現；第三層才是藥廠、CRO、醫院、檢測與製造端，能不能把結果一路接到臨床。 前兩層做得再漂亮，若第三層走不進患者，仍然只是很昂貴的科研展示。 ## 六、台灣可看兩個能力座標 台灣這一題，比較有用的問法是：本地哪些公司真的站在「資料變得可運算」與「研究結果可轉譯」的節點上？ 雲象科技－創（7803）做的是數位病理工作流程、病理影像標註與 AI 輔助分析，也把系統接進醫院病理科與藥廠研究流程。它對應的是 Bio Genesis 裡很關鍵的一塊：如何把原本散在玻片、人工判讀與醫院流程中的資料，轉成可供模型分析、又能回到臨床使用的數位工作流。 基米－創（4195）提供不同世代定序、全基因體、RNA、甲基化、單細胞、微生物體、蛋白質體與生資分析，也延伸到核酸與胜肽 CRDMO。它比較靠近另一段：多體學資料如何被生成、整理，並往生物標記與早期轉譯製造推進。 兩家公司代表的是不同模組。雲象偏臨床影像與病理工作流，基米偏基因體、多體學及早期轉譯服務。它們都不是美方目前公布的 Genesis Mission 合作夥伴，也沒有一級證據顯示拿到 Bio Genesis 訂單。 這個邊界不能省略。能力相近，和直接受惠，是兩件事。 對台灣更大的提醒反而是：單一公司不可能複製美國這種國家級閉環。台灣要補的是跨院資料標準、患者同意與治理、可信任運算環境、臨床驗證，以及研究結果接到製造和法規的協作方式。 *圖4｜台灣可觀察臨床資料治理與醫療 AI，以及基因體與精準醫療兩個能力座標；能力對位不等於取得美方合約。* ## 七、接下來看三張落地成績單 Bio Genesis 剛啟動，現在就宣告它會讓新藥研發翻倍，太早了。 第一張成績單會來自資金：NIH 新專項何時公布，六個方向各拿多少，新增預算又占多少。第二張來自平台：American Science and Security Platform 能否跑出一條可公開驗證的「資料—模型—實驗」工作流。 最關鍵的第三張，仍在臨床與監管。敏感資料如何授權與追蹤來源？AI 找到的靶點有多少走進前瞻性實驗？FDA 又會用什麼標準判斷 AI 生成的證據？ 投影片上快了幾倍沒有意義。患者少等了多久，才是最後那張成績單。 宏大的計畫名稱不重要。Bio Genesis 若真有價值，最後會反映在一件很具體的事：一項可靠發現，能否更快走到患者面前。 ## 免責聲明 本文為生技醫藥、政策與產業資訊整理，不構成醫療、投資或個別產品建議。政策資金、項目狀態與合作名單仍可能更新，應以官方最新公告為準。 --- ## After 20 Years, Have HDAC Inhibitors Finally Found Their Role? Why ABT-301 Could Become a Next-Generation Combination Backbone URL: https://drugnews.com.tw/articles/2026-08-11-anbogen-abt301-next-generation-hdaci-backbone-en.html Date: 2026-08-11 Language: en Summary: Chidamide has brought HDAC inhibition back into solid-tumor development. ABT-301 must now show that a wider therapeutic window, triplet combinability, and human immune-mechanism data can support a next-generation combination backbone. > From two decades of clinical setbacks and Chidamide’s two solid-tumor validations to Anbogen’s bid to unlock pMMR colorectal cancer through a differentiated therapeutic window, immune remodeling, and the HDAC3–NRF2 axis *Data cutoff: August 5, 2026* > **11.7 months and 44% have brought HDAC inhibitors back to the solid-tumor stage after two decades. Anbogen’s 66-patient colorectal cancer triplet trial could open a path for ABT-301 to become a next-generation combination backbone.** Two numbers — 11.7 months and 44% — have reset the conversation around HDAC inhibitors in solid tumors. In July 2026, HUYABIO reported topline Phase III results for HBI-8000 (Chidamide/tucidinostat) plus nivolumab in advanced melanoma. Across 404 patients in 15 countries, median progression-free survival (PFS) reached 11.7 months versus 7.4 months in the control arm, meeting the primary endpoint. The full hazard ratio, safety profile, overall survival (OS), and subgroup analyses have yet to be reported. CAPability-01 had already provided another important proof point. In MSS/pMMR colorectal cancer, the Chidamide, sintilimab, and bevacizumab triplet delivered a 44% objective response rate (ORR) and a median PFS of 7.3 months. Together, the two studies have put HDAC inhibition back on the map in solid tumors. That does not mean ABT-301 is de-risked. The story now comes down to three questions: Has Chidamide reduced the class risk for HDAC inhibitors? Can ABT-301 translate a differentiated therapeutic window into a next-generation combination backbone in its 66-patient pMMR colorectal cancer triplet study? And, if it can, can that biology travel across cold tumors and partner therapies? ## 1. The HDACi Problem Was Never Lack of Activity — It Was Lack of Durability and Combinability HDAC inhibitors can reactivate gene programs silenced by cancer while influencing the cell cycle, protein stability, angiogenesis, and immune signaling. Since vorinostat became the first FDA-approved HDAC inhibitor in 2006, romidepsin, belinostat, and panobinostat have validated the class in hematologic malignancies. What the field has struggled to do is carry that success into solid tumors. Over the past two decades, the class has repeatedly run into three walls. First, bone-marrow suppression, fatigue, nausea, diarrhea, thrombocytopenia, and cardiac risk can drive dose reductions or discontinuation. Second, with earlier broad-spectrum agents, the active exposure often sits uncomfortably close to the toxicity ceiling, making combinations hard to sustain. Third, early tumor shrinkage has not reliably translated into durable disease control or survival benefit. Two Phase II studies of vorinostat plus pembrolizumab capture the problem. In randomized NSCLC, ORR rose from 28% to 44%, but PFS did not improve (4.3 vs. 4.5 months), and 49% of patients required a vorinostat dose reduction. In a separate metastatic squamous-cell carcinoma study, the combination produced a 26% ORR and a 9.7-month median DOR, yet 66% of patients required dose reduction or interruption because of vorinostat toxicity and 39% permanently discontinued treatment. Solid tumors also bring additional barriers: abnormal vasculature, hypoxia, fibrotic stroma, tumor heterogeneity, T-cell exclusion, and myeloid suppression. Opening chromatin does not automatically get immune cells into the tumor, and a higher response rate does not necessarily mean the response will last. The missing piece, then, is not another mechanism diagram. It is an HDAC inhibitor that can convert multi-layered biology into durable clinical benefit at a tolerable exposure — and remain on board long enough to work with other backbone therapies. *Figure 1 | Three barriers have constrained HDAC inhibitors in solid tumors for two decades: safety and combinability, a narrow therapeutic window, and the failure of early responses to consistently translate into PFS and OS.* ## 2. Chidamide Has Reopened the Class — and Raised the Bar For clarity, Chidamide is the same active molecule as tucidinostat; HUYABIO uses the HBI-8000 code in its ex-Greater China clinical program. The drug primarily inhibits HDAC1, HDAC2, HDAC3, and HDAC10. It was developed by Chipscreen Biosciences, which licensed rights outside Greater China to HUYABIO in 2006. Its importance in solid tumors rests on two complementary proof points: CAPability-01 asked whether HDAC inhibition could help turn an immunologically cold tumor more responsive, while HBI-8000-303 asked whether that concept could hold up in a global confirmatory trial. First proof point: opening an immunotherapy-resistant MSS colorectal cancer population CAPability-01 enrolled patients with MSS/pMMR colorectal cancer after at least two prior lines of therapy. Among 25 patients treated with Chidamide, sintilimab, and bevacizumab, ORR was 44% and median PFS was 7.3 months. In the 23-patient Chidamide plus sintilimab arm, ORR was 13% and median PFS was 1.5 months. The read-through is important: MSS colorectal cancer may require three barriers to be addressed at once. Bevacizumab helps normalize abnormal vasculature, Chidamide remodels the epigenetic and immune microenvironment, and PD-1 blockade releases the immune brake. Tumor RNA analyses showing greater CD8 T-cell infiltration in the triplet arm provide biological support for that treatment logic. The caveats are just as important. The triplet arm included only 25 patients, so the 44% ORR still needs validation in larger, multi-regional studies. Proteinuria, thrombocytopenia, neutropenia, anemia, diarrhea, and treatment-related deaths also make clear that efficacy came with meaningful hematologic and combination-treatment burden. Second proof point: a 4.3-month PFS gain in global Phase III melanoma HBI-8000-303 then took the concept into a global randomized Phase III study of 404 patients with advanced melanoma. Median PFS was 11.7 months with HBI-8000 plus nivolumab versus 7.4 months in the control arm — a 4.3-month absolute gain — and the study met its primary endpoint. That matters because the HDACi-plus-PD-1 thesis has now moved from small proof-of-concept studies into a global confirmatory setting. With approvals in China and Japan and exposure in more than 90,000 patients, Chidamide is one of the most clinically mature benchmarks in the class. Still, topline data are not the final word: the full hazard ratio, safety dataset, OS, and subgroup analyses remain pending. Chidamide’s biggest contribution is class-level de-risking. It shows that HDAC inhibition can add clinical value in solid tumors and, in doing so, raises the bar for the next generation. The benchmark now goes well beyond ORR: sustained dose intensity, dose-reduction and discontinuation rates, DOR, the shape of the PFS tail, and human immune biomarkers all matter. Those unanswered questions are where ABT-301 has room to differentiate. *Figure 2 | Chidamide has delivered two solid-tumor proof points: CAPability-01 achieved a 44% ORR in MSS metastatic colorectal cancer, while HBI-8000 plus nivolumab met the primary PFS endpoint in global Phase III melanoma. Hematologic toxicity and combination burden remain key points of comparison.* ## 3. ABT-301’s Differentiation: The Thesis Is the Therapeutic Window, Not ‘16× Potency’ De-risking the class does not de-risk the asset. For ABT-301 to emerge as a best-in-class contender, its in vitro potency has to translate into three clinically meaningful advantages: a wider therapeutic window, a more combination-friendly immune profile, and an additional route to address tumor resistance. 1. Therapeutic window: potency only matters if it translates ABT-301 is an oral HDAC1/2/3 inhibitor with a class I focus, particularly on HDAC3. In Anbogen’s assay, its HDAC3 IC50 was 6.6 nM — about one-sixteenth that of Chidamide. That is not, by itself, evidence of a better drug. The clinically relevant question is whether ABT-301 can achieve sufficient target engagement at a lower, tolerable exposure and still preserve dose intensity once PD-1 and anti-VEGF therapy are added. 2. Immune remodeling: suppress the suppressors, preserve the effectors Preclinical data suggest that ABT-301 can reduce monocytic myeloid-derived suppressor cells while increasing CD8 T cells and granzyme B. In the CT26 cold-tumor model, ABT-301 plus avelumab drove complete tumor regression in 7 of 8 animals; after rechallenge, 6 of 7 remained tumor-free, supporting a T-cell-memory hypothesis. For a combination backbone, that is the more important question: can an early response become durable control? 3. Resistance biology: extend the role beyond the tumor microenvironment ABT-301 also targets a second biological axis: HDAC3–NRF2–ferroptosis. In KRAS-mutant pancreatic cancer with elevated NRF2 activity, NRF2 activates defenses such as NQO1, SLC7A11, and GPX4 that help tumor cells withstand oxidative stress from gemcitabine. Anbogen’s preclinical data suggest ABT-301 may suppress downstream NRF2 signaling, increase iron-dependent lipid peroxidation, and enhance tumor control with gemcitabine. Taken together, these three mechanisms create a differentiated development thesis — but they remain hypotheses until proven in patients. A best-in-class claim will not be earned by a 6.6 nM IC50, a single animal model, or a compelling mechanism graphic. It will be earned if human combination studies show tolerability, reproducible efficacy, and durable benefit. *Figure 3 | ABT-301’s differentiation thesis links HDAC3 targeting, immune-microenvironment remodeling, and resistance biology to human exposure and, ultimately, combination-treatment validation.* > **ABT-301’s differentiation is not a single 6.6 nM figure. It is whether the therapeutic window, immune remodeling, and resistance-reversal mechanisms can translate into clinically verifiable benefit in humans.** ## 4. Phase I: 23 Patients Put the Mechanism Into Humans — Enough to Advance, Not Enough to Call the Outcome ABT-301’s first-in-human Phase I study enrolled 23 patients with advanced solid tumors and tested once-daily monotherapy doses from 50 to 250 mg. The maximum tolerated dose (MTD) was 150 mg. Most adverse events were Grade 1–2; Grade 3 abdominal pain and Grade 4 thrombocytopenia emerged at 200 mg, while Grade 3 vomiting and Grade 4 thrombocytopenia occurred at 250 mg. In other words, platelet and gastrointestinal toxicity became more limiting at the top end of the dose range. The study did not highlight the cardiac abnormalities, neutropenia, or lymphopenia often associated with the class, although the sample remains small. On efficacy, 1 of 23 patients achieved a partial response and 8 had stable disease, for an overall disease control rate of approximately 39%. The partial response occurred in sweat-gland ductal carcinoma and lasted 24 weeks. Five other patients maintained disease control for at least 16 weeks, including hepatocellular carcinoma for 53 weeks and thymic carcinoma for 49 weeks. At 150 mg, 1 of 9 patients achieved a partial response and 6 had stable disease, for a DCR of approximately 78%. Increased histone H3 acetylation was observed across dose cohorts. What does Phase I establish? Three things. First, ABT-301 achieved measurable target engagement in humans. Second, it produced early — and in some cases durable — antitumor signals across multiple tumor types. Third, it identified a dose range that can now be optimized in combination studies for finding recommended phase 2 dose (RP2D). What it does not establish is comparative efficacy: a 23-patient, single-arm, mixed-tumor Phase I cannot prove superiority over another HDAC inhibitor. Its value is that it earns ABT-301 the right to move into the next test. *Figure 4 | Phase I established the human foundation for ABT-301 — safety, preliminary activity, target engagement, and a workable dose range — but not yet comparative proof.* ## 5. The Triplet Is the Inflection Point: Four Gates Matter More Than One ORR Number Anbogen is working with BeOne Medicine on a global Phase Ib/II triplet study of ABT-301, tislelizumab, and bevacizumab. The study plans to enroll approximately 66 patients in Taiwan and Australia with pMMR/non-MSI-H advanced colorectal cancer after at least two prior lines of therapy. It is testing once-daily and every-12-hour ABT-301 schedules, with the near-term goal of defining a recommended Phase II dose that can be maintained over time before moving into efficacy and durability readouts. Each drug has a distinct job. Bevacizumab helps normalize tumor vasculature and improve immune-cell access. ABT-301 is intended to remodel the epigenetic and immune microenvironment, reduce suppressive myeloid cells, and increase T-cell activity. Tislelizumab releases the PD-1 brake. The design builds directly on the biological logic of CAPability-01; the key question, however, is not whether another triplet can be assembled, but whether ABT-301 can make the combination more sustainable. Gate 1 is combinability: Grade ≥3 thrombocytopenia and GI toxicity, dose reductions, interruptions, permanent discontinuations, and the relative dose intensity patients can actually maintain. Gate 2 is reproducible efficacy: confirmed partial responses, depth of tumor shrinkage, DCR, consistency across sites, and activity in heavily pretreated patients and those with liver metastases. CAPability-01’s 44% ORR is a useful external reference, but differences in patient mix, line of therapy, and baseline risk make a simple cross-trial comparison inappropriate. Gate 3 is durability: DOR, the proportion of responses still ongoing at 6 and 12 months, and whether the PFS curve develops a meaningful tail. Gate 4 is human mechanistic validation. If biopsies or peripheral blood show lower suppressive myeloid cells, higher CD8 T-cell infiltration and granzyme B, expansion of memory T-cell populations, and a relationship between histone acetylation and clinical benefit, ABT-301 could move from being the ‘third drug’ in a triplet to a reusable combination backbone across tumor types and partner therapies. A headline ORR will not be enough to validate the backbone thesis. The real bar is four things happening together: patients can stay on treatment, efficacy is reproducible, responses last, and the mechanism can be measured in humans. The study is ongoing, and actual enrollment and initial clinical data have not yet been disclosed. *Figure 5 | The pMMR colorectal cancer triplet creates four sequential value inflection points: combinability, reproducible efficacy, durability, and human mechanistic validation — the evidence needed to determine whether ABT-301 can become a true combination backbone.* > **The triplet does not pass on response rate alone. Combinability, reproducible efficacy, durability, and human mechanistic validation must all be demonstrated.** ## 6. ABT-301’s Market Expansion Path: Set the Valuation Floor, Add Cross-Tumor Optionality, Earn the Platform Premium Anbogen chose MSS/pMMR colorectal cancer as ABT-301’s first human triplet battleground because the clinical hurdle is high — and therefore highly informative. Success here would provide one of the clearest tests of whether ABT-301 can function as a backbone across partner therapies and then expand into other hard-to-treat tumors. Layer 1: MSS colorectal cancer sets the valuation floor Roughly 95% of metastatic colorectal cancers are pMMR/non-MSI-H and generally respond poorly to PD-1 monotherapy. If the ABT-301 triplet establishes credible human proof of concept, it could open a late-line colorectal cancer opportunity estimated at roughly US$9.0–9.5 billion and show that PD-1, PD-L1, and PD-1/VEGF bispecific therapies can be extended into an immunologically resistant population. This is the closest-in and most tangible layer of ABT-301’s valuation today. Layer 2: pancreatic and brain tumors add cross-tumor optionality Pancreatic cancer sits at the intersection of KRAS-driven biology, dense stroma, MDSC-mediated suppression, chemotherapy resistance, and strong HDAC3 dependence. Revolution Medicines’ daraxonrasib recently delivered significant OS and PFS improvements versus standard chemotherapy in the Phase III RASolute 302 study, potentially reshaping future treatment guidelines. If ABT-301 can weaken tumor defenses through the HDAC3–NRF2–ferroptosis axis, it may complement pan-RAS inhibitors and become part of a next-generation RAS-targeted combination strategy in a pancreatic cancer market projected to reach approximately US$12.05 billion. Brain tumors represent a higher-risk, higher-upside option. Anbogen’s preclinical data show an approximately 40% brain-to-blood concentration ratio for ABT-301 — about four times the BBB-penetration benchmark cited for traditional HDAC inhibitors — together with inhibitory activity across multiple glioblastoma models. The associated brain-tumor treatment market is estimated at roughly US$8 billion. But total drug concentration in an animal brain is not the same as free-drug exposure or target engagement in the human CNS. Until human CNS pharmacokinetics and intracranial biomarkers are available, this opportunity should remain outside the core valuation. Layer 3: repeatable success across cold tumors and modalities creates the platform premium Platform value is not created by listing more indications; it is created by repeatability. If ABT-301 can generate benefit with different PD-1, PD-L1, PD-1/VEGF bispecific antibodies, T-cell engagers, cancer vaccines, or CAR-T therapies — and do so across multiple cold tumors — the asset could evolve from a single drug into a true Combination Backbone. The broader treatment need in ICI-refractory and cold tumors is estimated at roughly US$96 billion, providing the strategic backdrop for that platform opportunity. Those market figures are context, not ABT-301’s directly addressable market and not a revenue forecast. The asset still has to de-risk each clinical milestone with data before the next layer of value can be underwritten. The value-unlocking sequence is therefore straightforward: establish human proof of concept in MSS colorectal cancer; show that the biology travels into pancreatic cancer or another tumor type; then demonstrate repeatability with a second partner therapy. Each step increases the potential upside, but unproven optionality should be discounted accordingly. *Figure 6 | ABT-301’s value-unlocking path: MSS colorectal cancer sets the valuation floor; pancreatic and brain tumors add cross-tumor optionality; and repeatable success with multiple partner therapies is what ultimately supports a platform premium. Market figures are shown for demand context only.* ## Conclusion: Chidamide Has De-Risked the Class. ABT-301 Still Has to De-Risk the Asset Twenty years ago, HDAC inhibitors competed on potency and maximum tolerated dose. Today, the winning profile looks very different: a workable therapeutic window, long-term usability, preservation of immune effector cells, and the ability to pair cleanly with the right partner therapy. Chidamide has already done something important for the field. CAPability-01 established the triplet concept in pMMR colorectal cancer, while the global Phase III HBI-8000 program moved HDAC inhibition plus PD-1 blockade into a confirmatory setting. That de-risks the class — but it does not de-risk ABT-301. For ABT-301, three questions now matter most. Can it maintain dose intensity in the triplet? Can it generate reproducible, durable responses in pMMR colorectal cancer? And can human biomarkers line up with clinical benefit in a way that validates the mechanism? If the answer to all three is yes, ABT-301 could move from a promising sensitizer to a true combination backbone. A second tumor type and a second partner therapy would then be the evidence needed to turn the asset story into a platform story. > **Chidamide has validated the class; ABT-301 must prove asset-level differentiation with human triplet data, then validate the platform with a second tumor type and a second partner therapy.** ### Disclaimer This article analyzes drug mechanisms, clinical development, and industry trends. The safety and efficacy of the ABT-301 triplet in humans remain under evaluation in the ongoing Phase I/II study. Preclinical findings should not be interpreted as evidence of human efficacy, and market-size estimates are not equivalent to directly addressable market or revenue forecasts. Nothing in this article constitutes investment advice. This article does not constitute medical advice. --- ## 20年後，HDAC抑制劑終於找到真正的位置？ABT-301能否成為下一代聯合治療骨幹 URL: https://drugnews.com.tw/articles/2026-08-11-anbogen-abt301-next-generation-hdaci-backbone.html Date: 2026-08-11 Language: zh-Hant Summary: Chidamide以pMMR大腸癌44% ORR與黑色素瘤第三期PFS達標，重新降低HDAC抑制劑的類別風險；安邦ABT-301接下來必須用治療窗口、三聯可組合性與人體免疫機轉，證明資產差異。 > 從二十年臨床卡關 、Chidamide（西達本胺）兩項實體瘤驗證，到安邦如何以治療窗口、免疫重塑與HDAC3–NRF2路徑，挑戰pMMR大腸癌 *資料基準日｜2026年8月5日* > **11.7個月、44%，讓沉寂20年的HDAC抑制劑重回到實體瘤舞台。安邦ABT-301的66人的大腸癌三聯試驗打開下一代HDAC抑制劑成為聯合治療骨幹的可能** 11.7個月與44%，是HDAC抑制劑二十年實體瘤開發的兩個新座標。2026年7月，HUYABIO公布HBI-8000（即Chidamide／西達本胺）加Nivolumab治療晚期黑色素瘤的全球第三期初步結果：404位病人、15個國家，聯合組中位無惡化存活期為11.7個月，對照組7.4個月，主要終點達標；完整風險比、安全性、整體存活期與次族群資料仍待發表。 更早之前，CAPability-01以Chidamide、Sintilimab與Bevacizumab治療MSS／pMMR大腸直腸癌，三聯組客觀緩解率達44%，中位無惡化存活期7.3個月。這兩項研究把HDAC抑制劑重新帶回實體瘤開發的核心舞台。 但「類別重新成立」不等於「ABT-301已經成功」。真正需要拆開的是三層問題：Chidamide是否已替HDAC抑制劑降低類別風險；ABT-301能否用更好的治療窗口，用66位病人的pMMR晚期大腸癌三聯試驗證明成為下一代聯合治療骨幹，打開冷腫瘤與多種藥物聯合的天花板。 ## 一、HDAC抑制劑真正卡住的，不是藥效，而是可持續性 HDAC抑制劑能重新打開被癌細胞關閉的基因程式，並同時影響細胞週期、蛋白質穩定、血管生成與免疫反應。自2006年vorinostat獲美國FDA核准後，romidepsin、belinostat與panobinostat也陸續在血液腫瘤證明此類標靶可成藥；問題在於，血液腫瘤的成功始終沒有順利複製到實體瘤。 二十年來主要撞上三道牆。第一，骨髓抑制、疲倦、噁心、腹瀉、血小板下降與心臟風險，常迫使病人降劑量或停藥；第二，早期廣效型藥物的有效劑量過度靠近毒性劑量，一旦加入另一顆主藥，整套方案便難以長期維持；第三，早期縮瘤未必能轉化為更長的疾病控制與存活。 Vorinostat與pembrolizumab的兩項臨床二期研究都指出：HDACi雖可能提高免疫反應，卻未必能轉化為持久獲益。隨機非小細胞肺癌試驗中，聯合組ORR由28%升至44%，但PFS未改善（4.3個月對4.5個月），且49%病人需降低vorinostat劑量。 另一項轉移性鱗狀細胞癌試驗雖取得26%整體ORR與9.7個月中位DOR，但66%因vorinostat毒性減量或中斷、39%永久停藥。 實體瘤還多了異常血管、缺氧、纖維化基質、腫瘤異質性、T細胞被排斥在腫瘤外，以及髓系抑制細胞等障礙。打開染色質，不代表免疫細胞就能進入腫瘤；提高反應率，也不代表反應能持久。 因此，HDAC抑制劑真正缺的從來不是更多機制，而是一款能在可承受暴露下，把多重藥理作用轉成持續臨床利益、並可與其他主藥長期聯用的產品。 *圖1｜HDAC抑制劑二十年真正卡住的三道牆：安全性與可組合性、治療窗口太窄，以及早期反應無法穩定轉化為無惡化存活與整體存活。* ## 二、Chidamide完成類別去風險，但也留下升級空間 Chidamide（西達本胺）其國際非專利名為Tucidinostat，HUYABIO在海外臨床使用HBI-8000代號。四個名稱指向同一活性成分，不應被理解為不同藥物。Chidamide主要抑制HDAC1、2、3與10，由微芯生物開發，大中華區以外權利於2006年授權HUYABIO。 它改變實體瘤敘事，來自兩項互補證據：CAPability-01回答「能否打開冷腫瘤」，HBI-8000-303則回答「能否走到全球確認性試驗」。 第一次驗證：在MSS大腸癌打開免疫治療缺口 CAPability-01納入接受過至少兩線治療的MSS／pMMR（微衛星穩定／錯配修復功能正常）大腸直腸癌病人。Chidamide、Sintilimab與Bevacizumab三聯組25位病人的客觀緩解率為44%、中位無惡化存活期7.3個月；Chidamide加Sintilimab雙聯組23位病人則為13%與1.5個月。 這項結果支持一個關鍵判斷：MSS大腸癌可能必須同時處理三道障礙——Bevacizumab改善異常血管，Chidamide重整表觀遺傳與免疫微環境，PD-1抗體再解除免疫煞車。腫瘤RNA顯示三聯組有較多CD8 T細胞浸潤，也讓這套分工不只停留在臨床數字。 限制同樣清楚：三聯組只有25人，反應率仍待更大、跨地區研究重現；蛋白尿、血小板與中性球下降、貧血、腹瀉及治療相關死亡，也提醒市場，療效改善仍伴隨骨髓與聯用負擔。 第二次驗證：全球黑色素瘤第三期延長4.3個月無惡化存活 HBI-8000-303把訊號推進到404位晚期黑色素瘤病人的全球隨機第三期。HBI-8000加Nivolumab的中位無惡化存活期為11.7個月，對照組7.4個月，絕對延長4.3個月，主要終點達標。 這代表HDAC抑制劑加PD-1抗體的增益，已從小型概念驗證走到全球確認性研究。加上中國、日本核准與逾9萬人使用經驗，Chidamide已成為此類藥物在實體瘤開發中最成熟的臨床標杆之一；但公司尚未公布完整風險比、安全性、整體存活期與次族群結果，現階段仍不宜把初步結果解讀成終局答案。 Chidamide完成的是「類別去風險」：它證明HDAC抑制劑可以在實體瘤產生臨床增益，也把下一代競爭標準拉高。後續不能只比較客觀緩解率，而要比較可維持劑量、降劑量與停藥率、緩解持續時間、無惡化存活曲線是否形成長尾，以及人體免疫生物標記。這些尚未完全解答的問題，才是ABT-301的切入點。 *圖2｜Chidamide（西達本胺）完成兩項實體瘤驗證：CAPability-01在MSS轉移性大腸直腸癌取得44%客觀緩解率；HBI-8000加nivolumab則在黑色素瘤全球第三期達成主要無惡化存活終點，但骨髓毒性與聯用負擔仍需持續比較。* ## 三、ABT-301的差異化：不是「強16倍」，而是臨床窗口 類別已被去風險，不代表單一資產也已被去風險。ABT-301要成立，必須把體外抑制強度轉成三項臨床可驗證的差異：較寬的治療窗口、較好的免疫聯用條件，以及對腫瘤抗藥機制的額外作用。 1. 治療窗口：低抑制濃度只是起點 ABT-301是一款以第一型HDAC、尤其HDAC3為主的口服HDAC1/2/3抑制劑，對HDAC3的半數抑制濃度為6.6 nM，約為Chidamide的1／16。這個數字本身不能證明藥更好；真正價值在於，能否以較低且可耐受的暴露達到足夠標靶作用，並在加入PD-1與抗VEGF藥物後，仍維持劑量強度。 2. 免疫重塑：降低抑制，同時保留攻擊部隊 ABT-301臨床前資料顯示，單核細胞樣髓源性抑制細胞下降，CD8 T細胞與顆粒酶B增加。CT26冷腫瘤模型中，ABT-301與avelumab聯用使8隻動物中7隻完全退縮；重新植入腫瘤後，7隻中6隻仍無腫瘤，支持形成T細胞記憶的假說。這比單次縮瘤更重要，因為聯合治療骨幹最終必須把早期反應轉成持久控制。 3. 抗藥逆轉：從腫瘤外部走進癌細胞內部 另一條路徑是HDAC3–NRF2–鐵死亡。在KRAS突變、NRF2活性較高的胰臟癌中，NRF2會啟動NQO1、SLC7A11與GPX4等防禦，抵抗gemcitabine造成的氧化壓力。安邦資料顯示，ABT-301可能降低NRF2下游訊號，增加鐵與脂質過氧化，進而提升gemcitabine的腫瘤控制。 這三條路徑構成一張完整的差異化地圖，卻仍只是可驗證假說。ABT-301能否成為同類最佳，不能由6.6 nM、單一動物模型或機制圖決定，而要由人體聯合試驗同時證明耐受性、可重現療效與反應持久性。 *圖3｜ABT-301的差異化證據鏈：由HDAC3標靶、免疫微環境與抗藥路徑，逐步連到人體暴露與聯合治療驗證。* > **ABT-301的差異不在單一6.6 nM，而在能否把治療窗口、免疫重塑與抗藥逆轉，轉成人體可驗證的臨床利益。** ## 四、23位一期病人：人體證據足以推進，但不足以定調 ABT-301首度人體第一期共納入23位晚期實體腫瘤病人，測試50至250 mg每日一次的單藥劑量。最高耐受劑量（MTD, Maximum Tolerated Dose）定為150 mg，多數不良事件為1至2級；200 mg出現3級腹痛與4級血小板減少，250 mg出現3級嘔吐與4級血小板減少，顯示高劑量區間的主要限制仍是血小板與腸胃毒性。試驗未突出顯示同類藥物常見的明顯心臟異常、嗜中性白血球或淋巴球下降，但樣本數仍小。 療效方面，23位病人中有1位部分緩解、8位疾病穩定，整體疾病控制率約39%。部分緩解出現在汗腺導管癌病人，維持24週；另有5位病人的疾病控制至少16週，最長為肝細胞癌53週與胸腺癌49週。在150 mg治療的9位病人中，1位部分緩解、6位疾病穩定，疾病控制率約78%；各劑量組也都觀察到組蛋白H3乙醯化增加。 這批資料支持三項結論：第一，藥物已在人體達到可量測的標靶作用；第二，跨癌種出現初步且部分持久的抗腫瘤訊號；第三，已找到可供聯合試驗探索二期推薦劑量（RP2D, Recommended Phase 2 Dose）的劑量區間。但單臂、23人、混合癌種的第一期試驗，不能證明療效，也不能直接與其他HDAC抑制劑比較。它的真正價值，是讓ABT-301有資格進入下一場考試。 *圖4｜ABT-301第一期人體訊號：已建立安全性、初步活性與治療窗口的推進依據，但尚未形成比較性證據。* ## 五、三聯試驗是決勝點：不只看反應率，還要過四道關 安邦正與BeOne Medicine合作，推進ABT-301、tislelizumab與bevacizumab的全球臨床Ib／II期三聯試驗，預計在台灣與澳洲收案約66位接受過至少兩線治療的pMMR／non-MSI-H晚期大腸直腸癌病人。試驗比較ABT-301每日一次與每12小時一次的給藥方式，先找出可長期維持的建議第二期劑量，再評估初步療效與持久性。 三顆藥的分工清楚：bevacizumab改善異常血管，讓免疫細胞更容易進入腫瘤；ABT-301重塑表觀遺傳與免疫微環境，降低抑制性髓系細胞並提高T細胞活性；tislelizumab解除PD-1免疫煞車。這套設計直接承接CAPability-01，但更重要的不是複製三聯形式，而是證明ABT-301能否提供更好的可持續性。 第一關是可組合性：3級以上血小板與腸胃毒性、降劑量、中斷、永久停藥率，以及實際維持的劑量強度。第二關是療效能否跨中心重現：確認後的部分緩解、縮瘤深度、疾病控制率，並特別觀察肝轉移與多線治療病人。CAPability-01的44%可作外部標杆，但兩項試驗的族群、線別與基線風險不同，不能直接橫向比高低。 第三關是反應能否持久：緩解持續時間、6個月與12個月持續反應率，以及無惡化存活曲線是否形成長尾。第四關是人體是否複製臨床前機制：若腫瘤切片或周邊血液同時顯示抑制性髓系細胞下降、CD8 T細胞浸潤與顆粒酶B上升、記憶T細胞增加，且組蛋白乙醯化與療效相關，ABT-301才可能由三聯中的「第三顆藥」升級為可跨癌種、跨搭配藥物使用的聯合治療骨幹。 因此，一個漂亮的反應率並不足以讓骨幹論成立。真正的及格線是四項條件同時出現：病人能持續用藥、療效可重現、反應夠持久、機制能在人體量化。目前試驗正在進行，實際收案與初步數據仍待公布。 *圖5｜pMMR大腸直腸癌三聯試驗的四階段評估：先證明可組合性，再看療效、持久性與人體機制，決定ABT-301能否升級為聯合治療骨幹。* > **三聯試驗的及格線不是單一反應率，而是可組合性、可重現療效、持久性與人體機制同時成立。** ## 六、ABT-301市場滲透路徑：適應症底價、跨癌種擴張與平台溢價 安邦把MSS／pMMR大腸癌放在第一個人體三聯戰場，臨床門檻高，卻最能驗證ABT-301能否成為跨主藥使用的骨幹，進一步透過多種難治型腫瘤的跨癌腫擴張，逐步形成。 第一層：MSS大腸直腸癌建立資產底價 約95%的後期轉移性大腸直腸癌屬於pMMR／non-MSI-H，通常對PD-1單藥反應有限。若ABT-301三聯療法建立可信的人體PoC，便可先切入約90至95億美元級的晚期大腸直腸癌市場，並證明它能把PD-1、PD-L1及PD-1／VEGF雙抗帶進原本難以治療的免疫黑區。這將構成ABT-301最接近現階段的資產底價。 第二層：胰臟癌與腦腫瘤提供跨癌種擴張 胰臟癌同時具有KRAS驅動、緻密基質、MDSC壓制與高度化療抗藥，並高度依賴HDAC3。近期Revolution Medicines的Daraxonrasib已在臨床三期RASolute 302試驗中，對比標準化療組，顯著改善整體存活與無惡化存活，可能改變未來胰臟癌的治療指引。ABT-301若能透過HDAC3–NRF2–ferroptosis路徑削弱腫瘤防禦，便可能與泛RAS抑制劑形成互補，成為下一代泛RAS標靶治療的聯合骨幹，滲透未來120.5億美元級的全球胰臟癌治療市場。 腦腫瘤則屬高風險選擇權。公司臨床前資料顯示，ABT-301在動物的腦部／血液濃度比約40%，相對傳統HDACi的血腦障壁穿透力提升400%，並在多個膠質母細胞瘤模型觀察到抑制活性，背後對應約80億美元級的腦腫瘤治療機會。但動物腦部總濃度不等於人體腦內游離藥物暴露，也不能證明腦內標靶作用。在人體中樞藥物動力學與腦內生物標記出現前，不宜先計入主要估值。 第三層：冷腫瘤與其他模態藥物聯用成功，形成平台溢價 真正的平台價值，來自ABT-301能否與不同PD-1、PD-L1、PD-1／VEGF雙抗、T-cell engager、癌症疫苗或CAR-T聯用，並在多種冷腫瘤中重複產生效果。全球ICI抗藥與冷腫瘤治療需求約達960億美元級。當ABT-301能跨癌種、跨partner與跨modality驗證可複製性，它才會從單一資產升級為Combination Backbone，進一步產生平台溢價。 *圖中所預估適應症的全球治療市場的缺口屬於背景需求，ABT-301仍須提出更多具備降低風險的臨床數據，逐步撬開每個去風險的臨床節點，形成多層次堆疊的平台溢價。* 換句話說，ABT-301的價值解鎖順序應是：先以MSS大腸癌證明人體概念，再由胰臟癌或其他癌種驗證機制外推，最後用第二個搭配藥物證明可複製性。越往後，潛在價值越大，但折現率也越高。 *圖6｜ABT-301的價值解鎖路徑：MSS大腸癌建立資產底價，胰臟癌與腦腫瘤提供跨癌種選擇權，多種搭配藥物重現後才形成平台溢價；圖中市場數字僅作需求估算。* ## 結語：Chidamide證明類別，ABT-301必須證明資產 二十年前，HDAC抑制劑競爭的是抑制強度與最大耐受劑量；今天真正決定勝負的，是治療窗口、長期可用性、免疫效應細胞能否保留，以及能否與正確主藥形成穩定組合。 Chidamide已替這個類別完成重要去風險：CAPability-01在pMMR大腸癌建立三聯概念，HBI-8000全球第三期又把HDAC抑制劑加PD-1的增益推進到確認性研究。但類別成功不會自動轉化為ABT-301成功。 ABT-301下一步只看三件事：能否在三聯中維持劑量強度、能否在pMMR大腸癌產生可重現且持久的反應、能否讓人體生物標記與臨床利益指向同一條機制。三項同時成立，它才有資格從有潛力的增敏藥，走向真正的聯合治療骨幹；第二個癌種與第二個搭配藥物再成功，平台故事才算成立。 > **Chidamide證明類別可行；ABT-301必須以三聯人體數據證明資產差異，再由第二個癌種與第二個搭配藥物證明平台。** ### 免責聲明 本文為藥物機制、臨床研發與產業趨勢分析。ABT-301三聯療法的人體安全性與療效，仍待進行中的第一期／第二期試驗驗證；文中臨床前結果不應直接視為人體療效，市場規模亦不等於可及市場或營收預測，本文不構成投資建議。 --- ## 禮來領頭，大藥廠半年豪砸 1,300 億美元搶時間！ URL: https://drugnews.com.tw/articles/2026-08-10-biopharma-ma-buying-time-2026h1.html Date: 2026-08-10 Language: zh-Hant Summary: 2026 年上半年生技製藥併購總額約 1,300 億美元，買家偏向臨床後期與接近上市的資產。這一輪不是全面牛市，而是大藥廠在專利懸崖前，用交易結構與現金購買可計算的開發時間。 禮來、默沙東、艾伯維、吉利德、GSK，全都在買。 2026 年上半年，IQVIA 統計 42 筆生技製藥併購，總額約 1,300 億美元，幾乎追平 2025 全年；BioSpace 的完整公司收購口徑則數到 52 筆，明顯高於去年同期約 30 筆。最積極的禮來，一家公司就完成 9 筆、承諾金額超過 250 億美元。 有趣的是，上半年沒有標的價值超過 300 億美元的巨型合併，卻有 9 筆交易達 50 億美元以上。藥廠把目光移到臨床後期、即將讀出或接近上市的資產。 因為本十年有超過 3,000 億美元品牌藥收入暴露在市場獨占權到期風險。這波併購搶的，是一段能被計算、也趕得上專利懸崖的時間。 *圖 1｜半年千億美元併購的核心，不只是買公司，而是在專利懸崖前買進臨床與商業化時間。* ## 01｜52 筆與 42 筆，來自兩套統計口徑 BioSpace 依 S&P Capital IQ 統計到 52 筆交易，納入的是至少 1,000 萬美元、收購完整治療藥物公司的交易，交易價值還可能包含債務與其他調整；IQVIA 的口徑則是取得生技製藥標的多數股權、交易值至少 2.5 億美元，因此得到 42 筆、合計約 1,300 億美元。 IQVIA 的原文寫得很清楚：2026 上半年約 1,300 億美元，幾乎追平 2025 全年的 1,330 億美元。 ### 52 筆與 42 筆，是兩種不同交易口徑 BioSpace 的 52 筆，是一張「完整公司收購」清單。它顯示 2025 年同期約 30 筆，2026 年確實明顯變多；其中禮來以 9 筆居首，承諾金額超過 250 億美元。 IQVIA 的 42 筆則設了更高的 2.5 億美元門檻。這 42 筆平均交易規模約 31 億美元，高於 2025 年的 27 億美元；上半年共有 9 筆交易達 50 億美元以上，卻沒有一筆標的價值高於 300 億美元的「超大型合併」。 市場並非靠一樁 Celgene 等級的巨獸交易撐場，而是湧進一群 50 億到 110 億美元左右、科學與時程都相對具體的標的。 PwC 從另一個角度看到同一件事。2026 年第一季生技製藥交易額已超過 650 億美元，宣布的 10 億美元級交易有 16 筆；其年中報告直接把這一波描述為「更少由規模、更多由精準科學驅動」。買家偏好中型、可嵌入現有產品線的收購，也更常把或有價值權（CVR）、里程碑或其他條件式付款塞進交易結構。 *圖 2｜52 筆與 42 筆來自不同門檻與交易定義；同一市場、不同資料庫，不可直接加總。* ## 02｜專利時鐘倒數，藥廠搶臨床時間 答案先從一個數字開始：PwC 估計，本十年有超過 3,000 億美元的品牌藥收入暴露在市場獨占權到期風險之下。 3,000 億美元指的是本十年的收入風險曝險。大藥廠必須趕在舊產品失速前，補上一條夠大的新曲線。 早期平台當然可能很值錢，可是從動物實驗、一期、二期一路走到關鍵試驗，時間、失敗率與追加資本都很難算。當專利倒數只剩幾年，買家會更在意三件事：臨床證據走到哪裡、下一個讀出還要多久、拿進來之後能否立刻接上自己的開發與商業系統。 這就是「買時間」的意思。 默沙東 3 月宣布以約 67 億美元股權價值收購 Terns Pharmaceuticals，5 月 5 日完成交易。核心資產 TERN-701 仍在慢性骨髓性白血病一期／二期試驗；公司收購已完成，不代表已買到確定收入。 艾伯維 6 月宣布以約 109 億美元股權價值收購 Apogee Therapeutics，取得以異位性皮膚炎與氣喘為主的免疫管線。交易預計第三季完成，仍要經股東與監管條件；目前狀態停在簽約，尚未交割。 兩筆交易都在做同一件事：把一個已有臨床輪廓的資產，直接塞進既有治療領域與商業機器。 *圖 3｜買下臨床輪廓較清楚的資產，可以縮短補管線的時間，但公司交割不等於產品風險消失。* ## 03｜禮來掃貨，整合與淘汰才是驗收 2026 上半年最積極的買家仍是禮來。 依 IQVIA 與 BioSpace 的完整公司收購口徑，禮來是 9 筆；IQVIA 計算投入約 260 億美元，BioSpace 則寫承諾超過 250 億美元。「投入」「承諾」與最終實付採用不同口徑。 GLP-1 帶來的現金，正被禮來灑向肥胖與糖尿病以外的戰場：免疫、神經、體內細胞治療、疫苗都出手。IQVIA 也把 5 月一次收購三家疫苗公司視為代表性案例，三筆合計 38 億美元。 禮來正把 GLP-1 帶來的現金分散到免疫、神經、體內細胞治療與疫苗。現在只能確認它買得快；第二支柱能否長成，要看收購後的試驗、排序、整合與淘汰紀律。 併購盤點也不能把收購、授權、投資與合作混在同一張「戰利品清單」；數量會很好看，判斷卻會失真。 ## 04｜吉利德與 GSK，押的是接近商業化的腫瘤資產 吉利德的 Arcellx 交易最能看出時間的價值。 吉利德在 2026 年 4 月 28 日完成收購 Arcellx，現金加一項每股 5 美元的或有價值權，完成時隱含股權價值約 78 億美元。標的是仍在開發中的 BCMA CAR-T anito-cel；交易完成讓吉利德取得完整控制權，也消除原合作中的分潤、里程碑與權利金義務。 公司收購已完成，anito-cel 仍待監管核准；產品風險沒有隨交割一起消失。 吉利德也在 5 月 21 日完成 Tubulis 收購。條件是 31.5 億美元前期現金，加上最高 18.5 億美元里程碑；新聞只寫「50 億美元收購」會把確定支付與未來達標才支付的金額混在一起。交易已完成，主力資產 TUB-040 卻仍在 I／IIa 期試驗，產品與里程碑風險都沒有消失。 GSK 6 月宣布以 106 億美元總股權價值收購 Nuvalent，並在 7 月 15 日完成交易，取得兩項後期肺癌標靶藥 zidesamtinib 與 neladalkib，以及一期的 HER2 抑制劑 NVL-330。前兩項資產仍在 FDA 審查，目標決定日分別是 2026 年 9 月 18 日與 11 月 27 日。 因此，Nuvalent 能寫成「公司收購已完成、買進接近上市的資產」，不能寫成「產品已上市、2027 年一定貢獻獲利」。GSK 自己也把未來推出產品綁在 FDA 核准前提上。 三筆交易排出一條風險光譜：Arcellx 的產品待核准；Tubulis 的主力資產仍在早期臨床，部分價金綁里程碑；Nuvalent 的兩項後期產品仍待 FDA 決定。把「公司已收購」翻成「產品即將貢獻收入」，會把真正的風險擦掉。 ## 05｜最大一筆，也不能只看公告金額 上半年最大型交易之一，是 Sun Pharma 與 Organon。公司公告的企業價值為 117.5 億美元，預計 2027 年初左右交割；BioSpace 的 126 億美元則含債務等資料庫調整，不能當成公司公告的現金收購價。Sun Pharma 買的是女性健康、成熟藥、生物相似藥與全球通路，顯示只要能補上明確缺口，成熟產品組合也可能是合理標的。 *圖 4｜公告金額不等於今天付現；看懂付款節點與交割條件，才能看見交易真正保留的風險。* ## 06｜併購升溫，只照亮一小群 Biotech 對賣方而言，併購變多當然是好消息。它提供 IPO 以外的退出路徑，也讓有資料、有現金壓力的公司多一個選項。 這波熱度只照到少數資料成熟、能補專利缺口的公司；多數早期 Biotech 的融資環境仍沒有一起轉暖。 PwC 觀察到的是更「外科手術式」的交易：資產要能填專利缺口、切進高成長領域，最好有近期臨床或商業事件；買家也會用 CVR、里程碑把部分失敗風險推回賣方。這其實比全面牛市更挑剔。 接下來可能讓熱度降溫的因素也很具體：美國藥價政策、最惠國藥價構想、IRA 議價擴張、關稅、跨境審查與美中政策不確定性，都可能改變估值與付款條件。另一方面，如果熱門資產價格被少數大藥廠搶高，董事會也可能選擇等待資料，避開高點追價。 所以「下半年會不會繼續熱」不能只看上半年總額外推。更可靠的追蹤方式是： 1. 十億美元級交易的數量是否延續，而非只看一筆巨額交易； 2. 前期現金占總交易價值的比例，有沒有被里程碑愈拉愈低； 3. 買家是否持續偏好三期、申請前或 FDA 審查中的資產； 4. 已宣布交易能否從簽約新聞走到如期交割； 5. 收購後一年內，資產究竟被加速、延後，還是直接砍掉。 ## 07｜台灣讀者該看什麼？ 這一波沒有足夠一級證據能把台灣上市櫃公司列為直接受益者，因此不硬湊概念股。台灣 Biotech 更該注意買方偏好：臨床資料要能回答差異化、關鍵試驗路徑清楚、全球權利乾淨，CMC 與供應能接上大型藥廠。平台可以有想像力，交易文件卻必須經得起盡職調查。 對投資人而言，看到「最高交易價值」時，先拆成現金、股權價值、企業價值、CVR 與里程碑；看到「完成交易」時，再回公司公告確認究竟是已宣布、已簽約還是已交割。這兩個動作，會過濾掉不少看似驚人的數字。 ## 結語｜半年千億，買的是時間 BioSpace 的 52 筆、IQVIA 的 42 筆，都指向併購轉熱；但不同口徑不能拼成一個數字，更不能只用總額宣布資本寒冬結束。 這一輪買家搶的是臨床時間，而且必須趕在專利懸崖前抵達。越接近讀出、申請或上市，價格越容易被抬高；風險還沒消失，只是被重新分配到前期現金、CVR、里程碑與交割條件裡。 大藥廠不是突然變得浪漫。它們只是看著專利時鐘，開始用現金買時間。 ## 免責聲明 本文為生技醫藥與產業趨勢整理，不構成醫療診斷、治療建議或投資建議。交易金額可能因股權價值、企業價值、債務、現金、CVR 與里程碑口徑而不同，投資判斷應回到公司公告與監管文件。 --- ## $130 Billion in Six Months: Big Pharma Is Buying Time URL: https://drugnews.com.tw/articles/2026-08-10-biopharma-ma-buying-time-2026h1-en.html Date: 2026-08-10 Language: en Summary: Biopharma M&A reached approximately $130 billion in the first half of 2026 as buyers favored late-stage and near-launch assets. This is not a broad biotech bull market; it is Big Pharma using cash and contingent deal structures to buy measurable development time before the patent cliff. Eli Lilly, Merck, AbbVie, Gilead and GSK are all buying. In the first half of 2026, IQVIA counted 42 biopharma mergers and acquisitions worth approximately $130 billion, nearly matching the total for all of 2025. BioSpace, using a methodology focused on acquisitions of entire companies, counted 52 deals, well above roughly 30 in the same period last year. Lilly was the most active buyer, completing nine deals and committing more than $25 billion on its own. The striking part is that the first half produced no megamerger involving a target valued above $30 billion, yet nine transactions were worth at least $5 billion. Drugmakers have shifted their attention toward late-stage assets, programs approaching a data readout, and products nearing launch. That is because more than $300 billion in branded-drug revenue is exposed to loss of market exclusivity during this decade. What this acquisition wave is chasing is time: time that can be measured, and time that may still arrive before the patent cliff. *Figure 1 | The central purpose of this acquisition wave is not simply to buy companies, but to secure clinical and commercialization time before the patent cliff.* ## 01 | The 52-Deal and 42-Deal Totals Use Different Methodologies BioSpace, drawing on S&P Capital IQ, identified 52 transactions. Its list includes purchases of entire therapeutic drug companies worth at least $10 million, and transaction values may include debt and other adjustments. IQVIA instead counts transactions that secure a majority interest in a biopharma target and are worth at least $250 million. Under that definition, it recorded 42 deals totaling about $130 billion. IQVIA states the comparison clearly: approximately $130 billion in the first half of 2026, nearly equal to the $133 billion recorded for all of 2025. ### The 52 and 42 figures describe two different deal universes BioSpace's 52 transactions form a list of whole-company acquisitions. That list shows a clear increase from approximately 30 deals in the same period of 2025. Lilly ranked first with nine transactions and more than $25 billion in committed value. IQVIA applies a higher threshold of $250 million. Its 42 deals averaged approximately $3.1 billion, up from $2.7 billion in 2025. Nine transactions in the first half were worth at least $5 billion, but none involved a target valued above $30 billion, the threshold for a megamerger in this analysis. The market was not propped up by one Celgene-scale giant. Instead, it filled with targets in roughly the $5 billion to $11 billion range whose science and development timelines were relatively concrete. PwC sees the same pattern from another angle. Biopharma deal value exceeded $65 billion in the first quarter of 2026, with 16 announced transactions at or above $1 billion. Its midyear report describes this cycle as being driven less by scale and more by precision science. Buyers prefer midsize acquisitions that can fit into existing portfolios, and they increasingly use contingent value rights, or CVRs, milestones and other conditional payments in deal structures. *Figure 2 | The 52- and 42-deal counts use different thresholds and definitions. They describe the same market through different databases and should not be added together.* ## 02 | As the Patent Clock Runs Down, Drugmakers Buy Clinical Time The answer begins with one number: PwC estimates that more than $300 billion in branded-drug revenue is exposed to loss of market exclusivity during this decade. That $300 billion represents revenue at risk over the decade. Large pharmaceutical companies must build a sufficiently large new growth curve before older products lose momentum. Early platforms can certainly be valuable. But the time, failure rate and additional capital required to move from animal studies through Phase 1, Phase 2 and pivotal trials are difficult to predict. When only a few years remain on the patent clock, buyers focus more closely on three questions: How far has the clinical evidence progressed? How long until the next readout? Can the acquired asset connect immediately to the buyer's development and commercial systems? That is what it means to buy time. Merck announced in March that it would acquire Terns Pharmaceuticals for an equity value of approximately $6.7 billion and completed the transaction on May 5. The core asset, TERN-701, remains in a Phase 1/2 trial for chronic myeloid leukemia. Completion of the corporate acquisition does not mean Merck purchased guaranteed revenue. AbbVie announced in June that it would acquire Apogee Therapeutics for an equity value of approximately $10.9 billion, adding an immunology pipeline centered on atopic dermatitis and asthma. The transaction is expected to close in the third quarter and remains subject to shareholder approval and regulatory conditions. Its current status is signed, not closed. Both transactions are doing the same thing: placing an asset with an emerging clinical profile directly into an established therapeutic franchise and commercial engine. *Figure 3 | Acquiring assets with clearer clinical profiles can shorten the time needed to replenish a pipeline, but closing a corporate transaction does not eliminate product risk.* ## 03 | Lilly Is Shopping Aggressively, but Integration and Attrition Are the Real Tests Lilly remained the most active buyer in the first half of 2026. Under the whole-company acquisition methodologies used by IQVIA and BioSpace, Lilly completed nine deals. IQVIA calculates approximately $26 billion of investment, while BioSpace reports more than $25 billion in committed value. Investment, commitment and eventual cash paid are different measures. Cash generated by GLP-1 products is being deployed beyond obesity and diabetes. Lilly has moved into immunology, neuroscience, in vivo cell therapy and vaccines. IQVIA highlights Lilly's acquisition of three vaccine companies in May as a representative example; the three transactions totaled $3.8 billion. Lilly is diversifying GLP-1-generated cash across immunology, neuroscience, in vivo cell therapy and vaccines. For now, the only conclusion is that it is buying quickly. Whether a second pillar can grow will depend on post-acquisition trials, portfolio prioritization, integration and disciplined attrition. An M&A review also cannot combine acquisitions, licensing deals, investments and collaborations into one trophy list. The total will look impressive, but the analysis will be distorted. ## 04 | Gilead and GSK Are Betting on Oncology Assets Closer to Commercialization Gilead's Arcellx transaction shows the value of time particularly clearly. Gilead completed its acquisition of Arcellx on April 28, 2026. The consideration consisted of cash plus one CVR worth $5 per share, implying an equity value of approximately $7.8 billion at closing. The target asset is anito-cel, a BCMA-directed CAR T-cell therapy still in development. Closing gave Gilead full control and removed the profit-sharing, milestone and royalty obligations in the companies' previous collaboration. The company acquisition is complete, but anito-cel still awaits regulatory approval. Product risk did not disappear at closing. Gilead also completed its acquisition of Tubulis on May 21. The terms included $3.15 billion in upfront cash plus up to $1.85 billion in milestones. Calling it simply a "$5 billion acquisition" combines a certain payment with amounts that will be paid only if future conditions are met. The transaction is complete, but the lead asset, TUB-040, remains in a Phase 1/2a trial. Both product risk and milestone risk remain. GSK announced in June that it would acquire Nuvalent for a total equity value of $10.6 billion and completed the transaction on July 15. It gained two late-stage targeted lung cancer therapies, zidesamtinib and neladalkib, as well as NVL-330, a Phase 1 HER2 inhibitor. The first two assets remain under FDA review, with target decision dates of September 18 and November 27, 2026, respectively. Nuvalent can therefore be described as a completed corporate acquisition that brought in assets approaching potential launch. It cannot be described as a case in which the products are already on the market or certain to contribute profit in 2027. GSK itself conditions future launches on FDA approval. The three transactions form a risk spectrum. Arcellx's product awaits approval. Tubulis's lead asset remains in early clinical development, with part of the consideration tied to milestones. Nuvalent's two late-stage products still await FDA decisions. Translating "the company has been acquired" into "the product will soon generate revenue" erases the actual risk. ## 05 | Even One of the Largest Deals Cannot Be Read from the Headline Value Alone One of the largest transactions in the first half was the Sun Pharma and Organon deal. The companies announced an enterprise value of $11.75 billion and expect the transaction to close around early 2027. BioSpace's $12.6 billion figure includes debt and other database adjustments, so it should not be treated as the cash acquisition price announced by the companies. Sun Pharma is buying women's health products, established medicines, biosimilars and global distribution. The deal shows that a mature product portfolio can also be a rational target when it fills a clearly defined gap. *Figure 4 | An announced value is not the same as cash paid today. Payment stages and closing conditions reveal where transaction risk remains.* ## 06 | The M&A Recovery Illuminates Only a Small Group of Biotechs For sellers, more acquisitions are clearly welcome. M&A provides an exit route beyond an initial public offering and gives companies with data but limited cash another option. Yet this wave of activity reaches only a small group of companies with mature data and assets that can fill patent-cliff gaps. Financing conditions for most early-stage biotechs have not improved alongside it. PwC is seeing more surgical transactions. Assets must fill patent gaps, enter high-growth categories and preferably offer a near-term clinical or commercial catalyst. Buyers also use CVRs and milestones to shift part of the failure risk back to sellers. This environment is more selective than a broad bull market. The factors that could cool the market are equally concrete: US drug-pricing policy, most-favored-nation pricing proposals, expansion of Inflation Reduction Act negotiations, tariffs, cross-border reviews and uncertainty in US-China policy could all change valuations and payment terms. If a small number of large drugmakers bid popular assets to high prices, boards may also choose to wait for more data rather than chase the market. Whether the market remains hot in the second half therefore cannot be projected from first-half aggregate value alone. More reliable indicators include: 1. Whether the number of billion-dollar transactions continues, rather than whether one enormous deal appears; 2. Whether upfront cash represents a shrinking share of total transaction value as milestones grow; 3. Whether buyers continue to prefer Phase 3, pre-filing or FDA-reviewed assets; 4. Whether announced transactions move from signing headlines to closing on schedule; 5. Whether acquired assets are accelerated, delayed or discontinued within one year of the transaction. ## 07 | What Should Readers in Taiwan Watch? There is not enough primary-source evidence to identify a Taiwan-listed or over-the-counter company as a direct beneficiary of this wave, so there is no reason to force a thematic stock connection. Taiwan biotechs should pay closer attention to buyer preferences: clinical data must establish differentiation, the pivotal-trial path must be clear, global rights must be clean, and chemistry, manufacturing and controls, or CMC, plus supply capabilities must connect to a large pharmaceutical company. A platform may support an ambitious vision, but transaction documents still have to withstand due diligence. For investors, any claim about a "maximum transaction value" should first be separated into cash, equity value, enterprise value, CVRs and milestones. Any claim that a transaction is "complete" should then be checked against the company's announcement to determine whether it was merely announced, signed or actually closed. Those two steps will filter out many numbers that appear more dramatic than they are. ## Conclusion | A Hundred-Billion-Dollar Half-Year Spent on Time BioSpace's 52 transactions and IQVIA's 42 transactions both point to a hotter M&A market. But different methodologies cannot be merged into one number, and aggregate deal value alone cannot establish that the biotech funding winter is over. In this cycle, buyers are competing for clinical time, and they need it to arrive before the patent cliff. The closer an asset is to a readout, filing or launch, the more easily its price can rise. Risk has not disappeared. It has been redistributed across upfront cash, CVRs, milestones and closing conditions. Large pharmaceutical companies have not suddenly become romantic. They are watching the patent clock and beginning to buy time with cash. ## Disclaimer This article summarizes biopharma and industry trends. It does not constitute medical diagnosis, treatment advice or investment advice. Transaction values may differ depending on whether they refer to equity value, enterprise value, debt, cash, CVRs or milestones. Investment decisions should be based on company announcements and regulatory filings. --- ## J&J's Oral Psoriasis Drug Logged 18,000 Prescriptions in Its First Full Quarter. Is It Coming for Injectables? URL: https://drugnews.com.tw/articles/2026-08-09-icotyde-oral-il23r-launch-en.html Date: 2026-08-09 Language: en Summary: ICOTYDE logged more than 18,000 prescriptions in its first full quarter. The oral IL-23R peptide has earned entry into the high-efficacy psoriasis market, but durable value still depends on paid fills, persistence, payer quality, safety, and additional indications. One of the hardest rules to break in psoriasis may be starting to loosen: patients who want biologic-like efficacy have generally had to accept an injection. In the first full quarter after launch, Johnson & Johnson's ICOTYDE (icotrokinra) accumulated more than 18,000 prescriptions, reached roughly 11,000 patients, and was prescribed by about 6,000 healthcare professionals. A once-daily oral medicine is beginning to enter a high-efficacy market that injectables have long dominated. The U.S. Food and Drug Administration approved ICOTYDE on March 18, 2026. It is the first and currently the only FDA-approved targeted oral peptide that directly blocks the interleukin-23 receptor, an important inflammatory signaling switch in psoriasis. It is indicated for patients aged 12 years and older, weighing at least 40 kilograms, with moderate-to-severe plaque psoriasis who are candidates for systemic therapy or phototherapy. What patients want is not complicated: skin that clears sufficiently, an effect that lasts, and preferably no repeated injections. Historically, oral convenience came with an efficacy gap versus IL-23 and IL-17 biologics. Injectable biologics can deliver high levels of skin clearance, but patients must manage needles, clinic logistics, storage requirements, and the burden of long-term treatment. ICOTYDE gives an oral therapy a credible place in the high-efficacy conversation for the first time. Eighteen thousand prescriptions demonstrate willingness to try it. Commercial success will depend on refill behavior, actual reimbursement, and long-term persistence. It has reached for the injectables' lunch. For now, it has only earned a seat at the table. ## 01 | One Pill a Day: How Did J&J Get a Peptide into the Body? ICOTYDE's generic name is icotrokinra. It is taken as a 200 mg tablet once daily on an empty stomach. The drug blocks the IL-23 receptor and suppresses downstream signaling associated with psoriatic inflammation. J&J and the FDA describe it as a targeted oral peptide. That term matters. From a regulatory perspective, ICOTYDE is an oral peptide new drug reviewed through the conventional new drug application pathway. It is not a monoclonal antibody placed inside a capsule. Oral peptides face degradation and absorption barriers in the gastrointestinal tract. ICOTYDE's ability to directly engage IL-23R demonstrates that at least one molecular design can overcome those barriers. It validates this molecule, not every oral macromolecule platform, and it should not be explained with unsupported stories about disulfide bonds, permeation enhancers, or membrane penetration that do not appear in formal documents. Patients ultimately care about one question: can one pill a day maintain clear skin for one year, two years, or longer, with an acceptable safety profile? ## 02 | Eighteen Thousand Prescriptions Ignited the Launch, but Payment Quality Still Needs Testing On J&J's second-quarter 2026 earnings call, management disclosed that ICOTYDE had accumulated more than 18,000 prescriptions, approximately 11,000 patients, and about 6,000 prescribers. More than half of the prescribers were advanced practice providers, and roughly 40% were dermatologists. Within 90 days of launch, U.S. commercial insurance coverage exceeded 50%. Those figures tell us at least three things. First, clinicians are willing to try it. Six thousand prescribers extend beyond a small number of major academic centers, indicating that product education is spreading into a broader range of care settings. Second, patients have a real appetite for a high-efficacy oral option. Eleven thousand patients and 18,000 prescriptions suggest that some refills have begun, but dividing the two numbers to produce 1.6 prescriptions per patient does not prove adherence. The count may include initial prescriptions, refills, unfilled prescriptions, free trials, or patient-assistance supply. Public data do not separate those categories. Third, payer access did not start slowly. More than 50% commercial coverage within 90 days can help move a prescription toward an actual fill. But coverage does not mean unconditional reimbursement. It does not disclose prior authorization, step-therapy restrictions, patient out-of-pocket costs, rebates, or the share of prescriptions that were actually paid. That is why 18,000 prescriptions cannot be translated directly into sales success. J&J's second-quarter financial report did not disclose ICOTYDE sales separately, and the earnings call did not provide a full split among free samples, support programs, and commercially paid prescriptions. A new medicine can rapidly build prescription volume through trials and market education. The harder commercial questions arrive three to six months later: how many patients continue filling it, how far reimbursement expands, and whether net pricing holds under competition. The start is strong. The scorecard is not complete. ## 03 | An Oral Drug Finally Has an Efficacy Case Oral psoriasis medicines have never lacked the convenience argument. What they lacked was evidence strong enough to place them inside a high-efficacy treatment sequence. ICOTYDE was compared directly with deucravacitinib in the ICONIC-ADVANCE head-to-head program. According to J&J's approval announcement, approximately 70% of patients achieved an Investigator's Global Assessment score of 0 or 1 at Week 16, indicating clear or almost-clear skin, and approximately 55% achieved PASI 90. Both outcomes favored ICOTYDE. A randomized comparison within the same study is more persuasive than placing slides from unrelated trials next to each other. One-year data addressed a second question: does the effect last? J&J reported in 2025 that adult patients who had already achieved PASI 90 at Week 24 were re-randomized. Among these responders, 84% of those who continued icotrokinra maintained PASI 90 through Week 52, compared with 21% of those switched to placebo. The 84% figure is easy to misuse. It applies only to adult responders who had achieved PASI 90 by Week 24 and then continued treatment. It is not a one-year response rate for the entire enrolled population. The 21% result after treatment withdrawal also underscores the importance of continuous treatment in maintaining efficacy. Some comparisons place the 84% maintenance rate next to an 82% result for Skyrizi and conclude that an oral medicine has matched a leading monoclonal antibody. That conclusion is not valid. The 84% denominator consists of Week 24 responders, while the other percentage comes from a different trial and analysis population. Numbers that differ by only two percentage points may still answer entirely different questions. A more defensible conclusion is that ICOTYDE has materially raised the efficacy credibility of oral treatment. It outperformed deucravacitinib head to head and has supportive one-year maintenance data. Whether it can displace established IL-23 biologics at scale in routine care still requires longer-term safety, persistence, reimbursement, and direct comparative evidence. ## 04 | Once Daily Still Has to Survive the Real World There is still a gap between being oral and being taken reliably for years. The disadvantages of injectable biologics are obvious: needle aversion, training or clinic visits, and storage requirements for some products. Their dosing intervals, however, may be several weeks or longer, so patients do not have to remember treatment every day. ICOTYDE is taken once daily and its label requires administration on an empty stomach. For some patients this will be easier than an injection. For others with irregular schedules, multiple medicines, or busy routines, daily fasting administration may create a different kind of friction. Adherence therefore has to be assessed through real persistence: refill rates, discontinuation reasons, missed doses, treatment satisfaction, maintained skin clearance, and interruptions in insurance coverage. Safety should not be summarized simply as better than traditional small molecules. ICOTYDE does not carry the same boxed warning used for JAK inhibitors, but its FDA label still contains precautions involving infection, tuberculosis assessment, and immunization. ## 05 | A $5 Billion Ambition Requires Multiple Indications J&J has listed ICOTYDE as an asset with more than $5 billion in potential sales. That is a forward-looking expectation for a multi-indication asset. Peak sales in psoriasis alone and second-quarter revenue have not been disclosed. As of July 15, 2026, J&J's development pipeline showed icotrokinra in Phase 3 for psoriatic arthritis and ulcerative colitis and in Phase 2 for Crohn's disease. If those indications succeed, the commercial opportunity for oral IL-23R inhibition could expand from dermatology into rheumatology and gastroenterology. Psoriasis efficacy cannot simply be copied into other indications. Disease biology, endpoints, background therapies, competitors, and reimbursement differ. Each program has to answer its own clinical question. ICOTYDE will also grow in a crowded market. Established injectable IL-23 and IL-17 biologics have accumulated long-term data, physician familiarity, and payer contracts. Other oral small molecules and peptides are advancing. ICOTYDE must demonstrate efficacy, convenience, safety, and access simultaneously; weakness in any one dimension could cause early launch momentum to fade. ## 06 | One Successful Product Does Not Yet Validate an Entire Platform ICOTYDE's success will encourage investment in oral peptides. That industry inference is reasonable. One approval, however, does not validate every platform for oral macromolecules. Each peptide has different requirements for stability, absorption, exposure, target affinity, therapeutic window, and manufacturing. Cardiovascular, metabolic, and autoimmune diseases involve different biology and dosing demands. Oral semaglutide showed that certain peptides can be commercialized, and ICOTYDE adds another successful design. Those examples prove that paths exist; they do not prove that every path is equivalent. Calling this the eve of an oral-biologics boom by piling together different mechanisms, dosage forms, and secondary-market transaction values obscures the four measures that matter most for ICOTYDE: paid prescriptions, refill persistence, quality of insurance coverage, and efficacy and safety beyond one year. Only when those four hold up does platform spillover become a serious claim. ## 07 | What Should Taiwan-Based Readers Watch Next? First, separate prescription quantity from prescription quality. If J&J begins disclosing net sales, paid prescriptions, refills, and payer detail, readers can estimate how much of the 18,000-prescription start converts into durable revenue. Second, watch persistence beyond Week 52. Whether daily fasting administration holds up in routine care cannot be answered by launch enthusiasm. Third, follow psoriatic arthritis, ulcerative colitis, and Crohn's disease. The more-than-$5-billion narrative requires success across indications. Fourth, track Taiwan regulatory status and patient access. U.S. FDA approval does not mean that a medicine is approved, launched, or reimbursed in Taiwan. Patients should not change or stop treatment based on overseas news. The public-market company most directly connected to this analysis is Johnson & Johnson (NYSE: JNJ). At the evidence cut-off, Drugnews did not identify a Taiwan-listed company with a verifiable direct role in ICOTYDE product rights, core manufacturing, pivotal clinical development, or commercialization. We therefore do not force an unrelated Taiwan concept-stock angle. ## The Bottom Line ICOTYDE has brought an oral therapy into the high-efficacy competitive set. It has head-to-head evidence and supportive one-year maintenance data. One pill has reached for the injectables' lunch. Whether it actually takes the meal will depend on refills, reimbursement, long-term outcomes, and new indications. ## Disclaimer This article summarizes public medical and industry information. It does not constitute medical advice, a diagnosis or treatment recommendation, or investment advice. U.S. approval does not imply approval or reimbursement in Taiwan. Treatment decisions should be made with a qualified healthcare team based on the individual patient's condition. --- ## 嬌生口服藥首季 1.8 萬張處方，針劑飯碗要被搶了？ URL: https://drugnews.com.tw/articles/2026-08-09-icotyde-oral-il23r-launch.html Date: 2026-08-09 Language: zh-Hant Summary: ICOTYDE 上市後第一個完整季度累積逾 18,000 張處方。口服 IL-23R 胜肽已取得進入高療效市場的資格，真正的商業答案仍取決於續方、支付品質、長期留存與多適應症。 乾癬市場最難撼動的一條規則，可能開始鬆了：想追上生物製劑的療效，通常就得接受注射。 嬌生（J&J）的 ICOTYDE（icotrokinra）上市後第一個完整季度，就累積超過 18,000 張處方、約 11,000 位病人與 6,000 位開方者。一顆每天服用的口服藥，開始把手伸進原本由針劑主導的高療效市場。 ICOTYDE 於 2026 年 3 月 18 日獲美國食品藥物管理局（FDA）核准。它是第一款、也是目前唯一獲 FDA 核准、直接阻斷 IL-23 受體（乾癬發炎的重要訊號開關）的標靶口服胜肽，適用於 12 歲以上、體重至少 40 公斤，且適合全身性治療或光照治療的中重度斑塊型乾癬患者。 乾癬病人想要的其實不複雜：皮膚清得夠乾淨、效果維持得夠久，而且最好不要一直打針。過去的問題是，口服藥方便，療效常和 IL-23、IL-17 生物製劑有一段差距；注射型生物製劑能把皮膚廓清率拉得很高，卻要面對針劑、回診、保存與長期治療負擔。 ICOTYDE 讓口服藥第一次有資格進入高療效的對話。18,000 張處方證明市場願意試；商業成績要等續方、實際支付與長期留存接棒。 它正在搶針劑的飯碗，現在只是拿到入場券。 ## 01｜每天一顆，嬌生怎麼把胜肽送進體內？ ICOTYDE 的學名是 icotrokinra，每日一次、空腹口服 200 mg。它會阻斷 IL-23 receptor，抑制與乾癬發炎相關的下游訊號。 J&J 與 FDA 使用的定位是 targeted oral peptide，也就是標靶口服胜肽。這個名詞很重要。 監管上，ICOTYDE 屬於口服胜肽新藥，走的是一般新藥申請途徑；它並非膠囊版單株抗體。 胜肽口服會遇到胃腸分解與吸收難題；ICOTYDE 能直接作用 IL-23R，證明至少有一條設計路徑能跨過障礙。但這只驗證了這個分子，不代表所有大分子都能口服化，也不該靠未經正式文件支持的二硫鍵、促滲透或穿膜故事解釋成功。 病人只在乎一件事：每日一顆，能不能在安全性可接受下，把皮膚清除維持一年、兩年，甚至更久？ ## 02｜18,000 張處方點火，付費品質還要驗收 J&J 2026 年第二季電話會議披露：ICOTYDE 上市後已累積超過 18,000 張處方、約 11,000 位病人與約 6,000 位開方者。開方者中，超過一半是進階執業醫療人員，約 40% 是皮膚科醫師；上市 90 天內，美國商業保險覆蓋率超過 50%。 這些數字至少說明三件事。 第一，醫療端願意試。6,000 位開方者已跨出少數大型醫學中心，產品教育正擴散到更廣的診療場域。 第二，病人對口服高療效選項有真實需求。11,000 位病人對 18,000 張處方，顯示已有部分續方，但不能僅用 1.6 張處方／人的簡單除法就宣布依從性成功。處方可能包含起始、續方、未實際取藥、免費試用或患者援助，公開資料沒有拆開。 第三，保險准入起步不慢。90 天內商業保險覆蓋超過 50%，有助於從開方走到取藥；但覆蓋不代表無條件支付，也沒有告訴我們事前授權、先試用其他藥物的限制、病人自付、折讓與實際付費處方的比例。 這也是為什麼 18,000 張處方不能直接翻譯成銷售成功。J&J 第二季財報沒有單獨揭露 ICOTYDE 銷售額，電話會議也沒有完整說明免費樣品、援助方案與商業付費處方的組成。 一款新藥上市初期，可以靠試用與市場教育快速拉高處方數。三到六個月後還有多少病人持續取藥、保險支付能否擴大、淨價扛不扛得住競爭，才是商業價值的硬考題。 開局很強，成績單還沒寫完。 ## 03｜口服藥，終於拿出療效說服力 乾癬口服藥從來不缺「不用打針」這個賣點。缺的是讓醫師願意把它放到高療效治療序列裡的數據。 ICOTYDE 在 ICONIC-ADVANCE 頭對頭研究中，和 deucravacitinib 比較。依 J&J 核准公告，第 16 週約 70% 病人達到 IGA 0/1，也就是皮膚清除或幾乎清除；約 55% 達到 PASI 90，兩個結果都優於對照。兩組患者在同一研究中直接比較，說服力比拿兩張不同試驗的簡報拼在一起更高。 一年資料也回答了另一個問題：效果能不能維持。 J&J 2025 年公告顯示，第 24 週已達 PASI 90 的成人反應者重新隨機分組後，持續使用 icotrokinra 的一組有 84% 到第 52 週仍維持 PASI 90；停藥改用安慰劑的一組則為 21%。 這個 84% 很容易被寫錯。 84% 只適用於第 24 週已達 PASI 90、之後持續用藥的成人反應者。這個分母不能當成全體患者的一年反應率；停藥組僅 21%，也凸顯持續治療對維持效果的重要性。 有人把 ICOTYDE 持續使用組第 52 週「84% 維持 PASI 90」直接和 Skyrizi 的 82% 放在一起，得出「口服已和一線單抗持平」。這個比較不成立：84% 的分母是第 24 週反應者，另一個數字來自不同試驗、不同分析族群。跨試驗百分比看起來只差兩個點，實際回答的是兩個問題。 比較合理的說法是：ICOTYDE 已經把口服療法的療效可信度往上推，頭對頭優於 deucravacitinib，且一年維持資料有支持；它是否能在真實世界中大規模取代成熟 IL-23 生物製劑，仍需要更長安全性、留存、支付與直接比較資料。 ## 04｜每日一顆，真實世界留存才是答案 「口服」和「長期按時用藥」，中間仍隔著真實生活。 注射型生物製劑的缺點很直觀：怕針、需要訓練或就醫、部分產品有保存要求。但它們的給藥間隔可能是數週或更久，病人不需要每天記得服藥。 ICOTYDE 是每日一次，而且標示要求空腹服用。對某些病人，這比注射容易；對另一些忙碌、服藥多或生活作息不規律的病人，每天固定空腹反而會形成新的摩擦。 所以依從性要回到實際留存。處方續領率、停藥原因、漏服、治療滿意度、皮膚清除維持與保險中斷，才是上市後需要回答的指標。 安全性也不能只用「優於傳統小分子」一筆帶過。ICOTYDE 沒有 JAK 抑制劑同一套黑框警告，FDA 標示仍列出感染、結核評估與免疫接種等注意事項。 ## 05｜50 億美元，要靠多適應症兌現 J&J 把 ICOTYDE 列為具有 50 億美元以上銷售潛力的資產。這是公司對多適應症資產的前瞻期待；乾癬單一適應症的峰值與第二季銷售仍未揭曉。 截至 2026 年 7 月 15 日，J&J 開發管線顯示，icotrokinra 在乾癬性關節炎與潰瘍性結腸炎進入第三期，克隆氏症則在第二期。若後續適應症成功，口服 IL-23R 的商業場景才可能從皮膚科擴到風濕與腸胃領域。 乾癬療效無法直接複製到其他適應症。疾病位置、終點、背景治療、競品與支付邏輯都不同，每項試驗仍要各自作答。 它也不會在沒有競爭的真空裡成長。成熟的注射型 IL-23、IL-17 生物製劑已累積長期資料、醫師習慣與保險合約；其他口服小分子與口服胜肽正在推進。ICOTYDE 必須同時證明療效、便利、安全與支付，少一塊都可能讓開局熱度掉下來。 ## 06｜先跑出一個產品，平台價值還要慢慢累積 ICOTYDE 的成功會鼓勵更多公司投入口服胜肽，這個產業推論合理；但不能把一個產品的核准，直接寫成所有口服大分子平台都已被驗證。 每個胜肽的穩定性、吸收、暴露、標靶親和力、治療窗與製造條件都不同。心血管、代謝、自體免疫的疾病與劑量需求也完全不同。口服 semaglutide 已經證明某些多肽可以商業化，ICOTYDE 又替另一種設計增加一個成功案例；這些案例說明路可以走，不代表每一條路都一樣。 把不同機制、不同劑型與二手交易金額堆成「口服生物藥爆發前夜」，反而容易看不見 ICOTYDE 最重要的四個指標：付費處方、續方留存、保險覆蓋品質，以及一年以上的療效與安全性。這四項能站穩，才有資格談平台外溢。 ## 07｜台灣讀者接下來看什麼？ 第一，拆解處方的「質」。未來若 J&J 開始揭露淨銷售、實際付費處方、續方與保險細節，才能判斷 18,000 張處方有多少轉成可持續收入。 第二，看第 52 週以後的留存。每日空腹服用能否在真實世界維持，不能只靠上市初期熱情回答。 第三，看乾癬性關節炎、潰瘍性結腸炎與克隆氏症。50 億美元以上的敘事，需要多適應症成功才能兌現。 第四，看台灣監管與可近性。美國 FDA 核准不等於台灣已核准、已上市或已給付；個別病人也不能因為看到國外新聞就自行停換藥。 股票方面，本文最直接的公開市場公司是 Johnson & Johnson（NYSE：JNJ）。截至查證截止日，沒有找到台灣上市櫃公司在 ICOTYDE 的產品權利、核心製造、關鍵臨床或商業化上具可驗證的直接角色，因此不硬列台股概念股。 ## 最後一句 ICOTYDE 已把口服療法帶進高療效競爭區：有頭對頭數據，也有一年維持資料。 一顆藥已經把手伸向針劑的飯碗。它能否真的端走那碗飯，接下來就看續方、支付、長期結果與新適應症。 ## 免責聲明 本文為公開醫藥與產業資料整理，不構成醫療診斷、治療建議或投資建議。美國核准不等於台灣核准或給付；用藥選擇應由專業醫療團隊依個人病況判斷。 --- ## Why These Four Biotechs Can Afford Not to Sell URL: https://drugnews.com.tw/articles/2026-08-08-biotech-independence-not-selling-en.html Date: 2026-08-08 Language: en Summary: argenx, Insmed, Incyte, and Revolution Medicines show why biotech independence depends on product evidence, cash generation, a second growth curve, global execution, and financing without surrendering control. > Evidence cut-off: August 7, 2026. This analysis puts product evidence, cash generation, a second growth curve, global execution, and financing options on the same scorecard to examine what it really costs a biotech to remain independent. For years, the market compressed a biotech company's future into two outcomes. One was clinical failure, followed by a rapid collapse in asset value. The other was clinical success, followed by a sale to a large pharmaceutical company. Both outcomes are real. But treating an acquisition as the only successful finish line misses a question that is becoming increasingly important: If a company can create more value by continuing to operate independently than it can capture in a sale today, does it still need to sell? This is not a story about whether a founder has enough courage, nor is it a romantic tale of a small company standing up to a multinational pharmaceutical company. The ability to say no rests on five practical capabilities: product evidence, cash flow, a second growth curve, global commercialization, and access to capital without surrendering control. argenx, Insmed, Revolution Medicines, and Incyte illustrate four different stages. The first three do not represent a single formula for success, and Incyte is not a failed company. Putting them side by side instead reveals the dividing lines that are often missed when an asset begins to turn into an operating company. Drugnews' central view is this: Biotech independence is not the courage to reject an MNC. It is the ability to retain core rights and reinvest the first success into a second one. Without a second growth curve, independence may simply delay a sale. Without capital and commercial capabilities, rejecting an acquisition can turn scientific value into execution risk. ## Test One: Can One Drug Become an Internal Capital Engine? argenx has crossed this threshold. The VYVGART franchise generated $4.2 billion in global net product sales in 2025, up 90% year over year. The company also reached full-year operating profitability for the first time, with $1.1 billion in operating profit. In the second quarter of 2026, VYVGART global net product sales reached $1.5 billion, up 60% year over year, bringing first-half sales to $2.8 billion. The significance of those figures is not simply that the product is selling well. They show that argenx can use its own product revenue to fund three expensive activities at the same time: expanding the disease and delivery footprint of VYVGART, advancing follow-on molecules such as empasiprubart and adimanebart, and maintaining a cross-market medical, regulatory, and commercial organization. The first product is therefore no longer only an asset waiting to be valued by a buyer. It has become an internal capital-allocation tool. That does not make argenx risk-free. VYVGART remains the source of the company's product revenue. New indications and formulations may extend the growth of the same franchise, but they do not prove that a second independent molecule has achieved commercial validation. The next question for the market is no longer simply how many additional labels VYVGART can secure. It is whether the cash generated by VYVGART can produce the next product outside efgartigimod. That is the difference between having cash flow and having a repeatable platform. ## Test Two: A Second Drug Shows Whether the First Success Was Repeatable Insmed looks more like a replication test. For a long time, Insmed was effectively synonymous with ARIKAYCE. The product generated $433.8 million in global revenue in 2025 and demonstrated that the company could develop and commercialize a therapy in the highly specialized market for Mycobacterium avium complex lung disease. Yet one successful product still leaves the market unable to tell whether the result came from organizational capability or from finding one unusually attractive asset. [[Insmed FY2025](https://investor.insmed.com/2026-02-19-Insmed-Reports-Fourth-Quarter-and-Full-Year-2025-Financial-Results-and-Provides-Business-Update)] BRINSUPRI changes the question. In the second quarter of 2026, BRINSUPRI generated $309.2 million, up 49% sequentially, while ARIKAYCE generated $116.3 million, up 8% year over year. Insmed raised its full-year BRINSUPRI revenue guidance from at least $1.0 billion to $1.25 billion to $1.4 billion, while maintaining ARIKAYCE guidance of $450 million to $470 million. This is not merely an exercise in adding two revenue streams together. If BRINSUPRI scales as planned, it would show that Insmed's regulatory, payer, medical-affairs, and commercial organizations can support a second and different product. The first success would no longer look like a one-off event. The next layer is TPIP. Insmed was continuing enrollment in the Phase 3 PALM-ILD study in PH-ILD and initiated the Phase 3 PALM-PAH study in PAH in April 2026. A Phase 3 study in PPF was expected to begin by the end of 2026, while a Phase 3 study in IPF was expected to begin in the first half of 2027. The company also maintained an internal objective of filing, on average, one to two investigational new drug applications each year. That creates a three-part structure: ARIKAYCE is the established base, BRINSUPRI is the second commercial validation, and TPIP plus the early-stage pipeline must show whether a third growth curve can emerge. [[Insmed Q1 2026](https://investor.insmed.com/2026-05-07-Insmed-Reports-First-Quarter-2026-Financial-Results-and-Provides-Business-Update)] The path still consumes substantial capital. Insmed narrowed its second-quarter net loss to $13.2 million, but research and development expense reached $210.0 million and selling, general and administrative expense reached $247.5 million. As of the end of June, it still held approximately $1.2 billion in cash, cash equivalents, and marketable securities. The rapid growth of a second product increases the value of remaining independent, but it does not mean the organization has achieved risk-free self-funding. A second drug is a birth certificate for a biopharma. It is not a graduation certificate. ## The Incyte Warning: Profitability Does Not Guarantee Belief in the Next Curve Incyte is easy to oversimplify. It is not an operating failure. The company reported $5.1412 billion in total revenue and $1.2867 billion in net income in 2025. Jakafi generated $3.0925 billion in net product revenue and Opzelura generated $678.5 million. In the first quarter of 2026, total revenue reached $1.27 billion, Jakafi net product revenue was $758 million, other hematology and oncology products were growing, and the company had ten Phase 3 trials under way. [[FY2025](https://investor.incyte.com/news-releases/news-release-details/incyte-reports-fourth-quarter-and-full-year-2025-financial); [Q1 2026](https://investor.incyte.com/news-releases/news-release-details/incyte-reports-first-quarter-2026-financial-results-and-provides)] The pressure comes from timing. In its 2025 Form 10-K, Incyte states that it expects Jakafi product sales to begin declining after patent exclusivity expires in 2028 and that new products will need to offset that decline. [[Incyte 2025 Form 10-K](https://www.sec.gov/Archives/edgar/data/879169/000087916926000010/incy-20251231.htm)] That makes Incyte a useful stress test: - Revenue does not mean that the revenue base is sufficiently diversified. - Profitability does not mean that the second growth curve is large enough to cross a patent cliff. - A long list of clinical programs does not guarantee that one of them will become the next product capable of redefining the company. The right question is not whether Incyte is a healthy business today. It is whether its current cash flow can create a replacement curve that is both large enough and early enough to arrive before Jakafi declines. That is the most difficult coming-of-age test for a mature biotech. The first success solves survival. The second proves that the organization can create value again. ## Revolution Medicines: It Did Not Sell the Company, but It Did Sell a Smaller Claim Revolution Medicines carries the highest risk of the four companies and offers the clearest example of how financing can change the control equation. The status needs to be anchored to August 7, 2026. daraxonrasib had not received approval from any regulatory authority, and Revolution Medicines had no revenue from an approved product. The program, however, was no longer just an early clinical story. The Phase 3 RASolute 302 study enrolled 500 patients with previously treated metastatic pancreatic ductal adenocarcinoma. In the intention-to-treat population, median overall survival was 13.2 months with daraxonrasib and 6.7 months with chemotherapy, with an overall-survival hazard ratio of 0.40. Progression-free survival and other primary and key secondary endpoints were also met. [[RASolute 302](https://ir.revmed.com/news-releases/news-release-details/revolution-medicines-announces-asco-plenary-presentation)] The European Medicines Agency started a phased review on July 7, 2026. On July 22, the U.S. Food and Drug Administration accepted the daraxonrasib new drug application for review in previously treated metastatic pancreatic ductal adenocarcinoma. The application was included in the FDA Commissioner's National Priority Voucher pilot program, which may accelerate the review process. Acceptance means the application has formally entered review; it does not constitute approval or prove commercial success. [[EMA phased review](https://ir.revmed.com/news-releases/news-release-details/european-medicines-agency-expedites-assessment-revolution); [FDA accepts NDA](https://ir.revmed.com/news-releases/news-release-details/revolution-medicines-new-drug-application-daraxonrasib-accepted/)] The financing side had also advanced. Revolution held $1.9 billion in cash, cash equivalents, and marketable securities as of March 31, 2026. An equity and convertible-notes offering in April brought in approximately $2.1 billion in additional net proceeds. At the same time, the company raised its 2026 GAAP operating-expense guidance to $1.7 billion to $1.8 billion and reported a first-quarter net loss of $453.8 million. [[Q1 2026 results](https://ir.revmed.com/news-releases/news-release-details/revolution-medicines-reports-first-quarter-2026-financial/); [Form 10-Q](https://ir.revmed.com/static-files/e7a99744-380a-4fa9-93b6-2fb71793c8ee); [April offering 8-K](https://www.sec.gov/Archives/edgar/data/1628171/000119312526159289/d95226d8k.htm)] This is the real price of independence: the balance sheet is stronger, but the organization has already grown toward the cost structure of a large commercial-stage company. The 2025 transaction with Royalty Pharma makes the trade-off even clearer. The agreement provides up to $2.0 billion in financing. Up to $1.25 billion takes the form of a synthetic royalty on daraxonrasib, while up to $750 million is available through a senior secured term loan. After the Phase 3 result met the contractual conditions, Royalty Pharma triggered the second $250 million payment. Together with the first $250 million funded at signing, the first two tranches total $500 million. [[Funding agreement](https://www.royaltypharma.com/news/royalty-pharma-and-revolution-medicines-enter-into-funding-agreements-for-up-to-2-billion/); [2026-04-13 8-K](https://www.sec.gov/Archives/edgar/data/1802768/000119312526152145/d871473d8k.htm)] In return, Royalty Pharma receives a 15-year, tiered claim based on annual worldwide net sales. Under the rates corresponding to the first two tranches, the royalty is 4.55% on annual net sales from zero to $2 billion, 2.50% from $2 billion to $4 billion, 1.00% from $4 billion to $8 billion, and zero on net sales above $8 billion. The rates would increase if Revolution elects to draw all remaining royalty tranches. [[Royalty Pharma 2026-04-13 8-K](https://www.sec.gov/Archives/edgar/data/1802768/000119312526152145/d871473d8k.htm)] The package also includes up to $750 million in senior secured term loans. The first loan draw is tied to specified FDA approval conditions for daraxonrasib. The interest rate is SOFR plus 5.75%, subject to a 3.5% SOFR floor. Revolution therefore did not sell nothing. It sold a portion of the product's future cash flows for 15 years and accepted the possibility of debt costs. What it retained was corporate control, leadership of global development and commercialization, and the remaining upside in the RAS platform. The structure matters because the capital market now offers alternatives to an MNC acquisition. Equity, convertible debt, synthetic royalties, and milestone-based loans can turn “sell the company because it needs money” into a set of smaller, configurable decisions. Every option has a price. Equity dilutes shareholders. Debt carries interest. A royalty removes part of successful product cash flow for the agreed 15-year period. If the lead program is delayed or commercialization underperforms, the global organization built in advance will amplify the loss. ## The Five Tests of Whether a Biotech Has Earned the Option Not to Sell Putting the four companies on one page produces a more useful framework than asking which one might be acquired. ### 1. Product evidence Does the lead asset only have attractive early data, or has it crossed pivotal trials, regulatory review, and real market adoption? Revolution's Phase 3 result reduced risk, but the product remained unapproved at the evidence cut-off. argenx and Insmed already had commercial readbacks. ### 2. Cash engine Can the first product fund the next round of research, or does it only extend the company's life? argenx had reached full-year operating profitability. Insmed was expanding revenue with a second product but remained loss-making. Revolution continued to rely on the capital markets and structured financing. ### 3. Second growth curve Does the next success come from additional indications for the same molecule, or from a separate molecule? Both can create value, but they offer different levels of evidence that the platform and organization are repeatable. ### 4. Global execution Can the company manage multinational registration, supply, medical affairs, payer access, and commercialization itself? If every critical capability remains outsourced, retaining global rights does not necessarily mean that the company can capture global value. ### 5. Financing options and the capacity to absorb failure Can the company secure sufficient capital without surrendering control? More importantly, if a Phase 3 trial fails or a launch is delayed by a year, can its balance sheet and organization survive? Real independence is not the ability to reject one offer. It is the capacity to endure a bad outcome after saying no. ## What Taiwan's Biotech Sector Should Take Away Is Not “Never Sell” Taiwanese biotech companies do not need to copy argenx, Insmed, or Revolution Medicines. Nor should overseas licensing, co-development, or an acquisition be treated as failure. For companies with smaller capital bases and still-developing global commercial experience, an early license can be a rational allocation of risk. Transferring regional rights to a more capable partner may also move a drug to patients faster. The more useful questions are: - Does a business-development deal merely monetize one asset, or does it buy time for the next one? - Which geographies, indications, manufacturing rights, or co-development rights does the company retain? - After the first product succeeds, has a second independent molecule entered the clinic? - Does “internationalization” mean licensing assets out, or building an organization that can lead global trials and commercialization? - Do financing tools distribute risk, or carve up too much of the future cash flow too early? At the evidence cut-off, we did not have sufficient primary-source support to place a specific Taiwan-listed company directly into the same maturity framework. We therefore do not force a list of Taiwan stock tickers into the article. The framework is better used to examine each company's retained rights, cash runway, and second growth curve than to manufacture a concept-stock basket. ## Conclusion: Being Acquired Is Not a Defeat, and Remaining Independent Is Not a Medal An acquisition can give a product faster access to capital, manufacturing capacity, and global distribution. Independence preserves more long-term upside for the original company, but it also leaves the clinical, regulatory, manufacturing, commercial, and financing risks inside that company. The most accurate dividing line is therefore not whether a biotech sells. It is whether the biotech has a real choice. An acquisition can be the successful realization of an asset's value. Choosing independence requires proof that the company can turn one success into a second. Only when the risk-adjusted value of remaining independent exceeds the acquisition premium—and when the company has the capital, organization, and resilience to absorb one failure—does “not selling” become a strategy rather than a posture. ## Disclaimer This article is an analysis of biotechnology, corporate strategy, and capital structure. It does not constitute medical advice, a diagnosis or treatment recommendation, or investment advice. Investigational products remain subject to clinical, regulatory, manufacturing, and commercialization risks. Readers should make investment decisions independently. --- ## 不賣給大藥廠，4 家生技公司憑什麼自己做大？ URL: https://drugnews.com.tw/articles/2026-08-08-biotech-independence-not-selling.html Date: 2026-08-08 Language: zh-Hant Summary: argenx、Insmed、Incyte 與 Revolution Medicines 選擇保留獨立經營權。真正的不賣資格，來自產品證據、現金引擎、第二曲線、全球執行與不交出控制權也能取得資金的能力。 生技公司做出好藥，最常見的結局是被大藥廠買走。但 argenx、Insmed、Incyte 與 Revolution Medicines，選擇了另一條更難的路：自己把藥推到全球、自己養下一條管線，也自己承擔失敗。 它們的底氣都可以換算成數字。argenx 靠 VYVGART（降低致病性 IgG 抗體的自體免疫暢銷藥）在 2025 年賣出 42 億美元，首度全年營業獲利；Insmed 的第二款藥 BRINSUPRI 在 2026 年第二季又賣出 3.092 億美元，季增 49%；Revolution 還沒有核准產品，卻在 4 月透過股票與可轉債取得約 21 億美元淨收入。 它們用現金與產品，替自己換到拒絕整家公司出售的底氣。時間，則是這份底氣最昂貴的成本。 這正是生技公司最難跨過的一關。臨床成功後，被大型藥廠收購通常是最快的變現路；但如果第一款藥已經能養研發、第二款產品開始成形、全球商業化有人做，公司繼續獨立，可能比一次出售更值錢。 判斷一家公司是否真的有「不賣資格」，不能只看創辦人的態度，而要看五件現實的事：產品證據夠不夠、第一款藥能不能變成現金引擎、第二條曲線有沒有出現、全球商業化能不能自己做，以及公司能否在不交出控制權的前提下拿到資金。 四家公司走在不同位置，剛好把生技公司從一項資產長成一家企業的分水嶺攤在桌上。第二曲線決定獨立能走多遠；資金與商業化能力，決定科學價值能否活著抵達市場。 ## 第一關｜一款藥能不能從產品，變成資本引擎？ argenx 已經跨過這一關。 VYVGART 系列 2025 年全球產品淨銷售 42 億美元，較前一年成長 90%；公司同年首度達到全年營業獲利，營業利益 11 億美元。到了 2026 年第二季，VYVGART 全球產品淨銷售再達 15 億美元，年增 60%；上半年累計 28 億美元。 42 億美元一進帳，argenx 的身分就變了。它已能用產品收入支付三種昂貴工作：擴大 VYVGART 的疾病與給藥布局、推進 empasiprubart、adimanebart 等後續分子，以及維持跨市場的醫學、法規與商業化組織。 第一個產品已從「等待別人出價的資產」，變成公司手上的資本引擎。 但 argenx 並未因此免除風險。VYVGART 仍是收入主體；多適應症與不同劑型能延長同一平台的成長，第二個獨立分子卻還沒完成商業驗證。下一張成績單很清楚：VYVGART 賺進來的現金，能不能養出下一個非 efgartigimod 的產品？ 這是「有現金流」與「可重複平台」之間的差別。 ## 第二關｜第二款藥，驗收組織能不能複製成功 Insmed 的案例更像一張能力複製測驗。 過去很長一段時間，Insmed 幾乎等同 ARIKAYCE。2025 年 ARIKAYCE 全球收入 4.338 億美元，證明公司能在 MAC 肺病這類高度專業市場完成開發與商業化；但單一產品成功，市場仍很難判斷，那是組織能力，還是剛好遇到一個好資產。 BRINSUPRI 改變了這個問題。 2026 年第二季，BRINSUPRI 收入已達 3.092 億美元，較第一季再成長 49%；ARIKAYCE 為 1.163 億美元，年增 8%。公司因此把 BRINSUPRI 全年收入指引，從至少 10 億美元上修至 12.5 億至 14 億美元；ARIKAYCE 指引維持 4.5 億至 4.7 億美元。 BRINSUPRI 的意義超過兩個產品收入相加。它如果照計畫放量，代表 Insmed 的法規、支付方、醫學事務與商業團隊，能把成功複製到第二個不同產品。 更下一層是 TPIP。公司正推進 PH-ILD、PAH、PPF、IPF 等第三期計畫，並維持每年平均申報一到兩個 IND 的內部研發目標。這形成三段式結構：ARIKAYCE 是已建立的底座，BRINSUPRI 是第二次商業化驗證，TPIP 與早期管線則負責回答第三條曲線能不能形成。 不過，這條路仍然燒錢。Insmed 第二季雖把淨損壓到 1,320 萬美元，但研發費用 2.10 億美元、銷售與管理費用 2.475 億美元，兩者都在增加；截至 6 月底，現金、約當現金與有價證券仍約 12 億美元。第二款藥拉高獨立價值，組織距離無風險自我供血仍有一段路。 第二款藥是出生證明，不是畢業證書。 ## Incyte 的提醒｜會賺錢，市場為何還不買單？ Incyte 最容易被過度簡化。 Incyte 仍是一家會賺錢的公司。2025 年公司總收入 51.412 億美元、淨利 12.867 億美元；Jakafi 淨產品收入 30.925 億美元，Opzelura 6.785 億美元。2026 年第一季總收入 12.7 億美元，Jakafi 7.58 億美元，其他血液與腫瘤產品也在成長，公司同時有 10 項第三期試驗進行中。 問題是，Incyte 自己在 2025 年 10-K 明確寫出：Jakafi 專利獨占到期後，預期產品銷售將自 2028 年開始下滑，公司必須靠新產品抵銷缺口。 Incyte 現在會賺錢，收入卻仍高度依賴 Jakafi。十項第三期試驗能否在 2028 年前補上缺口，才是市場願不願意重估它的關鍵。 這也是所有成熟生技公司最難的成人禮：第一個成功解決生存，第二個成功才證明組織能重複創造價值。 ## Revolution｜賣掉一部分未來，換回控制權 Revolution Medicines 是四家公司中風險最高、也最能解釋「融資如何改變控制權」的一家。 先更新到 2026 年 8 月 3 日的狀態。 daraxonrasib 還沒有獲任何主管機關核准，Revolution 也還沒有來自核准產品的收入；但它已不再只是早期臨床故事。第三期 RASolute 302 納入 500 位先前接受過治療的轉移性胰臟導管腺癌患者。在整體意向治療族群，daraxonrasib 的中位總生存期為 13.2 個月，化療為 6.7 個月，OS HR 0.40；PFS 與其他主要、關鍵次要終點也達標。 歐洲藥品管理局（EMA）已在 2026 年 7 月 7 日啟動分階段審查（phased review）；7 月 22 日，美國食品藥物管理局（FDA）也接受 daraxonrasib 的新藥申請（NDA）進入審查，適應症為先前接受過治療的轉移性胰臟導管腺癌。這項申請納入 FDA「國家優先審查券」試行計畫，審查程序可能加快；NDA 受理只代表正式進入審查，距核准與商業化還有兩關。 資本端同樣往前推了一大步。Revolution 在 2026 年 3 月 31 日持有 19 億美元現金、約當現金與有價證券；4 月的股票與可轉債融資又帶來約 21 億美元淨收入。與此同時，公司把 2026 年 GAAP 營運費用指引提高到 17 億至 18 億美元，第一季淨損 4.538 億美元。 這就是獨立的真實價格：資本更充足，組織成本也已經先長到接近大型商業化公司的尺度。 2025 年與 Royalty Pharma 的交易，更能說明 Revolution 到底做了什麼。 雙方安排最高 20 億美元資金，其中最高 12.5 億美元是 daraxonrasib 的合成式權利金（synthetic royalty），最高 7.5 億美元是有擔保貸款。第三期結果達標後，Royalty Pharma 已觸發第二筆 2.5 億美元；加上簽約時的 2.5 億美元，前兩筆合計 5 億美元。 代價是：Royalty Pharma 取得 15 年期、按全球年銷售分級的權利。以目前前兩筆資金對應的級距，年銷售 0–20 億美元部分為 4.55%，20–40 億美元部分為 2.50%，40–80 億美元部分為 1.00%，80 億美元以上為零。若 Revolution 未來選擇把其餘權利金分批款（royalty tranches）全部提領，分潤比例還會上升。 另有最高 7.5 億美元的優先順位有擔保定期貸款（senior secured term loan）；首筆需在 daraxonrasib 符合約定的 FDA 核准條件後提領，利率為擔保隔夜融資利率（SOFR）加 5.75%，並設 3.5% SOFR 下限。 Revolution 先讓出未來 15 年的一部分產品現金流，也接受可能的債務成本，換回公司控制權、全球開發主導權與 RAS 平台的上行空間。 這個結構之所以重要，是因為資本市場開始提供跨國大型藥廠（MNC）以外的選項：股權、可轉債、合成式權利金、里程碑式貸款，可以把「沒錢就必須賣掉整家公司」改成多個可組合的決策。 但每個選項都有價格。股權會稀釋，債務要付利息，權利金會在約定的 15 年期間切走一部分成功後的現金流；若核心產品延遲或商業化不如預期，提前搭起的全球組織也會放大虧損。 ## 想說「不賣」，先過這五關 把四家公司放在一起，可以得到一個比「誰會被收購」更實用的框架。 1. 產品證據 核心資產只是有漂亮的早期數據，還是已跨過關鍵試驗、法規與真實市場？Revolution 的第三期成功降低了風險，但截至查證日仍未核准；argenx 與 Insmed 已有商業化讀回。 2. 現金引擎 第一款產品能否支付下一輪研發，而不是只能延長公司壽命？argenx 已有全年營業獲利；Insmed 正在用第二款產品擴大收入，但仍有淨損；Revolution 主要靠資本市場與結構化融資。 3. 第二曲線 第二個成功來自同一分子的適應症擴張，還是另一個獨立分子？兩者都能創造價值，但對「平台可重複性」的證明強度不同。 4. 全球執行 公司是否能自己完成多國註冊、供應、醫學事務、支付方與商業化？若所有關鍵能力仍外包，保留全球權利不必然等於有能力把全球價值收回來。 5. 融資選項與失敗承受力 公司能不能在不交出控制權的前提下拿到足夠資金？更重要的是，若一次第三期失敗、上市延遲一年，現金與組織還撐不撐得住？獨立公司的硬門檻，是拒絕報價後仍撐得過第三期失敗或上市延誤。 ## 台灣生技更該學會替自己保留選擇權 台灣生技公司不需要複製 argenx、Insmed 或 Revolution，也不該把海外授權、共同開發或併購視為失敗。 對資本規模較小、全球商業化經驗仍在累積的公司來說，早期授權可能是最合理的風險分配；把區域權利交給更有能力的夥伴，也可能讓藥更快走到患者面前。 台灣公司該追問的是： - 一筆 BD 是把單一資產變現，還是替下一個資產買到時間？ - 公司保留哪些地區、適應症、製造或共同開發權？ - 首款產品成功後，有沒有第二個獨立分子進入臨床？ - 所謂「國際化」，是簽出資產，還是逐步建立能主導全球試驗與商業化的組織？ - 融資工具是否把風險分散，還是把未來現金流過早切得太薄？ 台灣市場目前很難找到成熟度完全相同的上市櫃公司。與其硬湊概念股，這套框架更適合用來檢查每家公司的權利保留、現金跑道與第二曲線。 ## 結論｜被收購能兌現，獨立要能重複成功 併購可以讓產品更快取得資金、產能與全球通路；獨立則讓原公司保留更大的長期上行，但也把臨床、法規、製造、商業化與融資風險全部留在自己身上。 最後的分界很簡單：公司手上還有沒有選擇權。 被收購，可能是一項資產成功兌現；選擇獨立，則必須證明公司能把一次成功重複成第二次。當繼續獨立的風險調整後價值高於收購溢價，而且公司有錢、有組織、也承受得起一次失敗時，「不賣」才從態度變成策略。 ## 免責聲明 本文為生技醫藥、公司策略與資本結構整理，不構成醫療診斷、治療建議或投資建議。文中候選藥仍可能面臨臨床、法規、製造與商業化風險；投資決策應自行評估。 --- ## 22 億美元，買下一個自體免疫藥王？ URL: https://drugnews.com.tw/articles/2026-08-07-argenx-forte-fb102-acquisition.html Date: 2026-08-07 Language: zh-Hant Summary: argenx 擬以 22 億美元現金收購 Forte，把早期抗 CD122 抗體 FB102 納入管線。真正的考題不是一組一期 p 值，而是 VYVGART 的單品成功能否被複製成可重複的公司能力。 靠 VYVGART（可降低致病性 IgG 抗體、用於重症肌無力等自體免疫疾病的暢銷藥）崛起的 argenx，現在拿出約 22 億美元現金，準備收購 Forte Biosciences。 買方與賣方的座位，換了。 這個轉身的底氣，來自 VYVGART。2026 年第二季，VYVGART 產品淨銷售達 15.16 億美元；上半年累計 28.13 億美元。到 6 月底，argenx 持有 36 億美元現金與約當現金，加上 16 億美元流動金融資產，合計流動性約 52 億美元。 但這筆交易最值得看的，並不是「一家有錢的新藥公司開始買東西」這麼簡單。 Forte 沒有上市產品，核心資產 FB102 還停在早期臨床。argenx 卻願意用相當於自身期末流動性約四成的價格，提前買走它的未來。 所以，22 億美元買的究竟是什麼？ 是一支抗 CD122 抗體（CD122 是會影響 T 細胞與自然殺手細胞的免疫受體元件）？兩組一期訊號？還是 argenx 相信，自己已經把 VYVGART 從一顆藥的成功，提煉成一套可以再做一次的公司能力？ 先把答案說在前面： 1️⃣ 已經證明的，是 argenx 能把一個新機制做成全球大藥。 2️⃣ 目前買到的，是 FB102 在白斑與乳糜瀉出現的早期概念驗證訊號。 3️⃣ 還沒證明的，是這些訊號能否在第二期重現，並長成另一個「一藥多適應症」平台。 第一顆大藥，證明一家公司會贏。 第二顆，才證明它知道自己為什麼贏。 *圖 1｜交易已宣布、仍待交割；22 億美元買到的是 FB102 的早期選擇權，不是第二個 VYVGART 的完成證明。* ## 01｜交易還沒完成，價格已先把期待買滿 2026 年 7 月 27 日，argenx 宣布與 Forte 簽署確定性合併協議，擬以每股 77 美元現金收購 Forte，總股權價值約 22 億美元。 這裡有一個很重要的時態：截至 7 月 31 日，雙方是「宣布收購、等待交割」，不是 Forte 已經被收入囊中。 交易仍須取得過半流通股應賣，並完成美國反壟斷審查等一般條件；雙方預期在 2026 年第三季完成。argenx 會以手頭現金支付，交易沒有融資條件。 公司採用的官方溢價口徑，是相較 Forte 自 7 月 9 日公布白斑 Phase 1b 數據以來的成交量加權平均價，高出約 86%。 這個價格顯示，argenx 沒打算等所有答案都出來才下注。 等第二期成功、適應症排序明朗、皮下注射劑型與商業模型都成形後，風險會變小，但價格也不會還在原地。argenx 選擇在證據仍薄、想像空間仍大的時候進場，等於同時買下控制權與開發主導權。 交易也不是完全沒有鋪墊。Forte 在 4 月進行約 1.5 億美元公開募資，argenx 是其中一名策略投資人。不過，一級文件沒有披露 argenx 個別認購多少，不能把整筆募資額直接算在 argenx 頭上。 這些差別看似小，卻決定我們是在讀交易，還是在替交易補故事。 *圖 2｜VYVGART 已證明 argenx 能把單一機制做成全球大藥；FB102 要驗證的是這套能力能否再複製一次。* ## 02｜VYVGART 給了 argenx 一張昂貴的入場券 argenx 敢提 22 億美元現金，真正的支票簿是 VYVGART。 VYVGART 的成分 efgartigimod 透過阻斷新生兒 Fc 受體（FcRn），加速體內 IgG 降解。許多自體免疫疾病由異常 IgG 抗體驅動，這讓一個機制有機會跨進多個適應症，而不是只服務一種疾病。 它的商業價值也不只來自「第一個適應症賣得好」。更關鍵的是，argenx 已經建立一套重複開發的方法：先找到具強生物學理由的患者族群，再安排適應症順序、劑型、全球臨床與商業化路徑，把同一顆藥的價值一層一層打開。 公司提出的 Vision 2030，要在 2030 年以前讓全球 5 萬名患者接受治療、累積 10 個核准標示適應症，並推動 5 個候選藥進入註冊性開發。 這才是 Forte 對 argenx 的吸引力。 FB102 被選中的理由，並不限於白斑或乳糜瀉。argenx 看上的，是 CD122 背後可能橫跨多種疾病的免疫軸。 argenx 想複製的其實是一套開發方法：找出共同病理，先在一個疾病建立訊號，再把一支藥展開成多個適應症。它當然不會複製 VYVGART 的分子。 這也是這筆交易最硬的一道考題。 如果 FB102 最後只在一個小適應症出現有限效果，22 億美元會顯得非常昂貴；若它能在多種由 T 細胞、NK 細胞驅動的疾病中反覆做出臨床差異，這支早期資產才有機會長成第二台成長引擎。 *圖 3｜CD122 是 IL-2／IL-15 受體共用的 β 次單元；Treg 相對保留仍是早期設計假說，長期安全窗尚未證明。* ## 03｜先改正一個關鍵：FB102 切的是 IL-2／IL-15 FB102 是 Forte 開發的抗 CD122 單株抗體候選藥。 CD122 是 IL-2 與 IL-15 受體共用的 β 次單元，不是 IL-12 受體。這個機制若寫錯，後面的免疫學推論也會跟著偏掉。 IL-15 對 NK 細胞以及部分記憶 T 細胞的存活與活化很重要；IL-2 則同時牽涉效應 T 細胞與調節型 T 細胞（Treg）。直接碰 CD122 的難處在於：既要降低致病 T／NK 細胞的訊號，又不能把維持免疫平衡的 Treg 一起過度壓掉。 Forte 對 FB102 的設計主張，就是尋找這個治療窗。 公司的體外資料顯示，FB102 可在 IL-2 或 IL-15 刺激下抑制人類 T 細胞與 NK 細胞的增殖或活化；在其 IL-2 刺激的實驗裡，Treg 增殖相對被保留。 但這裡不能跳太快。 「盡量保留 Treg」目前主要來自體外、非人靈長類與早期人體資料，還不是大型臨床已證明的選擇性，更不能直接寫成「只打壞細胞、不傷正常免疫」或「一定比 JAK 抑制劑安全」。 早期藥效學告訴我們這個想法值得測；只有更大、更久的隨機試驗，才能告訴我們這個治療窗是否真的存在。 ## 04｜兩個一期訊號，值不值得 22 億美元？ argenx 決定出手前，FB102 已經在乳糜瀉與白斑各交出一組早期人體資料。 先看乳糜瀉。 乳糜瀉是一種由麩質觸發的免疫疾病。目前治療核心仍是嚴格無麩質飲食，但實際生活裡很難完全避免交叉污染，也不是每位患者都能靠飲食把症狀與腸黏膜損傷控制好。 FB102 的 Phase 1b 是小型、隨機、雙盲、安慰劑對照研究，共 32 人；其中 24 人接受 FB102、8 人接受安慰劑。受試者接受 4 劑 10 mg/kg 治療，接著進行 16 天麩質挑戰。 Forte 報告，在 VCIEL 複合組織學指標上，安慰劑組相較基線平均變化為 -1.849，FB102 組為 0.079，p=0.0099；CD3+ 上皮內淋巴球密度則是安慰劑增加 13.3，FB102 下降 1.5，p=0.0035。 這組數據的價值，在於人工麩質挑戰下同時看見組織學方向與免疫細胞訊號，證據比單一症狀問卷更紮實一些。 它的限制也很清楚：總樣本只有 32 人、挑戰期 16 天，不能直接外推成患者長期自由飲食後仍能受到保護。研究裡 32 人都完成第 32 天活檢，治療期間不良事件多為第 1 級；但小型一期也不足以排除低頻或長期安全風險。 再看白斑。 Forte 7 月公布的是 43 人早期 Phase 1b 分析，32 人接受 FB102、11 人接受安慰劑。較保守的 efficacy-evaluable 口徑為 32 對 10 人。 第 24 週、中央判讀的臉部白斑面積與嚴重度指數（F-VASI）平均改善，FB102 組為 29.6%，安慰劑組為 7.9%，安慰劑校正差 21.7 個百分點，p=0.020。 這是「臉部」指標，不是全身皮膚改善，也不能拿去與其他藥物的不同試驗直接比輸贏。 在基線 F-VASI 至少 0.75 的子群裡，FB102 17 人、安慰劑只有 4 人；第 24 週平均改善為 43.2% 對 0.5%。數字亮眼，但安慰劑只有 4 人，應被視為探索性訊號，不是放大後就能當作確證療效的答案。 這兩組早期研究，足以解釋為什麼 argenx 願意進場；還不足以證明 argenx 已經找到下一個 VYVGART。 真正會改變評價的，是下一層證據：乳糜瀉 Phase 2 能否重現組織學與臨床訊號、白斑能否在更大樣本維持效果、斑禿等下一個適應症是否出現相同方向，以及更長追蹤是否仍保有可接受的安全窗。 截至 7 月 31 日，乳糜瀉 Phase 2 尚未公布結果。公司仍稱預期 2026 年下半年讀出，但 ClinicalTrials.gov 列出的預估主要完成時間是 2027 年 2 月。兩個時程存在落差，所以現在能寫的是「公司預期」，不是確定的揭盲日期。 *圖 4｜乳糜瀉與白斑 Phase 1b 是起點；Phase 2 重現、更大樣本與長期安全，才是第二個商業支柱的真正門檻。* ## 05｜故事夠大，三道風險仍得過關 22 億美元對一支早期候選藥來說，價格不低。 它把三件還沒發生的事，先寫進交易價裡。 第一，Phase 1b 的訊號要能重現。 早期小樣本容易受基線差異、分析族群與單一受試者影響。尤其白斑的探索子群只有 17 對 4 人，下一階段若效果縮水，不一定代表藥物完全無效，卻會明顯改變可接受的商業估值。 第二，CD122 的安全窗要能拉長。 T 細胞與 NK 細胞不是只在疾病裡工作，也負責感染與免疫監控。短期沒有第 3 級以上事件，不等於長期大樣本沒有免疫抑制、感染或其他低頻風險。FB102 的成敗，會取決於它能否在有效劑量下維持足夠選擇性。 第三，argenx 要選對適應症順序。 乳糜瀉有清楚的未滿足需求，但臨床終點、飲食背景與支付路徑都不簡單；白斑患者多，卻已有外用 JAK 等治療競爭；斑禿的開發也正快速擁擠。哪個疾病先做、用什麼終點、是否發展皮下劑型，都會影響時間與資本效率。 VYVGART 的成功，讓市場容易相信 argenx 已經有答案。 但公司能力不是一張永久通行證。它能提高選對分子與適應症的機率，不能替 Phase 2 和 Phase 3 給出結果。 ## 06｜台灣不是沒有角色，但不能亂湊「概念股」 台灣讀者可以從兩條真實產業鏈看這件事。 第一條是 argenx 既有產品的商業化鏈。 再鼎醫藥（NASDAQ：ZLAB；港股：9688）持有 efgartigimod 大中華區權利。健喬（4114）則已從再鼎台灣取得包含 VYVGART 在內六項療法、為期 15 年的台灣代理與商業化權利。 所以健喬與 argenx 的連結，是 VYVGART 台灣商業化生態；不是 Forte 收購案參與方，也不是 FB102 的台灣夥伴。現有一級資料也不足以讓本文直接宣稱 VYVGART 已在台灣核准或上市。 第二條是白斑同適應症的新藥研發。 安立璽榮‑KY（7871）的 EI‑001／Indemakitug 是 anti‑IFN‑γ 白斑單株抗體；安成生技（6610）的 AC‑1101 則是外用 JAK 抑制劑，已布局白斑與圓禿等免疫皮膚疾病。 這兩家公司與 FB102 沒有合作關係。它們的價值，是提供台灣市場自己的研發座標：同樣處理免疫驅動的色素或毛髮疾病，卻選擇不同靶點、不同劑型與不同開發風險。 反過來說，台康、永昕、中裕等公司雖有抗體製造或 CDMO 能力，目前沒有一級證據顯示參與 FB102。只因為它是一支抗體，就推論台灣所有抗體公司會受惠，並不負責任。 ## 結語｜VYVGART 證明會做大；FB102 要證明能再做一次 這筆交易最有意思的地方，不是 argenx 終於有錢買公司，也不是 FB102 某一個 p 值特別漂亮。 它把一家成功新藥公司的下一個難題，攤在市場面前。 VYVGART 已經證明 argenx 能找出一條有價值的免疫機制，完成全球臨床、劑型與商業化，把一顆藥推成數十億美元產品。 FB102 則要回答另一題：這套方法能不能離開 FcRn，轉到 CD122，再做一次？ 22 億美元買到的，是一張進入考場的門票，不是畢業證書。 下一個真正的分水嶺，不會是另一句「pipeline in a product」，而是乳糜瀉 Phase 2 能否重現、白斑更大試驗能否站穩，以及 CD122 的安全窗能否撐過更長時間。 第一顆大藥讓 argenx 有了主動出價的底氣。 第二顆大藥，才會決定它究竟是一家擁有超級單品的公司，還是一台能反覆產生新藥的免疫學平台。 --- ## $2.2 Billion for the Next Autoimmune Blockbuster? URL: https://drugnews.com.tw/articles/2026-08-07-argenx-forte-fb102-acquisition-en.html Date: 2026-08-07 Language: en Summary: argenx plans to pay $2.2 billion in cash for Forte and its early anti-CD122 antibody FB102. The real test is not one Phase 1b p-value, but whether VYVGART's product success reflects a development engine the company can repeat. argenx built its rise on VYVGART, a blockbuster that reduces pathogenic immunoglobulin G antibodies and is used across diseases such as generalized myasthenia gravis. Now the company is preparing to spend approximately $2.2 billion in cash to acquire Forte Biosciences. The buyer and seller have changed seats. VYVGART made that reversal possible. The product generated $1.516 billion in net sales in the second quarter of 2026 and $2.813 billion in the first half. At the end of June, argenx held $3.6 billion in cash and cash equivalents plus $1.6 billion in current financial assets, for total liquidity of about $5.2 billion. But this transaction is more than a newly wealthy biotech going shopping. Forte has no commercial product, and its lead asset, FB102, remains in early clinical development. Even so, argenx is willing to commit an amount equal to roughly 42% of its quarter-end liquidity to buy the program before the decisive trials have read out. What, exactly, does $2.2 billion purchase: an anti-CD122 antibody, two Phase 1b signals, or the belief that argenx has distilled VYVGART's success into a repeatable corporate capability? Three distinctions matter. argenx has already shown that it can turn a novel mechanism into a global medicine. FB102 has produced early proof-of-concept signals in vitiligo and celiac disease. What remains unproven is whether those signals can survive Phase 2 and become another product with credible multi-indication potential. The first blockbuster proves that a company can win. The second shows whether it understands why it won. *Figure 1. The acquisition has been announced but has not closed. The price buys an early clinical option, not proof of a second VYVGART.* ## 01 | The Deal Has Not Closed, but the Price Already Includes Ambition On July 27, 2026, argenx announced a definitive merger agreement to acquire Forte for $77 per share in cash, representing approximately $2.2 billion in total equity value. The tense is important. As of the July 31 research cutoff, the companies had announced an acquisition and were working toward closing; Forte had not yet become part of argenx. The tender still required a majority of outstanding shares and the completion of customary conditions, including the U.S. antitrust waiting period. Both companies said they expected the transaction to close in the third quarter of 2026. It is funded with cash on hand and is not subject to a financing condition. The official premium was approximately 86% relative to Forte's volume-weighted average share price since it released its Phase 1b vitiligo data on July 9. That is not the same as an 86% premium to the previous day's closing price, and the distinction matters when investors assess how much of the clinical story was already reflected in the market. The valuation says argenx does not intend to wait for every uncertainty to disappear. Once a successful Phase 2, a clear indication sequence, a subcutaneous formulation strategy and a commercial path are visible, the scientific risk would be lower—but the asset would no longer be priced as an early-stage option. argenx is paying early to secure both control and development leadership. The relationship also did not begin with the merger agreement. In April, Forte raised approximately $150 million in a public offering, and argenx participated as a strategic investor. The public filings do not disclose how much argenx itself purchased, so the entire financing cannot be attributed to the company. These may sound like technical qualifications. They are what separate reading a transaction from filling its gaps with a story. ## 02 | VYVGART Gave argenx an Expensive Admission Ticket VYVGART is the cheque book behind the acquisition. Efgartigimod, VYVGART's active ingredient, blocks the neonatal Fc receptor, or FcRn, and accelerates the breakdown of IgG. Because abnormal IgG antibodies drive multiple autoimmune diseases, one mechanism can potentially support development across several indications instead of remaining confined to one disease. The commercial achievement is not simply that the first indication sold well. argenx developed a repeatable method around the molecule: identify patient groups with strong biological rationale, sequence indications, develop formulations, execute global trials and commercialization, and unlock the same asset layer by layer. Under its Vision 2030 plan, the company aims to treat 50,000 patients globally, secure ten labeled indications and advance five candidates into registrational development by 2030. That operating model is what makes Forte strategically attractive. argenx is not trying to copy the molecular design of VYVGART. It is trying to reproduce the development logic: identify a shared disease pathway, establish a signal in one setting, then expand one drug across multiple indications. *Figure 2. VYVGART has already proved that argenx can scale one mechanism. FB102 asks whether that success reflects a repeatable platform capability.* That is also the transaction's hardest test. If FB102 ultimately produces only modest benefit in a narrow indication, $2.2 billion will look expensive. If it repeatedly creates clinically meaningful differences across diseases driven by pathogenic T cells and natural killer cells, the asset could become a second growth engine. The acquisition price is therefore a wager on organizational repeatability as much as it is a wager on one antibody. ## 03 | Get the Biology Right: FB102 Targets the IL-2/IL-15 Axis FB102 is Forte's anti-CD122 monoclonal antibody candidate. CD122 is the beta subunit shared by the interleukin-2 and interleukin-15 receptors. It is not part of the IL-12 receptor. Getting that detail wrong sends the rest of the immunology in the wrong direction. IL-15 supports the survival and activation of natural killer cells and subsets of memory T cells. IL-2 participates in signaling for both effector T cells and regulatory T cells, or Tregs. That creates the central challenge of targeting CD122: suppress enough pathogenic T-cell and NK-cell activity to produce clinical benefit without excessively impairing the Treg populations that help maintain immune balance. Forte's design thesis is that FB102 can find that therapeutic window. In company-reported in vitro experiments, the antibody inhibited proliferation or activation of human T cells and NK cells after IL-2 or IL-15 stimulation. Under its IL-2 experimental conditions, Treg proliferation was relatively preserved. The qualification is essential. Relative Treg preservation is supported mainly by in vitro work, nonhuman-primate studies and early human evidence. It is not yet selective activity proven in large clinical trials. The data do not justify claims that FB102 attacks only harmful immune cells, spares normal immunity, or is certain to be safer than JAK inhibitors. *Figure 3. Dampening pathogenic T-cell and NK-cell signaling is the goal. Relative Treg preservation remains an early hypothesis, not a settled clinical advantage.* Early pharmacodynamic evidence tells us that the idea deserves testing. Only larger and longer randomized trials can tell us whether the proposed therapeutic window exists in patients. ## 04 | Are Two Phase 1b Signals Worth $2.2 Billion? Before argenx moved to acquire Forte, FB102 had produced early human signals in celiac disease and vitiligo. Celiac disease is an immune-mediated disorder triggered by gluten. Strict gluten avoidance remains the foundation of treatment, yet cross-contamination is difficult to eliminate in daily life, and diet alone does not fully control symptoms and intestinal injury for every patient. FB102's Phase 1b celiac study was a small randomized, double-blind, placebo-controlled trial with 32 participants: 24 received FB102 and eight received placebo. Participants received four doses at 10 mg/kg and then underwent a 16-day gluten challenge. Forte reported a mean change in the composite VCIEL histology measure of -1.849 from baseline in the placebo group versus 0.079 in the FB102 group, with p=0.0099. CD3-positive intraepithelial lymphocyte density increased by 13.3 in the placebo group and declined by 1.5 with FB102, with p=0.0035. The value of this result is that a controlled gluten challenge produced directionally consistent histology and immune-cell signals, which are more informative than a symptom questionnaire alone. Its limits are equally clear. The study enrolled only 32 people, and the challenge lasted 16 days. It cannot establish that patients would remain protected during long-term, real-world exposure to gluten. All 32 completed the day-32 biopsy, and treatment-emergent adverse events were mostly grade 1, but a small Phase 1b trial cannot exclude uncommon or long-term safety risks. The vitiligo evidence came from an early Phase 1b analysis of 43 people: 32 received FB102 and 11 received placebo. The more conservative efficacy-evaluable comparison included 32 and 10 patients, respectively. At week 24, centrally read mean improvement in the facial Vitiligo Area Scoring Index, or F-VASI, was 29.6% with FB102 and 7.9% with placebo. The placebo-adjusted difference was 21.7 percentage points, with p=0.020. F-VASI is a facial measure, not evidence of the same response across the entire body, and the result cannot be used for a direct cross-trial ranking against other medicines. In the subgroup with baseline F-VASI of at least 0.75, 17 FB102-treated patients and only four placebo patients were evaluable. Mean improvement at week 24 was 43.2% versus 0.5%. The magnitude is interesting, but a four-person placebo group makes the analysis exploratory rather than confirmatory. These two studies explain why argenx was willing to act. They do not prove that it has already found the next VYVGART. The next valuation-changing evidence is whether the celiac Phase 2 reproduces both biological and clinical benefit, whether vitiligo holds up in a larger study, whether alopecia areata or another indication points in the same direction, and whether longer follow-up preserves an acceptable immune-safety window. At the research cutoff, the celiac Phase 2 had not reported results. Forte said it expected a readout in the second half of 2026, while ClinicalTrials.gov listed an estimated primary completion date in February 2027. The two timelines are not identical. The company expectation can be reported as an expectation; it should not be presented as a guaranteed data date. *Figure 4. Celiac and vitiligo Phase 1b studies are the beginning. Replication, scale and durability still separate FB102 from a second blockbuster.* ## 05 | A Large Story Still Has Three Gates to Clear For an early-stage candidate, $2.2 billion is a substantial price. It capitalizes three outcomes that have not yet occurred. First, the Phase 1b signals must replicate. Small early studies are vulnerable to baseline imbalances, analysis-population choices and the influence of individual participants. The vitiligo subgroup, in particular, compared 17 treated patients with four placebo patients. A smaller effect in the next trial would not necessarily mean the drug has no activity, but it would materially change the commercial value investors should assign to the program. Second, the CD122 safety window must persist over time. T cells and NK cells do not operate only in autoimmune disease; they also contribute to infection defense and immune surveillance. The absence of severe events in a short early study does not establish that large, longer studies will avoid immunosuppression, infection or uncommon complications. FB102's value depends on maintaining useful selectivity at an effective dose. Third, argenx must sequence the indications well. Celiac disease has clear unmet need, but endpoints, dietary background and reimbursement are complex. Vitiligo affects a large population but already has competition from topical JAK therapies. Alopecia areata is also becoming crowded. Which disease moves first, which endpoint is used, and whether a subcutaneous formulation is developed will determine development time and capital efficiency. VYVGART's record understandably encourages confidence in argenx. Corporate capability is not a permanent pass. It may improve the odds of selecting the right molecule and indication, but it cannot answer Phase 2 and Phase 3 in advance. ## 06 | Taiwan Has Real Connections, but Not a Ready-Made “Concept Stock” Basket Taiwanese readers can locate the transaction through two genuine industry links. The first is VYVGART's existing commercialization network. Zai Lab, listed on Nasdaq as ZLAB and in Hong Kong as 9688, holds rights to efgartigimod in Greater China. Synmosa Biopharma, listed in Taiwan as 4114, obtained 15-year Taiwan commercialization rights from Zai Lab Taiwan for six therapies, including VYVGART. Synmosa's connection to argenx therefore sits in VYVGART's Taiwan commercialization ecosystem. It is not a party to the Forte acquisition and is not a disclosed Taiwan partner for FB102. The available primary sources also do not support using this article to claim that VYVGART has already been approved or launched in Taiwan. The second link is Taiwan's own drug development in vitiligo. Elixiron Immunotherapeutics, listed as 7871, is developing EI-001, or indemakitug, an anti-interferon-gamma antibody for vitiligo. TWi Biotechnology, listed as 6610, is developing the topical JAK inhibitor AC-1101 across immune dermatology indications including vitiligo and alopecia areata. Neither company has a disclosed relationship with FB102. Their relevance is comparative: Taiwan-based developers are approaching immune-driven pigment and hair disorders through different targets, modalities and development risks. Conversely, Taiwanese antibody manufacturers and contract development companies should not be labeled acquisition beneficiaries merely because FB102 is an antibody. No primary evidence links companies such as EirGenix, Mycenax or United BioPharma to this program. Capability adjacency is not a supply agreement. ## Conclusion | VYVGART Proved argenx Can Scale; FB102 Must Prove It Can Repeat The most interesting part of this transaction is not that argenx finally has enough cash to buy a company, nor that one FB102 p-value looks attractive. The deal exposes the next challenge for a successful biotech. VYVGART proved that argenx could identify a valuable immune mechanism, execute global clinical development and formulation strategy, commercialize it, and turn one molecule into a multibillion-dollar product. FB102 asks a different question: can that method leave FcRn, move to CD122, and work again? The $2.2 billion price buys admission to the examination room, not a diploma. The real dividing lines will be Phase 2 replication in celiac disease, stability of the vitiligo signal in larger trials, and a CD122 safety window that holds over longer exposure. The first blockbuster gave argenx the resources to become an acquirer. A second will determine whether it is a company with one extraordinary product or an immunology platform capable of repeatedly producing medicines. --- ## Investors, Take Note: Hair Loss Could Be Biotech's Next Weight-Loss-Drug-Sized Blockbuster URL: https://drugnews.com.tw/articles/2026-08-06-anhorn-ah001-hair-loss-blockbuster-market-en.html Date: 2026-08-06 Language: en Summary: Lilly has placed a $40 million bet on next-generation hair-loss drugs. With roughly 80 million U.S. patients still constrained by two legacy therapies, AH-001's Phase 2 data could become the next global asset-revaluation trigger. > Lilly has placed a $40 million bet on hair-loss innovation. With roughly 80 million patients in the United States and only two legacy therapies, could AnHorn's AH-001 place topical AR degradation at the center of the next global revaluation of hair-loss drug assets? The next biotech market capable of being propelled at the same time by global pharmaceutical companies, Wall Street, and consumers may not be in oncology or rare disease. It may be hiding in something people see in the mirror every day: their hair. In 2026, a series of signals began to emerge around next-generation hair-loss drugs, and the industry had never seen all of them at the same time. Phase 2/3 data for Veradermics' extended-release oral minoxidil pushed the company's market capitalization close to $5 billion. Lilly first expressed an interest in purchasing shares in the Veradermics IPO, then invested $40 million directly in Absci to secure a position around ABS-201, an AI-designed antibody for hair loss. Global pharmaceutical companies are no longer viewing hair loss only as a market for shampoos, transplants, or aesthetic services. They are beginning to treat it as a chronic prescription-drug market capable of producing blockbuster products. What makes this market remarkable is not only the roughly 80 million patients in the United States, but also the fact that most of the demand has never been unlocked. The mainstay drugs have been on the market for decades, yet patients still have to choose among concerns about sexual function and hormones, cardiovascular risk, scalp irritation, daily application, and limited efficacy. When a new drug can improve efficacy, safety, and convenience at the same time, the upside is not limited to taking share from older therapies. It can bring people who never started treatment back into the market. Taiwan-based AnHorn Medicines and its AH-001 program are already at the most important table to watch. AH-001 is not another minoxidil, nor is it a conventional androgen blocker. It is a topical small-molecule AR protein degrader (PROTAC) designed to directly remove the androgen receptor in hair follicles. The program has already cleared the U.S. Phase 1 human safety threshold. The next Phase 2 readout will determine whether it can become a transaction-ready asset at the center of the global revaluation of hair-loss drugs. > **CORE INVESTMENT THESIS** > > The upside in next-generation hair-loss drugs will not come only from switching patients off legacy therapies. It will come from bringing back the much larger population that never started treatment, discontinued it, or avoided it altogether. If a product can improve efficacy, safety, and convenience at the same time, low penetration could expand into a $10 billion-scale prescription market. ## 01 | Why Hair-Loss Drugs Have the Same Commercial Setup That Fueled the Weight-Loss Boom Hair-loss drugs already have a commercial structure that closely resembles the weight-loss drug market just before its breakout: a very large patient population, a condition that requires long-term management, results that can be seen directly, patients who actively seek treatment, and a treatment penetration rate that has been held down for years by side effects and inconvenience. In its U.S. securities filings, Veradermics estimated that approximately 80 million Americans have pattern hair loss and that roughly 74 million fall within a commercially addressable population. Yet only about 15 million are currently treated on an ongoing basis. Another 59 million are untreated, have discontinued therapy, or remain caught between shampoos, supplements, aesthetic procedures, and intermittent prescription use. The company estimated that the currently treated population alone represents an annual commercial opportunity of about $9 billion.[1] The more important point is that demand is not absent; it is being held back by the limitations of the available products. Veradermics' patient research found that only about 20% of patients actively pursue treatment, and among those who are treated, only about 9% are satisfied with the results. In other words, the largest source of growth is not simply switching a finasteride user to another drug. It is unlocking the large population that remains untreated, has stopped treatment, or is unwilling to use the available options.[1] | Key Metric | Market Estimate | Investment Relevance | | --- | --- | --- | | U.S. patient population | Approximately 80 million | A patient base large enough to support a major chronic prescription market | | Commercially addressable population | Approximately 74 million | The opportunity extends far beyond a small group of severe patients | | Currently treated | Approximately 15 million | Existing annual commercial opportunity estimated at roughly $9 billion | | Untreated / lapsed / other | Approximately 59 million | The largest upside comes from reactivating suppressed demand | | Treatment satisfaction | Approximately 9% | A materially better product profile could re-rate the entire category | Table 1 | The hair-loss market is not short of demand. Treatment penetration and satisfaction have remained low for years. Source: Veradermics S-1. ## 02 | What the Market Underestimates Is Not the Patient Count, but the Demand That Has Never Been Unlocked The two legacy mainstays are not entirely ineffective, but nearly every treatment path requires patients to compromise. Topical minoxidil must be applied every day, and its stickiness, scalp irritation, drying time, and interference with hairstyling can erode adherence. Oral finasteride works by lowering DHT, but long-standing concerns about sexual function, mood, and use in women of childbearing potential continue to affect prescribing and patient willingness. Low-dose oral minoxidil is widely used off-label in dermatology, but it is still fundamentally a cardiovascular drug, requiring physicians and patients to manage edema, palpitations, blood pressure, and other systemic risks. These limitations create an unusual market dynamic: patients are willing to spend, but many are unwilling to remain on the products currently available. In the United States, patients may spend anywhere from several hundred dollars to more than $10,000 a year across supplements, lasers, platelet-rich plasma, hair transplantation, and prescription drugs. Yet the market still lacks a standard product that combines visible efficacy, long-term safety, convenience, and sustainable use.[1] Under a conservative scenario, if a new drug brought back only 5% of the 59 million potential users, the market would gain nearly 3 million new patients. At an annual drug cost of $1,200 to $1,800 per patient, that would create approximately $3.5 billion to $5.3 billion in additional annual U.S. market opportunity. This illustrates why, once side-effect and adherence barriers are reduced, the ceiling for the hair-loss market is not today's prescription volume. It is the population that has not yet been activated. Women represent another underestimated source of growth. Roughly 30 million women in the United States have female pattern hair loss, but hormonal risks, pregnancy concerns, and the limited number of available prescription therapies have left the market chronically underdeveloped. Any new product that can demonstrate low systemic exposure, a nonhormonal profile, or a more practical path to long-term use could raise the market ceiling again. Taken together, these conditions closely resemble the weight-loss drug market before it took off. The same pre-breakout setup may be taking shape again. ## 03 | Wall Street Has Already Voted: One Hair-Loss Drug Supported a Valuation Near $5 Billion In 2026, Veradermics gave capital markets a direct answer to a simple question: should next-generation hair-loss drugs be treated as a major biotech category? The company went public in February at $17 per share, with an IPO valuation of roughly $600 million. After it reported positive results from VDPHL01 Study 302 on April 27, the stock gained about 48% in a single session, and the company's market value later approached $5 billion.[2][3] Study 302 enrolled 519 men with mild-to-moderate androgenetic alopecia. After six months, target-area non-vellus hair count increased by 30.3 hairs/cm2 in the once-daily group and 33.0 hairs/cm2 in the twice-daily group, compared with 7.3 hairs/cm2 for placebo. Using the stricter patient assessment of "improved" or "much improved," response rates were 48.4% and 62.9% in the two active arms, versus 13.4% for placebo.[2] VDPHL01 did not create new biology. It simply reformulated minoxidil as an extended-release oral drug, changing peak blood concentration, exposure time, and dosing convenience. Yet capital markets were willing to assign it a multibillion-dollar valuation because it demonstrated one point: when the product profile is strong enough, a hair-loss drug can move beyond the consumer and aesthetic categories and become a major prescription-drug asset. Figure 1 | Study 302 showed separation in both objective hair counts and patient-reported outcomes. Results from different clinical trials should not be compared directly. Source: Veradermics. ## 04 | Lilly Has Entered the Market: From Olumiant and Veradermics to ABS-201 The more important signal for investors is that Eli Lilly, the leader in weight-loss drugs, has entered the hair-loss market directly. Few global pharmaceutical companies understand better than Lilly how to turn chronic-disease and consumer-health demand into a large market. Lilly is now building multiple layers of exposure to hair loss. The first is Olumiant (baricitinib), which is FDA-approved for adults with severe alopecia areata. That means Lilly is not entering hair loss for the first time; it already has clinical, regulatory, and commercialization experience in an immune-mediated hair-loss indication.[4] The second appeared in Veradermics' 2026 IPO filing, where Lilly publicly indicated an interest in purchasing up to approximately 4.9% of the post-offering shares. The indication was non-binding, but it still reflected Big Pharma's interest in a late-stage oral asset for pattern hair loss.[5] The third layer was the most concrete. In June 2026, Absci completed a $100 million equity offering, with 5,398,111 shares, or approximately $40 million, specifically allocated to Lilly. The proceeds are intended to support ABS-201, a long-acting, AI-designed antibody targeting the prolactin receptor (PRLR) for indications that include androgenetic alopecia and endometriosis.[6] ABS-201 has entered the Phase 1/2a HEADLINE study, with initial human proof-of-concept data expected in the second half of 2026.[7] The investment is not a license and it is not an acquisition, but it effectively gives Lilly a seat close to the data and the team. When the company that reshaped the obesity and metabolic-disease markets with Zepbound and Mounjaro begins monitoring both an oral hair-loss drug and a novel biologic, the market should no longer view hair loss as a niche aesthetic category. It should view it as a pool of potential blockbuster assets that major pharmaceutical companies may pursue. | Asset / Timing | Lilly Action | Strategic Meaning | | --- | --- | --- | | Olumiant / alopecia areata | FDA-approved for adults with severe alopecia areata | Existing clinical, regulatory, and commercial experience in hair loss | | Veradermics IPO | Indicated non-binding interest in up to approximately 4.9% of post-offering shares | Monitoring a late-stage oral asset for pattern hair loss | | Absci / ABS-201 | $40 million of shares specifically allocated to Lilly | A strategic equity position around a novel PRLR antibody mechanism | Table 2 | Lilly's hair-loss strategy now spans a commercialized product, a late-stage oral asset, and an early-stage biologic with a new mechanism. ## 05 | The Next Wave Could Be Triggered by Three New Treatment Classes at Once Hair-loss drug development is moving beyond two legacy drugs and into three paths that could establish new standards of care. The first is new engineering applied to an old drug: VDPHL01 uses extended-release oral minoxidil to improve efficacy and the risk-benefit profile. The second is new biology: ABS-201 uses a long-acting PRLR antibody in an effort to change the hair-follicle cycle. The third is the localized, precise control of androgen signaling through topical AR antagonists and AR degraders. Figure 2 | Global Competitive Landscape for Hair-Loss Drugs. Source: publicly available information and company websites. The key catalysts over the next 12 to 24 months are clear. Veradermics' 52-week Study 304 must show that efficacy can be reproduced and maintained, while Study 306 in women will help determine the commercial ceiling. Human proof-of-concept data for ABS-201 will show whether a long-acting injectable biologic can truly promote hair growth. KX-826, GT20029, and other topical antiandrogen programs will continue to educate regulators and physicians that localized control of androgen signaling in the hair follicle is a clinically developable path.[11][12] Success in any one of these approaches would expand the entire category, not just benefit one company. The weight-loss drug market did not break open because the first product monopolized every patient. Semaglutide, tirzepatide, and later oral, long-acting, and combination products collectively expanded patient education, reimbursement, and treatment capacity. Next-generation hair-loss drugs are moving toward a similar multi-product expansion phase. ## 06 | AH-001's Distinctive Position: At the Intersection of a New Mechanism, Topical Delivery, and Scalability In this new competitive landscape, AnHorn's AH-001 is a potential blockbuster competitor worth watching. It sits at a rare product intersection: topical delivery, a small-molecule format, and direct degradation of the androgen receptor. It has also completed a U.S. Phase 1 study with favorable company-reported safety and tolerability results. This early profile supports further study of whether topical delivery can limit systemic exposure, but it does not yet establish freedom from the adverse effects associated with oral therapies. AH-001 does not lower DHT upstream, as finasteride does, and it does not simply compete for the receptor site like a conventional antagonist. It uses the ubiquitin-proteasome system to tag the androgen receptor (AR) for clearance and send it to the proteasome for degradation. Conventional inhibitors must continue occupying the receptor. Protein degraders are designed to work through an event-driven process: after effective recruitment, they directly reduce the amount of AR protein inside the cell. Its product logic also differs from Lilly-backed ABS-201. ABS-201 is a long-acting injectable antibody that may trade less frequent dosing for sustained activity. AH-001 is a patient-applied topical small molecule that could offer advantages in manufacturing cost, distribution, home use, and localized exposure. If it can achieve low irritation, rapid absorption, and less frequent dosing, it would more closely resemble an everyday dermatology product that can be scaled broadly. Figure 3 | Hair-loss treatment is moving from follicle stimulation and androgen blockade toward localized degradation of the androgen receptor. ClinicalTrials.gov shows that the U.S. Phase 1 study of AH-001 enrolled 64 participants and evaluated single and multiple doses at 0.2%, 0.5%, 1%, and 2%. AnHorn reported favorable safety and tolerability across all dose levels, with no drug-related adverse events observed.[8][9] This does not mean that hair growth has been proven, but it moves AH-001 from a concept molecule into the small global group of topical AR-degradation assets with U.S. human data. ## 07 | Why AH-001 Could Create Major Value: The Key Is the Product Profile, Not AI For large pharmaceutical companies, the most valuable candidate is often not the one with the most sophisticated mechanism. It is the one that strikes the best balance among clinical benefit, long-term safety, convenience, and commercial cost. AH-001's potential advantage is that it has room to differentiate across all four dimensions. First, if localized AR degradation can produce deeper or more durable signaling suppression than receptor blockade alone, AH-001 could establish a new efficacy category. Second, if systemic exposure remains very low, it could appeal to younger men and some women who are reluctant to use oral hormonal drugs. Third, a topical small-molecule formulation is generally easier than an injectable antibody to bring into a large self-pay market. Fourth, follicle stimulation and AR reduction act at different points in the pathway, leaving open the possibility of combining AH-001 with minoxidil to broaden patient coverage. Phase 2 must now establish three connected layers of evidence: that AR is actually reduced in the scalp or hair follicle; that target-area hair count, standardized photography, and patient-reported outcomes improve; and that these effects occur while maintaining low systemic exposure, good local tolerability, and a profile suitable for long-term use. If all three layers are established, AH-001 could move from a technology demonstration to a blockbuster transaction asset in the eyes of Big Pharma. ## 08 | AnHorn's Second Opportunity: One Hair-Loss Drug Could Validate the Entire Drug-Development Platform AH-001 is more than a potential blockbuster for pattern hair loss. It is also the first full human test of AnHorn's AIMCADD® x Targeted Protein Degradation platform. From disease and target selection, AI molecular design, wet-lab screening, formulation and toxicology, through the U.S. IND and Phase 1, AnHorn has moved multiple candidates into human development. If AH-001 begins to show signs of success, the value of the company's other assets could rise with it. In June 2026, AnHorn's second asset, AH-008, received FDA IND clearance and Taiwan CDE Index Case designation.[10] AH-008 is being developed to prevent chemotherapy-induced neuropathy, an addressable market with more than $10 billion in potential because its scope could extend across chemotherapy regimens worldwide. It may also expand into other neuropathic pain indications; even an early sign of success could position it as a potential megablockbuster. AH-008 and AH-001 address different diseases through different mechanisms, so progress in one cannot be treated as proof that the other will succeed. But the arrival of a second clinical asset gives the market a way to judge whether AnHorn is a company built around a single hair-loss program or an AI drug-development platform capable of repeatedly generating clinical assets. That distinction matters to investors because it sets AnHorn apart from a conventional single-asset biotech. One early-stage asset may support one licensing transaction. A platform that can repeatedly generate clinically testable assets can support regional licenses, co-development agreements, pipeline options, and even multi-asset partnerships centered on a specific E3 ligase or disease area. This year has already produced several transactions with potential values above $10 billion, including Innovent-Takeda at $11.2 billion, CSPC-AstraZeneca at $18.5 billion, and Hengrui-Bristol Myers Squibb at $15.2 billion. If AH-001 succeeds in Phase 2, AnHorn would gain more than the value of one hair-loss drug. It would gain a credibility premium for the entire platform. Figure 4 | AH-001's second layer of value is the human validation it could provide for AnHorn's AIMCADD® x protein-degradation drug-development flywheel. ## 09 | What Should Investors Watch? Three Milestones Will Determine the Next Revaluation The first milestone is the Phase 2 design. Can the selected doses produce a clear response? Does the study collect target-area hair count, standardized imaging, patient-reported outcomes, and target-engagement data at the same time? A trial designed only to achieve a statistically significant result may fail to show why the product is differentiated, leaving licensing value constrained. The second milestone is productization. Odor, drying time, residue, redness, itching, and dosing frequency will all be magnified in real-world use. Hair loss is a chronic market that may require years of treatment. No matter how strong the mechanism, a poor user experience will make it difficult to reproduce the high persistence and brand loyalty seen with GLP-1 drugs. The third milestone is the partnership model. Lilly's $40 million position in ABS-201 and Veradermics' near-$5 billion market value show that capital is creating new valuation reference points for hair-loss drugs that are both effective and scalable. Once AH-001 produces human efficacy and mechanism data, the key question will be whether AnHorn can convert that data package into a global license, a co-development agreement, or a strategic investment. The risks are equally clear. AH-001 has not yet demonstrated hair regrowth in humans. Full Phase 1 results have not been published in a peer-reviewed journal. Competitors with related mechanisms are further ahead in development, and a female indication would require separate reproductive-toxicology work and clinical validation. Investors should follow the data rather than assume that any novel mechanism is destined to succeed. | Value Milestone | Evidence the Market Needs | Potential Revaluation | | --- | --- | --- | | Phase 1 complete | Safety, tolerability, PK, and low systemic exposure | From a concept molecule to a clinically developable option | | Positive Phase 2 | TAHC, imaging, patient-reported outcomes | Asset-level proof of concept and entry into global licensing or strategic-investment discussions | | Partnership / Phase 3 | Reproducible efficacy, long-term tolerability, and a commercial formulation | Expansion from single-asset value toward platform and global commercial value | Table 3 | AH-001 will not be revalued in a single step. Value will build sequentially through safety, proof of concept, and commercial partnering. ## CONCLUSION | Lilly Has Bought Its Ticket. AnHorn Is Waiting for the Readout That Could Ignite the Market Next-generation hair-loss drugs could become the next market on the scale of weight-loss drugs, not because hair loss and obesity have the same medical severity, but because they share the same market multipliers: a very large patient base, chronic long-term treatment, highly visible results, patients willing to pay out of pocket, and low market penetration that has been suppressed for years by the side effects and inconvenience of legacy products. Veradermics has already shown that Wall Street is willing to value a hair-loss asset in the billions of dollars. Lilly's progression from Olumiant, to its indicated interest in Veradermics, to a $40 million investment in Absci shows that Big Pharma is building strategic options across the category. Lilly's entry also raises expectations for what a global pharmaceutical company can do: build a true megablockbuster, just as the weight-loss drug market was pushed to its current peak. The opportunity for AH-001 in the next stage is to prove that topical small-molecule AR degradation can deliver hair growth that patients can see, safety that physicians can trust, and a product profile that a global pharmaceutical company can scale. If it can do that, AH-001 could become one of the few Taiwan-origin clinical assets with a credible position in a global clinical opportunity at the $10 billion scale. AnHorn would also demonstrate the strength of a platform capable of repeatedly creating blockbuster drugs, and Taiwan's capital market could be looking at a genuine AI-drug-discovery breakout company. ### IMPORTANT DISCLOSURE This article is provided solely for industry research and market observation. It does not constitute investment, trading, medical, fundraising, or securities advice. All drug candidates discussed remain subject to clinical, regulatory, partnering, commercialization, and competitive risks. Market scenarios are illustrative and do not represent sales forecasts. --- ## 投資人請注意，下一個減肥藥級別賽道來了！ URL: https://drugnews.com.tw/articles/2026-08-06-anhorn-ah001-hair-loss-blockbuster-market.html Date: 2026-08-06 Language: zh-Hant Summary: 禮來用 4,000 萬美元卡位掉髮新藥；當美國約 8,000 萬名患者仍受限於兩款老藥，安宏 AH-001 的二期數據可能成為下一次全球資產重估關鍵。 > 禮來用 4,000 萬美元卡位：當 8,000 萬名美國患者只有兩款老藥，安宏 AH-001 為什麼可能站在掉髮新藥重估中心？ 如果要找下一個可能被全球藥廠、華爾街與消費者同時推動的生醫賽道，答案未必還在癌症或罕病，也可能藏在每個人每天照鏡子都看得見的地方：頭髮。 2026 年，掉髮新藥開始出現一連串過去從未同時發生的訊號。Veradermics 的一顆緩釋口服 minoxidil，靠 Phase 2/3 數據把公司市值推向近 50 億美元；禮來先在 Veradermics IPO 文件中表達認購意向，接著又以 4,000 萬美元實際買進 Absci 股份，卡位 AI 設計的掉髮抗體 ABS-201。全球大藥廠已不再只把掉髮當成洗髮精、植髮或醫美服務，而是開始把它當成一個可以誕生 blockbuster 的慢性處方藥市場。 這個市場真正驚人的地方，不只是美國有約 8,000 萬名患者，而是大多數需求根本還沒有被釋放。現有藥物上市數十年，患者仍必須在性功能與荷爾蒙疑慮、心血管風險、頭皮刺激、每日塗抹與效果有限之間做選擇。當一款新藥能把療效、安全性與便利性同時往前推，增量不只是搶走舊藥市占，而是讓原本不治療的人重新走進市場。 而台灣的安宏生醫 AH-001 已經在最值得關注的牌桌上。它不是另一款 minoxidil，也不是傳統的雄激素阻斷劑，而是一款外用小分子 AR protein degrader（PROTAC），試圖直接把毛囊中的雄激素受體「拆掉」。一期已經跨過美國人體安全門檻；下一張二期臨床的成績單，將決定它究竟能否站上全球掉髮新藥重估中心的交易資產。 > **核心投資邏輯** > > 掉髮新藥的增量，不只來自舊藥換藥，而是把未治療、停藥與不敢用藥的人重新帶回市場。只要產品同時改善療效、安全與便利性，低滲透率就可能轉化為百億美元級處方市場。 ## 01｜為什麼說掉髮新藥有「減肥藥級別」的商業爆發結構？ 掉髮藥其實早已具備一套接近減肥藥爆發前夜的商業結構——患者人數極大、疾病需要長期管理、療效可以直接被肉眼看見、患者願意主動尋找治療，而且現有藥物的副作用與使用不便，長期壓抑了真正的治療滲透率。 Veradermics 在美國證券文件中估算，美國約有 8,000 萬名 pattern hair loss（型態性落髮）患者，其中約 7,400 萬人屬於可觸及的商業人群；但目前真正持續治療者約 1,500 萬人，仍有約 5,900 萬人沒有治療、已經停藥，或只停留在洗髮精、保健品與醫美程序之間。公司估算，僅現有治療人群就形成約 90 億美元的年度商業機會。[1] 更關鍵的是，這不是一個需求不存在的市場，而是一個需求被產品限制壓住的市場。Veradermics 的患者研究顯示，整體只有約兩成患者積極接受治療，而接受治療者中只有約 9% 對結果滿意。換句話說，掉髮市場最大的增量，不只是把使用 finasteride 的患者換到另一顆藥，而是把龐大的未治療、停藥與不敢用藥人口重新釋放出來。[1] | 關鍵指標 | 市場數字 | 投資意義 | | --- | --- | --- | | 美國患者 | 約 8,000 萬 | 患者基數足以支撐大型慢性處方市場 | | 可觸及人群 | 約 7,400 萬 | 不是只服務少數重度患者 | | 目前治療者 | 約 1,500 萬 | 現有年度商業機會估計約 90 億美元 | | 未治療／其他 | 約 5,900 萬 | 新藥最大增量來自重新啟動需求 | | 治療滿意度 | 約 9% | 產品輪廓改善可帶來類別重估 | 表 1｜掉髮新藥市場不是需求不足，而是治療滲透率與滿意度長期偏低。資料來源：Veradermics S-1。 ## 02｜真正被低估的，不是患者數，而是「沒有被釋放的需求」 現有兩款主流老藥並非完全無效，但幾乎每一條路都要求患者妥協。外用 minoxidil 需要每天塗抹，黏膩、頭皮刺激、乾燥時間與造型衝突會侵蝕依從性；口服 finasteride 透過降低 DHT 發揮作用，卻長期伴隨性功能、情緒與育齡女性使用限制等疑慮；低劑量口服 minoxidil 在皮膚科被大量 off-label 使用，但它本質上仍是心血管藥物，醫師與患者必須管理水腫、心悸與血壓等風險。 這些限制形成一個很特殊的市場現象：患者願意花錢，卻不願意長期使用現有產品。美國患者目前可能在保健品、雷射、PRP、植髮與處方藥之間支出每年數百到上萬美元，但仍缺少一個同時兼顧明顯療效、長期安全、使用便利與可持續性的標準產品。[1] 若用保守情境試算：只要約 5% 的 5,900 萬名潛在人群被新藥喚回，便是接近 300 萬名新增使用者；若每人每年藥費為 1,200 至 1,800 美元，就可能新增約 35 億至 53 億美元的美國年度市場。這說明當副作用與依從性障礙被改善，掉髮市場的天花板並不在今天的處方量，而在尚未被啟動的人口。 女性更是下一個被低估的增量。美國約有 3,000 萬名女性型態性落髮患者，但荷爾蒙風險、妊娠顧慮與可用處方藥不足，使女性市場長期沒有被完整開發。任何能證明低系統暴露、非荷爾蒙或更友善長期使用的新產品，都有機會把市場再往上推一層。 這些種種都與當年的減肥藥市場有異曲同工之妙，減肥藥的爆發前夜似曾相似場景似乎又要來臨。 ## 03｜華爾街已經先投票：一顆掉髮藥，撐起近 50 億美元估值 2026 年的 Veradermics，讓資本市場第一次用非常直接的方式回答：掉髮新藥值不值得被當成大型生技賽道？ 公司在 2 月以每股 17 美元上市，IPO 定價時估值約 6 億美元；4 月 27 日公布 VDPHL01 Study 302 正向數據後，股價單日再上漲約 48%，其後市值一度接近 50 億美元。[2][3] Study 302 納入 519 名輕至中度男性患者。治療六個月後，VDPHL01 每日一次與每日兩次組的目標區非絨毛毛髮計數，分別較基線增加 30.3 與 33.0 根/cm²，安慰劑組為 7.3 根/cm²；採較嚴格的患者自評「改善或明顯改善」，兩個治療組為 48.4% 與 62.9%，安慰劑僅 13.4%。[2] VDPHL01 並沒有創造全新的生物學，它只是把 minoxidil 做成緩釋口服製劑，重新配置血中峰值、暴露時間與用藥便利性。可是資本市場願意給出數十億美元估值，因為它證明一件事：只要產品輪廓足夠好，掉髮藥可以從消費品與醫美題材，升級為大型處方藥資產。 圖 1｜Study 302 同時在客觀毛髮計數與患者自評建立差異；不同試驗不可直接跨試驗比較。資料來源：Veradermics。 ## 04｜禮來已經下場：從 Olumiant、Veradermics 到 ABS-201 更值得投資人注意的訊號，減肥藥巨頭 禮來 已經親自下場卡位掉髮新藥市場！ 全球最懂得慢性疾病與消費醫療需求做成大市場的禮來，已經開始在掉髮領域建立多層次卡位。 第一層，禮來的 Olumiant（baricitinib）已獲 FDA 核准治療成人嚴重圓禿，代表禮來不是第一次進入掉髮市場，而是已經具備免疫性掉髮的臨床、監管與商業化經驗。[4] 第二層，在 Veradermics 2026 年 IPO 文件中，禮來曾公開表示有意認購上市後最高約 4.9% 的股份；這項意向並非具約束力，但已反映大藥廠對晚期型態性落髮資產的關注。[5] 第三層，也是最具體的一步：2026 年 6 月，Absci 完成 1 億美元增資，其中 5,398,111 股、合計 4,000 萬美元明確配置給禮來。資金將推進 ABS-201，一款由 AI 設計、鎖定 prolactin receptor（PRLR）的長效抗體，適應症包括雄性素性脫髮與子宮內膜異位症。[6] ABS-201 已進入 Phase 1/2a HEADLINE 試驗，預計在 2026 年下半年取得初步人體概念驗證資料。[7] 這筆投資不是授權，也不是收購，但它等於讓禮來先買到一張靠近資料與團隊的入場券。當一家靠 Zepbound 與Mounjaro 重塑減重與代謝慢性病市場的公司，開始同時觀察口服掉髮藥與新機制生物製劑，市場就不應再把掉髮視為小眾醫美題材，而應該把它視為可能巨頭獵取重磅新藥資產池。 | 時間／資產 | 禮來動作 | 戰略含義 | | --- | --- | --- | | Olumiant／圓禿 | 已取得成人嚴重圓禿適應症 | 已具備掉髮領域的臨床、監管與商業化經驗 | | Veradermics IPO | 曾表示有意認購上市後最高約 4.9% 股份（非具約束力） | 觀察晚期口服型態性落髮資產 | | Absci／ABS-201 | 4,000 萬美元股份明確配置給禮來 | 以策略股權取得新機制 PRLR 抗體的場內座位 | 表 2｜禮來的掉髮布局已從商業化產品，延伸至晚期口服資產與早期新機制生物製劑。 ## 05｜下一輪引爆點，會從三種「新治療類別」同時出現 掉髮新藥正在從兩款老藥，分裂成三條可能建立新標準的路線。第一條是老藥新工程：VDPHL01 用緩釋口服 minoxidil 改善療效與風險收益比；第二條是新生物學：ABS-201 嘗試用長效 PRLR 抗體改變毛囊週期；第三條則是局部精準處理雄激素訊號，包括外用 AR antagonist 與 AR degrader。 圖 2｜全球掉髮新藥競爭地圖。資料來源：公開資訊與公司官網。 未來 12 至 24 個月的引爆點非常清楚。Veradermics 的 52 週 Study 304 要驗證療效能否複製並維持，女性 Study 306 則決定商業天花板；ABS-201 的人體概念驗證，將回答長效注射生物製劑能否真正促進毛髮生長；KX-826、GT20029 與其他外用抗雄激素產品，則持續教育監管與醫師：毛囊局部處理雄激素訊號，是一條可以正式開發的臨床路徑。[11][12] 任何一條路成功，都會擴大整個賽道，而不只是單一公司受益。減肥藥市場真正爆發，也不是因為第一款產品壟斷所有患者，而是 semaglutide、tirzepatide 與後續口服、長效與複方產品共同把患者教育、支付與醫療供給做大。掉髮新藥正在走向相似的多產品擴張階段。 ## 06｜安宏 AH-001 的特殊地位：站在「新機制、局部用藥、可規模化」交叉點 在這張新競爭版圖中，安宏生醫的 AH-001 是值得關注的潛在重磅競爭者。他站在一個少見的產品交叉點：外用、小分子、直接降解雄激素受體，而且在已完成美國人體一期與展現高度安全成果。 AH-001 不像 finasteride 在上游降低 DHT，也不像傳統 antagonist 只在受體位置競爭。它利用 ubiquitin-proteasome system，讓細胞把 androgen receptor（AR）貼上清除標記並送入蛋白酶體分解。傳統抑制劑需要持續占住受體；蛋白降解劑追求的是事件驅動——完成一次有效招募後，直接降低細胞內 AR 蛋白量。 它與禮來支持的 ABS-201 也有不同產品邏輯。ABS-201 是長效注射抗體，可能以較低頻率給藥換取持久效果；AH-001 則是患者可自行使用的外用小分子，理論上具有製造成本、分銷、居家使用與局部暴露上的優勢。若能做到低刺激、快速吸收與較低給藥頻率，它更接近皮膚科能大規模推廣的日常產品。 ClinicalTrials.gov 顯示，AH-001 美國 Phase I 實際納入 64 人，以 0.2%、0.5%、1% 與 2% 進行單次與多次給藥；安宏公告各劑量安全性與耐受性良好，未觀察到藥物相關不良事件，大幅避免原本口服藥物常見的副作用。[8][9] 這並不等於已證明生髮，但它使 AH-001 從概念分子，進入全球少數已有美國人體資料的局部 AR 降解資產行列。 圖 3｜掉髮治療由刺激毛囊、阻斷雄激素訊號，進一步走向局部降解雄激素受體。 ## 07｜為什麼 AH-001 有機會取得巨大價值？關鍵不是「AI」，而是產品輪廓 對大型藥廠而言，最有價值的候選藥物往往不是機制最花俏，而是能在臨床效益、長期安全、便利性與商業成本之間形成最佳平衡。AH-001 的潛在優勢，正是在這四個維度同時存在想像空間。 第一，若局部 AR 降解能比單純阻斷帶來更深或更持久的訊號抑制，AH-001 可能在療效上建立新類別；第二，若血中暴露維持極低，便可能吸引不願使用口服荷爾蒙藥物的年輕男性與部分女性人群；第三，小分子外用製劑通常比抗體注射更容易進入大規模自費市場；第四，刺激毛囊與降低 AR 位於不同節點，未來仍存在與 minoxidil 組合使用、擴大患者覆蓋的可能。 接下來是 Phase II 能否建立三層連續證據：頭皮或毛囊中的 AR 確實被降低；TAHC、標準化攝影與患者自評確實改善；同時維持低系統暴露、良好局部耐受與可長期使用。三層證據若同時成立，AH-001 將從技術展示升級為大藥廠眼中的重磅交易資產。 ## 08｜安宏的第二層機會：一款掉髮藥，可能驗證整個成藥平台 AH-001 不僅僅是型態性落髮的潛在重磅炸彈。它也是安宏 AIMCADD® × Targeted Protein Degradation 平台第一張完整的人體成績單：從疾病與靶點選擇、AI 分子設計、濕實驗篩選、製劑與毒理，到美國 IND 與 Phase I，安宏已經把多個候選物真正推進人體，而一旦AH-001成功跡象逐步顯現，其他的資產價值也將水漲船高。 2026 年 6 月，安宏第二項資產 AH-008 取得美國 FDA IND clearance，並獲台灣 CDE Index Case designation。[10] AH-008申請的適應症為預防化療引起的周邊神經性病變，更是一個注定為百億美元級別以上的市場，全球的化療藥物都將進入其市場範圍空間想像力巨大，此外他亦有極高的可能性再次擴展到其他神經痛領域，一但有成功苗頭出現必將成爲超重磅新藥。 當然AH-008 與 AH-001 的疾病和機制不同，不能直接視為複製成功；但第二個臨床資產出現，使市場開始有機會判斷：安宏是一家只有單一掉髮題材的公司，還是一個可以持續產生臨床資產的 AI製藥技術平台。 在投資人心中，這會讓安宏與一般傳統生技公司有所差異。過往單一早期資產只能談一筆授權；然而能持續產生可驗證臨床資產的平台，則可以談區域授權、共同開發、管線選擇權，甚至針對特定 E3 ligase 或疾病領域建立多資產合作（今年多起百億美元級別交易案，如信達x武田112億美元, 石藥x阿斯特捷利康185億美元, 恆瑞x 必治妥施貴寶 152億美元）。AH-001 若在 Phase II 成功，安宏得到的不只是一款掉髮藥的價值，而是整個平台的可信度溢價。 圖 4｜AH-001 的第二層價值，是讓 AIMCADD® × Protein Degradation 成藥飛輪取得人體驗證。 ## 09｜投資人應該看什麼？三個節點決定下一次價值重估 第一個節點是 Phase II 設計。劑量是否能形成清楚反應？是否同時收集 TAHC、標準化影像、患者自評與 target engagement？試驗若只追求一個統計顯著數字，無法回答產品差異，授權價值仍會受限。 第二個節點是產品化能力。外用藥的氣味、乾燥速度、殘留、紅癢與給藥頻率，都會在真實世界放大。掉髮是需要多年使用的慢性市場，再強的機制，若輸在使用體驗，也無法複製 GLP-1 的高續用與品牌黏著。 第三個節點是合作模式。禮來用 4,000 萬美元卡位 ABS-201、Veradermics 市值一度接近 50 億美元，說明資本正在為「有效且可規模化的掉髮新藥」建立新估值座標。AH-001 一旦取得人體療效與機制證據，能否把資料包轉成全球授權、共同開發或策略投資，將成為下一步重估的關鍵。 當然，風險同樣明確：AH-001 仍未證明人體生髮效果；一期完整結果尚未以同儕審查形式公開；同機制附近已有競爭者領先；女性適應症仍需獨立生殖毒理與臨床驗證。投資人應該追蹤數據，而不是把任何新機制直接當成成功。 | 價值節點 | 市場要看的證據 | 可能帶來的重估 | | --- | --- | --- | | Phase I 完成 | 安全性、耐受性、PK／低系統暴露 | 從概念分子變成可進一步開發的臨床選擇權 | | Phase II 正向 | TAHC、影像、PRO | 資產 proof-of-concept，進入全球授權與策略投資討論 | | 合作／Phase III | 可複製療效、長期耐受、商業製劑 | 從單一資產估值走向平台與全球商業價值 | 表 3｜AH-001 的價值重估不會一次完成，而是由安全性、概念驗證與商業合作逐層推進。 ## 結語｜禮來已經買票，安宏等待下一張真正能引爆市場的成績單 掉髮新藥有機會成為下一個「減肥藥級別」賽道，不是因為掉髮與肥胖具有相同醫療嚴重度，而是兩者共享同一套市場放大器：龐大患者基數、慢性長期用藥、結果高度可視化、患者願意主動支付，以及一個被舊產品副作用與不便壓抑多年的低滲透市場。 Veradermics 已經證明，掉髮資產可以被華爾街定價到數十億美元；禮來從 Olumiant、Veradermics 認購意向到 4,000 萬美元投資 Absci，則證明大藥廠正在建立自己的賽道選擇權。裡來的入局，我們更能夠期待的是大藥廠打造超級爆款的能力！就像當初的減肥藥一路被推向如今的巔峰狀態！ 安宏 AH-001 的機會是在下一階段證明：外用小分子 AR 降解，能否同時做到患者看得見的生髮、醫師信得過的安全性，以及大藥廠可以全球放大的產品輪廓，AH-001將成為台灣少數真正站上全球百億美元級臨床資產，而安宏將證明其持續打造重磅新藥的技術平台實力，台灣資本市場將迎來一家真正的的AI製藥超新星。 ### 重要聲明 本文僅供產業研究與市場觀察，不構成投資、買賣、醫療、募資或個股建議。文中候選藥物仍涉及臨床、法規、授權、商業化與競爭風險；所有市場情境試算均非銷售預測。 --- ## More Than $1 Billion in One Quarter: Why Did Alnylam Plunge 28%? URL: https://drugnews.com.tw/articles/2026-08-05-alnylam-amvuttra-ttr-growth-reset-en.html Date: 2026-08-05 Language: en Summary: Amvuttra surpassed $1 billion in quarterly sales, yet Alnylam cut its TTR guidance and lost 28% in one day. The reset was not about RNAi technology failing; it was about pent-up demand normalizing, revenue concentration and the timing gap before a second commercial pillar arrives. *Figure 1. Alnylam and Amvuttra: TTR growth shifts gears.* --- If a drug generates more than $1 billion in quarterly sales and grows 106% year over year, while its maker delivers $231 million in GAAP operating income in the same quarter, what could the market possibly be seeing when the stock still falls 28.5% in a single day? On July 30, RNA interference leader Alnylam reported its financial results for the second quarter of 2026. Product net revenues reached $1.172 billion, up 74% year over year. Amvuttra alone generated $1.012 billion for the quarter, while total TTR product revenues came to $1.030 billion. This was not a case of earnings falling apart. Yet on the same day, the company lowered its 2026 guidance for TTR product revenues from $4.4 billion-$4.7 billion to $4.2 billion-$4.5 billion. At the midpoint, the market lost $200 million from its expected revenue base, an amount equal to less than one-fifth of Amvuttra's revenue in a single quarter. Even so, Nasdaq data show that ALNY closed at $204.84, down from $286.62 the previous day, a one-day decline of approximately 28.5%. That disconnect is the heart of the story. The market was not rejecting RNAi technology. It was correcting an assumption: that the release of pent-up demand at the beginning of Amvuttra's launch could be extrapolated indefinitely as a linear growth trajectory. With the next commercial pillar capable of reducing Alnylam's dependence on TTR still not in place, a $200 million guidance reduction was enough to force a recalculation of the entire company's valuation model. In other words, Alnylam did not suddenly lose its ability to sell. It simply moved from a story in which every quarter was expected to beat expectations into a stage where it must prove the quality and durability of its growth. *Figure 2. Revenue concentration: Amvuttra accounted for approximately 86.3% of Alnylam's product net revenues in the second quarter of 2026.* ## 01 | Strong Results, but the Bad News Was in the Outlook Start by breaking down the numbers. In the second quarter of 2026, Amvuttra generated $1.012 billion in net product revenues for Alnylam, an increase of 106% from the same period a year earlier. Onpattro, another TTR product, contributed $18.46 million. Total TTR product revenues reached $1.030 billion, up 89% year over year. Givlaari and Oxlumo together generated $142 million, bringing company-wide product net revenues to $1.172 billion. Total revenue for the quarter was $1.291 billion. Alnylam reported GAAP operating income of $231 million and GAAP net income of $164 million; both measures had still been losses in the prior-year period. What unsettled the market on July 30, therefore, was not the quarter Alnylam had just completed. It was what the company said about the next stretch of the road. Alnylam acknowledged in its announcement that growth in U.S. demand for Amvuttra among second-line patients had slowed in early 2026 to what the company considered a "normalized" level. The reason was specific. When Amvuttra first launched for cardiomyopathy, there was a group of patients who were already taking TTR stabilizers, whose disease was continuing to progress, and who were waiting for a new treatment. Once the product was approved, those patients moved into treatment in a concentrated wave, releasing a backlog of pent-up demand. That demand was real, but an equivalent pool of waiting patients does not reappear every year. At launch, a new drug's revenue curve often benefits from two patient populations at once. The first is the accumulated "stock" of patients who have been waiting for a new therapy over an extended period. The second is the ongoing "flow" of patients who are newly diagnosed each year or who gradually enter treatment. The first group can make the initial launch curve exceptionally steep. The second determines how quickly the product can grow once it reaches a more normal operating state. Amvuttra is now moving from the first phase to the second. *Figure 3. Demand shifts from the release of pent-up second-line demand to a normalized flow of newly diagnosed and progressively treated patients.* This does not mean demand has collapsed. Alnylam disclosed that actual demand for its U.S. TTR business increased by approximately $129 million in the second quarter compared with the first, more than twice the demand increase recorded in the previous quarter. The growth visible in reported revenue was partly offset, however, by approximately $20 million of inventory movement and a modest decline in net price. Even after the guidance reduction, the midpoint of the revised TTR range still represented growth of approximately 75% over 2025. The problem is that capital markets do not pay only for a company that is "still growing." They pay for whether that growth can repeatedly exceed expectations that are already very high. When a launch curve has once been steep enough to look like a straight line, a return to normal speed can look like a stall. ## 02 | $200 Million Is Not Much; 86% Concentration Is Why did a guidance reduction of only $200 million trigger such a severe share-price reaction? Because Amvuttra has become large enough to stand in for Alnylam as a whole. Based on the company's second-quarter figures, Amvuttra accounted for approximately 86.3% of total product net revenues, while TTR products together represented approximately 87.9%. Givlaari and Oxlumo are both commercial products, and Alnylam also receives collaboration and royalty revenue from products including Leqvio and Qfitlia. It would be unfair to describe the company as having "only one drug." But if the question is where the next dollar of growth is primarily coming from, the answer remains heavily concentrated in Amvuttra. That concentration magnifies several variables tied to a single product into risks at the level of the entire company: the pace of new diagnoses, first-line adoption, reimbursement conditions, inventory, pricing, and competition. In its latest Form 10-Q, Alnylam also identified the concentration of revenue in Amvuttra as a material risk. If the company cannot maintain or expand sales, the effect on its operations could be significant. It is important to distinguish between "platform value" and "portfolio diversification." They are not the same thing. Alnylam has already proved that RNAi can travel from a scientific paper to global commercialization. Amvuttra is administered by subcutaneous injection once every three months. Under its FDA label, it can be used in adults with hereditary transthyretin-mediated amyloidosis with polyneuropathy, as well as wild-type or hereditary transthyretin-mediated amyloidosis with cardiomyopathy. This is not a proof-of-concept drug. It has crossed the barriers of approval, reimbursement, physician adoption, and manufacturing at scale. But owning a mature platform does not mean a company's revenue is automatically diversified. The scientific risk surrounding RNAi is declining, while Alnylam's portfolio risk remains high. The market's repricing on July 30 was asking a direct question: beyond Amvuttra, when will the next product arrive that can affect the income statement in billions of dollars? ## 03 | Wainua's Phase 3 Miss Should Not Be Rewritten as an RNAi Failure The part of the original market narrative most vulnerable to overstatement concerns the Phase 3 CARDIO-TTRansform results for eplontersen, marketed as Wainua, from Ionis and AstraZeneca. The bottom line should be stated first. The study did not meet its primary endpoint, and that does require the market to re-examine how TTR silencers and stabilizers should be used together. But it does not directly support the conclusion that "all TTR silencers are ineffective," much less that Amvuttra itself has failed. TTR stabilizers and TTR silencers intervene at different points in the disease process. Stabilizers such as tafamidis and acoramidis, marketed as Attruby, are designed to stabilize the TTR tetramer and reduce its dissociation and subsequent formation of amyloid fibrils. Eplontersen and vutrisiran instead reduce the liver's production of TTR at the RNA level. Eplontersen is an antisense oligonucleotide, while vutrisiran, the active ingredient in Amvuttra, is a GalNAc-conjugated double-stranded small interfering RNA. Both lower TTR, but they differ in molecule, therapeutic platform, dosing, and clinical-study design. CARDIO-TTRansform enrolled patients with ATTR cardiomyopathy. Its primary endpoint was a composite of cardiovascular death and recurrent cardiovascular events. According to Ionis, 57% of participants were already receiving a stabilizer when the study began, and another 24% started one during the trial. The overall study did not achieve statistical significance. In the prespecified monotherapy population that was not receiving a stabilizer, the company's announcement reported a nominal hazard ratio of 0.71. By contrast, no effect was observed among patients already taking a stabilizer at baseline. Because the complete data were expected to be presented at the European Society of Cardiology meeting in August 2026, it was too early, based on a topline announcement alone, to treat every subgroup finding as settled. The real question raised by the result is this: as more patients receive effective standard therapy, a new drug no longer has to prove only that it can lower TTR. It must show that, on top of modern background treatment, it can further reduce death, hospitalization, or functional decline. Hitting a molecular target, lowering a biomarker, and delivering clinical benefit are still separated by the test of a Phase 3 trial. This also explains why the market incorporated a competitor's setback into Alnylam's valuation model. It was not because Wainua's result had already invalidated Amvuttra. It was because the future TTR market is becoming more crowded, and every new product must clearly define its place in monotherapy, combination treatment, or sequential therapy. ## 04 | Nucresiran Is Next, but the Market Must Wait for Clinical Evidence, Not a Slogan The key asset on which Alnylam is betting for the next phase of TTR treatment is nucresiran. It is also an RNAi therapy, but the company aims to extend dosing to once every six months. The ongoing Phase 3 TRITON-CM trial is a randomized, quadruple-blind, placebo-controlled study. Following a July 23, 2026 update on ClinicalTrials.gov, estimated enrollment stood at 1,750 participants, and the primary endpoint was a composite of all-cause mortality and recurrent cardiovascular events. This study is not being conducted in a vacuum. Registry information shows that patients may participate while receiving approved stabilizers such as tafamidis or acoramidis, as well as standard heart-failure therapy. Those who have previously used or are currently using another TTR-lowering therapy are excluded. The estimated primary completion date was May 2030. In other words, if nucresiran is to become the next-generation product, it must prove more than the convenience of one injection every six months. It must deliver hard clinical outcomes on top of contemporary standard treatment. *Figure 4. The next leg of growth: nucresiran must demonstrate incremental clinical benefit on top of modern background therapy.* This is the timing gap Alnylam now faces. The company did have several data catalysts scheduled for the second half of 2026. ALN-HTT02 was expected to report preliminary Phase 1 results in Huntington's disease. ALN-6400 was expected to produce data from a Phase 1 study in healthy volunteers and a Phase 2 study in hereditary hemorrhagic telangiectasia. ALN-2232, an obesity candidate targeting ACVR1C, was also expected to report Phase 1 data. Mivelsiran had entered Phase 2 development in Down syndrome-associated Alzheimer's disease, and the regulatory application for cemdisiran had been accepted for review. These are important tests of whether the platform can expand. Most, however, remain early clinical, regulatory, or partnership milestones. In the near term, they cannot be treated as equivalent to another Amvuttra generating $1 billion in a single quarter. Alnylam's expanded collaboration with Inceptive on AI-enabled RNA drug design, and its deeper work with Komodo Health on real-world data analytics, may also improve the efficiency of drug discovery and commercial decision-making. But AI can shorten design and analysis cycles; it cannot replace human clinical trials, nor can it directly eliminate revenue concentration. The hardest question for Alnylam today is not whether RNAi still has a future. It is whether the next commercial pillar can arrive in time as Amvuttra's growth normalizes. ## 05 | Taiwan Has a Clinical Foothold, but No Evidence of a Direct Beneficiary Stock This topic should not be forced into a list of "Alnylam concept stocks" merely to tag Taiwanese oligonucleotide, contract development and manufacturing, or diagnostics companies on CMoney. As of July 31, 2026, publicly available primary sources did not show that any publicly traded Taiwanese company was a raw-material supplier, manufacturer, licensing partner, or commercialization partner for Amvuttra, nucresiran, or Alnylam's TTR products. The scope of the BeOne collaboration disclosed in Alnylam's announcement covered mainland China and Macau. It cannot be rewritten as "Greater China" in order to include Taiwan. Taiwan's more concrete links are instead found in clinical development and among companies with capabilities adjacent to RNAi. ClinicalTrials.gov listed TRITON-CM study locations in New Taipei, Taichung, and Taipei. This means Taiwan's medical system is participating directly in a global Phase 3 trial of next-generation TTR RNAi. The work can build experience in patient identification, genetic and imaging diagnosis, heart-failure care, and clinical-trial execution. That is a genuine medical and research role, but it is not an order that can be directly assigned to a Taiwanese stock. To map the real RNAi positions of Taiwanese public-market companies, it is useful to separate product development, delivery technology, and manufacturing services. Oneness Biotech (4743) and Microbio (4128) are co-developing the siRNA candidates SNS812 and SNS851 and are participating in a collaboration involving glycan-targeted delivery technology. These are named product, clinical, and delivery roles. But the two companies are working on the same set of jointly developed assets; they should not be counted twice as two separate pipelines. Nor is there any disclosed relationship between those assets and Alnylam or Amvuttra. IntelliGene (7832; Emerging Stock Board) sits more directly on the RNAi product and delivery side. Information from the Taipei Exchange and the company shows that it is developing the siRNA candidates IG-001 and IG-002, while also investing in lipid nanoparticle and exosome delivery systems. These programs remain in the research and development stage, with indications focused on coronavirus infections and influenza. No public information shows a collaboration or supply relationship with Alnylam or Amvuttra. Including IntelliGene in a map of Taiwan's RNAi landscape is reasonable; describing it as an Amvuttra beneficiary is not. Genomics (4195) occupies the services side. Its annual report disclosed capabilities in the synthesis and analysis of short-chain nucleic acids and siRNA active pharmaceutical ingredients, as well as contract research, development, and manufacturing organization services. It also participates in an LNP integration collaboration with a Japanese partner. This reflects Taiwan's development of upstream tools and manufacturing capacity for therapeutic nucleic acids. But the company has not disclosed Alnylam as a customer, and it does not separately report revenue from therapeutic siRNA. It therefore cannot be presented as part of Amvuttra's supply chain. If stocks are tagged for this article, the U.S. names can be identified precisely as Alnylam (NASDAQ: ALNY), Ionis (NASDAQ: IONS), AstraZeneca (NASDAQ: AZN), and BridgeBio (NASDAQ: BBIO). Taiwanese names can include Oneness Biotech (4743), Microbio (4128), IntelliGene (7832, Emerging Stock Board), and Genomics (4195), but both the article and the platform tags must state the same boundary: these companies have capabilities adjacent to RNAi or oligonucleotides and are not disclosed direct beneficiaries of Alnylam. If a platform does not support an Emerging Stock Board ticker, IntelliGene's actual role should remain in the body of the article rather than being replaced with an unrelated stock tag. For Taiwan's industry, three questions are genuinely worth tracking. Which companies have verifiable oligonucleotide manufacturing and quality systems? Which have entered the supply chains of international pharmaceutical companies? And which can turn ATTR diagnostic and clinical-trial capabilities into durable medical services? Stock tagging should come only after those roles are publicly established. ## 06 | What to Watch Next: Do Not Focus Only on Next Quarter's Revenue Alnylam's next phase can be tracked through six questions. First, will "actual demand" in the U.S. TTR business continue to increase, or will reported revenue be driven mainly by inventory and pricing fluctuations? Second, after the initial backlog of second-line patients dissipates, can newly diagnosed patients and those treated earlier in the disease course take over as the next source of growth? Third, can the revised $4.2 billion-$4.5 billion guidance be maintained, or will it be adjusted again? Fourth, will the complete CARDIO-TTRansform data clarify the differences between monotherapy and treatment on a stabilizer background, instead of leaving the market with only the label "Phase 3 failure"? Fifth, can TRITON-CM prove nucresiran's incremental clinical benefit on top of modern background therapy? Sixth, can readouts from the early-stage pipeline gradually convert platform value into a second and third genuinely commercializable product? The first three questions will determine the near-term slope of Alnylam's financial performance. The last three will determine whether Alnylam is ultimately a company with one mega-blockbuster drug or an RNAi platform capable of repeatedly producing new products. ## Conclusion | The Technology Did Not Break; the Valuation Shifted Gears First The most important point in Alnylam's second-quarter results was neither how frightening a 28% decline looked nor how impressive 106% growth appeared. It was that both figures could be true at the same time. Amvuttra has proved that RNAi can become a commercial product generating several billion dollars a year. But its share of product net revenues, at more than 86%, also shows that Alnylam's platform value has not yet been fully converted into a diversified product portfolio. The surge created by patients accumulated before launch was always going to normalize. That is not drug failure. It is a shift in the growth profile that every blockbuster new therapy eventually faces. The real test is whether new diagnoses, first-line adoption, and the next generation of pipeline assets can support the curve as it flattens. On July 30, the market did not declare RNAi a failure. It simply reminded Alnylam, with little courtesy, that the science of the platform has passed its test, while the economics of the platform still has another one to take. --- ## Disclaimer This article is an overview of biopharmaceutical and industry trends and does not constitute medical diagnosis, treatment, or investment advice. Data from different clinical trials should not be compared directly. Treatment decisions should be made by qualified healthcare professionals based on approved labeling, each patient's individual condition, and the latest clinical evidence. --- ## 一季賣破 10 億，Alnylam 為何崩 28%？ URL: https://drugnews.com.tw/articles/2026-08-05-alnylam-amvuttra-ttr-growth-reset.html Date: 2026-08-05 Language: zh-Hant Summary: Amvuttra 單季銷售突破 10 億美元，Alnylam 卻下修 TTR 指引並重挫 28%。真正被重估的不是 RNAi 技術，而是積壓需求退潮、收入集中與下一個商業支柱的時間差。 如果一款藥單季賣破 10 億美元、年增 106%，公司同季也交出 2.31 億美元 GAAP 營業利益，股價卻在一天內跌掉 28.5%，市場究竟看見了什麼？ 7 月 30 日，RNAi 龍頭 Alnylam 公布 2026 年第二季財報。產品淨收入 11.72 億美元、年增 74%；其中 Amvuttra 單季 10.12 億美元，TTR 產品合計 10.30 億美元。這不是一份「業績暴雷」的成績單。 但同一天，公司把 2026 年 TTR 產品收入指引從 44 億至 47 億美元，下修到 42 億至 45 億美元。以區間中位數計算，市場少掉的是 2 億美元，不到 Amvuttra 一季收入的五分之一；Nasdaq 資料卻顯示，ALNY 收盤從前一日 286.62 美元跌到 204.84 美元，單日下跌約 28.5%。 這個反差才是整件事的核心。 市場砍的不是 RNAi 技術，而是把 Amvuttra 上市初期的積壓需求，錯當成可以永久線性外推的成長斜率。當下一個能分攤 TTR 依賴的商業支柱還沒接上，2 億美元的指引下修，就足以讓整家公司的估值模型重新計算。 換句話說，Alnylam 沒有突然賣不動；它只是從「每一季都要超預期」的故事，進入必須證明成長品質的階段。 *圖 1｜Amvuttra 單季銷售突破 10 億美元，但收入增長並不必然帶來估值擴張。* ## 01｜財報很好，壞消息藏在未來式 先把數字拆開。 2026 年第二季，Alnylam 的 Amvuttra 淨產品收入為 10.12 億美元，比去年同期增加 106%；同屬 TTR 產品的 Onpattro 為 1,846 萬美元，TTR 合計 10.30 億美元、年增 89%。Givlaari 與 Oxlumo 合計 1.42 億美元，讓全公司產品淨收入來到 11.72 億美元。 公司同季總營收為 12.91 億美元，GAAP 營業利益 2.31 億美元、GAAP 淨利 1.64 億美元；去年同期兩項都還是虧損。 所以 7 月 30 日真正讓市場不安的，不是這一季，而是公司對下一段路的說法。 Alnylam 在公告裡坦白指出，美國 Amvuttra 二線患者的需求成長，已在 2026 年初放緩到公司認定的「正常化」水準。原因也很具體：Amvuttra 的心肌病變適應症上市初期，有一批已使用 TTR 穩定劑、疾病仍持續惡化，而且正等待新療法的患者。產品一核准，這群人集中轉入，形成一波積壓需求釋放。 這種需求是真的，卻不會每年重新長出同樣一批。 新藥剛上市時，營收常同時吃到兩種患者：一種是長期累積、等待新療法的「存量」；另一種是之後每年新診斷、或逐步進入治療的「流量」。前者可以把上市曲線拉得很陡，後者才決定產品進入常態後能跑多快。 Amvuttra 現在正從前者切到後者。 這不等於需求崩塌。Alnylam 揭露，美國 TTR 業務第二季相較第一季，實際需求增加約 1.29 億美元，是前一季需求增量的兩倍以上；只是報表上的成長被約 2,000 萬美元庫存變動，以及小幅淨價下降抵銷了一部分。公司下修後的 TTR 指引中位數，仍相當於比 2025 年成長約 75%。 問題是，資本市場買的從來不只「還在成長」，而是成長能不能一再超過已經很高的預期。 當一條上市曲線曾經陡到像直線，後來只是回到正常速度，看起來也會像失速。 *圖 2｜Amvuttra 約占 Alnylam 2026 年第二季產品淨收入 86.3%，TTR 產品合計約占 87.9%。* ## 02｜2 億美元不大，86% 的集中度很大 為什麼指引只少 2 億美元，股價反應會這麼激烈？ 因為 Amvuttra 已經大到足以代表 Alnylam。 按公司第二季數字計算，Amvuttra 約占全公司產品淨收入 86.3%，TTR 產品合計占約 87.9%。Givlaari 與 Oxlumo 都是已上市產品，Alnylam 也有 Leqvio、Qfitlia 等合作與權利金收入；把它說成「只有一款藥」並不公平。 但若問下一美元成長主要從哪裡來，答案仍高度集中在 Amvuttra。 這讓一項產品的幾個變數——新診斷速度、一線採用、支付條件、庫存、價格與競品——都會被放大成整家公司層級的風險。公司在最新 10-Q 也把 Amvuttra 的收入集中列為重要風險：若無法維持或擴大銷售，可能對營運造成重大影響。 這裡要分清楚「平台價值」與「產品組合」是兩件事。 Alnylam 已經證明 RNAi 可以從科學論文走到全球商業化；Amvuttra 每三個月皮下注射一次，依 FDA 標籤已可用於成人遺傳型 ATTR 多發性神經病變，以及野生型或遺傳型 ATTR 心肌病變。它不是概念藥，而是已經跨過核准、給付、醫師採用與規模化製造的產品。 但一家公司擁有成熟平台，不代表它的收入已經自然分散。 RNAi 的科學風險正在下降，Alnylam 的組合風險卻仍然很高。7 月 30 日的市場定價，正是在追問：除了 Amvuttra，下一個能用「十億美元」為單位影響財報的產品，何時才會出現？ *圖 3｜積壓患者釋放推高上市初期曲線，後續增長更仰賴新增診斷與前線滲透。* ## 03｜Wainua 三期失利，不能被寫成 RNAi 失效 原始市場敘事裡，最容易被寫過頭的，是 Ionis 與 AstraZeneca 的 eplontersen（Wainua）三期 CARDIO-TTRansform 結果。 先講結論：這項研究沒有達到主要終點，確實會讓市場重新檢查 TTR 沉默劑與穩定劑之間的搭配邏輯；但它不能直接推論為「所有 TTR silencer 都沒用」，更不能直接推論 Amvuttra 失效。 TTR 穩定劑與 TTR 沉默劑處理的是不同環節。 Tafamidis、acoramidis（Attruby）等穩定劑，目標是穩住 TTR 四聚體，減少它解離後形成澱粉樣纖維；eplontersen 與 vutrisiran 則從 RNA 層級降低肝臟製造 TTR。Eplontersen 是反義寡核苷酸；Amvuttra 的成分 vutrisiran 則是 GalNAc 偶聯的雙股 siRNA。兩者都會降低 TTR，分子、作用平台、劑量與臨床設計卻不同。 CARDIO-TTRansform 納入 ATTR 心肌病變患者，主要終點是心血管死亡與反覆心血管事件的複合結果。Ionis 公告顯示，57% 受試者在研究開始時已使用穩定劑，另有 24% 在試驗期間開始使用；整體研究未達統計顯著。 預先設定、未使用穩定劑的單藥族群，公告則報告名目風險比 0.71；相反地，基線已使用穩定劑者沒有觀察到效果。由於完整資料預計在 2026 年 8 月歐洲心臟學會公布，現在還不能只憑一份 topline 公告把所有次族群寫成定論。 這項結果真正提出的問題是：當愈來愈多患者已在使用有效的標準治療，新藥要證明的已不是「能不能降低 TTR」，而是「在現代背景治療上，還能不能把死亡、住院或功能惡化再往下推」。 命中標的、生物標誌下降與臨床獲益，中間仍隔著一張三期考卷。 這也解釋了市場為什麼會把競品的失利，納入 Alnylam 的估值模型。不是因為 Wainua 的結果已否定 Amvuttra，而是未來 TTR 市場愈來愈擁擠，每一個新產品都要說清楚自己在單用、聯合或序貫治療中的位置。 ## 04｜下一棒 nucresiran，要等的是臨床而不是口號 Alnylam 為 TTR 下一階段押注的關鍵資產，是 nucresiran。 它同樣是 RNAi 療法，但公司希望把給藥頻率拉長到每六個月一次。正在進行的第三期 TRITON-CM，是一項隨機、四盲、安慰劑對照研究；ClinicalTrials.gov 在 2026 年 7 月 23 日更新後，預估收案 1,750 人，主要終點為全因死亡與反覆心血管事件的複合結果。 這項試驗不是在真空環境裡跑。登錄資料顯示，患者可以在 tafamidis 或 acoramidis 等核准穩定劑，以及心衰竭標準治療的背景下參與；曾經或目前使用其他 TTR-lowering therapy 者則被排除。主要完成時間目前估計在 2030 年 5 月。 換言之，nucresiran 若要成為下一代產品，必須證明的不只是半年打一針很方便，還要在當代標準治療上拿出硬臨床結果。 這正是 Alnylam 的時間差。 2026 年下半年，公司確實有多個資料節點：ALN-HTT02 預計公布亨丁頓舞蹈症一期初步結果；ALN-6400 將有健康受試者一期與遺傳性出血性微血管擴張症二期資料；靶向 ACVR1C 的肥胖症候選藥 ALN-2232 也預計讀出一期資料。Mivelsiran 已在唐氏症相關阿茲海默症啟動二期，cemdisiran 的監管申請也已獲受理。 這些是平台能否擴張的重要證據，卻多半仍是早期臨床、監管或合作節點，不能在短期內等同另一個單季十億美元的 Amvuttra。 公司與 Inceptive 擴大 AI RNA 藥物設計合作、與 Komodo Health 深化真實世界資料分析，也可能提高發現與商業決策效率。但 AI 可以縮短設計與分析，不能替代人體試驗，更不能直接填補收入集中。 Alnylam 現在最難回答的，不是「RNAi 還有沒有未來」，而是下一個商業支柱能否在 Amvuttra 成長正常化後及時接上。 *圖 4｜nucresiran 的真正考題，是在現代背景治療上證明額外臨床獲益。不同分子與試驗設計不可直接類推。* ## 05｜台灣有臨床座標，沒有直接受惠股證據 這個題目不能為了 CMoney 標股，就硬把台灣寡核苷酸、CDMO 或檢測公司全部塞成「Alnylam 概念股」。 截至 2026 年 7 月 31 日，公開一級資料沒有顯示台灣公開市場公司是 Amvuttra、nucresiran 或 Alnylam TTR 產品的原料、製造、授權或商業化夥伴。Alnylam 這次公告的 BeOne 合作範圍是中國大陸與澳門，也不能自行擴寫成「大中華區」而把台灣納入。 台灣目前較具體的連結，反而在臨床端與「RNAi 能力相鄰」公司。 ClinicalTrials.gov 列出的 TRITON-CM 台灣試驗地點涵蓋新北、台中與台北。這代表台灣醫療體系實際參與下一代 TTR RNAi 的全球三期試驗，能累積患者辨識、基因／影像診斷、心衰照護與臨床執行經驗；它是一個真實的醫療與研發角色，但不是一張可以直接標記的台股訂單。 若要看台灣公開市場公司的真實 RNAi 座標，可以分成產品、遞送與製造服務端。 合一（4743）與中天（4128）共同開發 siRNA 候選藥 SNS812、SNS851，並參與醣靶向遞送技術合作。這是具名產品、臨床與遞送角色；但兩家公司對應的是同一組共同開發資產，不能重複算成兩套管線，也與 Alnylam／Amvuttra 沒有已揭露關係。 智新生技（7832，興櫃）則更直接站在 RNAi 產品與遞送技術端。櫃買中心及公司資料顯示，智新正在開發 siRNA 候選藥 IG-001、IG-002，並投入 LNP 與外泌體遞送系統；目前仍屬研發階段，適應症集中在冠狀病毒感染與流感，也沒有公開資料顯示它與 Alnylam／Amvuttra 有合作或供應關係。把它納入台灣 RNAi 版圖是合理的，把它寫成 Amvuttra 受惠股則不合理。 基米－創（4195）則在服務端。公司年報揭露短鏈核酸與 siRNA 原料藥的合成、分析與 CRDMO 能力，並參與日本夥伴的 LNP 整合合作。它反映台灣正在建立治療性核酸的上游工具與製造能力；但公司沒有揭露 Alnylam 是客戶，也沒有把治療用 siRNA 的營收單獨拆出，不能寫成 Amvuttra 供應鏈。 所以這篇若標股票，美股可以精準標記 Alnylam（NASDAQ：ALNY）、Ionis（NASDAQ：IONS）、AstraZeneca（NASDAQ：AZN）與 BridgeBio（NASDAQ：BBIO）；台灣標的可列合一（4743）、中天（4128）、智新生技（7832，興櫃）與基米－創（4195），但文章與平台標記都必須同步寫明「RNAi／寡核苷酸能力相鄰，非 Alnylam 直接受惠」。若平台不支援興櫃代號，正文仍保留智新的真實角色，不得改掛其他股票。 對台灣產業真正值得追的，是三件事：誰有可驗證的寡核苷酸量產與品質系統、誰進入國際藥廠供應鏈、誰能把 ATTR 的診斷與臨床試驗能力轉成長期醫療服務。等角色公開落地，再談標股。 ## 06｜接下來，不要只盯下一季營收 Alnylam 的下一階段，可以用六個問題追蹤： 第一，美國 TTR 的「實際需求」是否持續增加，還是報告收入主要受庫存與價格擾動？ 第二，上市初期的二線積壓需求退去後，新診斷與較前線患者能否接棒？ 第三，42 億至 45 億美元的新指引能否守住，還會不會再次調整？ 第四，CARDIO-TTRansform 完整資料能否說清楚單用與穩定劑背景下的差異，而不是只剩一個「三期失敗」標籤？ 第五，TRITON-CM 能否在現代背景治療上證明 nucresiran 的臨床增益？ 第六，早期管線讀出能否逐步把平台價值，變成第二、第三個真正可商業化的產品？ 這六題裡，前三題決定近期財報斜率；後三題決定 Alnylam 究竟是一家擁有一支超級重磅藥物的公司，還是一個能不斷複製產品的 RNAi 平台。 ## 結語｜技術沒有暴雷，估值先換檔 Alnylam 第二季最值得看的，不是 28% 跌幅有多嚇人，也不是 106% 成長有多漂亮，而是兩個數字同時成立。 Amvuttra 已經證明 RNAi 可以長成一年數十億美元的商業產品；但它在產品收入裡占比超過 86%，也說明 Alnylam 的平台價值，還沒有完全轉成分散的產品組合。 上市初期積壓患者帶來的爆發，終究會回到常態。這不是藥物失效，而是每一款重磅新藥都會經過的成長換檔。真正的考驗，是公司能不能讓新增診斷、一線採用與下一代管線，接住那條變平的曲線。 7 月 30 日，市場沒有宣判 RNAi 失敗。 它只是很不客氣地提醒 Alnylam：平台科學已經及格，平台經濟還要再考一次。 --- ## 免責聲明 本文僅為生技醫藥與產業趨勢整理，不構成醫療診斷、治療或投資建議。跨試驗數據不可直接比較；患者用藥應由專業醫療團隊依核准標籤、個別病況與最新臨床證據判斷。 --- ## A $400 Billion Pharma Merger? AstraZeneca and Bristol Myers Squibb Are Reportedly Talking URL: https://drugnews.com.tw/articles/2026-08-04-astrazeneca-bms-400b-merger-talks-en.html Date: 2026-08-04 Language: en Summary: AstraZeneca and BMS reportedly explored a combination with a combined market value approaching $400 billion. There is no announced transaction. The real questions are AstraZeneca's 2030 growth target, BMS's Eliquis and Opdivo exclusivity timeline, oncology-pipeline overlap, antitrust and whether integration can actually buy time. The global pharmaceutical industry may be considering a combination on a scale it has never seen before. The Financial Times, citing people familiar with the matter, reported that AstraZeneca and Bristol Myers Squibb (BMS) had discussed a possible combination in recent months. At the share prices prevailing when the report emerged, the two companies had a combined market value approaching $400 billion. If the discussions ever turn into a transaction, this would not simply be a case of two large drugmakers becoming one even larger company. Tagrisso, Imfinzi, Enhertu, Opdivo, Yervoy, Eliquis, Reblozyl, Breyanzi and Abecma are only the beginning of a product list spanning lung cancer, breast cancer, liver cancer, hematologic malignancies, anticoagulation and CAR-T therapy. All of those assets could end up on the same balance sheet. Adding their publicly reported 2025 revenue produces a combined annual total of more than $100 billion. AstraZeneca's depth in precision medicines for solid tumors and antibody-drug conjugates, paired with BMS's immuno-oncology, hematology, cell therapy and U.S. commercial infrastructure, could create a platform covering nearly every major therapeutic modality in oncology. That is why the report immediately attracted so much attention. But a potential combination would be much more complicated than a straightforward union of two strong companies. AstraZeneca is pursuing a goal of at least $80 billion in revenue by 2030. BMS is racing to establish its next growth engines before Eliquis and Opdivo reach major exclusivity inflection points. If the two companies are sitting at the same table, what they may really be trying to buy is not scale, but time. And in pharmaceutical mergers, time is often the hardest asset to buy. *Figure 1 | The nearly $400 billion figure refers to the companies' combined market value around the time of the report. It is not an offer price or a completed transaction value.* ## 01 | Why now? Start with AstraZeneca. The company is not in crisis. On the contrary, it is in the middle of a powerful growth phase. Total revenue reached $58.739 billion in 2025, and assets including Tagrisso, Imfinzi, Calquence, Lynparza and the Daiichi Sankyo-partnered Enhertu have made it one of the most important drugmakers in global solid-tumor oncology. The pressure comes from the destination AstraZeneca has set for itself. The company wants to generate at least $80 billion in annual revenue by 2030. That means it must add more than $20 billion over the next several years, and existing products cannot be expected to do all of the work through organic expansion alone. AstraZeneca has announced plans to invest $50 billion in U.S. research and manufacturing through 2030. In February 2026, its ordinary shares also began trading directly on the New York Stock Exchange while the company retained its London and Stockholm listings. Those actions do not mean that AstraZeneca is moving its headquarters to the United States. They do, however, make the strategic direction clear: the U.S. market, capital base and policy environment are becoming increasingly important to the company's next phase. BMS could fill much of that gap. It brings an established U.S. commercial network, substantial cash flow, and deep capabilities in hematology and cell therapy. Building those assets from scratch would take AstraZeneca years. In theory, integrating BMS could provide products, teams, hospital relationships and commercial channels in a single transaction. Now look at BMS. It is not a company standing at the edge of a cliff waiting to be rescued. In the second quarter of 2026, BMS reported revenue of $12.973 billion, up 6% year over year, and raised its full-year revenue guidance to $49 billion to $50 billion. Eliquis generated $4.481 billion in the quarter, Opdivo generated $2.485 billion, and the subcutaneous formulation Opdivo Qvantig contributed another $261 million. That strength is also the problem. Together, those three products represented about 55.7% of quarterly revenue. BMS's biggest medicines remain highly productive, but the company is still heavily dependent on a small number of large franchises. In its annual report, BMS estimated that minimum U.S. market exclusivity for both Eliquis and Opdivo extends through 2028. That does not mean both products will fall to zero on the same day in 2028, and patent timelines will not be identical in every country. But the strategic timetable is unmistakable: growth products such as Reblozyl, Breyanzi, Camzyos and Opdualag must take over before erosion of the old pillars becomes material. One company needs faster growth. The other needs its next sufficiently large growth curve. That is why the rumor is not difficult to understand. ## 02 | How extraordinary would the combined oncology map be? Viewed strictly as a product portfolio, AstraZeneca and BMS make for a compelling combination. AstraZeneca's core strengths are precision therapy for solid tumors, ADCs, lung cancer and increasingly complex combination regimens. Tagrisso, Imfinzi, Enhertu and Datroway already reach deeply into lung, breast and gastrointestinal cancers. BMS has a different foundation. Opdivo, Yervoy and Opdualag form an immuno-oncology backbone. Reblozyl, Pomalyst and other products provide depth in hematology. Breyanzi and Abecma have also taken the company into commercial CAR-T therapy. A combined company could simultaneously hold major positions in: - precision medicines for large solid-tumor markets such as lung and breast cancer; - antibody-drug conjugates; - PD-1/PD-L1 and CTLA-4 dual immunotherapy; - hematologic malignancies; - CD19 and BCMA CAR-T therapies; and - rare diseases and cardiovascular, renal and metabolic medicine. The value would not come only from owning a longer list of products. If BMS's U.S. commercial organization and AstraZeneca's solid-tumor development capabilities could truly be integrated, targeted therapies, ADCs, immunotherapies and cell therapies in the same tumor type could share clinical-development infrastructure, physician networks and global market resources. This portfolio would also be more than a collection of medicines sold to the same physicians. For a large drugmaker, each cancer type increasingly resembles a full operating chain. Biomarkers identify patient groups; clinicians choose among targeted therapy, ADCs, immunotherapy and cell therapy; then the company must address resistance and later-line treatment after the first option fails. The more modalities a company controls, the greater its chance of remaining involved throughout the patient's treatment journey. AstraZeneca is already expanding across EGFR, HER2, TROP2, CLDN18.2 and other solid-tumor biomarkers. BMS has mature immunotherapy combinations, hematology franchises and CAR-T products. If integration worked, the new company would not have to bet on a single technology. It could design multiple trials and combinations within the same cancer type. That vision of a tumor-specific platform is what makes a combination approaching $400 billion genuinely consequential. For a major pharmaceutical company, the attraction is not simply having more products. It is the ability to deploy assets across an entire patient journey. Yet the more complete the map becomes, the more likely the borders are to collide. *Figure 2 | One company needs faster growth and greater U.S. scale. The other must build its next major revenue curve before core-product exclusivity turns.* ## 03 | The overlaps in liver cancer and CAR-T are already visible The clearest overlap is dual immunotherapy in liver cancer. AstraZeneca has Imfinzi plus Imjudo. BMS has Opdivo plus Yervoy. Both pair a PD-L1 or PD-1 agent with a CTLA-4 antibody, and both have entered first-line treatment for unresectable or metastatic hepatocellular carcinoma. Today, those regimens compete for physicians, patients, budgets and clinical mindshare. If they eventually sit inside the same company, management would have to answer difficult questions. Should both regimens continue to receive full commercial support? How should pricing differ across countries? Which follow-on trials deserve capital and operating resources? The second overlap is CAR-T therapy for multiple myeloma. BMS already markets the BCMA-directed CAR-T Abecma. AstraZeneca acquired the investigational BCMA/CD19 dual-targeting asset GC012F, now AZD0120, through its acquisition of Gracell. One is a marketed product and the other is a dual-target clinical-stage program, so it would be misleading to treat them as identical assets. But they would still compete for attention around BCMA, multiple myeloma, manufacturing capacity, treatment centers and clinical-trial resources. This is the part of pharmaceutical consolidation that presentation decks tend to beautify. The deck says "synergy." In practice, integration can mean two medicines in the same tumor type competing for patients, two development plans competing for budget, and externally licensed assets competing with internally discovered programs for priority. Eventually, some programs remain at the center. Others are delayed, sold or discontinued. If this transaction ever happens, the most important question will not be how many pipeline assets the new company owns. It will be whether management is willing to state clearly which programs are strategic and which ones become the price of consolidation. ## 04 | Where are the talks now? The reporting is real, but there is no transaction The tense of this story matters. As of August 3, 2026, the public record supports one conclusion: the Financial Times, citing unnamed people familiar with the matter, reported that the companies had explored a possible combination in recent months. Reuters subsequently reported the development as well. Neither company has formally announced a merger. There is no published price, exchange ratio, transaction structure or signed agreement. AstraZeneca declined to comment to the media, while BMS did not immediately respond. Public SEC filings also do not contain a formal transaction that could accurately be described as a $400 billion acquisition. The $400 billion figure is therefore not a bid and not a purchase price. It is the approximate combined market capitalization of the two companies around the time of the report. If a future transaction were structured as an all-stock merger, the real economics would lie in the exchange ratio, control and shareholder dilution. If one party acquired the other, the analysis would need to examine the premium, financing and debt. In other words, a pharmaceutical merger of historic scale may have been discussed, but no formal document has yet converted possibility into fact. That boundary does not weaken the story. It makes every subsequent signal more important. *Figure 3 | The portfolios are complementary, but they also create overlaps that would need to be managed in areas such as dual immunotherapy for liver cancer and CAR-T for multiple myeloma.* ## 05 | Even an agreed deal would still face antitrust scrutiny Regulators would not evaluate a transaction of this size by looking only at current product sales. They would also examine late-stage pipelines, future competition and whether the transaction could eliminate assets that might otherwise have competed with one another. When BMS acquired Celgene in 2019, the U.S. Federal Trade Commission required the divestiture of Otezla because of potential competition with BMS's then-investigational TYK2 asset. Amgen ultimately acquired Otezla for $13.4 billion, creating one of the most prominent divestiture precedents in modern pharmaceutical M&A. That precedent does not mean an AstraZeneca-BMS combination would necessarily be blocked, and it cannot predict which specific asset might have to be sold. It does mean that dual immunotherapy in liver cancer, hematology, CAR-T, and combinations involving immunotherapy and ADCs across multiple solid tumors would likely receive detailed review. Antitrust is only one layer. The integration itself may be even harder to quantify. Who would be chief executive? Where would research leadership sit? How would U.K. and U.S. listings and governance be organized? How much dilution would AstraZeneca shareholders accept? What premium would BMS shareholders demand before giving up control, particularly before the patent-cliff pressure fully materializes? A spreadsheet can add two market capitalizations in seconds. Preventing two enormous organizations from slowing each other down is the genuinely expensive part. ## 06 | Taiwan has a real role, but this is not a license to manufacture concept stocks When news like this breaks, the most common reaction in Taiwan's equity market is to pull every company associated with ADCs, antibodies or CDMO services into the narrative. There is currently insufficient primary-source evidence to establish that any Taiwan-listed company holds direct rights, a confirmed order or a defined economic benefit tied to these reported discussions. Taiwan's verifiable role is in clinical trials. Trials associated with BMS's iza-bren list National Taiwan University Hospital, Taipei Veterans General Hospital, Taichung Veterans General Hospital, National Cheng Kung University Hospital, E-Da Hospital and Taipei Medical University Hospital among Taiwan sites. AstraZeneca's AZD0901 trial also lists a Taipei study location. Historical registration for AstraZeneca's HIMALAYA liver-cancer study likewise shows Taiwan participation. Taiwan is therefore not merely watching global oncology development from the outside. Its hospitals, investigators and patients are directly involved in multinational trials, enrollment and data generation. If the companies were eventually to combine, the most practical questions for Taiwan would not be which stock rises first. The questions would be whether existing trials continue, whether externally licensed assets are reprioritized, and whether clinical-development and supply-chain decisions become more concentrated. For Taiwan biotechnology companies that work with multinational pharmaceutical partners, a larger partner does not automatically mean more resources. Sometimes it only means that a local company's program becomes smaller inside a much larger global pipeline. *Figure 4 | Media reporting is not a completed transaction. Formal terms, governance, regulatory review and asset prioritization all remain unresolved.* ## Conclusion | $400 billion can buy scale. Can it buy time? There is a coherent industrial logic behind a possible AstraZeneca-BMS combination. One company is pursuing an $80 billion revenue target for 2030 and wants to accelerate U.S. expansion and portfolio growth. The other has substantial cash flow and an established commercial network, but must complete its product transition before exclusivity on its largest medicines turns. If completed, the combination could redraw the global oncology landscape and become one of the largest pharmaceutical mergers ever attempted. But a more complete portfolio also creates harder overlaps. A larger company can make integration slower and decision-making more difficult. Acquiring BMS could give AstraZeneca many capabilities at once. Preserving those capabilities after the acquisition would be a separate challenge. The next information worth waiting for is not another adjective from an unnamed source. It is four concrete answers in a formal announcement: how the transaction would be structured, who would control the combined company, which assets would be divested, and which pipeline programs would receive priority. $400 billion can buy one of the most comprehensive drug portfolios in the world. Whether it can buy the next decade of growth remains unanswered. ## Disclaimer This article is provided for biotechnology, pharmaceutical-industry and market analysis. It does not constitute medical diagnosis, treatment advice or investment advice. Media reports, corporate strategy, transaction negotiations, regulatory review, drug development and commercialization can change. Readers should rely on the latest public information from companies, regulators and primary-source documents when making decisions. --- ## 超級大藥廠 4,000 億美元合併？阿斯特捷利康＋BMS！ URL: https://drugnews.com.tw/articles/2026-08-04-astrazeneca-bms-400b-merger-talks.html Date: 2026-08-04 Language: zh-Hant Summary: Financial Times 報導 AstraZeneca 與 BMS 曾探討結合，合計市值接近 4,000 億美元。這不是已簽交易；真正的產業問題是 2030 成長、Eliquis／Opdivo 獨占期、腫瘤管線重疊、反壟斷與整合能否買到時間。 全球製藥業，可能正在醞釀一場史上最大級的合併。 Financial Times 引述知情人士報導，AstraZeneca（阿斯特捷利康）與 Bristol Myers Squibb（BMS，百時美施貴寶）近幾個月曾討論結合的可能性。 以報導當時的股價計算，兩家公司合計市值接近 4,000 億美元。 如果真的走到合併，這不只是兩家大藥廠變成一家更大的藥廠。 Tagrisso、Imfinzi、Enhertu、Opdivo、Yervoy、Eliquis、Reblozyl、Breyanzi、Abecma……一整排橫跨肺癌、乳癌、肝癌、血液腫瘤、抗凝血與 CAR-T 的產品，可能被放進同一張資產負債表。 以 2025 年公開財報粗略相加，兩家公司年度營收超過 1,000 億美元。阿斯特捷利康在實體瘤精準治療與 ADC 的深度，若接上 BMS 的免疫治療、血液腫瘤、細胞治療與美國商業網絡，幾乎可以拼出一個覆蓋腫瘤治療主要模態的超級平台。 這就是消息一出，市場立刻炸鍋的原因。 但這場可能的合併，並不是單純的強強聯手。 阿斯特捷利康想追的是 2030 年 800 億美元營收目標；BMS 要趕的是 Eliquis 與 Opdivo 走向獨占期轉折前的產品接棒。兩家公司坐上同一張桌，真正想買的可能都不是規模，而是時間。 而藥廠合併最難買到的，偏偏也是時間。 *圖 1｜接近 4,000 億美元的是報導當時兩家公司的合計市值量級，不是出價或成交價。* ## 01｜為什麼可能是現在？ 先看阿斯特捷利康。 這家公司不是遇到危機，反而正處在一段很強的成長期。2025 年總營收達 587.39 億美元，Tagrisso、Imfinzi、Calquence、Lynparza，加上與 Daiichi Sankyo 合作的 Enhertu 等資產，已經讓它成為全球實體瘤最有存在感的大藥廠之一。 真正的壓力，是它替自己訂下了一個很高的終點。 阿斯特捷利康希望 2030 年營收至少達 800 億美元。也就是說，未來幾年還得再長出超過 200 億美元，而且不能只靠既有產品自然放量。 公司已宣布至 2030 年在美國投資 500 億美元，布局研發與製造；2026 年 2 月也讓普通股直接在紐約證交所交易，同時保留倫敦與斯德哥爾摩上市。這些動作不等於「把總部搬去美國」，卻清楚顯示：美國市場、資本與政策環境，在阿斯特捷利康下一階段的權重愈來愈高。 BMS 剛好能補上這塊。 它擁有成熟的美國商業網絡、大體量現金流、血液腫瘤與細胞治療能力。阿斯特捷利康若從零建立，需要時間；透過整合 BMS，理論上可以一次買到產品、團隊、醫院關係與商業通路。 再看 BMS。 它也不是一家公司走到懸崖邊，只能等人救援。 2026 年第二季，BMS 營收 129.73 億美元、年增 6%，還把全年營收指引上調至 490 億至 500 億美元。Eliquis 單季賣出 44.81 億美元，Opdivo 24.85 億美元，皮下注射版本 Opdivo Qvantig 也有 2.61 億美元。 問題就在這裡。 這三項合計約占單季營收 55.7%。BMS 的大藥仍然很會賺，公司對少數大藥的依賴也仍然很深。 BMS 在年報裡估計，Eliquis 與 Opdivo 在美國的最低市場獨占期都到 2028 年。這不代表兩款藥會在 2028 年同一天歸零，更不代表全球市場使用同一個專利日期；但對 BMS 而言，時間表已經很清楚：Reblozyl、Breyanzi、Camzyos、Opdualag 等成長產品，必須在舊支柱受到侵蝕前接起來。 所以一方缺的是更快的增長，一方缺的是下一條夠大的曲線。 這場傳聞會出現，並不難理解。 ## 02｜合併後，會是一張多誇張的腫瘤地圖？ 如果只看產品拼圖，阿斯特捷利康與 BMS 確實非常有想像力。 阿斯特捷利康的核心能力，在實體瘤精準治療、ADC、肺癌與多種聯合療法。它手上的 Tagrisso、Imfinzi、Enhertu、Datroway，已經深入肺癌、乳癌與消化道腫瘤。 BMS 的底盤則不同。Opdivo、Yervoy、Opdualag 建立的是免疫治療骨架；Reblozyl、Pomalyst 等產品累積了血液腫瘤深度；Breyanzi 與 Abecma 則讓它已經跨進商業化 CAR-T。 兩家公司若真的放在一起，可以同時握住： 肺癌與乳癌等大型實體瘤的精準標靶； ADC； PD-1／PD-L1 與 CTLA-4 雙免疫； 血液腫瘤； CD19 與 BCMA CAR-T； 罕病與心腎代謝。 更重要的是，這不是只有藥物名稱變多。BMS 的美國商業化網絡與阿斯特捷利康的實體瘤開發能力若能真正整合，理論上可以讓同一個癌種裡的標靶、ADC、免疫治療與細胞治療，共享臨床開發、醫師網絡與全球市場資源。 這張地圖的價值，也不只是把藥賣給同一批醫師。對大藥廠來說，一個癌種愈來愈像一整條經營鏈：先用生物標記分出患者，再決定標靶、ADC、免疫治療或細胞治療；第一線失敗後，還要接住耐藥與後線治療。誰手上的模態夠多，誰就比較有機會在不同治療階段持續留在患者旅程裡。 阿斯特捷利康已經在 EGFR、HER2、TROP2、CLDN18.2 等實體瘤標誌物上擴張；BMS 則握有成熟免疫組合、血液腫瘤產品與 CAR-T。若整合順利，新公司不必只押一種技術的成敗，還能在同一癌種裡安排多種試驗與聯合療法。這種「癌種平台」想像，才是接近 4,000 億美元規模真正讓人屏息的地方。 對大型藥廠來說，這種組合最誘人的地方，不是「產品多」，而是可以圍繞患者整段治療旅程配置資產。 但越完整的地圖，也越容易在邊界上撞車。 *圖 2｜一方需要更快的成長與美國規模，另一方需要在核心產品獨占期轉折前補上下一條大曲線。* ## 03｜肝癌與 CAR-T，已經看得到正面重疊 第一個最明顯的重疊，是肝癌雙免疫。 阿斯特捷利康有 Imfinzi 加 Imjudo，BMS 有 Opdivo 加 Yervoy。兩組都是 PD-L1／PD-1 加 CTLA-4 的雙免疫策略，也都已經進入不可切除或轉移性肝細胞癌第一線治療。 它們原本在同一個市場競爭醫師、患者、預算與臨床心智。若有一天被放進同一家公司，管理層就得回答：兩套療法都繼續推嗎？不同國家如何定價？後續試驗與商業資源如何分配？ 第二個重疊，是多發性骨髓瘤 CAR-T。 BMS 已有上市的 BCMA CAR-T Abecma；阿斯特捷利康收購 Gracell 後，取得 BCMA／CD19 雙靶向在研資產 GC012F（AZD0120）。一邊是上市產品，一邊是雙靶點臨床管線，成熟度不同，也不能粗暴畫成同一支產品。 但它們確實會在 BCMA、多發性骨髓瘤、製造產能、治療中心與臨床試驗資源上碰到彼此。 這也是大型藥廠合併最容易被簡報美化的地方。 簡報會寫「協同效應」；真正的整合，卻可能是同一癌種有兩支藥搶病人、兩個開發方案搶預算、外部授權資產與內部管線搶順位。 最後一定有人留下，也一定有人被延後、出售或停止。 所以這筆交易若真發生，最重要的問題不是新公司有多少管線，而是它敢不敢說清楚：誰是核心，誰成為代價。 ## 04｜現在走到哪一步？報導是真的，交易還沒成立 看到這裡，要把新聞的時態說清楚。 截至 2026 年 8 月 3 日，公開資訊支持的是：Financial Times 引述匿名知情人士，報導兩家公司近月曾探討結合。Reuters 隨後也報導這項消息。 但兩家公司沒有正式宣布合併，也沒有公布價格、換股比例、交易架構或簽署協議。阿斯特捷利康對媒體拒絕評論，BMS 沒有立即回覆；SEC 公開文件中，也沒有一筆可以稱作「4,000 億美元收購案」的正式交易。 因此，4,000 億美元不是出價，也不是成交價。 它是兩家公司在報導當時的合計市值量級。若未來採換股合併，真正的代價會藏在換股比例、控制權與股東稀釋裡；若由其中一方收購另一方，才需要進一步看溢價、融資與債務。 換句話說，史上最大級藥廠合併「有可能」正在被討論，但還沒有一份正式文件，能把可能改寫成已經。 這個界線不會削弱故事，反而讓接下來的每個訊號更值得看。 *圖 3｜產品拼圖既有互補，也在肝癌雙免疫與多發性骨髓瘤 CAR-T 等領域出現需要管理的重疊。* ## 05｜就算談成，還要過反壟斷這一關 這種規模的交易，監管機關不可能只看兩家公司今天各賣多少藥。 它還會看後期管線、未來競爭，以及一樁交易會不會讓原本可能互相挑戰的資產，提前消失在同一家公司內。 2019 年 BMS 收購 Celgene 時，美國 FTC 就因 Otezla 與 BMS 當時的在研 TYK2 資產可能形成競爭，要求剝離 Otezla。Amgen 最後以 134 億美元接手，成為當時極具代表性的藥廠合併剝離案例。 這個先例不能直接推論阿斯特捷利康與 BMS 一定被擋，也不能預測哪支藥一定會被賣。 但肝癌雙免疫、血液腫瘤、CAR-T，以及多個實體瘤裡的免疫與 ADC 組合，很可能逐項被檢視。 反壟斷之外，還有更難量化的整合。 誰當執行長？研發總部在哪裡？英國與美國的上市、治理怎麼安排？阿斯特捷利康股東願意承擔多少稀釋？BMS 股東又要多少溢價，才願意在專利懸崖真正發生以前交出控制權？ 市值相加，一個試算表就能完成。 讓兩個龐大組織不拖慢彼此，才是真正昂貴的部分。 ## 06｜台灣有真實角色，但不是亂貼概念股 這種消息一出，台股最常見的反應，是把「做 ADC」「會做抗體」「有 CDMO」的公司全部拉進來。 目前沒有足夠的一級資料，能證明哪一家台灣上市公司直接擁有這樁傳聞的權利、確定訂單或合併受益。 真正可驗證的台灣角色，在臨床試驗。 BMS 的 iza-bren 相關試驗列有台大、北榮、中榮、成大、義大與北醫等台灣中心；阿斯特捷利康的 AZD0901 試驗也列有台北研究地點。阿斯特捷利康的 HIMALAYA 肝癌研究，歷史登錄同樣能看到台灣中心。 這代表台灣不是站在全球腫瘤開發外面看新聞，而是真正參與跨國試驗、患者收案與資料產生。 如果兩家公司未來合併，對台灣最實際的觀察不是「哪一檔先漲」，而是：既有試驗會不會延續、外部授權資產會不會重新排序，以及臨床開發與供應鏈決策會不會更集中。 對與大型藥廠合作的台灣生技公司來說，合作方變大不一定代表資源變多。有時候，它只代表自己的專案，在對方更大的管線裡變得更小。 *圖 4｜媒體報導存在不等於交易成立；正式條款、治理安排、監管審查與資產排序都還沒有答案。* ## 結語｜4,000 億美元，買得到規模，買得到時間嗎？ 阿斯特捷利康與 BMS 的組合，確實有產業邏輯。 一方要追 2030 年 800 億美元營收目標，想縮短美國化與產品擴張的時間；另一方握著龐大現金流與成熟商業網絡，卻必須趕在核心大藥獨占期轉折前完成接棒。 如果成真，它可能改寫全球腫瘤版圖，也可能成為史上最大規模的藥廠合併之一。 但產品越完整，重疊也越難處理；公司越龐大，整合越容易拖慢決策。買下 BMS，可以替阿斯特捷利康一次補上很多能力；買下之後能不能保留這些能力，則是另一回事。 接下來真正值得等的，不是更多「知情人士」形容詞，而是正式公告裡的四個答案：交易怎麼做、誰掌權、哪些資產要賣，以及合併後哪些管線會被放在最前面。 4,000 億美元能買到一張全球最完整的藥物地圖。 它能不能買到下一個十年的成長，現在還沒有答案。 ## 免責聲明 本文為生技醫藥與產業趨勢整理，不構成醫療診斷、治療建議或投資建議。媒體報導、公司策略、交易談判、監管審查、藥物研發與商業化皆可能改變，實際判斷應以公司、監管機關與一級文件的最新公開資訊為準。 --- ## Tau Earned a Retest: Why Biogen Is Taking Diranersen Into Phase 3 After CELIA Missed Its Primary Endpoint URL: https://drugnews.com.tw/articles/2026-08-03-diranersen-celia-tau-phase3-en.html Date: 2026-08-03 Language: en Summary: CELIA missed its prespecified dose-response primary endpoint, yet the lowest dose produced a 0.54-point CDR-SB difference, or 26% slowing, while CSF total tau and tau PET moved in the expected direction. That is enough to justify Phase 3 planning, but not enough to declare tau a validated second disease-modifying pathway. A Phase 2 study missed its primary endpoint, yet the sponsor says it intends to move the drug into confirmatory Phase 3 testing. That can sound like an exercise in explaining away failure. But after Biogen presented the complete CELIA dataset for diranersen at the 2026 Alzheimer's Association International Conference, the discussion did not end with the word “failed.” It reopened a more consequential question: can lowering tau genuinely alter the course of early Alzheimer's disease? The most defensible interpretation sits between two extremes. Tau has not made a clinical comeback, but the failed primary endpoint did not erase every other signal in the dataset. Drugnews' central judgment is this: Tau has not been validated. CELIA failed to show that higher or more frequent dosing produced greater benefit, so it missed the prespecified dose-response primary endpoint. At the same time, the lowest dose was associated with 26% less worsening on CDR-SB, while cerebrospinal-fluid tau and tau PET moved in the direction predicted by the mechanism. That combination gives Biogen and Ionis a rational basis to plan Phase 3. It does not establish a second disease-modifying pathway beyond amyloid-beta; one confirmatory trial is still missing. The statistical problem, the clinical signal and the biomarker consistency all appear in the same dataset. The primary endpoint did not somehow turn positive. What CELIA did was clarify the questions the next study must answer. ## 01 | Start With the Right Question: CELIA Tested Dose Response Diranersen, also known as BIIB080 and IONIS-MAPTRx, is an antisense oligonucleotide developed by Biogen and Ionis. It binds MAPT messenger RNA with the aim of reducing tau production before the protein is made. That is different from an antibody designed to capture a selected form of tau that already exists. To reach the central nervous system, diranersen is administered intrathecally through lumbar puncture into the cerebrospinal fluid. The route bypasses the blood-brain barrier, but it also introduces procedural burden and procedure-related adverse events. CELIA was a global, randomized, double-blind, placebo-controlled Phase 2 dose-finding study. It enrolled 416 people with mild cognitive impairment due to Alzheimer's disease or mild Alzheimer's dementia. The placebo-controlled period lasted 76 weeks, or roughly 18 months. None of the participants had previously received anti-amyloid therapy. The three diranersen regimens were: - 60 mg every six months; - 115 mg every six months; and - 115 mg every three months. The primary endpoint asked whether the Week 76 change in CDR-SB demonstrated a dose response. Under the original hypothesis, a higher dose or more frequent dosing should have produced greater clinical benefit. The result went the other way. The best-performing group received the lowest and least frequent regimen: 60 mg every six months. CELIA therefore did not meet its primary endpoint. Biogen explicitly acknowledged that statistical conclusion in its formal disclosure. *Figure 1 | CELIA was a 416-participant, 76-week, randomized, double-blind, placebo-controlled dose-finding study. Higher dosing was not better, so the prespecified dose-response endpoint was missed.* ## 02 | Where Did 26% Come From? Read the Point Difference First CDR-SB is a common Alzheimer's disease scale that integrates cognitive and functional domains. Higher scores indicate more severe disease. The study compares how much each group worsens over time. In the 60 mg every-six-months group, Biogen reported 0.54 points less worsening on CDR-SB than placebo at Month 18. Expressed as a relative effect, that corresponds to 26% slowing. The dose group included 60 participants, while the placebo group included 115. The two other regimens also moved in the favorable direction, but by smaller amounts: - 115 mg every six months: 0.28 points less worsening, or 14% slowing; - 115 mg every three months: 0.18 points less worsening, or 9% slowing. That ordering is precisely why the primary endpoint failed. The study showed a possible drug signal, but it contradicted the prespecified assumption that greater exposure would produce greater CDR-SB benefit. Several explanations are possible. Tau lowering may have an optimal dose window. The pattern may also reflect sample size, patient heterogeneity, measurement variability or chance. The most direct alternative is that the apparent advantage of the lowest dose may not reproduce in a larger study. Phase 2 has exposed those uncertainties rather than resolved them. “The lowest dose works best” remains a hypothesis, not an established biological explanation. Conversely, the fact that all three groups moved in the same direction should not be discarded simply because the trend was not linear. Biogen also reported several cognitive and composite outcomes. In the 60 mg group, ADAS-Cog13, MMSE, modified iADRS and ADCOMS generally favored diranersen, and the company said most comparisons reached nominal statistical significance. The word “nominal” matters. With many secondary comparisons, p-values that are not protected by a complete multiplicity-control strategy do not carry the confirmatory status of a successful primary endpoint. At the same time, ADCS-ADL-MCI did not separate from placebo in any of the three dose groups. The cognitive signal was therefore not reproduced uniformly across functional measures. The 26% figure is reportable, but it must remain attached to three qualifications: the primary dose-response endpoint was missed; the signal came from the lowest dose; and functional outcomes were not consistently aligned. *Figure 2 | The 60 mg every-six-months group showed a 0.54-point difference, or 26% slowing. Higher doses produced smaller differences, which is the reason the primary endpoint failed.* ## 03 | Why the Tau Debate Reopened: Three Evidence Layers Moved Together If CELIA had produced only one attractive clinical percentage, the case for Phase 3 would be much weaker. The dataset drew attention because mechanism, biomarkers and part of the clinical readout pointed in the same direction. Diranersen is designed to reduce MAPT messenger RNA so that cells produce less tau. If the drug engages its target, the most direct evidence should be a reduction in tau-related cerebrospinal-fluid markers. Across the three CELIA dose groups, mean CSF total tau fell by approximately 50% to 65% from baseline. That result indicates that the drug altered its intended molecular marker. The tau PET substudy adds another layer. It included 131 participants, and Biogen reported reductions from baseline across all assessed brain regions in each of the three dose groups. Tau PET measures a signal associated with tau pathology in the brain. A change in both imaging and cerebrospinal-fluid markers provides more information than either fluid or imaging alone. The images in the accompanying graphic are scientific illustrations, not scans from individual CELIA participants. Biomarker improvement does not automatically mean that patients live or function better. Alzheimer's drug development offers many examples in which a biological marker moved while clinical function did not move to the same degree. CELIA nevertheless provides a chain worth testing: diranersen reduced tau production, CSF total tau and tau PET declined, and selected cognitive and CDR-SB outcomes moved in a favorable direction. That chain supports further validation. It should not be promoted by commentators into regulatory confirmation, nor does it establish tau as a validated second disease-modifying pathway. Biogen used “proof of concept” in its own interpretation of the data; responsible analysis must preserve the attribution to the company. *Figure 3 | MAPT mRNA, CSF total tau, tau PET and selected clinical outcomes form a coherent chain. Biomarker reduction still cannot substitute for confirmatory evidence of patient benefit.* ## 04 | Reducing Tau Production Is Different From Clearing Existing Tau Tau has been an important Alzheimer's target for years. The protein helps stabilize neuronal microtubules. In Alzheimer's disease and other tauopathies, tau becomes abnormally modified, aggregates and spreads with disease progression. Many earlier programs used antibodies against extracellular tau in an effort to interrupt propagation, but those approaches did not reliably translate into reproducible clinical benefit. An antisense oligonucleotide acts further upstream by reducing production of tau isoforms through a shared messenger-RNA target. In principle, that can affect both intracellular and extracellular tau without requiring a separate antibody for each conformation or aggregate. Being upstream creates a different set of questions. Tau has physiological functions in healthy neurons. How much long-term reduction is safe? Is more reduction necessarily better? The fact that CELIA's lowest dose produced the strongest apparent clinical signal is a reminder that biomarker depth and clinical benefit may not have a simple linear relationship. Intrathecal administration also requires repeated lumbar puncture. If the 60 mg every-six-months regimen reproduces in Phase 3, the treatment burden may be manageable for some patients. If development ultimately requires a loading regimen, a different maintenance schedule or more intensive monitoring, real-world access will look different. The optimal dose window, cumulative safety and the treatment burden patients are willing to accept all require longer and larger studies. ## 05 | Safety Was Acceptable So Far, but ARIA Cannot Be Declared Impossible During the 76-week placebo-controlled period, Biogen characterized diranersen as generally well tolerated. Most adverse events were mild to moderate and non-serious, and none resulted in treatment discontinuation or study withdrawal. Common events included procedural pain, post-lumbar-puncture syndrome and confusional state. Most confusional events occurred within days of dosing and resolved within a week. Those events are closely related to the intrathecal route. If the program enters Phase 3, investigators will need to assess not only the molecule but also the capacity to deliver repeated lumbar punctures, patient acceptance and cumulative procedural risk. The company also said that amyloid-related imaging abnormalities associated with anti-amyloid antibodies were not expected based on a tau-targeting mechanism, and that CELIA was consistent with that expectation. This is still the company's interpretation based on mechanism and the available dataset. It cannot guarantee that every tau-directed drug will always be free of ARIA. Monitoring methods, study population and duration of exposure all shape what a trial can detect. We have also excluded a secondary-conference claim about “11 studies” and “0.9% severe ARIA” because no official abstract was available for verification. High-risk medical statements should not be built from untraceable meeting summaries. ## 06 | Do Not Rank CELIA's 26% Beside Leqembi and Kisunla The 26% figure invites an immediate comparison with relative slowing reported for approved anti-amyloid therapies. A side-by-side ranking may be visually appealing, but it creates a false equivalence. CELIA was a 416-participant Phase 2 dose-finding study with a dose-response primary endpoint. Other programs differ in patient eligibility, background therapy, follow-up duration, scale analysis, missing-data handling and statistical hierarchy. The fact that two trials use CDR-SB does not make their relative percentages interchangeable. The clinical meaning of a 0.54-point difference also needs longer follow-up and more functional evidence. Alzheimer's disease is progressive, so a slower rate of decline could accumulate into more time living independently. But if the difference remains confined to scales and is not reflected consistently in functions patients and caregivers can recognize, its value will be limited. The more useful question is whether Phase 3 can reproduce the 0.54-point difference, demonstrate aligned cognitive and functional benefit, and do so with acceptable safety and treatment burden. ## 07 | Why Is Biogen Willing to Proceed? Biogen said it plans to advance diranersen into a confirmatory Phase 3 program. That is a company development decision. The FDA has not publicly endorsed a study design, and approval is not assured. From a development perspective, the rationale is understandable: - all three CDR-SB dose groups favored diranersen, even though higher doses were not better; - the lowest-dose signal appeared across several cognitive and composite measures; - CSF total tau and tau PET demonstrated effects on the intended molecular and pathological markers; - 60 mg every six months provides a plausible regimen to test further; and - more than 90% of participants who completed the controlled period elected to enter the extension, which at least indicates willingness to continue. Phase 3 will still be difficult to design. If Biogen selects 60 mg, it must show that the apparent low-dose advantage was not chance. If it includes additional doses, it must explain why CELIA did not produce a clear dose response. The primary endpoint, sample size, follow-up duration, permitted background anti-amyloid therapy, and the way cognition and function are combined can all change the answer. Phase 3 must remove the most important uncertainties from Phase 2 rather than replay the same presentation. *Figure 4 | Advancing to Phase 3 is a development decision, not FDA approval. A confirmatory trial must resolve the central uncertainty left by CELIA.* ## 08 | What Should Taiwan-Based Readers Watch? As of July 19, 2026, first-party public disclosures did not show any Taiwan-listed company holding direct co-development, manufacturing, delivery or commercialization rights to diranersen. The most direct public-market exposures are Biogen, listed on Nasdaq as BIIB, and the molecule's originator Ionis Pharmaceuticals, listed as IONS. Biogen exercised its licensing option in 2019 and obtained exclusive worldwide, royalty-bearing development and commercialization rights. That is why both companies have a direct relationship to Phase 3 execution, milestones and commercial value. Taiwan's industry can still follow three capability areas: long-term manufacturing and quality for central-nervous-system antisense oligonucleotides, standardization of cerebrospinal-fluid biomarkers and tau PET, and large-trial execution that can support repeated intrathecal dosing. A local company should enter the investment narrative only after a formal collaboration, trial-site, supply or licensing disclosure appears. ## 09 | Conclusion: Tau Earned a Retest CELIA's failed primary endpoint cannot be erased. Neither should the 0.54-point CDR-SB difference and 26% relative slowing in the 60 mg every-six-months group be treated as if they did not exist. The most defensible conclusion is: Diranersen has given the tau-lowering strategy a reason to enter Phase 3. Clinical validation still depends on a confirmatory trial. Five items now matter: whether the Phase 3 design receives constructive regulatory feedback, which dose is selected, whether the primary framework captures both cognition and function, how the trial accommodates an evolving Alzheimer's treatment landscape, and whether repeated lumbar puncture and sustained tau reduction remain acceptable over time. Alzheimer's development does not lack elegant mechanisms. It lacks effects that reproduce in large trials and preserve a period of life that patients and families can recognize. Tau has received a serious retest, not a diploma. ## Disclaimer This article provides biotech, pharmaceutical and industry analysis. It does not constitute medical diagnosis, treatment advice or investment advice. Diranersen is investigational and has not been approved for marketing; clinical-trial results cannot be extrapolated to an individual patient. --- ## Tau 拿到補考資格：Diranersen 主終點失敗，為什麼 Biogen 仍要進三期？ URL: https://drugnews.com.tw/articles/2026-08-03-diranersen-celia-tau-phase3.html Date: 2026-08-03 Language: zh-Hant Summary: CELIA 未達預設劑量反應主要終點，最低劑量卻在 CDR-SB 看到 26% 惡化減緩，CSF total tau 與 Tau PET 也同向下降。這些訊號足以讓 Biogen 與 Ionis 規劃三期，還不足以宣布 Tau 已成為第二條獲確認的疾病修飾路線。 一個二期試驗沒有達到主要終點，公司卻說要把藥推進確認性三期。 這通常會被解讀成硬拗。但 2026 年阿茲海默症協會國際會議（AAIC）上，Biogen 公布 diranersen 的 CELIA 完整資料後，市場討論沒有停在「失敗」，反而重新問起：降低 Tau，是否真的可能改變早期阿茲海默症的病程？ 目前最合理的讀法介於兩個極端之間：Tau 尚未翻身；主要終點失敗，也沒有抹掉其餘訊號。 藥時事的主判斷是： Tau 還沒有正式翻身。CELIA 沒有證明劑量愈高、效果愈好，因而未達預設的劑量反應主要終點；最低劑量在 CDR-SB 看到 26% 的惡化減緩，腦脊髓液與 Tau PET 又同時往靶點預期方向移動。這組訊號足以支持 Biogen／Ionis 規劃三期；要宣布「Aβ 之外的第二條疾病修飾路線成立」，還少一場確認性試驗。 同一份資料裡，統計問題、臨床訊號與生物標記一致性同時出現。主終點失敗沒有突然變成功；它讓下一步要回答的問題變得更清楚。 ## 01｜先把試驗問題說對：CELIA 問的是劑量反應 Diranersen 也叫 BIIB080、IONIS-MAPTRx，是 Biogen 與 Ionis 共同開發的反義寡核苷酸（ASO）。它和 MAPT mRNA 結合，目的是在蛋白質製造之前降低 Tau 生成；作用位置不同於抓取既有 Tau 聚集體的抗體。 這類藥物要進入中樞神經系統，目前採用鞘內注射，也就是透過腰椎穿刺把藥送進腦脊髓液。它繞過血腦障壁，卻也帶來程序負擔與相關不良反應。 CELIA 是全球 Phase 2、隨機、雙盲、安慰劑對照、劑量探索研究，納入 416 名因阿茲海默症造成的輕度認知障礙，或輕度阿茲海默症失智患者。安慰劑對照期 76 週，也就是約 18 個月。所有入組者先前都沒有接受抗類澱粉蛋白治療。 三種 diranersen 給藥方案是： - 60 mg，每 6 個月一次； - 115 mg，每 6 個月一次； - 115 mg，每 3 個月一次。 試驗的主要終點是看第 76 週 CDR-SB 的變化是否呈現劑量反應。照原本假設，更高或更頻繁的劑量，應該帶來更大的臨床效益。 結果偏偏相反。表現最好的是最低、最不頻繁的 60 mg 每 6 個月組。 因此，CELIA 沒有達到主要終點。Biogen 已在正式公告明確承認這個統計結論。 *圖1｜CELIA 是 416 人、76 週的隨機雙盲安慰劑對照劑量探索研究；更高劑量沒有帶來更大效益，預設劑量反應主要終點未達。* ## 02｜26% 從哪裡來？先看點數，再看百分比 CDR-SB 是阿茲海默症研究常見量表，綜合認知與日常功能面向。分數愈高代表疾病愈嚴重；研究比較的是一段時間後，各組惡化多少。 在 60 mg 每 6 個月組，Biogen 報告相較安慰劑，第 18 個月 CDR-SB 少惡化 0.54 分，換算成相對減緩幅度是 26%。這一組 n=60，對照組 n=115。 另外兩組也往有利方向移動，但幅度較小： - 115 mg 每 6 個月：少惡化 0.28 分，減緩 14%； - 115 mg 每 3 個月：少惡化 0.18 分，減緩 9%。 這個排列正是主要終點失敗的原因：藥物看似有訊號，結果卻違反「劑量愈高、效果愈大」的預設。更高劑量沒有帶來更大 CDR-SB 效益。 為什麼會出現這種結果？可能性很多。也許 Tau 降低存在最佳劑量窗；也可能是樣本數、患者差異、量表波動或偶然性造成。還有一種更直接的可能：最低劑量的效果在更大研究裡無法重現。 二期試驗把這些不確定性攤開。「低劑量最好」仍是假說，不是已被證實的生物學解釋；三組同方向的臨床訊號，也不該因趨勢不像直線而被抹掉。 Biogen 還公布多個認知與複合終點。60 mg 組在 ADAS-Cog13、MMSE、modified iADRS、ADCOMS 等指標多數往有利方向，公告稱大部分差異達到名目統計顯著（nominal statistical significance）。「名目」兩字不能省略，因為在多重終點裡反覆比較，若沒有完整控制多重檢定，p 值不能等同主要終點的確證性證據。 同時，ADCS-ADL-MCI 這個日常功能量表在三個劑量組都沒有和安慰劑分離。也就是說，認知訊號並沒有在所有功能指標上整齊複製。 所以 26% 可以寫，但必須和三句話綁在一起：主要劑量反應終點未達；這是最低劑量組；日常功能指標並非全面一致。 *圖2｜60 mg 每 6 個月組少惡化 0.54 分、相對減緩 26%；更高劑量的差異反而較小，正是主要終點失敗的原因。* ## 03｜三層證據同方向，Tau 討論才會升溫 如果 CELIA 只有一個漂亮的臨床百分比，Biogen 要推進三期的說服力會弱很多。這次更受關注的原因，是藥物機制、生物標記與部分臨床結果出現在同一條方向上。 Diranersen 的設計，是降低 MAPT mRNA，讓細胞少製造 Tau。若藥物真的命中靶點，最直接的證據應該是腦脊髓液中的 Tau 下降。 CELIA 三個劑量組的腦脊髓液 total tau，相較基線平均下降約 50%–65%。這表示藥物確實改變了預定的分子標記。 更進一步，Tau PET 子研究納入 131 人。Biogen 表示，三種劑量在所有評估腦區都看到 Tau PET 相較基線下降。PET 觀察的是腦內 Tau 病理訊號；它與腦脊髓液標記同時下降，比單一體液標記提供更多資訊。配圖中的 PET 圖像為科學示意，不是 CELIA 個別患者掃描影像。 標記下降無法自動換算成患者活得更好。阿茲海默症研發史上，生物標記改善、臨床功能卻沒有相應變化的例子並不少。CELIA 提供了一個值得確認的鏈條：藥物降低 Tau 生成，CSF total tau 與 Tau PET 下降，部分認知與 CDR-SB 指標同方向。 這條鏈條可以叫「支持進一步驗證」，不能由我們替監管機關升格成已完成概念驗證，更不能說 Tau 已成為獲確認的第二條疾病修飾路線。Biogen 在公告中使用 proof of concept，是公司的資料詮釋；公開文章要把這個歸屬寫清楚。 *圖3｜MAPT mRNA、CSF total tau、Tau PET 與部分臨床訊號形成一致鏈條；生物標記下降仍不能取代確認性三期的患者效益證據。* ## 04｜降低生成，和過去清除 Tau 有何不同？ Tau 早已是阿茲海默症研發的重要靶點。它和神經元骨架穩定有關；在阿茲海默症等 tauopathy 裡，Tau 會異常修飾、聚集並隨疾病擴散。過去許多計畫想用抗體清除細胞外 Tau，希望阻止病理傳播，卻沒有順利轉成可重複的臨床效益。 ASO 從共同上游減少各種 Tau 亞型生成，不需挑選單一 Tau 構形。理論上，這可以同時影響細胞內與細胞外 Tau，也不必為每種聚集形態各做一支抗體。 但「更上游」也帶來新的問題。Tau 在正常神經元有生理功能，長期降低多少才安全？降得越多是否越好？CELIA 最低劑量反而表現最好，正提醒我們：標記降低幅度和臨床效益未必是簡單線性關係。 另外，鞘內注射需要腰椎穿刺。60 mg 每 6 個月一次的頻率若能在三期重現，給藥負擔可能還算可管理；若未來需要不同起始療程、維持策略或更密集監測，真實世界可近性會改變。 最佳劑量窗、累積安全性與患者願意接受的治療負擔，都要交給更長期、更大規模研究回答。 ## 05｜安全性目前可接受，ARIA 仍不能保證為零 在 76 週安慰劑對照期，Biogen 表示 diranersen 整體耐受性良好。多數不良事件是輕至中度、非嚴重，且沒有造成停藥或退出。常見事件包括程序性疼痛、腰椎穿刺後症候群與意識混亂狀態；多數意識混亂事件在給藥後數日內出現，並在一週內緩解。 這些事件和鞘內給藥路徑密切相關。未來產品若進三期，除了藥物本身，腰椎穿刺的執行量能、患者接受度與重複給藥風險都要一起評估。 公司另稱，依 Tau 標靶機制，預期不會出現抗 Aβ 抗體相關的 ARIA，CELIA 結果也符合這個預期。這仍是公司基於機制與現有資料的判讀，無法保證所有 Tau 藥物永遠不會出現 ARIA。 不同試驗的監測方法、族群與暴露時間都會影響安全性觀察。本文也不引用沒有官方摘要可核對的「11 項試驗、0.9% 嚴重 ARIA」數字，避免把會議二手轉述混進高風險醫療主張。 ## 06｜別把 26% 和 Leqembi、Kisunla 直接排隊 看到 26%，很多人會立刻把它和已核准抗 Aβ 藥物的相對減緩百分比放在同一張表。這種圖很吸睛，卻容易製造錯誤等號。 CELIA 是 416 人的二期劑量探索研究，主要終點是劑量反應；不同藥物的患者條件、背景治療、追蹤時間、量表分析、缺失資料處理與統計階層都不同。即使同樣使用 CDR-SB，相對百分比也不能直接排成療效名次。 更重要的是，0.54 分代表什麼臨床感受，仍需要更長追蹤與更多功能資料回答。阿茲海默症是持續惡化的疾病，較慢惡化可能隨時間累積成更長的獨立生活；但若差異只停留在量表、沒有穩定反映到患者與照顧者可感受的功能，價值也會受限。 更有意義的判讀是：三期能否重現 0.54 分差異，能否同時改善認知與功能，以及治療負擔和安全性是否可接受。 ## 07｜Biogen 為什麼還敢進三期？ Biogen 表示，將把 diranersen 推進確認性 Phase 3。這是公司開發決策；FDA 尚未公開同意試驗設計，核准更沒有保證。 從開發邏輯看，理由可以理解： - 三個劑量的 CDR-SB 都往有利方向，雖然高劑量沒有更好； - 最低劑量的臨床訊號跨多個認知／複合指標出現； - CSF total tau 與 Tau PET 證實靶點和病理標記確實被改變； - 60 mg 每 6 個月一次提供可進一步驗證的劑量候選； - 完成對照期的患者中，超過 90% 選擇進入延伸研究，至少顯示持續參與意願。 但三期也會非常難設計。若選 60 mg，必須證明二期的最低劑量優勢不是偶然；若加入其他劑量，又要解釋二期為何沒有清楚劑量反應。主要終點、樣本數、追蹤時間、是否允許背景抗 Aβ 治療，以及認知和功能如何共同判定，都可能改變試驗答案。 三期要消除二期最重要的不確定性，而非重播同一份簡報。 *圖4｜進入三期是公司開發決策，不是 FDA 核准；確認性試驗必須消除二期留下的核心不確定性。* ## 08｜台灣讀者該看什麼？ 查證截至 2026 年 7 月 19 日，公開一級資料沒有顯示台灣上市櫃公司直接擁有 diranersen 的共同開發、製造、遞送或商業化權利。 最直接的資本市場標的是 Biogen（Nasdaq: BIIB）與藥物發現者 Ionis Pharmaceuticals（Nasdaq: IONS）。Biogen 在 2019 年行使授權選擇權，取得全球獨家、附權利金的開發與商業化權利；這也是為什麼兩家公司都和後續三期、里程碑與商業價值直接相關。 台灣產業可追三種能力：中樞 ASO 的長期製造與品質、腦脊髓液生物標記與 Tau PET 的標準化，以及可支撐重複鞘內給藥的大型臨床執行。只有出現正式合作、試驗中心、供應或授權公告，才適合把本地公司名字放進來。 ## 09｜結論：Tau 拿到補考資格 CELIA 的主要終點失敗，不能擦掉；60 mg 每 6 個月組在 CDR-SB 少惡化 0.54 分、相對減緩 26%，也不能因為主要終點失敗就當作不存在。 最合理的解讀是： Diranersen 讓 Tau 降低策略拿到進三期的理由，臨床驗證仍待確認性試驗。 接下來請盯五件事：三期設計是否獲得監管溝通支持、選哪個劑量、主要終點是否同時涵蓋認知與功能、能否容納新的阿茲海默症治療背景，以及腰椎穿刺與長期 Tau 降低的安全性。 阿茲海默症研發不缺漂亮機制，缺的是能在大試驗重現、讓患者多保留一段生活能力的效果。Tau 這次拿到的是一張值得認真看的補考單，離畢業還有一場大試驗。 ## 免責聲明 本文為生技醫藥與產業趨勢整理，不構成醫療診斷、治療建議或投資建議。Diranersen 為研究中藥物，尚未獲准上市；臨床結果不能外推至個別患者。 --- ## Bempikibart in Alopecia Areata: A 33-Patient Signal Is Not Yet a Win Over JAK Inhibitors URL: https://drugnews.com.tw/articles/2026-08-02-bempikibart-alopecia-areata-en.html Date: 2026-08-02 Language: en Summary: Q32 Bio nearly doubled after a 33-patient, single-arm Phase 2a readout. The investable questions are whether efficacy survives randomization, whether responses persist after treatment stops, and whether a larger safety database can support a differentiated place beside JAK inhibitors. A Phase 2 trial with just 33 patients and no placebo group nearly doubled the market value of a small biotechnology company in a single day. On July 13, 2026, Q32 Bio reported topline results from Part B of SIGNAL-AA, evaluating bempikibart in severe alopecia areata. QTTB shares closed 90.72% higher that day. The next day, the company priced an offering at $18.25 per share, targeting approximately $200 million in gross proceeds to fund further clinical development. Hair regrew, the stock surged and the company raised capital. It is tempting to turn that sequence into a simple headline: a non-JAK therapy has finally beaten JAK inhibitors. That is not what the data establish. They establish that bempikibart has earned a more demanding test. Drugnews' central judgment is straightforward: Q32's near-doubling reflects how strongly the market wants a non-JAK treatment for alopecia areata. But bempikibart currently rests on a 33-patient, single-arm, open-label Phase 2a dataset. That evidence can move the program into a stricter controlled trial; it cannot show that the drug has already outperformed JAK inhibitors. The distinction matters because the next competitive round will ask more than whether hair grew back. It will ask whether efficacy replicates in a larger population, whether responses persist after treatment stops and whether the safety profile is sufficiently differentiated to change the clinical position already established by JAK inhibitors. *Figure 1 | Thirty-three patients were enough to change market expectations, not enough to establish mature comparative evidence.* ## 01 | First, understand what SALT measures Alopecia areata is an autoimmune disease in which the immune system attacks hair follicles. Its burden extends well beyond appearance. Disease can range from coin-sized patches to near-total loss of scalp hair, eyebrows, eyelashes or body hair, leaving patients with long-term uncertainty, social pressure and substantial psychological burden. Clinical trials commonly use the Severity of Alopecia Tool, or SALT, to measure the proportion of scalp affected by hair loss. SALT 0 means no scalp hair loss; SALT 100 means complete scalp hair loss. SIGNAL-AA Part B enrolled patients with severe or very severe disease, defined by baseline SALT scores of 50 to 100. Accordingly, SALT less than or equal to 20 means that no more than 20% of the scalp lacks hair at the assessment, or roughly 80% or more scalp coverage. It does not mean a 20% improvement, and it does not mean that every patient regrew 80% of their hair. Part B enrolled 33 patients, 36.4% of whom had previously received an oral JAK inhibitor. Everyone knew they were receiving bempikibart, and there was no concurrent placebo arm. The subcutaneous regimen began with four weekly 200 mg doses, followed by 200 mg every two weeks through 36 weeks, after which patients entered an off-treatment follow-up through Week 52. The prespecified primary analysis assessed mean percentage change from baseline in SALT at Week 36 in the modified intent-to-treat, or mITT, population. The proportion achieving SALT less than or equal to 20 was another prespecified efficacy analysis. Those design details define what the numbers can answer and what they cannot. ## 02 | Forty percent is striking, but the denominator was 25 Three Week 36 figures matter most: 1. In the prespecified 25-patient mITT population, mean SALT fell 35.3% from baseline. 2. Ten of 25 mITT patients, or 40.0%, achieved SALT less than or equal to 20. 3. Counting all 33 enrolled patients in the ITT population, 10 of 33, or 30.3%, achieved the same threshold. How can the same 10 responders produce 40.0% in one analysis and 30.3% in another? The mITT set excluded some participants who did not meet the rules for the primary analysis. That population was prespecified, but investors who remember only that “40% of patients regrew 80% of their hair” lose sight of the eight enrolled patients included in ITT but not in the mITT denominator. In a small trial, each patient materially changes the percentage. If one fewer patient had responded in the mITT set, 10/25 would become 9/25, or 36.0%. In the ITT set, 10/33 would become 9/33, or 27.3%. That is why an early Phase 2 signal cannot be read from a headline percentage alone: the numerator, denominator, missing-data rules and analysis population all belong in the same sentence. *Figure 2 | The same 10 responders produce 40.0% in mITT and 30.3% in ITT because the denominators differ.* More importantly, Part B was single-arm and open-label. Without a concurrent placebo, the study cannot exclude natural fluctuation, assessment bias, patient selection or other background effects. Without a head-to-head comparison against an approved JAK inhibitor, results from different trials cannot be arranged as if they were a comparative league table. The correct conclusion is that bempikibart showed an activity signal worth testing further. Whether it can outperform JAK inhibitors remains unanswered. ## 03 | Why a non-JAK mechanism excited the market Bempikibart is a fully human monoclonal antibody against IL-7 receptor alpha. IL-7Rα participates in both IL-7 and TSLP signaling, pathways involved in T-cell survival, differentiation and multiple immune-inflammatory responses. Q32 is attempting to modulate adaptive immunity upstream at IL-7Rα; JAK inhibitors act intracellularly across several cytokine signals. The mechanisms are distinct. That mechanism is attractive, but “more targeted” does not automatically mean “safer.” Safety differentiation must be established through enough patients, enough follow-up, controlled trials and, eventually, postmarketing experience. The initial Part B safety readout was relatively clean. Q32 reported no treatment-related serious adverse events and no treatment-related Grade 3 or higher adverse events. Injection-site reactions were most common, affecting 36.3% of patients; across all administered doses, their incidence was 4%. All were mild and resolved without intervention. Those observations are encouraging, but the safety database contains only 33 patients. If an important adverse event truly occurs at 1%, a 33-patient trial will often see none. Even a 3% event may not appear in a sample this small. “Not observed in this trial” is not the same statement as “the risk does not exist.” Larger exposure and longer follow-up sit between those two claims. *Figure 3 | A clean early safety signal is useful, but a small study is poorly equipped to detect uncommon, clinically important risks.* JAK inhibitors already have approved products, much larger exposure and more developed risk-management experience. AbbVie's upadacitinib offers a useful evidence-maturity benchmark. As of April 28, 2026, AbbVie had submitted a US FDA application for severe alopecia areata based on two randomized, double-blind, placebo-controlled Phase 3 studies totaling 1,399 patients. At that time the indication remained under review and should not be described as approved. The two evidence packages are fundamentally different. Directly comparing a 30.3% or 40.0% response rate from a 33-patient single-arm study with a historical figure from 1,399 Phase 3 patients would convert trial-design differences into an illusion of drug-effect differences. ## 04 | “Hair keeps growing after treatment stops” remains a question Another claim at risk of being overstated is durability. Q32's language was appropriately restrained. The company described “early durability signals” during off-treatment follow-up, with several patients maintaining or deepening their responses and one patient reaching SALT 0, meaning no scalp hair loss. Follow-up continues through Week 52, and full data are expected to be presented later at a medical meeting. One SALT 0 patient makes a compelling story, but one case cannot establish population-level durability. To answer whether benefit persists after withdrawal, readers need to know how many responders entered the off-treatment period, how long they had been off therapy, how many maintained response, how many relapsed, whether outcomes differed by baseline state, and whether patients received rescue therapy or entered an extension study. Q32's own risk disclosure states that additional data may not support its current durability expectations. That is not boilerplate to skip. It identifies the exact question the study has not yet answered. Part A extension data add another wrinkle by suggesting that maintenance treatment may be important. That points toward a different product proposition: if long-term injections are needed, convenience, pricing, adherence and reimbursement will look very different from a one-time or short-course therapy. The most accurate statement today is that early individual cases suggest response may be maintained or deepen off treatment, but durable benefit across the population has not been demonstrated. ## 05 | After the stock nearly doubled, the real test begins QTTB gained 90.72% on the day of the announcement and priced a $200 million public offering the next day. The market expressed expectation through price, and the company converted that expectation into clinical runway. Share price measures expectations, not therapeutic effect. A small biotech stock also reflects starting valuation, short interest, float, financing needs and future optionality. The magnitude of this move tells us how low the market's prior probability of success may have been. Drug efficacy still has to be answered by randomized controlled evidence. The financing also has two sides. Approximately $200 million in gross proceeds can support further bempikibart development, which is financially positive for the program. For existing shareholders, new shares and pre-funded warrants also mean dilution. Good data can allow a company to finance itself; they do not guarantee that every dollar raised becomes an approved product. Q32 plans to advance bempikibart into a registration-directed program in the first half of 2027. The operative word is plans. At the verification cutoff, no Phase 3 trial had been announced as initiated, and no regulator had approved the product. The next study needs to close at least four gaps: 1. Randomization, blinding and placebo or active control to reduce single-arm bias. 2. A larger sample with prespecified statistical testing of SALT less than or equal to 20 and other patient-relevant endpoints. 3. Sufficient on-treatment and off-treatment follow-up to separate continued dosing from true post-treatment durability. 4. A broader safety database, particularly for infection, immune-related risks and tolerability of long-term subcutaneous administration. *Figure 4 | The next test is not another headline: it is Week 52 durability, randomized control, reproducible efficacy and long-term safety.* ## 06 | The next competitive round in alopecia areata Interest in a non-JAK approach does not mean the market will simply replace one dominant mechanism with another. Nektar's rezpegaldesleukin takes a regulatory T-cell approach and is currently in Phase 2b for alopecia areata, not Phase 3. Other developers are probing additional T-cell costimulatory and cytokine nodes. Together, these programs show an industry trying to find a better balance among efficacy, long-term safety, recurrence and treatment burden. A new mechanism is only an entry ticket. Whether it becomes a clinical option depends on which patient-relevant problem it actually improves over existing therapies. One response percentage cannot support a claim that a drug will “replace JAK.” For patients, differentiation might mean less infection-risk tradeoff, fewer relapses after withdrawal or freedom from daily tablets. Clinicians need a predictable onset, manageable injection reactions and a clear switching strategy after failure. Payers need annual treatment cost, duration and evidence of persistence after stopping. If bempikibart offers similar efficacy but requires long-term injections at a higher price, it may fit only selected patients. Conversely, even without overwhelming average efficacy, consistent benefit in people who cannot use JAK inhibitors or accept their risks could create a genuine clinical position. Future studies should therefore prespecify JAK-experienced subgroups, time to relapse, patient-reported outcomes and maintenance-treatment requirements, not merely count how many patients reach SALT less than or equal to 20. Those data will determine whether bempikibart remains a new-mechanism story or becomes a treatment physicians can place in sequence. ## 07 | Industry coordinates for Taiwan investors The most direct listed exposure is US-listed Q32 Bio (Nasdaq: QTTB). As of July 19, 2026, no primary public source showed a Taiwan-listed company directly participating in bempikibart development, manufacturing, clinical execution or commercialization rights. AbbVie (NYSE: ABBV) and Nektar (Nasdaq: NKTR) are useful coordinates for different mechanisms and evidence maturity, not a single basket of supposed beneficiaries. ## 08 | Conclusion: separate “worth validating” from “already proven” Bempikibart's Week 36 data have clear value. In a 33-patient early study, 10 patients reaching SALT less than or equal to 20, a 35.3% mean SALT reduction in the primary analysis, and no observed treatment-related serious or Grade 3 or higher adverse events are enough to justify advancing the program. Its most important value today is that it supports a larger test. A claim that it has beaten JAK inhibitors still requires controlled evidence. Four items now deserve attention: complete Week 52 off-treatment data, whether the registration-directed trial actually begins, whether SALT less than or equal to 20 replicates after randomization, and whether safety remains favorable as exposure grows. Thirty-three patients nearly doubled the stock. Changing the treatment standard will require the next exam. ## Disclaimer This article provides biotechnology, pharmaceutical and industry analysis and does not constitute medical or investment advice. Bempikibart is investigational and has not been approved for marketing. Clinical-trial results do not predict an individual patient's outcome; treatment decisions should be made by qualified healthcare professionals using approved labeling and the patient's circumstances. --- ## 鬼剃頭新藥，先別封神 URL: https://drugnews.com.tw/articles/2026-08-02-bempikibart-alopecia-areata.html Date: 2026-08-02 Language: zh-Hant Summary: 33 人、單臂、開放標籤的 Phase 2a 資料讓 Q32 Bio 股價近乎翻倍；真正的問題是療效能否重現、停藥後能否維持，以及安全性是否足以改變 JAK 抑制劑的臨床位置。 一個只有 33 人、沒有安慰劑對照的二期試驗，讓一家小型生技公司的股價一天近乎翻倍（+90.72%）。 2026 年 7 月 13 日，Q32 Bio 公布 bempikibart 治療重度斑禿的 SIGNAL-AA Part B 頂線結果。當天 QTTB 收盤上漲 90.72%。隔天，公司就以每股 18.25 美元定價增資，預計募得約 2 億美元毛額，資金將支應 bempikibart 後續臨床開發。 頭髮長回來、股價漲上去、錢也募到了。這個故事很容易被寫成：「非 JAK 新藥終於打敗 JAK 抑制劑。」 資料真正說的是：這只是一張進入下一階段的門票。 藥時事的主判斷是： Q32 股價近乎翻倍，反映市場有多渴望非 JAK 的斑禿治療。Bempikibart 目前只有 33 人、單臂、開放標籤的 Phase 2a 資料；這張門票能把它送進更嚴格的對照試驗，還無法證明它已勝過 JAK。 這個區別很重要。因為斑禿的下一場競爭，不會只問「有沒有長頭髮」，而會問三件更難的事：療效能不能在大樣本重現？停藥後能不能維持？安全性是否真的足以換掉已經建立臨床位置的 JAK 抑制劑？ *圖1｜33 人讓股價近乎翻倍，但證據成熟度仍只支持進入下一階段。* ## 01｜先看清楚：SALT 代表什麼？ 斑禿是免疫系統攻擊毛囊造成的疾病，影響遠超過美容。它可能表現為硬幣大小的禿髮，也可能進展到幾乎整個頭皮、眉毛、睫毛甚至全身毛髮脫落。嚴重患者同時承受外觀改變、長期不確定性、社交壓力與心理負擔。 臨床試驗常用 SALT（Severity of Alopecia Tool）評估頭皮掉髮範圍。SALT 0 代表頭皮沒有掉髮；SALT 100 代表頭皮完全掉髮。SIGNAL-AA Part B 納入的是基線 SALT 50–100 的重度或極重度患者。 因此，新聞裡常見的「SALT≤20」，正確意思是試驗評估時頭皮掉髮範圍不超過 20%，也就是至少約 80% 頭皮有毛髮覆蓋。請勿把它誤讀成「改善 20%」，也不能說每位患者都恢復八成頭髮。 Q32 在這個 Part B 一共收了 33 人；其中 36.4% 曾接受口服 JAK 抑制劑。所有人都知道自己接受 bempikibart，研究也沒有同期安慰劑組。給藥方式是皮下注射：先每週 200 mg、共四劑，再每兩週 200 mg，總治療期 36 週，之後停藥追蹤至第 52 週。 試驗預先設定的主要分析，是 mITT 族群在第 36 週相較基線的 SALT 平均百分比變化；SALT≤20 比例屬於其他預設療效分析。 這些設計細節，決定了數字能回答什麼、不能回答什麼。 ## 02｜40% 很醒目，但分母只有 25 Q32 公布的第 36 週結果有三個最重要的數字： - 在預設的 mITT 族群 25 人中，SALT 相較基線平均下降 35.3%； - mITT 族群有 40.0%，也就是 10/25，達到 SALT≤20； - 把全部 33 名入組患者都算進 ITT，達到 SALT≤20 的比例是 30.3%，也就是 10/33。 為什麼同樣是 10 個人，比例會從 40% 變成 30.3%？ 因為 mITT 排除了部分未符合主要分析規則的受試者。這是預先設定的分析集；投資人若只記得「四成患者長回八成頭髮」，就會忽略八名被納入 ITT、卻沒有進入 mITT 分母的患者。 在小型研究裡，每一個人都會明顯改變比例。10/25 少一位應答者，就會從 40% 掉到 36%；而 10/33 若少一位，則會降到 27.3%。這也是為什麼二期早期訊號不能只看漂亮百分比，還要把分子、分母與缺失資料處理方式一起攤開。 *圖2｜同樣 10 名患者達標，分母從 25 人改成 33 人，答案就從 40.0% 變成 30.3%。* 更關鍵的是，Part B 是單臂、開放標籤。沒有同期安慰劑，就無法排除自然波動、評估偏差、患者選擇或其他背景因素對結果的影響；沒有和既有 JAK 抑制劑頭對頭，也不能把不同試驗的 SALT 數字排成排行榜。 所以，這批資料支持 bempikibart 出現值得進一步驗證的活性訊號；療效能否超越 JAK，現階段仍沒有答案。 ## 03｜非 JAK 為什麼讓市場這麼興奮？ Bempikibart 是全人源 IL-7Rα 單株抗體。IL-7Rα 同時參與 IL-7 與 TSLP 訊號，這兩條路徑和 T 細胞存活、分化，以及多種免疫發炎反應有關。Q32 選擇從 IL-7Rα 上游調節適應性免疫；JAK 抑制劑則在細胞內抑制多條細胞激素訊號，兩者路徑不同。 這個機制有吸引力，但「比較精準」無法自動換算成「比較安全」。安全性仍要靠足夠人數、足夠時間、對照試驗與上市後資料建立。 目前 Part B 的安全訊號相對乾淨：公司表示沒有治療相關的嚴重不良事件，也沒有治療相關的 Grade 3 以上不良事件。最常見的是注射部位反應，發生在 36.3% 的患者；若以所有給藥次數計算，發生率是 4%。所有注射部位反應都屬輕度，且不需處置便緩解。 這是好消息，但樣本只有 33 人。若某個重要不良事件的真實發生率是 1%，33 人試驗很可能完全看不到；即使真實發生率是 3%，也可能因樣本小而沒有出現。「本試驗未觀察到」和「沒有這種風險」之間，隔著更大樣本與更長追蹤；醫藥內容不能跨過這道證據缺口。 *圖3｜安全訊號乾淨是好消息；小樣本仍難以看見低頻但重要的長期風險。* JAK 抑制劑已有核准產品、更多患者暴露與較長期的風險管理經驗。以 AbbVie 的 upadacitinib 為例，截至 2026 年 4 月 28 日，公司已向美國 FDA 申請擴增重度斑禿適應症，資料來自兩個隨機、雙盲、安慰劑對照的第三期研究，總計 1,399 人；但當時該斑禿適應症仍在審查，不能寫成已核准。 兩套證據的成熟度根本不同。拿 33 人單臂研究的 30.3% 或 40%，直接和 1,399 人第三期的歷史數字比高低，會把試驗設計差異假裝成藥效差異。 ## 04｜「停藥後還在長」目前只是一個問號 這次最容易被誇大的另一句，是 bempikibart 可能具有「持久性」。 Q32 的原文很克制：停藥追蹤期間出現「早期持久性訊號」，多名患者維持或加深反應，其中一人達到 SALT 0，也就是頭皮完全沒有掉髮。完整的停藥追蹤仍在進行，預計到第 52 週，完整資料之後才會在醫學會議公布。 一位患者的 SALT 0 很有故事性，單一個案卻無法證明整體持久性。要回答停藥後能否維持，至少要知道：有多少應答者進入停藥期、停藥多久、多少人維持、多少人復發、不同起始狀態是否有差異，以及是否有救援治療或進入延伸研究。 Q32 自己也在風險揭露中明說，額外資料可能不支持公司目前對持久性的期待。這句法律文字不該被當成例行公事略過，因為它正好點出試驗尚未回答的核心。 更有意思的是，Part A 延伸資料提示「維持治療可能很重要」。這指向另一個產品命題：若未來需要長期注射，便利性、定價、依從性與支付條件，都會和一次性或短期療程的想像不同。 因此目前最準確的說法是：看到了停藥後維持或加深反應的早期個案訊號，但尚未證明整體族群有持久效益。 ## 05｜股價近乎翻倍以後，真正的考試才開始 QTTB 在資料公布當天上漲 90.72%，隔天就完成 2 億美元公開發行定價。市場用價格表達期待，公司也把期待換成了臨床跑道。 股價反映市場期待，無法代替療效判讀。小型生技股價會同時受原先估值、空頭部位、流通股數、融資需求與未來選擇權影響。這次反應可以告訴我們「市場原本把成功機率估得多低」；藥效仍要交給隨機對照試驗回答。 這筆增資同樣要兩面看。對公司而言，約 2 億美元毛額能支應 bempikibart 後續試驗，是正面財務事件；對既有股東而言，新股與預付認股權證也代表稀釋。好資料讓公司有能力募資，並不等於每一美元資金都會轉成核准產品。 Q32 表示，計畫在 2027 年上半年把 bempikibart 推進註冊導向（registration-directed）計畫。關鍵字是「計畫」。截至查證日，公司尚未宣布第三期已啟動，也沒有監管機關核准這個產品。 下一個研究至少要補上四個缺口： 1. 隨機、盲性與安慰劑或有效對照，確認效益能排除單臂研究偏差； 2. 更大的樣本與預先設定的統計檢定，確認 SALT≤20 與其他患者重要終點； 3. 足夠長的治療與停藥追蹤，拆清「持續給藥」和「停藥持久」； 4. 更完整的安全性暴露，特別是感染、免疫相關風險與長期皮下注射耐受性。 *圖4｜下一場考試要補上第 52 週持久性、隨機對照、療效重現與長期安全。* ## 06｜斑禿治療的下一輪競爭 非 JAK 路線之所以值得關注，並不代表市場會從單一機制換成另一個單一機制。 Nektar 的 rezpegaldesleukin 走調節型 T 細胞路線，目前斑禿開發期別是 Phase 2b，尚未進入第三期。其他公司也在探索不同的 T 細胞共刺激或細胞激素節點。這些計畫共同反映：產業想在療效、長期安全、復發與給藥負擔之間，找到比現在更好的平衡。 「新機制」只是進場券。最後能否成為臨床選項，要看它和既有治療相比，究竟改善了哪一個患者真正感受得到的問題。 要談「替代 JAK」，單一百分比遠遠不夠。對患者而言，可能是少做一次感染風險權衡、停藥後少一次復發，或不必每天服藥；對醫師而言，則要有可預測的起效時間、能管理的注射反應，以及清楚的失敗後轉換策略；對支付方而言，還要看一年療程成本、需要治療多久、停藥後是否維持。若 bempikibart 只有類似療效，卻需要長期皮下注射且價格更高，它仍可能只適合特定族群。反過來說，即使平均療效沒有壓倒性優勢，只要在某些不適合 JAK、或無法接受其風險的患者中建立穩定效益，也可能找到真正的臨床位置。 所以後續研究不只要問「多少人 SALT≤20」，還應預先定義 JAK 經驗族群、復發時間、患者回報結果與維持治療需求。這些資料才會決定它是新機制新聞，還是一個能被醫師真正放進治療順序的產品。 ## 07｜台灣讀者的產業座標 目前最直接的市場標的是美股 Q32 Bio（Nasdaq: QTTB）。查證截至 2026 年 7 月 19 日，台灣上市櫃公司尚無一級公開資料可證明直接參與 bempikibart 的共同開發、製造、臨床或商業化權利。AbbVie（NYSE: ABBV）與 Nektar（Nasdaq: NKTR）則是不同證據成熟度與機制路線的產業座標，不能混成同一串「受惠股」。 ## 08｜結論：先把「值得驗證」和「已經證明」分開 Bempikibart 的 36 週資料有明確價值。對一個 33 人的早期研究來說，10 名患者達到 SALT≤20、主要分析看到 35.3% 的平均 SALT 降幅，而且沒有觀察到治療相關嚴重或 Grade 3 以上不良事件，足以讓這個計畫繼續往前。 它目前最重要的價值，是支持一場更大的試驗；「勝過 JAK」仍待未來對照資料證明。 接下來請盯住四件事：完整第 52 週停藥資料、註冊導向試驗是否真的啟動、隨機對照後的 SALT≤20 是否重現，以及安全性樣本擴大後能否維持。 33 人讓股價近乎翻倍，很精彩；能不能讓治療標準改變，要等下一場考試。 ## 免責聲明 本文為生技醫藥與產業趨勢整理，不構成醫療診斷、治療建議或投資建議。Bempikibart 尚未獲准上市；臨床試驗結果不代表個別患者療效，實際治療應由醫療專業人員依核准標籤與個別狀況判斷。 --- ## AnBogen Therapeutics: Building a “Starlink” for Cancer Immunotherapy URL: https://drugnews.com.tw/articles/2026-07-31-anbogen-immunotherapy-starlink-en.html Date: 2026-07-31 Language: en Summary: Can a small molecule from Taiwan extend the $30 billion PD-1 ecosystem into the 95% of metastatic colorectal cancer patients still beyond immunotherapy’s reach? > Can a small molecule from Taiwan extend the $30 billion PD-1 ecosystem into the 95% of metastatic colorectal cancer patients still beyond immunotherapy’s reach? | $30B+ | ~95% | 66 Patients | | --- | --- | --- | | PD-1 Blockbuster Ecosystem | of mCRC is pMMR / non-MSI-H | Planned Phase 1/2 Enrollment | As SpaceX stepped into the public-market spotlight this year, Starlink stood out as the asset generating its most important source of commercial cash flow. Terrestrial broadband already reaches much of the globe, but large pockets remain underserved. Starlink uses a constellation of low-Earth-orbit satellites to bring connectivity to remote regions, ships at sea, disaster zones, and increasingly, aircraft. [1] Cancer immunotherapy has its own coverage gap. PD-1 therapies have transformed cancer care, but they do not work across the board. Some tumors are visible to the immune system and readily infiltrated by T cells. Others suppress antigen presentation, exclude immune cells, and build a deeply immunosuppressive microenvironment. The drugs exist. The signal simply cannot get through. ABT-301, AnBogen Therapeutics’ lead asset, was designed to address that disconnect. The oral small molecule is intended to remodel the tumor microenvironment and broaden the reach of existing immunotherapies. If its mechanism translates in the clinic, ABT-301 could become a reusable combination partner across multiple drugs and tumor types—a “Starlink” for cancer immunotherapy. > Starlink extends the reach of broadband. ABT-301 is trying to extend the reach of immunotherapy. *Figure 1 | A “Starlink” for cancer immunotherapy: ABT-301 is designed to carry immune signaling into tumors that remain beyond the reach of current treatment.* ## 01 | PD-1 Is a Blockbuster Success—But Much of the Market Still Has No Signal Merck’s Keytruda generated $31.68 billion in global sales in 2025—more than NT$1 trillion from a single product. [2] AnBogen has selected tislelizumab—marketed as TEVIMBRA/百澤安 by BeOne Medicines—as the PD-1 backbone for its current study. The drug generated $737 million in global sales in 2025, up 19% year over year, and has secured approvals in multiple markets. [3] That gives ABT-301 a real-world combination setting inside an established global immuno-oncology ecosystem. If the study shows repeatable sensitization and tumor-microenvironment remodeling, AnBogen will have a path into a treatment infrastructure that is already operating at scale. The commercial opportunity, however, extends well beyond tislelizumab. Most major oncology companies already own PD-1 or PD-L1 assets, and many are advancing PD-1/VEGF bispecifics or next-generation immune combinations. What remains scarce is a sensitizing agent that can open cold tumors and expand the addressable patient pool for drugs already on the market. If ABT-301 demonstrates a reproducible human effect, its partnering potential could extend across multiple PD-1/PD-L1 products and combination strategies. The boundaries of the current relationship are clear. BeOne is supplying study drug so the combination can be tested in patients; the arrangement does not yet include licensing or commercialization. Drug supply is the fuel that gets the first satellite off the ground. A transaction that changes AnBogen’s valuation will require human data strong enough to support a real negotiation. The real opportunity lies with patients current PD-1 therapies still fail to reach. About 95% of metastatic colorectal cancers are pMMR/non-MSI-H, a biology that generally responds poorly to immune-checkpoint inhibitor monotherapy. Using GlobalData estimates for second-line-or-later patients across the United States, China, Japan, and the five largest European markets, the annual patient pool is roughly 370,000. Applying reference treatment-pricing assumptions implies a potential market of about $9 billion. [4] Why are these tumors so difficult? pMMR/non-MSI-H cancers typically carry a lower mutational burden and produce fewer immune-visible antigens. They are also surrounded by abnormal vasculature and suppressive myeloid cells. T cells struggle to recognize the target—and often cannot enter the tumor even when they do. Simply increasing the PD-1 dose will not solve that problem. The immune-excluding microenvironment has to be changed first. The $9 billion figure is an upper-bound market scenario, not a revenue forecast. Patient selection, line of therapy, competing regimens, pricing, treatment duration, clinical success, and commercial rights will all narrow the opportunity. Even after those discounts, the market remains fundamentally different from a small orphan indication: it is a large, long-standing population that PD-1 therapy has largely left outside its coverage area. Chemotherapy and targeted agents remain the main treatment backbone for these patients, with fewer options and diminishing benefit in later lines. Immunotherapy has repeatedly struggled here, which is precisely why combinations capable of changing the tumor microenvironment are drawing attention. AnBogen is entering one of the hardest markets first. That raises clinical risk—but success would carry far more weight across other cold tumors. Nearly every major oncology company already has a PD-1 asset or is betting on the next generation of PD-1/VEGF combinations. What they lack is a partner that can unlock new tumor types for those franchises. If ABT-301 can do that, it will naturally move onto the partnering radar of multiple companies. That is where its capital-markets upside begins. ## 02 | How ABT-301 Could Reconnect a Cold Tumor? ABT-301, also known as imofinostat, is an oral HDAC1/2/3 inhibitor. AnBogen describes it as an epigenetic tumor-microenvironment reprogrammer—a role intended to distinguish it from conventional HDAC inhibitors. [4] Epigenetics can be thought of as the set of switches that determines which genes are read and which remain silent. Cancer cells can manipulate those switches without changing their DNA sequence, reducing antigen presentation, escaping immune recognition, and creating conditions that prevent T cells from entering or surviving. In preclinical studies, ABT-301 in combo with PD-1 blockade increased antigen presentation, enhanced CD8+ cytotoxic T-cell infiltration and activity, reduced monocytic myeloid-derived suppressor cells, and influenced apoptosis, angiogenesis, and tumor metabolism. Together, these effects point toward one goal: converting a tumor that is invisible, inaccessible, and unresponsive into one that immunotherapy can engage. ABT-301 was built from the outset as a combination asset. If it can expand the reach of multiple approved drugs, investors will view it less as a single molecule and more as a platform node. Starlink brings connectivity to places the terrestrial network cannot reach. ABT-301 is trying to make existing cancer therapies work where they currently cannot. Two data points in AnBogen’s research package stand out. ABT-301 showed an HDAC3 IC50 of 6.6 nM versus 106.9 nM for the comparator chidamide—roughly a 16-fold difference. In an anti-PD-1 combination model, 2 of 7 animals in the ABT-301 arm achieved complete responses, versus 0 of 8 in the chidamide arm. [6] The first result speaks to molecular potency. The second suggests the possibility of deeper antitumor activity in vivo. A complete response in a mouse is not the same as an objective response in a patient. These findings do not prove clinical efficacy; they provide a biologically credible reason to test the combination. The next questions are whether tumors shrink in patients, whether the benefit lasts, and whether human samples show the expected immune-microenvironment changes. Safety is the other gate. AnBogen’s first-in-human monotherapy study enrolled 23 participants, and company materials reported no clear signal of neutropenia or cardiac toxicity—issues often watched closely with the HDAC class. [4] That provides a rationale for moving into triple therapy, but monotherapy safety does not guarantee that three drugs can be combined safely. Bevacizumab and PD-1 antibodies carry their own adverse-event profiles, and the therapeutic window will have to be established again in the 66-patient study. HDAC inhibitors have been used clinically for years, including approved indications in hematologic malignancies. Results in solid tumors have been inconsistent, while marrow suppression, fatigue, and other systemic toxicities have constrained prolonged combination use. AnBogen is therefore focusing ABT-301 on HDAC1/2/3 and microenvironment remodeling, with the goal of preserving the sensitizing effect of an oral small molecule while identifying a dose that can coexist with PD-1 and anti-VEGF therapy over time. That is the most practical product test for the triple regimen. *Figure 2 | ABT-301 is designed to reprogram the tumor microenvironment—making the tumor visible again and reactivating an immune response.* ## 03 | Triple Therapy: Reprogram the Environment, Open the Vasculature, Release the Brakes AnBogen is advancing a triple regimen of ABT-301, bevacizumab, and tislelizumab. ABT-301 is intended to reprogram the epigenetic and immune microenvironment. Bevacizumab inhibits VEGF, addressing abnormal vasculature and VEGF-driven immune suppression. Tislelizumab blocks PD-1 and releases the brakes on T-cell activity. > First, make the tumor visible again. > Then, create a path for immune cells to enter. > Finally, remove the restraints on T-cell attack. AnBogen has launched an international multicenter Phase 1/2 study in Taiwan and Australia, with planned enrollment of 66 patients with locally advanced or metastatic pMMR/non-MSI-H colorectal cancer. The trial is evaluating safety, tolerability, a recommended Phase 2 dose, and preliminary efficacy. The first patient was treated in Australia in November 2025, and BeOne is supplying tislelizumab under a clinical drug-supply arrangement. [4] The study now has to answer four questions: Can a workable dose be established? Are tumor responses reproducible? Do they last? And do patient samples show the expected changes in the immune microenvironment? The early portion of the study will escalate dose levels to determine whether the three agents can be administered together before moving into expansion. Investigators will look beyond tumor shrinkage to depth of response, duration of response, and progression-free survival, while monitoring hematologic toxicity, infection, bleeding, hypertension, proteinuria, and immune-related adverse events. Mechanistic synergy could produce deeper activity; overlapping toxicity could erase that value just as quickly. Objective response rate will likely grab the first headline, but durability will determine the program’s longer-term value. A tumor that shrinks briefly and then progresses will not support a compelling combination strategy. If duration of response and progression-free survival improve in parallel—and treatment discontinuations are not driven by toxicity—the regimen will have a stronger case for advancing into a larger randomized trial. AnBogen’s future data package needs to tell all three stories together. A 66-patient study is not a pivotal trial, but it is large enough to reveal the first meaningful clinical profile of ABT-301. If responses appear across patients and centers, and paired biopsies show increased antigen presentation and CD8+ T-cell infiltration, AnBogen will have more than a mechanistic narrative—it will have a clinically weighted dataset that global partners can begin to diligence. Patient mix will matter. Prior therapies, baseline tumor burden, and the proportion of patients with liver metastases can materially affect response rates. A topline ORR alone will not be enough; baseline characteristics, duration of response, and reasons for discontinuation must be read together. In an early-stage study, one or two additional responders or progressors can move the percentage sharply, which makes consistency across centers and over time especially important. *Figure 3 | A sequential network of the ABT-301 triple therapy: reprogram the microenvironment, improve immune access, and release the immune brake.* ## 04 | The Same Path Has Already Produced a Human Signal ABT-301’s triple-combination strategy is not starting without clinical precedent. The randomized Phase 2 CAPability-01 study, published in Nature Medicine in 2024, combined another HDAC inhibitor—chidamide—with the PD-1 antibody sintilimab and bevacizumab in chemotherapy-refractory MSS/pMMR metastatic colorectal cancer. The study placed a regimen similar to AnBogen’s directly into patients and paired the clinical readout with tumor-immune profiling. [5] The results support the therapeutic concept, but they cannot be used as a proxy for ABT-301’s probability of success. The numbers were notable. The triplet achieved an 18-week progression-free survival rate of 64.0% versus 21.7% for the doublet, an objective response rate of 44.0% versus 13.0%, and median progression-free survival of 7.3 months versus 1.5 months. [5] Tumor analyses also showed greater CD8+ T-cell infiltration and a more activated immune microenvironment. This is the closest human evidence to AnBogen’s current design and moves the “reprogram first, amplify PD-1 second” thesis beyond animal models. The safety trade-off also matters. Common adverse events included proteinuria, thrombocytopenia, leukopenia, neutropenia, anemia, and diarrhea, and the study reported two treatment-related deaths. That leaves room for ABT-301 to differentiate. If AnBogen can preserve a meaningful efficacy signal while delivering a cleaner safety profile, convenient oral dosing, or a more precise biomarker strategy, the asset could still stand out to potential partners. CAPability-01 used chidamide; AnBogen’s study uses ABT-301. The trials are independent and have not been compared head to head. CAPability-01 validates the pathway—not AnBogen’s product. ABT-301 must answer the same efficacy and safety questions with its own 66 patients. When early data emerge, one key question will be whether responses cluster in a narrow subset of patients. If activity is confined to a particular biomarker group or to patients with lower disease burden, the next study may need a tighter enrollment strategy. If responses are more broadly distributed, the platform case becomes stronger. This is why tumor-sample analysis matters: it can show not only whether the drug works, but who is most likely to benefit—and prevent later trials from enrolling patients who are unlikely to respond. AnBogen now has to show that ABT-301 can preserve the mechanistic synergy of the triplet while establishing a safety window suitable for sustained combination therapy. If it succeeds, ABT-301 could move from “another HDAC inhibitor” to a new sensitizing backbone for cancer immunotherapy. ## 05 | The First Market Is $9 Billion. The Real Upside Is the “Constellation Effect” Starlink becomes more valuable as more satellites are added and more use cases come online. ABT-301 faces the same valuation divide. The first coverage area is the estimated $9 billion second-line-or-later pMMR/non-MSI-H metastatic colorectal cancer market. The second layer is whether ABT-301 can work with multiple PD-1 or PD-L1 antibodies rather than remaining tied to a single product. The third is whether the asset can move beyond immune sensitization into reversal of chemotherapy resistance. AnBogen has reported preclinical data in KRAS-mutant pancreatic cancer suggesting that ABT-301 may improve chemotherapy sensitivity through the HDAC3–NRF2 axis. [6] The fourth is expansion into brain, liver, head-and-neck, and other solid “cold” tumors that respond poorly to immunotherapy—potentially opening access to a checkpoint-inhibitor-refractory cancer market estimated at approximately $95.5 billion. [7] If the same mechanism can be reproduced across drugs, treatment backbones, and tumor types, ABT-301 could become a tumor-microenvironment reprogramming platform: one mechanism × multiple drugs × multiple cancers × multiple partners. At that point, the valuation framework shifts from “one drug, one indication” to a treatment-coverage network that can keep adding nodes. “Platform” is one of the most overused words in biotech. The market ultimately asks one question: Can the same mechanism be repeated? One tumor type and one PD-1 partner would make ABT-301 a valuable combination asset. A second PD-1, a second cancer, or a chemotherapy-resistance setting would begin to turn that single satellite into a constellation. For large pharmaceutical companies, that possibility addresses a very practical problem. Many already own PD-1, PD-L1, or PD-1/VEGF assets, and their portfolios are becoming increasingly difficult to distinguish. The next step is finding partners that can expand the patient population. A small molecule that works across multiple immune therapies could bring several companies to the table and give AnBogen greater partnering leverage. From a licensing perspective, the first successful indication would establish ABT-301’s floor value. A second tumor type would let a buyer spread development costs across a broader commercial opportunity. Reproducibility with another PD-1 would reduce dependence on any single partner. Each additional human dataset could therefore change not just the pipeline diagram, but the structure of potential regional deals, co-development arrangements, and global licensing discussions. *Figure 4 | From one satellite to a full constellation: ABT-301’s expansion blueprint across drugs, tumor types, and partners.* ## 06 | The Second Starlink Node: ABT-500 Connects Diagnosis, Patient Selection, and Therapy Beyond ABT-301, the second platform worth watching is AnBogen’s ABT-500 theranostics franchise. The ABT-500 series uses the LHRH receptor as a precision-guidance entry point. ABT-502 is designed for imaging and patient selection; ABT-501 carries the cytotoxic payload DM1; and ABT-505 extends the same targeting concept to boron delivery and boron neutron capture therapy, or BNCT. [8] The clinical logic is sequential. Patients would first undergo ABT-502 fluorine-18 PET imaging to determine whether their tumors express enough LHRH receptor. Those who qualify could then receive ABT-501, which uses an LHRH peptide to direct DM1 toward receptor-high tumors. [8] For ABT-502, the key questions are PET–IHC concordance, the SUV cutoff, tumor-to-background ratio, and whether the imaging signal predicts subsequent treatment benefit. ABT-501 must establish biodistribution, toxicology, and a therapeutic window; delivering DM1 to the tumor is not enough if normal-tissue exposure remains high. Animal data in triple-negative breast cancer, ovarian cancer, and pancreatic cancer provide a starting point. Human dose and response will determine whether the approach is truly differentiated. [8] Company materials suggest that approximately 50% of patients with triple-negative breast cancer may overexpress the LHRH receptor. [8] Receptor expression is only the first gate. The imaging signal must be strong enough, normal-tissue exposure must be controlled, and the payload must be released in the right place. If ABT-502 can identify the right patients up front, it may improve the hit rate for ABT-501. ABT-505 attaches a boron payload to the same targeting approach and extends the platform toward BNCT. That turns a single peptide-drug conjugate into a modular architecture that can swap imaging and therapeutic payloads. The upside is substantial, but so is the engineering burden: PET, PDC, and BNCT are three distinct development systems, and a failure at any link would leave the theranostic concept stranded on a slide. ABT-505 remains earlier stage. In company-reported HCC1806 cell-line data, boron uptake was approximately 20 times higher than with BPA. [8] That is an intriguing cellular-uptake signal, but it answers only whether boron can be delivered into cells. Before BNCT can become a product, AnBogen still needs to establish in vivo distribution, tumor-to-normal tissue ratios, irradiation conditions, and normal-tissue safety. If the model works, imaging would screen out patients whose tumors lack sufficient receptor expression before treatment begins. That could reduce trial size and improve response rates. Commercially, however, the diagnostic and therapeutic workflows must also connect inside the hospital, including isotope supply, image-interpretation standards, and payload manufacturing. ABT-500’s ultimate value will depend as much on the care-delivery system as on the drug science. For now, these assets should be viewed as long-dated options. The market is unlikely to reclassify AnBogen from a single-HDAC company to a multi-node precision-oncology platform until a second program produces reproducible human clinical data. ## 07 | Four Signal Upgrades Will Determine AnBogen’s Next Valuation Step AnBogen remains a clinical-stage biotechnology company with no approved product. Its valuation will therefore move primarily with changes in the probability of clinical success, not current operating cash flow. The first upgrade is confirmation that the triple regimen has a manageable safety profile and a viable recommended Phase 2 dose. The second is evidence of objective responses that are both reproducible and durable—not a small number of isolated responders. The third is human validation of the mechanism: increased antigen presentation, greater CD8+ infiltration, reduced immune suppression, and a biomarker that can identify patients most likely to benefit. The fourth is reproduction of the same mechanism in a second tumor type or treatment combination—or a licensing or co-development agreement with a global pharmaceutical company. > Candidate molecule > → Human data supporting the mechanism > → A platform that can extend across tumor types > → A strategic asset global pharma is willing to transact on These four upgrades would move the clinical program forward—and change how the capital markets classify the entire company. Not every press release will carry the same weight. First-patient dosing, completion of dose escalation, reproducible objective responses, biomarker discovery, and a license containing upfront cash and milestones each remove a different layer of risk. Markets may react to news, but durable valuation changes require new evidence. Every new datapoint must connect back to the same treatment thesis before it becomes a genuine new node in the constellation. Capital requirements cannot be ignored. International multicenter studies, a second indication, and the imaging and payload programs within ABT-500 will all require continued investment. Until a major licensing transaction occurs, cash runway, financing timing, and dilution will influence how much of the asset value ultimately accrues to existing shareholders. Drug success and shareholder return are separated by the cost and structure of financing—a central reality of early-stage biotech investing. Deal terms matter just as much as the partner’s name. A drug-supply relationship may show that two companies are willing to work together, but it does not establish the value of the asset. Upfront payments, development and sales milestones, royalty rates, and geographic rights are the inputs that determine future economics. Licensing ABT-301 after early data could share development costs but limit leverage; carrying the program further could support a higher price, while requiring AnBogen to absorb more expense and more clinical risk. *Figure 5 | AnBogen’s capital-markets trajectory: from the first signal in 66 patients to a cross-tumor platform and a globally partnerable asset.* ## Conclusion | The First Clinical Signal Will Decide Whether AnBogen Is One Satellite—or a Constellation AnBogen’s next task is straightforward: bring existing immunotherapies into patient populations they still do not reach. Keytruda has already become a $30 billion-class blockbuster, and the commercial infrastructure for PD-1 is firmly established. AnBogen is betting on the patients still left outside that network. The first checkpoint is ABT-301. The program must show that a small molecule from Taiwan can carry a commercially proven class of immunotherapies into tumors that have remained beyond reach. The first destination is the pMMR/non-MSI-H population, which represents roughly 95% of metastatic colorectal cancer. The first major test is the 66-patient Phase 1/2 study. The first quantifiable commercial opportunity is the company’s estimated $9 billion market scenario. If the first human signal is credible, AnBogen will have opened a new treatment-coverage map—one that can expand across drugs, tumor types, and partners. The company must first make sure the first satellite sends back a clear signal before talking about the rest of the constellation. If the 66-patient study can connect safety, efficacy, and human mechanism in one coherent dataset, ABT-301 may move from an early-stage Taiwan asset into the combination strategy of global oncology companies. Only then will the $30 billion PD-1 ecosystem, the 95% treatment blind spot, and the $9 billion market opportunity begin to form a single investable story. > Whether the first satellite can return a stable signal > will determine whether AnBogen remains an early candidate-drug company > or grows into a constellation that expands the reach of cancer immunotherapy. > **AnBogen Therapeutics: Building a Starlink for Cancer Immunotherapy** > The first satellite is in orbit. The valuation story begins when the signal comes back. ### Disclaimer This article is based on public information available as of July 24, 2026, together with the research materials provided. It is intended solely for industry analysis and does not constitute investment advice. Preclinical and early-stage clinical findings do not guarantee human efficacy, regulatory approval, or commercial success. --- ## 安邦生技：免疫治療的「星鏈計畫」 URL: https://drugnews.com.tw/articles/2026-07-31-anbogen-immunotherapy-starlink.html Date: 2026-07-31 Language: zh-Hant Summary: 一顆台灣小分子，有機會把三百億美元級PD-1藥王送進九成五患者仍無法受益的治療盲區嗎？ > 一顆台灣小分子，能否把300億美元級PD-1生態，連進95%的大腸癌免疫盲區？ | 300億美元級 | 約95% | 66名 | | --- | --- | --- | | PD-1藥王生態 | mCRC為pMMR／non-MSI-H | Phase 1/2預計收案 | 馬斯克旗下的SpaceX今年上市首秀，然而多數虧損的資產中，星鏈Starlink帶來了其目前最重要的商業現金流。即便當今的網路基站已經遍佈全球大部分的角落，但是依然留有大片面積是無法被覆蓋的。星鏈就是透過低軌衛星把訊號送偏遠地區、海上船舶到災區通訊。[1]如今美國多數的航空公司都透過星鏈的服務，提供人們在飛機上的網路服務，將網際網路的世界無時無刻地傳遞覆蓋全球。 而在癌症免疫治療也存在類似的「通訊覆蓋缺口」。 眾所週知的PD-1藥物已經改寫全球腫瘤治療，但它並不是對所有腫瘤都有效。部分腫瘤能被免疫系統辨認，T細胞也能順利進入；另一些腫瘤卻會降低抗原呈現、排斥免疫細胞，甚至在腫瘤內建立強烈的免疫抑制環境。藥物已經存在，訊號卻送不進去。 安邦生技的核心藥物ABT-301，試圖處理的正是這個問題。 透過重塑腫瘤微環境、擴大既有免疫療法治療範圍的小分子藥物。若這顆藥物與機制能在人體臨床中成功，安邦的價值將成為與不同藥物連用於多種癌種的「免疫治療星鏈」。 > Starlink擴張的是網路覆蓋；ABT-301挑戰擴張的，是免疫治療覆蓋。 *圖1｜免疫治療的「星鏈計畫」：ABT-301嘗試把免疫訊號送進治療盲區* ## 01｜PD-1已經很成功，但仍有大片市場沒有訊號 PD-1默沙東Keytruda在2025年全球銷售額達316.8億美元，單品一年便創造超過新台幣兆元營收。[2] 安邦這次選用的PD-1，是BeOne Medicines（原百濟神州）旗下的tislelizumab（TEVIMBRA／百澤安）。這款藥2025年全球銷售額達7.37億美元，年增19%，並已在多個市場取得藥證[3]，ABT-301因此直接進入一個真實的連用環境。後續若能看到增敏與腫瘤微環境重塑訊號，安邦就有機會接上這套已在全球運轉的免疫治療體系。 對安邦而言，ABT-301的商業想像不只停在tislelizumab。全球腫瘤藥廠多數已布局PD-1、PD-L1，或正在推進PD-1／VEGF雙抗與下一代免疫組合，真正稀缺的是能替既有產品打開冷腫瘤與新患者市場的增敏分子。若ABT-301的人體資料證明其微環境重塑能力可重複，未來合作想像可延伸至更多PD-1／PD-L1產品與不同免疫組合。 目前的合作邊界很清楚：BeOne提供臨床試驗用藥，雙方先把組合送進人體測試；合作尚未延伸到授權或商業化。供藥是讓衛星升空的燃料，真正能改變公司估值的交易，要等ABT-301拿出足以談判的人體數據。 真正的缺口，在PD-1尚未有效覆蓋的患者。 轉移性大腸直腸癌中，約95%的患者屬於pMMR／non-MSI-H類型，通常難以從現有免疫檢查點抑制劑獲得明顯效益。根據GlobalData估算美國、中國、日本與歐洲五大市場的病患數，每年約有37萬名二線以上患者，推算出潛在市場約90億美元。[4] 為什麼這95%的患者特別難治？pMMR／non-MSI-H腫瘤通常突變負荷較低，可被免疫系統辨認的抗原訊號不夠，腫瘤周圍又布滿異常血管與髓系抑制細胞。T細胞看不清楚目標，也很難走進腫瘤裡面。因此，單純把PD-1劑量加大沒有用，必須先改造那個排斥免疫反應的腫瘤環境。 90億美元只是潛在市場，患者分層、治療線別、競爭藥物、定價與臨床成功率，都會把這個數字一層一層往下折。即使如此，這仍是一塊長年被PD-1留在覆蓋區外的大市場。ABT-301若能走進去，商業量級與少數患者的孤兒適應症完全不同。 這群患者目前仍以化療與標靶治療為主要骨幹，進入後線之後，藥物選擇與療效空間會逐漸縮小。免疫治療過去在這裡多次碰壁，反而讓能夠改變腫瘤微環境的組合更受關注。ABT-301選擇從這個最難打開的市場切入，臨床風險較高；一旦成功，對其他冷腫瘤也會產生更強的說服力。 大藥廠手上幾乎都有PD-1，或正在押注下一代PD-1／VEGF。真正稀缺的是能替現有產品打開新病種的搭檔。ABT-301若能做到，自然會進入各家公司爭取合作的名單。這正是它最有資本市場想像力的位置。 ## 02｜ABT-301如何讓腫瘤重新「連線」？ ABT-301，又名imofinostat，是一款口服HDAC1／2／3抑制劑。安邦將它定位為「表觀遺傳腫瘤微環境重塑劑」，希望與傳統HDAC抑制劑拉開差異。[4] 表觀遺傳可以理解為控制基因是否被讀取的一組開關。癌細胞即使沒有改變DNA序列，也能透過這些開關降低抗原呈現、躲避免疫辨識，並創造不利於T細胞進入與存活的環境。依已公布的臨床前資料，ABT-301聯合PD-1抗體提升抗原呈現、增加CD8⁺毒殺型T細胞的浸潤與活性、降低單核性髓源抑制細胞，並影響癌細胞凋亡、血管新生與腫瘤代謝。把這些作用放在一起看，方向很清楚：先讓「看不見、進不去、打不動」的腫瘤，變成免疫療法有機會發揮的狀態。ABT-301從一開始就把舞台放在組合治療。它若能替多款上市藥物打開患者覆蓋，市場自然會用平台來看待。Starlink把訊號送進原本連不上的地方；ABT-301想做的，是讓各類腫瘤治療藥物重新產生作用。 安邦公布的研究資料裡，有兩組數字值得特別看。ABT-301對HDAC3的IC50為6.6 nM，對照分子chidamide則為106.9 nM，相差約16倍；在與anti-PD-1聯用的動物模型中，ABT-301組有2／7出現完全反應，chidamide組則是0／8。[6] 前一組數字看的是分子活性，後一組看的是動物腫瘤有沒有出現更深的反應。 不過，老鼠身上的完全反應不能直接寫成人體的客觀反應率。這些資料能做的是替三聯療法提供一個合理起點：ABT-301確實碰得到它想碰的標的，也在動物裡看到與PD-1合用的可能性。接下來還是要看患者的腫瘤會不會縮小、能維持多久，以及人體檢體裡是否真的出現免疫微環境改變。 安全性則是另一道門。安邦先前完成的ABT-301單藥一期共有23名受試者，公司資料顯示，沒有觀察到嗜中性白血球低下或心臟毒性異常等其他HDAC抑制劑常被關注的問題。[4]這讓ABT-301有機會繼續往三聯療法走，但單藥安全不代表三顆藥放在一起也一定安全。bevacizumab與PD-1抗體各自都有不良反應，治療窗口必須在66名患者裡重新找。 HDAC抑制劑已有多年的臨床使用經驗，過去也有產品獲准用於血液腫瘤；實體腫瘤的成果卻不穩定，骨髓抑制、疲倦與其他全身性副作用也限制了長期合併治療。安邦因此把ABT-301的重點放在HDAC1／2／3與腫瘤微環境，希望用口服小分子保留增敏效果，同時找出可長期與PD-1、抗VEGF藥物共存的劑量。這會是三聯試驗最實際的產品考驗。 *圖2｜ABT-301透過腫瘤微環境重塑，嘗試讓腫瘤重新被看見、讓免疫重新被啟動* ## 03｜三聯療法：重塑環境、打開通道、解除煞車 安邦目前推進的是ABT-301、bevacizumab與tislelizumab三聯療法。 ABT-301負責重塑表觀遺傳與腫瘤免疫微環境；bevacizumab透過抑制VEGF，處理異常血管與相關免疫抑制屏障；tislelizumab則封鎖PD-1，解除T細胞的免疫煞車。 > 先讓腫瘤重新被辨認， > 再讓免疫細胞進入， > 最後解除T細胞的攻擊限制。 安邦已啟動Phase 1/2國際多中心試驗，預計在台灣與澳洲收納66名pMMR／non-MSI-H局部晚期或轉移性大腸直腸癌患者，評估安全性、耐受性、建議第二期劑量與初步療效。資料顯示，首位患者已於2025年11月在澳洲接受治療，tislelizumab由BeOne依臨床藥物供應合作提供。[4]接下來的66名患者將回答四件事：安全劑量能否確立、腫瘤反應能否重複、療效能否持續，以及人體樣本是否出現預期的免疫微環境變化。 這項試驗前段會先做劑量爬升，確認三顆藥能不能一起使用，再進入後續擴增。臨床除了看腫瘤是否縮小，還要追蹤反應深度、反應持續時間與無惡化存活期；同時盯住血液毒性、感染、出血、高血壓、蛋白尿，以及免疫相關不良事件。三種機轉如果能合作，可能帶來更深療效；如果毒性也一起疊上來，臨床價值就會被吃掉。 早期數據公布時，ORR通常最先搶走目光，真正決定後續價值的卻是反應能維持多久。腫瘤短暫縮小後很快惡化，組合療法很難說服大藥廠；DOR與PFS若能同步拉長，而且停藥原因沒有被毒性主導，才有條件走進更大規模的隨機試驗。安邦未來的數據包，至少要把這三條線一起交代清楚。 66人的規模還不到大型樞紐試驗，卻足以讓市場第一次看見ABT-301在人類身上的輪廓。倘若不同患者、不同中心都能看到反應，而且治療前後檢體出現抗原呈現上升、CD8⁺T細胞浸潤增加，安邦拿去和國際藥廠談的，將是一份開始有臨床重量的資料，機轉故事也因此有了人體證據。 試驗收進的是局部晚期或轉移性患者，先前接受過哪些治療、腫瘤負荷多高、肝轉移比例多少，都會影響最後的反應率。讀取頂線數據時，一個ORR百分比遠遠不夠，基線特徵、反應持續時間與停藥原因都得放在一起看。早期試驗樣本不大，只要多一兩名患者有反應或惡化，比例就可能大幅變動，持續且跨中心重複的訊號因此更重要。 *圖3｜三聯療法的順序式邏輯：重塑環境、打開通道、解除煞車* ## 04｜同一條路，已在人類臨床看見過訊號 ABT-301三聯策略已有外部臨床資料可供參照。 2024年刊登於《Nature Medicine》的CAPability-01隨機第二期試驗，使用另一款HDAC抑制劑chidamide，搭配PD-1抗體sintilimab與bevacizumab，治療化療難治的MSS／pMMR晚期大腸直腸癌。這項研究把與安邦相近的「HDAC調節＋PD-1＋抗VEGF」組合真正放進人體，也同步分析腫瘤免疫微環境。[5] 結果支持這條治療路徑值得追蹤，但不能直接換算成ABT-301的成功率。 如果把數據拆開來看，CAPability-01三聯組的18週無惡化存活率為64.0%，雙聯組只有21.7%；客觀反應率為44.0%對13.0%；中位無惡化存活期則是7.3個月對1.5個月。[5]腫瘤分析也看到較高的CD8⁺T細胞浸潤與較活化的免疫微環境。這是目前最接近安邦三聯設計的人體證據，也把「先重塑環境，再放大PD-1作用」從動物實驗推進到人體訊號。 療效亮眼，代價也不能省略。這項研究常見的不良事件包括蛋白尿、血小板與白血球下降、嗜中性白血球低下、貧血與腹瀉，並報告兩例治療相關死亡。安邦可以從其他地方建立差異。若ABT-301能維持療效訊號，同時交出較乾淨的安全性、更便利的口服使用方式或更清楚的患者標記，同樣可能成為大藥廠願意接手的差異。 CAPability-01使用chidamide，安邦試驗使用ABT-301，兩項研究各自獨立，也沒有做頭對頭比較。這份資料能支持治療路徑，無法替安邦預先宣布成功。ABT-301仍要用自己的66名患者，回答同樣的療效與安全問題。 若未來安邦公布早期數據，最值得注意的是反應有沒有集中在少數特殊患者。倘若所有反應都來自某一種生物標記或較輕的腫瘤負荷，下一階段試驗就要重新調整選人方式；倘若反應分布較廣，平台延伸的空間才會變大。這也是腫瘤檢體分析的重要性：它可以告訴公司藥物對誰有效，也能避免下一輪試驗繼續把不會反應的患者全部收進來。 安邦接下來要用ABT-301證明，它能在保留機轉協同的同時，建立更適合長期組合治療的安全窗口。若能做到，這顆分子便有機會從另一款HDAC抑制劑，升級為免疫治療的增敏新底座。 ## 05｜第一個市場是90億美元，真正的上限在「星座效應」 Starlink的價值隨著衛星數量增加而放大，覆蓋範圍與服務場景也持續擴張。ABT-301同樣面臨這條估值分水嶺。 第一個覆蓋區，是約90億美元的pMMR／non-MSI-H後線轉移性大腸直腸癌市場。 第二層，看它能否搭配不同PD-1／PD-L1抗體，擺脫單一藥物的限制。 第三層，是能否從免疫增敏，延伸到化療抗藥逆轉。安邦已公布ABT-301在KRAS突變胰臟癌中，透過HDAC3–NRF2路徑提升化療敏感性的臨床前研究。[6] 第四層，則是能否向腦癌、肝癌、頭頸癌及其他免疫治療反應有限的實體冷腫瘤延伸，潛在滲透約955億美金的冷腫瘤市場。[7] 如果ABT-301同一機制能跨藥物、跨治療骨幹、跨癌種重現，安邦將變成腫瘤微環境重塑平台。一個機制 × 多款藥物 × 多個癌種 × 多家合作夥伴。資本市場可能給予的價值，也會從「一顆藥乘上一個適應症」，轉向一張能持續增加節點的治療覆蓋網路。 平台兩個字，在生技公司簡報裡早已被用到發白。市場其實只問一件事：同一套機轉能不能重複。ABT-301先在一個癌種、搭配一款PD-1站穩，還只能算一項有價值的組合藥；等到它換PD-1、換癌種，甚至接上化療抗藥逆轉，星鏈才真正從一顆衛星長成星座。 對大型藥廠來說，這正好打中一個現實難題。很多公司手上已經有PD-1、PD-L1或PD-1／VEGF雙抗，但產品彼此愈來愈相似，下一步都在尋找可以擴大患者範圍的組合夥伴。一顆能跨多款免疫藥使用的小分子，將吸引多家公司走上談判桌，安邦的授權選擇也會因此增加。 從授權角度看，第一個成功適應症會先決定ABT-301的底價。第二個癌種若能延續相同機轉，買方就能把開發成本分攤到更多市場；另一款PD-1也能重現時，藥物對單一合作夥伴的依賴會下降。因此安邦未來每增加一組人體資料，管線頁面會多一個箭頭，區域權利、共同開發與全球授權的談判方式也可能跟著改變。 *圖4｜從一顆衛星到整個星座：ABT-301的跨藥物、跨癌種、跨夥伴擴張藍圖* ## 06｜第二組星鏈節點：ABT-500把診斷、分層與治療連成閉環 除ABT-301外，安邦目前最值得追蹤的第二條資產主軸是ABT-500診療一體化系列： ABT-500系列以LHRH受體為精準導引入口，先透過ABT-502進行影像定位與患者篩選，再由ABT-501攜帶DM1執行治療，並由ABT-505延伸至含硼載荷與BNCT應用。[8] ABT-500系列走的是診療一體化。病人先接受ABT-502氟-18 PET顯影，確認腫瘤有沒有足夠的LHRH受體；篩選合適後，再由ABT-501以LHRH胜肽導引DM1載荷，把細胞毒性藥物送到受體高表現的腫瘤。[8] ABT-502這一關，重點會落在PET與IHC的一致性、SUV cutoff、腫瘤對背景比，以及影像訊號能不能預測後續療效。ABT-501則要回答體內分布、毒理與治療窗；把DM1送到腫瘤還不夠，正常組織暴露也得壓低。三陰性乳癌、卵巢癌與胰臟癌的動物資料提供起點，真正分出高下的仍是人體劑量與反應。[8] 公司資料指出，約50%的三陰性乳癌患者可能有LHRH受體過度表現。[8]受體只是第一道門，影像訊號還要足夠清楚、正常組織暴露也要能控制，載荷也必須在正確位置釋放。ABT-502若能先把適合的患者找出來，才有機會提高ABT-501後續治療的命中率。 ABT-505則把同一套導航方式接上含硼載荷，往硼中子捕獲治療延伸。安邦也因此把一顆PDC延伸成可以替換影像與治療載荷的設計架構。想像力很大，開發難度也同樣大：PET、PDC與BNCT是三套不同工程，任何一段接不起來，診療一體化就只會停在簡報上。 ABT-505目前還在更早期的位置。公司資料顯示，在HCC1806細胞株中，硼吸收通量約為BPA的20倍。[8]這個數字很吸睛，回答的仍是細胞能否把硼帶進去。BNCT走進臨床前，還要跨過體內分布、腫瘤對正常組織比、照射條件與正常組織毒性。這幾道關卡都成立，含硼載荷才有產品意義。 這套模式若能成立，影像會先替治療做一次篩選，減少把藥物用在受體表現不足的患者身上。對臨床試驗而言，收案人數可能因此縮小，反應率也有機會提高；對商業化而言，診斷與治療必須在醫院端順利銜接，還要處理放射性同位素供應、影像判讀標準與載荷製造。ABT-500的價值最後會同時取決於藥物科學與整套醫療流程。 現階段，這些資產比較像長期選擇權。等到第二條管線也走出人體臨床訊號，市場才會開始把安邦從單一HDAC公司，改看成多節點精準腫瘤公司。 ## 07｜四次訊號升級，決定安邦的下一段估值 安邦目前仍是尚無核准產品的臨床階段生技公司。市場會跟著臨床成功機率調整定價，當期現金流暫時不列為主要指標。 第一次升級，是三聯療法確認可管理的安全性與建議第二期劑量。 第二次升級，要看到可重複、可持續的客觀反應，排除少數患者偶然反應的可能。 第三次，是人體樣本驗證抗原呈現、CD8⁺浸潤或免疫抑制解除等機轉，並找到可用於選人的生物標記。 第四次，是同一機制在第二癌種或另一治療組合中重現，或取得跨國藥廠的授權與共同開發合作。 > 候選分子 > → 人體數據支持的機制 > → 可跨癌種的平台 > → 全球藥廠願意交易的戰略資產 這四次升級會推進臨床里程碑，也會改變資本市場對整家公司的分類。 對投資人來說，往後每一則公告的含金量其實不一樣。第一位患者入組、完成劑量爬升、出現可重複的客觀反應、找到生物標記，以及簽下有upfront與里程碑的授權，每一步降低的風險都不同。市場每天都有新聞，估值卻只認新證據。每一個新證據都要接回同一套治療邏輯，星鏈的節點才算真的增加。 另一個不能忽略的是錢。國際多中心試驗、第二癌種與ABT-500的影像及載荷開發都要持續投入。在大型授權交易出現以前，現金跑道、募資時間與股本稀釋都會影響股東最後拿到多少價值。藥物成功與股價報酬之間，還隔著募資成本與資金結構，這也是早期生技投資最現實的一面。 交易條款也要看細。合作對象和供藥關係只說明雙方已坐上談判桌，還算不出資產價值。簽約金、開發里程碑、銷售里程碑、權利金比例與地區權益，才是估算公司未來收益的材料。ABT-301若在早期數據後就授權，可以分攤開發成本，但議價空間較小；公司若自行推進到更成熟階段，潛在價格較高，也要承擔更多試驗支出與失敗風險。 *圖5｜安邦的資本市場想像：從66名患者的第一組訊號，走向跨癌種平台與全球合作資產* ## 結語｜第一組臨床訊號，決定安邦是一顆衛星，或是一個星座 安邦接下來要做的事，其實很清楚：讓現有免疫療法走進尚未覆蓋的患者市場。Keytruda已經做成300億美元級藥王，PD-1的商業軌道也早已鋪好；安邦押注的是那些至今仍被留在門外的患者。 這條路的第一個檢查點落在ABT-301。它要證明一顆台灣小分子，能把已有巨大商業規模的免疫療法送進過去難以抵達的腫瘤。 第一站，是約占轉移性大腸直腸癌95%的pMMR／non-MSI-H治療盲區；第一場驗證，是66名患者的Phase 1/2試驗；第一個可量化市場，則是公司引用資料估算的90億美元。 若第一組人體訊號成立，安邦將打開一張新的治療覆蓋圖，並沿著不同藥物、不同癌種與不同合作夥伴繼續擴張。 所以安邦接下來要先讓第一顆衛星回傳夠清楚的訊號，再談更多衛星。66名患者如果能把安全性、療效與人體機轉同時接起來，ABT-301才有資格從台灣早期新藥，走進全球藥廠的組合策略裡。到那個時候，300億美元、95%與90億美元，才會開始成為同一個故事。 > 第一顆衛星能否穩定回傳訊號， > 將決定安邦最終停在早期候選藥物， > 或長成一個能替全球免疫治療擴大覆蓋的星座。 > **安邦生技：免疫治療的星鏈計畫** > 第一顆衛星已升空，真正的估值，等訊號回傳。 ### 免責聲明 本文依截至2026年7月24日公開資訊與所提供研究資料整理，僅供產業分析，不構成投資建議；臨床前與早期臨床結果不代表最終人體療效、藥證核准或商業化成果。 --- ## Xtandi 之後，安斯泰來靠誰？從單一明星藥到腫瘤平台的四道驗證門 URL: https://drugnews.com.tw/articles/2026-07-23-astellas-oncology-platform.html Date: 2026-07-30 Language: zh-Hant Summary: Xtandi 仍主導安斯泰來腫瘤收入；Padcev、Vyloy 與下一代 CLDN18.2 資產能否在 LOE 後形成可重複的產品平台，才是下一段考題。 2026 年 7 月 10 日，美國 FDA 再度擴大 Padcev（enfortumab vedotin-ejfv）聯合 Keytruda（pembrolizumab）的使用範圍：這次進入的是肌層浸潤性膀胱癌（MIBC）手術前後的圍手術期治療，而且不再受患者是否適合 cisplatin 限制。 對安斯泰來而言，這不只是一張新適應症標籤。它真正回答的是更大的問題：當 Xtandi 的專利與市場獨占壓力逐步逼近，安斯泰來是否已經從「一款超級產品撐起腫瘤業務」，跨進能持續產生下一款產品的腫瘤平台？ 答案不是簡單的「是」或「否」。 Padcev、Vyloy 與 Xospata 已證明安斯泰來不再只有 Xtandi；但從產品組合、權益結構、臨床採納到下一批管線的兌現速度來看，平台雛形已經出現，平台能力仍在驗證途中。 > 本文所稱 FY2025，是安斯泰來截至 2026 年 3 月 31 日的會計年度，不等於曆年 2025。所有產品數字均按公司 FY2025 補充財務資料口徑整理。 ## 第一層：四款產品已成形，但營收重心仍高度集中 安斯泰來 FY2025 的四款主要腫瘤產品全球銷售分別為： - Xtandi：9,608 億日圓 - Padcev：2,212 億日圓 - Xospata：718 億日圓 - Vyloy：631 億日圓 Padcev、Xospata 與 Vyloy 三款非 Xtandi 產品簡單合計為 3,561 億日圓，相當於 Xtandi 的 37.1%。這個計算可以用來觀察產品集中度，但不能被包裝成安斯泰來正式披露的「腫瘤部門營收」。 這組數字同時說明兩件事。 第一，安斯泰來已經跨過「第二支柱不存在」的階段。Padcev 在泌尿上皮癌與膀胱癌持續擴張，Vyloy 打開 CLDN18.2 精準治療市場，Xospata 則在帶 FLT3 突變的復發或難治性急性骨髓性白血病（AML）維持穩定貢獻。 第二，Xtandi 依然大到足以左右整家公司。三個非 Xtandi 產品加總仍不到 Xtandi 的四成，因此「已有多產品組合」不等於「已完成營收替代」。接下來真正要看的，不是新品名單有多長，而是非 Xtandi 產品能否以更快速度放量，並在獲利品質上接住 Xtandi 之後的缺口。 ## 第二層：產品銷售不等於安斯泰來可以完整留下的經濟利益 判斷安斯泰來的腫瘤轉型，不能只把四個產品的銷售額加總。 公司 FY2025 補充資料明確註記，Padcev 採用 in-market sales 口徑，美洲數字依 Pfizer 入帳資料計算。這代表 Padcev 的市場規模與安斯泰來最終可保留的收入、分潤及利潤並不是同一件事。它是一項重要資產，但也是合作產品，不能把全部終端銷售直接視為安斯泰來的全額經濟利益。 Xtandi 也有相似的判讀問題。安斯泰來管理層在 2026 年摩根大通醫療健康大會簡報講稿中表示，美國 Xtandi 共同推廣安排需要向 Pfizer 支付接近一半的銷售額；管理層同時強調，多數策略品牌的權益結構與毛利條件優於 Xtandi。這是公司對產品經濟性的說明，不等於外部投資人可以僅憑終端銷售額推算各產品淨利。 因此，安斯泰來轉型是否成功，至少要同時看兩張表： 1. 產品在市場端賣了多少； 2. 公司實際保留多少權益、收入與現金流。 若只看第一張表，容易高估合作產品對利潤的即時貢獻；若只看第二張表，又會低估 Padcev 等產品建立臨床地位與商業規模的戰略價值。 ## 第三層：Padcev 已跨進治癒意圖場景，下一關是臨床採納 Padcev 是目前最有機會加快產品組合去集中化的資產。 2026 年 7 月 10 日的美國核准，建立在第三期 EV-304／KEYNOTE-B15 研究上。Padcev 聯合 Keytruda 用於術前新輔助治療與術後輔助治療，相較 gemcitabine 加 cisplatin，在事件無惡化存活期（EFS）呈現 HR 0.53，整體存活期（OS）呈現 HR 0.65。兩年 EFS 率為 79.4% 對 66.2%，病理完全緩解率為 55.8% 對 32.5%。 這些 HR 描述的是相對風險，不是治癒率，也不能直接換算成每位患者的絕對獲益。 安全性同樣不能被省略。公司公告顯示，試驗中 Grade 3 以上治療期間不良事件發生率為 75.7%，對照組為 67.2%。這不代表組合不可用，但提醒市場：從核准走向廣泛採納，不能只看療效曲線，也要看醫院能否管理毒性、患者完成術前與術後療程的比例，以及多專科團隊如何安排手術、免疫治療與 ADC 的順序。 這次核准也要和 2025 年 11 月的版本分開理解。前一次圍手術期核准只適用於不適合 cisplatin 的成人 MIBC 患者；2026 年 7 月的新核准則把適用範圍擴展至不論 cisplatin 資格的患者。 但「可使用」與「被廣泛使用」仍是兩個階段。新標籤剛落地，官方資料尚不能回答治療中心滲透率、支付方覆蓋、實際完成療程比例與真實世界停藥率。這些不是目前已證實的成果，而是下一輪應追蹤的採納指標。 換句話說，Padcev 的臨床價值已經再向前一步，商業價值仍需要靠真實採納來兌現。 ## 第四層：Vyloy 先拿到門票，但 CLDN18.2 不會永遠只有一位玩家 Vyloy（zolbetuximab-clzb）是安斯泰來第二條最值得追蹤的成長曲線。 美國 FDA 核准的邊界非常清楚：Vyloy 必須搭配 fluoropyrimidine 與 platinum 化療，用於經 FDA 核准檢測確認 CLDN18.2 陽性、HER2 陰性的局部晚期不可切除或轉移性胃癌／胃食道接合部（GEJ）腺癌成人第一線治療。它不是「所有胃癌都能用」，也不是單藥廣泛適用。 在 SPOTLIGHT 研究中，Vyloy 組中位無惡化存活期為 10.6 個月，對照組為 8.7 個月，HR 0.751；中位整體存活期為 18.2 個月對 15.5 個月，HR 0.750。GLOW 研究也達成無惡化存活期與整體存活期主要終點。這些結果建立了 CLDN18.2 作為治療標誌物的臨床基礎，但商業化還取決於檢測覆蓋、患者篩選與醫師對化療組合的採用。 Vyloy FY2025 銷售為 631 億日圓，較前一年增加 509 億日圓、增幅 415.6%。這是高速成長，但也帶有低基期效應；判斷它是否成為真正支柱，應持續看絕對銷售額、地區擴張及新組合，而不是只看百分比。 安斯泰來正用不同模態補強 CLDN18.2 版圖。ASP2138 是公司正在推進的 CLDN18.2／CD3 雙特異性抗體；2025 年又從信諾維引進 CLDN18.2 ADC XNW27011（後續代號 ASP546C）。該交易包含 1.30 億美元 upfront、最高 7,000 萬美元近期款，以及最高 13.4 億美元開發與商業里程碑加權利金。里程碑是條件式上限，不是已付款交易總額；而且安斯泰來取得的權利明確排除中國大陸、香港、澳門與台灣。 先發優勢也不等於永久壟斷。AstraZeneca 的 CLDN18.2 ADC AZD0901 已有第二線以上胃癌／GEJ 癌第三期研究；信達生物的 CLDN18.2 ADC IBI343（arcotatug tavatecan）則在已治療胃癌／GEJ 癌第三期研究達成主要終點，並於 2026 年 6 月在中國獲受理新藥上市申請。 這些資產與 Vyloy 的美國第一線化療合併標籤處在不同治療線別與市場，不能直接視為同條件頭對頭競爭；但它們說明 CLDN18.2 的競爭正在從單抗擴展到 ADC，並跨越治療線別與地區。安斯泰來能否守住優勢，最終要看檢測入口、治療排序、聯合方案與下一代資產，而不只是「第一個上市」。 ## 第五層：Xospata 提供底座，卻不能被誤寫成更廣泛的 AML 平台 Xospata（gilteritinib）讓安斯泰來在血液腫瘤保有一個穩定商業化支點。其美國核准適用於經核准檢測確認帶 FLT3 突變的成人復發或難治性 AML，不應延伸成所有 AML，更不能寫成初治 AML 的普遍標準。 Xospata FY2025 銷售為 718 億日圓，年增 5.7%。它的角色比較像穩定底座，而不是足以單獨填補 Xtandi 缺口的爆發型第二曲線。對平台判斷而言，Xospata 的價值在於證明安斯泰來能在實體瘤之外經營精準血液腫瘤產品，但後續仍要觀察適應症拓展與新資產接續。 台灣供應鏈的敘事也必須留在證據邊界內。台灣神隆（1789）在量產原料藥清單中將 enzalutamide 列為美國申報、台灣生產的商業化原料藥，這可以作為 Xtandi 學名藥 API 與專利懸崖供應鏈的觀察座標。 但公開資料不能證明台灣神隆是安斯泰來或 Xtandi 原廠供應商，也不能外推成其參與 Padcev、Vyloy 或 Xospata。這篇文章因此不硬湊「安斯泰來概念股」；台灣公司若沒有可驗證的直接角色，就應誠實標示沒有直接映射。 ## 第六層：公司已為 Xtandi 之後準備資本，但目標不是成果 安斯泰來在官方文件中已把克服 Xtandi 喪失獨占權（LOE）列為企業級任務，並表示相關影響將從 2027 年起加速。由於不同地區的專利、法規獨占與訴訟安排不同，本文不硬寫一個全球通用的單一到期日。 CSP2026 將 FY2026 至 FY2030 設為五年執行期，目標包括： - 策略品牌銷售相較 FY2025 成長至兩倍； - 期間啟動 10 項以上第三期或關鍵試驗，其中截至 FY2027（含）啟動 5 項以上； - 以 ADC／iADC、T-cell engager、標靶蛋白降解、小分子、細胞及基因／mRNA 等工具推進管線； - 建立到 2030 年代中期約 1 兆日圓的管線收入潛力。 這些都是公司目標與潛力值，不是已實現營收，更不能等同於所有模態均已被臨床驗證。 資源端也有一個值得追蹤的交換。安斯泰來表示 FY2024 至 FY2025 已完成約 650 億日圓成本優化，並在 FY2026 至 FY2030 追求約 2,000 億日圓的經常性成本優化，釋出的資源將再投資於管線與成長。這不只是「省錢」故事；真正的考題是，節省下來的資本能否提高研發成功率、加快關鍵試驗並轉成更高品質的產品權益。 ## 四道驗證門：什麼時候才可以說平台真的成立？ 現階段最精準的說法，是安斯泰來已經形成多產品、多癌種、多模態的腫瘤平台雛形，但還沒有完全證明可重複產出下一個成長支柱。 後續可以用四道門檢驗： ### 1. 產品組合與經濟利益 非 Xtandi 產品的絕對銷售額與占比是否持續提升？公司最終保留的收入、權利與毛利，能否比只看 in-market sales 更扎實地接住 LOE 缺口？ ### 2. Padcev 的臨床採納 圍手術期核准能否轉化為治療中心滲透、支付覆蓋、療程完成與可管理的安全性？批准本身已完成，採納仍待證明。 ### 3. Vyloy 與下一批關鍵資產 CLDN18.2 的檢測、治療排序和聯合方案能否形成護城河？ASP2138、ASP546C 及其他關鍵項目能否在競爭者加速時按期跨過概念驗證、第三期、核准與商業化？ ### 4. 資本效率與平台再生能力 2,000 億日圓成本優化、10 項以上關鍵試驗與約 1 兆日圓中期管線潛力，能否從目標表變成核准、收入與現金流？只有當第一代產品賺到的錢，能穩定長出下一代產品，平台才算真正成立。 ## 結論：安斯泰來已不是「只有 Xtandi」，但還不能提前封王 Padcev 把 ADC 從晚期泌尿上皮癌帶進膀胱癌手術前後；Vyloy 建立 CLDN18.2 第一線精準治療入口；Xospata 提供血液腫瘤的穩定底座。這些成果足以推翻「安斯泰來只有 Xtandi」的舊印象。 但一家成熟的腫瘤平台，不只要有多款產品，還要能重複做到發現或引進、臨床驗證、核准、全球商業化與再投資。安斯泰來目前已走到中段：產品組合存在，策略與資本也到位；接下來要交的成績單，是更分散且更高品質的經濟利益、Padcev 的真實採納、CLDN18.2 的競爭防守，以及下一批資產能否按時撞線。 真正的問題因此不是「日本是不是又出了一家腫瘤巨頭」，而是安斯泰來能否把一次次單點成功，變成可以重複的公司能力。 --- ## 一級來源 1. [Astellas FY2025 Financial Results Presentation](https://www.astellas.com/content/dam/astellas-com/global/en/confidential-documents/financial-results/4q2025_pre_en.pdf) 2. [Astellas FY2025 Supplementary Materials](https://www.astellas.com/content/dam/astellas-com/global/en/confidential-documents/financial-results/4q2025_sup_en.pdf) 3. [Astellas／Pfizer：Padcev＋Keytruda 圍手術期 MIBC 核准公告，2026-07-13](https://newsroom.astellas.com/2026-07-13-us-fda-approves-padcev-plus-keytruda-as-neoadjuvant-and-adjuvant-treatment-for-muscle-invasive-bladder-cancer-regardless-of-cisplatin-eligibility) 4. [FDA：2025 年 Padcev＋Keytruda 圍手術期 MIBC 核准](https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-pembrolizumab-enfortumab-vedotin-ejfv-muscle-invasive-bladder-cancer) 5. [FDA：Vyloy 核准與 SPOTLIGHT／GLOW 摘要](https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-zolbetuximab-clzb-chemotherapy-gastric-or-gastroesophageal-junction-adenocarcinoma) 6. [Astellas／Evopoint：XNW27011 授權公告](https://newsroom.astellas.com/2025-05-30-CLDN18-2-XNW27011) 7. [ClinicalTrials.gov：AZD0901 第二線以上胃癌／GEJ 癌第三期研究 NCT06346392](https://clinicaltrials.gov/study/NCT06346392) 8. [Innovent：IBI343 第三期達成主要終點與中國 NDA 獲受理，2026-06-04](https://en.innoventbio.com/InvestorsAndMedia/PressReleaseDetail?key=601) 9. [FDA：Xospata 核准](https://www.fda.gov/drugs/fda-approves-gilteritinib-relapsed-or-refractory-acute-myeloid-leukemia-aml-flt3-mutation) 10. [Astellas：CSP2026 Presentation](https://www.astellas.com/content/dam/astellas-com/global/en/confidential-documents/Investors-Meeting/astellas_csp2026_presentation_20260526_en.pdf) 11. [Astellas：2026 JPM Healthcare Conference Script](https://www.astellas.com/content/dam/astellas-com/global/en/confidential-documents/Investors-Meeting/44th_Annual_JPM_Healthcare_Conference_meeting_script_Astellas.pdf) 12. [台灣神隆：量產原料藥清單](https://www.scinopharm.com/tw/products-detail/commercialAPI/) > 本文僅為公開資訊整理與產業研究，不構成醫療建議、個別診療建議或投資建議。藥品適用對象、療效與風險應以各地主管機關核准標籤及醫療專業人員判斷為準。 --- ## After Xtandi: Can Astellas Build an Oncology Platform? URL: https://drugnews.com.tw/articles/2026-07-23-astellas-oncology-platform-en.html Date: 2026-07-30 Language: en Summary: Xtandi still dominates Astellas oncology sales. The real test is whether Padcev, Vyloy and the next CLDN18.2 assets can become a repeatable post-LOE platform. # What Comes After Xtandi? Four Tests for Astellas’ Emerging Oncology Platform On July 10, 2026, the US Food and Drug Administration expanded the use of Padcev (enfortumab vedotin-ejfv) plus Keytruda (pembrolizumab) once again. The combination can now be used before and after surgery for muscle-invasive bladder cancer (MIBC), regardless of whether a patient is eligible for cisplatin. For Astellas, this is more than another indication on a label. It speaks to a larger question: as patent and market-exclusivity pressure builds around Xtandi, has Astellas moved from an oncology business supported by one exceptionally large product to a platform capable of producing successive growth engines? The answer is not a simple yes or no. Padcev, Vyloy and Xospata show that Astellas is no longer a one-product oncology story. Yet the composition of sales, underlying economic rights, clinical adoption and the delivery of the next pipeline wave all point to the same conclusion: a platform is taking shape, but its repeatability is still being tested. > In this article, FY2025 refers to Astellas’ fiscal year ended March 31, 2026, not calendar year 2025. All product figures use the company’s FY2025 supplementary financial reporting conventions. ## Layer one: four products are on the market, but the portfolio remains concentrated Astellas reported the following FY2025 global sales for its four major oncology products: - Xtandi: JPY 960.8 billion - Padcev: JPY 221.2 billion - Xospata: JPY 71.8 billion - Vyloy: JPY 63.1 billion Padcev, Xospata and Vyloy generated JPY 356.1 billion in aggregate, equal to 37.1% of Xtandi sales. That calculation is useful as a measure of product concentration, but it is not a company-reported “oncology segment revenue” figure. These numbers establish two points at once. First, Astellas has moved beyond the stage in which no second pillar existed. Padcev continues to expand across urothelial and bladder cancer. Vyloy has opened a precision-treatment market around CLDN18.2. Xospata provides a stable contribution in adults with relapsed or refractory FLT3-mutated acute myeloid leukemia (AML). Second, Xtandi remains large enough to shape the entire company. The three non-Xtandi products combined still represent less than 40% of Xtandi sales. A multi-product portfolio is therefore not the same as a completed revenue replacement. The key issue is no longer how many new product names Astellas can list. It is whether the non-Xtandi portfolio can scale quickly enough—and with sufficient earnings quality—to absorb the gap that follows Xtandi. ## Layer two: product sales are not the same as the economics Astellas retains Adding four product sales figures is not sufficient to evaluate Astellas’ oncology transition. Astellas’ FY2025 supplementary materials explicitly present Padcev on an in-market-sales basis and state that the Americas figure is based on sales booked by Pfizer. The market size of Padcev is therefore not identical to the revenue, profit share or cash economics ultimately retained by Astellas. Padcev is a strategically important asset, but it is also a partnered product; total end-market sales should not be described as Astellas’ wholly retained economics. Xtandi requires a similar distinction. In its script for the 2026 JPM Healthcare Conference, Astellas management said that its US Xtandi co-promotion arrangement requires a payment to Pfizer of almost half of sales. Management also argued that most of the company’s Strategic Brands carry more favorable ownership and gross-margin characteristics than Xtandi. This is management’s description of product economics; it does not allow outside investors to derive product-level net profit from end-market sales alone. A credible assessment of the transition therefore requires two scorecards: 1. how much each product sells in the market; and 2. how much ownership, revenue and cash flow Astellas ultimately retains. Looking only at the first may overstate the immediate profit contribution of partnered assets. Looking only at the second may understate the strategic value of Padcev and other products that establish clinical position and commercial scale. ## Layer three: Padcev has entered a curative-intent setting; adoption is the next test Padcev is the asset with the greatest near-term potential to reduce portfolio concentration. The July 10, 2026 US approval was based on the phase 3 EV-304/KEYNOTE-B15 trial. Padcev plus Keytruda was given before surgery as neoadjuvant treatment and after surgery as adjuvant treatment. Compared with gemcitabine plus cisplatin, the regimen produced a hazard ratio of 0.53 for event-free survival (EFS) and 0.65 for overall survival (OS). Two-year EFS was 79.4% versus 66.2%, while the pathologic complete response rate was 55.8% versus 32.5%. Those hazard ratios describe relative risk. They are not cure rates and should not be converted into a presumed absolute benefit for every patient. Safety also belongs in the analysis. According to the company announcement, grade 3 or higher treatment-emergent adverse events occurred in 75.7% of patients in the Padcev-plus-Keytruda arm and 67.2% in the comparator arm. That does not mean the combination should not be used. It does mean that broad adoption will depend on more than an efficacy curve: hospitals must be able to manage toxicity, patients must be able to complete the pre- and postoperative course, and multidisciplinary teams must coordinate surgery, immunotherapy and an antibody-drug conjugate. The new approval should also be distinguished from the November 2025 decision. The earlier perioperative approval applied only to adults with MIBC who were ineligible for cisplatin. The July 2026 label expanded the eligible population regardless of cisplatin eligibility. However, availability and broad utilization are different milestones. Because the expanded label is new, official sources do not yet answer questions such as treatment-center penetration, payer coverage, real-world completion of the full regimen or discontinuation rates. These are not established achievements; they are the next adoption metrics to monitor. Padcev has advanced clinically. Its additional commercial value must now be earned through real-world adoption. ## Layer four: Vyloy has first-mover advantage, but CLDN18.2 will not remain a one-player field Vyloy (zolbetuximab-clzb) is Astellas’ second major growth curve to watch. Its US label has precise boundaries. Vyloy must be combined with fluoropyrimidine- and platinum-containing chemotherapy for the first-line treatment of adults with locally advanced unresectable or metastatic gastric or gastroesophageal junction (GEJ) adenocarcinoma that is CLDN18.2-positive and HER2-negative, as determined by an FDA-approved test. It is not a treatment for every gastric cancer patient, nor is it broadly approved as monotherapy. In SPOTLIGHT, median progression-free survival was 10.6 months with the Vyloy regimen and 8.7 months in the control group, for a hazard ratio of 0.751. Median overall survival was 18.2 months versus 15.5 months, for a hazard ratio of 0.750. GLOW also met its primary progression-free and overall-survival endpoints. These trials established the clinical foundation for CLDN18.2 as a treatment biomarker. Commercial execution still depends on testing access, patient identification and physician acceptance of the chemotherapy combinations. Vyloy generated JPY 63.1 billion in FY2025 sales, an increase of JPY 50.9 billion, or 415.6%, from the prior year. The growth rate is substantial but also reflects a small base. The better tests of whether Vyloy becomes a durable pillar will be absolute sales, geographic expansion and progress with new combinations, rather than the percentage growth rate alone. Astellas is building around CLDN18.2 with multiple modalities. ASP2138 is the company’s investigational CLDN18.2/CD3 bispecific antibody. In 2025, Astellas also licensed the CLDN18.2 antibody-drug conjugate XNW27011, later designated ASP546C, from Evopoint. The agreement included a $130 million upfront payment, up to $70 million in near-term payments, and as much as $1.34 billion in development and commercial milestones plus royalties. Milestones are contingent maximums, not money already paid. The licensed territory also explicitly excludes mainland China, Hong Kong, Macau and Taiwan. First-mover status does not confer a permanent monopoly. AstraZeneca’s CLDN18.2 ADC AZD0901 is being studied in a phase 3 trial in second-line or later gastric and GEJ cancer. Innovent’s CLDN18.2 ADC IBI343, or arcotatug tavatecan, met the primary endpoint in a phase 3 trial in previously treated gastric and GEJ cancer, and a new drug application was accepted in China in June 2026. These programs operate in different treatment lines and markets from Vyloy’s US first-line chemotherapy-combination label. They should not be described as direct, like-for-like head-to-head competitors. They do show that CLDN18.2 competition is expanding across modalities, treatment lines and geographies. Astellas’ moat will ultimately depend on diagnostic access, treatment sequencing, combination strategies and follow-on assets—not only on being first. ## Layer five: Xospata provides a base, but it should not be overstated as a broad AML platform Xospata (gilteritinib) gives Astellas a stable commercial position in hematologic oncology. Its US approval covers adults with relapsed or refractory AML carrying a FLT3 mutation, as detected by an approved test. The indication should not be expanded in writing to all AML patients or presented as a general first-line standard. Xospata generated JPY 71.8 billion in FY2025 sales, up 5.7%. Its role is closer to that of a durable base than a rapidly expanding second curve capable of replacing Xtandi by itself. For the platform thesis, Xospata demonstrates that Astellas can commercialize a precision hematology product outside solid tumors. Its longer-term contribution will depend on indication development and the assets that follow it. The Taiwan supply-chain discussion also needs to remain within the evidence. ScinoPharm Taiwan (1789) lists enzalutamide as a commercial active pharmaceutical ingredient manufactured in Taiwan and filed in the United States. This can be monitored as one data point in the generic-API supply chain around Xtandi’s loss of exclusivity. Public information does not establish that ScinoPharm supplies Astellas or the originator Xtandi product, and it does not support a role in Padcev, Vyloy or Xospata. We therefore do not force a “Taiwan Astellas concept stock” narrative. When a Taiwan-listed company has no verifiable direct role, the responsible conclusion is that no direct mapping has been established. ## Layer six: Astellas has prepared capital for life after Xtandi, but targets are not results Astellas has identified overcoming Xtandi’s loss of exclusivity as an enterprise-level task and has said the impact will accelerate from 2027. Patent terms, regulatory exclusivities and litigation arrangements vary across markets, so this article does not impose a single worldwide patent-expiry date. CSP2026 covers FY2026 through FY2030. Its objectives include: - doubling Strategic Brand sales from the FY2025 base; - initiating more than ten phase 3 or pivotal studies during the plan, including more than five by the end of FY2027; - advancing the pipeline with a toolkit that includes ADCs and immune-stimulating ADCs, T-cell engagers, targeted protein degradation, small molecules, cell therapies and gene or mRNA approaches; and - building approximately JPY 1 trillion in pipeline revenue potential for the mid-2030s. These are management targets and potential values, not realized revenue. They do not show that every modality has already been clinically validated. There is also a capital-allocation trade. Astellas says it completed approximately JPY 65 billion of cost optimization in FY2024 and FY2025 and is targeting roughly JPY 200 billion of recurring cost optimization from FY2026 through FY2030, with resources reinvested in the pipeline and growth. This is more than a cost-cutting story. The real test is whether the released capital improves R&D productivity, accelerates pivotal development and produces assets with better ownership economics. ## Four gates: when can the platform be considered proven? The most precise conclusion today is that Astellas has the outline of a multi-product, multi-tumor and multi-modality oncology platform, but it has not yet fully demonstrated that it can repeatedly generate the next growth pillar. Four gates can be used to track that transition. ### 1. Product mix and retained economics Do absolute sales and the share of non-Xtandi products continue to rise? After partnerships and co-promotion arrangements, can the revenue, ownership and gross profit retained by Astellas absorb the loss-of-exclusivity gap more convincingly than headline in-market sales alone? ### 2. Clinical adoption of Padcev Does the perioperative approval translate into treatment-center penetration, payer coverage, regimen completion and manageable safety? Approval is established; adoption remains to be demonstrated. ### 3. Vyloy and the next pivotal assets Can CLDN18.2 testing, sequencing and combinations create a defensible position? Can ASP2138, ASP546C and other pivotal programs move through proof of concept, phase 3 development, approval and commercialization on schedule while competitors accelerate? ### 4. Capital efficiency and platform regeneration Can JPY 200 billion of targeted cost optimization, more than ten pivotal starts and approximately JPY 1 trillion of mid-2030s pipeline potential be converted from a planning slide into approvals, revenue and cash flow? A platform becomes durable only when cash generated by one product can repeatedly create the next. ## Conclusion: Astellas is no longer “only Xtandi,” but it is too early to declare victory Padcev has carried an ADC from advanced urothelial cancer into treatment before and after bladder-cancer surgery. Vyloy has created a first-line precision-treatment entry point around CLDN18.2. Xospata provides a stable base in hematologic oncology. Together, these achievements are enough to retire the old idea that Astellas has only Xtandi. A mature oncology platform, however, needs more than several marketed products. It must repeatedly discover or acquire assets, validate them clinically, obtain approval, commercialize globally and reinvest the proceeds. Astellas is in the middle of that transition: the portfolio exists, and the strategy and capital are in place. The next scorecard is whether it can produce more diversified and higher-quality economics, real adoption of Padcev, a defensible CLDN18.2 position and timely delivery from the next pipeline wave. The central question is therefore not whether Japan has produced another oncology giant. It is whether Astellas can turn a series of individual successes into a repeatable organizational capability. --- --- ## Pelacarsen's HORIZON Test: Does Lowering Lp(a) Prevent Cardiovascular Events? URL: https://drugnews.com.tw/articles/2026-07-19-pelacarsen-lpa-horizon-en.html Date: 2026-07-29 Language: en Summary: Pelacarsen can substantially lower Lp(a), but HORIZON must show whether biomarker reduction on top of standard care translates into fewer cardiovascular deaths, myocardial infarctions, strokes and urgent revascularizations. Start with the timing, because this is a topic where one premature verb can distort the entire investment and clinical thesis. As of July 19, 2026, the phase 3 Lp(a) HORIZON trial of pelacarsen has not reported results. ClinicalTrials.gov still lists the study as active, not recruiting, and contains no results section. Novartis' latest quarterly investor presentation places the event-driven readout in the second half of 2026. Ionis says results are expected later this year. That means it is not accurate to write that the trial has unblinded, that the result is imminent within a specific number of weeks, or that the endpoint has already been met. HORIZON nevertheless matters far beyond a single corporate catalyst. It is designed to answer one of the most consequential unresolved questions in cardiovascular drug development: If lipoprotein(a), or Lp(a), is reduced substantially in patients who already have cardiovascular disease and are receiving optimized standard care, will they experience fewer cardiovascular deaths, myocardial infarctions, strokes and urgent coronary revascularizations? The answer will determine whether Lp(a) moves from being a compelling causal risk factor to a druggable outcomes target. It will also shape the development path for a growing field of antisense oligonucleotides, siRNA agents and oral small molecules. Are these programs simply lowering an attractive biomarker, or are they building a new cardiovascular treatment category? ## 01 | Why Lp(a) and LDL-C Are Not the Same Problem Lp(a) can be understood as an apoB-containing lipoprotein particle with an additional apolipoprotein(a), or apo(a), component. Its concentration is strongly influenced by genetics and is associated with atherosclerotic cardiovascular disease and calcific aortic-valve stenosis. The clinical problem is residual risk. A patient's LDL cholesterol may be brought down through statins, ezetimibe, drugs targeting the PCSK9 pathway and other evidence-based therapies. That does not mean the Lp(a) component of cardiovascular risk has been resolved. This is the background against which HORIZON should be read. Pelacarsen is not being tested in a therapeutic vacuum. It is being added to optimized standard cardiovascular care. The trial is therefore not asking whether Lp(a) management should replace LDL-C management. It is asking whether directly addressing Lp(a) can reduce events after the established parts of risk reduction are already in place. Genetic studies, epidemiology and biomarker analyses can make the biological direction persuasive. They cannot substitute for a randomized cardiovascular outcomes trial. A causal association does not automatically tell clinicians how much pharmacologic lowering is required, when intervention should begin, which patients benefit most, or whether the effect is large enough to change practice. That evidentiary gap is precisely what HORIZON was designed to close. ## 02 | Pelacarsen Has Shown That It Can Lower Lp(a), Not Yet That It Can Prevent Events Pelacarsen is an antisense oligonucleotide, or ASO, directed at the apo(a) production pathway. In simplified terms, it reduces the hepatic production of apo(a), leaving less material available for Lp(a) assembly. In a randomized, double-blind, placebo-controlled phase 2 trial involving 286 patients, the tested dose regimens produced mean Lp(a) reductions of approximately 35% to 80%. The most intensive regimen achieved a mean reduction of about 80%. The most common adverse events were injection-site reactions. In that study, there were no significant between-group differences in platelet counts, liver or kidney laboratory measures, or influenza-like symptoms. These findings are important proof of mechanism. They are not an outcomes victory. The phase 2 study primarily answered whether the drug engaged its target, how large the biomarker effect could be and what the short- to medium-term safety profile looked like. HORIZON asks whether patients ultimately experience fewer major cardiovascular events. The distinction is not semantic. An 80% reduction in Lp(a) must never be translated into an 80% reduction in cardiovascular risk. The percentage change in a biomarker and the relative risk reduction in clinical events are not interchangeable. ## 03 | The Trial Design Determines How the Result Should Be Read HORIZON enrolled 8,323 patients with established cardiovascular disease and Lp(a) of at least 70 mg/dL. Established disease includes a history of myocardial infarction, ischemic stroke or clinically significant symptomatic peripheral artery disease. Participants were assigned 1:1 to receive pelacarsen 80 mg by subcutaneous injection once monthly or placebo, while continuing optimized standard cardiovascular treatment. The primary composite endpoint includes: - cardiovascular death; - nonfatal myocardial infarction; - nonfatal stroke; and - urgent coronary revascularization requiring hospitalization. The study will test the overall population and a higher-Lp(a) population defined at screening by Lp(a) of at least 90 mg/dL, with multiplicity handled prospectively. That detail matters. If the clearest result emerges only in patients with higher baseline Lp(a), the finding cannot be reframed as proof that every person with an elevated measurement benefits equally. HORIZON is also event driven. The design calls for 993 primary cardiovascular events confirmed by an independent clinical events committee, with a minimum follow-up of 2.5 years. Reaching a calendar date does not make the data mature automatically. Event accumulation, data cleaning, database lock, statistical analysis and corporate disclosure timing can all affect when a result becomes public. Novartis' “H2 2026, event-driven” language should therefore be treated as the company's current readout window, not as a guaranteed publication date and certainly not as evidence that the result is already known publicly. Three different milestones are often collapsed into one. The first is event maturity: the trial reaches its target event count and the dataset moves toward lock and analysis. The second is a corporate topline announcement: usually a concise disclosure of the primary endpoint and a limited number of safety findings. The third is a full scientific presentation or publication, which may provide baseline characteristics, individual endpoint curves, subgroups, discontinuations and complete adverse-event data. Even when a sponsor says a trial is complete or a readout has been achieved, public readers may not yet have enough information to judge every important question. Conversely, a shift in timing should not be interpreted as success or failure without a primary-source explanation. Event-driven trials can move because events accrue faster or more slowly than expected. That is why the verification date belongs at the beginning of this article. HORIZON is a thesis that could be rewritten instantly by new data. Any future headline must be checked again against the trial registry, the sponsor's formal release and the complete scientific dataset. ## 04 | When the Result Arrives, Do Not Stop at “Met” or “Missed” A press release can reduce a complex cardiovascular outcomes trial to positive or negative. A useful reading of HORIZON requires at least five layers. ### 1. Did the prespecified hypothesis pass? Start with the overall population and the prespecified Lp(a) at least 90 mg/dL population. Examine how multiplicity was controlled. A nominally significant subgroup does not necessarily mean the trial succeeded statistically as designed. ### 2. How large is the effect? The hazard ratio matters, but so do its 95% confidence interval, the absolute difference in event rates, follow-up duration and the interpretability of the number needed to treat. A very large trial can produce an attractive p value from a modest difference. Clinicians and payers will ask whether the magnitude is sufficient to change care. ### 3. Which events drive the composite? Cardiovascular death, myocardial infarction, stroke and urgent revascularization do not carry identical clinical weight. If the composite is driven primarily by a component more sensitive to medical decision-making, the result will be interpreted differently from a consistent reduction in death, myocardial infarction and stroke. ### 4. Does the biomarker effect connect coherently to the outcomes effect? Readers should look at the Lp(a) reduction achieved in the treatment group, the baseline distribution, whether higher-Lp(a) patients show clearer benefit and whether the relationship between biomarker reduction and events makes biological and clinical sense. This is the bridge between hitting the target and helping the patient. ### 5. What is the long-term cost in safety, persistence and treatment burden? Injection-site reactions were the most common adverse events in phase 2, but phase 3 is much larger and longer. The result must be examined for serious adverse events, liver, kidney or platelet signals, discontinuation rates and the effect of monthly injections on long-term adherence. Efficacy does not automatically make a chronic therapy easy to adopt. Together, these five layers define the real HORIZON result. ## 05 | Why One Readout Could Reprice the Entire Lp(a) Field Pelacarsen is not the only program waiting for an outcomes answer. Amgen's olpasiran is being studied in the phase 3 OCEAN(a)-Outcomes trial. The study uses subcutaneous dosing every 12 weeks in patients with Lp(a) of at least 200 nmol/L and established atherosclerotic cardiovascular disease. Eli Lilly's lepodisiran is also in a phase 3 cardiovascular outcomes program, ACCLAIM-Lp(a). The same company is developing the oral small molecule muvalaplin in the phase 3 MOVE-Lp(a) outcomes trial. The useful comparison is not a leaderboard of which phase 2 program lowers Lp(a) by the highest percentage. The strategic question is whether the first persuasive outcomes result establishes a clinically acceptable bar for the target as a whole. If HORIZON shows a robust event benefit, with a meaningful effect size, convincing component events and an acceptable safety profile, competition will shift. The debate will move from whether Lp(a) is druggable to which mechanism, formulation, dosing interval, patient segment and price delivers the strongest overall product. If the result is neutral, the field will need to separate several possible explanations: target biology, timing of intervention, enrollment threshold, follow-up duration, degree of Lp(a) lowering, background therapy and trial design. One molecule missing an endpoint would not prove that every Lp(a) strategy is biologically dead. But a deeper biomarker reduction by a competitor would not make the outcomes question disappear either. This is why HORIZON is both a product catalyst and a target-validation event. ## 06 | What Taiwan Readers Should Watch The first issue is measurement. HORIZON uses mg/dL thresholds, while the design paper also provides approximate nmol/L equivalents. Because apo(a) particle size varies, values should not be converted for every patient with one fixed ratio. Read the original unit and assay context. The second issue is population. This is a secondary-prevention trial. Participants already have a history of myocardial infarction, ischemic stroke or symptomatic peripheral artery disease. Even a positive study could not be applied automatically to otherwise healthy people with no previous cardiovascular event, and it would not justify self-directed testing or treatment. The third issue is local adoption. A global trial result would still need to pass through regulatory labeling, test availability, specialist care pathways, reimbursement and pricing in Taiwan. Global success is a necessary condition for local use, not next-day access. The fourth issue is investment discipline. This analysis did not identify primary evidence that a Taiwan-listed company directly shares pelacarsen rights or HORIZON economics. Pelacarsen's central development relationship is between Novartis and Ionis; the major competing outcomes programs are associated with Amgen and Eli Lilly. Attaching an Lp(a) label to a Taiwan company through a loose manufacturing, diagnostic or thematic connection would obscure the actual clinical question. For investors, the most useful checklist is straightforward: 1. Which prespecified tests pass: the overall population, the Lp(a) at least 90 mg/dL population, or both? 2. What are the hazard ratio, 95% confidence interval and absolute event-rate difference? 3. Which component—cardiovascular death, myocardial infarction, stroke or urgent revascularization—drives the composite? 4. Does the achieved Lp(a) reduction connect coherently to the event benefit? 5. What do long-term safety, discontinuation and monthly-injection persistence look like? 6. Is the sponsor's next step a broad filing, a narrower subgroup strategy or additional research? ## Conclusion | The Test Is Not Whether Pelacarsen Lowers a Number HORIZON's value does not lie in proving once again that pelacarsen can lower Lp(a). Phase 2 has already provided a clear signal on that question. The harder threshold is converting the reduction of an inherited residual-risk factor, on top of standard therapy, into fewer myocardial infarctions, strokes, cardiovascular deaths and urgent revascularizations. As of July 19, 2026, that answer has not been reported. Novartis currently points to an event-driven readout in the second half of 2026. The most valuable preparation is not guessing the day of the press release. It is defining the evidentiary standard before the data arrive. Markets may take seconds to decide whether a headline is bullish or bearish. Medicine and industry still have to answer the more important question: can changing one laboratory measurement change a patient's cardiovascular fate? --- This article is provided for industry research and educational purposes only. It does not constitute investment, medical, fundraising or securities advice. Individual testing and treatment decisions should be made with a qualified healthcare professional based on the complete clinical context. --- ## HORIZON：Lp(a) 降了，心梗會少嗎？ URL: https://drugnews.com.tw/articles/2026-07-19-pelacarsen-lpa-horizon.html Date: 2026-07-29 Language: zh-Hant Summary: Pelacarsen 已證明能大幅降低 Lp(a)，但 HORIZON 真正要回答的是：在標準治療之上，生物標記下降能否轉成更少的心血管死亡、心肌梗塞、中風與緊急再血管化。 先把時間講清楚。 截至 2026 年 7 月 19 日，pelacarsen 的第三期 Lp(a) HORIZON 試驗還沒有公布結果。ClinicalTrials.gov 顯示研究仍是「Active, not recruiting」，也沒有結果資料；Novartis 最新一季投資人資料把事件驅動讀出放在 2026 年下半年，Ionis 則說結果預計在今年稍晚公布。 所以，現在不能寫「正式揭盲」、不能寫「結果出爐」，也不能把「2026 下半年」改寫成「未來幾週」。 但這不代表 HORIZON 只是又一個等待中的藥廠催化劑。它真正要回答的，是心血管藥物開發裡一個拖了很多年的問題： 把 Lp(a) 大幅壓低，能不能在既有心血管疾病、而且已接受標準治療的人身上，進一步減少心血管死亡、心肌梗塞、中風與緊急冠狀動脈再血管化？ 這個答案會決定 Lp(a) 是從「很有說服力的風險因子」，走到「值得被藥物直接治療的事件標的」；也會影響後面一整批 ASO、siRNA 與口服小分子，究竟是在追一個漂亮的生物標記，還是在建立新的心血管治療類別。 資料狀態：ClinicalTrials.gov NCT04023552（2026 年 7 月 19 日重查）；Novartis 2026 Q1 investor presentation；Ionis 2026 Q1 financial results。 ## 01｜Lp(a) 為什麼和 LDL-C 不是同一題？ Lp(a)，也就是脂蛋白(a)，可以先把它想成一顆帶有額外 apo(a) 結構的含 apoB 脂蛋白。它的濃度很大程度受遺傳影響，和動脈粥樣硬化性心血管疾病及鈣化性主動脈瓣狹窄風險有關。 臨床上麻煩的地方在於：一個人的 LDL-C 已經因為 statin、ezetimibe、PCSK9 路徑藥物與其他標準治療降下來，不代表 Lp(a) 這條殘餘風險也被處理掉。 這正是 HORIZON 的重要背景。它不是把 pelacarsen 放在沒有治療的真空裡，而是加在「最佳化的心血管標準照護」之上。換句話說，試驗要找的不是取代 LDL-C 管理的方法，而是問：當該做的標準治療已經做了，針對 Lp(a) 還能不能再把事件風險往下推？ 這個問題不能只靠遺傳研究、觀察性研究或生物標記降幅回答。那些證據可以告訴我們方向很合理，卻不能代替隨機對照的心血管事件試驗。 HORIZON 設計論文：https://pubmed.ncbi.nlm.nih.gov/40185318/ ## 02｜Pelacarsen 已經證明「能降」，還沒證明「能救事件」 Pelacarsen 是針對 apo(a) 生成路徑的反義寡核苷酸（ASO）。簡化來說，它在肝臟端降低 apo(a) 的製造，讓 Lp(a) 的組裝來源變少。 在 286 人的第二期隨機、雙盲、安慰劑對照試驗中，不同劑量方案讓 Lp(a) 平均下降約 35% 到 80%；最高強度方案的平均降幅約 80%。最常見的不良事件是注射部位反應，該試驗沒有看到血小板、肝腎指標或類流感症狀出現顯著組間差異。 這是很重要的 proof of mechanism，卻不是臨床終點的勝利。 因為第二期主要回答的是「藥有沒有打中標的、降幅有多大、短中期安全性如何」，HORIZON 才是在回答「病人最後有沒有少發生重大心血管事件」。兩個問題不能混寫。 尤其不能把「Lp(a) 降 80%」直接翻譯成「心血管風險降 80%」。生物標記的百分比與臨床事件的相對風險下降，從來不是一比一換算。 Pelacarsen 第二期論文：https://pubmed.ncbi.nlm.nih.gov/31893580/ ## 03｜HORIZON 的設計，決定我們之後該怎麼讀結果 HORIZON 共納入 8,323 位已有心血管疾病、且 Lp(a) 至少 70 mg/dL 的患者。這裡的既有心血管疾病，包括心肌梗塞、缺血性中風或有症狀的周邊動脈疾病病史。 受試者以 1:1 分派，每月皮下注射 pelacarsen 80 mg 或安慰劑，同時繼續接受最佳化的心血管標準治療。 主要複合終點包括： - 心血管死亡 - 非致死性心肌梗塞 - 非致死性中風 - 需要住院的緊急冠狀動脈再血管化 試驗會同時檢驗整體族群，以及篩檢時 Lp(a) 至少 90 mg/dL 的較高族群，並預先處理多重檢定問題。這一點很關鍵：如果結果只在更高 Lp(a) 族群清楚，市場與臨床解讀就不能偷換成「所有 Lp(a) 偏高者都受益」。 它也是一個事件驅動試驗。設計要求累積 993 件經獨立委員會確認的主要心血管事件，且最短追蹤 2.5 年。達到日曆上的某一天，不代表資料就自動成熟；事件累積速度、資料清理、統計鎖庫與公司發布節奏都會影響讀出時間。 因此 Novartis 所寫的「H2 2026（event-driven）」應該被理解為公司目前的讀出窗口，不是保證日期，更不是已公布結果。 這裡還有三個很容易混在一起的時間點。 第一是事件數成熟：試驗累積到預定事件、資料開始進入鎖庫與分析。第二是公司公布 topline：通常先交代主要終點與少數關鍵安全訊號。第三才是完整學會報告或論文：基線特徵、各終點曲線、子群、停藥與完整不良事件，往往要更晚才能看見。 所以，即使公司未來說「trial completed」或「data readout achieved」，也不代表公開讀者已經拿到足以判斷全部問題的資料。反過來，讀出窗口往後移，也不應在沒有一級說明時自行解讀成成功或失敗；事件驅動試驗的時程，本來就可能受事件累積速度影響。 這也是為什麼這篇把查證截止日寫在最前面。HORIZON 是一個會被新資料瞬間改寫的題目，任何結果新聞都必須回到試驗登錄、公司正式公告與完整科學資料重新核對，不能用舊稿直接續貼。 試驗登錄：https://clinicaltrials.gov/study/NCT04023552 Novartis 2026 Q1 資料：https://www.novartis.com/sites/novartis_com/files/q1-2026-investor-presentation.pdf ## 04｜結果公布時，先別只找「有沒有達標」 一篇新聞稿很容易只剩下 positive 或 negative，但 HORIZON 真正有用的讀法至少分五層。 第一層：主要終點有沒有過？ 先看整體族群與預先設定的 Lp(a) ≥90 mg/dL 子群各自結果，以及多重檢定如何控制。只看到某個子群名目上顯著，不等於試驗在統計上已完整成功。 第二層：效果有多大，不只看 p 值 要看 hazard ratio、95% 信賴區間、絕對事件率差、追蹤時間與 number needed to treat 的可解讀性。大型試驗可以用很小的差距得到漂亮 p 值，但臨床與支付者會追問：差距是否足以改變照護？ 第三層：是哪一種事件在拉動結果？ 心血管死亡、心肌梗塞、中風與緊急再血管化的重要性不完全相同。如果複合終點主要由較軟、較容易受醫療行為影響的項目推動，說服力會和死亡、心肌梗塞或中風一致下降不同。 第四層：Lp(a) 降幅和事件獲益是否有可理解的關係？ 要看治療組實際達到的 Lp(a) 降幅、基線分布、較高 Lp(a) 者是否出現更清楚的效益，以及降幅與事件之間是否呈現合理趨勢。這一層是把「標的有被打中」和「病人真的受益」接起來。 第五層：安全性、停藥率與每月注射的代價 第二期最常見的是注射部位反應，但第三期規模更大、時間更長。結果公布時，還要看嚴重不良事件、肝腎與血小板訊號、停藥率，以及每月注射是否影響長期依從性。有效，不等於一定容易被真實世界採用。 這五層合在一起，才是 HORIZON 的完整答案。 ## 05｜為什麼這一讀出會影響整條 Lp(a) 賽道？ 因為後面不只 pelacarsen。 Amgen 的 olpasiran 正在 OCEAN(a)-Outcomes 第三期事件試驗中，以每 12 週皮下注射的方式，研究 Lp(a) 至少 200 nmol/L、已有特定動脈粥樣硬化性心血管疾病的患者。Lilly 的 lepodisiran 也已進入 ACCLAIM-Lp(a) 第三期事件試驗；同一家公司另有口服小分子 muvalaplin 的 MOVE-Lp(a) 第三期事件研究。 這裡不需要把每一個分子排成「誰降得最多」的排行榜。真正的產業問題是：第一個清楚的事件證據，會不會替整個標的建立可接受的臨床門檻？ 如果 HORIZON 顯示事件獲益，而且效果量、重要事件組成與安全性都站得住腳，後續競爭就會從「Lp(a) 能不能成為藥物標的」，轉向「哪種機制、劑型、給藥頻率、適用族群與價格最有優勢」。 如果結果中性，則要拆解是標的本身、介入時機、納入門檻、追蹤時間、降幅、背景治療，還是特定試驗設計出了問題。不能因為一個分子未達標就直接宣布整個 Lp(a) 生物學死亡，也不能拿其他分子的更高降幅假裝事件問題已經不存在。 競品試驗登錄： - Olpasiran OCEAN(a)-Outcomes：https://clinicaltrials.gov/study/NCT05581303 - Lepodisiran ACCLAIM-Lp(a)：https://clinicaltrials.gov/study/NCT06292013 - Muvalaplin MOVE-Lp(a)：https://clinicaltrials.gov/study/NCT07157774 ## 06｜台灣讀者最該注意的，不是「有沒有一檔概念股」 第一，看到結果時先確認檢驗單位與納入門檻。HORIZON 的門檻用 mg/dL 表示，試驗設計論文也提供約當 nmol/L，但 Lp(a) 的粒子大小有差異，不應把所有人的數值用單一固定比例草率互換。 第二，這是一個次級預防試驗。受試者已經有心肌梗塞、缺血性中風或有症狀周邊動脈疾病等病史。即使試驗成功，也不能直接外推到沒有心血管事件的一般健康族群，更不能自行據此決定檢測或用藥。 第三，台灣臨床落地還會經過核准標示、檢測可近性、專科路徑、給付與定價。全球試驗成功只是必要條件，不等於隔天就能在台灣開立。 第四，本文沒有建立足以支持「台股直接受惠」的一級證據，因此不硬列台股標的。Pelacarsen 的核心權利與開發關係在 Novartis 與 Ionis；競爭事件試驗主要來自 Amgen 與 Eli Lilly。把任何只有鬆散製造、檢測或題材關聯的台灣公司貼上 Lp(a) 概念股，反而會模糊真正的臨床問題。 對投資人來說，最值得看的也不是新聞稿第一行，而是這張檢查表： 1. 整體族群與 Lp(a) ≥90 mg/dL 子群，誰通過預設檢定？ 2. hazard ratio、95% 信賴區間與絕對事件差是多少？ 3. 心血管死亡、心肌梗塞、中風、緊急再血管化，誰在拉動複合終點？ 4. 實際 Lp(a) 降幅和事件獲益能否接得起來？ 5. 長期安全性、停藥率與每月注射的接受度如何？ 6. 公司所說的下一步，是完整申請、特定子群，還是仍需更多研究？ ## 最後的主判斷 HORIZON 的價值不在於再證明一次 pelacarsen 可以把 Lp(a) 壓低。這件事第二期已經給過明確訊號。 它要跨過的是更難的一道門：在標準治療之上，把一個遺傳性殘餘風險因子的下降，轉成病人真正少發生心肌梗塞、中風、心血管死亡與緊急再血管化。 截至 2026 年 7 月 19 日，這個答案仍未公布；Novartis 目前指向 2026 年下半年、事件驅動讀出。 等待結果時，最有價值的準備不是猜哪一天發新聞稿，而是先把判讀標準訂好。因為當數據真的來了，市場只需要幾秒鐘決定漲跌，但醫療與產業真正要回答的，是那個結果能不能把「降低一個數字」變成「改變一個人的心血管命運」。 --- 本文僅為公開臨床與產業資料整理，不構成醫療診斷、治療建議、募資或投資建議。個人檢測與治療應由專業醫療人員依病史與完整臨床資料判斷。 --- ## T-cell Exhaustion Across PD-1, CAR-T and TCE: Immunotherapy's Next Contest Is Durability URL: https://drugnews.com.tw/articles/2026-07-28-t-cell-exhaustion-immunotherapy-durability-en.html Date: 2026-07-28 Language: en Summary: PD-1 blockade, CAR-T and T-cell engagers ultimately depend on T cells that can expand, renew and persist. This analysis connects exhaustion biology, product engineering and clinical evidence to the next competitive threshold in immunotherapy. A tumor cell being killed makes for a striking image. The harder question is what remains on day 30, day 90, and beyond: how many T cells can still expand, find the tumor, and complete another round of killing? PD-1 antibodies, CAR-T cells, and T-cell engagers, or TCEs, take three different routes. All three ultimately rely on the functional state, differentiation quality, and persistence of T cells. 1. PD-1 blockade releases an inhibitory signal and reactivates T cells that still retain the capacity to respond. 2. CAR-T therapy removes T cells, engineers them with a tumor-recognition receptor, and returns them to the patient. 3. A TCE binds a tumor antigen at one end and CD3 at the other, drawing the patient's existing T cells into close proximity with the tumor. Drugnews' central judgment is that these modalities depend on the same underlying asset: a pool of T cells with sustained proliferative capacity. The next round of competition in immunotherapy will be decided by which product can preserve this "immune capital" rather than spending it all on the first day. This does not mean that every treatment failure should be attributed to exhaustion. Antigen loss, abnormal antigen presentation, poor tumor infiltration, inadequate drug exposure, and alternative escape pathways can all stop a response. The value of an exhaustion framework is that it changes how the data are read. Beyond asking how much the tumor shrank, investors and developers should ask which T-cell population performed the killing, whether that population can be replenished, and how long its function lasted. Similar early response rates do not imply similar durability. Distinguishing products requires response duration, cellular kinetics, and antigen status at relapse to be examined together. A single scan can show that a tumor became smaller; it cannot show which T-cell population sustained that effect. ## 01 | Exhaustion Is Not Simply Fatigue; It Is a Differentiation Trajectory When a tumor persists, antigen stimulation does not disappear after one immune response. T cells repeatedly encounter antigen while also facing hypoxia, nutrient deprivation, high lactate, inhibitory cytokines, regulatory T cells, and myeloid-derived suppressor cells. Their function changes progressively under these pressures. Common features include reduced proliferative capacity, altered secretion of IL-2, TNF, and IFN-gamma, and co-expression of inhibitory receptors such as PD-1, LAG-3, TIM-3, and TIGIT. The underlying biology includes TOX-associated exhaustion programs, epigenetic remodeling, and metabolic reprogramming. TCF1 is frequently used to identify a progenitor-like population that retains self-renewal capacity.[1][2] Exhaustion, however, is not one uniform endpoint and cannot be diagnosed from one marker. The same tumor can contain progenitor exhausted T cells that retain proliferative potential and more deeply exhausted, terminal populations. The former often express TCF1, can self-renew, and can generate downstream effector cells. The latter have less proliferative potential and more fixed functional and epigenetic states. That distinction matters. The clinical reality is not an army aging in synchrony. It is a mixed cellular reservoir containing a renewable reserve, cells being consumed, and cells that are becoming increasingly difficult to replenish. Exhaustion is not necessarily a biological error message. When antigen persists, reducing some effector functions can limit uncontrolled immune injury. The trade-off is that tumors can exploit this regulatory state and hold cytotoxic activity below the level required for disease clearance. This is why researchers should avoid describing cells only as "young" or "old." Antigen history, differentiation position, function, and plasticity need to be read together.[1] The sampling site changes the answer as well. Peripheral blood is easy to collect repeatedly and useful for tracking numbers and kinetics. Tumor tissue is required to determine whether cells actually enter the lesion, which cells they are near, and how intense local inhibitory signals are. A single pretreatment blood draw and one PD-1 measurement cannot characterize the full intratumoral T-cell reservoir. ## 02 | PD-1 Releases the Brake, but the Remaining Engine Still Matters A 2016 Nature study identified a TCF1-positive progenitor exhausted CD8 T-cell population in a chronic infection model. After PD-1 blockade, the major proliferative burst came from this self-renewing population, not from the restoration of every PD-1-positive T cell.[2] A Science study published the same year blocked the PD-1/PD-L1 axis in a chronic LCMV mouse model. It found that the epigenetic architecture of exhausted T cells was relatively stable. Treatment could improve some functions temporarily, but it did not fully reset deeply exhausted cells into conventional memory T cells. When antigenic pressure persisted, dysfunction could recur.[3] High PD-1 expression therefore cannot be translated directly into either "definitely exhausted" or "definitely sensitive to a PD-1 antibody." PD-1 is also expressed on activated T cells. What matters is its combination with TCF1, TIM-3, CD39, tumor specificity, spatial location, and functional state. This helps explain a clinical contrast in checkpoint therapy. Two patients may both have visible T cells and visible PD-1, yet possess very different reserves of cells that can still be mobilized. No single biomarker can resolve the whole problem. Response also depends on antigen presentation, tumor infiltration, the tumor microenvironment, tumor burden, and other resistance mechanisms. The clinically important question is where a response comes from. Tumor shrinkage after treatment may reflect reinvigoration of cells already inside the tumor, or the recruitment of new clones from the periphery. Subsequent progression can also arise through several routes: the proliferative reservoir may be depleted, the tumor may lose antigen, or the lesion may develop a new exclusionary environment. Calling every case of acquired resistance "exhaustion" risks selecting the wrong combination strategy. Good translational design therefore requires paired, longitudinal sampling. Baseline samples define the starting reservoir. On-treatment samples show whether particular clones expand. Samples at progression reveal whether the cells remain present and whether antigen has been retained. Only that timeline can distinguish a brake being reapplied from a change in the vehicle, the road, or the destination. ## 03 | CAR-T Exhaustion Risk May Be Engineered Before Infusion CAR-T therapy has a clear advantage: tumor-recognition capacity is installed directly into the T cell. Its limitation arises from the same intervention. Receptor architecture, signaling intensity, and manufacturing can determine how far the cell will travel before it ever enters the patient. A 2015 Nature Medicine study showed that some CARs aggregate and generate persistent, antigen-independent activity known as tonic signaling. T cells were repeatedly stimulated during ex vivo expansion, expressed more PD-1, LAG-3, and TIM-3, and subsequently showed impaired killing and in vivo persistence.[4] In the particular models used in that study, CD28 costimulation aggravated exhaustion caused by tonic signaling, while 4-1BB alleviated it. That finding cannot be simplified into a claim that every 4-1BB CAR is superior to every CD28 CAR. The antigen, antibody fragment, hinge, transmembrane domain, receptor density, and manufacturing process all influence the result. Each product still has to be validated on its own data. CAR-T development should therefore not be judged only by how quickly the first round of killing occurs. - Does the receptor aggregate without antigen? - How many stem-memory or central-memory T cells remain at the end of manufacturing? - Can the product expand and persist after infusion, and retain function when it encounters antigen again? These questions are directly connected to the long clinical tail. Strong short-term killing in vitro does not guarantee durable activity in vivo. The number of viable cells is only one component of release testing. Starting material, culture duration, cytokine conditions, and expansion pressure can all change the final differentiation composition. Memory phenotype, exhaustion markers, metabolic state, and polyfunctionality may be useful research characterizations, but not all are statutory release specifications. Clinical studies have associated characteristics of the preinfusion product with subsequent response, but those findings come from particular diseases and products and cannot be converted into universal pass-fail thresholds.[17] When a company claims that a shorter process or a more stem-like phenotype is an advantage, that claim still needs to be validated by batch consistency, in vivo expansion, and the durability of remission. Clinically, cellular dysfunction also has to be separated from antigen escape. Declining CAR-T numbers, declining cellular function, and loss of the target antigen can all cause relapse, but they require different development responses. The first two may point toward cell quality, signaling design, or the microenvironment. Antigen loss may require dual targeting, sequential therapy, or another antigen strategy. The word "relapse" alone does not reveal where the product failed. ## 04 | A T-cell Engager Creates Proximity, but Excessive Signaling Has a Cost A TCE does not require manufacturing a new batch of patient-specific cells. It uses the patient's existing T cells and brings them next to malignant cells through the CD3-binding arm. That enables rapid killing, but it also makes the patient's baseline T-cell condition part of the efficacy equation. The same molecule can behave differently after multiple prior therapies, in a patient with low lymphocyte counts or impaired cellular function, or in a tumor with few usable T cells. Tighter CD3 binding is not automatically better. A 2021 study spanning several tumor targets showed that tuning CD3 affinity simultaneously altered in vitro cytotoxicity, cytokine release, and in vivo biodistribution. This is a therapeutic-window problem, not a simple contest of signal strength.[5] Dosing cadence matters too. A 2022 Blood study observed declining T-cell function during continuous blinatumomab infusion in some patients with relapsed or refractory B-cell precursor acute lymphoblastic leukemia. The investigators then tested prolonged exposure using AMG 562 models. Continuous stimulation progressively induced exhaustion features, while treatment-free intervals restored portions of function, metabolism, and killing capacity.[6] This work provides valuable proof of concept, but it cannot be generalized to every TCE. Molecular format, half-life, target, disease, dose, and regimen differ. The more precise question is how each product balances tumor killing, cytokine-release syndrome, exposure time, and T-cell durability. Two timescales are often conflated. Cytokine-release syndrome is usually an early-treatment concern and can be managed through step-up dosing, pretreatment, and other strategies. Exhaustion concerns whether function can be maintained after repeated or sustained stimulation. Lower early cytokine release does not prove the absence of long-term exhaustion. A strong initial response does not prove that the usable T-cell reservoir will remain available. Convincing TCE data place drug concentration, cytokines, T-cell activation and differentiation, and tumor response on one timeline. A first response scan cannot show whether a molecule is mobilizing the immune system efficiently or exchanging greater intensity for a short-lived peak. Dosing intervals, half-life, and functional recovery should therefore be interpreted together. ## 05 | Exhaustion Management Is Becoming Part of Product Design "Reversing exhaustion" sounds like switching something off. The biology is better understood as layered management. The first layer is the starting cellular reservoir. Research programs can examine CD8 T-cell differentiation, TCF1, the co-expression of PD-1, LAG-3, TIM-3, and TIGIT, and cytokine secretion, proliferation, and cytotoxic capacity. These measurements have not become a universal routine assay across cancers. No single panel should be marketed as a mature companion diagnostic without disease-specific validation. The second layer is product engineering. CAR-T programs must manage tonic signaling, costimulatory architecture, receptor density, and manufacturing phenotype. TCE programs must manage CD3 affinity, molecular valency and format, half-life, dose, and schedule. The third layer is clinical context. Tumor burden, the microenvironment, lymphocyte damage caused by prior therapy, and the availability of tumor-specific T cells can all change efficacy. Debulking disease before T-cell-directed therapy may be biologically rational in some settings, but it must be determined by each disease and clinical study. It is not a universal treatment recommendation. The language of immunotherapy development will need to change accordingly. Saying that a product "activates T cells" is no longer enough. Developers should specify which population is activated, for how long, what phenotype remains, and which cells sustain the response. Even the most sophisticated flow cytometer provides only a snapshot if a program samples one time point. Exhaustion is a process. The data should at least define the starting state, the peak, and the decline: which clones expand, which phenotypes disappear, and whether function returns after treatment interruption or lower stimulation. Longitudinal data are more expensive, but they move a project from a plausible mechanism to evidence that the product actually behaves as intended. ## 06 | Taiwan Has More Than One Relevant Company: Four Direct Product Coordinates Focusing on Pell Biomedical alone would omit the clinical-stage TCE most directly related to this framework in Taiwan. A product-based view, rather than a loosely assembled "concept stock" list, yields at least four public-market coordinates. The first is SOA101 from Shine-On Biomedical (6926). It is a nanobody-based PD-L1 x HLA-G x CD3 trispecific T-cell engager. The formal title of ClinicalTrials.gov study NCT07055594 identifies it as a Phase I/IIa trial with Recruiting status in advanced or metastatic solid tumors.[10][11] It is the most direct Taiwan TCE in this analysis: two arms engage immune checkpoints while the CD3 arm recruits T cells. The next evidence should show whether dose escalation, cytokine release, exposure, antitumor activity, and T-cell function can coexist within an acceptable therapeutic window. The second is CAR001 from Ever Supreme (6712). NCT06150885 describes it as an allogeneic gamma-delta T-cell product combining an HLA-G CAR with a BiTE for relapsed or refractory solid tumors. On July 1, 2026, Ever Supreme reported that the program had passed its Phase I safety monitoring committee review, completed the first dose-escalation and safety stage, and formally initiated Phase IIa.[12][13] This product combines cellular engineering and an engager in one design. Repeat dosing, in vivo changes in CAR-positive gamma-delta T cells, and the balance between transient and sustained signaling deserve more attention than a generic CAR-T label. The third is PL001 from Pell Biomedical (6949). The company's current product page says that the program has entered Phase II, while ClinicalTrials.gov NCT05326243 lists a Phase I/II study with Recruiting status in relapsed or refractory B-cell lymphoma.[7][9] The company and Taiwan Stock Exchange materials emphasize a short manufacturing process and a higher proportion of stem-memory T cells, but these remain process claims.[8] The essential validation is whether release specifications and batch consistency, post-infusion expansion, cellular persistence, and the proposed phenotypes translate into durable remission. The fourth is AP203 from AP Biosciences (6945). AP203 is a PD-L1 x CD137 bispecific antibody without an anti-CD3 arm and is therefore not a CD3-directed TCE. According to the company product page and preclinical research, the design activates CD137 under PD-L1-binding and clustering conditions. Whether that conditional design reduces systemic activation risk in humans remains a clinical question. NCT05473156 is registered as a Recruiting Phase I/II study covering lung, head-and-neck, esophageal, and other solid tumors. The company page still describes a Taiwan Phase I program. Those sources reflect different fields and update points and should not be simplified into a claim that the company has already initiated Phase II.[14][15] ACE1831 from AcePodiA-KY (6976) is another direct T-cell product coordinate. It uses antibody-cell conjugation technology to attach a CD20 antibody to allogeneic gamma-delta T cells. It is neither CAR-T nor a CD3 TCE and is better treated as an adjacent comparison rather than merged mechanistically with the four products above.[16] None of these five companies should be described as having "proven reversal of T-cell exhaustion." They represent five different design questions: CD3 recruitment, a CAR/BiTE hybrid, CAR-T manufacturing, conditional costimulation, and antibody-armed allogeneic gamma-delta T cells. The clinical evidence still needs to establish whether effective cells remain, renew, and persist within an acceptable safety range. They should not be placed in one efficacy ranking. SOA101 first needs to show that its three binding functions create a manageable window at human exposure. CAR001 must separate the persistence of the infused gamma-delta T cells from the contribution of immune cells recruited by the BiTE. PL001 must show that its manufacturing phenotype translates into in vivo expansion and durable remission. AP203 must show how much conditional CD137 costimulation adds inside the tumor and how much systemic risk remains. These evidence lines will mature on different timelines and cannot be interpreted with one early-news yardstick. ## 07 | Four Data Sets Investors Should Watch Next Immunotherapy narratives are easily dominated by early response rates. Once exhaustion is connected to product value, at least four additional groups of data become essential. - **Starting phenotype:** What proportions of the cell population are progenitor-like, stem-memory, or terminally differentiated? - **Expansion and persistence:** What is the post-infusion or post-dose expansion peak, how long do cells remain detectable, and what function remains after re-encountering antigen? - **Therapeutic window:** How do cytokine-release syndrome, neurotoxicity, dose interruption, and dosing cadence relate to stable long-term exposure? - **Depth and durability of remission:** Does short-term killing translate into durable remission, progression-free survival, or overall survival rather than one attractive early time point? For the Taiwan programs, SOA101 should be read through dose, cytokine, and efficacy data that mature together. CAR001 should be read through repeat dosing and changes in CAR-positive gamma-delta T cells. PL001 should be read through batch characteristics, expansion, and persistence. AP203 should be read through whether conditional costimulation creates an adequate therapeutic window. Those variables reveal more about product value than whether immunotherapy is a fashionable market theme. Percentages in early trials always need a time dimension. Dose-escalation cohorts are small, follow-up differs across patients, and one response rate can combine newly treated patients with people followed for months. For products that depend strongly on T cells, when responses appear, how long they last, whether cells remain detectable, and under what toxicity or interruption conditions patients complete treatment can be more informative than one cutoff. When a company updates its data, find the median follow-up and evaluable population before reading the response percentage. A useful sequence is to ask whether the mechanism operated as expected, whether exposure, batch characteristics, cellular kinetics, and toxicity can be reproduced, and only then whether response depth and durability justify a valuation change. Without the intermediate product evidence, several responses do not prove that the durability problem has been solved. Consider two hypothetical products. One produces a high expansion peak and substantial early toxicity, followed by rapid cellular disappearance. The other has a lower peak but more persistent cells. A single peak cannot establish which is better; the answer depends on disease, dose, tumor burden, and remission durability. That is the analytical value of an exhaustion framework: it forces the market to replace "the drug produced a response" with "which cells produced the response, and how long did they remain functional?" ## Conclusion | The Scarce Asset Is a Renewable T-cell Force PD-1 releases the brake. CAR-T changes the steering system. A TCE forces two lanes to merge. All three still depend on the same underlying asset: T cells that retain sustained proliferative capacity and can execute cytotoxic activity over time. Cancer is a long war. Killing rapidly in the first round does not solve the later problem of an unreplenished cellular reservoir, declining function, and continuing microenvironmental pressure. The next generation of immunotherapy competition will increasingly move from "how to make the signal stronger" toward "how to keep effective cells present, renewable, and durable." That is the product threshold rewritten by T-cell exhaustion. ## Disclaimer This article is an analysis of biopharmaceutical science and industry trends. It does not constitute medical diagnosis, treatment advice, investment advice, or a recommendation to buy or sell securities. Treatment decisions should be made by qualified healthcare professionals based on each patient's condition and the full clinical evidence. --- ## T 細胞耗竭卡住 PD-1、CAR-T 與 TCE：免疫治療下半場，拼的是誰能撐到最後 URL: https://drugnews.com.tw/articles/2026-07-28-t-cell-exhaustion-immunotherapy-durability.html Date: 2026-07-28 Language: zh-Hant Summary: PD-1、CAR-T 與 TCE 最後都仰賴可增殖、可補充且能持久的 T 細胞戰力。本文從耗竭分化、產品工程與臨床資料，拆解免疫治療下半場的真正競爭門檻。 一顆腫瘤細胞被殺掉，畫面很有衝擊力。 真正困難的是第 30 天、第 90 天，甚至更久以後，還剩多少 T 細胞能繼續擴增、找到腫瘤，再完成下一輪殺傷。 PD-1 抗體、CAR-T 與 T-cell engager（TCE，T 細胞銜接器）走的是三條路，最後都得靠 T 細胞的功能狀態、分化品質與持久性完成殺傷。 1️⃣ PD-1 阻斷解除抑制訊號，讓仍有反應能力的 T 細胞重新動起來。 2️⃣ CAR-T 把 T 細胞取出改造，裝上辨識腫瘤的受體，再送回體內作戰。 3️⃣ TCE 一端抓腫瘤抗原、一端抓 CD3，把患者現有的 T 細胞拉到腫瘤旁邊。 藥時事團隊的判斷是：這三類療法共同依賴同一種底層資產——具有持續增殖能力的 T 細胞戰力。免疫治療下一輪的產品差距，會落在誰能把這筆「免疫資本」用得久，而不是第一天就燒完。 這個判斷不等於把所有治療失敗都歸因於耗竭。抗原消失、抗原呈現異常、T 細胞進不了腫瘤、藥物暴露不足，以及腫瘤以其他路徑逃脫，都可能讓療效停下來。耗竭真正改變的是我們讀數據的方式：除了問腫瘤縮了多少，還要追問是哪一群 T 細胞完成殺傷、這群細胞能不能補充，以及功能維持了多久。 早期反應率相近，不代表產品續航相同。要分辨差異，得把反應維持時間、細胞動力學與復發時的抗原狀態放在一起看；一次掃描只看得到腫瘤縮小，看不到是哪一批 T 細胞把效果撐住。 ## 01｜耗竭不是「累了」而已，而是一條分化路徑 腫瘤長期存在，抗原刺激不會在一次免疫反應後消失。T 細胞反覆接觸抗原，同時承受缺氧、養分不足、高乳酸、抑制性細胞激素、Treg 與髓源抑制細胞等壓力，功能會逐步改變。 常見現象包括增殖能力下降，IL-2、TNF、IFN-γ 等細胞激素分泌受影響，並共同表現 PD-1、LAG-3、TIM-3、TIGIT 等抑制性受體。背後牽涉 TOX 相關的耗竭程式、表觀遺傳與代謝重塑；TCF1 則常用來辨識仍保有自我更新能力的前驅樣族群。[1][2] 但「耗竭」不是一個整齊的終點，也不能只靠一個標記下判斷。 同一個腫瘤裡，可以同時存在保有增殖能力的前驅耗竭 T 細胞，也有更深、更終末的耗竭族群。前者常見 TCF1 表現，仍能自我更新並產生後續效應細胞；後者的增殖潛力較低，功能與表觀遺傳狀態也更固定。 這個差別很重要。臨床上看到的不是一支整齊老化的軍隊，而是一個仍有後備、正在消耗、以及幾乎難以補充的混合細胞庫。 耗竭本身也不是生物學的「故障訊息」。在抗原長期存在時，降低部分效應功能可以減少失控的免疫傷害；代價是腫瘤可能利用這套調節，讓殺傷反應停在不足以清除疾病的程度。這也是為什麼研究者不宜只用「年輕／老化」描述細胞，而要把抗原歷史、分化位置、功能與可塑性放在一起看。[1] 採樣位置同樣會改變答案。外周血容易重複取得，適合追蹤數量與動態；腫瘤組織才能告訴我們細胞是否真正進入病灶、靠近哪些細胞，以及局部抑制訊號有多強。若只在治療前抽一次血、只量一個 PD-1，就很難替整個腫瘤內的 T 細胞庫下結論。 ## 02｜PD-1 鬆開煞車，還要看引擎剩多少力 2016 年的 Nature 研究在慢性感染模型中辨識出一群 TCF1 陽性的前驅耗竭 CD8 T 細胞。PD-1 阻斷後，主要的增殖爆發來自這群仍有自我更新能力的細胞，而非所有 PD-1 陽性 T 細胞一起恢復。[2] 同年的 Science 研究在慢性 LCMV 小鼠模型中阻斷 PD-1／PD-L1 軸，結果指出耗竭 T 細胞的表觀遺傳圖譜相對穩定。治療可以短期改善部分功能，卻不會把深度耗竭的細胞完整重置成典型記憶 T 細胞；抗原壓力持續存在時，也可能再次走向功能失調。[3] 所以，PD-1 高表現不能直接翻譯成「一定耗竭」，更不能當成「一定對 PD-1 抗體敏感」。PD-1 也會出現在活化中的 T 細胞上，真正有意義的是它和 TCF1、TIM-3、CD39、腫瘤特異性、空間位置與功能狀態的組合。 這也解釋了免疫檢查點治療的一個臨床反差：同樣看見 T 細胞、同樣看見 PD-1，患者能被動員的後備細胞庫可能完全不同。 單一生物標記很難把這件事講完。療效還會被抗原呈現、T 細胞能不能進入腫瘤、腫瘤微環境、腫瘤負荷與其他抗藥機制共同決定。 把這段機制帶回臨床，最值得注意的是「反應來自哪裡」。治療後腫瘤縮小，可能代表原有腫瘤內 T 細胞被重新動員，也可能有新的克隆從周邊進入病灶。若後來再次進展，原因也不只一種：可增殖的細胞庫可能被耗盡，腫瘤可能失去抗原，或病灶形成新的排斥環境。把所有續發性抗藥都叫作耗竭，反而會讓下一個聯合療法選錯靶點。 因此，好的轉譯設計需要前後配對。治療前知道起始細胞庫，治療中看到克隆是否擴增，進展時再判斷細胞還在不在、抗原是否保留。只有這條時間軸，才能分辨「煞車重新踩下去」與「車、路或目的地已經變了」。 ## 03｜CAR-T 的耗竭風險，可能在輸注前就被設計進去了 CAR-T 的優勢，是把辨識能力直接裝進 T 細胞。它的限制也來自同一件事：受體結構、訊號強度與製程，可能在細胞進入患者體內前，先決定它能跑多遠。 2015 年 Nature Medicine 的研究顯示，部分 CAR 會因受體聚集產生不依賴抗原的持續訊號，也就是 tonic signaling。細胞在體外擴增階段就被反覆刺激，PD-1、LAG-3、TIM-3 等標記上升，後續殺傷與體內持久性也受到影響。[4] 該研究的特定模型中，CD28 共刺激加重持續訊號造成的耗竭，4-1BB 則有所緩解。但這不能簡化成「所有 4-1BB CAR 都優於 CD28」。抗原、抗體片段、鉸鏈區（hinge）、跨膜區、受體密度與製程都會改變結果，產品仍要逐一驗證。 CAR-T 開發因此不能只比較第一輪殺得多快。 🔧 受體會不會在沒有抗原時自行聚集？ 🧬 製程最後留下多少幹記憶型或中央記憶型 T 細胞？ 📈 輸注後能否擴增、持續存在，並在再次遇到抗原時保留功能？ 這三題直接連到臨床長尾。體外做出很強的短期殺傷，不代表體內能維持同樣的戰力。 活細胞數量只是放行的一部分。起始原料、培養天數、細胞激素條件與擴增壓力，都可能改變最終分化組成；記憶表型、耗竭標記、代謝狀態與多功能性可以作為研究性產品表徵，但未必都是法定放行規格。已有臨床研究觀察到輸注前細胞產品的表型與功能特徵和後續反應相關，但結果來自特定疾病與產品，仍不能直接跨產品設定及格線。[17] 若公司主張短製程或較高幹性帶來優勢，還是要用批次一致性、體內擴增與緩解維持驗證。 臨床上還要把細胞耗竭和抗原逃脫分開。CAR-T 數量下降、功能下降，與腫瘤失去標的抗原，都會造成復發，研發對策卻完全不同。前者可能指向細胞品質、訊號設計或微環境；後者則可能需要雙標的、序列治療或其他抗原策略。只看「復發」兩個字，判斷不了產品真正卡在哪裡。 ## 04｜TCE 把細胞拉到一起，訊號太猛也會付出代價 TCE 不需要先替患者製造一批新細胞。它借用患者體內既有的 T 細胞，透過 CD3 端把它們拉到腫瘤細胞旁邊。 這讓藥物可以快速啟動殺傷，也把患者原本的 T 細胞狀態帶進療效方程式。經過多線治療、淋巴球數量偏低、功能較差或腫瘤內缺乏可用 T 細胞時，同一個分子可能得到不同結果。 CD3 端也不是抓得越緊越好。2021 年一項跨多個腫瘤標的的雙特異性抗體研究顯示，調整 CD3 親和力會同時改變體外殺傷、細胞激素釋放與體內分布。這是一個治療窗問題，而非單純的強弱競賽。[5] 給藥節奏同樣重要。Blood 在 2022 年發表的研究觀察到，接受 blinatumomab 連續輸注的部分復發／難治性 B 細胞前驅急性淋巴性白血病患者，其 T 細胞功能在治療期間下降；研究團隊再以 AMG 562 的長期暴露模型測試，持續刺激會逐步出現耗竭特徵，加入停藥間隔後，部分功能、代謝與殺傷能力得到恢復。[6] 這項結果提供很好的概念驗證，卻不能直接外推到所有 TCE。分子格式、半衰期、標的、疾病、劑量與療程都不同。更精確的問題應該是：每一款 TCE 如何在殺傷、CRS、暴露時間與 T 細胞續航之間找到自己的區間？ 這裡很容易把兩種時間尺度混在一起。CRS 多半在治療早期成為焦點，開發上會使用階梯式給藥、前處置或其他方式管理；耗竭討論的是反覆或持續刺激之後，細胞能否維持功能。早期細胞激素較低，不足以證明長期不耗竭；初期反應強，也不保證可用的 T 細胞庫能一路撐到後面。 一款 TCE 真正有說服力的資料，會把藥物濃度、細胞激素、T 細胞活化與分化、腫瘤反應放在同一條時間線。若只有第一個療效掃描，還看不出分子是在有效調動免疫系統，或只是用較高強度換到短暫峰值。這也是給藥間隔、半衰期與可恢復性值得一起追的原因。 ## 05｜耗竭管理，會變成產品設計的一部分 「逆轉耗竭」聽起來像把一個開關關掉，真實生物學更接近分層管理。 第一層是起始細胞庫。研究可以評估 CD8 T 細胞分化組成、TCF1、PD-1、LAG-3、TIM-3、TIGIT 的共同表現，以及細胞激素、增殖與殺傷能力。這些指標尚未形成跨癌別通用的常規檢測，不能把任一組合包裝成成熟伴隨式診斷。 第二層是產品工程。CAR-T 要管理 tonic signaling、共刺激結構、受體密度與製程表型；TCE 要管理 CD3 親和力、分子價態、半衰期、劑量與給藥節奏。 第三層是臨床情境。腫瘤負荷、微環境、前線治療造成的淋巴球影響，以及是否存在足夠的腫瘤特異性 T 細胞，都會改變療效。先降低腫瘤負荷再導入 T 細胞療法，在部分疾病可能有生物學合理性，但仍要由各癌別與臨床試驗決定，不能當成普遍治療建議。 免疫治療的研發語言也會跟著改變。未來只說「活化 T 細胞」已經不夠，還要交代活化哪一群、刺激多久、留下什麼表型，以及療效靠哪一批細胞維持。 如果一個專案只在單一時間點收樣，再精密的流式細胞儀也只能提供快照。耗竭是一段過程，資料至少要回答起點、峰值與回落：哪些克隆擴增、哪些表型消失、停藥或降低刺激後能否恢復。這種縱向資料成本更高，卻能把「機制看起來合理」推進到「產品確實照預期運作」。 ## 06｜台灣不是只有一家公司：四個直接產品座標 若只寫沛爾生醫，會漏掉台灣最貼近本文的臨床期 TCE。依產品、技術與臨床角色，而不是用「概念股」硬湊，至少有四個公開市場座標。 第一個是聖安生醫（6926）的 SOA101。它是以奈米抗體打造的 PD-L1 × HLA-G × CD3 三特異性 T-cell engager，ClinicalTrials.gov NCT07055594 的正式試驗名稱標示為 Phase I/IIa，狀態為 Recruiting，適應症是晚期或轉移性實體瘤。[10][11] 這是本文最直接的台灣 TCE：它一邊抓雙免疫檢查點，一邊透過 CD3 銜接 T 細胞。接下來要看劑量遞增、細胞激素釋放、暴露、療效與 T 細胞功能能否一起落在可接受的治療窗。 第二個是長聖（6712）的 CAR001。NCT06150885 將它登錄為結合 HLA-G CAR 與 BiTE 的異體 γδT 細胞產品，治療復發或難治性實體瘤；長聖在 2026 年 7 月 1 日公告，已通過第一期安全監測委員會（SMC）審查、完成第一階段劑量遞增與安全性評估，並正式啟動 Phase IIa。[12][13] 這個產品把細胞工程與 engager 放在同一套設計裡，因此重複給藥、CAR 陽性 γδT 細胞的體內變化、短暫與持續訊號之間的平衡，都比一句「CAR-T 題材」更值得追。 第三個是沛爾生醫（6949）的 PL001。公司現行產品頁稱已進入第二期，ClinicalTrials.gov NCT05326243 則登錄為第一／二期、Recruiting，適應症為復發或難治性 B 細胞淋巴癌。[7][9] 公司與臺灣證券交易所把短製程、較高比例幹記憶型 T 細胞列為技術重點，但這仍是製程主張。[8] 真正要驗證的是放行規格與批次一致性、輸注後擴增、細胞持續時間，以及這些表型能否轉成持久緩解。 第四個是圓祥生技（6945）的 AP203。它是 PD-L1 × CD137 雙特異性抗體，不含 anti-CD3 作用端，因此不屬於 CD3 型 TCE。依公司產品頁與臨床前研究，其設計是在 PD-L1 結合與聚集條件下活化 CD137；能否在人體降低全身性活化風險，仍待臨床資料驗證。NCT05473156 登錄為第一／二期、Recruiting，涵蓋肺癌、頭頸癌與食道癌等實體瘤；公司產品頁仍標示台灣第一期進行中，兩個資料來源反映的欄位與更新節點不同，不能直接寫成公司已啟動第二期。[14][15] 育世博-KY（6976）的 ACE1831 也是直接的 T 細胞產品座標，但它用抗體細胞連結技術把 CD20 抗體接到異體 γδ T 細胞，並不是 CAR-T 或 CD3 TCE，適合列為延伸觀察，不和前四家公司混成同一機制。[16] 這五家公司都不能被寫成「已證明逆轉 T 細胞耗竭」。它們代表的是五種不同設計題：CD3 銜接、CAR／BiTE 混合、CAR-T 製程、條件式共刺激，以及抗體武裝的異體 γδT。臨床資料最後要回答的，仍是有效細胞能不能留下、補上，並在安全範圍內撐得久。 它們也不適合被塞進同一張療效排名表。SOA101 要先證明三個作用端在人體暴露下形成可管理的治療窗；CAR001 要分清輸注的 γδT 細胞自身持久性，以及 BiTE 招募其他免疫細胞各自貢獻多少；PL001 的核心證據是製程表型能否轉成體內擴增與緩解維持；AP203 則要看條件式 CD137 共刺激在腫瘤內提供多少增益，又留下多少全身性風險。這四條資料線成熟的時間不會一致，早期新聞也不能用同一把尺解讀。 ## 07｜投資人接下來要盯的四組資料 免疫治療文章很容易被早期反應率帶走。若把耗竭放回產品價值，至少還要追四組指標。 📍 起始細胞表型：前驅樣、幹記憶型與終末分化細胞各占多少？ 📈 擴增與持久性：輸注或給藥後的擴增峰值、持續時間，以及再次遇到抗原時的功能。 ⚠️ 治療窗：CRS、神經毒性、劑量中斷與給藥節奏，是否換來更穩定的長期暴露。 ⏳ 緩解深度與維持：短期殺傷能不能轉成持久緩解、PFS 或 OS，而非只在第一個時間點漂亮。 放回台灣公司，SOA101 要看劑量、細胞激素與療效是否同步成形；CAR001 要看重複給藥與 CAR 陽性 γδT 細胞的變化；PL001 要看批次、擴增與持久性；AP203 要看條件式共刺激能否換來足夠的治療窗。這些資料比「免疫治療題材熱不熱」更能回答產品價值。 讀早期臨床時，百分比一定要和時間一起看。劑量遞增階段人數少、追蹤長短不同，同一個反應率可能混合剛開始治療與已追蹤多月的患者。對高度依賴 T 細胞的產品，反應出現多久、維持多久、細胞是否仍可測得，以及患者在什麼毒性與停藥條件下完成治療，往往比單一截點更有資訊。當公司更新資料，先找中位追蹤時間與可評估人數，再看反應率，較不容易被漂亮但尚未成熟的分子帶走。 公司更新可先看機制是否照預期運作，再查暴露、批次、細胞動力學與毒性能否重複，最後才判斷緩解深度與維持；少了中間的產品證據，幾個反應個案還不能證明續航問題已解決。 例如，一項治療同時出現高擴增峰值與明顯早期毒性，後續細胞迅速消失；另一項峰值較低，細胞卻維持更久。兩者誰更好，不能靠單一峰值回答，還得放回疾病、劑量、腫瘤負荷與緩解維持。耗竭視角的價值就在這裡：它迫使市場把「很有反應」改問成「反應由誰完成，又維持了多久」。 ## 結語｜免疫治療下半場，真正稀缺的是能補上的兵 PD-1 鬆開的是煞車，CAR-T 換的是方向盤，TCE 做的是強制併線。三條路最後都依賴同一種底層資產：具有持續增殖能力、能長期執行殺傷的 T 細胞戰力。 腫瘤是長期戰。第一輪殺得很快，解決不了後續細胞補不上、功能掉下來、微環境持續施壓的問題。 下一代免疫治療的競爭，會逐漸從「如何把訊號做強」轉向「如何讓有效細胞留得住、補得上、撐得久」。這才是耗竭問題真正改寫的產品門檻。 ## 免責聲明 本文僅為生技醫藥與產業趨勢整理，不構成醫療診斷、治療建議、投資建議或買賣建議。實際治療選擇應由專業醫療團隊依個別病況與臨床證據判斷。 --- ## 半年千億，買的是時間 URL: https://drugnews.com.tw/articles/2026-07-27-h1-2026-biopharma-ma-buying-time.html Date: 2026-07-27 Language: zh-Hant Summary: 2026 上半年生醫併購轉熱，但 BioSpace 的 52 筆與 IQVIA 的 42 筆不能硬拼。真正的主線，是專利懸崖迫使大藥廠用現金購買可計算的臨床時間。 2026 年上半年，生技併購真的回來了。 但先別急著把「超過 50 筆、接近 1,300 億美元」排在同一句，做成一個看起來很完整的大數字。這兩個數字來自不同資料庫，篩選門檻也不同。 BioSpace 依 S&P Capital IQ 統計到 52 筆交易，納入的是至少 1,000 萬美元、收購完整治療藥物公司的交易，交易價值還可能包含債務與其他調整；IQVIA 的口徑則是取得生技製藥標的多數股權、交易值至少 2.5 億美元，因此得到 42 筆、合計約 1,300 億美元。 兩個數字都可以用，不能把它們拼成同一份統計。 而且 IQVIA 的原文是：2026 上半年約 1,300 億美元，幾乎追平 2025 全年的 1,330 億美元，不是「已經超過去年全年」。 這個小小的口徑差，正好提醒我們：併購熱度固然回來了，真正值得追的卻不是一個越喊越大的總額，而是藥廠正在買什麼。 藥時事的主判斷是： 2026 年的生醫併購不是單純從冷轉熱，而是專利懸崖把大藥廠的選擇題改了。買家不再優先追求「公司越大越好」，而是用更高價格搶臨床後期、接近讀出或接近上市的資產，買下一段可以被計算的時間。 ## 52 筆和 42 筆，差的不是 10 筆而已 BioSpace 的 52 筆，是一張「完整公司收購」清單。它告訴我們，2025 年同期約 30 筆，2026 年確實明顯變多；其中禮來以 9 筆居首，承諾金額超過 250 億美元。原始文章寫成 11 筆，並不符合這份統計。 IQVIA 的 42 筆則設了更高的 2.5 億美元門檻。這 42 筆平均交易規模約 31 億美元，高於 2025 年的 27 億美元；上半年共有 9 筆交易達 50 億美元以上，卻沒有一筆標的價值高於 300 億美元的「超大型合併」。 換句話說，市場不是靠一樁 Celgene 等級的巨獸交易撐起來，而是出現一群 50 億到 110 億美元左右、科學與時程都相對具體的標的。 PwC 從另一個角度看到同一件事。2026 年第一季生技製藥交易額已超過 650 億美元，宣布的 10 億美元級交易有 16 筆；其年中報告直接把這一波描述為「更少由規模、更多由精準科學驅動」。買家偏好中型、可嵌入現有產品線的收購，也更常把 CVR、里程碑或其他條件式付款塞進交易結構。 所以別只問「併購潮回來了嗎？」更精準的問題是：為什麼大藥廠突然願意為這些資產加價？ ## 藥廠買的不是故事，是專利時鐘 答案先從一個數字開始：PwC 估計，本十年有超過 3,000 億美元的品牌藥收入暴露在市場獨占權到期風險之下。 這不是 3,000 億美元會在某一天同時消失，也不是每一塊錢都一定流失；它代表的是，大型藥廠必須在舊產品失速前，找到足夠大的新曲線補上。 早期平台當然可能很值錢，可是從動物實驗、一期、二期一路走到關鍵試驗，時間、失敗率與追加資本都很難算。當專利倒數只剩幾年，買家會更在意三件事：臨床證據走到哪裡、下一個讀出還要多久、拿進來之後能否立刻接上自己的開發與商業系統。 這就是「買時間」的意思。 默沙東 3 月宣布以約 67 億美元股權價值收購 Terns Pharmaceuticals，並在 5 月 5 日完成交易。核心資產 TERN-701 是治療慢性骨髓性白血病的口服變構 BCR::ABL1 抑制劑；它仍在一期／二期 CARDINAL 試驗，不是上市藥。因此，公司收購雖已完成，卻不能寫成已經買到一段確定收入。 艾伯維 6 月宣布以約 109 億美元股權價值收購 Apogee Therapeutics，取得以異位性皮膚炎與氣喘為主的多項免疫管線。帶頭的 zumilokibart（APG777）是一款延長半衰期的 IL-13 單株抗體。這筆交易預計第三季才完成，仍要經股東與監管條件；「宣布收購」不是「已完成收購」。 兩筆交易的共同點很清楚：不是為了合併兩間巨型藥廠的後台，而是為了把一個有臨床輪廓的資產，塞進既有治療領域與商業機器。 ## 禮來掃貨，但不是原文寫的 11 筆 2026 上半年最積極的買家仍是禮來。 依 IQVIA 與 BioSpace 的完整公司收購口徑，禮來是 9 筆，不是 11 筆；IQVIA 計算投入約 260 億美元，BioSpace 的表述則是承諾超過 250 億美元。兩者接近，但「投入」「承諾」與最終實付仍不是完全相同的概念。 禮來的特色不是只守著肥胖與糖尿病，而是用 GLP-1 帶來的現金能力向外擴張：免疫、神經、體內細胞治療、疫苗等不同領域都出手。IQVIA 也把 5 月一次收購三家疫苗公司視為代表性案例，三筆合計 38 億美元。 這不代表禮來已經把每一個新領域都做成第二支柱。頻繁買進只是資產配置，真正的驗收要看收購後的試驗設計、優先順序、整合速度與淘汰紀律。藥廠有錢買十個項目，不等於十個都應該進三期。 原始文章還列入幾筆無法由禮來一級公告直接核對的公司與金額；這些名稱本文不保留。併購盤點最怕把收購、授權、投資與合作混在同一張「戰利品清單」，數量會很好看，判斷卻會失真。 ## 吉利德與 GSK，押的是接近商業化的腫瘤資產 吉利德的 Arcellx 交易最能看出時間的價值。 吉利德在 2026 年 4 月 28 日完成收購 Arcellx，現金加一項每股 5 美元的或有價值權，完成時隱含股權價值約 78 億美元。標的是仍在開發中的 BCMA CAR-T anito-cel；交易完成讓吉利德取得完整控制權，也消除原合作中的分潤、里程碑與權利金義務。 注意這裡可以用「已完成」，因為公司公告明確寫了 successful completion；但 anito-cel 仍待監管核准。完成公司交易和產品成功，不是同一件事。 吉利德也在 5 月 21 日完成 Tubulis 收購。條件是 31.5 億美元前期現金，加上最高 18.5 億美元里程碑；新聞只寫「50 億美元收購」會把確定支付與未來達標才支付的金額混在一起。交易已完成，主力資產 TUB-040 卻仍在 I／IIa 期試驗，產品與里程碑風險都沒有消失。 GSK 6 月宣布以 106 億美元總股權價值收購 Nuvalent，並在 7 月 15 日完成交易，取得兩項後期肺癌標靶藥 zidesamtinib 與 neladalkib，以及一期的 HER2 抑制劑 NVL-330。前兩項資產仍在 FDA 審查，目標決定日分別是 2026 年 9 月 18 日與 11 月 27 日。 因此，Nuvalent 能寫成「公司收購已完成、買進接近上市的資產」，不能寫成「產品已上市、2027 年一定貢獻獲利」。GSK 自己也把未來推出產品綁在 FDA 核准前提上。 這三個案例剛好排出一條風險光譜： - Arcellx：公司收購已完成，但產品仍待監管核准； - Tubulis：公司收購已完成，但 TUB-040 仍在早期臨床，部分價金也綁里程碑； - Nuvalent：公司收購已完成，但兩項後期產品仍待 FDA 決定。 把「公司已收購」直接翻成「產品即將貢獻收入」，會把真正的投資風險擦掉。 ## 最大一筆，也不能只看 headline 數字 2026 上半年最大型交易之一，是印度 Sun Pharma 與 Organon。 公司正式公告的交易企業價值是 117.5 億美元，每股現金 14 美元，並預計在監管與股東條件完成後，於 2027 年初左右交割。BioSpace 依 S&P Capital IQ、包含債務與其他因素的口徑記為 126 億美元；兩個數字差異來自統計方法，不應把 126 億美元說成公司公告的現金收購價。 這筆交易也和前面的管線型交易不同。Organon 有女性健康、成熟藥與生物相似藥的現成產品與全球商業網路；Sun Pharma 買的比較像是產品組合、地理通路與現金流平台，而不是單押一個臨床事件。 這提醒我們，2026 的主流雖然偏向精準、後期資產，市場仍不是只有一種模板。只要能回答「買進之後補哪個缺口」，成熟產品組合也可能是合理標的。 ## 併購熱，為什麼不等於 Biotech 寒冬結束？ 對賣方而言，併購變多當然是好消息。它提供 IPO 以外的退出路徑，也讓有資料、有現金壓力的公司多一個選項。 但大藥廠開始掃貨，不代表每一家 Biotech 的資金成本都回到低點，更不代表只要有平台故事就會被買。 PwC 觀察到的是更「外科手術式」的交易：資產要能填專利缺口、切進高成長領域，最好有近期臨床或商業事件；買家也會用 CVR、里程碑把部分失敗風險推回賣方。這其實比全面牛市更挑剔。 接下來可能讓熱度降溫的因素也很具體：美國藥價政策、最惠國藥價構想、IRA 議價擴張、關稅、跨境審查與美中政策不確定性，都可能改變估值與付款條件。另一方面，如果熱門資產價格被少數大藥廠搶高，董事會也可能選擇等待資料，而不是在高點追價。 所以「下半年會不會繼續熱」不能只看上半年總額外推。更可靠的追蹤方式是： 1. 十億美元級交易的數量是否延續，而非只看一筆巨額交易； 2. 前期現金占總交易價值的比例，有沒有被里程碑愈拉愈低； 3. 買家是否持續偏好三期、申請前或 FDA 審查中的資產； 4. 已宣布交易能否如期完成，而不是停在簽約新聞； 5. 收購後一年內，資產究竟被加速、延後，還是直接砍掉。 ## 台灣讀者該看什麼？ 這一波沒有足夠一級證據，能把某一家台灣上市櫃公司直接標成上述交易的產品、製造、檢測或商業化受益者，因此本文不硬湊「併購概念股」。 真正值得台灣 Biotech 留意的，是買方偏好已變得更具體：臨床資料要能回答差異化、關鍵試驗路徑要清楚、全球權利要整理乾淨、CMC 與供應要能接上大型藥廠。平台可以有想像力，交易文件卻必須能被盡職調查拆開。 對投資人而言，看到「最高交易價值」時，先拆成現金、股權價值、企業價值、CVR 與里程碑；看到「完成交易」時，再回公司公告確認究竟是 announced、signed 還是 completed。這兩個動作，會過濾掉不少看似驚人的數字。 ## 結語：半年千億，買的是時間 2026 上半年生醫併購回暖，這件事沒有問題。BioSpace 的 52 筆與 IQVIA 的 42 筆，都指向交易明顯轉熱；PwC 的第一季 650 億美元與 16 筆十億美元級交易，也提供另一個旁證。 有問題的是，把不同口徑拼成一個完美的大標題，再用總額直接宣告資本寒冬結束。 這一輪買家真正搶的，是能在專利懸崖前趕上的臨床時間。越接近讀出、申請或上市，價格越容易被抬高；風險還沒消失，只是被重新分配到前期現金、CVR、里程碑與交割條件裡。 大藥廠不是突然變得浪漫。它們只是看著專利時鐘，開始用現金買時間。 ## 免責聲明 本文為生技醫藥與產業趨勢整理，不構成醫療診斷、治療建議或投資建議。交易金額可能因股權價值、企業價值、債務、現金、CVR 與里程碑口徑而不同，投資判斷應回到公司公告與監管文件。 --- ## $130 Billion in Six Months: Big Pharma Is Buying Time URL: https://drugnews.com.tw/articles/2026-07-27-h1-2026-biopharma-ma-buying-time-en.html Date: 2026-07-27 Language: en Summary: Biopharma M&A accelerated in the first half of 2026, but BioSpace's 52 deals and IQVIA's 42 deals cannot be combined. The deeper story is that patent cliffs are pushing Big Pharma to pay for measurable clinical time. Biopharma M&A is back. But before placing “more than 50 deals” and “nearly $130 billion” in the same sentence and presenting them as one neat market statistic, there is an important methodological problem: those figures come from different databases with different inclusion rules. BioSpace, using S&P Capital IQ, counted 52 transactions. Its screen includes acquisitions of whole drug-development companies valued at no less than $10 million, and reported deal value can include debt and other adjustments. IQVIA applied a different threshold: a buyer had to obtain majority ownership of a biopharma target, and the transaction had to be worth at least $250 million. That screen produced 42 transactions with an aggregate value of approximately $130 billion. Both figures are useful. They are not one dataset, and they cannot be added together. IQVIA's original wording also matters. The approximately $130 billion announced in the first half of 2026 nearly matched the $133 billion recorded for all of 2025. It did not already exceed the previous full year. That small methodological distinction points to the larger issue. The M&A market has clearly accelerated, but the most important question is not how large a headline number can become. It is what pharmaceutical companies are buying. Drugnews' central judgment is this: Biopharma M&A in 2026 is not simply moving from cold to hot. The patent cliff has changed the set of choices facing large drugmakers. Buyers are no longer prioritizing size for its own sake. They are paying more for late-stage, near-readout, near-filing and near-launch assets because those assets provide something that can be valued with greater confidence: time. ## The Difference Between 52 and 42 Is Not Merely Ten Deals BioSpace's 52-transaction count is a list of whole-company acquisitions. It shows that activity increased sharply from approximately 30 transactions in the comparable 2025 period. Eli Lilly led the group with nine acquisitions and more than $25 billion in committed value. An earlier circulating account listed 11, but that does not match this dataset. IQVIA's 42-transaction count applies the higher $250 million threshold. The average deal size in this group was approximately $3.1 billion, up from $2.7 billion in 2025. Nine transactions reached at least $5 billion, yet not one target was valued above $30 billion. In other words, the market was not carried by one Celgene-scale megamerger. It was built from a cluster of transactions broadly worth $5 billion to $11 billion, involving assets whose science and development timelines were relatively concrete. PwC reached a similar conclusion from another angle. Biopharma deal value exceeded $65 billion in the first quarter of 2026, with 16 announced transactions worth at least $1 billion. Its midyear outlook described the market as being driven less by scale and more by precision science. Buyers have favored midsize acquisitions that can be integrated into existing franchises, while using contingent value rights, milestones and other conditional payments more frequently. So the better question is not merely whether the M&A market has returned. It is why large pharmaceutical companies are suddenly prepared to pay premiums for these particular assets. ## Drugmakers Are Buying a Patent Clock, Not a Story The answer begins with one number. PwC estimates that more than $300 billion in branded-drug revenue is exposed to loss-of-exclusivity risk during this decade. That does not mean $300 billion disappears on one date, nor does it mean every dollar will be lost. It means large pharmaceutical companies must establish sufficiently large new revenue curves before older products decelerate. Early platforms can be valuable, but the path from animal studies through Phase I, Phase II and a pivotal trial is difficult to price. Time, failure probability and additional capital requirements remain highly uncertain. When a major patent has only a few years left, buyers care intensely about three things: how far the clinical evidence has advanced, how long it will take to reach the next readout, and whether the asset can immediately connect to the buyer's development and commercial system. That is what “buying time” means. Merck announced in March that it would acquire Terns Pharmaceuticals for approximately $6.7 billion in equity value and completed the transaction on May 5. The lead asset, TERN-701, is an oral allosteric BCR::ABL1 inhibitor for chronic myeloid leukemia. It remains in the Phase I/II CARDINAL trial and is not an approved medicine. The company acquisition is complete, but Merck has not purchased guaranteed revenue. AbbVie announced in June that it would acquire Apogee Therapeutics for approximately $10.9 billion in equity value, gaining an immunology pipeline led by atopic dermatitis and asthma programs. The lead asset, zumilokibart (APG777), is a half-life-extended IL-13 monoclonal antibody. The transaction is expected to close in the third quarter, subject to shareholder and regulatory conditions. An announced acquisition is not a completed acquisition. The common logic is clear. These transactions are not designed to combine the back offices of two giant pharmaceutical companies. They are designed to place clinically legible assets inside established therapeutic-area and commercial engines. ## Lilly Went Shopping, but the Count Is Nine, Not Eleven Eli Lilly was the most active buyer in the first half of 2026. Using the whole-company acquisition definition applied by IQVIA and BioSpace, the count is nine, not 11. IQVIA estimated that Lilly invested approximately $26 billion, while BioSpace described more than $25 billion in committed value. The two figures are close, but invested capital, committed value and eventual cash paid are not identical concepts. Lilly's defining feature is that it is not simply defending obesity and diabetes. It is using the cash-generating power of GLP-1 products to expand into immunology, neuroscience, in vivo cell therapy, vaccines and other areas. IQVIA highlighted Lilly's acquisition of three vaccine companies in May as one representative move; the three transactions totaled $3.8 billion. That does not mean Lilly has already turned every new field into a second growth pillar. Frequent acquisition is an allocation decision. The real test comes afterward: trial design, program priority, integration speed and the discipline to terminate weak programs. A company can afford to buy ten assets without having a scientific or commercial reason to move all ten into Phase III. Earlier accounts also included several companies and transaction values that could not be reconciled with Lilly's primary announcements. Those names are not repeated here. M&A reviews become misleading when acquisitions, licenses, investments and collaborations are presented as one trophy list. The count becomes more impressive while the analysis becomes less reliable. ## Gilead and GSK Are Betting on Oncology Assets Near Commercialization Gilead's acquisition of Arcellx shows the value of time especially clearly. Gilead completed the acquisition on April 28, 2026. The combination of cash and a contingent value right worth $5 per share implied an equity value of approximately $7.8 billion at closing. The target asset was anito-cel, a BCMA-directed CAR-T therapy still under development. Completing the transaction gave Gilead full control and eliminated profit-sharing, milestone and royalty obligations that had existed under the prior partnership. It is accurate to say the acquisition was completed because the company announcement explicitly reported successful completion. Anito-cel, however, still awaits regulatory approval. Completing the purchase of a company and proving the success of its product are different events. Gilead also completed its acquisition of Tubulis on May 21. The terms included $3.15 billion in upfront cash and as much as $1.85 billion in milestones. Describing the transaction only as a “$5 billion acquisition” combines guaranteed payment with amounts that will be paid only if future conditions are met. The company transaction is complete, but lead asset TUB-040 remains in a Phase I/IIa trial. Product risk and milestone risk did not disappear at closing. GSK announced in June that it would acquire Nuvalent for total equity value of $10.6 billion and completed the transaction on July 15. It gained two late-stage targeted lung cancer agents, zidesamtinib and neladalkib, as well as the Phase I HER2 inhibitor NVL-330. The first two assets remain under FDA review, with target action dates of September 18 and November 27, 2026. Nuvalent can therefore be described as a completed company acquisition involving assets close to market. It cannot be described as a portfolio of launched products that will certainly contribute profit in 2027. GSK's own expectations for product launch remain conditional on FDA approval. Together, the three transactions form a useful risk spectrum: - Arcellx: the company acquisition is complete, but the product still awaits regulatory approval. - Tubulis: the company acquisition is complete, but TUB-040 remains in early clinical development and part of the consideration is milestone-based. - Nuvalent: the company acquisition is complete, but two late-stage products still await FDA decisions. Translating “the company was acquired” into “the product is about to generate revenue” removes the most important investment risk from the story. ## Even the Largest Deal Cannot Be Read from the Headline Alone One of the largest transactions of the first half involved India's Sun Pharma and Organon. The companies announced an enterprise value of $11.75 billion and cash consideration of $14 per share. The transaction is expected to close around early 2027 after regulatory and shareholder conditions are satisfied. BioSpace, using S&P Capital IQ and a methodology that incorporates debt and other factors, listed the value as $12.6 billion. The difference comes from methodology. The $12.6 billion figure should not be presented as the cash purchase price announced by the companies. This transaction also differs from the pipeline-centered deals above. Organon brings existing women's-health, established-medicine and biosimilar products, together with a global commercial network. Sun Pharma is acquiring a portfolio, geographic reach and a cash-flow platform rather than placing a single bet on one clinical event. That distinction matters. Precision and late-stage assets may dominate the 2026 narrative, but there is no single acquisition template. A mature portfolio can still be strategically rational when the buyer can clearly identify the gap it will fill. ## Why a Hot M&A Market Does Not End the Biotech Winter For sellers, more acquisitions are clearly constructive. M&A provides an exit path outside the IPO market and gives data-rich, cash-constrained companies another strategic option. But large pharmaceutical companies shopping more aggressively does not return the cost of capital for every biotech to its previous low. It also does not mean a platform story alone will attract a buyer. PwC is seeing more surgical transactions. An asset must fill a patent gap, open a high-growth market or provide a near-term clinical or commercial event. Buyers are also using CVRs and milestones to move part of the failure risk back to sellers. That market is more selective than a broad bull cycle, not less. The forces that could cool activity are equally specific. US drug-pricing policy, most-favored-nation proposals, expansion of Inflation Reduction Act negotiations, tariffs, cross-border review and US-China policy uncertainty can all alter valuation and payment structure. If a small group of large pharmaceutical companies drives prices too high, boards may also prefer to wait for additional data rather than chase assets at the top of the market. That is why the first-half total cannot simply be extrapolated into the second half. A more reliable monitoring framework asks: 1. Does the number of transactions worth at least $1 billion remain elevated, rather than being carried by one exceptionally large deal? 2. Is upfront cash becoming a smaller share of total potential value as more consideration moves into milestones? 3. Do buyers continue to prefer Phase III, pre-filing or FDA-review assets? 4. Do announced transactions close on schedule, rather than remaining press-release events? 5. One year after acquisition, are the assets accelerated, delayed or terminated? ## What Should Readers in Taiwan Watch? There is not enough primary-source evidence to identify a Taiwan-listed company as a direct product, manufacturing, testing or commercialization beneficiary of the transactions reviewed here. This article will not manufacture a list of “M&A concept stocks.” The more relevant lesson for Taiwan biotech companies is that buyer preference has become more concrete. Clinical data must establish differentiation. The pivotal-trial pathway must be legible. Global rights must be organized cleanly. CMC and supply must be capable of connecting to a large pharmaceutical system. A platform can support imagination, but the transaction file must survive due diligence line by line. Investors should perform two simple checks. When a release cites “maximum transaction value,” separate cash, equity value, enterprise value, contingent value rights and milestones. When a release says a deal is “complete,” return to the company announcement and verify whether the transaction was announced, signed or actually closed. Those two steps eliminate many apparently spectacular but analytically weak numbers. ## Conclusion: $130 Billion Is Really a Purchase of Time There is no question that biopharma M&A recovered in the first half of 2026. BioSpace's 52 transactions and IQVIA's 42 transactions both point to materially higher activity. PwC's more than $65 billion in first-quarter deal value and 16 billion-dollar transactions provide additional support. The problem begins when different methodologies are assembled into one perfect headline and the aggregate value is used to declare that the capital winter is over. What buyers are competing for is clinical time that can arrive before the patent cliff. The closer an asset is to readout, filing or launch, the more likely its price is to rise. Risk has not disappeared. It has been redistributed among upfront cash, CVRs, milestones and closing conditions. Big Pharma has not suddenly become romantic. It is looking at a patent clock and using cash to buy time. ## Disclaimer This article provides biotechnology, pharmaceutical and industry-trend analysis. It does not constitute medical diagnosis, treatment advice or investment advice. Transaction values can differ depending on whether a source reports equity value, enterprise value, debt, cash, contingent value rights or milestones. Investment decisions should be based on company announcements and regulatory documents. --- ## Roche's Hidden Succession Engine: How 19 New Molecules Could Power the Next Growth Cycle URL: https://drugnews.com.tw/articles/2026-07-25-roche-19-molecule-succession-engine-en.html Date: 2026-07-25 Language: en Summary: Roche generated CHF 47.669 billion in pharmaceutical sales in 2025, with Ocrevus, Hemlibra, and Vabysmo already accounting for roughly one-third of the total. The number that matters is not simply 19 potential new molecules, but whether giredestrant, fenebrutinib, obesity, Alzheimer's, and protein-degradation assets can inherit Roche's existing clinical, diagnostic, and commercial infrastructure on time. When one company is placing bets across breast cancer, multiple sclerosis, Alzheimer's disease, obesity, and targeted protein degradation, the easy description is "portfolio diversification." It is a safe phrase. It is also nearly devoid of information. Put Roche's moves over the past year on a single timeline and the picture looks less like diversification than a relay race. Roche's pharmaceutical division generated CHF 47.669 billion in sales in 2025. Three non-oncology products — Ocrevus, Hemlibra, and Vabysmo — contributed CHF 15.866 billion, or roughly one-third of pharmaceutical revenue. In 2026, two applications for the oral breast cancer drug giredestrant entered FDA review; the multiple sclerosis candidate fenebrutinib completed three positive phase 3 trials; and additional legs of the race began to take shape in obesity, Alzheimer's disease, and protein degradation. At its 2025 Pharma Day, Roche said that as many as 19 new molecules could have launch potential before 2030. That is a large number, but it is not the answer that matters most. The more important calculation is how many of those runners can take the baton on time. - In the near term, watch two FDA applications and three positive phase 3 trials. - In the middle distance, ask whether obesity and Alzheimer's assets can become new pillars. - Further out, test whether recently acquired technologies can scale through Roche's existing channels. ## 01 | The Transformation Has Already Happened. The Income Statement Said So First. When investors discuss Roche, they often start with HER2-positive breast cancer, hematologic malignancies, and a roster of classic antibodies. That impression is not wrong. It simply overlooks half of the income statement. Ocrevus generated CHF 7.010 billion in 2025 sales, Hemlibra CHF 4.754 billion, and the ophthalmology drug Vabysmo CHF 4.102 billion. Roche's five leading growth drivers contributed a combined CHF 21.4 billion, and only Phesgo was an oncology product. The others spanned neuroscience, hematology, ophthalmology, and immunology. Roche's revenue engine is already diversified. The more useful question in 2026 is whether the next products can enter the same specialties, reach the same physicians, and use the same commercial system before the current engines begin to slow. That timing gap will determine whether Roche's growth cycle forms one continuous curve or a series of isolated peaks. Big Pharma does not fear aging drugs. The painful part is when the next runner has not reached the bend. ## 02 | The Two Legs Closest to the Finish: Giredestrant and Fenebrutinib In breast cancer, Roche is betting on giredestrant, an oral selective estrogen receptor degrader. The FDA has accepted two applications. The first covers giredestrant plus everolimus for ESR1-mutated, ER-positive, HER2-negative advanced breast cancer, with a PDUFA target date of December 18, 2026. The second covers adjuvant treatment in ER-positive, HER2-negative early breast cancer. It received priority review and has a target date of November 30. Together, the applications span advanced and early disease. Roche now has an opportunity to hand more than three decades of breast cancer clinical development, regulatory experience, and commercial infrastructure to a next-generation oral endocrine therapy. The succession logic in neuroscience is equally clear. Ocrevus is already Roche's largest product, with CHF 7.010 billion in 2025 sales. The drug behind it is fenebrutinib, an oral BTK inhibitor designed to penetrate the brain. FENhance 1, FENhance 2, and FENtrepid all met their primary endpoints across relapsing and primary progressive multiple sclerosis. Roche has said it will submit the full data package to regulators. The commercial attraction is straightforward: the same multiple sclerosis centers and physician networks could potentially support both injectable Ocrevus and oral fenebrutinib, while giving patients a wider range of choices. This leg of the race still carries risk. Roche's complete April release reported seven deaths from different causes in the fenebrutinib groups during the reporting period across the two relapsing multiple sclerosis studies, plus one after the reporting period, compared with one death in the teriflunomide groups. Further analysis remains underway. Positive phase 3 results make the path to filing clearer; the safety profile will influence labeling, physician adoption, and the eventual pace of uptake. ## 03 | The Next Layer of Options: Obesity, Alzheimer's, and BTK Degradation Assets approaching regulatory review protect the near term. Roche is also buying time for the revenue curve further out. Its heaviest commitment is in obesity. Roche partnered with Zealand Pharma on petrelintide for $1.65 billion upfront and as much as $5.3 billion including development and commercial milestones. In phase 2 data released in 2026, participants achieved mean weight loss of up to 10.7% at week 42, versus 1.7% with placebo. Roche is now moving petrelintide and enicepatide, formerly CT-388, toward phase 3 development and is planning a phase 2 combination study. Alzheimer's disease puts Roche's pharmaceutical and diagnostic capabilities on the same map. Two phase 3 studies of trontinemab, TRONTIER 1 and TRONTIER 2, are underway. A third study, PrevenTRON, is designed for asymptomatic people whose biomarkers indicate a high risk of disease. Roche also has the CE-marked Elecsys pTau217 blood test to help identify amyloid pathology. As treatment moves into earlier populations, finding the right patient becomes as important as the drug's effect. Roche has a rare end-to-end pathway: diagnostics find the patient, clinical trials test the medicine, and the same healthcare system can support commercialization. Its June agreement with Nurix Therapeutics added another cross-disease option. Roche paid $700 million upfront for co-development and commercialization rights to bexobrutideg, also known as NX-5948, a BTK degrader being developed across B-cell malignancies, immunology, and neurological disease. The potential transaction value is as high as $2.3 billion. These programs appear dispersed on the surface. Behind them, the capabilities overlap heavily: antibody and molecular engineering, patient stratification, global phase 3 execution, diagnostics, and specialist channels that already exist. ## 04 | Roche's Most Valuable Asset Is the Time It Does Not Need to Spend Rebuilding a Market Roche's 19 potential new molecules constitute a large option pool. They will not all succeed, and they do not need to. Our team is more interested in the "succession rate." The most expensive work for a large drugmaker often begins after approval: educating physicians again, establishing testing, negotiating reimbursement, and building a commercial organization. Roche's two closest-to-market programs already reconnect with its breast cancer and multiple sclerosis franchises, while trontinemab links into an existing diagnostics entry point. Whether those assets can avoid the time and cost of rebuilding a market will affect the next revenue cycle sooner than the headline count of 19. Succession rate asks a more complete question than which individual drug will win. It must also estimate when established products will slow, when a successor can enter, how much time existing physician networks and diagnostics can save, and how much Roche must still spend on milestones, trials, and launch execution after acquiring an asset. Both giredestrant review dates fall near the end of 2026. Fenebrutinib still has to pass regulatory and safety scrutiny. Obesity is crowded with exceptionally well-funded competitors. Trontinemab ultimately has to prove value on clinical functional outcomes. If any runner loses pace, the revenue curve could still develop a gap. Pipeline size determines how long the story can be. Succession rate determines whether the revenue curve breaks. ## 05 | A Taiwan Comparison: Can BRIM Biotechnology Hand Ophthalmology Assets to a Mature Channel? Taiwan's BRIM Biotechnology, ticker 6885, offers a smaller-scale comparison. BRIM concentrates its main research resources in ophthalmology. Its website lists BRM421 for dry eye disease, BRM424 for neurotrophic keratitis, BRM411 for glaucoma, and BRM412 for neovascular eye disease. A Taiwan phase 2 dose-ranging study of BRM421 has received approval in principle to proceed. BRM424 has completed a protocol change for its US phase 2 study and begun enrollment. BRIM is in the first half of the relay: advancing candidates toward clinical proof of concept that could support licensing, then handing them to a partner with late-stage development and commercialization capabilities. Roche sits at the other end of the value chain, showing how a mature pharmaceutical company can amplify ophthalmology assets with specialist physicians, imaging diagnostics, reimbursement access, and global distribution. The companies differ greatly in scale and commercial position. The comparison is useful because it shows how value changes hands along the same chain. Vabysmo's CHF 4.102 billion in 2025 sales demonstrates that an ophthalmology product can become a major Big Pharma pillar once it connects with specialists, imaging, reimbursement, and global channels. BRIM's investment question can be reduced to three verbs: prove, license, scale. - **Prove:** Can BRM421's new formulation and dosing support a phase 3 design? Can BRM424 establish a clear clinical effect in phase 2? - **License:** After proof of concept, can BRIM secure an international partner with ophthalmology channels and late-stage development expertise? - **Scale:** Can manufacturing, formulation, and supply support multinational trials and eventual commercialization? Roche's current strategy offers a direct lesson for Taiwanese drug developers: an asset's value has a chance to expand only after it connects with a mature disease franchise. Science opens the door. Clinical execution, licensing, and commercial channels determine how far that door leads. ## Conclusion | Roche's Deepest Moat Is Preventing the Revenue Baton From Being Dropped Roche's next growth cycle may not be supported by one suddenly dominant blockbuster. Giredestrant reconnects with breast cancer, fenebrutinib with multiple sclerosis centers, and trontinemab with pTau217 testing. Obesity and BTK degradation provide options further out. Every program carries risk, but each has an existing place to land. Drugnews' assessment is that Roche has already completed the first stage of diversification. Two FDA decisions near the end of 2026, the safety review of fenebrutinib, and the transition of obesity assets into phase 3 will deliver the first meaningful report card for this operating system. What is hardest to replicate in Big Pharma often sits outside the headline. It includes physician relationships, trial networks, diagnostic entry points, regulatory experience, and commercial teams. Those capabilities rarely appear beside a drug's name. Their weight becomes visible when established and emerging products exchange the baton. A blockbuster can illuminate a few years. A succession system can keep the light across the entire revenue statement. ## Disclaimer This article is provided solely for industry research and market observation. It does not constitute investment advice, a recommendation to buy or sell securities, or an endorsement of any company. Biopharmaceutical investing involves clinical, regulatory, licensing, commercialization, currency, and capital-market risks. Readers must make their own decisions and bear responsibility for the outcomes. --- ## 羅氏的隱藏底牌：19 個新分子排隊接棒，下一個成長週期怎麼長出來？ URL: https://drugnews.com.tw/articles/2026-07-25-roche-19-molecule-succession-engine.html Date: 2026-07-25 Language: zh-Hant Summary: 羅氏 2025 年製藥銷售 476.69 億瑞士法郎，Ocrevus、Hemlibra 與 Vabysmo 已撐起約三分之一收入。真正值得追的不是 19 個新分子的總數，而是 giredestrant、fenebrutinib、減重、阿茲海默症與蛋白質降解資產能否沿既有臨床、診斷與商業通路準時接棒。 一家公司同時在乳癌、多發性硬化症、阿茲海默症、減重與蛋白質降解下注，很容易被寫成「多元布局」。 這四個字很安全，也幾乎沒有資訊量。 把羅氏近一年的動作排在一起，畫面更像一場接力賽。 2025 年，羅氏製藥事業銷售 476.69 億瑞士法郎。Ocrevus、Hemlibra、Vabysmo 三款非腫瘤產品合計 158.66 億瑞士法郎，已經撐起約三分之一的製藥收入。到了 2026 年，乳癌口服新藥 giredestrant 有兩項申請進入 FDA 審查；多發性硬化症新藥 fenebrutinib 連續完成三個陽性三期；減重、阿茲海默症與蛋白質降解療法也往後排出新的棒次。 羅氏在 2025 年藥品日提出，2030 年前最多有 19 個新分子具備上市潛力。這個數字很大，卻還不是最關鍵的答案。 真正值得算的，是下一棒有多少能準時接上。 - 近端，看兩項 FDA 申請與三個陽性三期。 - 中段，看減重和阿茲海默症能否長成新支柱。 - 遠端，看剛買進的技術能否沿著既有通路放大。 ## 01｜轉型早已發生，收入表先說了 市場談羅氏，第一個想到的仍常是 HER2 乳癌、血液腫瘤與幾款經典抗體藥。這個印象沒有錯，只是少看了半張收入表。 2025 年，Ocrevus 銷售 70.10 億瑞士法郎，Hemlibra 為 47.54 億，眼科藥 Vabysmo 也來到 41.02 億。羅氏前五大成長動能合計貢獻 214 億瑞士法郎，其中只有 Phesgo 屬於腫瘤；其餘還包括神經、血液、眼科與免疫產品。 羅氏的收入引擎早已分散。2026 年更值得追的問題，是新藥能否在舊引擎放慢之前，沿著同一個科別、同一批醫師與同一套商業系統接進來。 這個時間差，會決定成長週期是一條連續曲線，還是幾座彼此分離的高峰。 大藥廠不怕藥會變老。真正難熬的，是接棒的人還沒跑到彎道。 ## 02｜最靠近終點的兩棒：giredestrant 與 fenebrutinib 乳癌這一棒，羅氏押的是口服選擇性雌激素受體降解劑 giredestrant。 FDA 已受理兩項申請。第一項用於 ESR1 突變、ER 陽性、HER2 陰性的晚期乳癌，搭配 everolimus，預定審查期限為 2026 年 12 月 18 日；第二項則用於 ER 陽性、HER2 陰性的早期乳癌輔助治療，獲優先審查，預定審查期限為 11 月 30 日。 兩個申請橫跨晚期與早期病人。羅氏累積三十多年的乳癌臨床、法規與商業網路，現在有機會交給下一代口服內分泌療法繼續使用。 神經科學的接棒同樣清楚。 Ocrevus 已是羅氏最大的產品，2025 年銷售 70.10 億瑞士法郎。接在後面的 fenebrutinib，是可進入腦部的口服 BTK 抑制劑。FENhance 1、FENhance 2 與 FENtrepid 三個關鍵三期均達主要終點，涵蓋復發型與原發進展型多發性硬化症，羅氏表示將把完整資料送交監管機關。 商業上的吸引力很直接：同一套多發性硬化症中心與醫師網路，有機會同時承接注射型 Ocrevus 與口服 fenebrutinib，病人選擇也會變多。 這一棒仍有風險。羅氏四月公布的完整資料顯示，在兩個復發型多發性硬化症試驗中，fenebrutinib 組於報告期內出現七例不同原因的死亡，報告期後另有一例；teriflunomide 組為一例，進一步分析仍在進行。三期陽性讓申請資格變得更清楚，安全性資料則會直接影響標籤、醫師接受度與最終放量速度。 ## 03｜下一層選擇權：減重、阿茲海默症與 BTK 降解 接近審查的產品守住近端，羅氏也在替更遠的收入曲線買時間。 減重是出手最重的一塊。羅氏與 Zealand Pharma 合作開發 petrelintide，簽約金 16.5 億美元，開發與銷售里程碑計入後，總潛在價值最高 53 億美元。2026 年公布的二期試驗中，受試者在第 42 週平均減重最高 10.7%，安慰劑組為 1.7%。羅氏目前正把 petrelintide 與 enicepatide（CT-388）推向三期開發，也規劃兩者的二期組合試驗。 阿茲海默症則把羅氏的藥品與診斷能力放到同一張圖上。Trontinemab 的 TRONTIER 1、2 兩個三期試驗正在進行；另一個 PrevenTRON 三期，準備研究尚未出現症狀、但生物標記顯示高風險的人。羅氏同時擁有已取得 CE 標誌的 Elecsys pTau217 血液檢測，能協助辨認類澱粉蛋白病理。 藥物往更早期病人移動時，先找出誰適合治療，會變得和藥效一樣重要。羅氏在這裡擁有一條少見的完整路徑：檢測先找到人，臨床試驗再驗證藥，商業化之後仍由同一套醫療體系接手。 六月與 Nurix Therapeutics 的合作，又添了一張跨領域選擇權。羅氏支付 7 億美元簽約金取得 BTK 降解劑 bexobrutideg（NX-5948）的共同開發與商業化權利，總潛在交易價值最高 23 億美元，開發範圍橫跨 B 細胞惡性腫瘤、免疫與神經疾病。 這些題目表面分散，後台卻高度共用：抗體與分子工程、病人分層、全球三期執行、診斷能力，以及已經存在的科別通路。 ## 04｜羅氏最值錢的，是省下重建市場的時間 羅氏的 19 個潛在新分子，代表一個很大的選擇權池。它們不會全部成功，也不需要全部成功。 我們團隊更在意「接棒率」。大藥廠最昂貴的工作，常在核准之後：重新教育醫師、建立檢測、談給付，再拉起一支商業團隊。最靠近上市的兩條線，已經清楚接回羅氏原有的乳癌與多發性硬化症版圖；trontinemab 則接上診斷入口。它們能否省下重建市場的時間，會比 19 這個總數更早影響下一段收入。 接棒率要回答的，比哪一顆藥會成功更完整。它還要估算：現有大藥何時放緩，下一顆藥何時能進場，原有的醫師網路與診斷入口能替它省下多少重建市場的時間，以及買進資產後還要負擔多少里程碑、試驗與上市費用。 Giredestrant 的兩個審查期限集中在 2026 年底，fenebrutinib 還要跨過監管與安全性檢視；減重市場擠滿資金雄厚的對手；trontinemab 最終仍得用臨床功能終點證明價值。任何一棒慢下來，都可能讓收入曲線留下空檔。 因此，管線數量只決定故事有多長；接棒率才決定收入曲線有沒有斷點。 ## 05｜台灣看全福生技：眼科資產能不能接上成熟通路 台灣可以用全福生技（6885）做一個小型對照。 全福把主要研發資源放在眼科。公司官網列出的產品包括乾眼症新藥 BRM421、神經營養性角膜炎新藥 BRM424、青光眼新藥 BRM411，以及新生血管性眼疾新藥 BRM412。BRM421 的台灣二期劑量探索試驗已獲原則同意進行；BRM424 則完成美國二期設計變更並啟動收案。 全福目前站在接力賽的前半段：把候選藥推到可授權的臨床概念驗證，再交給具備後期開發與商業化能力的夥伴。羅氏站在價值鏈另一端，展示成熟藥廠如何用專科醫師、影像診斷、給付與全球通路放大眼科資產。兩者規模與商業位置不同，放在一起看的意義，是同一條價值鏈如何交棒。 Vabysmo 在 2025 年已賣出 41.02 億瑞士法郎，說明眼科產品只要接上專科醫師、影像診斷、給付與全球通路，也能長成大型藥廠的重要支柱。全福的投資問題可以壓縮成三個動詞：證明、授權、放大。 - 證明：BRM421 的新配方與劑量能否支持下一步三期設計？BRM424 二期能否建立清楚的臨床效果？ - 授權：概念驗證完成後，能否找到有眼科通路與後期開發能力的國際夥伴？ - 放大：製程、配方與供應能否承受多國臨床和未來商業化？ 羅氏這一輪布局給台灣新藥公司最直接的提醒，是一項資產接進成熟疾病版圖之後，價值才有機會被放大。科學先把門打開，後面的臨床、授權與通路，決定這扇門能通到哪裡。 ## 結語｜羅氏最深的護城河，是不讓收入斷棒 羅氏的下一個成長週期，未必會由某一顆突然爆紅的藥單獨撐起。 Giredestrant 接回乳癌版圖，fenebrutinib 接回多發性硬化症中心，trontinemab 與 pTau217 把診斷和治療串在一起，減重與 BTK 降解劑則提供更遠的選擇權。每一條線都有風險，卻都能找到一個已經存在的落點。 藥時事團隊的判斷是：羅氏的多元化早已完成第一段路。2026 年底的兩項 FDA 決定、fenebrutinib 的安全性檢視，以及減重資產的三期推進，會交出這套系統的第一批成績單。 大藥廠真正難複製的，往往藏在新聞標題以外。它包括醫師關係、試驗網路、診斷入口、法規經驗和商業團隊。這些能力平常不會出現在藥名旁邊，到了新舊產品交棒時，才看得出重量。 一顆明星藥會照亮幾年；一套接棒系統，才可能把光留在整張收入表上。 ## 免責聲明 本文僅為產業研究與市場觀察，不構成任何投資建議、買賣建議或個股推薦。生技醫藥投資涉及臨床試驗、法規審查、授權談判、商業化、匯率與資本市場波動等多重風險，投資人應自行判斷並自負盈虧。 --- ## SpaceX Goes to Orbit. Is the Money in the Crystals? URL: https://drugnews.com.tw/articles/2026-07-24-spacex-microgravity-drug-crystals-en.html Date: 2026-07-24 Language: en Summary: SpaceX is making low-Earth-orbit round trips more routine, but rockets do not discover drugs. From Keytruda crystallization and SpaceMD to Varda, UK regulation, and a Taiwan mission, this analysis asks whether microgravity-enabled manufacturing can pass the tests of repetition, scale-up, and regulation. > Evidence cut-off: July 19, 2026. SpaceX supplies transportation to and from low Earth orbit; it does not develop medicines. The commercial question is whether a better crystal grown in microgravity can be converted into a repeatable, scalable, and regulator-ready process on Earth. On June 17, 2026, a SpaceX Dragon spacecraft splashed down in the Pacific Ocean carrying biological, materials-science, and medical research samples. It was NASA's 34th SpaceX commercial resupply mission. Launching a payload, entering the International Space Station, completing an experiment, and returning the samples to Earth was beginning to look less like a one-off expedition and more like a scheduled route. Three weeks later, Redwire's SpaceMD appointed two new advisers. Paul Reichert, a former Merck research leader, had worked on early crystallization research involving Keytruda. Niki Werkheiser, a former NASA executive, had managed more than $1.5 billion in technology investments. Reichert understands crystals. Werkheiser understands how space technology moves beyond an experiment. The gap SpaceMD is trying to close is the one between “we can do it” and “a customer will pay to do it repeatedly.” - SpaceX and Dragon provide a round trip: send an experiment up, then bring the result back down. - Microgravity suppresses convection and sedimentation, allowing some proteins and small molecules to form more uniform crystals. - Commercial success is still settled on Earth: how much better is the crystal, can the process be scaled, and is the improvement worth the cost of going to orbit? Drugnews' view is that the first meaningful revenue in space-based pharmaceutical manufacturing is most likely to come from the smallest, highest-value, and hardest-to-reproduce step on Earth: solid-form screening, seed-crystal growth, and formulation-condition selection. Rockets may drive down the cost per kilogram. A medicine still has to drive up the value per gram. ## 01 | SpaceX Is Turning Orbit into a Laboratory That Can Be Revisited SpaceX's role in this value chain is clear: transportation to and from low Earth orbit. NASA's commercial resupply program allows Dragon to deliver scientific hardware to the International Space Station and return samples that require analysis on Earth. For pharmaceutical research, coming back is as important as going up. If crystals, cells, or tissue samples cannot be recovered quickly, kept within their temperature requirements, and transferred to an analytical laboratory, an orbital result cannot readily enter the next development cycle. Redwire's PIL-BOX grew out of this logistics system. It is a miniature crystallization platform that can be activated remotely on the International Space Station. A pharmaceutical customer loads a candidate molecule before launch, crystallization occurs after the device reaches the station, and the hardware returns to Earth aboard Dragon. In a July 2026 corporate release, Redwire said that 54 PIL-BOX units had traveled to the International Space Station since the platform's first flight in November 2023 and that 45 compounds had been crystallized. The disclosed work included insulin and molecules associated with cancer, cardiovascular disease, obesity, and diabetes. These are company-reported figures, and they remain far removed from drug approval or commercial-scale production. What they do show is that one hardware platform can be used repeatedly. Space crystallization is moving from an exceptional mission that takes years to arrange toward a recurring experimental cycle. [[Redwire/SpaceMD](https://ir.rdw.com/news-events/press-releases/detail/235/former-merck-and-nasa-leaders-bring-strategic-expertise-to)] That is the most practical pharmaceutical effect of the space activity associated with SpaceX. A rocket does not discover a drug. Reliable, recoverable transport can nevertheless turn the microgravity environment roughly 400 kilometers above Earth into a research tool that a company can book, test, fail with, and use again. ## 02 | Keytruda's Trip to Space Became a Lesson in Formulation Physics Why might microgravity change a medicine? The answer lies in the invisible fluid motion around a growing crystal. Liquids on Earth are affected by gravity, creating convection and sedimentation. As protein molecules move toward a crystal surface, the concentration field can be disturbed. A crystal may also settle to the bottom of a vessel before it has finished growing, collide with other crystals, aggregate, or develop a broad and defective size distribution. In low Earth orbit, molecules move mainly through diffusion. With less bulk motion around them, crystals may assemble more steadily. The advantage is not universal. Some molecules form larger crystals, some show improvements only in defects or particle-size distribution, and some do not exhibit a difference large enough to matter for manufacturing. Merck's pembrolizumab, the active ingredient in Keytruda, provides one of the clearest examples. On February 19, 2017, researchers launched a pembrolizumab crystallization experiment aboard SpaceX CRS-10. The hardware operated in an incubator on the space station for 18 days and returned to Earth aboard Dragon on March 19. A 2019 paper in *npj Microgravity* reported that the ground controls produced two widely separated crystal-size populations, centered at approximately 13 and 102 micrometers. The microgravity group instead produced a single distribution with an average size of approximately 39 micrometers. At the same concentration, the viscosity of the space-grown crystal suspension was also approximately 2 cP lower, while binding activity after redissolution remained comparable with the reference material. [[npj Microgravity](https://www.nature.com/articles/s41526-019-0090-3)] The researchers then brought two variables revealed by the space experiment back to Earth: sedimentation and temperature gradients. By combining slow rotation with temperature control, they produced a more uniform, injectable crystal suspension in a conventional laboratory. The causal chain has to be described carefully. Crystals grown in orbit were not placed directly into a later commercial vial. In a January 2026 retrospective, NASA said the microgravity work provided early insight into particle size and structure relevant to a subcutaneous formulation. Keytruda Qlex, which the FDA approved in 2025, still went through formulation development, clinical comparison, and regulatory review. [[NASA](https://www.nasa.gov/missions/station/iss-research/space-station-research-informs-new-fda-approved-cancer-therapy/); [FDA](https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-pembrolizumab-and-berahyaluronidase-alfa-pmph-subcutaneous-injection)] Space does not complete a clinical-development program. At most, it may help certain molecules escape a physical constraint that is difficult to solve on Earth. ## 03 | One Orbit, Three Emerging Business Models SpaceMD is selling access to seed crystals. It sends small quantities of a candidate molecule into orbit, grows seeds in microgravity, and then sells or licenses those seeds to a pharmaceutical company. New solid forms, formulations, and commercial manufacturing processes are developed back on Earth. Varda Space Industries goes one step further. It builds uncrewed reentry capsules in which small molecules can be heated, cooled, and crystallized in orbit. As of July 2026, Varda's official mission page listed six reentry missions, W-1 through W-6. W-1 processed ritonavir, an HIV medicine. In three vials containing approximately 150 milligrams each, researchers heated stable Form II and converted it to Form III, a more soluble polymorph that can also transform more readily into other crystal forms. A peer-reviewed paper published in 2026 reported that Form III could be recovered after orbital processing, a long flight, and atmospheric reentry. Chemical analysis did not identify unusual impurities attributable to the space environment. [[Varda mission platform](https://www.varda.com/platform); [npj Microgravity](https://www.nature.com/articles/s41526-026-00594-0)] This result is best understood as a logistics and process qualification milestone. It shows that a high-value small molecule can be processed in orbit and returned in a condition that remains suitable for analysis. The United Kingdom is investing in rules and infrastructure. In March 2026, the UK Space Agency and the Medicines and Healthcare products Regulatory Agency announced a support program for pharmaceutical manufacturing in space. It included regulatory guidance, case studies, a regulatory sandbox for atmospheric reentry, and supply-chain coordination. The government was also supporting BioOrbit's scale-up research in crystallizing biologic drugs in space, with the aim of enabling high-concentration formulations that might be more suitable for home administration. By July 19, 2026, two additional policy milestones were visible. The MHRA said it expected to publish an end-to-end regulatory roadmap for in-orbit pharmaceutical manufacturing in autumn 2026. The UK Space Agency's 2025–2026 annual report, published on July 14, confirmed that feasibility-study contracts for BioOrbit, OrbiSky, and Space Forge had been awarded in February. These were regulatory preparations and feasibility studies. They did not mean that a medicine manufactured in space had received approval. All three models still have to settle their accounts on Earth. Can the additional quality created in orbit support new intellectual property, a better route of administration, a more stable product, or a lower-cost terrestrial process? A pharmaceutical company's next purchase order will provide a more useful answer than a launch count. ## 04 | The Added Value from Orbit Must Pass Four Commercial Gates Science-fiction imagery attracts attention. Pharmaceutical development still accepts only evidence. ### Gate One: Molecule selection Microgravity is most likely to matter for a molecule with a defined solid-form problem, a high dose, poor solubility, high viscosity, or repeated failures in a terrestrial process. If a conventional process already crystallizes a molecule reliably, sending it into orbit adds an expensive travel bill without creating enough value. ### Gate Two: Transportation and timing Samples must survive launch vibration, radiation, temperature changes, waiting time in orbit, and atmospheric reentry. Any delay can alter the crystallization conditions specified in the experimental design. The ritonavir work by Varda matters because it begins to turn “can the material come back intact?” into measurable process data. ### Gate Three: Scale-up on Earth A small quantity of attractive crystals grown in space still has to enter a kilogram- or tonne-scale process. The most plausible model is for orbit to identify a solid form or grow a seed, while Earth handles commercial production. If every commercial batch has to make a round trip to orbit, only extremely high-value products in very small volumes can bear the cost. ### Gate Four: Pharmaceutical regulation Regulators will ask about starting-material provenance, instrument calibration, process deviations, aseptic control, chain of custody, quality after reentry, and responsibility for the orbital step across the product life cycle. The UK's regulatory sandbox is intended to surface these questions before a company submits a formal application. A high mission count only proves that the hardware is busy. More useful signals are whether the same pharmaceutical customer pays again, whether a recovered solid form can be licensed, and whether the terrestrial process can be scaled. If those three events occur, the next question is whether regulators will accept the orbital evidence in a marketing application. ## 05 | Taiwan Has Already Sent a Protein-Crystallization Experiment into Orbit Taiwan has one concrete coordinate on this map. In December 2021, Taiwan's National Synchrotron Radiation Research Center sent virus-like particles produced in *E. coli* to the International Space Station. Taiwan-based HelioX Cosmos and Japan's Space BD coordinated the mission, which flew aboard SpaceX CRS-24. The samples crystallized in microgravity and returned safely on January 24, 2022. [[Space BD](https://www.space-bd.com/dam/include-content/PUBLIC/en/news_archive/storage/news20220303-001/pdf/20220520en-4.pdf)] Space BD's contemporaneous release was careful. It said that the returned material would undergo X-ray analysis to study the structure of the virus-like particles and possible mechanisms of infection. The available announcement did not disclose a final analytical result. This article therefore confirms only that the mission, orbital crystallization, and sample return were completed. The mission still made two Taiwanese capabilities tangible. HelioX handled mission integration and international coordination. NSRRC operates protein-crystallography beamlines that can support subsequent X-ray diffraction and structural analysis. HelioX's website also lists microgravity life science and in-orbit services among its activities. HelioX is not a listed company, so this development is better followed as an industrial capability than forced into a stock-market theme. If Taiwan wants to turn a single experiment into a business, three signals matter: repeated paid missions from pharmaceutical customers, a workflow that connects an orbital crystal to terrestrial drug design, and a repeatable service combining mission integration with crystal analysis. Taiwan's strengths in structural biology, precision instrumentation, crystallography, and international project integration sit exactly where a space-returned sample needs help after landing. ## 06 | Around 2030, Orbital Laboratory Capacity Will Also Be Repriced NASA currently plans to operate the International Space Station through 2030 while advancing commercial space-station programs intended to carry low Earth orbit research into the next phase. If the transition leaves a gap, microgravity experiments may face pressure on station volume, electrical power, astronaut time, and return schedules. China is advancing a different model through a national research platform. In July 2025, the Shanghai Institute of Materia Medica sent a nucleic-acid lipid-nanocarrier experiment to China's space station. The stated objective was to study how microgravity affects the intracellular transport of nucleic-acid medicines. Subsequent official reporting said that dosing, microscopic imaging, and sample fixation had been completed in orbit and that the samples had been moved into low-temperature storage to await return. As of July 19, 2026, we had not found a complete formal paper or public result that supported a comparison of clinical efficacy. In the future, a pharmaceutical customer will buy an entire service chain: reliable orbital capacity, standardized hardware, timely return, terrestrial analysis, and a regulatory pathway. Remove any one of those components, and a space mission becomes difficult to place inside an ordinary research-and-development budget. This helps explain why SpaceMD recruited both a Merck crystallization scientist and a former NASA technology-investment executive. Microgravity science has existed for decades. What remains scarce is the ability to turn one successful experiment into a product that customers purchase repeatedly. ## Conclusion | The Next Formulation Answer May Take a Detour through Space The most compelling feature of space-enabled pharmaceutical research is that it changes the environment around a familiar problem. A molecule repeatedly disturbed by gravity on Earth may reveal a different solid form, a different assembly pattern, or a different formulation path in orbit. After an attractive crystal returns, it still has to pass three tests: repetition, scale-up, and regulation. Without all three, it remains an elegant experiment. The real endpoint is still a treatment that is more practical or more effective for a patient. SpaceX is helping orbital round trips look more like a service with a schedule. Pharmaceutical developers now have to show how much billable scientific value each trip leaves behind. The first things sent to orbit will not be tonnes of finished medicine. They will be a seed crystal, a small vial of candidate material, and a problem that has remained stuck on Earth. ## Disclaimer This article is provided for biotechnology and pharmaceutical-industry information and research discussion only. It does not constitute medical advice, investment advice, a recommendation to buy or sell securities, or a guarantee of returns. Drug development, in-orbit processing, and commercialization remain highly uncertain. Readers should rely on official documents and regulatory records and make independent decisions consistent with their own risk tolerance. --- ## SpaceX 上太空，錢在晶體裡？微重力製藥的第一桶金 URL: https://drugnews.com.tw/articles/2026-07-24-spacex-microgravity-drug-crystals.html Date: 2026-07-24 Language: zh-Hant Summary: SpaceX 讓低軌往返逐漸有了班次，但火箭不是新藥公司。本文從 Keytruda 結晶、SpaceMD、Varda、英國監管與台灣任務，拆解微重力製藥能否通過重複、放大與監管三道商業考驗。 > 查證截止：2026-07-19。SpaceX 提供的是往返低軌的交通，不是新藥研發；真正的商業考題，是微重力能否把更好的晶體轉化成可重複、可放大、可監管的地面製程。 2026 年 6 月 17 日，一艘 SpaceX Dragon 太空船濺落太平洋，帶回滿艙的生物、材料與醫學研究樣本。這是 NASA 第 34 次 SpaceX 商業補給任務。上太空、進入國際太空站、完成實驗、再把樣本送回地球，已經慢慢有了固定班次的樣子。 三週後，Redwire 旗下 SpaceMD 找來兩位新顧問：曾參與 Keytruda 早期結晶研究的 Merck 前研究主持人 Paul Reichert，以及曾管理逾 15 億美元技術投資的 NASA 前主管 Niki Werkheiser。 Reichert 懂晶體，Werkheiser 懂如何讓太空技術走出實驗。SpaceMD 要補的，正是從「做得出來」到「有人願意反覆買」之間那段路。 - SpaceX 與 Dragon 打通了「上去做實驗、下來看結果」的往返路線。 - 微重力讓對流與沉降安靜下來，某些蛋白質與小分子能長出更均一的晶體。 - 商業成敗仍要回到地面：晶體好多少、製程能否放大、改善值不值得支付太空成本。 藥時事團隊的判斷是，太空製藥的第一桶金，最可能落在體積最小、價值最高、最難在地球完成的那一步：晶型探索、晶種培養與製劑條件篩選。 火箭把每公斤成本往下壓，藥物還得把每一克價值往上推。 ## 01｜SpaceX 把軌道變成可以反覆往返的實驗室 SpaceX 在這條產業鏈的位置很清楚：提供往返低地球軌道的交通。 NASA 的商業補給計畫讓 Dragon 可以把科學設備送進國際太空站，再把需要分析的樣本帶回地面。對製藥研究而言，「回得來」和「上得去」同樣重要。晶體、細胞與組織樣本若不能迅速回收、維持溫控並交給實驗室分析，太空中的結果就很難進入下一輪研發。 Redwire 的 PIL-BOX 正是在這套物流上長出來的產品。它是一組可在國際太空站遠端啟動的微型結晶設備，藥廠先把候選分子裝進去，設備到站後進行結晶，再跟著 Dragon 回到地球。 Redwire 2026 年 7 月的官方新聞稿稱，自 2023 年 11 月首飛以來，已有 54 套 PIL-BOX 前往國際太空站，完成 45 種化合物的結晶，合作範圍涵蓋胰島素，以及癌症、心血管疾病、肥胖與糖尿病相關分子。這些數字屬於公司披露，距離藥品核准與商業量產仍有很長一段路；它至少證明了一件事：太空結晶已能以同一套硬體反覆執行，從多年等一次的特殊任務走向固定迭代。［[Redwire／SpaceMD](https://ir.rdw.com/news-events/press-releases/detail/235/former-merck-and-nasa-leaders-bring-strategic-expertise-to)］ SpaceX 帶起的太空熱潮，對製藥業最實際的影響也在這裡。火箭本身不負責發現新藥，固定而可回收的運輸能力，卻把四百公里外的微重力變成一項可以預約、測試、失敗後再來一次的研發工具。 ## 02｜Keytruda 的太空旅程，留下了一堂製劑物理課 微重力為什麼可能改變藥物？答案藏在一顆晶體成長時，看不見的水流裡。 地球上的液體受到重力影響，會產生對流與沉降。蛋白質分子朝晶體表面移動時，濃度分布容易被攪動；晶體也可能在尚未長好前沉到容器底部，彼此碰撞、聚集，最後留下大小不一或帶有缺陷的結構。 在低地球軌道，分子主要靠擴散移動。周圍安靜下來後，晶體有機會按照較穩定的節奏排列。這項優勢並不會平均落在所有分子上，有些實驗得到較大晶體，有些只改善缺陷或粒徑分布，也有分子看不出足以改變製程的差異。 Merck 的 pembrolizumab，也就是 Keytruda 的有效成分，提供了目前最容易理解的案例。 2017 年 2 月 19 日，研究團隊把 pembrolizumab 結晶實驗搭上 SpaceX CRS-10。設備在太空站的培養櫃中運作 18 天，3 月 19 日再隨 Dragon 返回地球。2019 年發表於《npj Microgravity》的結果顯示，地面對照組形成兩群大小差異很大的晶體，主要分布約在 13 與 102 微米；太空組則集中成平均約 39 微米的單一分布。相同濃度下，太空組晶體懸浮液的黏度也低了約 2 cP，溶解後的抗體結合活性仍與參考品相近。［[npj Microgravity](https://www.nature.com/articles/s41526-019-0090-3)］ 研究團隊後來把太空實驗揭露的兩個關鍵變數帶回地面：沉降與溫度梯度。他們利用緩慢旋轉與溫控，在一般實驗室製作出更均一、可注射的晶體懸浮液。 這段因果關係要說得精準。太空長出的晶體沒有直接裝進後來的藥瓶。NASA 在 2026 年 1 月回顧時，用的是「提供早期洞見」：微重力研究幫助科學家理解適合皮下注射製劑的粒徑與結構。FDA 於 2025 年核准的 Keytruda Qlex，仍經過完整的製劑開發、臨床比較與監管審查。［[NASA 回顧](https://www.nasa.gov/missions/station/iss-research/space-station-research-informs-new-fda-approved-cancer-therapy/)；[FDA](https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-pembrolizumab-and-berahyaluronidase-alfa-pmph-subcutaneous-injection)］ 太空不會替藥物完成臨床，它只可能替某些分子解開地球上的物理限制。 ## 03｜同一片軌道，長出了三種生意 SpaceMD 賣的是晶種。它把少量候選分子送上軌道，培養出晶種後出售或授權給藥廠；新的晶型、配方與量產製程，回到地面繼續完成。 Varda Space Industries 往前多走了一步。它建造可返回地球的無人太空艙，讓小分子在軌道內加熱、冷卻與結晶。Varda 官方任務頁截至 2026 年 7 月已列出 W-1 到 W-6 六次返回任務。 W-1 做的是 HIV 藥物 ritonavir。研究團隊在三個各約 150 毫克的小瓶中，把穩定的 Form II 加熱後轉成溶解度較高、也更容易轉回其他晶型的 Form III。2026 年發表的同儕審查論文顯示，Form III 經過在軌處理、長時間飛行與再入大氣後仍能回收，化學分析也沒有發現太空環境額外帶來的特殊雜質。［[Varda 任務頁](https://www.varda.com/platform)；[npj Microgravity](https://www.nature.com/articles/s41526-026-00594-0)］ 這個結果更像一張物流與製程的驗收單：高價值小分子可以在軌道上處理，並以可分析的狀態回到地面。 英國押的是規則與基礎設施。2026 年 3 月，英國航天局與藥品監管機構 MHRA 公布太空製藥支持方案，內容包括監管指南、案例研究、再入大氣監管沙盒與供應鏈協作。英國同時資助 BioOrbit 進行生物藥太空結晶的規模化研究，希望把高濃度生物藥推向更方便的居家給藥。 截至 2026 年 7 月 19 日，這條線又多了兩個可以追蹤的制度節點。MHRA 表示，預計在 2026 年秋季發布端到端的在軌藥品製造監管路線圖；英國航天局 7 月 14 日公開的 2025–26 年報則確認，BioOrbit、OrbiSky 與 Space Forge 的三項在軌製造可行性研究合約已於 2 月授出。這仍是監管準備與可行性研究，不等於已有太空製造藥品取得核准。 三條路都得回到地球結帳。軌道上多出來的品質，究竟能否換成新的專利、較好的給藥方式、較穩定的產品，或更省成本的製程？藥廠會用下一張採購單回答。 ## 04｜太空附加值，要過四道商業關卡 科幻感很容易帶來流量，藥物開發仍然只認證據。 ### 第一關：分子選擇 晶型問題明確、劑量高、溶解度差、黏度高或地面製程反覆失敗的分子，才有機會把微重力優勢放大。普通分子若在地面就能順利結晶，送上太空只會多一張昂貴的旅費帳單。 ### 第二關：往返與時程 樣本要經歷發射震動、輻射、溫度變化、在軌等待與再入大氣。任何一次延誤，都可能改變原本設計的結晶條件。Varda 對 ritonavir 的研究之所以重要，正因為它開始把「能不能完整帶回來」變成可量測的資料。 ### 第三關：地面放大 少量晶體在太空表現漂亮，仍要回到公斤級、噸級製程。最理想的模式是太空找到晶型或晶種，地面接手量產。若每一批商業產品都要往返軌道，只有單位價值極高、體積極小的產品能承擔成本。 ### 第四關：藥品監管 監管者會追問原料來源、設備校正、製程偏差、無菌控制、樣本交接、再入大氣後的品質，以及軌道步驟在整個藥品生命週期中的責任歸屬。英國搶先做監管沙盒，目標就是在企業送出正式申請前，把這些問題逐一攤開。 發射次數只能證明設備很忙。更值得追的是同一藥廠有沒有再次付費、回收晶型能否授權、地面製程能否放大。前三件事真的發生後，再看監管機構是否接受這段資料進入藥證。 ## 05｜台灣把蛋白質結晶實驗送上了太空 台灣在這張地圖上已留下過一個很具體的座標。 2021 年 12 月，國家同步輻射研究中心（NSRRC）把由大腸桿菌生成的病毒樣顆粒送上國際太空站。任務由台灣的陽翼先進科技（HelioX Cosmos）與日本 Space BD 協調，搭乘 SpaceX CRS-24。樣本在微重力環境中結晶，並於 2022 年 1 月 24 日安全返回。［[Space BD](https://www.space-bd.com/dam/include-content/PUBLIC/en/news_archive/storage/news20220303-001/pdf/20220520en-4.pdf)］ Space BD 當時的官方資料寫得很保守：回收後將以 X 光分析，研究病毒樣顆粒的結構與可能的感染機制。現有公告沒有揭露這次分析的最終結果，因此本文只確認任務、在軌結晶與樣本回收已完成。 這趟任務仍讓台灣的兩項能力變得很具體。陽翼負責太空任務整合與國際協作；NSRRC 擁有蛋白質晶體學光束線，能承接後續 X 光繞射與結構分析。陽翼官網目前也把「微重力生命科學」與「在軌服務」列為業務之一。 陽翼是未上市公司，這條線現階段更適合用產業能力來觀察。台灣若要把單次實驗做成生意，可以追三件事：是否出現藥廠付費的重複任務、是否能把太空晶體接回地面藥物設計流程，以及任務整合與晶體解析能否形成可複製的服務。 台灣的優勢集中在結構生物學、精密儀器、晶體解析與國際專案整合，恰好位在太空樣本回到地球後最需要的那一段。 ## 06｜2030 前後，軌道實驗空間也會被重新定價 NASA 目前計畫讓國際太空站運作至 2030 年，同時推進下一階段商業太空站計畫，希望新平台能接續低軌研究。交接若出現空檔，微重力實驗就可能面臨艙位、電力、太空人時間與返回班次的供給壓力。 中國的路線則以國家科研平台推進。2025 年 7 月，中國科學院上海藥物研究所把核酸脂質奈米載體研究送入中國太空站，研究目標是觀察微重力下核酸藥物在細胞內的轉運規律。官方後續披露已完成在軌給藥、顯微成像與樣品固定，樣品轉入低溫儲存等待返地分析；截至 2026 年 7 月 19 日，仍未查到完整正式論文或可支持臨床療效比較的公開結果。 藥廠往後買的會是一整套服務：穩定艙位、標準化設備、快速返回、地面分析與監管路徑。少掉其中一環，太空任務就很難被編進正常研發預算。 這也解釋了 SpaceMD 為何同時找來 Merck 的結晶科學家與 NASA 的技術投資主管。微重力科學已經做了數十年，現在缺的是把一次成功，變成可以反覆採購的產品。 ## 結語｜下一個製劑答案，可能先繞一趟太空 太空製藥最迷人的地方，是它把一個熟悉的問題換了環境。地球上總被重力干擾的分子，到了軌道上，可能露出另一種晶型、另一種排列方式，甚至另一條製劑路徑。 漂亮晶體回到地面後，先過三關：重複、放大、監管。少一關，都只是好看的實驗結果。真正的終點，仍是讓病人獲得更方便或更有效的治療。 SpaceX 讓往返太空逐漸有了班次。製藥業接下來要證明的，是這趟旅程能留下多少可計價的科學成果。 最先上軌道的，會是一顆晶種、一小瓶候選分子，以及一個在地球上卡了很久的問題。 ## 免責聲明 本文僅供生技醫藥產業資訊與研究交流，不構成醫療建議、投資建議、證券買賣建議或任何形式的報酬保證。藥物研發、在軌製程與商業化仍具有高度不確定性，讀者應依官方文件、監管資料與自身風險承受能力獨立判斷。 --- ## When a Biotech Says No to a Sale: Five Tests of Whether It Can Become a Biopharma URL: https://drugnews.com.tw/articles/2026-07-24-biotech-independence-five-engines-en.html Date: 2026-07-24 Language: en Summary: An acquisition is not the only successful outcome for a biotech. Using argenx, Insmed, Incyte, and Revolution Medicines as four operating snapshots, this analysis tests product evidence, cash generation, a second growth curve, global execution, and financing options—and shows what retaining control really costs. > Evidence cut-off: July 15, 2026. This analysis puts product evidence, cash generation, a second growth curve, global execution, and financing options on the same scorecard to examine what it really costs a biotech to remain independent. For years, the market compressed a biotech company's future into two outcomes. One was clinical failure, followed by a rapid collapse in asset value. The other was clinical success, followed by a sale to a large pharmaceutical company. Both outcomes are real. But treating an acquisition as the only successful finish line misses a question that is becoming increasingly important: If a company can create more value by continuing to operate independently than it can capture in a sale today, does it still need to sell? This is not a story about whether a founder has enough courage, nor is it a romantic tale of a small company standing up to a multinational pharmaceutical company. The ability to say no rests on five practical capabilities: product evidence, cash flow, a second growth curve, global commercialization, and access to capital without surrendering control. argenx, Insmed, Revolution Medicines, and Incyte illustrate four different stages. The first three do not represent a single formula for success, and Incyte is not a failed company. Putting them side by side instead reveals the dividing lines that are often missed when an asset begins to turn into an operating company. Drugnews' central view is this: Biotech independence is not the courage to reject an MNC. It is the ability to retain core rights and reinvest the first success into a second one. Without a second growth curve, independence may simply delay a sale. Without capital and commercial capabilities, rejecting an acquisition can turn scientific value into execution risk. ## Test One: Can One Drug Become an Internal Capital Engine? argenx has crossed this threshold. The VYVGART franchise generated $4.2 billion in global net product sales in 2025, up 90% year over year. The company also reached full-year operating profitability for the first time, with $1.1 billion in operating profit. In the first quarter of 2026, VYVGART global net product sales reached another $1.3 billion, an increase of approximately 63% from the prior-year period. [[FY2025](https://argenx.com/news/2026/press-release-3245199); [Q1 2026](https://argenx.com/news/2026/press-release-3289577)] The significance of those figures is not simply that the product is selling well. They show that argenx can use its own product revenue to fund three expensive activities at the same time: expanding the disease and delivery footprint of VYVGART, advancing follow-on molecules such as empasiprubart and adimanebart, and maintaining a cross-market medical, regulatory, and commercial organization. The first product is therefore no longer only an asset waiting to be valued by a buyer. It has become an internal capital-allocation tool. That does not make argenx risk-free. VYVGART remains the source of the company's product revenue. New indications and formulations may extend the growth of the same franchise, but they do not prove that a second independent molecule has achieved commercial validation. The next question for the market is no longer simply how many additional labels VYVGART can secure. It is whether the cash generated by VYVGART can produce the next product outside efgartigimod. That is the difference between having cash flow and having a repeatable platform. ## Test Two: A Second Drug Shows Whether the First Success Was Repeatable Insmed looks more like a replication test. For a long time, Insmed was effectively synonymous with ARIKAYCE. The product generated $433.8 million in global revenue in 2025 and demonstrated that the company could develop and commercialize a therapy in the highly specialized market for Mycobacterium avium complex lung disease. Yet one successful product still leaves the market unable to tell whether the result came from organizational capability or from finding one unusually attractive asset. [[Insmed FY2025](https://investor.insmed.com/2026-02-19-Insmed-Reports-Fourth-Quarter-and-Full-Year-2025-Financial-Results-and-Provides-Business-Update)] BRINSUPRI changes the question. In the first quarter of 2026, BRINSUPRI generated $207.9 million, while ARIKAYCE generated $98.1 million. For full-year 2026, Insmed guided to at least $1.0 billion in BRINSUPRI revenue and $450 million to $470 million in ARIKAYCE revenue. [[Insmed Q1 2026](https://investor.insmed.com/2026-05-07-Insmed-Reports-First-Quarter-2026-Financial-Results-and-Provides-Business-Update)] This is not merely an exercise in adding two revenue streams together. If BRINSUPRI scales as planned, it would show that Insmed's regulatory, payer, medical-affairs, and commercial organizations can support a second and different product. The first success would no longer look like a one-off event. The next layer is TPIP. Insmed was continuing enrollment in the Phase 3 PALM-ILD study in PH-ILD and initiated the Phase 3 PALM-PAH study in PAH in April 2026. A Phase 3 study in PPF was expected to begin by the end of 2026, while a Phase 3 study in IPF was expected to begin in the first half of 2027. The company also maintained an internal objective of filing, on average, one to two investigational new drug applications each year. That creates a three-part structure: ARIKAYCE is the established base, BRINSUPRI is the second commercial validation, and TPIP plus the early-stage pipeline must show whether a third growth curve can emerge. [[Insmed Q1 2026](https://investor.insmed.com/2026-05-07-Insmed-Reports-First-Quarter-2026-Financial-Results-and-Provides-Business-Update)] The path still consumes substantial capital. Insmed recorded a net loss of $163.6 million in the first quarter of 2026. As of the end of March, it held approximately $1.2 billion in cash, cash equivalents, and marketable securities. The rapid growth of a second product increases the value of remaining independent, but it does not mean the organization has achieved risk-free self-funding. [[Insmed Q1 2026 Form 10-Q](https://www.sec.gov/Archives/edgar/data/1104506/000110450626000032/insm-20260331.htm)] A second drug is a birth certificate for a biopharma. It is not a graduation certificate. ## The Incyte Warning: Profitability Does Not Guarantee Belief in the Next Curve Incyte is easy to oversimplify. It is not an operating failure. The company reported $5.1412 billion in total revenue and $1.2867 billion in net income in 2025. Jakafi generated $3.0925 billion in net product revenue and Opzelura generated $678.5 million. In the first quarter of 2026, total revenue reached $1.27 billion, Jakafi net product revenue was $758 million, other hematology and oncology products were growing, and the company had ten Phase 3 trials under way. [[FY2025](https://investor.incyte.com/news-releases/news-release-details/incyte-reports-fourth-quarter-and-full-year-2025-financial); [Q1 2026](https://investor.incyte.com/news-releases/news-release-details/incyte-reports-first-quarter-2026-financial-results-and-provides)] The pressure comes from timing. In its 2025 Form 10-K, Incyte states that it expects Jakafi product sales to begin declining after patent exclusivity expires in 2028 and that new products will need to offset that decline. [[Incyte 2025 Form 10-K](https://www.sec.gov/Archives/edgar/data/879169/000087916926000010/incy-20251231.htm)] That makes Incyte a useful stress test: - Revenue does not mean that the revenue base is sufficiently diversified. - Profitability does not mean that the second growth curve is large enough to cross a patent cliff. - A long list of clinical programs does not guarantee that one of them will become the next product capable of redefining the company. The right question is not whether Incyte is a healthy business today. It is whether its current cash flow can create a replacement curve that is both large enough and early enough to arrive before Jakafi declines. That is the most difficult coming-of-age test for a mature biotech. The first success solves survival. The second proves that the organization can create value again. ## Revolution Medicines: It Did Not Sell the Company, but It Did Sell a Smaller Claim Revolution Medicines carries the highest risk of the four companies and offers the clearest example of how financing can change the control equation. The status needs to be anchored to July 15, 2026. daraxonrasib had not received approval from any regulatory authority, and Revolution Medicines had no revenue from an approved product. The program, however, was no longer just an early clinical story. The Phase 3 RASolute 302 study enrolled 500 patients with previously treated metastatic pancreatic ductal adenocarcinoma. In the intention-to-treat population, median overall survival was 13.2 months with daraxonrasib and 6.7 months with chemotherapy, with an overall-survival hazard ratio of 0.40. Progression-free survival and other primary and key secondary endpoints were also met. [[RASolute 302](https://ir.revmed.com/news-releases/news-release-details/revolution-medicines-announces-asco-plenary-presentation)] As of July 7, 2026, the European Medicines Agency had started a phased review, and the company said the rolling submission of its new drug application to the U.S. Food and Drug Administration was nearing completion. These developments materially reduced the clinical risk around the lead asset. They did not constitute approval, and they did not prove commercial success. [[EMA phased review](https://ir.revmed.com/news-releases/news-release-details/european-medicines-agency-expedites-assessment-revolution)] The financing side had also advanced. Revolution held $1.9 billion in cash, cash equivalents, and marketable securities as of March 31, 2026. An equity and convertible-notes offering in April brought in approximately $2.1 billion in additional net proceeds. At the same time, the company raised its 2026 GAAP operating-expense guidance to $1.7 billion to $1.8 billion and reported a first-quarter net loss of $453.8 million. [[Q1 2026 results](https://ir.revmed.com/news-releases/news-release-details/revolution-medicines-reports-first-quarter-2026-financial/); [Form 10-Q](https://ir.revmed.com/static-files/e7a99744-380a-4fa9-93b6-2fb71793c8ee); [April offering 8-K](https://www.sec.gov/Archives/edgar/data/1628171/000119312526159289/d95226d8k.htm)] This is the real price of independence: the balance sheet is stronger, but the organization has already grown toward the cost structure of a large commercial-stage company. The 2025 transaction with Royalty Pharma makes the trade-off even clearer. The agreement provides up to $2.0 billion in financing. Up to $1.25 billion takes the form of a synthetic royalty on daraxonrasib, while up to $750 million is available through a senior secured term loan. After the Phase 3 result met the contractual conditions, Royalty Pharma triggered the second $250 million payment. Together with the first $250 million funded at signing, the first two tranches total $500 million. [[Funding agreement](https://www.royaltypharma.com/news/royalty-pharma-and-revolution-medicines-enter-into-funding-agreements-for-up-to-2-billion/); [2026-04-13 8-K](https://www.sec.gov/Archives/edgar/data/1802768/000119312526152145/d871473d8k.htm)] In return, Royalty Pharma receives a 15-year, tiered claim based on annual worldwide net sales. Under the rates corresponding to the first two tranches, the royalty is 4.55% on annual net sales from zero to $2 billion, 2.50% from $2 billion to $4 billion, 1.00% from $4 billion to $8 billion, and zero on net sales above $8 billion. The rates would increase if Revolution elects to draw all remaining royalty tranches. [[Royalty Pharma 2026-04-13 8-K](https://www.sec.gov/Archives/edgar/data/1802768/000119312526152145/d871473d8k.htm)] The package also includes up to $750 million in senior secured term loans. The first loan draw is tied to specified FDA approval conditions for daraxonrasib. The interest rate is SOFR plus 5.75%, subject to a 3.5% SOFR floor. Revolution therefore did not sell nothing. It sold a portion of the product's future cash flows for 15 years and accepted the possibility of debt costs. What it retained was corporate control, leadership of global development and commercialization, and the remaining upside in the RAS platform. The structure matters because the capital market now offers alternatives to an MNC acquisition. Equity, convertible debt, synthetic royalties, and milestone-based loans can turn “sell the company because it needs money” into a set of smaller, configurable decisions. Every option has a price. Equity dilutes shareholders. Debt carries interest. A royalty removes part of successful product cash flow for the agreed 15-year period. If the lead program is delayed or commercialization underperforms, the global organization built in advance will amplify the loss. ## The Five Tests of Whether a Biotech Has Earned the Option Not to Sell Putting the four companies on one page produces a more useful framework than asking which one might be acquired. ### 1. Product evidence Does the lead asset only have attractive early data, or has it crossed pivotal trials, regulatory review, and real market adoption? Revolution's Phase 3 result reduced risk, but the product remained unapproved at the evidence cut-off. argenx and Insmed already had commercial readbacks. ### 2. Cash engine Can the first product fund the next round of research, or does it only extend the company's life? argenx had reached full-year operating profitability. Insmed was expanding revenue with a second product but remained loss-making. Revolution continued to rely on the capital markets and structured financing. ### 3. Second growth curve Does the next success come from additional indications for the same molecule, or from a separate molecule? Both can create value, but they offer different levels of evidence that the platform and organization are repeatable. ### 4. Global execution Can the company manage multinational registration, supply, medical affairs, payer access, and commercialization itself? If every critical capability remains outsourced, retaining global rights does not necessarily mean that the company can capture global value. ### 5. Financing options and the capacity to absorb failure Can the company secure sufficient capital without surrendering control? More importantly, if a Phase 3 trial fails or a launch is delayed by a year, can its balance sheet and organization survive? Real independence is not the ability to reject one offer. It is the capacity to endure a bad outcome after saying no. ## What Taiwan's Biotech Sector Should Take Away Is Not “Never Sell” Taiwanese biotech companies do not need to copy argenx, Insmed, or Revolution Medicines. Nor should overseas licensing, co-development, or an acquisition be treated as failure. For companies with smaller capital bases and still-developing global commercial experience, an early license can be a rational allocation of risk. Transferring regional rights to a more capable partner may also move a drug to patients faster. The more useful questions are: - Does a business-development deal merely monetize one asset, or does it buy time for the next one? - Which geographies, indications, manufacturing rights, or co-development rights does the company retain? - After the first product succeeds, has a second independent molecule entered the clinic? - Does “internationalization” mean licensing assets out, or building an organization that can lead global trials and commercialization? - Do financing tools distribute risk, or carve up too much of the future cash flow too early? At the evidence cut-off, we did not have sufficient primary-source support to place a specific Taiwan-listed company directly into the same maturity framework. We therefore do not force a list of Taiwan stock tickers into the article. The framework is better used to examine each company's retained rights, cash runway, and second growth curve than to manufacture a concept-stock basket. ## Conclusion: Being Acquired Is Not a Defeat, and Remaining Independent Is Not a Medal An acquisition can give a product faster access to capital, manufacturing capacity, and global distribution. Independence preserves more long-term upside for the original company, but it also leaves the clinical, regulatory, manufacturing, commercial, and financing risks inside that company. The most accurate dividing line is therefore not whether a biotech sells. It is whether the biotech has a real choice. An acquisition can be the successful realization of an asset's value. Choosing independence requires proof that the company can turn one success into a second. Only when the risk-adjusted value of remaining independent exceeds the acquisition premium—and when the company has the capital, organization, and resilience to absorb one failure—does “not selling” become a strategy rather than a posture. ## Disclaimer This article is an analysis of biotechnology, corporate strategy, and capital structure. It does not constitute medical advice, a diagnosis or treatment recommendation, or investment advice. Investigational products remain subject to clinical, regulatory, manufacturing, and commercialization risks. Readers should make investment decisions independently. --- ## Biotech 不賣，憑什麼？五張門票決定它能否長成 Biopharma URL: https://drugnews.com.tw/articles/2026-07-24-biotech-independence-five-engines.html Date: 2026-07-24 Language: zh-Hant Summary: 被收購不是 Biotech 唯一的成功終點。本文以 argenx、Insmed、Incyte 與 Revolution Medicines 為四個截面，檢驗產品證據、現金引擎、第二曲線、全球執行與融資選項，並拆解保留控制權的真實成本。 > 查證截止：2026-07-15。本文將產品證據、現金引擎、第二曲線、全球執行與融資選項放在同一張表上，拆解 Biotech 保留獨立經營權的真正代價。 過去談 Biotech，市場常把結局壓成兩種： 一種是臨床失敗，資產價值快速歸零；另一種是臨床成功，被大型藥廠買走。 這兩條路都是真實的。但把「被收購」當成成功的唯一終點，會漏看一個正在變重要的問題： 如果公司繼續獨立，能創造的價值已經高於今天一次賣掉的價格，它還需要急著出售嗎？ 這不是創辦人有沒有骨氣，也不是「小公司對抗 MNC」的浪漫故事。真正能讓 Biotech 說不的，是五項很現實的能力：產品證據、現金流、第二條成長曲線、全球商業化，以及不讓渡控制權也能拿到資金的選項。 argenx、Insmed、Revolution Medicines 與 Incyte，剛好展示了四種不同狀態。前三家不是同一套成功模板，Incyte 也不是「失敗公司」；把它們並排，反而能看出一家公司從資產走向企業時，最容易被忽略的分水嶺。 藥時事的主判斷是： Biotech 的獨立，不是拒絕 MNC 的勇氣，而是保留核心權利後，仍有能力把第一個成功再投資成下一個成功。沒有第二曲線，獨立可能只是在延後出售；沒有資金與商業化能力，拒絕收購甚至會把科學價值變成執行風險。 ## 第一關：一款藥能不能從產品，變成資本引擎？ argenx 已經跨過這一關。 VYVGART 系列 2025 年全球產品淨銷售 42 億美元，較前一年成長 90%；公司同年首度達到全年營業獲利，營業利益 11 億美元。到了 2026 年第一季，VYVGART 全球產品淨銷售再達 13 億美元，年增約 63%。［[FY2025](https://argenx.com/news/2026/press-release-3245199)；[Q1 2026](https://argenx.com/news/2026/press-release-3289577)］ 這組數字的重要性，不只是「賣得很好」。它代表 argenx 已經能用自己的產品收入，同時支付三種昂貴工作：擴大 VYVGART 的疾病與給藥布局、推進 empasiprubart、adimanebart 等後續分子，以及維持跨市場的醫學、法規與商業化組織。 換句話說，第一個產品不再只是等待被估價的資產，而開始成為公司內部的資本配置工具。 但 argenx 也沒有因此免除風險。VYVGART 仍是收入主體；多適應症與不同劑型能延長同一平台的成長，卻不等於第二個獨立分子已經完成商業驗證。因此，市場接下來要看的不是 VYVGART 還能多一個標籤，而是 VYVGART 產生的現金，能不能真正養出下一個非 efgartigimod 的產品。 這是「有現金流」與「可重複平台」之間的差別。 ## 第二關：第二款藥，才知道第一次是不是偶然 Insmed 的案例更像一張能力複製測驗。 過去很長一段時間，Insmed 幾乎等同 ARIKAYCE。2025 年 ARIKAYCE 全球收入 4.338 億美元，證明公司能在 MAC 肺病這類高度專業市場完成開發與商業化；但單一產品成功，市場仍很難判斷，那是組織能力，還是剛好遇到一個好資產。［[Insmed FY2025](https://investor.insmed.com/2026-02-19-Insmed-Reports-Fourth-Quarter-and-Full-Year-2025-Financial-Results-and-Provides-Business-Update)］ BRINSUPRI 改變了這個問題。 2026 年第一季，BRINSUPRI 收入 2.079 億美元，ARIKAYCE 為 9,810 萬美元；Insmed 對 2026 年的公司指引，是 BRINSUPRI 至少 10 億美元，ARIKAYCE 4.5 億至 4.7 億美元。［[Insmed Q1 2026](https://investor.insmed.com/2026-05-07-Insmed-Reports-First-Quarter-2026-Financial-Results-and-Provides-Business-Update)］ 這不是單純把兩個產品收入加起來。BRINSUPRI 如果照計畫放量，代表 Insmed 的法規、支付方、醫學事務與商業團隊能服務第二個不同產品，第一次成功不再只是一次性事件。 更下一層是 TPIP。Insmed 正持續招募 PH-ILD 第三期 PALM-ILD，並已在 2026 年 4 月啟動 PAH 第三期 PALM-PAH；PPF 第三期預計在 2026 年底前啟動，IPF 第三期則預計在 2027 年上半年啟動。公司也維持每年平均申報一到兩個 IND 的內部研發目標。這形成三段式結構：ARIKAYCE 是已建立的底座，BRINSUPRI 是第二次商業化驗證，TPIP 與早期管線則負責回答第三條曲線能不能形成。［[Insmed Q1 2026](https://investor.insmed.com/2026-05-07-Insmed-Reports-First-Quarter-2026-Financial-Results-and-Provides-Business-Update)］ 不過，這條路仍然燒錢。Insmed 2026 年第一季淨損 1.636 億美元；截至 3 月底，現金、約當現金與有價證券約 12 億美元。第二款藥的快速成長，確實提高獨立價值，卻不代表組織已進入無風險自我供血。［[Insmed Q1 2026 Form 10-Q](https://www.sec.gov/Archives/edgar/data/1104506/000110450626000032/insm-20260331.htm)］ 第二款藥是出生證明，不是畢業證書。 ## Incyte 的提醒：會賺錢，不等於市場會相信下一次 Incyte 最容易被過度簡化。 它不是失敗案例。2025 年公司總收入 51.412 億美元、淨利 12.867 億美元；Jakafi 淨產品收入 30.925 億美元，Opzelura 6.785 億美元。2026 年第一季總收入 12.7 億美元，Jakafi 7.58 億美元，其他血液與腫瘤產品也在成長，公司同時有 10 項第三期試驗進行中。［[FY2025](https://investor.incyte.com/news-releases/news-release-details/incyte-reports-fourth-quarter-and-full-year-2025-financial)；[Q1 2026](https://investor.incyte.com/news-releases/news-release-details/incyte-reports-first-quarter-2026-financial-results-and-provides)］ 問題是，Incyte 自己在 2025 年 10-K 明確寫出：Jakafi 專利獨占到期後，預期產品銷售將自 2028 年開始下滑，公司必須靠新產品抵銷缺口。［[Incyte 2025 Form 10-K](https://www.sec.gov/Archives/edgar/data/879169/000087916926000010/incy-20251231.htm)］ 這讓 Incyte 成為很好的壓力測試： - 有營收，不代表收入來源已充分分散； - 有獲利，不代表第二曲線足以穿過專利懸崖； - 有很多臨床試驗，也不等於其中一定會出現下一個能重新定義公司的產品。 所以真正要問的，不是「Incyte 活得好不好」，而是現有現金流能否在 Jakafi 下滑前，生出規模足夠、時間也來得及的替代曲線。 這也是所有成熟 Biotech 最難的成人禮：第一個成功解決生存，第二個成功才證明組織能重複創造價值。 ## Revolution：不是沒賣，而是把較小的一部分先賣掉 Revolution Medicines 是四家公司中風險最高、也最能解釋「融資如何改變控制權」的一家。 先更新到 2026 年 7 月 15 日的狀態。 daraxonrasib 還沒有獲任何主管機關核准，Revolution 也還沒有來自核准產品的收入；但它已不再只是早期臨床故事。第三期 RASolute 302 納入 500 位先前接受過治療的轉移性胰臟導管腺癌患者。在整體意向治療族群，daraxonrasib 的中位總生存期為 13.2 個月，化療為 6.7 個月，OS HR 0.40；PFS 與其他主要、關鍵次要終點也達標。［[RASolute 302](https://ir.revmed.com/news-releases/news-release-details/revolution-medicines-announces-asco-plenary-presentation)］ 截至 2026 年 7 月 7 日，EMA 已啟動 phased review；公司並表示，美國 FDA 的 NDA 滾動送件接近完成。這些都大幅降低了核心資產的臨床風險，卻仍不是上市核准，更不是商業化成功。［[EMA phased review](https://ir.revmed.com/news-releases/news-release-details/european-medicines-agency-expedites-assessment-revolution)］ 資本端同樣往前推了一大步。Revolution 在 2026 年 3 月 31 日持有 19 億美元現金、約當現金與有價證券；4 月的股票與可轉債融資又帶來約 21 億美元淨收入。與此同時，公司把 2026 年 GAAP 營運費用指引提高到 17 億至 18 億美元，第一季淨損 4.538 億美元。［[Q1 2026 results](https://ir.revmed.com/news-releases/news-release-details/revolution-medicines-reports-first-quarter-2026-financial/)；[Form 10-Q](https://ir.revmed.com/static-files/e7a99744-380a-4fa9-93b6-2fb71793c8ee)；[April offering 8-K](https://www.sec.gov/Archives/edgar/data/1628171/000119312526159289/d95226d8k.htm)］ 這就是獨立的真實價格：資本更充足，組織成本也已經先長到接近大型商業化公司的尺度。 2025 年與 Royalty Pharma 的交易，更能說明 Revolution 到底做了什麼。 雙方安排最高 20 億美元資金，其中最高 12.5 億美元是 daraxonrasib 的 synthetic royalty，最高 7.5 億美元是有擔保貸款。第三期結果達標後，Royalty Pharma 已觸發第二筆 2.5 億美元；加上簽約時的 2.5 億美元，前兩筆合計 5 億美元。［[Funding agreement](https://www.royaltypharma.com/news/royalty-pharma-and-revolution-medicines-enter-into-funding-agreements-for-up-to-2-billion/)；[2026-04-13 8-K](https://www.sec.gov/Archives/edgar/data/1802768/000119312526152145/d871473d8k.htm)］ 代價是：Royalty Pharma 取得 15 年期、按全球年銷售分級的權利。以目前前兩筆資金對應的級距，年銷售 0–20 億美元部分為 4.55%，20–40 億美元部分為 2.50%，40–80 億美元部分為 1.00%，80 億美元以上為零。若 Revolution 未來選擇把其餘 royalty tranches 全部提領，分潤比例還會上升。［[Royalty Pharma 2026-04-13 8-K](https://www.sec.gov/Archives/edgar/data/1802768/000119312526152145/d871473d8k.htm)］ 另有最高 7.5 億美元的 senior secured term loan；首筆需在 daraxonrasib 符合約定的 FDA 核准條件後提領，利率為 SOFR 加 5.75%，並設 3.5% SOFR 下限。 所以 Revolution 並不是「什麼都沒賣」。它賣的是未來 15 年部分產品現金流，也接受可能的債務成本；保留下來的，是公司控制權、全球開發與商業化主導權，以及 RAS 平台其餘上行空間。 這個結構之所以重要，是因為資本市場開始提供 MNC 以外的選項：股權、可轉債、synthetic royalty、里程碑式貸款，可以把「沒錢就必須賣掉整家公司」改成多個可組合的決策。 但每個選項都有價格。股權會稀釋，債務要付利息，royalty 會在約定的 15 年期間切走一部分成功後的現金流；若核心產品延遲或商業化不如預期，提前搭起的全球組織也會放大虧損。 ## 真正的「不賣資格」，要同時過五關 把四家公司放在一起，可以得到一個比「誰會被收購」更實用的框架。 ### 1. 產品證據 核心資產只是有漂亮的早期數據，還是已跨過關鍵試驗、法規與真實市場？Revolution 的第三期成功降低了風險，但截至查證日仍未核准；argenx 與 Insmed 已有商業化讀回。 ### 2. 現金引擎 第一款產品能否支付下一輪研發，而不是只能延長公司壽命？argenx 已有全年營業獲利；Insmed 正在用第二款產品擴大收入，但仍有淨損；Revolution 主要靠資本市場與結構化融資。 ### 3. 第二曲線 第二個成功來自同一分子的適應症擴張，還是另一個獨立分子？兩者都能創造價值，但對「平台可重複性」的證明強度不同。 ### 4. 全球執行 公司是否能自己完成多國註冊、供應、醫學事務、支付方與商業化？若所有關鍵能力仍外包，保留全球權利不必然等於有能力把全球價值收回來。 ### 5. 融資選項與失敗承受力 公司能不能在不交出控制權的前提下拿到足夠資金？更重要的是，若一次第三期失敗、上市延遲一年，現金與組織還撐不撐得住？真正的獨立，不是能拒絕一份報價，而是能承受拒絕後的壞情境。 ## 對台灣 Biotech，最值得帶走的不是「別賣」 台灣生技公司不需要複製 argenx、Insmed 或 Revolution，也不該把海外授權、共同開發或併購視為失敗。 對資本規模較小、全球商業化經驗仍在累積的公司來說，早期授權可能是最合理的風險分配；把區域權利交給更有能力的夥伴，也可能讓藥更快走到患者面前。 真正值得追問的是： - 一筆 BD 是把單一資產變現，還是替下一個資產買到時間？ - 公司保留哪些地區、適應症、製造或共同開發權？ - 首款產品成功後，有沒有第二個獨立分子進入臨床？ - 所謂「國際化」，是簽出資產，還是逐步建立能主導全球試驗與商業化的組織？ - 融資工具是否把風險分散，還是把未來現金流過早切得太薄？ 截至查證日，本文沒有足夠一級證據可把特定台灣上市櫃公司直接放進同一成熟度框架，因此不硬湊台股代號。這套框架更適合用來讀每一家公司的權利保留、現金跑道與第二曲線，而不是製造一串概念股。 ## 結論：被買不是輸，能不賣也不是勳章 併購可以讓產品更快取得資金、產能與全球通路；獨立則讓原公司保留更大的長期上行，但也把臨床、法規、製造、商業化與融資風險全部留在自己身上。 因此，最準確的分界不是「賣或不賣」，而是公司有沒有選擇權。 被收購，可能是一項資產成功兌現；選擇獨立，則必須證明公司能把一次成功重複成第二次。當繼續獨立的風險調整後價值高於收購溢價，而且公司有錢、有組織、也承受得起一次失敗時，「不賣」才從態度變成策略。 ## 免責聲明 本文為生技醫藥、公司策略與資本結構整理，不構成醫療診斷、治療建議或投資建議。文中候選藥仍可能面臨臨床、法規、製造與商業化風險；投資決策應自行評估。 --- ## Madrigal's MASH Platform Strategy: Rezdiffra as the Cornerstone, siRNA as the Next Growth Engine URL: https://drugnews.com.tw/articles/2026-07-20-madrigal-mash-sirna-platform-en.html Date: 2026-07-20 Language: en Summary: Rezdiffra has opened the MASH market and built its commercialization infrastructure. Madrigal is now adding six Ribo siRNA programs and Arrowhead's ARO-PNPLA3 to create a cornerstone-plus-precision-combination treatment matrix. MASH, once regarded as a drug-development graveyard where global pharmaceutical companies repeatedly failed, is being redefined. The company that truly opened the market was Madrigal Pharmaceuticals, with Rezdiffra (resmetirom). Rezdiffra is a once-daily, oral, liver-directed THR-beta agonist. In March 2024, the FDA granted accelerated approval for adults with noncirrhotic MASH accompanied by moderate to advanced liver fibrosis, or stages F2 to F3. The decision was based on histologic surrogate endpoints; continued approval remains contingent on confirmatory studies verifying and describing clinical benefit. It was the first FDA-approved treatment for MASH. For a vast market that had long relied on lifestyle intervention without an approved drug therapy, Rezdiffra finally created a commercial entry point. The launch moved faster than the market had initially expected. Rezdiffra generated $958.4 million in full-year 2025 net sales. Strictly speaking, that was still just short of $1 billion, but first-quarter 2026 sales reached $311.3 million. Trailing 12-month sales exceeded $1.1 billion, officially taking the product into blockbuster territory. By the end of March 2026, more than 42,250 patients had been treated with Rezdiffra. The significance goes beyond sales. Madrigal is using Rezdiffra to clear a path through MASH. From clinical trials and FDA approval to physician education, patient diagnosis, insurance coverage and commercial execution, the company is building an integrated infrastructure. Once a company has successfully navigated the most difficult path in a new disease market, the usual next step is not to defend a single product. It is to turn that product into the center of a treatment matrix. That is why Madrigal has begun making substantial bets on small interfering RNA, or siRNA. ## 01 | Rezdiffra Is Highly Successful, but It Is Not the End Point for MASH MASH stands for metabolic dysfunction-associated steatohepatitis, formerly known as NASH. It is the inflammatory and fibrotic stage that can develop as metabolic dysfunction-associated steatotic liver disease progresses. This is not simply an accumulation of fat. MASH can cause hepatocellular injury, chronic inflammation and liver fibrosis, and can progress to cirrhosis, liver cancer and liver failure, potentially requiring liver transplantation. Available evidence also indicates that MASH is a leading cause of liver transplantation among women in the United States, the second-leading cause overall in the United States, and one of the fastest-growing indications for liver transplantation in Europe. Rezdiffra's success demonstrates how long the treatment needs of people with MASH had been suppressed. But it is not the final answer. In the phase 3 MAESTRO-NASH trial, Rezdiffra met two key histologic endpoints: - MASH resolution without worsening of fibrosis. - Fibrosis improvement without worsening of MASH. Published results showed that MASH resolution was achieved in approximately 25.9% of patients receiving resmetirom 80 mg and 29.9% of those receiving 100 mg, compared with approximately 9.7% in the placebo group. This was a major clinical breakthrough, but it also means that many patients still cannot achieve the desired histologic improvement with Rezdiffra alone. That is the essence of MASH: it is not a single-pathway disease. Lipotoxicity, insulin resistance, hepatic inflammation, fibrosis, the gut-liver axis, genetic risk and cardiometabolic dysfunction may all contribute. A single drug can open the market, but it is unlikely to address every patient. For Madrigal, Rezdiffra's ideal position is therefore not to be the only drug, but to become the cornerstone therapy for MASH. The analogy is the role of PD-1 therapy in oncology: establish a treatment backbone, then combine it with agents that work through different mechanisms to expand the addressable patient population layer by layer. ## 02 | Why Is Madrigal Betting on siRNA? The major competitive approaches in MASH each offer different advantages and limitations. GLP-1 receptor agonists, including Novo Nordisk's Wegovy (semaglutide) and Eli Lilly's tirzepatide, a dual GLP-1/GIP receptor agonist, can address obesity, diabetes and systemic metabolic dysfunction. In 2025, Wegovy also received FDA accelerated approval for MASH, becoming the first drug in the GLP-1 class approved for this indication. FGF21 analogs, including Akero's efruxifermin and 89bio's pegozafermin, are more directly focused on improving hepatic metabolism, lipid handling and fibrotic signaling. This approach is also attractive in moderate to advanced MASH. Roche's acquisition of 89bio reflected the perceived potential of pegozafermin in MASH and metabolic disease. Madrigal's interest in siRNA follows a different logic. Rather than simply improving the metabolic environment, siRNA uses RNA interference to silence specific disease-causing genes. That is particularly relevant in MASH because disease progression is closely associated with genetic risk. Genes such as PNPLA3, HSD17B13 and MARC1 have been linked to hepatic fat accumulation, inflammation, fibrosis or protective mechanisms. If Rezdiffra modulates the metabolic environment, siRNA is a precision intervention at the genetic level. - One adjusts the liver's metabolic rhythm. - The other switches off a specific disease-driving gene. This is the cornerstone-plus-combination model Madrigal wants to build. ## 03 | Ribo Life Science and Ribocure: Why Did Madrigal License Six siRNA Programs at Once? In February 2026, Madrigal entered into an exclusive global licensing collaboration with Ribo Life Science and its Swedish subsidiary, Ribocure Pharmaceuticals, obtaining rights to develop, manufacture and commercialize six MASH siRNA programs. The transaction included a $60 million upfront payment and up to $4.4 billion if all milestones are achieved, as well as royalties on future sales. This was not an ordinary early-stage licensing deal. The key number is six. MASH is highly heterogeneous. Different patients may be driven by different combinations of genetic risk, metabolic status, fibrosis severity and comorbidities. A single siRNA target may not cover the entire population. By licensing six programs at once, Madrigal is building a MASH siRNA toolbox that can support patient segmentation, combinations and long-term expansion beyond its Rezdiffra cornerstone. The principal attraction of Ribo and Ribocure is their proprietary GalSTAR liver-targeted siRNA platform. The value of this type of GalNAc-conjugated siRNA lies in its ability to direct the drug preferentially to hepatocytes, reduce systemic exposure and potentially support long-acting dosing. Ribocure has also said that the collaboration will use the GalSTAR platform to develop innovative MASH therapies and could eventually extend to bispecific siRNA designs capable of silencing two disease-causing genes at the same time. Equally important, the collaboration progressed quickly. In July 2026, Ribo and Madrigal announced that they had reached the first development-candidate nomination milestone. The liver-targeted siRNA program will now enter IND-enabling studies, the preclinical work required to support a clinical trial application. The partnership has therefore moved beyond the press-release stage and into preparation for clinical development. This matters to Madrigal. The company does not merely need to add pipeline assets. It needs to convert Rezdiffra's commercial lead into a clinical-development advantage for future combination programs as rapidly as possible. ## 04 | Madrigal Did Not Stop with Ribo: Arrowhead Confirms siRNA as a Second Growth Engine If Ribo were Madrigal's only transaction, investors could still interpret it as a diversified early-stage bet. Madrigal's subsequent acquisition of exclusive global rights to ARO-PNPLA3 from Arrowhead Pharmaceuticals makes the strategy much clearer. ARO-PNPLA3 is a clinical-stage siRNA designed to silence PNPLA3 in hepatocytes. Madrigal has noted that the PNPLA3 I148M variant is an important genetic driver of MASH progression and is associated with hepatic fat, inflammation, fibrosis, cirrhosis and liver-cancer risk. The transaction clarifies how the pieces fit together: - Ribo provides a multi-target siRNA matrix that is early stage but expandable. - Arrowhead provides a clinical-stage PNPLA3 asset with a clearly defined genetic segment. Together, they allow Madrigal to evolve Rezdiffra from the first MASH drug into the foundation of a MASH platform company. This is also central to Madrigal's potential progression into a mid-sized biopharma company. The opportunity does not arise simply because one product is selling well. It arises because Madrigal can build second and third layers of growth around that first product. ## 05 | Madrigal's Most Important Asset May Be Its MASH Commercial Infrastructure Many investors look only at Rezdiffra's sales figures. Madrigal's more formidable asset is the MASH commercialization infrastructure it has already built. The difficulty in MASH is not limited to drug discovery. Diagnosis and patient identification are also major barriers. Many patients do not know that they have MASH, and physicians need new workflows for fibrosis assessment, noninvasive testing, referrals and prescribing. By developing and launching Rezdiffra first, Madrigal has educated the market, built a prescribing network, advanced insurance coverage and accumulated real-world patient data. More than 42,250 patients had received Rezdiffra by the first quarter of 2026. That gives Madrigal real-world treatment experience, an established physician-interaction network, patient-segmentation data and payer-access expertise. For partners such as Ribo and Arrowhead, this infrastructure is a powerful multiplier. Future siRNA products will not need to educate the market from zero, explain the concept of MASH to physicians all over again or rebuild a liver-disease prescribing network. If the clinical data hold up, these pipeline assets can connect directly to the market access Rezdiffra has already opened. That is the appeal of a cornerstone drug. Once the first therapy is established, every subsequent combination agent can build on its momentum. ## 06 | Several Taiwan MASH and Liver-Disease Programs Still Merit Attention The first company to watch is TaiwanJ Pharmaceuticals (6549). TaiwanJ has invested in liver disease and MASH for many years. Its corporate history shows that JKB-121 previously received FDA clearance to proceed into a phase 2 NASH trial. JKB-122 also made clinical progress in NAFLD, NASH and autoimmune hepatitis. The company has also disclosed a NASH licensing collaboration involving TJC-265, an autotaxin inhibitor. This is not the same THR-beta approach used by Madrigal, nor is it an siRNA strategy. But TaiwanJ remains one of the relatively few Taiwanese companies with a long-standing focus on liver disease and steatohepatitis. The second company is Sinew Pharma (6634). Sinew's SNP-630 previously received clearance from Taiwan's TFDA to enter a phase 1 clinical trial for the treatment of nonalcoholic steatohepatitis, then known as NASH. Company materials also describe SNP-630 as a second-generation product derived from SNP-610, with patents in the United States and South Africa. At the time, Sinew planned to advance clinical development and pursue international licensing discussions. Neither company is a direct competitor to Rezdiffra, and neither should be forced into a Madrigal comparison. ## Conclusion | MASH Is Not a Single-Drug War; It Is a Battle of Cornerstone Therapies and Combination Matrices Rezdiffra's success has begun to transform Madrigal from a liver-focused biotech into a mid-sized biopharma company. But Madrigal understands that Rezdiffra alone is not enough. MASH is a multi-mechanism disease, and one THR-beta agonist cannot address every patient indefinitely. The outcome will ultimately be decided by the combination matrix. - Rezdiffra is the cornerstone. - GLP-1 therapies address obesity and systemic metabolism. - FGF21 analogs target hepatic metabolism and fibrosis more directly. - siRNA approaches disease-driving pathways such as PNPLA3, HSD17B13 and MARC1 at the genetic level. That is the real significance of Madrigal's alliances with Ribo and Arrowhead. The company is not adding pipeline assets at random. It is building a precision-treatment platform for MASH. Ribo's opportunity also comes from this platform-level position. If its GalSTAR siRNA technology can demonstrate translation, combination potential and long-acting dosing in MASH, its value could move beyond a single licensing transaction and become an important validation of a global siRNA platform. MASH was once a graveyard for drug development, but Rezdiffra has opened the first door. The next battle is no longer about who enters first. It is about who can build the most complete and precise treatment matrix around a cornerstone therapy, covering the widest range of patient subtypes. --- This article is an analysis of publicly available clinical and industry information. It does not constitute medical diagnosis, treatment advice, fundraising advice, securities advice or investment advice. Individual medical decisions should be made by qualified healthcare professionals based on the full clinical context. --- ## 聯手巨頭卡位 MASH：Madrigal 把 Rezdiffra 做成基石藥，小核酸開始接棒 URL: https://drugnews.com.tw/articles/2026-07-20-madrigal-mash-sirna-platform.html Date: 2026-07-20 Language: zh-Hant Summary: Rezdiffra 已把 MASH 市場與商業化基礎設施打開；Madrigal 再以 Ribo 六條 siRNA 管線與 Arrowhead ARO-PNPLA3，建立基石藥加精準聯用矩陣。 MASH 這個曾經讓全球藥廠屢戰屢敗的「藥物墳場」，正在被重新定義。 第一個真正把市場打開的，是 Madrigal Pharmaceuticals 的 Rezdiffra（resmetirom）。Rezdiffra 是一款每日一次口服、肝臟導向的 THR-beta agonist，也就是甲狀腺激素受體 beta 促效劑。2024 年 3 月，FDA 透過加速核准途徑，核准 Rezdiffra 用於成人非肝硬化 MASH，且已伴隨中度至重度肝纖維化，也就是 F2 至 F3 族群。這項核准依據組織學替代終點；持續核准仍取決於確認性研究能否驗證並描述臨床效益。 這是 MASH 領域第一個獲 FDA 核准的藥物，也讓一個長期只有生活型態介入、卻沒有真正藥物治療的巨大市場，終於看見商業化入口。 上市後的速度，比市場原先想像更快。Madrigal 2025 年 Rezdiffra 全年淨銷售額達 9.584 億美元，雖然嚴格來說還差一點才到 10 億美元，但到 2026 年第一季，Rezdiffra 單季銷售已達 3.113 億美元。過去 12 個月銷售額已突破 11 億美元，正式站上重磅藥門檻；截至 2026 年 3 月底，已有超過 42,250 名患者使用 Rezdiffra。 這件事的重要性不只是銷售額。 而是 Madrigal 正在用 Rezdiffra 打通 MASH 這條路。從臨床試驗、FDA 核准、醫師教育、患者診斷、保險給付到商業化落地，Madrigal 正在建立一套完整基礎設施。當一家公司先跑通這條最難的路，接下來它要做的，通常不是守著單一產品，而是把這個產品變成治療矩陣的核心。 這也是為什麼 Madrigal 開始大手筆押注 siRNA，也就是小干擾 RNA。 ## 01｜Rezdiffra 很成功，但它不是 MASH 的終點 MASH，也就是 metabolic dysfunction-associated steatohepatitis，代謝功能障礙相關脂肪性肝炎，過去稱為 NASH。它是代謝功能異常相關脂肪肝病進展後的發炎與纖維化階段。 這不是單純脂肪堆積。它會造成肝細胞損傷、慢性發炎、肝纖維化，進一步走向肝硬化、肝癌、肝衰竭，甚至需要肝移植。相關資料也指出，MASH 是美國女性肝移植的主要原因，也是美國整體肝移植第二大原因，並且是歐洲成長最快的肝移植適應症之一。 Rezdiffra 的成功，證明 MASH 患者的治療需求被壓抑太久。但它不是終極答案。 在 MAESTRO-NASH 三期研究中，Rezdiffra 達到兩個關鍵組織學終點： - MASH resolution without worsening of fibrosis，也就是 MASH 消退且纖維化未惡化。 - Fibrosis improvement without worsening of MASH，也就是纖維化改善且 MASH 未惡化。 研究整理指出，resmetirom 80 mg 與 100 mg 組的 MASH 消退率分別約 25.9% 與 29.9%，安慰劑組約 9.7%。這是非常重要的臨床突破，但也意味著仍有大量患者無法單靠 Rezdiffra 達到理想組織學改善。 這正是 MASH 的本質：它不是單一路徑疾病。 脂肪毒性、胰島素阻抗、肝臟發炎、纖維化、腸肝軸、遺傳風險、心血管代謝異常，都可能參與其中。單一藥物能打開市場，但很難吃下所有病人。 所以，Rezdiffra 對 Madrigal 來說，最理想的位置不是「唯一藥物」，而是 MASH 的基石藥。就像腫瘤領域的 PD-1，先成為治療底盤，再和不同機制藥物聯用，把患者族群一層一層擴出去。 ## 02｜為什麼 Madrigal 看上 siRNA？ MASH 目前幾條主要競爭路線，各有優缺點。 GLP-1 receptor agonist，也就是 GLP-1 受體促效劑，例如 Novo Nordisk 的 Wegovy（semaglutide，司美格魯肽）與 Eli Lilly 的 tirzepatide（GLP-1／GIP 雙重受體促效劑），能處理肥胖、糖尿病與全身代謝問題。2025 年，Wegovy 也獲 FDA 加速核准用於 MASH，成為 GLP-1 類別中第一個獲准進入 MASH 的藥物。 FGF21 analog，也就是 FGF21 類似物，例如 Akero 的 efruxifermin、89bio 的 pegozafermin，則更偏向直接改善肝臟代謝、脂質與纖維化訊號。這條線對中重度 MASH 也有吸引力，Roche 先前收購 89bio，就是看中 pegozafermin 在 MASH 與代謝疾病中的潛力。 但 Madrigal 對 siRNA 的興趣，來自另一種邏輯。 siRNA 不是單純改善代謝環境，而是透過 RNA interference，也就是 RNA 干擾，直接沉默特定致病基因。對 MASH 來說，這特別有意思，因為 MASH 的進展和遺傳風險高度相關。像 PNPLA3、HSD17B13、MARC1 這些基因，都和肝脂肪堆積、發炎、纖維化或疾病保護機制有關。 如果 Rezdiffra 是代謝微環境調節，那 siRNA 就是基因層面的精準干預。 - 一個處理肝臟代謝節奏。 - 一個關掉特定疾病驅動基因。 這正是 Madrigal 想要的「基石＋聯用」模式。 ## 03｜Ribo Life Science 與 Ribocure：Madrigal 為什麼一次買六條 siRNA？ 2026 年 2 月，Madrigal 與 Ribo Life Science（瑞博生物）及其瑞典子公司 Ribocure Pharmaceuticals 達成全球獨家授權合作，取得六個 MASH siRNA 專案的開發、製造與商業化權利。交易首付款為 6,000 萬美元，若里程碑全部達成，總金額最高可達 44 億美元，Ribo 還可獲得未來銷售分成。 這不是一筆普通的早期授權。它的重點在於「六條」。 MASH 是高度異質性疾病。不同患者可能由不同基因風險、代謝狀態、纖維化程度與共病組合驅動。單一 siRNA 靶點未必能覆蓋所有人。Madrigal 一次拿下六個專案，等於是在 Rezdiffra 這個基石藥之外，建立一個可分型、可組合、可長期擴張的 MASH siRNA 工具箱。 Ribo／Ribocure 吸引 Madrigal 的關鍵，在於它具備自有的 GalSTAR liver-targeted siRNA platform，也就是 GalSTAR 肝臟導向 siRNA 平台。 這類 GalNAc 偶聯 siRNA 的價值，在於能把藥物高度導向肝細胞，降低全身暴露，並帶來長效給藥可能。Ribocure 也表示，雙方合作將利用 GalSTAR 平台發展 MASH 創新療法，未來甚至可擴展到雙特異性 siRNA，也就是同時沉默兩個致病基因的設計。 更值得注意的是，合作落地很快。 2026 年 7 月，Ribo 與 Madrigal 宣布達成第一個候選藥物提名里程碑，該肝臟導向 siRNA 專案將進入 IND-enabling studies，也就是支持臨床試驗申請的臨床前研究。這代表合作不是停留在新聞稿，而是已經進入臨床前申報準備。 這對 Madrigal 很重要。它需要的不只是補管線，而是快速把 Rezdiffra 的商業化領先優勢，轉化成聯用管線的臨床開發優勢。 ## 04｜Madrigal 不只找 Ribo，也找 Arrowhead：siRNA 已成為第二成長曲線 如果只有 Ribo 一筆合作，市場還可以理解成分散押注。 但 Madrigal 後續又從 Arrowhead Pharmaceuticals 取得 ARO-PNPLA3 全球獨家權利，這就代表策略非常明確。 ARO-PNPLA3 是一款臨床階段 siRNA，目標是沉默肝細胞中的 PNPLA3。Madrigal 官方指出，PNPLA3 I148M 是 MASH 進展的重要遺傳因子，與肝脂肪、發炎、纖維化、肝硬化與肝癌風險有關。 這筆合作讓 Madrigal 的策略更清楚： - Ribo 給它多靶點、早期但可擴張的 siRNA 矩陣。 - Arrowhead 給它一條臨床階段、遺傳分型明確的 PNPLA3 資產。 兩者合在一起，Madrigal 正在把 Rezdiffra 從「第一款 MASH 藥」升級成「MASH 平台公司」。 這也是它有機會進階為中型 Biopharma 的關鍵。不是因為單一產品賣得好，而是因為它能圍繞第一款產品，建出第二層、第三層成長曲線。 ## 05｜Madrigal 的最大資產，其實是 MASH 商業化基礎設施 很多人只看 Rezdiffra 的銷售數字。 但 Madrigal 真正可怕的地方，是它已經建立了 MASH 商業化基礎設施。MASH 的難點不只是藥物研發，也包括診斷與患者篩選。許多患者不知道自己有 MASH，醫師也需要重新建立肝臟纖維化評估、非侵入式檢測、轉診與用藥流程。 Madrigal 先做出 Rezdiffra，等於先教育市場、建立處方網絡、推動保險給付，也累積真實世界患者資料。到 2026 年第一季，超過 42,250 名患者使用 Rezdiffra，這代表 Madrigal 手上已經有真實世界治療經驗、醫師互動網絡、患者分型資料與保險准入經驗。 這對 Ribo、Arrowhead 這類合作夥伴而言，是很強的放大器。因為未來 siRNA 不需要自己從零教育市場，不需要重新說服醫師什麼是 MASH，也不需要從頭建立肝病處方網絡。只要臨床資料成立，這些管線可以直接接上 Rezdiffra 已經打開的市場通路。 這就是「基石藥」最迷人的地方。一旦第一款藥站穩，後面每一款聯用藥都能借力。 ## 06｜台灣對 MASH 與肝病仍有幾個觀察點 第一個可以看的是景凱生技（6549）。 景凱長期布局肝病與 MASH。公司沿革顯示，JKB-121 曾獲美國 FDA 核可進行 NASH 二期臨床，JKB-122 也曾在 NAFLD、NASH、自體免疫性肝炎等方向取得臨床進展。公司也有 autotaxin inhibitor，也就是 autotaxin 抑制劑 TJC-265 相關 NASH 授權合作紀錄。 這不是 Madrigal 式的 THR-beta，也不是 siRNA，但它代表台灣少數長期深耕肝病與脂肪肝炎的公司。 第二個是欣耀生醫（6634）。 欣耀的 SNP-630 曾獲 TFDA 同意進行治療非酒精性脂肪肝炎，也就是 NASH 的一期臨床試驗。 公司資料也提到，SNP-630 是 SNP-610 的第二代產品，具備美國與南非專利，當時規劃進行臨床並洽談國際授權。 這兩家公司都不是 Rezdiffra 的直接競爭者，也不該硬拿來類比 Madrigal。 ## 結語｜MASH 不是單藥戰爭，而是基石藥與聯用矩陣的戰爭 Rezdiffra 的成功，讓 Madrigal 從一家專注肝病的 Biotech，開始走向中型 Biopharma。 但 Madrigal 很清楚，單靠 Rezdiffra 還不夠。MASH 是多機制疾病，不可能永遠只靠一款 THR-beta agonist 解決所有患者。未來的勝負，會在聯用矩陣裡決定。 - Rezdiffra 是基石。 - GLP-1 處理肥胖與全身代謝。 - FGF21 analog 瞄準更直接的肝臟代謝與纖維化。 - siRNA 則從基因層面切入 PNPLA3、HSD17B13、MARC1 等疾病驅動通路。 這就是 Madrigal 聯手 Ribo、Arrowhead 的真正意義。它不是隨便補管線，而是在建立一個 MASH 精準治療平台。 Ribo 的機會，也來自這個平台級位置。若它的 GalSTAR siRNA 能在 MASH 中證明可轉化、可聯用、可長效給藥，它的價值就會從單一授權案，升級為全球 siRNA 平台的一次重要驗證。 MASH 曾經是藥物研發的墳場，但現在，Rezdiffra 已經把第一道門打開。 下一場戰爭，不再是誰第一個進場，而是誰能圍繞基石藥，建立最完整、最精準、最能覆蓋不同患者分型的治療矩陣。 --- 本文僅為公開臨床與產業資料整理，不構成醫療診斷、治療建議、募資、證券或投資建議。個別醫療決策應由合格醫療專業人員依完整臨床情境判斷。 --- ## Animal Studies Are Not Disappearing. Drug Development Rules Are Changing. URL: https://drugnews.com.tw/articles/2026-07-19-nams-regulatory-shift-en.html Date: 2026-07-19 Language: en Summary: China's CDE PIONEER program opens a regulatory pathway for new approach methodologies. The real test is whether human-relevant evidence becomes reproducible, validated and reusable. On July 7, 2026, China's Center for Drug Evaluation (CDE) opened a consultation on its program to propel the research and application of New Approach Methodologies, or NAMs. The English title, *Propel Innovation and Operation of NAMs to Enhance Efficiency in R&D*, gives the program its PIONEER acronym. The point is not that drug developers can suddenly abandon animal studies. It is that nonclinical evidence is beginning to be judged more explicitly by its fitness for purpose. Organoids, organ-on-a-chip systems, human-cell assays, in vitro tests, computational models, AI prediction and real-world data may all have a role when they produce evidence that is verifiable, reproducible and interpretable in a defined context of use. That distinction matters. Drugs can look safe or effective in animals and fail in people; some cell, gene, RNA and antibody therapies have no animal model that faithfully captures the human biology. NAMs are not about doing less science. They are about building evidence that is closer to the patient. ## 01 | CDE PIONEER has two doors and concrete use cases The first door is for drug developers that have used, or plan to use, NAMs data to support a registration package. The second is for method developers whose organoid, chip, AI toxicology or in-vitro platform needs to be evaluated jointly with regulators. The doors are not mutually exclusive: a sponsor can bring a product program and later add a method-validation application. CDE has also named the kinds of questions at stake. They include supplementing conventional nonclinical work, mechanism studies, toxicity prediction, proof of concept and efficacy assessment where an appropriate animal model does not exist. The program also contemplates selected antibody toxicology questions, local eye and skin tolerance, pro-arrhythmia risk, pharmacokinetic endpoints, and products such as cell and gene therapies or rare-disease medicines where conventional animal evidence has clear limits. The operating design is consequential. The pilot will recruit for five years without a fixed quota, offers early engagement and follow-up, and can convene expert consultation. For joint validation, CDE and the applicant will agree the validation plan. A method that is validated for a particular context should, in principle, not need to be revalidated from zero every time it is used in that same context. That is the start of a platform business model for NAMs. A useful method is not merely an impressive experiment: it becomes reusable regulatory infrastructure. ## 02 | Regulators are moving, but they are not moving in the same way NAMs have become a competition in regulatory science. The U.S. FDA has taken the most visible policy stance, announcing in 2025 that it would work toward reducing, refining or replacing certain animal-testing expectations for monoclonal antibodies and other drug-development programs. Its roadmap starts with settings in which animal models have limited predictive value and expands from there. The European Medicines Agency has generally emphasized qualification, safe-harbor discussions, voluntary submissions and incremental recognition. That route can look slower, but it is designed to accumulate regulatory trust around specific methods and use cases. The UK's MHRA has added a more operational timetable. The UK roadmap describes targets for moving away from certain animal regulatory tests, while its early-review mechanism gives companies a way to seek non-binding feedback on non-animal evidence before a marketing application. That matters because the largest risk for a NAMs developer is often not whether its technology is elegant, but whether a regulator will accept the evidence when a product decision is on the line. ## 03 | CDE's program is meaningful, but it is a starting point PIONEER gets three important things right. First, it creates a collaboration framework across industry, academia, research institutions and regulators. Organoid and chip validation needs clinical samples, engineering platforms, high-quality data and repeated testing across sites and conditions; no single company can build that evidence base alone. Second, its dual-track structure lets product sponsors and technology developers enter together. That reduces the risk that NAMs remain an academic publication rather than a usable development tool. Third, the principle that a method validated for a defined use case should not require repeated validation is essential. Without it, every program has to reopen the same regulatory argument, and a method can never become a scalable platform. There are still material gaps. The program is an administrative pilot, not yet a legal reform. It does not provide the kind of detailed timetable used in some other jurisdictions. It also leaves the technical threshold to be worked out: sponsors need clarity on context of use, human relevance, model characterization, reproducibility, applicability domain, positive and negative controls, and cross-laboratory consistency. Funding and shared validation capacity are equally important. Who supplies samples? Who pays for multi-center validation? How are data rights and intellectual property handled after validation? A strong technology can still stall at the stage of “promising, but too expensive to validate.” The meaningful milestone will not be the publication of another policy document. It will be the first product whose NAMs evidence demonstrably helps move through CDE review or supports a filing. ## 04 | The value chain will be repriced around validated, reusable evidence Organoid and organ-on-a-chip companies are the most direct beneficiaries because their historic bottleneck has been an unclear regulatory adoption path. China has large clinical-sample resources, a substantial hospital system and strong translational demand. If disease models, liver toxicity, cardiac toxicity or immune-toxicity use cases can be validated, adoption could accelerate. Cell and gene therapy programs may benefit especially where species specificity makes conventional models poor proxies. AI drug-discovery companies also gain an opening, but not a shortcut: regulators will still ask about data provenance, reproducibility, interpretability, external validation, limits of use and failures. A black-box prediction will not become filing-grade evidence by itself. Conventional CROs face pressure, but CROs that can combine GLP animal work with organoids, chips, AI toxicology and regulatory strategy may become more valuable. The industry is not simply replacing one test with another. It is changing the capabilities required to assemble a credible nonclinical package. ## Conclusion | NAMs are about bringing drug development closer to people CDE's PIONEER program is an important move, but it does not mean animal studies will vanish tomorrow. In the near term, NAMs will often be supporting evidence alongside conventional nonclinical work, and some programs may initially carry the cost of both systems. The opportunity is greatest where human biology is difficult to model in animals: cell and gene therapies, rare diseases and human-specific targets. The lasting change will come only when NAMs data can travel a full path from a validated method, to a defined regulatory use case, to a product decision. That is when the field moves from “animal-to-human extrapolation” toward evidence designed around people from the start. --- ## 動物實驗不是消失，而是藥物研發規則開始換代 URL: https://drugnews.com.tw/articles/2026-07-19-nams-regulatory-shift.html Date: 2026-07-19 Language: zh-Hant Summary: CDE 的 NAMs 先鋒計畫為藥企與方法開發者開出雙軌入口；真正的考題是可重複、可驗證且被監管接受的人體相關證據。 2026 年 7 月 7 日，中國國家藥品監督管理局藥品審評中心，也就是 CDE，一次公開徵求「推進新方法（NAMs）研究和應用試點計畫」相關文件意見。 這項計畫英文全名是 Propel Innovation and Operation of NAMs to Enhance Efficiency in R&D，簡稱 PIONEER，中文則被稱為「先鋒計畫」。 CDE 同步釋出介紹、申報指南、申請表與實施框架，徵求意見期為一個月。 官方文件也明確指出，計畫目標是促進 organoids（類器官）、organ-on-a-chip（器官晶片）等 New Approach Methodologies（NAMs，新方法學）在藥物研發中的研究與應用。 這件事的意義，不只是多了一個試點計畫，它代表中國藥品監管正式進入全球 NAMs 競賽。NAMs 說白了，就是藥物安全性與有效性評估不再只能依賴動物模型。類器官、器官晶片、人體細胞模型、體外檢測、in silico 模型、AI 預測、真實世界資料，只要能在明確使用情境下提供可驗證、可重複、可解釋的證據，都可能成為非臨床評估的一部分。這不是「反動物實驗」的口號。而是藥物研發正在面對一個老問題：動物不是人。 很多藥在動物身上看起來安全有效，進入人體卻失敗；也有些創新療法本來就找不到合適動物模型。當細胞治療、基因治療、RNA 藥物、抗體藥物與 AI 藥物越來越多，傳統動物實驗已經無法完全支撐新一代藥物開發。NAMs 的核心，不是少做實驗，而是做更接近人體的實驗。 【01｜CDE 先鋒計畫怎麼玩？兩個入口，八個場景】 這次 CDE 的設計有兩個入口。 第一個入口，是藥企。如果藥物開發中已經使用，或計畫使用 NAMs 資料支持註冊申報，可以申請納入試點。 第二個入口，是 NAMs 方法開發者。如果類器官、器官晶片、AI 毒性模型、體外檢測平台等方法需要和監管端共同驗證，也可以申請進入試點。 這兩個入口不互斥。換句話說，藥企在開發過程中如果發現某個 NAMs 方法需要聯合驗證，也可以再補充申請。更重要的是，CDE 不是泛泛說「鼓勵創新」，而是列出具體使用方向。包括傳統非臨床評價的補充、機制研究、毒性預測、概念驗證、缺乏合適動物模型時的有效性評估；也包括支援部分抗體類藥物簡化長週期重複給藥毒性試驗，特定毒性終點如眼部與皮膚局部耐受性、致心律不整風險，特定藥代終點，以及細胞與基因治療、罕見病藥物等傳統動物模型不適用的品種。 這些方向很有針對性。因為傳統動物模型最弱的地方，剛好就是 NAMs 最有價值的地方。 例如抗體藥物長期猴子毒性試驗成本高、倫理壓力大，而且動物模型對人類免疫反應的可預測性有限。再例如細胞與基因治療，很多產品只在人類細胞或特定人類靶點上有作用，硬做動物模型反而可能得到不具轉譯價值的資料。CDE 這次也給出幾個重要制度設計。 試點招募期五年，不設名額限制。 納入後強調早期介入、及時溝通、全程跟蹤。 必要時可啟動專家諮詢。 對於方法學聯合驗證項目，CDE 將與申請方共同確定驗證方案；若某一 NAMs 方法通過驗證，後續用於特定場景時，原則上不再重複驗證。 這句話很關鍵，因為它等於給 NAMs 方法一張監管通行證。一旦某個器官晶片模型、類器官毒性模型或 AI 預測模型，在特定場景被驗證，未來其他藥物開發就可能重複使用，而不必每次從零證明。 這才是 NAMs 真正產業化的起點。 【02｜全球監管都在動：FDA 最激進，EMA 最細緻，MHRA 最有時間表】 NAMs 已經是全球監管科學競賽。 美國 FDA 動作最大。 2025 年 4 月，FDA 宣布將逐步減少、精簡或替代單株抗體與其他藥物開發中的動物實驗要求，並鼓勵在 IND 申請中納入 AI 毒性模型、細胞模型、類器官與器官晶片等 NAMs 資料。FDA 也明確說，這套做法將從單株抗體開始，逐步擴展到其他藥物類型。FDA 的打法很清楚：先用政策宣示開路，再用 roadmap 鋪方向，接著用具體指引告訴產業如何做。 FDA 在 NAMs 專頁也提到，2025 年發布的 roadmap 採取分階段策略，先從動物模型預測性特別有限的單株抗體開始，再擴展到其他生物分子與新化學實體；2026 年也發布「General Considerations for the Use of New Approach Methodologies in Drug Development」草案，提供 NAMs 在藥物開發中替代或減少動物毒理試驗的驗證框架與監管期待。 歐洲 EMA 則走另一條路。 它沒有像 FDA 那樣用強烈政策語言大幅宣示，但在具體方法上更重視 qualification、safe harbor、自願提交與逐步認可。這種做法比較慢，但更適合累積監管信任。 英國 MHRA 則很有執行感。 英國政府 2025 年公布替代動物實驗 roadmap，投入約 7,500 萬英鎊推動替代方法發展，並提出明確時間表： 2026 年底前停止皮膚與眼部刺激、皮膚致敏等領域的動物監管測試。 2027 年前推動肉毒桿菌毒素小鼠 potency test 的替代。 2030 年前減少狗與非人靈長類動物藥代研究。 更重要的是，MHRA 2026 年提出提前審查機制。企業若使用非動物資料開發藥物，可在上市申請前先提交 Module 4、試驗主持人手冊與至少一項臨床試驗最終報告，取得 MHRA 的非約束性書面意見。這對企業非常重要。因為使用 NAMs 最大風險，不是技術不美，而是不知道監管最後收不收。如果監管能提前審查、提前給意見，企業就敢投入。 【03｜CDE 這一步很重要，但還不是終點】 CDE 先鋒計畫值得肯定，它至少做對三件事。 第一，搭起產、學、研、監的合作框架。 NAMs 不是單一家企業能完成的事。類器官需要臨床樣本，器官晶片需要工程平台，AI 模型需要高品質資料，驗證又需要多中心、多批次、多條件重複。這必須靠產業、學研與監管共同搭建。 第二，採取雙軌制。 藥企可以帶著產品進來，方法開發方也可以帶著技術進來。這能避免 NAMs 只停留在學術發表，也避免藥企缺乏可用工具。 第三，給出「驗證通過後不再重複驗證」的承諾。 這對 NAMs 方法開發者非常重要。否則每個項目都要重新說服監管，方法學就無法形成平台價值。 但 CDE 這套框架也有幾個尚待補齊的地方。 首先，它仍是行政試點，不是法律層級改革。這和 FDA Modernization Act 2.0 後美國在制度層面放寬動物試驗強制要求，穩定性不同。 其次，它還缺少明確時間表。FDA 說先從單株抗體開始，英國說 2026 年底前在哪些毒性評價取消動物試驗。CDE 目前說五年內招募、不設名額，態度是「歡迎你來」，但還不是「我推著你走」。 第三，技術驗證標準還不夠細。企業最想知道的不是表格怎麼填，而是什麼樣的 NAMs 資料算合格。使用場景、人體相關性、模型表徵、可重複性、適用邊界、陽性陰性對照、跨實驗室一致性，這些都需要更細的技術指引。 第四，缺少資金與公共驗證平台。NAMs 驗證很燒錢。誰提供樣本？誰付多中心驗證成本？驗證成功後智慧財產權與資料使用權怎麼分配？如果沒有公共資助或國家級驗證網路，很多方法會卡在「好技術，但沒錢驗證」的階段。 所以，先鋒計畫是起點，不是答案，真正的里程碑，不是文件發布。而是第一個靠 NAMs 資料真正推進 CDE 審評、甚至支持申報的藥物案例出現。 【04｜誰會受益？類器官、器官晶片、CGT、AI 製藥、CRO 都會被重新定價】 NAMs 最直接受益者，是類器官與器官晶片公司。 過去這些公司最大痛點，是技術看起來前沿，但監管接受路徑不清楚。沒有監管採納，再漂亮的類器官模型也很難成為藥物申報資料的一部分。現在 CDE 把類器官、器官晶片寫進試點計畫，等於給產業一個入口。中國在類器官與器官晶片上有獨特優勢：臨床樣本量大、醫院體系龐大、轉化醫學需求強。如果能先在腫瘤藥敏、肝毒性、心臟毒性、免疫毒性等具體場景跑通，未來可能在某些 NAMs 應用上走得很快。 第二類受益者，是 CGT，也就是細胞與基因治療公司。 CGT 最大問題之一，就是非臨床評價經常找不到合適動物模型。很多基因療法只對人類序列有效，很多細胞治療的免疫互動也難以在動物中重現。NAMs 若能被接受，會讓 CGT 的機制驗證、安全性評估與有效性推論更接近人體。 第三類，是 AI 製藥公司。 CDE 把 computer simulation、AI prediction 這類方法納入 NAMs 範圍，等於承認 AI 可以進入非臨床證據鏈。但 AI 公司不能只說模型很準。監管要看的會是資料來源、模型可重複性、可解釋性、外部驗證、適用邊界與失敗案例。AI 製藥若想真正走進申報資料，未來必須搭配體外實驗、類器官或臨床樣本資料，而不是只做黑箱預測。 第四類，是 CRO，也就是委託研究機構。 傳統動物實驗 CRO 會有壓力，但能同時提供 GLP 動物實驗、類器官模型、器官晶片驗證、AI 毒性模型與法規策略的 CRO，反而會變得更有價值。NAMs 不會讓 CRO 消失，只會讓 CRO 能力結構重新洗牌。 【結語｜NAMs 不是少做動物，而是讓藥物研發更接近人】 CDE 啟動 NAMs 先鋒計畫，是重要一步。但不能把它理解成動物實驗明天就要被全面取代。短期內，NAMs 多半仍是支持性資料，企業仍要同時準備傳統非臨床資料。甚至在過渡期，成本可能上升，因為企業要兩套系統一起做。 但長期看，方向已經很明確。 藥物研發會越來越依賴人體相關模型。 類器官會成為疾病建模與藥效評估的重要工具。 器官晶片會進入毒性、藥代與器官交互作用研究。 AI 模型會協助預測風險，但必須接受嚴格驗證。 CGT、罕見病與人類專一性靶點藥物，會最先感受到 NAMs 的價值。 這不是科學倫理的口號，而是研發效率、轉譯成功率與監管科學的升級。真正值得期待的，不是更多文件。而是第一批用 NAMs 資料跑通審評路徑的藥物。當那一天出現，NAMs 才不只是新名詞。它會成為新藥研發從「動物推人」走向「以人為本」的真正轉折點。 ----------- --- ## When ADCs Enter Autoimmune Disease: Precision Immune Modulation, Not More Killing URL: https://drugnews.com.tw/articles/2026-07-18-autoimmune-adc-lfd200-en.html Date: 2026-07-18 Language: en Summary: The opportunity for autoimmune ADCs is not to transplant oncology cytotoxins into immune cells. It is to deliver powerful glucocorticoids, small molecules, ASOs or siRNAs selectively to the cells that drive disease. LFD-200 offers an early human proof-of-mechanism signal for that thesis. Over the past several years, antibody-drug conjugates, or ADCs, have become almost synonymous with oncology drugs. Mention an ADC and most people think of HER2, TROP2 or Nectin-4. They think of linkers, payloads, bystander effects and the engineering required to deliver a cytotoxic drug more precisely into a tumor cell. Yet if the platform is stripped back to its core, an ADC is not an oncology-only technology. Its fundamental capability is antibody-mediated targeted delivery. Once the platform is understood as a delivery system, its boundaries no longer have to stop at oncology. Autoimmune disease is emerging as the next important testing ground for ADC technology. But an ADC that enters autoimmune medicine cannot simply import oncology's “precision chemotherapy” logic. An oncology ADC is designed to kill malignant cells more thoroughly. An autoimmune ADC must answer a different question: can a potent immunomodulatory effect be delivered more precisely to the immune cells and inflamed tissues that actually require intervention? This is not a simple spillover from oncology. It is a redefinition of what the ADC platform can be: from selective killing to selective immune modulation. ## 01 | Autoimmune Treatment Is Moving From Suppressing Inflammation to Reshaping Immunity Autoimmune ADCs cannot be understood by looking at ADC technology alone. The more important context is that autoimmune treatment itself has entered a new stage. Historically, autoimmune diseases were managed primarily with broad immunosuppression. Glucocorticoids, methotrexate and cyclophosphamide can be effective, but long-term use creates substantial costs, including infection, metabolic abnormalities, osteoporosis and endocrine disruption. Treatment later entered an era of pathway blockade. TNF inhibitors, IL-6 inhibitors, IL-17 inhibitors, JAK inhibitors, BTK inhibitors and FcRn inhibitors made intervention more precise and allowed many patients to avoid chronic dependence on high-dose steroids. Most of these therapies, however, still work by turning down inflammatory signaling. They can control disease without necessarily resetting it. The leading edge of the field now involves cell-selective intervention and, in some cases, immune remodeling. Interest in CD19 CAR-T, B-cell depletion, plasma-cell targeting, T-cell engagers and CD38 antibodies is rising for the same underlying reason: Can treatment do more than suppress inflammation and instead address the immune cells that actually drive the disease? Recent research also suggests that conventional protein-based B-cell-depleting therapies often require continued dosing and rarely produce true drug-free remission. A more durable immune reset may require deeper depletion of B cells residing in tissue. This is the setting in which the ADC opportunity has appeared. It is not an isolated new dosage form. It is part of a broader shift in autoimmune medicine from chronic inflammatory suppression toward cell-selective immune remodeling. ## 02 | Two Paths for Autoimmune ADCs: Deplete the Cell or Modulate It Autoimmune ADCs are not yet a mature therapeutic category, but the technology is already separating into distinct development paths. The first is the cell-depleting ADC. This route most closely resembles an oncology ADC. The antibody recognizes an antigen on a defined immune-cell population, the conjugate is internalized and the released payload eliminates the targeted cell. In principle, this approach could fit autoimmune diseases in which a clearly identifiable pathogenic cell population drives the mechanism. Examples include B cells and plasma cells in autoantibody-mediated disease, as well as some chronic inflammatory states dominated by activated T cells or myeloid cells. Potential targets include CD19, CD22, CD38, BCMA, SLAMF7, CD74 and CD163. The problem is equally clear: immune cells in autoimmune disease are not cancer cells. B cells, plasma cells, T cells and macrophages can contribute to pathology while also carrying essential responsibilities in normal immune defense. The deeper the depletion, the greater the potential for infection, weaker vaccine responses, impaired immune memory and a prolonged immune deficit. The central question for a cell-depleting autoimmune ADC is therefore not whether it can kill. It is whether the therapy can eliminate the cells that drive disease without weakening the normal immune system at the same time. The second path is the immune-modulating ADC. This approach may be even more consequential. Rather than treating cell elimination as the primary objective, an immune-modulating ADC uses targeted delivery to carry a glucocorticoid, an immunomodulatory small molecule, an antisense oligonucleotide, a small interfering RNA or another non-cytotoxic payload into selected immune cells or inflamed tissue. The underlying logic is compelling: Many autoimmune diseases do not require a cell to be deleted. They require its state to be changed. Glucocorticoids offer the clearest example. They can be highly effective in autoimmune and inflammatory disease, but prolonged systemic exposure can cause a broad range of adverse effects. If an antibody can deliver a steroid payload preferentially into the immune cells most relevant to disease, it may preserve anti-inflammatory activity while reducing systemic toxicity. The most imaginative use of an autoimmune ADC may therefore not be more precise cell killing. It may be the more precise use of powerful immune-modulating drugs. ## 03 | LFD-200 Is the First Important Clinical Signal The autoimmune ADC field remains early, and few programs with public clinical data are visible today. The program most worth watching is LFD-200 from Lifordi Immunotherapeutics. LFD-200 is a subcutaneously administered glucocorticoid ADC. Its antibody targets VISTA, or V-domain immunoglobulin suppressor of T-cell activation, on immune cells and carries eight glucocorticoid payloads. The design objective is unusually clear: The antibody is not intended to contribute an additional biological function. Its job is to carry the glucocorticoid payload into VISTA-positive immune cells. That is the core idea of an autoimmune ADC. Instead of exposing the entire body to a steroid, the system attempts to place the drug inside the cells that should be modulated. In preliminary phase 1 data from healthy participants presented at EULAR 2026, single LFD-200 doses up to 12 mg/kg and four weekly doses of 5.5 mg/kg were reported as safe and well tolerated. The study also showed dose-related anti-inflammatory activity without an effect on cortisol levels. Cortisol is one of the sensitive indicators used to assess systemic glucocorticoid toxicity. These data do not establish that LFD-200 is effective in patients with rheumatoid arthritis. They establish something earlier but still important: An ADC can deliver an immunomodulatory payload to immune cells, produce a pharmacodynamic signal in humans and avoid an immediate signal of the cortisol suppression associated with systemic steroids. This is proof of mechanism. It is also the first step from an attractive autoimmune ADC concept toward a clinically testable development thesis. ## 04 | The Strategic Position: Between Biologics and Cell Therapy Autoimmune medicine already has biologics, JAK inhibitors, BTK inhibitors, FcRn inhibitors, CAR-T therapies and T-cell engagers. What position could an ADC occupy? The answer may be the space between conventional biologics and cell therapy. Compared with CAR-T, an ADC is lighter operationally, easier to standardize and more compatible with manufacturing at scale. CAR-T can achieve deep depletion and may create durable drug-free remission. Its costs are high, however, and manufacturing is complex. Treatment may require inpatient management and can involve cytokine release syndrome, immune effector cell-associated neurotoxicity syndrome, infection and prolonged B-cell depletion. Compared with a T-cell engager, an ADC does not have to force T-cell activation and does not depend on the patient's T-cell state. Its theoretical cytokine release syndrome risk may therefore be lower. The risk does not disappear. It shifts to the payload, linker, target-expression profile and release kinetics. Compared with a conventional monoclonal antibody, an ADC may deliver a stronger effect and may localize a drug that would be unsuitable for systemic exposure. The price is a narrower safety window and greater CMC complexity. Compared with a small molecule or systemic steroid, an ADC may reduce systemic exposure, but cost, dosing convenience and long-term safety remain to be demonstrated. The best position for an autoimmune ADC may therefore not be to replace every existing treatment. It may provide a scalable targeted immune intervention in the middle ground where a conventional biologic is not deep enough and cell therapy is too heavy. That is the platform's most credible commercial position. ## 05 | Autoimmune ADC Development Requires a New Rulebook The development logic for an oncology ADC is comparatively direct. Identify an antigen that is highly expressed on the tumor, expressed less in normal tissue and capable of internalization. Attach a potent cytotoxic payload and create an efficacy-toxicity window. Autoimmune disease has no equally clean population of “bad cells.” The same B-cell, T-cell or myeloid-cell population may contain cells that drive disease and cells that maintain normal immune function. Target too broadly and the therapy becomes another form of systemic immunosuppression. Target too narrowly and it may fail to cover the biological heterogeneity of the disease. The true difficulty lies in delivery and release. An immune-modulating ADC does not succeed merely because it enters a cell. It must release the right amount of payload in the right cell, at the right location and at the right time. Release too little and the treatment may be ineffective. Release too much and it may cause systemic immunosuppression. Payload selection must also be redesigned for chronic disease. The potent cytotoxic payloads commonly used in oncology ADCs may be poorly suited to rheumatoid arthritis, systemic lupus erythematosus, inflammatory bowel disease or atopic dermatitis. Autoimmune ADCs are more likely to require glucocorticoids, immunomodulatory small molecules, antisense oligonucleotides, siRNAs or other functional payloads with lower systemic toxicity. The objective is not to kill harder. It is to modulate more precisely. ## The Next Boundary for ADCs May Be Precision Immune Modulation The success of ADCs in oncology has taught the industry to understand the platform as precision chemotherapy. That definition must change if ADCs move into autoimmune disease. The value of an autoimmune ADC is not in transferring a cytotoxic oncology payload from tumor cells to immune cells. It is in using the localization capability of an antibody to deliver powerful immunomodulatory drugs, whose systemic adverse effects would otherwise limit their use, into the cells or tissue environments where modulation is needed. If the defining phrase for an oncology ADC is selective killing, the defining phrase for an autoimmune ADC may be selective modulation. Whether the category succeeds will not be decided by the letters ADC. It will be decided by whether the technology can demonstrate three things: Deliver a powerful immune drug to the correct cells. Reduce unnecessary systemic exposure. Establish sufficiently clear safety and efficacy for long-term treatment of chronic disease. If those conditions are met, the boundary of the ADC platform will no longer be oncology. It could become a new therapeutic form within the next generation of autoimmune treatment, positioned between small molecules, biologics and cell therapies. In retrospect, the importance of bringing ADCs into autoimmune disease may not be that the industry added one more indication. It may be that the move forced us to understand the platform differently: An ADC is not only a weapon for delivering a toxin to a cancer cell. It can also be a navigation system for moving the immune system back toward balance. --- This article synthesizes public clinical and industry information and does not constitute medical advice, a treatment recommendation, fundraising advice or investment advice. Public human data for LFD-200 remain primarily from an early study in healthy participants and cannot establish efficacy in rheumatoid arthritis or any other autoimmune disease. --- ## 當 ADC 走進自體免疫：不是殺得更狠，而是調得更準 URL: https://drugnews.com.tw/articles/2026-07-18-autoimmune-adc-lfd200.html Date: 2026-07-18 Language: zh-Hant Summary: 自體免疫 ADC 的價值不在把腫瘤細胞毒 payload 搬到免疫細胞，而在把 glucocorticoid、ASO、siRNA 等強效免疫調節作用定向送到真正需要干預的細胞；LFD-200 的早期人體訊號，是這條路第一次重要的機制驗證。 過去幾年，ADC（antibody-drug conjugate，抗體藥物複合體）幾乎被等同於腫瘤藥。 一談 ADC，大家想到的是 HER2、TROP2、Nectin-4。想到的是 linker（連接子）、payload（藥物載荷）、bystander effect（旁觀者效應），以及如何把細胞毒藥物更精準地送進腫瘤細胞。 但如果把 ADC 拆回本質，它其實不是「腫瘤專用技術」。 ADC 的真正核心，是抗體介導的定向遞送。既然是定向遞送，它的邊界就不必停在 oncology，也就是腫瘤領域。 自體免疫疾病，正在成為 ADC 技術下一個值得觀察的試驗場。 只是，ADC 一旦進入自體免疫，就不能再照搬腫瘤裡「精準化療」的邏輯。 腫瘤 ADC 追求的是把癌細胞殺得更乾淨；自體免疫 ADC 要回答的問題，則是能否把強效免疫調節作用，更精準地送到真正需要被干預的免疫細胞與發炎組織。 這不是腫瘤 ADC 的簡單外溢。 這是 ADC 技術本質的一次重新定義： 從選擇性殺傷，走向選擇性免疫調節。 ## 01｜自體免疫治療，正在從壓制發炎走向重塑免疫 要理解自體免疫 ADC，不能只從 ADC 談起。 更重要的是，自體免疫治療本身已經走到新階段。 過去，自體免疫疾病主要靠廣譜免疫抑制。Glucocorticoids（糖皮質素）、methotrexate（甲氨蝶呤）、cyclophosphamide（環磷醯胺）這些藥有效，但長期使用會帶來感染、代謝異常、骨質疏鬆、內分泌失衡等問題。 後來，治療進入通路阻斷時代。TNF inhibitor（TNF 抑制劑）、IL-6 inhibitor（IL-6 抑制劑）、IL-17 inhibitor（IL-17 抑制劑）、JAK inhibitor（JAK 抑制劑）、BTK inhibitor（BTK 抑制劑）、FcRn inhibitor（FcRn 抑制劑）等藥物，讓治療更精準，也讓許多患者不必長期依賴高劑量 steroid（類固醇）。 但這些藥物多數仍是在「壓低發炎訊號」。它們可以控制疾病，卻未必能重置疾病。 現在最前沿的方向，是細胞選擇性干預，甚至免疫重塑。 CD19 CAR-T、B 細胞耗竭、漿細胞靶向、T cell engager（T 細胞接合器）、CD38 antibody（CD38 抗體）之所以在自體免疫領域升溫，背後都是同一個問題： 能不能不只是壓低發炎，而是更深層處理那些真正驅動疾病的免疫細胞？ 相關研究也指出，傳統蛋白類 B 細胞耗竭療法通常需要持續給藥，較少帶來真正無藥緩解；若要達到更持久的免疫重置，可能需要更深層清除組織中的 B 細胞。 ADC 的機會，正是在這個背景下出現。 它不是孤立的新劑型，而是自體免疫治療從長期壓制發炎，走向細胞選擇性重塑的一部分。 ## 02｜自體免疫 ADC 有兩條路：清除細胞，或調節細胞 自體免疫 ADC 目前還不是成熟賽道，但技術路線已經開始分化。 第一條路，是「細胞清除型 ADC」。 這條路比較接近腫瘤 ADC。抗體辨識特定免疫細胞表面抗原，ADC 被內吞後釋放 payload，進而清除目標細胞。 理論上，這適合那些疾病機制中有明確致病細胞族群的自體免疫疾病，例如自身抗體驅動疾病中的 B 細胞與漿細胞，或某些由活化 T 細胞、髓系細胞主導的慢性發炎狀態。 潛在靶點包括 CD19、CD22、CD38、BCMA、SLAMF7、CD74、CD163 等。 但這條路的問題也很清楚：自體免疫疾病中的免疫細胞不是癌細胞。 B 細胞、漿細胞、T 細胞、巨噬細胞既可能參與疾病，也承擔正常免疫防禦。 清除越深，越可能帶來感染、疫苗反應下降、免疫記憶受損與長期免疫缺口。 所以細胞清除型 ADC 在自體免疫中的核心問題，不是「能不能殺」。 而是能不能只清除真正驅動疾病的細胞，而不是把正常免疫系統一起削弱。 第二條路，是「免疫調節型 ADC」。 這條路更值得注意。 它不以清除細胞為主要目標，而是把 glucocorticoid（糖皮質素）、免疫調節小分子、ASO（antisense oligonucleotide，反義寡核苷酸）、siRNA（small interfering RNA，小干擾 RNA）或其他非細胞毒 payload，定向送到特定免疫細胞或發炎組織中。 這條路的邏輯很漂亮： 自體免疫疾病很多時候不需要把細胞刪掉，而是需要調整細胞狀態。 Glucocorticoid 就是最典型的例子。它在自體免疫與發炎疾病中非常有效，但全身長期暴露會帶來一連串副作用。 如果可以用抗體把 steroid payload 送進更相關的免疫細胞，也許就能保留抗發炎效果，同時降低全身副作用。 所以，自體免疫 ADC 最有想像力的方向，可能不是更精準地殺細胞。 而是更精準地使用強效免疫調節藥物。 ## 03｜LFD-200：自體免疫 ADC 的第一個重要訊號 目前自體免疫 ADC 還很早期，公開可見的臨床項目不多。 其中最值得看的是 Lifordi Immunotherapeutics 的 LFD-200。 LFD-200 是一款皮下注射的 glucocorticoid ADC，也就是糖皮質素 ADC。 抗體靶向免疫細胞表面蛋白 VISTA（V-domain immunoglobulin suppressor of T cell activation），並連接 8 個 glucocorticoid payload。 它的設計重點很明確： 抗體本身不追求額外生物功能，而是負責把 glucocorticoid payload 送進 VISTA 陽性的免疫細胞。 這正是自體免疫 ADC 的核心精神。 不是把 steroid 丟到全身，而是讓它進到更該被調節的細胞。 在 EULAR 2026 公布的健康受試者 Phase 1 初步資料中，LFD-200 單次劑量最高 12 mg/kg，以及每週 5.5 mg/kg 連續 4 次給藥均安全且耐受性良好。 研究也顯示，LFD-200 具有劑量相關抗發炎活性，且未影響 cortisol，也就是皮質醇水平。 Cortisol 是評估全身 glucocorticoid 毒性的敏感指標之一。 這組資料還不能證明 LFD-200 在類風濕性關節炎患者中有效。 但它先證明了一件很重要的事： ADC 可以把免疫調節 payload 送到免疫細胞，並在人體中看到藥效訊號，同時沒有立刻出現典型全身 steroid 抑制 cortisol 的問題。 這是 proof of mechanism，也就是機制驗證。 也是自體免疫 ADC 從概念進入臨床邏輯的第一步。 ## 04｜自體免疫 ADC 的位置：介於生物製劑與細胞治療之間 如果自體免疫領域已經有 biologics（生物製劑）、JAK inhibitor、BTK inhibitor、FcRn inhibitor、CAR-T、TCE，ADC 憑什麼有位置？ 答案可能是：它介於傳統生物製劑與細胞治療之間。 相較於 CAR-T，ADC 更輕、更容易標準化，也更可能規模化。 CAR-T 的優勢是深度清除，有機會帶來長期無藥緩解；缺點是成本高、製造複雜、需要住院管理，還有 CRS（細胞激素釋放症候群）、ICANS（免疫效應細胞相關神經毒性症候群）、感染與長期 B 細胞耗竭風險。 相較於 T cell engager，ADC 不需要強行啟動 T 細胞，也不依賴患者 T 細胞狀態，因此理論上 CRS 風險可能較低。 但它把風險轉移到 payload、linker、靶點表達與釋放動力學上。 相較於傳統 monoclonal antibody，也就是單株抗體，ADC 可能更強效，也能把原本不適合全身暴露的藥物局部送達，但安全窗口與 CMC 複雜度會更高。 相較於小分子或 steroid，ADC 可能降低全身暴露，但成本、給藥便利性與長期安全性仍需要被證明。 所以自體免疫 ADC 的最佳位置，不一定是取代所有現有療法。 它更可能是在傳統生物製劑不夠深、細胞治療又太重的中間地帶，提供一種可規模化的定向免疫干預方案。 這是它最合理的商業定位。 ## 05｜自體免疫 ADC 不是換適應症，而是換一套開發規則 ADC 在腫瘤領域的開發邏輯相對直接。 找一個腫瘤高表達、正常組織低表達、能內吞的抗原，接上強細胞毒 payload，打開療效與毒性窗口。 但自體免疫疾病沒有那麼清楚的「壞細胞」。 同一群 B 細胞、T 細胞、髓系細胞，可能一部分在致病，一部分在維持正常免疫。 靶點太廣，就會變成另一種系統性免疫抑制； 靶點太窄，又可能覆蓋不了疾病異質性。 真正的難點，在於遞送與釋放。 免疫調節型 ADC 不是只要「送進去」就好，而是要在正確細胞、正確位置、正確時間，釋放正確劑量的 payload。 釋放太少，可能無效。 釋放太多，可能造成全身免疫抑制。 Payload 的選擇也必須重新設計。 腫瘤 ADC 常用的強細胞毒 payload，未必適合 rheumatoid arthritis（類風濕性關節炎）、systemic lupus erythematosus（全身性紅斑性狼瘡）、inflammatory bowel disease（發炎性腸道疾病）、atopic dermatitis（異位性皮膚炎）這些慢性疾病。 自體免疫 ADC 更可能需要 glucocorticoid payload、免疫調節小分子、ASO、siRNA，或其他低系統毒性的功能性 payload。 它追求的不是殺得更狠。 而是調得更準。 ## ADC 的下一個邊界，可能不是更強殺傷，而是更準免疫調節 ADC 在腫瘤領域的成功，讓產業習慣把它理解成精準化療。 但如果 ADC 進入自體免疫疾病，這個定義需要改寫。 自體免疫 ADC 真正有價值的地方，不是把細胞毒 payload 從腫瘤細胞轉移到免疫細胞。 而是利用抗體的定位能力，把強效但副作用大的免疫調節藥物，送到更應該被調節的細胞或組織環境中。 如果 oncology ADC 的關鍵字是選擇性殺傷，那 autoimmune ADC 的關鍵字更可能是選擇性調節。 最後決定這條路能不能成立的，不是 ADC 這三個字，而是它能不能證明三件事： 把強效免疫藥物送到正確細胞。 減少不必要的全身暴露。 在慢性病長期治療中建立足夠清楚的安全性與療效證據。 如果答案成立，ADC 的邊界就不再只是 oncology。 它可能成為下一代自體免疫治療中，介於小分子、生物製劑與細胞治療之間的一種新技術形態。 也許未來回頭看，ADC 進入自體免疫的真正意義，不是多開了一個適應症。 而是讓我們重新理解 ADC： 它不只是把毒藥送到癌細胞的武器。 它也可以是把免疫系統調回正軌的導航系統。 --- 本文僅為公開臨床與產業資料整理，不構成醫療診斷、治療建議、募資或投資建議。LFD-200 的公開人體資料目前主要來自健康受試者早期試驗，不能據此推定其在類風濕性關節炎或其他自體免疫疾病中的療效。 --- ## 又一次RevMed 直接掀桌，胰臟癌進入 RAS 組合療法時代 URL: https://drugnews.com.tw/articles/2026-07-17-revmed-ras.html Date: 2026-07-17 Language: zh-Hant Summary: Daraxonrasib 已用第三期總生存期資料抬高胰臟癌 RAS 賽道門檻；zoldonrasib 與 pan-RAS 或化療的早期組合訊號，則把競爭從單藥推向治療生態系。 胰臟癌這個「癌王」，可能真的要被 RAS 藥物撬開一道縫了。 過去，胰臟癌最可怕的地方，不只是死亡率高，而是長期缺乏真正能改寫治療路徑的標靶藥。 KRAS 突變明明出現在超過九成胰臟導管腺癌（PDAC）患者身上，卻因為蛋白結構太滑、缺乏好口袋，長期被視為「不可成藥」。 直到 Revolution Medicines，也就是 RevMed，把 RAS(ON) 抑制劑推到三期陽性。 它的核心藥物 daraxonrasib（RMC-6236），是一款口服 RAS(ON) multi-selective inhibitor，也就是 RAS 活化狀態多選擇性抑制劑。在先前 RASolute-302 三期試驗中，daraxonrasib 用於既往治療過的轉移性 PDAC，將中位整體存活期拉到 13.2 個月，對照化療約 6.7 個月，死亡風險顯著下降，幾乎把後線胰臟癌治療標準直接重寫。 但 RevMed 想要的，顯然不只是二線胰臟癌。 7 月 2 日，RevMed 在 2026 ESMO GI 公布另一組更有想像力的資料：KRAS G12D 抑制劑 zoldonrasib（RMC-9805）聯合 daraxonrasib，在 RAS G12D 轉移性胰臟癌患者中展現強勁訊號。 二線患者中，zoldonrasib 加 daraxonrasib 的 ORR，也就是客觀緩解率，達 50%；DCR，也就是疾病控制率，達 97%；中位 PFS，也就是無惡化存活期，達 9.6 個月。三線以上患者 ORR 也有 47%，DCR 達 90%。同時，在一線患者中，zoldonrasib 聯合標準化療也做出高反應率： 搭配 modified FOLFIRINOX 的 ORR 達 82%。 搭配 gemcitabine plus nab-paclitaxel，也就是吉西他濱加白蛋白結合型紫杉醇的 ORR 達 61%。 這組資料真正重要的地方，不是又多一款 KRAS G12D 藥物。而是 RevMed 正在用 RAS 矩陣管線，試圖把胰臟癌從一線到後線全部吃下來。 ## 01｜為什麼胰臟癌這麼難？KRAS 是關鍵開關 要理解這組聯合療法，必須先理解 KRAS。 KRAS 在正常細胞裡像一個訊號開關。當外部有生長訊號時，KRAS 結合 GTP，進入活化狀態，通知細胞增殖；當訊號結束後，GTP 被水解成 GDP，KRAS 回到關閉狀態。這個「開—關」循環，原本非常精密。問題是，KRAS G12D 突變會讓 KRAS 蛋白結構改變，使它難以正常關閉。結果是 KRAS 長期卡在活化狀態，不斷向下游送出兩條訊號： 一條透過 RAF-MEK-ERK 告訴細胞繼續增殖。 另一條透過 PI3K-AKT 告訴細胞不要死亡。 這也是為什麼 KRAS 突變癌細胞很難打。你不是在處理一個單點訊號，而是在處理一個被鎖死的生長引擎。在胰臟癌中，KRAS G12D 又是最常見的突變亞型之一。 RevMed 公告指出，PDAC 是主要癌種中最典型的 RAS-driven malignancy，也就是 RAS 驅動惡性腫瘤，超過 90% 患者帶有 RAS 突變，其中 RAS G12D 約占 40%。所以，誰能真正壓住 G12D，誰就有機會改寫胰臟癌治療。 ## 02｜Zoldonrasib 加 Daraxonrasib：不是 1+1，而是「深度＋廣度」 這次最值得玩味的，不是 zoldonrasib 單藥，而是它和 daraxonrasib 的組合。 Zoldonrasib 是口服 RAS(ON) G12D-selective covalent inhibitor，也就是 RAS 活化狀態 G12D 選擇性共價抑制劑，主打深度。它用共價方式鎖住 KRAS G12D，針對最大亞型做高強度壓制。 Daraxonrasib 則是 RAS(ON) multi-selective inhibitor，主打廣度。 它不是只打 G12D，而是覆蓋多種活化 RAS 訊號，包括 G12D、G12V、G12R，以及部分癌細胞繞路後重新啟動的 RAS 訊號。兩者聯用的邏輯，不是簡單把兩個藥加在一起，而是在 RAS 通路上做雙重封鎖。 第一層，是縱向深度。Zoldonrasib 對 KRAS G12D 做高強度、持久性壓制，把最主要的驅動突變先封住。 第二層，是橫向廣度。Daraxonrasib 則負責堵住癌細胞的逃生路。癌細胞若透過 NRAS、HRAS 旁路，或 HER2、MET 等上游受體重新啟動野生型 RAS 訊號，daraxonrasib 仍有機會把這些活化 RAS 訊號一起攔下來。 這就是為什麼這組聯合療法有意思。 KRAS G12D 單藥面對的最大問題，是癌細胞會逃。它可以擴增 KRAS G12D 拷貝數，製造更多靶蛋白來稀釋抑制劑效果；也可以啟動其他 RAS 家族成員，繞過 G12D；還可以透過上游 RTK 訊號重建下游 ERK、AKT 活性。 Zoldonrasib 解決的是「主幹突變」。 Daraxonrasib 解決的是「旁路反撲」。 這就是 RAS 雙封鎖的精髓。 ## 03｜資料真的強，但還不能忘記它仍是早期樣本 從療效來看，zoldonrasib 加 daraxonrasib 的訊號相當亮眼。 二線 RAS G12D 轉移性 PDAC 患者 ORR 50%、DCR 97%、中位 PFS 9.6 個月； 三線以上 ORR 47%、DCR 90%、中位 PFS 7.6 個月。 對照傳統後線胰臟癌治療，這組數字非常有衝擊力。更重要的是，這不是挑最漂亮的患者族群。RevMed 公布資料顯示，這些患者多數已有後線治療經驗，且相當高比例存在肝轉移。胰臟癌肝轉移通常代表預後更差、進展更快、化療滲透更困難，能在這樣的人群中打出接近五成反應率，確實不容易。安全性方面，zoldonrasib 加 daraxonrasib 的 3 級以上治療相關不良事件發生率為 35%。常見 3 級以上事件包括： 皮疹 12%。 貧血 10%。 口腔炎／黏膜炎 7%。 因不良事件停用 zoldonrasib 與 daraxonrasib 的比例分別為 2% 與 5%。這組安全性比市場原先想像更好。但也必須保持冷靜。 第一，樣本數仍不大。二線 30 人、三線以上 30 人，訊號很強，但還不能替代三期結論。 第二，daraxonrasib 劑量強度只有 76%。這代表臨床上可能仍需要在療效與皮疹、口腔炎等毒性之間找平衡。 第三，真正決定這個組合價值的，仍是後續 OS、PFS 長期追蹤與三期驗證。OS 是整體存活期，PFS 是無惡化存活期，這些才是胰臟癌治療能否真正改寫標準的關鍵。 所以，這不是可以直接宣告勝利的資料。但它足以宣告一件事： RAS 治療不再是單藥故事，而是進入組合療法時代。 ## 04｜RevMed 正在做的，是把 RAS 做成一個 franchise RevMed 最可怕的地方，不只是有 daraxonrasib。而是它正在把 RAS 做成一套產品矩陣。 Daraxonrasib 負責廣譜 RAS(ON) 抑制，是平台骨架。 Zoldonrasib 負責 KRAS G12D，是胰臟癌最大突變族群。 Elironrasib（RMC-6291）負責 G12C。 RMC-5127 則瞄準 G12V。 後面還有 Q61H、G13C 等更細分突變亞型。 這種布局的意義在於，RevMed 不是只想做一款胰臟癌藥，而是想重建 RAS-driven cancers，也就是 RAS 驅動癌症的治療版圖。從二線胰臟癌，到一線胰臟癌，再到肺癌、結直腸癌與其他 RAS 突變實體瘤，它正在把不同 RAS 分型一塊一塊接起來。這也會改變整個 KRAS G12D 賽道的估值邏輯。 過去市場看 G12D，常把它當成單一突變抑制劑故事。但現在，daraxonrasib 成功後，G12D 抑制劑的價值不只是單藥，而是能不能成為 RAS backbone，也就是 RAS 治療骨架上的加成模組。換句話說，G12D 不再只是自己打。而是要看它能不能跟 pan-RAS、化療、PRMT5 inhibitor、MAT2A inhibitor、EGFR inhibitor 或免疫療法一起打。 也就是： G12D 能不能搭 pan-RAS？ 能不能搭化療？ 能不能搭 PRMT5 抑制劑？ 能不能搭 MAT2A 抑制劑？ 能不能搭 EGFR 抑制劑？ 能不能搭免疫療法？ 這就是 RevMed 掀桌的地方。它把一條原本單點競爭的賽道，改造成矩陣競爭。 ## 05｜國際玩家都會重新被定價 RevMed 的資料一出來，整個 RAS 賽道都會被重新看待。 Bristol Myers Squibb 旗下的 MRTX1133 曾是 KRAS G12D 領域早期焦點之一，目前仍有 G12D 實體瘤臨床試驗登錄紀錄。 Eli Lilly 的 LY4066434 則是 pan-KRAS inhibitor，也就是泛 KRAS 抑制劑，正在 KRAS 突變實體瘤中評估安全性與耐受性。 Incyte 的 INCB161734 也是口服 KRAS G12D inhibitor，也就是口服 KRAS G12D 抑制劑，已有早期臨床結果發表，並開始推進胰臟癌相關研究。 這些公司面對的問題會變得更尖銳： 你的藥能不能單藥打出差異？ 能不能安全地和化療、pan-RAS 或其他機制聯用？ 能不能在胰臟癌這種高死亡率疾病裡做出真正 OS？ 能不能建立自己的組合療法主場？ RevMed 已經把標準拉高了。未來 RAS 賽道的競爭，不只是誰最快進臨床，而是誰能在組合療法裡站住位置。 ## 結語｜癌王沒有被終結，但治療邏輯真的變了 RevMed 這次資料，真正改變的不是一個 ORR 數字。 它改變的是胰臟癌研發的想像。過去，胰臟癌治療多半圍繞化療組合微調；現在，RAS 標靶終於開始走進後線三期陽性、一線聯合、G12D 雙封鎖與多突變矩陣布局。 Daraxonrasib 證明，廣譜 RAS(ON) 抑制可以在胰臟癌做出 OS。 Zoldonrasib 證明，G12D 抑制可以與 pan-RAS 或化療形成更強組合。 接下來真正要看的，是三期試驗能否把這些早期訊號轉成標準治療。 胰臟癌還沒有被攻克。但 RevMed 至少做成一件事：它讓市場第一次相信，RAS 不再只是不可成藥的老故事，而可能成為胰臟癌下一個治療骨架。如果 daraxonrasib 成功上市，RevMed 也許不只是 Biotech。 它會成為下一代 RAS oncology company，也就是 RAS 腫瘤治療公司的原型。而整個 RAS 賽道，也會從「誰先做出單藥」進入「誰能建立組合療法生態」的新階段。 --- --- ## RevMed Upends Pancreatic Cancer Again: The RAS Combination Era Has Arrived URL: https://drugnews.com.tw/articles/2026-07-17-revmed-ras-en.html Date: 2026-07-17 Language: en Summary: Daraxonrasib has already raised the bar with phase 3 overall-survival data. Early zoldonrasib combinations now suggest that pancreatic cancer's RAS competition is shifting from single agents to treatment ecosystems. For years, pancreatic cancer's most frightening feature was not only its mortality. It was the absence of a targeted therapy capable of changing the treatment path in a meaningful way. More than 90% of pancreatic ductal adenocarcinomas carry a RAS mutation, and KRAS G12D is the largest single molecular subgroup. Yet RAS spent decades labeled “undruggable.” Its surface offered few stable pockets, its signaling changed shape dynamically, and even successful target engagement could be followed by pathway rebound. That old framework is now breaking apart. Revolution Medicines, better known as RevMed, has already shown in the phase 3 RASolute 302 trial that daraxonrasib can extend median overall survival to 13.2 months, compared with 6.7 months for investigator's choice, in previously treated metastatic pancreatic cancer. That result did more than validate one molecule. It established that broad RAS(ON) inhibition can produce a survival benefit in a disease long dominated by chemotherapy. The next step is even more consequential. Data presented around ESMO GI 2026 suggest that the KRAS G12D-selective inhibitor zoldonrasib can be paired with daraxonrasib or frontline chemotherapy to create a deeper and broader attack on RAS-driven disease. Drugnews' view is that the competitive question has changed. The market is no longer asking only which company can produce the best single KRAS inhibitor. It is asking which company can build the treatment backbone, combination logic and mutation-specific franchise around RAS. ## 01 | Why Pancreatic Cancer Is So Difficult: KRAS Is the Central Switch KRAS behaves like a molecular switch. In its GTP-bound state, it is active and transmits growth signals through downstream pathways such as RAF-MEK-ERK and PI3K-AKT. After GTP is converted to GDP, the normal protein can turn off. The G12D mutation interferes with that shutoff process. The switch remains biased toward the active state, allowing continuous proliferative and survival signaling. This matters especially in pancreatic cancer because RAS is not a peripheral feature of the disease. It is embedded in the biology of most tumors. Blocking the pathway therefore offers a potentially powerful point of intervention, but it also creates intense selective pressure. Cancer cells can recover signaling through wild-type RAS isoforms, receptor tyrosine kinases or parallel pathways. That is why a selective inhibitor can produce depth without necessarily producing breadth. It may suppress the dominant mutant driver very effectively while leaving the tumor opportunities to rewire around it. Conversely, a broad RAS inhibitor may cover multiple escape routes but face a narrower therapeutic window because it acts across a wider signaling network. The central challenge is therefore not merely to “hit KRAS.” It is to suppress the driver deeply enough, shut down the most relevant bypass mechanisms and still preserve a tolerable dose intensity. Pancreatic cancer has been one of the harshest tests of that balance. Its dense tumor microenvironment, rapid clinical progression and limited treatment window make incremental signals difficult to translate into durable benefit. A therapy that works here can carry implications far beyond one indication. ## 02 | Zoldonrasib Plus Daraxonrasib: Depth and Breadth, Not Just 1 + 1 Zoldonrasib and daraxonrasib play different roles. Zoldonrasib is designed to inhibit KRAS G12D selectively. Its strategic value is depth: it concentrates pressure on the mutation that drives a large share of pancreatic tumors. Daraxonrasib is a multi-selective RAS(ON) inhibitor. Its value is breadth: it can intercept active RAS signaling across multiple nodes and may constrain escape through other RAS proteins or reactivated upstream signaling. The biological logic behind the combination is straightforward. When the G12D driver is inhibited, a tumor may reactivate signaling through NRAS, HRAS, HER2, MET or other receptor-driven pathways. Adding daraxonrasib aims to keep those rescue routes from restoring enough pathway output for the tumor to survive. This is not simply two inhibitors attacking the same protein twice. It is a division of labor. One agent locks down the principal mutant driver; the other attempts to limit the network's ability to reorganize. The same reasoning explains why zoldonrasib is also being explored with chemotherapy. Modified FOLFIRINOX and gemcitabine plus nab-paclitaxel remain central first-line regimens. If selective RAS inhibition can deepen response without making the regimen impossible to deliver, the addressable opportunity moves from salvage treatment toward the much larger first-line setting. That step is commercially important. A later-line targeted therapy can establish proof of concept. A tolerable first-line combination can redefine the treatment architecture, expand duration of therapy and make the drug part of the default regimen rather than a rescue option. ## 03 | The Signals Are Strong, but the Samples Are Still Early The early combination data are difficult to ignore. In a second-line metastatic pancreatic cancer cohort, zoldonrasib plus daraxonrasib produced a 50% objective response rate, a 97% disease-control rate and median progression-free survival of 9.6 months. In patients treated in the third line or later, the reported objective response rate was 47%, the disease-control rate was 90% and median progression-free survival was 7.6 months. In first-line combinations, zoldonrasib plus modified FOLFIRINOX produced an objective response rate of 82%, while zoldonrasib plus gemcitabine and nab-paclitaxel produced an objective response rate of 61%. Those numbers are impressive in the context of metastatic pancreatic cancer, but the interpretation still needs discipline. The second-line and third-line-or-later cohorts each included approximately 30 patients. Small, early-phase cohorts can overstate effect size, especially when enrollment, baseline prognosis and follow-up differ. Response rate is not the same as overall survival, and cross-cohort comparisons are descriptive rather than randomized evidence. The frontline chemotherapy combinations are even earlier. The key questions are not only whether tumors shrink, but whether the response remains durable, whether progression-free survival holds up and whether overall survival ultimately improves. Safety will also determine whether the combination can become a backbone. Grade 3 or higher treatment-related adverse events were reported in 35% of patients in the highlighted combination dataset. Common events included rash at 12%, anemia at 10%, and stomatitis or mucositis at 7%. Treatment discontinuation rates were reported at 2% and 5% across the relevant cohorts, while daraxonrasib dose intensity was approximately 76%. The 76% figure deserves attention. A combination can look compelling at the efficacy level yet become commercially difficult if dose reductions, interruptions or cumulative toxicity prevent sustained delivery. The important follow-up questions are whether dose intensity stabilizes, whether adverse events can be managed predictably and whether efficacy persists after real-world dose modification. The investment conclusion is therefore not that these cohorts have already established a new standard of care. It is that RevMed has generated a sufficiently strong early signal to justify phase 3 development and to force the rest of the RAS field to answer a harder combination-therapy question. ## 04 | RevMed Is Turning RAS Into a Franchise RevMed's strongest strategic asset may not be any single drug. It is the architecture of the portfolio. Daraxonrasib provides the broad RAS(ON) backbone. Zoldonrasib covers KRAS G12D, the largest mutation subgroup in pancreatic cancer. Elironrasib, also known as RMC-6291, addresses KRAS G12C. RMC-5127 targets KRAS G12V. Additional programs aimed at variants such as Q61H and G13C could extend the matrix further. This portfolio matters because it allows RevMed to connect several dimensions of development at once: mutation subtype, line of therapy, tumor type and combination partner. The company can pursue second-line pancreatic cancer with a validated broad RAS inhibitor, move into first-line treatment through chemotherapy combinations, and layer mutation-selective agents onto the same backbone. The approach can then extend into lung cancer, colorectal cancer and other RAS-driven solid tumors. That changes how a KRAS G12D asset should be valued. In the previous framework, a G12D inhibitor was largely a stand-alone product story. In the new framework, its value also depends on whether it can become a modular component of a broader RAS regimen. The strategic questions become: - Can a G12D inhibitor combine safely with a pan-RAS or RAS(ON) backbone? - Can it improve chemotherapy outcomes without making the regimen undeliverable? - Can it be paired with PRMT5 or MAT2A inhibitors to exploit additional vulnerabilities? - Can EGFR inhibition control feedback in the relevant tumor types? - Can immunotherapy contribute after RAS signaling and the tumor microenvironment are altered? RevMed has effectively turned a single-node race into a matrix competition. Companies will need not only a potent molecule, but also a defensible place inside a combination ecosystem. ## 05 | The International RAS Field Will Be Repriced RevMed's progress raises the standard for every international competitor. Bristol Myers Squibb's MRTX1133 was one of the earliest high-profile KRAS G12D programs and remains in clinical development for G12D-mutant solid tumors. Its challenge is no longer simply to demonstrate target activity. It must show differentiated clinical depth, a manageable safety profile and a credible combination path. Eli Lilly's LY4066434 takes a broader pan-KRAS approach. That creates the possibility of mutation breadth, but it also places therapeutic-window and dose-intensity questions at the center of the thesis. Incyte's INCB161734 is an oral KRAS G12D inhibitor with early clinical data and pancreatic-cancer development activity. It will be judged not only on monotherapy response, but also on whether it can combine with chemotherapy or complementary RAS-pathway agents. The field now faces four sharper questions. First, can a molecule create a meaningful monotherapy difference? Second, can it combine safely with chemotherapy, broad RAS inhibition or other pathway mechanisms? Third, can it produce overall-survival evidence in a disease where response rate alone is not enough? Fourth, can the company establish a combination setting in which it owns the treatment backbone rather than merely supplies an add-on component? RevMed has raised the bar from entering the clinic to controlling a strategic position in the regimen. ## Conclusion | The “King of Cancers” Has Not Been Defeated, but the Treatment Logic Has Changed The most important feature of these data is not a single response-rate number. It is the change in what pancreatic-cancer drug development can plausibly look like. For years, progress largely meant adjusting chemotherapy combinations. Now the RAS field has reached phase 3 overall-survival evidence in later-line disease, first-line combination development, dual blockade of KRAS G12D and broader RAS signaling, and a portfolio strategy spanning multiple mutation subtypes. Daraxonrasib has shown that broad RAS(ON) inhibition can generate an overall-survival benefit in metastatic pancreatic cancer. Zoldonrasib has shown early evidence that selective G12D inhibition can be combined with daraxonrasib or chemotherapy to produce deeper signals. The next decisive test is whether randomized phase 3 trials can convert those signals into a new standard of care. Pancreatic cancer has not been solved. But RevMed has made the old claim that RAS is simply undruggable much harder to defend. If daraxonrasib reaches the market and the combination programs validate, RevMed may no longer be valued as a biotech built around one successful molecule. It could become the prototype of a next-generation RAS oncology company. And the competitive era would shift from “who has the first single agent?” to “who can build the most durable RAS treatment ecosystem?” ## Disclaimer This article is provided for industry research and educational purposes only. It does not constitute investment, medical, fundraising or securities advice. --- ## Big Pharma's New $10 Billion M&A Wave: Why Differentiation Now Beats Scale and Hype URL: https://drugnews.com.tw/articles/2026-07-16-big-pharma-ma-new-aesthetic-en.html Date: 2026-07-16 Language: en Summary: Vertex's roughly $10 billion deal for Crinetics captures a wider change in Big Pharma's acquisition taste: away from scale and fashionable platforms, toward niche leadership, technological step-changes and commercial visibility. ## 00 | Big Pharma's acquisition taste has changed Another $10 billion pharmaceutical acquisition has arrived. On July 6, Vertex Pharmaceuticals agreed to acquire Crinetics Pharmaceuticals for $85 per share in cash, implying an equity value of approximately $10 billion. If completed, the transaction will give Vertex PALSONIFY, Crinetics' marketed oral therapy for acromegaly, together with atumelnant, a late-stage asset for congenital adrenal hyperplasia and other ACTH-driven endocrine disorders. Vertex estimates that the portfolio carries roughly $5 billion in peak-sales potential. The deal also turns endocrinology into the company's fifth therapeutic area. By this point, 2026 has already produced a succession of pharmaceutical transactions at or near the $10 billion mark. The renewed acquisition appetite is not surprising. Patent cliffs are approaching, internal research cannot deliver on schedule every time, and large drugmakers still hold substantial cash and financing capacity. M&A remains the most direct way to fill a looming growth gap. What matters is not simply that Big Pharma is buying again. It is buying differently. Over the past decade, pharmaceutical acquisition logic passed through two distinct phases. The first worshipped scale: bigger was presumed to be better. The second worshipped fashionable platforms: the hotter the modality, the greater the fear of missing out. By 2026, a different preference is taking shape. Buyers are no longer chasing only the loudest therapeutic category. They are looking for assets that face less direct crowding, possess a clear source of differentiation and can establish leadership in a well-defined specialist market. That is the more important message behind Vertex's move for Crinetics. ## 01 | A new $10 billion wave built on precision, not indiscriminate buying Two strategic lines run through the current transaction cycle. The first is strategic de-crowding. Instead of forcing another asset into the most fashionable and congested category, companies are moving toward specialist markets with durable clinical needs, fewer competitors and a realistic path to category leadership. Sun Pharma's agreement to acquire Organon at an enterprise value of approximately $11.75 billion illustrates this logic. The attraction is not a glamorous new target. It is a global commercial network spanning women's health, established brands, biosimilars and specialty products. The portfolio may not carry the excitement of immuno-oncology, but it offers mature channels, clearer competitive dynamics and existing cash flow. Eli Lilly's agreement to acquire Centessa Pharmaceuticals for up to $7.8 billion represents a different version of the same move. Lilly is stepping outside the gravitational pull of GLP-1 to enter sleep-wake disorders. Centessa's core program, cleminorexton, is an orexin receptor 2 agonist being developed for narcolepsy and idiopathic hypersomnia. This is not the largest mass-market chronic-disease opportunity. It is a specialist CNS market with identifiable patients, substantial unmet need and a more limited field of direct competitors. Vertex's acquisition of Crinetics follows the same strategic grammar. Acromegaly and congenital adrenal hyperplasia have never been the loudest areas in biopharma. Treatment has often depended on injections or imperfect alternatives that create adherence and disease-management burdens. Crinetics did not need to invent entirely new biology to create value. It differentiated mature endocrine mechanisms through product design, converting cumbersome treatment into convenient once-daily oral therapy. PALSONIFY is the first and only once-daily oral therapy approved for adults with acromegaly. Atumelnant is a once-daily oral ACTH receptor antagonist in phase 3 development for congenital adrenal hyperplasia, with additional potential in Cushing's syndrome. This is not the most theatrical scientific story. It is, however, closely matched to Vertex's established operating model: build a deep specialist position in serious disease, then use global development and commercialization capabilities to magnify its value. Vertex created a highly concentrated franchise in cystic fibrosis. Buying Crinetics is, in effect, an attempt to reproduce that playbook in rare endocrine disease. The new preference can be stated simply: do not chase the biggest market; pursue the market that can be controlled. ## 02 | If a company stays in a crowded category, the technology must move up a generation The second line is technological de-crowding. Big Pharma will still invest in oncology, immunology, metabolism and neuroscience. What has changed is its willingness to pay a premium for an ordinary follow-on asset. Remaining in a large and competitive market now requires technology that can produce a generational difference from existing therapy. AbbVie's proposed acquisition of Apogee Therapeutics for approximately $10.9 billion is a bet on extended-duration immunology. Apogee's lead asset, zumilokibart, is a half-life-extended anti-IL-13 antibody being developed for atopic dermatitis and other type 2 inflammatory diseases. AbbVie does not lack immunology products. What it needs is a next-generation profile: fewer administrations, longer biological activity and a potentially better treatment experience that supports adherence. Novartis' approximately $12 billion acquisition of Avidity Biosciences adds a muscle-directed antibody oligonucleotide conjugate platform and three late-stage neuromuscular programs. The strategic value is not a single molecular target. Avidity's AOC architecture attempts to move RNA therapeutics beyond the liver and into muscle, addressing one of the central delivery constraints in oligonucleotide drug development. Merck's approximately $9.2 billion acquisition of Cidara Therapeutics follows the same logic. Cidara's CD388, now MK-1406, is a long-acting drug-Fc conjugate in phase 3 development for influenza prevention in people at higher risk of complications. It is neither a vaccine nor a conventional monoclonal antibody. The candidate links copies of a neuraminidase-inhibiting small molecule to an engineered Fc fragment, seeking to provide season-long protection from a single administration without dependence on vaccine-strain matching. These deals appear to span unrelated fields: immunology, neuromuscular disease, influenza and endocrinology. Their underlying language is remarkably consistent. Either leave the crowded arena and build dominance in a specialist market, or remain in a major category and enter with a technological step-change. Big Pharma has not stopped buying. It has become less willing to buy an ordinary story. ## 03 | A decade ago, scale itself was the thesis Turn the clock back a decade and the acquisition logic looks very different. The governing belief was that scale created safety. As blockbuster patents approached expiry, the most direct response was to acquire a larger company, absorb a broader product portfolio, eliminate duplicated costs and consolidate revenue immediately. The year 2019 marked the peak of this model. Bristol Myers Squibb acquired Celgene for roughly $74 billion. AbbVie acquired Allergan for about $63 billion. Takeda took on significant leverage to buy Shire and expand its rare-disease and plasma-derived-therapy businesses. These transactions reflected confidence in the idea of "too big to fail." More products, wider channels, removable costs and immediately consolidated profits appeared to offer protection against the patent cliff. The costs of mega-mergers were equally clear. Organizations became more cumbersome. Pipelines proved difficult to integrate. Debt burdens rose. Amortization of acquired intangible assets weighed on financial statements. When subsequent clinical programs or commercial launches underperformed, an acquisition designed to solve one growth problem could become a multi-year operating burden. Antitrust scrutiny also tightened globally, reducing the room for transformational mergers valued in the tens or hundreds of billions. Scale alone was no longer an adequate investment thesis. ## 04 | After the pandemic, fashionable platforms became the thesis Between 2020 and 2023, acquisition taste changed again. The pandemic created liquidity and a powerful fear of missing the next technology cycle. Drugmakers supported by COVID-19 windfalls or mature-product cash flow became anxious about being absent from the next major therapeutic platform. Antibody-drug conjugates, mRNA, cell therapy, gene therapy, new CNS mechanisms and targeted protein degradation all attracted escalating attention. Gilead paid approximately $21 billion for Immunomedics and its TROP2 ADC Trodelvy. Pfizer paid roughly $43 billion for Seagen in an attempt to rebuild its oncology business around an ADC platform. AbbVie paid approximately $10.1 billion for ImmunoGen and the FRalpha-directed ADC Elahere. The platform chase extended beyond ADCs. AbbVie bought Cerevel Therapeutics for about $8.7 billion, gaining emraclidine, an M4 receptor positive allosteric modulator then viewed as a promising approach to schizophrenia. Pfizer acquired Biohaven's migraine franchise and Global Blood Therapeutics' sickle-cell portfolio. The implicit rule was that buying the wrong fashionable asset might be painful, but missing the platform altogether could be worse. Once the cycle cooled, however, the costs became visible. Too many companies pursued similar pipelines. Clinical failures accumulated. Commercial performance frequently fell short of expectations. ADCs became crowded, CNS development risk reasserted itself, and some rare-disease assets encountered serious post-approval safety problems. Pfizer's Seagen portfolio has faced multiple setbacks. In June 2026, the phase 3 SigVie-002 study of sigvotatug vedotin in previously treated metastatic non-squamous non-small cell lung cancer failed to show a statistically significant overall-survival improvement over docetaxel in the overall population. The Global Blood Therapeutics acquisition delivered an even more severe lesson. Oxbryta, or voxelotor, was withdrawn globally over safety concerns. The FDA reported higher rates of vaso-occlusive crises and more deaths in postmarketing studies, concluding that the accumulating data no longer supported a favorable benefit-risk balance. The industry was forced to relearn three distinctions. A fashionable platform is not a competitive moat. Popularity is not differentiation. Paying a premium for technology cannot purchase clinical success in advance. ## 05 | The new preference is not to buy less, but to buy genuine differentiation By 2025 and 2026, pharmaceutical M&A was accelerating again, but the selection criteria had changed. Companies were no longer defaulting to another mega-merger or buying an entire fashionable platform simply to secure a seat at the table. They were searching more selectively for three types of asset. First, an asset with an established first-mover position in a specialist market. Second, a late-stage or marketed product with greater visibility into regulatory and commercial execution. Third, a platform capable of producing a generational difference rather than another closely related follow-on program. Crinetics fits this new preference. PALSONIFY is valuable not because the target is new, but because it turns acromegaly treatment into a once-daily oral regimen and enters a specialist market with clear unmet need and lower competitive density. Apogee is attractive not because IL-13 is novel, but because extended half-life could materially change dosing and the treatment experience in type 2 inflammatory disease. Avidity is not merely another RNA company. Its strategic proposition is the ability to deliver oligonucleotide therapy into muscle. Cidara is not simply another influenza program. Its drug-Fc conjugate may change how high-risk patients receive seasonal prevention. The new acquisition standard is therefore unforgiving. In a large market, an asset without a generational difference will struggle to command strategic value. In a small market, an asset without a credible path to leadership will face the same problem. ## 06 | What this means for Taiwan: build differentiation that international buyers can understand PharmaEssentia (6446) is the clearest Taiwan example of the specialist-franchise path. Its internally developed BESREMi, or ropeginterferon alfa-2b, is approved in the United States for adults with polycythemia vera. It is one of the few medicines originating from a Taiwan biotech that has established a real position in a global rare hematology market. Its significance does not come from addressing the largest possible market. It comes from using a long-acting interferon to build a differentiated global product position in myeloproliferative neoplasms, a highly specialized disease category. Foresee Pharmaceuticals (6576) is closer to the formulation and treatment-experience route. CAMCEVI, a long-acting leuprolide mesylate injectable, entered the hormonal-treatment market for advanced prostate cancer through a sustained-release formulation. The six-month product is marketed in the United States, while the three-month formulation previously entered FDA review. The strategic point is not a novel biological target. It is the use of formulation technology to improve dosing convenience and extend commercial options. That logic resembles Crinetics' effort to make endocrine treatment oral and easier to use. Ally Biotherapeutics, also known as Alar Pharmaceuticals (6785), belongs in the same discussion. Its long-acting buprenorphine injectable ALA-1000, now known as INDV-6001, was licensed to Indivior for development in opioid use disorder. Addiction medicine is not the most fashionable oncology narrative, but it contains substantial unmet need. A long-acting formulation could reduce the burden of frequent dosing, improve adherence and support patients with limited access to care. That is precisely the profile increasingly favored by the new acquisition environment: focused, specialist and built around a product difference that is easy to explain. For Taiwan biotech, the lesson is not to imitate the largest international platform or chase whichever target currently commands the highest valuation. The more durable question is whether a company's advantage can be translated into language a global partner understands. Does the asset dominate a defined patient segment? Does it materially improve administration, adherence or treatment logistics? Does the platform solve a delivery barrier? Is there enough clinical and regulatory visibility to reduce uncertainty? Can the company defend the economics and rights around the product? International value is created when scientific differentiation becomes commercial clarity. ## Conclusion | Big Pharma's taste has changed, and so has the standard for innovative drugs Big Pharma has spent hundreds of billions of dollars learning an expensive lesson over the past decade. The scale era demonstrated that bigger is not automatically better. The platform-hype era demonstrated that fashionable is not automatically right. The proliferation of similar pipelines demonstrated that buying a technology does not guarantee meaningful differentiation. The 2026 acquisition preference is therefore more disciplined. Large drugmakers remain willing to pay substantial premiums, but the asset must present a clear strategic case. It may dominate a specialist market. It may create a generational improvement in a large category. It may already be late-stage or marketed, improving visibility into execution. Or it may solve an important part of the clinical workflow rather than merely offering an elegant mechanism. Big Pharma is no longer buying scale or excitement alone. It is buying differentiation, certainty and the right to lead a market in which the product can actually survive. --- This article is provided for industry research and educational purposes only. It does not constitute investment, medical, fundraising or securities advice. --- ## 百億美元併購新狂潮 URL: https://drugnews.com.tw/articles/2026-07-16-big-pharma-ma-new-aesthetic.html Date: 2026-07-16 Language: zh-Hant Summary: Vertex 以約 100 億美元收購 Crinetics，映照 Big Pharma 併購審美的轉向：不再只買規模或熱門平台，而是買細分市場統治力、技術代差與商業確定性。 ## 00｜大藥廠的「審美」變了 又一筆百億美元級併購誕生。 7 月 6 日，Vertex Pharmaceuticals 宣布以每股 85 美元、總股權價值約 100 億美元的全現金交易，收購專注於內分泌罕見病的 Crinetics Pharmaceuticals。交易完成後，Crinetics 將為 Vertex 帶來已上市的肢端肥大症口服藥 PALSONIFY，以及處於後期開發階段的 atumelnant。 Atumelnant 瞄準先天性腎上腺增生症與其他 ACTH 相關內分泌疾病。Vertex 官方也表示，這組資產具備約 50 億美元峰值銷售潛力，並將內分泌學納入公司第五大治療領域。 至此，2026 年全球製藥產業的百億美元級併購已經一筆接一筆。 大型藥廠加大併購力度並不意外。專利懸崖逼近，內部研發不可能永遠準時交卷，手上又握有大量現金，併購本來就是 Big Pharma 填補成長缺口最直接的工具。 真正值得注意的，不是它們又開始買了，而是它們開始買得不一樣了。 過去十年，跨國藥廠的併購審美經歷過兩個階段：第一個階段是規模崇拜，越大越好；第二個階段是風口崇拜，越熱門越好。 到了 2026 年，新的審美正在成形：不再追逐最大聲量的賽道，而是尋找最不內卷、最具差異化、最能形成細分統治力的資產。 這才是 Vertex 收購 Crinetics 背後真正值得看的地方。 ## 01｜百億美元新狂潮：不是亂買，而是精準狙擊 這一輪跨國藥廠併購，有兩條很清楚的主線。 第一條，是策略去內卷。 也就是不再硬擠最紅、最擁擠、最被資本追捧的大賽道，而是轉向相對冷門、競爭較少、但臨床需求穩定的細分市場，追求絕對統治力。 Sun Pharma 以企業價值約 117.5 億美元收購 Organon，看中的不是一個炫目的新靶點，而是女性健康、成熟品牌、生物相似藥與專科產品組成的全球商業網路。這種資產不像腫瘤免疫那樣性感，但現金流穩定、通路成熟、競爭格局相對清楚。 Eli Lilly 以最高 78 億美元收購 Centessa Pharmaceuticals，則是跳出 GLP-1 舒適圈，切入睡眠清醒障礙市場。Centessa 的核心是 orexin receptor 2 agonist，也就是 orexin 2 受體促效劑 cleminorexton，瞄準 narcolepsy，也就是嗜睡症，以及 idiopathic hypersomnia，也就是特發性嗜睡症。這不是最熱的大眾慢病市場，卻是高度專科、需求明確、競品有限的 CNS，也就是中樞神經系統細分藍海。 Vertex 收購 Crinetics，也是同一個邏輯。 肢端肥大症與先天性腎上腺增生症，本來不是最吸睛的賽道。傳統治療長期依賴注射或不夠理想的替代方案，患者依從性差，醫療負擔也重。Crinetics 的價值，不在於發明了全新的生物學，而在於把成熟內分泌靶點做出產品差異化：把原本麻煩的治療，做成每日一次口服藥。 PALSONIFY 是全球首個且目前唯一每日一次口服肢端肥大症治療藥物。Atumelnant 則是口服 ACTH receptor antagonist，也就是 ACTH 受體拮抗劑，處於先天性腎上腺增生症後期開發階段。 這不是最華麗的科學故事，但非常符合 Vertex 的核心能力：在罕見病裡建立深度專科壁壘，再用全球商業化能力放大價值。 Vertex 過去靠囊性纖維化建立壟斷型商業模式。現在買 Crinetics，本質上是把這套模式複製到內分泌罕見病。 這是非常典型的新審美：不追最大市場，而追最能掌控的市場。 ## 02｜第二條路：如果留在大賽道，就必須技術升維 另一條主線，是技術去內卷。 如果 Big Pharma 仍然要留在腫瘤、免疫、代謝、神經這些大賽道，它們不再願意為普通跟隨型資產付高價。它們要的是能對現有療法形成代際差異的技術平台。 AbbVie 宣布以約 109 億美元收購 Apogee Therapeutics，看中的就是長效化免疫療法。Apogee 的核心資產 zumilokibart 是長效 IL-13 antibody，也就是長效 IL-13 抗體，主攻異位性皮膚炎、氣喘等 type 2 inflammation，也就是第二型發炎疾病。 AbbVie 不是缺免疫藥，它真正需要的是下一代免疫產品：更少給藥次數、更長作用時間、更好的患者黏著度。 Novartis 完成約 120 億美元收購 Avidity Biosciences，則是買下肌肉導向的 Antibody Oligonucleotide Conjugate，簡稱 AOC，也就是抗體寡核苷酸偶聯平台，與三個後期神經肌肉疾病管線。這筆交易的重點不是單一靶點，而是把 RNA 藥物從肝臟推向肌肉組織，解決過去小核酸藥物遞送受限的問題。 Merck 完成約 92 億美元收購 Cidara Therapeutics，也是類似邏輯。Cidara 的 CD388 是長效 Drug-Fc Conjugate，簡稱 DFC，也就是藥物 Fc 偶聯抗病毒候選藥，處於 Phase 3，用於高風險族群流感預防。它不是疫苗，也不是傳統單株抗體，而是小分子神經胺酸酶抑制劑與 Fc 片段結合，目標是一次注射提供整季保護，且不依賴疫苗株匹配。 這些交易表面上分散在免疫、神經肌肉、流感、內分泌，但背後語言一致：要嘛避開內卷，在細分市場建立壟斷；要嘛進入大賽道，但必須用技術升維打出差異化。 Big Pharma 不是不買了，而是不再願意買普通故事。 ## 03｜十年前的審美：規模即正義 如果把時間拉回十年前，Big Pharma 的併購邏輯完全不同。 那時候的核心信仰是：規模即正義。 重磅藥專利到期壓力浮現，藥廠最直接的解法，就是買更大的公司、吃下更大的產品矩陣、砍掉重複成本、快速並表收入。 2019 年，是大型併購的巔峰。Bristol Myers Squibb 以約 740 億美元收購 Celgene；AbbVie 以約 630 億美元收購 Allergan；Takeda 則背負巨大債務收購 Shire，押注罕見病與血液製品版圖。 這些交易背後，是對「大而不倒」的信仰。產品多、通路大、成本可砍、利潤能並表，看起來就能對抗專利懸崖。 但大型合併的代價也很清楚：組織臃腫、管線整合困難、債務壓力大、無形資產攤銷沉重。 一旦後續臨床或商業化不如預期，當初的併購很容易變成多年負擔。更重要的是，全球反壟斷監管開始收緊，千億美元級超大型合併的空間越來越小。 規模崇拜，不再那麼好用了。 ## 04｜疫情後的審美：風口即正義 2020 到 2023 年，Big Pharma 的審美又轉了一次。 疫情帶來流動性，也帶來 FOMO，也就是害怕錯過。手握 COVID-19 紅利和老藥現金流的大藥廠，開始害怕錯過下一代技術平台。 ADC、mRNA、細胞治療、基因治療、CNS 新機制、蛋白降解，一個又一個熱點被推高。 Gilead 以約 210 億美元收購 Immunomedics，拿下 TROP2 ADC Trodelvy。Pfizer 以約 430 億美元收購 Seagen，試圖用 ADC 平台重建腫瘤業務。AbbVie 以約 101 億美元收購 ImmunoGen，買下 FRα ADC Elahere。 風口不只在 ADC。AbbVie 也以約 87 億美元收購 Cerevel Therapeutics，押注精神分裂症 M4 receptor positive allosteric modulator，也就是 M4 受體正向變構調節劑 emraclidine。Pfizer 也買下 Biohaven 的偏頭痛管線，並收購 Global Blood Therapeutics，進入鐮刀型貧血市場。 那時候的邏輯是：寧可買錯，也不能缺席。 但風口過後，代價開始浮現。同質化管線太多，臨床失敗頻繁，商業化不如預期。ADC 變成紅海，CNS 高風險再次暴露，部分罕見病資產也在上市後面對真實世界安全性挑戰。 Pfizer 收購 Seagen 後，多個 ADC 管線接連遇挫。2026 年，sigvotatug vedotin 在既往治療過的非鱗狀非小細胞肺癌 Phase 3 試驗中，未能相較 docetaxel，也就是多西他賽，顯著改善整體存活期。 Pfizer 收購 Global Blood Therapeutics 的教訓更重。其鐮刀型貧血藥 Oxbryta，也就是 voxelotor，因安全性疑慮在全球撤市。FDA 也發布警示，指出 Pfizer 撤回該藥與疼痛併發症和潛在死亡風險有關。 這些案例讓產業重新清醒：風口不是壁壘，熱門不等於差異化；大手筆買下平台，也買不到臨床成功保證。 ## 05｜新審美的本質：不是不買，而是只買真正差異化 到了 2025、2026 年，Big Pharma 併購重新升溫，但審美明顯變了。 不是再追千億美元超級合併，也不是再看到熱門平台就全包，而是更精準地尋找三種資產： 第一，細分市場裡已建立先發位置的資產。 第二，後期臨床或已上市、商業化能見度高的資產。 第三，技術平台能帶來代際差異，而不是只做同質化跟隨的資產。 Vertex 買 Crinetics，就是這種新審美的代表。PALSONIFY 不是因為靶點新而值錢，而是因為它把肢端肥大症治療做成每日一次口服，切進一個高度專科、低內卷、未滿足需求明確的市場。 AbbVie 買 Apogee，也不是因為 IL-13 是新靶點，而是因為長效化可能改變 type 2 inflammation 的給藥體驗。 Novartis 買 Avidity，不是只買 RNA 藥物，而是買能把 RNA 藥物送進肌肉的遞送平台。 Merck 買 Cidara，不是買一個普通流感藥，而是買一種可能重新定義流感預防方式的長效 DFC 技術。 所以，Big Pharma 的新審美其實很殘酷：大賽道裡，沒有代際差異就不值錢；小賽道裡，沒有統治力也不值錢。 ## 06｜台灣可以怎麼看？不是追最大風口，而是做出可被國際理解的差異化 藥華藥（6446）是最典型的細分專科路線。 它的 BESREMi，也就是 ropeginterferon alfa-2b 長效干擾素，已在美國核准用於成人真性紅血球增多症，屬於台灣少數真正把自研藥物推進全球罕見血液疾病市場的公司。 它的意義不是「市場最大」，而是用長效干擾素在 MPN，也就是骨髓增生性腫瘤這種高度專科市場建立全球產品位置。 逸達（6576）則更接近劑型與給藥體驗革新的路線。 其 CAMCEVI，也就是 leuprolide mesylate 亮丙瑞林長效針劑，以長效緩釋注射劑切入前列腺癌荷爾蒙治療市場。美國六個月劑型已上市，三個月劑型也曾向 FDA 遞件並進入實質審查。 這類資產的重點不是全新靶點，而是用製劑平台改善治療便利性與商業延展性，和 Crinetics 把內分泌療法口服化的邏輯有相通之處。 昱展新藥（6785）也值得放進這個脈絡。 其長效 buprenorphine，也就是丁基原啡因注射劑 ALA-1000，現稱 INDV-6001，曾與 Indivior 達成授權合作，用於 opioid use disorder，也就是鴉片類藥物使用障礙。 這不是最熱門的癌症題材，但成癮醫學是明確未滿足需求，長效製劑又能改善依從性與復發風險，符合「小而深、專科化、產品差異清楚」的新審美。 ## 結語｜Big Pharma 的審美變了，創新藥的標準也變了 過去十年，Big Pharma 用數千億美元繳了一堂很貴的課。 規模崇拜告訴它們：大不等於好。 風口崇拜告訴它們：熱不等於對。 同質化平台告訴它們：買下技術，不代表買到差異化。 於是，2026 年的新審美開始變得冷靜。它們願意付高溢價，但前提是資產必須夠清楚：要嘛在細分市場形成統治力；要嘛在大賽道帶來代際升維；要嘛已經進入後期或上市，商業能見度足夠高；要嘛能真正解決臨床流程中的痛點，而不是只擁有一個漂亮機制。 Big Pharma 的審美變了。 它們不再只買規模，也不再只買熱點。 它們買的是差異化、確定性，以及一個細分市場裡真正能活下來的主導權。 --- --- ## After the Nobel Prize, Do Not Rush the First-Approved Treg Story: The Real Test Is Turning Immune Tolerance Into a Product URL: https://drugnews.com.tw/articles/2026-07-15-treg-immune-tolerance-productization-en.html Date: 2026-07-15 Language: en Summary: The Nobel Prize illuminated Treg biology, but commercial value will depend on whether immune tolerance can become a releasable, trackable, scalable and reimbursable product, not on an unverified first-approval headline. For some scientific themes, the most dangerous moment comes just after they are discovered by the wider market. That is when conclusions arrive too quickly. Regulatory T cells have returned to the spotlight. The 2025 Nobel Prize in Physiology or Medicine was awarded to Mary E. Brunkow, Fred Ramsdell and Shimon Sakaguchi for discoveries concerning peripheral immune tolerance. A term that once belonged mainly to the inner circle of immunology has now been pulled to the front of industry and capital-market conversations. Treg stands for regulatory T cell. Its story is very different from the aggressive language associated with CAR-T. CAR-T is described through finding, targeting and killing. Treg uses a quieter vocabulary: braking, tolerating and rebuilding order. Quiet does not mean unimportant. The underlying problem in many diseases is not that the immune system is too weak. It is that the immune system is too willing to damage the body it is meant to protect. Autoimmune disease, transplant rejection and uncontrolled inflammation cannot be solved simply by pressing harder on the accelerator. The more valuable question is whether the immune system can be taught to recover restraint. Our view is that the industrial value of Treg does not rest on the attention generated by a phrase such as "the first approved Treg drug." It rests on whether immune tolerance can be turned into a product that can be released, tracked, priced, delivered and reproduced. The Nobel Prize illuminated the science. The market's next questions concern manufacturing, quality, clinical setting and commercialization. ## 01｜What the Nobel Prize Illuminated: The Immune System's Brake An immune system that only knows how to attack does not make the body safer. Every day, the immune system must identify external pathogens while sparing the body's own tissues. That act of restraint is not passive. It is a highly organized biological system, and Treg cells are among its most important components. NobelPrize.org described the 2025 medicine award as recognizing discoveries concerning peripheral immune tolerance. In plain language, it concerns how the immune system avoids losing control in peripheral tissues and how it applies a brake when necessary. The drug-development implications are considerable. If CAR-T asks whether immune cells can be engineered into more precise weapons, Treg asks the opposite but equally consequential question: can immune cells be trained to maintain order? The idea sounds gentle. Product development is not. Once Treg becomes a therapeutic product, investors, pharmaceutical companies, physicians and regulators stop listening only to the immunology narrative. They ask harder questions. Where do the cells come from? How are they selected? How are they expanded? Is the phenotype stable? Can it change after infusion? Are batches consistent? Can a clinically meaningful endpoint be designed? The challenge in Treg has never been making the concept sound compelling. The challenge is that the modality operates close to the immune system's most precise control knobs. ## 02｜Why Treg Is Difficult: It Takes an Order-Engineering System, Not a Single Molecule For a small molecule or antibody, the hardest problems often involve target choice, affinity, half-life and toxicity. Treg cell therapy is different. The product is a living cell. Living cells change. They respond to their environment, are selected during manufacturing and encounter inflammatory signals in patients that differ radically from those in a controlled production system. Turning Treg into a product therefore involves at least five gates. The first is cell source. Is the product autologous or allogeneic? An autologous product is aligned with the patient's immune context, but manufacturing is expensive and slow, and batch variability can be substantial. An allogeneic product may be more industrializable, but it must address immune compatibility, safety and persistence. The second is Treg purity. A collection of T cells cannot be called a Treg product by assertion. Markers such as FOXP3, CD25 and low CD127 expression matter for the assessment of cellular identity and function. The closer a product moves toward the clinic, the less room there is to substitute a broad concept for specific characterization. The third is expansion and stability. Producing enough cells while preserving regulatory function is a manufacturing examination in its own right. Too little expansion leaves the program without an adequate dose. Expansion without sufficient control may loosen the identity of the product. The fourth is clinical setting. The choice of disease matters. Autoimmune disease, transplantation and inflammatory disease may all appear relevant, but each carries different endpoints, patient-selection criteria, dosing windows and competitive pressure. The fifth is pricing. If cell therapy is to enter chronic immune disease, it cannot simply inherit the high-price logic of oncology cell therapy. It must show payers what long-term medication, hospitalization, organ damage or recurrence it can reduce. Together, these questions form the real barrier to Treg development. Many companies can participate in a Treg theme. The companies most likely to be repriced are those that connect immunology, manufacturing, clinical development and a viable commercial model. ## 03｜The Global Field Is Splitting Into Three Product Routes The global Treg and immune-tolerance field can be organized into three broad routes. The first route is in vivo induction or expansion of Treg cells. Nektar's rezpegaldesleukin is a useful reference point for this model. It pursues immune modulation in vivo, using a Treg-biased mechanism to reshape immune balance inside the patient. The advantage is that the product remains closer to conventional drug development and can be understood through a more familiar clinical-development and scaling framework. Its challenge is equally clear. The broader the in vivo immune modulation, the more the program must prove selectivity, dose and therapeutic window. The immune system is not a single switch. Too little modulation may have no effect; too broad an effect may create new safety risks. The second route is ex vivo Treg cell therapy. Sonoma Biotherapeutics provides a strong example. Its pipeline emphasizes engineered Treg cell therapeutics for autoimmune and inflammatory disease. The route has substantial potential because the localization, function and intended indication of the cell product can be designed with greater precision. But the moment a program moves ex vivo, manufacturing, cost, logistics, batch release and the clinical administration workflow arrive together. A presentation may show one pipeline asset; a clinical site experiences an entire supply chain. The third route is engineered Treg. Quell Therapeutics represents a more engineered approach. The goal is to direct Treg cells more precisely toward selected tissues or antigen-associated settings and thereby address immune dysregulation. This route is both elegant and exceptionally demanding. It is elegant because it moves Treg beyond broad immune modulation toward a more precise product design. It is demanding because it must carry cell engineering, immune stability, manufacturing scale-up and clinical-endpoint design at the same time. The global Treg race should therefore be read through product-route selection and execution. The key question is not who uses the term Treg, but who can choose a route and advance it to verifiable evidence. ## 04｜The Taiwan Read-Through: Taiwan Bio Therapeutics Belongs on the Watchlist In Taiwan, Taiwan Bio Therapeutics belongs on the early productization watchlist for Treg and immune tolerance. That does not justify presenting the company as an analogue of an unverified overseas approval, nor is the label "Treg concept stock" sufficient analysis. The substantive questions are whether the company can explain cell source, manufacturing, identity markers, clinical setting and delivery capability. The company has publicly identified TRK-001 as an autologous, ex vivo-expanded Treg product intended to prevent rejection in living-donor kidney-transplant recipients. An official company disclosure stated that Taiwan's Ministry of Health and Welfare had agreed in principle to the Phase II study. This is a useful verified anchor, but it does not remove the need to evaluate execution. The market should not stop at asking whether a company "has Treg." It should ask: - Is the cell source and manufacturing process clearly described? - Can Treg identity and functional markers be followed consistently? - Do the expanded cells retain regulatory function? - Is the intended clinical setting autoimmune disease, transplantation or inflammatory disease? - Can the endpoint be understood by physicians, regulators and payers? - Can the product move from laboratory work to repeatable delivery? Those are the questions that matter when assessing Taiwan Bio Therapeutics. A better approach is to bring the same industrial framework back to Taiwan: Which capabilities are required for a Treg product? Which data can the market understand? Which developments deserve attention? Taiwan's biotech market can move quickly on a thematic phrase. Treg requires the opposite instinct. Its central promise is stability and control. ## 05｜Risk: A First-Approved Claim Cannot Rest on Secondary Narratives It is not surprising that the Treg theme is gaining attention. The Nobel Prize provided scientific recognition. Global companies are advancing different product routes, and cell therapy is moving beyond tumor killing toward immune regulation. These are real signals. The more real the signal, the less it should be distorted by an overextended headline. A claim such as "the world's first approved Treg drug" cannot be treated as fact without support from the FDA, Drugs@FDA, an official company release, a prescribing label or another formal regulatory document. Regulatory approval is among the highest-risk categories of information. It should not be reconstructed from a secondary article or screenshot. That is the reason for our caution. Treg can and should be analyzed. It relates to an important next direction for cell therapy: moving from tumor killing to rebuilding immune order. But the phrase "first approved" should not be rushed. When a primary source is available, the event can be treated as approval news. Until then, the better approach is to place Treg on a productization map and distinguish scientific attention, clinical progress, manufacturing capability and distance from commercialization. That does not weaken the story. It makes the story more consistent with how the industry actually develops. ## Conclusion｜The Next Treg Test Is Turning Immunology Into a Controlled Product The most compelling feature of Treg is that it changes the tone of cell therapy. The market has spent years hearing how cell therapies attack, kill and clear disease. Treg elevates a different question: when disease is caused by immune dysregulation, can the system be taught to recover order? The question is large and difficult. That is why the next stage for Treg is not to make the terminology hotter. It must turn immunology into a product. Can developers select the right cells? Can they expand stable cells? Can those cells preserve function inside patients? Can clinical endpoints be accepted by regulatory and reimbursement systems? Can cost be reduced enough for repeatable delivery? These questions matter more than who is first. The Nobel Prize made Treg visible to a much wider audience. The industrial examination is only beginning. For Taiwan, Taiwan Bio Therapeutics can be placed on the watchlist, but it should be evaluated using the same demanding framework. The market will ultimately reward companies that can turn immune tolerance into a controlled product. ## Disclaimer This article is industry research and market commentary. It does not constitute investment advice, a recommendation to buy or sell securities, medical advice or an endorsement of any individual company. Cell therapy and biotech investing involve risks related to clinical trials, regulatory review, manufacturing scale-up, commercial competition, reimbursement and capital-market volatility. Readers should conduct their own due diligence and bear responsibility for their investment decisions. --- ## 諾貝爾獎後，Treg 第一個獲批藥還不能急著喊：真正的考題是免疫耐受怎麼做成產品 URL: https://drugnews.com.tw/articles/2026-07-15-treg-immune-tolerance-productization.html Date: 2026-07-15 Language: zh-Hant Summary: 諾貝爾獎照亮 Treg 科學，但產業價值要看免疫耐受能否被做成可放行、可追蹤、可定價、可複製的產品，而不是急著追逐未經核實的第一款獲批敘事。 有些題材，最危險的時刻，是剛被看見的時候。 是大家太快替它下結論。 Treg 最近重新站回聚光燈。2025 年諾貝爾生理學或醫學獎頒給 Mary E. Brunkow、Fred Ramsdell 與 Shimon Sakaguchi，表彰他們在 peripheral immune tolerance 上的發現。這件事讓一個原本偏免疫學核心圈的名詞，被拉回產業和資本市場前台。 Treg，全名 regulatory T cell，調節性 T 細胞。 它不像 CAR-T 那樣有很強的攻擊敘事。CAR-T 的語言是找到、鎖定、殺掉；Treg 的語言安靜得多，它談的是煞車、容忍、重建秩序。 可是安靜不代表不重要。 很多疾病的底層問題，並非免疫系統不夠強，常常是它太會誤傷自己。自體免疫疾病、移植排斥、發炎失控，單純多踩一腳油門解不了。這時候，真正有價值的問題變成：能不能讓免疫系統重新學會分寸？ 藥時事團隊的判斷是：Treg 的產業價值，不在一句「第一個獲批藥」有多吸睛，而在免疫耐受能否被做成可放行、可追蹤、可定價、可複製的產品。 諾貝爾獎點亮的是科學。市場接下來要看的，是製程、品質、臨床場景和商業化。 ## 01｜諾貝爾獎照亮的，是免疫系統的煞車 免疫系統如果只會攻擊，人不會活得比較安全。 我們每天都靠免疫系統辨識外來病原，也靠它放過自己的組織。這個「放過」不是消極，它是一套高度精密的秩序。Treg 就是在這套秩序裡很重要的細胞之一。 NobelPrize.org 對 2025 年醫學獎的說法很清楚：這次獲獎，是因為 peripheral immune tolerance 的發現。白話講，就是免疫系統如何在身體周邊組織裡避免失控，如何在必要時踩住煞車。 這個概念放進藥物開發，想像空間很大。 如果 CAR-T 是把免疫細胞改造成更精準的攻擊武器，Treg 則像是在問另一個方向：能不能把免疫細胞訓練成秩序維持者？ 這句話看起來柔和，產品化卻一點也不柔和。 因為一旦要做成藥，市場就不會只聽免疫學故事。投資人、藥廠和醫師會追問更硬的細節：細胞從哪裡來？怎麼分選？怎麼擴增？表型穩不穩？進到病人體內後會不會變？批次之間能不能一致？臨床終點能不能設計得出來？ Treg 的難，從來不在概念漂亮。 難在它太接近免疫系統的精密旋鈕。 ## 02｜Treg 為什麼難做：一顆分子解不了，一套秩序工程才解得動 小分子藥或抗體藥，最難的問題常常是靶點、親和力、半衰期、毒性。 Treg 細胞療法不一樣。 它要處理的是活細胞。活細胞會變，會受環境影響，會在製程裡被篩選，也會在病人體內面對完全不同的發炎訊號。 所以 Treg 產品化至少有五道關卡。 第一，細胞來源。是自體，還是異體？自體比較貼近個人免疫狀態，但製程昂貴、速度慢、批次差異大。異體有機會工業化，但免疫相容性、安全性和持久性都要回答。 第二，Treg 純度。Treg 不是隨便抓一群 T 細胞就能叫 Treg。FOXP3、CD25、CD127 low 這些表型，會牽動細胞身份和功能判斷。產品越接近臨床，越不能靠概念帶過。 第三，擴增與穩定。要把細胞擴到足夠劑量，同時維持調節功能，本身就是製程考試。擴太少，沒有劑量；擴太粗，身份可能鬆掉。 第四，臨床場景。Treg 要打哪裡很關鍵。自體免疫疾病、移植、發炎疾病，看起來都能講，但每個場景的終點、病人選擇、給藥時間和競品壓力都不同。 第五，定價。細胞療法如果要進入慢性病或免疫疾病，不能只套用腫瘤細胞療法的高價邏輯。它要說服支付方：這個產品到底能少掉多少長期用藥、住院、器官損傷或疾病復發。 這些問題合在一起，才是 Treg 的真正門檻。 碰到 Treg 的公司很多，真正能被重估的，是把免疫學、製程、臨床和商業模式接起來的公司。 ## 03｜全球賽道正在分成三條路 現在全球 Treg / 免疫耐受賽道，大致可以看成三條路。 第一條，是體內誘導或擴增 Treg。 Nektar 的 rezpegaldesleukin 就屬於這類觀察座標。它選的是體內免疫調節，透過偏向 Treg 的作用方式，嘗試在人體內重塑免疫平衡。這條路的好處是比較接近藥物化、規模化，也比較容易被傳統臨床開發框架理解。 但它的問題也很清楚：體內調節越廣，越需要證明選擇性、劑量和安全窗。免疫系統不是單一開關，推得太弱沒有療效，推得太廣可能帶來新的安全性問題。 第二條，是 ex vivo Treg 細胞療法。 Sonoma Biotherapeutics 的管線，就是這條路很好的例子。它把重點放在 engineered Treg cell therapeutic，試圖用更精準的細胞工程去處理自體免疫與發炎疾病。這條路的想像空間大，因為它有機會把細胞的定位、功能和適應症設計得更精細。 可是只要走到 ex vivo，製程、成本、物流、批次放行和臨床給藥流程就會一起出現。它寫在簡報裡是一條管線，進到臨床現場就是一整套供應鏈。 第三條，是工程化 Treg。 Quell Therapeutics 代表的是更工程化的方向。它關注 engineered Treg cell therapy，核心想法是讓 Treg 更精準地到達特定組織或抗原相關場景，處理免疫失衡。 這條路很漂亮，也最硬。 漂亮，是因為它把 Treg 從「泛免疫調節」推向更精準的產品設計。硬，是因為它同時背負細胞工程、免疫穩定、製程放大和臨床終點四件事。 所以全球 Treg 賽道真正要看的，是誰能選對一條產品路徑，並且把它走到可驗證。 ## 04｜台灣怎麼看：台寶先放進觀察座標 放回台灣，台寶可以先放進 Treg / 免疫耐受產品化的觀察名單。 但這個觀察不能被講成海外未查證核准產品的同型藥，也不能只用一句「Treg 概念股」帶過。真正要看的是：它能不能把細胞來源、製程、身份標記、臨床場景與交付能力說清楚。 也就是說，市場不該只問「有沒有 Treg」。 更該問這幾件事： 📌 細胞來源與製程是否說得清楚？ 📌 Treg 身份與功能標記是否能穩定追蹤？ 📌 擴增後的細胞是否仍維持調節能力？ 📌 臨床場景是自體免疫、移植，還是發炎疾病？ 📌 終點設計能不能讓醫師、監管單位和支付方理解？ 📌 產品最後能不能從實驗室走向可重複交付？ 這才是看台寶時真正該回到的問題。 更好的做法，是把同一套產業題目拿回來檢查：Treg 如果要變成產品，台灣公司需要補上哪些能力，哪些資料可以被市場讀懂，哪些進展才值得追。 台灣生技市場很容易被一句題材帶動。Treg 這題更需要慢一點。 因為它的核心在穩。 ## 05｜風險：第一個獲批藥不能靠二手敘事 Treg 題材會熱，並不意外。 諾貝爾獎給了科學背書，全球公司開始推進不同產品路徑，細胞療法也正在從腫瘤攻擊走向免疫調節。這些都是真訊號。 但越是真訊號，越不能被一句過滿的話帶偏。 「全球第一個獲批 Treg 藥」這種說法，如果沒有 FDA、Drugs@FDA、公司官方新聞稿、仿單或正式監管文件支撐，就不能拿來當事實。監管核准是最高風險資訊，不適合用二手文章或截圖補。 這也是藥時事團隊這次更在意的點。 Treg 可以寫，而且應該寫。它關係到細胞療法下一個很重要的方向：從殺傷腫瘤，走向重建免疫秩序。 但「第一個獲批」不能急著喊。 等真正的一級來源出現，再把它當新聞事件處理。現在比較好的方式，是先把 Treg 放回產品化地圖裡看，分清楚科學熱度、臨床進度、製程能力和商業化距離。 這樣寫，沒有少掉故事。 反而更接近產業真正會發生的事。 ## 結語｜Treg 的下一關，是把免疫學做成可控產品 Treg 最迷人的地方，是它改變了細胞療法的語氣。 過去幾年，市場習慣聽細胞療法怎麼攻擊、怎麼殺傷、怎麼清除病灶。Treg 讓另一個問題變得更重要：當疾病來自免疫系統失控，我們能不能讓它重新學會秩序？ 這個問題很大，也很難。 所以 Treg 的下一關，不在把名詞講熱。 它要把免疫學做成產品。 能不能分選出對的細胞，能不能擴增出穩的細胞，能不能在病人體內保持功能，能不能設計出可被監管與支付系統接受的臨床終點，能不能把成本壓到可交付。 這些問題，比「誰第一個」更重要。 諾貝爾獎讓 Treg 被更多人看見。真正的產業考試，現在才開始。 對台灣來說，台寶可以先放進觀察座標，但也要用同一套嚴格問題檢查。市場最後會獎勵的，是最早把免疫耐受做成可控產品的公司。 ## 免責聲明 本文為產業研究與市場觀察，不構成任何投資建議、買賣建議、醫療建議或個股推薦。細胞療法與生技醫藥投資涉及臨床試驗、法規審查、製程放大、商業化競爭、支付環境與資本市場波動，投資人應自行判斷並自負盈虧。 --- ## The Next ADC War: Novartis Is Paying Up to $1.5 Billion for Myricx, but the Real Prize Is a New Payload Class URL: https://drugnews.com.tw/articles/2026-07-14-novartis-myricx-nmti-adc-en.html Date: 2026-07-14 Language: en Summary: Novartis is paying up to $1.5 billion for Myricx. The strategic prize is not merely two ADC assets, but an NMT-inhibitor payload platform that could reopen the questions of therapeutic window, resistance and platform value. The antibody-drug conjugate market has become crowded enough to feel overheated. HER2, TROP2, B7-H3 and Nectin-4 have been pursued by wave after wave of companies. Topoisomerase I inhibitors, microtubule inhibitors, bispecific ADCs and dual-payload ADCs now appear in almost every oncology R&D presentation. Over the past several years, ADCs have been one of the brightest tables in innovative drug development, drawing capital, business-development budgets, clinical programs and large-pharma attention. Yet the hotter the table becomes, the clearer the genuinely scarce resource is. The market does not need another ADC merely because it is an ADC. It needs enabling technology capable of reopening three stubborn questions: efficacy, therapeutic window and resistance. On July 6, Novartis announced an agreement to acquire the UK biotech Myricx Bio. The transaction could be worth up to $1.5 billion, including $1.1 billion upfront and as much as $400 million in potential milestones. Closing is expected in the second half of 2026, subject to regulatory approval and customary conditions. Novartis did not rush to follow the crowd. It appears to have waited until ADC competition entered a new phase, then bought a ticket to a different set of rules. Our view is that the visible assets are Myricx's B7-H3 and HER2 lead programs, but the strategic position lies in the novel payload. The next ADC war will move from the question of which target is hottest to whether a payload can speak a different clinical language. The platforms most likely to be repriced in the next round of business development will be those that can address resistance, widen the therapeutic window and demonstrate repeatability across assets. ## 01｜The Old ADC Problem: A Good Target Does Not Automatically Make a Good Drug ADCs are often described as biological missiles. The antibody finds the cancer cell, the linker controls release and the payload delivers the damage. The analogy is intuitive, but it can obscure a crucial point: accuracy does not guarantee that the warhead is good enough. Most of the persistent problems with current ADCs eventually return to three issues. First, the therapeutic window can be too narrow. If a payload is too toxic, the dose cannot be pushed. If the dose cannot be pushed, effective tumor exposure is constrained. Dose reductions, treatment delays and discontinuations are recurring reminders that target precision alone is insufficient. If the payload cannot be tolerated, the program remains dose-limited. Second, resistance emerges. Topoisomerase I inhibitors and microtubule inhibitors have clearly demonstrated value, but cancer cells do not remain passive. Lower target expression, drug efflux, altered DNA-repair pathways and TOP1-related resistance can all erode activity over time. Third, mechanisms are increasingly concentrated. When many companies combine similar targets, payloads and linkers in similar patient populations, the market eventually asks a harder question: which clinical problem did this program actually solve? That is why Myricx attracted Novartis. The company did not simply produce a more polished version of the familiar ADC presentation. It moved the contest one layer deeper, into payload biology. ## 02｜NMTi-ADCs: Novartis Is Buying a Different Logic of Tumor Killing Myricx's core technology uses an N-myristoyltransferase inhibitor, or NMTi, as an ADC payload. NMT is an intracellular enzyme involved in protein myristoylation. These modifications affect the localization, activity and signaling of selected proteins, and cancer-cell growth and survival can depend on those intracellular processes. Myricx's approach is to make an NMT inhibitor the payload of an ADC, use an antibody to deliver it to tumor cells and then disrupt critical cellular programs. The important distinction is that this mechanism differs from those of mainstream ADC payload classes. If a Topo I payload applies pressure to DNA replication and repair, and a microtubule inhibitor jams the machinery of cell division, an NMTi payload is designed to attack a different and potentially more upstream set of cellular dependencies. That mechanistic distance is central to why Novartis was willing to commit $1.1 billion upfront. The official announcement is explicit about the thesis. Myricx's NMTi payload is intended to address limitations of existing payload classes, and preclinical data suggest broad activity across multiple solid tumors, including Topo I-resistant models. This is also where restraint matters. What can currently be verified is preclinical potential. Whether NMTi-ADCs can translate resistance biology, tolerability and extended efficacy into a genuine clinical advantage still requires human data. ADC history contains no shortage of compelling preclinical stories. The hard part is entering patients and preserving both activity and tolerability. The significance of the transaction is therefore not that Myricx has already proved everything. It is that Novartis is willing to use a large upfront payment to secure a platform that could become a new payload class before its clinical profile is fully established. ## 03｜Novartis Waited for Differentiation Before Making Its Move This deal is particularly revealing in the context of Novartis. The company has not been among the most aggressive participants in the ADC deal frenzy. While transactions multiplied across the sector, Novartis did not rush to accumulate a collection of look-alike assets. Its choice of Myricx makes its standard clearer: the ADC label alone is not valuable enough. A differentiated mechanism is. Myricx gives Novartis three layers of value. The first is a pair of lead ADC assets. According to the announcement, the programs target B7-H3 and HER2, both intensely competitive ADC targets that still encompass substantial unmet clinical need. The second is the NMTi payload platform. This matters more than a single product. If the payload class works, it could potentially be coupled to additional antibodies, targets and solid-tumor settings. The third is an extension of Novartis's oncology-platform option set. In recent years, Novartis has emphasized repeatable capabilities in radioligand therapy, targeted oncology and platform assets. If successful, Myricx could become a new foundational module in the company's ADC portfolio. The transaction also makes the valuation logic for ADCs more demanding. The earlier market tended to begin with familiar questions: Is the target popular enough? Is the patient population large enough? Do early data show a response? The next round will add more difficult questions: - Is the payload truly differentiated from the dominant mechanisms? - Can the linker-payload combination support an adequate dose and therapeutic window? - Is there a credible clinical position in patients who failed or became resistant to existing ADCs? - Can the platform be replicated from one asset to several targets? - After acquisition, can a large pharmaceutical company integrate it into an established oncology-development and commercial system? Those questions matter far more than the ADC theme itself. ## 04｜Risk: A Novel Payload Does Not Guarantee a Clinical Win Novel payloads are attractive precisely because they are risky. The opportunity is that they may open patient populations current ADCs cannot adequately serve. As Enhertu, Trodelvy and Padcev increasingly establish ADCs as part of standard treatment, a next-generation asset that produces credible data in resistant disease, low-expression tumors, multiple cancer types or rational combinations could become extremely valuable. The danger comes from the same source. The payload is the killing core of an ADC, and a new mechanism also introduces new uncertainty. It may improve tumor killing, but it may create unexpected toxicity. It may bypass known resistance, yet encounter new problems in human exposure, tissue distribution, bystander effects, metabolism or safety. Myricx faces at least three major tests. First, the NMTi payload must demonstrate a real therapeutic window in humans. Preclinical models can indicate direction, but human safety will set the dose ceiling. Second, the B7-H3 and HER2 lead assets must produce persuasive early clinical signals. A target alone cannot carry an ADC. Antibody properties, linker, payload, drug-to-antibody ratio, tumor expression and patient selection all shape the outcome. Third, the platform must prove repeatable. Success in one asset would matter, but the $1.5 billion transaction is ultimately paying for a platform. If NMTi works only in a narrow set of programs, the value will be discounted. If it works across targets, the valuation logic changes substantially. The transaction is a useful reminder for the entire ADC industry: making the name of the modality more complicated is not enough. A next-generation ADC must solve a clinical problem more clearly. ## 05｜The Taiwan Read-Through: OBI Pharma, T-E Meds and HoneyBear Require Different Questions The Taiwan comparison begins by separating two common mistakes. OBI Pharma is not the only company worth watching, and not every antibody company belongs on an ADC list. Our current Taiwan map narrows the discussion to three names: OBI Pharma, T-E Meds and HoneyBear Biosciences. OBI Pharma is the most straightforward direct ADC company to verify through public information. Its official pipeline includes several ADC programs: the TROP2-targeted OBI-992, next-generation TROP2 ADC OBI-902, Nectin-4 ADC OBI-904 and the TROP2 x HER2 bispecific ADC OBI-201. The company also describes its Obrion ADC technology platform, including GlycOBI, HYPrOBI, ThiOBI and GlycOBI DUO capabilities related to site-specific conjugation, linkers and dual-payload design. The OBI Pharma thesis cannot stop at the phrase "has ADCs." The next evidence points are whether assets such as OBI-992 and OBI-902 can generate human safety, preliminary efficacy, dose and patient-selection data, and whether GlycOBI and Obrion can demonstrate platform repeatability through external partnerships, follow-on pipelines and accumulating evidence. The second name is T-E Meds. It is the company most clearly missing from many quick Taiwan ADC lists. A November 7, 2025 disclosure on Immunwork's website describes T-E Meds as focused on ADCs and antibody-radionuclide conjugates. Its proprietary multi-arm linker platform combines mono- or dual-drug bundles with modular conjugation, with the goal of producing high-DAR, high-yield, high-purity conjugates and greater payload flexibility. The same official page identifies two core candidates. TE-1146 is described as an anti-CD38 daratumumab conjugated with lenalidomide, with multiple myeloma and related diseases among its intended indications. The page said an IND filing was expected in the first half of 2026. It also stated that commercialization rights in Taiwan, Southeast Asia, Australia and New Zealand had been licensed to Orient EuroPharma for NT$270 million. TE-1182 is a dual-payload ADC for HER2-positive tumors, designed for HER2-low or later-line resistant settings. According to the official page, an IND filing was expected in the fourth quarter of 2026. T-E Meds matters because it shows an early Taiwan example combining linker technology, dual payloads, named candidates and regional licensing. Its biggest tests remain ahead: whether IND timing holds, whether CMC and toxicology support entry into the clinic, and whether first-in-human data can advance the story from "platform" to a clinically interpretable signal. The third name is HoneyBear Biosciences. It belongs on a reader watchlist, but we are not presenting it as a verified ADC pipeline at this stage. After a renewed review, we still could not identify a sufficiently citable official pipeline page, specified target, payload-linker design, IND milestone or licensing announcement. A company can be placed on an early watchlist without filling gaps with market impressions. The real question for HoneyBear is data completion. Has the company publicly disclosed an ADC product, target and payload architecture? Are there preclinical data, patents, presentations, partnerships or trial registrations? Once that information is available, it can enter a formal comparison. Until then, an early-watch classification is more responsible than a manufactured conclusion. These companies illustrate three distinct evidence states. For OBI Pharma, watch whether an established public pipeline and technology platform generate clinical data. For T-E Meds, watch whether an early ADC technology and its candidates reach human trials. For HoneyBear, watch whether public information becomes sufficient for the market to begin formal valuation. That is also the message from Novartis's acquisition of Myricx. A list of company names does not reveal value differences. ADCs must be separated into technology, evidence and partnership tracks before investors can see which companies may connect differentiated engineering, clinical data, platform repeatability and external validation. ## Conclusion｜ADCs Do Not Need Another Follower. They Need Another Answer Novartis's acquisition of Myricx Bio for up to $1.5 billion may look like one more ADC transaction. What it actually illuminates is the next dividing line in the sector. The first ADC wave was valued through targets, response rates, indications and large-pharma business development. The next wave will be judged more severely: Does the payload introduce a genuinely different mechanism? Is the therapeutic window wide enough? Can resistance be addressed? Can the platform be repeated across targets? The next ADC war will not be won by the company that merely connects the antibody more elegantly. It will be won by the company that turns the payload into a new clinical language. Myricx has not completed its clinical examination, so the transaction cannot be read as proof that NMTi-ADCs have already won. A more accurate interpretation is that Novartis has purchased an early option on a platform that could alter the ADC payload landscape. For global biotech, the message is that ADC competition is moving from participation in a fashionable modality to proof of foundational innovation. For Taiwan investors, the questions can now be sharper. Can OBI Pharma advance its platform and pipeline evidence? Can T-E Meds move early technology into human trials? Can HoneyBear provide enough public evidence to support real analysis? The companies whose data can be understood by the international market are the ones most likely to address the next set of ADC problems. As a field becomes crowded, ordinary assets are discounted. What remains scarce is a new answer compelling enough for large pharma to pay early. ## Disclaimer This article is industry research and market commentary. It does not constitute investment advice, a recommendation to buy or sell securities, medical advice or an endorsement of any individual company. Biotech investing involves risks related to clinical trials, regulatory review, licensing negotiations, commercialization, capital-market volatility and company execution. Readers should conduct their own due diligence and bear responsibility for their investment decisions. --- ## ADC 下一場戰爭：諾華 15 億美元買 Myricx，真正搶的是新型 payload URL: https://drugnews.com.tw/articles/2026-07-14-novartis-myricx-nmti-adc.html Date: 2026-07-14 Language: zh-Hant Summary: 諾華最高 15 億美元收購 Myricx，真正買下的不只是兩條 ADC 管線，而是可能改寫安全窗、耐藥與平台估值的新型 NMTi payload。 ADC 賽道已經熱到有點擁擠。 HER2、TROP2、B7-H3、Nectin-4，熱門靶點被一輪又一輪公司追著做；Topo I payload、微管抑制劑、雙抗 ADC、雙 payload ADC，也被放進幾乎每一場腫瘤研發簡報。過去幾年，ADC 像是創新藥產業裡最亮的牌桌，資本、BD、臨床與大藥廠預算都往這裡聚。 可是牌桌越熱，真正稀缺的東西反而越清楚。 現在不缺「又一個 ADC」。市場缺的是能把療效、安全窗、耐藥問題重新打開的底層技術。 7 月 6 日，諾華宣布簽署協議收購英國生技公司 Myricx Bio。這筆交易最高可達 15 億美元，其中 11 億美元是 upfront，另外最高 4 億美元為潛在里程碑付款。交易預計在 2026 年下半年完成，仍需監管核准與慣例成交條件。 諾華這次沒有急著追熱鬧。它像是等到 ADC 競爭進入下一階段，才出手買一張新規則的門票。 藥時事團隊的判斷是：諾華買 Myricx，表面上拿到 B7-H3 與 HER2 兩條 lead assets，真正的戰略位置在新型 payload。ADC 的下一場戰爭，會從「靶點誰更熱門」推進到「載荷能不能換一種臨床語言」。誰能把耐藥、安全窗與可複製平台一起講清楚，誰才可能在下一輪 BD 裡被重新定價。 ## 01｜ADC 的老問題：好靶點不等於好藥 ADC 常被形容成生物導彈：抗體負責找癌細胞，連接子負責控制釋放，payload 負責殺傷。 這個比喻很好懂，但也容易讓人忽略一件事：導彈打得準，不代表彈頭一定夠好。 現有 ADC 的痛點，大多最後都會繞回三個問題。 ⚠️ 第一，安全窗不夠寬。Payload 太毒，劑量上不去；劑量上不去，腫瘤裡的有效殺傷濃度就受限。臨床上常見的減量、延遲給藥、停藥，都在提醒投資人：靶點打得準還不夠，彈頭撐不住，臨床仍會被劑量卡住。 🧬 第二，耐藥會慢慢出現。Topo I inhibitor、微管抑制劑這些主流 payload 已經證明有價值，但癌細胞不會站著挨打。靶點表現下降、藥物外排、DNA repair 改變、TOP1 相關耐藥，都可能讓療效變薄。 📌 第三，機制太集中。當大量公司用相似靶點、相似 payload、相似 linker 去打相似病人，市場最後會追問：你到底多解了哪一個臨床問題？ 這也是 Myricx 會被諾華看上的原因。 它沒有拿一套更漂亮的簡報去說自己也是 ADC。它把戰場往 payload 底層移了一格。 ## 02｜NMTi-ADC：諾華買的是一種新的殺傷邏輯 Myricx 的核心，是以 N-myristoyltransferase inhibitor, NMTi 作為 payload 的 ADC 平台。 NMT 是細胞內蛋白功能運作的重要酶，參與蛋白肉豆蔻醯化修飾。這些修飾會影響部分蛋白的定位、活性與訊號傳遞，而癌細胞的生長與存活，往往依賴這些細胞內機制。Myricx 的想法，是把 NMTi 做成 ADC payload，透過抗體把它送到腫瘤細胞，再干擾癌細胞賴以運作的關鍵程序。 真正值得注意的地方，在於它和主流 ADC payload 的作用邏輯不同。 如果 Topo I payload 像是在 DNA 複製與修復上施壓，微管抑制劑像是在細胞分裂機器上卡住齒輪，NMTi payload 想切的是另一組更上游、更不一樣的細胞功能依賴。這種機制差異，才是諾華願意花 11 億美元 upfront 的核心。 官方公告也把話說得很清楚：Myricx 的 NMTi payload 是為了處理現有 payload 類別的限制，臨床前資料顯示它可能對多種實體瘤有廣泛活性，包含 Topo I resistant models。 但這裡要踩住煞車。 目前最能確認的是臨床前潛力。NMTi-ADC 能不能真正把耐藥、安全窗、療效延伸變成臨床優勢，還要等人體數據回答。ADC 產業史上不缺漂亮的臨床前故事，真正難的是進到病人身上，還能同時保住療效和耐受性。 所以這筆交易的精彩之處，不在於 Myricx 已經證明一切。 精彩的是：諾華願意在人體資料尚未完全展開前，用大額 upfront 先卡住一個可能成為新 payload 類別的平台。 ## 03｜諾華的 ADC 態度：等到差異化才下手 這筆交易放在諾華身上很有意思。 過去諾華追 ADC 的速度不算激進。當 ADC 交易滿天飛時，諾華沒有急著買一堆長相接近的資產。這一次它出手，反而更能看出選擇標準：ADC 三個字本身不夠值錢，差異化機制才值錢。 Myricx 讓諾華一次拿到三層東西。 📌 第一，是兩條 lead ADC assets。官方公告提到，Myricx 的兩個 lead assets 指向 B7-H3 與 HER2，兩個靶點都落在腫瘤 ADC 競爭密集、但臨床需求仍大的區域。 📌 第二，是 NMTi payload platform。這比單一產品更重要，因為 payload 若能成立，未來就可能接到更多抗體、更多靶點、更多實體瘤場景。 📌 第三，是 oncology platform 的延伸選擇權。諾華這幾年在放射配體療法、腫瘤標靶與平台型資產上都很重視可複製能力。Myricx 若成功，會變成諾華 ADC 版圖裡一個新的底層模組。 這也讓 ADC 的估值邏輯變得更尖銳。 以前市場看到 ADC，會先問靶點夠不夠熱門、腫瘤族群夠不夠大、早期數據有沒有 response。 下一輪會多問幾個問題： 📌 payload 是否真的和主流機制錯開？ 📌 linker 和 payload 的安全窗能不能支撐足夠劑量？ 📌 面對既有 ADC 失敗或耐藥病人，是否有合理的臨床位置？ 📌 平台能不能從一條管線複製到多個靶點？ 📌 大藥廠買下來之後，能不能放進既有腫瘤開發與商業化系統？ 這些問題，比 ADC 題材本身重要太多。 ## 04｜風險：payload 新，不代表臨床一定贏 新型 payload 很迷人，也很危險。 迷人的地方在於，它可能打開現有 ADC 解不了的病人族群。尤其當 Enhertu、Trodelvy、Padcev 這類產品逐漸把 ADC 推成標準治療的一部分，下一代資產如果能在耐藥、低表達、多癌種或聯用場景裡跑出數據，價值會非常大。 危險的地方也在這裡。 Payload 是 ADC 的殺傷核心，新的機制往往意味著新的不確定性。它可能帶來更好的腫瘤殺傷，也可能帶來未預期毒性；可能避開既有耐藥，也可能在人體暴露、組織分布、旁觀者效應、代謝與安全性上遇到新問題。 對 Myricx 來說，真正的考試至少有三道。 ⚠️ 第一，NMTi payload 在人體裡是否真的能拉開治療窗。臨床前模型能看方向，但人體安全性才會決定劑量天花板。 ⚠️ 第二，B7-H3 與 HER2 lead assets 是否能跑出足夠有說服力的早期臨床訊號。單靠靶點很難撐起一條 ADC，抗體、linker、payload、DAR、腫瘤表現與病人選擇都會一起決定結果。 ⚠️ 第三，平台能不能被複製。單一 asset 成功很重要，但 15 億美元交易真正想買的是平台。如果 NMTi 只適用少數管線，價值會被打折；如果能跨靶點成立，估值邏輯就完全不同。 這也是這筆交易對整個 ADC 產業的提醒：下一代 ADC 光把名稱變複雜不夠，臨床問題要解得更清楚。 ## 05｜台灣怎麼看：浩鼎、諦醫、嘉正，要問三種問題 放回台灣，這題先把兩件事分開：別把浩鼎當成唯一答案，也別把所有抗體公司塞進 ADC 名單。 這篇台灣段落，藥時事團隊收斂到三個名字：浩鼎、諦醫生技、嘉正。 浩鼎是最容易用公開資料核對的直接 ADC 公司。官方管線頁列出多條 ADC 相關管線，包括 TROP2-targeted ADC OBI-992、下一代 TROP2 ADC OBI-902、Nectin-4 ADC OBI-904，以及 TROP2 x HER2 bispecific ADC OBI-201。公司技術平台頁也說明 Obrion ADC technology platform，涵蓋 GlycOBI、HYPrOBI、ThiOBI、GlycOBI DUO 等 site-specific conjugation、linker 與 dual-payload 相關技術。 浩鼎這一格，不能停在「有 ADC」四個字。接下來要看 OBI-992、OBI-902 這類資產能不能交出人體安全性、初步療效、劑量與病人選擇資料；GlycOBI / Obrion 也要靠外部合作、後續管線與資料累積，證明平台可複製。 第二個名字，是諦醫生技。 這是這一輪台灣 ADC 名單裡最該補上的公司。免疫功坊官網 2025 年 11 月 7 日資訊提到，諦醫生技專注抗體藥物複合物 ADC 與抗體放射性核種複合物 ARC，核心是自主研發的 multi-arm linker 技術平台，搭配 mono- or dual-drug bundle 與模組化偶聯反應，目標是做出高 DAR、高產率、高純度、載藥彈性更大的新一代抗體偶聯藥物。 官方頁也列出兩條核心候選藥。 TE-1146 是抗 CD38 抗體 daratumumab 偶聯 Lenalidomide 的 ADC，適應症方向包括多發性骨髓瘤與相關疾病。該頁提到，TE-1146 在 2026 年上半年預計提交 IND，且未來在台灣、東南亞及紐澳地區商業化權利已以新台幣 2.7 億元授權予友華生技。 TE-1182 則是 HER2 腫瘤方向的 dual-payload ADC，設計目標放在 HER2 低表現或後線抗藥性場景，官方頁提到預計 2026 年第四季提交 IND。 諦醫放進這張地圖，意義很清楚：熟面孔之外，台灣 ADC 也開始出現 linker、dual payload、候選藥與區域授權組合在一起的早期創新案例。它現在最大的考題還在前面：IND 能否如期、CMC 與毒理能否支撐進臨床、第一個人體資料能不能把「平台故事」推進成「臨床可讀訊號」。 第三個名字，是嘉正。 嘉正先放進讀者雷達，但目前不把它寫成已確認 ADC 管線。這次重新查核後，藥時事團隊還沒有找到足以在正式公開稿引用的官方 ADC 管線頁、明確靶點、payload / linker、IND 節點或授權公告。這種公司可以提醒讀者追，但不能靠市場印象把它寫滿。 所以嘉正這一格，真正要追的是資料補齊：公司是否公開 ADC 產品、靶點與 payload 設計？是否有臨床前資料、專利、簡報、授權合作或臨床登錄？只要這些資料補上，它就能進入正式比較表；在那之前，放在早期雷達，比硬寫成結論更負責。 這三家公司剛好把台灣 ADC 的看法拉回正軌。 浩鼎，看公開管線與平台能不能交出資料。 諦醫，看早期 ADC 技術與候選藥能不能推進人體臨床。 嘉正，看公開資料能不能補到足以讓市場正式定價。 諾華買 Myricx 給台灣的提醒也在這裡：光列公司名稱，讀者看不到價值差異。ADC 要拆成技術、資料與合作三條線，才看得出哪家公司有機會把差異化技術、臨床資料、平台可複製性與外部合作接起來。 ## 結語｜ADC 不缺下一個跟隨者，缺下一種答案 諾華 15 億美元收購 Myricx Bio，表面看是又一筆 ADC 交易。 但這筆交易真正照亮的，是 ADC 產業的下一個分水嶺。 第一波 ADC 熱潮，市場看的是靶點、response rate、適應症與大藥廠 BD。下一波競爭會更挑剔：payload 是否有新機制，安全窗是否夠寬，耐藥問題是否能被打開，平台是否能跨靶點複製。 ADC 的下一場戰爭，不會只比誰把抗體接得更漂亮。 它會比誰能把 payload 換成新的臨床語言。 Myricx 還沒有走完人體臨床考試，所以這筆交易不能被解讀成 NMTi-ADC 已經勝利。比較準確的說法是：諾華提前買下了一個可能改變 ADC payload 版圖的選擇權。 對全球生技產業來說，這代表 ADC 競爭正在從「跟上熱門賽道」進入「證明底層創新」。 對台灣投資人來說，問題可以問得更準：浩鼎能不能把平台和管線資料往前推，諦醫能不能把早期技術送進人體臨床，嘉正能不能補出足夠公開資料。誰的資料能被國際市場讀懂，誰才有機會解決 ADC 下一階段的痛點。 當賽道開始擁擠，平庸資產會被折價。 真正稀缺的，是能讓大藥廠願意提早買單的新答案。 ## 免責聲明 本文為產業研究與市場觀察，不構成任何投資建議、買賣建議、醫療建議或個股推薦。生技醫藥投資涉及臨床試驗、法規審查、授權談判、商業化競爭、資本市場波動與公司營運風險，投資人應自行判斷並自負盈虧。 藥時事官網：https://drugnews.com.tw/