AAI Internal Wiki

ADC FIH Case Studies: B7-H3, HER2, and TROP-2

閱讀時間約 16 分鐘

ADC FIH Case Studies: B7-H3, HER2, and TROP-2

ADC 首次人體試驗案例：B7-H3、HER2 與 TROP-2

English

Concrete trials teach better than abstract frameworks. Three recent ADC first-in-human studies — targeting B7-H3, HER2, and TROP-2 respectively — each expose a different facet of the dose optimization challenge that cannot be learned from a toxicity table alone. Taken together, they illustrate why the three-box model of “find the DLT, select the MTD, proceed to phase 2” has been retired in favor of something more demanding: building a mechanistic case for every dose decision.

The YL201 B7-H3 ADC story: biomarker heterogeneity and the five-component audit

YL201 is a B7-H3-targeting ADC studied in a phase 1/1b trial published in Nature Medicine in 2025. B7-H3 (also known as CD276) is widely expressed across multiple solid tumor types, making it an attractive pan-tumor target. The trial enrolled 312 patients with advanced solid tumors and tested dose escalation plus expansion cohorts. The MTD was established at 2.8 mg/kg, and the recommended expansion doses were 2.0 and 2.4 mg/kg every 3 weeks. The most common grade 3 or higher treatment-related adverse events were neutropenia, leukopenia, and anemia — consistent with ADC-related myelosuppression. Interstitial lung disease (ILD) occurred at a low but clinically significant rate, signaling the class-specific pulmonary risk that all ADC programs must actively monitor rather than reactively document.

The finding that makes this trial particularly valuable for teaching is the absence of correlation between B7-H3 membrane expression by IHC and ORR. Students trained to expect that “antigen high = response high” will struggle with this result. But it is not actually surprising when you understand ADC biology. Target expression is only the first gate. The actual efficacy depends on internalization efficiency (how quickly the antigen-ADC complex is taken into the cell), linker stability in the tumor microenvironment, payload bystander effect (whether the released payload can kill adjacent cells that did not directly bind the ADC), and competition with immunosuppressive signals in the tumor. B7-H3 expression by IHC captures only the first variable. The clinical implication is that ADC biomarker programs cannot stop at a simple IHC positivity cutoff; they need to interrogate the full pathway from antigen binding to payload activity.

Ifinatamab deruxtecan: MTD not reached is not a license for escalation

Ifinatamab deruxtecan is another B7-H3-directed ADC but with a deruxtecan payload (topoisomerase I inhibitor) rather than a classic MMAE warhead. The IDeate-PanTumor01 trial, published in Lancet Oncology in 2026, enrolled 97 patients across multiple solid tumor types at doses ranging from 0.8 to 16.0 mg/kg every 3 weeks. Only 3 patients had dose-limiting toxicities. By protocol-defined criteria, the MTD was not reached.

This situation — no MTD, early efficacy signal, multi-tumor-type enrollment — is actually the hardest teaching case in modern ADC oncology, because it removes the natural brake that a DLT provides. The confirmed ORR in evaluable patients was 34%, suggesting meaningful activity. Yet 32% of patients experienced serious treatment-emergent adverse events, and there was one fatal treatment-related ILD. The ILD finding is critical because deruxtecan-class ADCs carry a known class risk of pulmonary toxicity: this is not an unexpected finding, it is a predictable class consequence that must be placed at the center of the dose selection framework, not relegated to a footnote in the safety table.

When reading this trial for a journal club or protocol review, the productive questions are: What were the intact ADC and unconjugated payload exposure-response relationships across the 0.8 to 16.0 mg/kg range? Was there an exposure threshold above which ILD risk increased? What dose modification rules and monitoring protocols were in place for early ILD detection? Did the investigators distinguish between tumor-type subgroups in their efficacy analysis, given that benefit-risk differs dramatically between a cancer with no other options and one with effective standard therapy? And critically: if the MTD was not reached, what affirmative evidence base supports the selection of any specific dose for the next phase?

SHR-A1811 HER2 ADC: pan-tumor expansion requires per-cohort benefit-risk

SHR-A1811 is a trastuzumab-backbone ADC with a topoisomerase I inhibitor payload targeting HER2. The HORIZON-X global phase 1 trial enrolled 396 patients across HER2-expressing and HER2-mutated advanced solid tumors at doses from 1.0 to 8.0 mg/kg every 3 weeks. The 2026 long-term follow-up analysis found grade 3 or higher treatment-related adverse events in 65.9% of patients and any-grade ILD in 2.5%. The efficacy landscape was dramatically heterogeneous across tumor types: HER2-positive breast cancer achieved a median PFS of approximately 25.0 months, HER2-low breast cancer approximately 11.0 months, and non-breast solid tumors ranged from 3.5 to 17.2 months.

This heterogeneity is the teaching lesson. A pan-tumor HER2 ADC FIH trial produces a pooled ORR that can look attractive, but that number is the average of very different clinical situations. Using one RP2D for a disease where median PFS is over two years and another where it is three and a half months demands that you ask: is the toxicity burden — including 65.9% grade 3+ adverse events and 2.5% ILD — acceptable in the context of a three-month median benefit? The answer is different for different patient populations, and the ADC community is still learning how to do this accounting properly. The companion diagnosis question also matters: HER2-positive, HER2-low, and HER2-mutated represent biologically distinct populations with different expected antibody binding, internalization, and payload delivery — they should not be assumed to require identical dosing.

SHR-A1921 TROP-2 ADC: target expression that does not predict response

SHR-A1921 is a TROP-2-targeted ADC that enrolled 391 patients with advanced or metastatic solid tumors across a three-stage phase 1 trial at doses from 1.5 to 6.0 mg/kg every 3 weeks. Grade 3 or higher treatment-related adverse events occurred in 33.8%, with stomatitis (oral mucositis) as the most common at 14.6% — a toxicity signature different from the myelosuppression and neuropathy profiles of other payload classes. Overall ORR was approximately 24.8%, with cohort-specific responses from 18.2% to 43.1% across cervical cancer, triple-negative breast cancer, small-cell lung cancer, NSCLC, and others. The investigators selected 3.0 mg/kg every 3 weeks for further development. Critically, they did not find a meaningful correlation between TROP-2 expression and efficacy.

The TROP-2 finding mirrors the B7-H3 lesson from YL201: target expression measured by IHC is not a reliable surrogate for the full delivery pathway. TROP-2 expression may vary with prior treatment, be heterogeneous within a tumor, differ between primary and metastatic sites, and be measured with assay platforms that are not equivalent. The bystander effect of the payload — meaning its ability to kill adjacent cells that did not directly capture the ADC — may also reduce the dependence on homogeneous target expression. When target expression does not predict response, the next layer of questions must be: which tumor types consistently respond, what is the payload’s known sensitivity pattern, is there a combination strategy that could sensitize, and what would a randomized dose-expansion comparison tell you about optimal dosing across responding subtypes?

中文

具體的臨床試驗比抽象框架教得更好。三個近期 ADC 首次人體試驗——分別靶向 B7-H3、HER2 和 TROP-2——各自揭示了僅從毒性表格無法學到的劑量最佳化挑戰的不同面向。整合來看，它們說明了為什麼「找到 DLT、選擇 MTD、進入二期」的三格模型已被一個更嚴格的要求所取代：為每個劑量決策建立機轉性論述。

YL201 B7-H3 ADC：生物標記異質性與五組分審計

YL201 是靶向 B7-H3 的 ADC，2025 年發表於 Nature Medicine 的一期/一期 b 研究中。B7-H3（又稱 CD276）在多種實體腫瘤類型中廣泛表達，使其成為吸引人的泛腫瘤標靶。試驗納入 312 位晚期實體腫瘤病人，測試劑量升量加擴增 cohort。MTD 確立為 2.8 mg/kg，建議擴增劑量為每 3 週 2.0 和 2.4 mg/kg。最常見的 3 級以上治療相關不良事件為嗜中性球低下、白血球低下和貧血——與 ADC 相關骨髓抑制一致。ILD 以低但臨床顯著的比例出現，標示所有 ADC 計畫必須主動監測而非被動記錄的類別肺毒性風險。

讓這個試驗對教學特別有價值的發現，是 B7-H3 膜表達（IHC）與 ORR 之間缺乏相關性。被訓練期待「抗原高 = 反應高」的學員會對此結果感到困惑。但當你理解 ADC 生物學時，這實際上並不令人驚訝。標靶表達只是第一道關卡。實際療效取決於內吞效率（抗原-ADC 複合物進入細胞的速度）、腫瘤微環境中的連接子穩定性、payload 旁觀者效應（釋放的 payload 是否能殺死未直接結合 ADC 的鄰近細胞），以及腫瘤免疫抑制訊號的競爭。IHC 的 B7-H3 表達只捕捉第一個變數。臨床意涵是：ADC 生物標記計畫不能停在簡單的 IHC 陽性截斷值；需要探究從抗原結合到 payload 活性的完整路徑。

Ifinatamab deruxtecan：MTD 未達不是升量的許可證

Ifinatamab deruxtecan 是另一個 B7-H3 標靶 ADC，但 payload 為 deruxtecan（拓樸異構酶 I 抑制劑）而非經典 MMAE 彈頭。IDeate-PanTumor01 試驗於 2026 年發表在 Lancet Oncology，在多種實體腫瘤類型中納入 97 位病人，劑量從 0.8 到 16.0 mg/kg 每 3 週。只有 3 位病人出現 DLT。按協議定義標準，MTD 未達到。

這種情況——無 MTD、早期療效訊號、多腫瘤類型入組——實際上是現代 ADC 腫瘤學中最難的教學案例，因為它去除了 DLT 提供的天然制動器。可評估病人的確認 ORR 為 34%，提示有意義的活性。然而 32% 病人出現嚴重治療緊急不良事件，並有 1 例治療相關 ILD 死亡。ILD 發現至關重要，因為 deruxtecan 類 ADC 具有已知的類別肺毒性風險：這不是意外發現，而是可預測的類別後果，必須放在劑量選擇框架的中心，而不是在安全性表格的腳注中。

閱讀這個試驗進行期刊俱樂部或方案審查時，有效的問題是：在 0.8 到 16.0 mg/kg 範圍內，完整 ADC 和游離 payload 的暴露—反應關係是什麼？是否存在 ILD 風險增加的暴露閾值？早期 ILD 偵測的劑量修改規則和監測方案是什麼？研究者是否在療效分析中區分了腫瘤類型亞組，考慮到不同癌種的效益—風險大相徑庭？以及關鍵問題：如果 MTD 未達到，什麼積極的證據基礎支持選擇特定劑量進入下一期？

SHR-A1811 HER2 ADC：泛腫瘤擴增需要每 cohort 的效益—風險評估

SHR-A1811 是靶向 HER2 的拓樸異構酶 I 抑制劑 payload ADC（使用 trastuzumab 骨架）。HORIZON-X 全球一期試驗在 HER2 表達和 HER2 突變的晚期實體腫瘤中納入 396 位病人，劑量每 3 週 1.0 到 8.0 mg/kg。2026 年長期追蹤分析發現 65.9% 病人有 3 級以上治療相關不良事件，任何級別 ILD 為 2.5%。療效格局在腫瘤類型間差異顯著：HER2 陽性乳癌中位 PFS 約 25.0 個月，HER2-low 乳癌約 11.0 個月，非乳癌實體腫瘤為 3.5 到 17.2 個月。

這種異質性是教學的核心。泛腫瘤 HER2 ADC FIH 試驗產生的匯總 ORR 看起來可能很吸引人，但那個數字是非常不同臨床情況的平均值。對中位 PFS 超過兩年的疾病和只有三個半月的疾病使用同一個 RP2D，要求你問：毒性負擔——包括 65.9% 3 級以上不良事件和 2.5% ILD——在三個月中位獲益的情境下可接受嗎？不同病人族群的答案不同，ADC 界仍在學習如何正確進行這種計算。伴隨診斷問題也很重要：HER2 陽性、HER2-low 和 HER2 突變代表生物學上不同的族群，具有不同的預期抗體結合、內吞和 payload 遞送——不應假設它們需要相同劑量。

SHR-A1921 TROP-2 ADC：不預測反應的標靶表達

SHR-A1921 是靶向 TROP-2 的 ADC，在三階段一期試驗中納入 391 位晚期或轉移性實體腫瘤病人，劑量每 3 週 1.5 到 6.0 mg/kg。3 級以上治療相關不良事件發生率 33.8%，最常見為口腔炎 14.6%——這是與其他 payload 類別的骨髓抑制和神經病變不同的毒性特徵。整體 ORR 約 24.8%，各 cohort 特異性反應為 18.2% 到 43.1%（宮頸癌、三陰性乳癌、小細胞肺癌、非小細胞肺癌等）。研究者選擇每 3 週 3.0 mg/kg 進一步開發。關鍵是，他們沒有發現 TROP-2 表達與療效之間有意義的相關性。

TROP-2 的發現與 YL201 的 B7-H3 教訓相呼應：IHC 測量的標靶表達不是完整遞送路徑的可靠替代物。TROP-2 表達可能因先前治療而改變、在腫瘤內部存在異質性、在原發和轉移部位之間不同，並且由不等同的分析平台測量。Payload 的旁觀者效應——即殺死未直接捕獲 ADC 的鄰近細胞的能力——也可能降低對均質標靶表達的依賴。當標靶表達不預測反應時，下一層問題必須是：哪些腫瘤類型持續有反應，payload 的已知敏感性模式是什麼，是否有能敏感化的聯合策略，以及隨機劑量擴增比較能告訴你在有反應亞型中的最佳劑量是什麼？

Key Concepts | 核心概念

  • B7-H3 IHC vs. ORR disconnect: Target expression alone does not predict ADC response; internalization, linker, payload bystander all matter
  • MTD not reached requires affirmative evidence | MTD 未達需積極證據: Especially when class toxicity (ILD) is known and lethal events occur
  • Pan-tumor HER2 expansion | 泛腫瘤擴增: Pooled ORR hides dramatically different benefit-risk by tumor type — must be read per cohort
  • TROP-2 expression-response disconnect: Bystander effect may reduce dependence on homogeneous target expression
  • ILD monitoring in deruxtecan-class ADCs | ILD 監測: New cough, dyspnea, or infiltrates → suspend ADC, rule out infection, early steroids if drug-related

Related Pages | 相關頁面

關係圖譜

反向連結