AAI Internal Wiki

ADC Dose Selection and Regimen Optimization

閱讀時間約 12 分鐘

ADC Dose Selection and Regimen Optimization

ADC 劑量選擇與給藥方案最佳化

English

The MTD is not the destination. For antibody-drug conjugates, Project Optimus — the FDA initiative that challenged the entire oncology field to stop treating the maximum tolerated dose as the automatically optimal dose — has found its clearest application. The argument is not that MTD is always too high. The argument is that for ADCs, where payload toxicities accumulate over cycles, where different schedules produce meaningfully different peak-to-trough ratios, and where the therapeutic window is often defined not by acute DLT but by cumulative tolerability over months, a dose selected only on the basis of a 28-day DLT window is almost certainly under-characterized.

The most instructive teaching case for this shift in thinking is zelenectide pevedotin (BT8009), a Nectin-4 targeting Bicycle Drug Conjugate with an MMAE payload. In the Duravelo-1 phase 1/2 study, the investigators did something unusual: they directly compared three schedules in 49 patients with Nectin-4-relevant advanced solid tumors. They tested 2.5, 5.0, and 7.5 mg/m2 given weekly; 7.5 mg/m2 on days 1 and 8 of a 21-day cycle; and 7.5 or 10.0 mg/m2 every two weeks. The MTD was identified as 7.5 mg/m2 every 2 weeks. But the recommended phase 2 doses were 5.0 mg/m2 weekly and 7.5 mg/m2 on days 1 and 8. This is the distinction that Project Optimus demands: the MTD and the RP2D are different clinical concepts, and the RP2D requires affirmative evidence of durable tolerability, not merely the absence of acute DLT. The overall ORR was 24%, with a 38% ORR in the urothelial carcinoma subgroup, providing enough efficacy signal to warrant the schedule-comparative design.

When the MTD is not reached, the dose selection problem becomes harder, not easier. SHR-A1904, a CLDN18.2-targeting ADC with a topoisomerase I inhibitor payload, enrolled 95 patients with previously treated CLDN18.2-positive advanced gastric or gastroesophageal junction cancer. Dose escalation from 0.6 to 8.0 mg/kg produced DLTs at 4.8 mg/kg (febrile neutropenia, bilirubin increase) and 6.0 mg/kg (gastric mucosal lesion), but the protocol-defined MTD was never reached. The investigators then enrolled expansion cohorts at 6.0 and 8.0 mg/kg. The confirmed ORRs were approximately 24% and 25% respectively — nearly identical — and drug-related grade 3 or higher adverse events occurred in 62.1% of patients. The pedagogical question this case forces is: when two doses produce equal efficacy and one produces substantially more toxicity, what additional evidence would justify taking the higher dose into phase 2? Not intuition and not protocol inertia, but explicit exposure-response data, dose modification rates, treatment duration, and patient-reported tolerability.

Model-informed drug development (MIDD) has emerged as a legitimate tool for filling these evidence gaps without running infinite parallel cohorts. The 2026 datopotamab deruxtecan (Dato-DXd) analysis in Clinical and Translational Science built a tumor growth inhibition-to-progression-free-survival model linking PK exposure, tumor size dynamics, and PFS outcomes in HR-positive/HER2-negative breast cancer. Virtual trial simulations from this model supported 6 mg/kg every 3 weeks over lower doses. The important teaching moment here is what the model can and cannot do: it can extend inference from limited clinical data and make dose comparisons explicit where head-to-head randomization is impractical. What it cannot do is replace actual tolerability data showing whether real patients can stay on 6 mg/kg over multiple cycles without dose reduction or treatment discontinuation for stomatitis, nausea, ocular toxicity, or pneumonitis. MIDD amplifies evidence; it does not substitute for it.

The path from FIH to FDA approval is longer than the MTD story implies. Telisotuzumab vedotin provides a complete teaching arc. The drug is a c-Met-directed ADC with an MMAE payload, approved by the FDA in May 2025 for previously treated, locally advanced or metastatic non-squamous NSCLC with high c-Met protein overexpression — defined as at least 50% of tumor cells with strong 3+ staining on a companion diagnostic. The approved dose, 1.9 mg/kg every 2 weeks (capped at 190 mg for body weight over 100 kg), was derived from phase 1 FIH escalation, phase 2 efficacy data in the high c-Met cohort (confirmed ORR approximately 35%, median duration of response approximately 7.2 months), and a 2026 population PK/exposure-response analysis integrating both phases. That analysis found that higher ADC conjugate exposure was associated with improved ORR, PFS, and OS — but also with grade 2 or higher peripheral neuropathy and corneal epitheliopathy. The dose reflects a balance: high enough for meaningful efficacy, low enough for chronic tolerability. The companion diagnostic requirement reflects the biomarker lesson: not all c-Met-expressing patients are equivalent, and dose optimization and patient selection are inseparable questions.

The FOR46 (FG-3246) case adds another dimension: translational endpoints within FIH. This CD46-targeting ADC with an MMAE payload in metastatic castration-resistant prostate cancer established an MTD of 2.7 mg/kg adjusted body weight, with neutropenia as the primary grade 3+ adverse event. PSA50 response in the evaluable adenocarcinoma subset was 36%, confirmed ORR 20%. But the investigators also conducted whole-blood mass cytometry and found that responders showed higher on-treatment circulating effector CD8 T cells. This signals a maturing expectation: ADC FIH trials are not purely toxicology experiments with incidental efficacy data. They can and should generate mechanism-relevant pharmacodynamic evidence that informs both the dose selection rationale and the enrichment strategy for subsequent trials.

中文

MTD 不是終點。對抗體藥物複合體（ADC）來說，Project Optimus——FDA 挑戰整個腫瘤學領域停止把最大耐受劑量（MTD）視為自動最佳劑量的倡議——找到了其最清晰的應用場合。這個論點不是說 MTD 一定太高；而是說對 ADC 而言，payload 毒性會隨週期累積、不同時程產生顯著不同的峰谷比、且治療窗口往往不由急性 DLT 而是由月份累積耐受性定義——僅基於 28 天 DLT 窗口選擇的劑量幾乎必然特徵描述不足。

這個思維轉變最有說服力的教學案例是 zelenectide pevedotin（BT8009），一個 Nectin-4 標靶 Bicycle Drug Conjugate，payload 為 MMAE。在 Duravelo-1 一期/二期研究中，研究者做了一件不尋常的事：在 49 位 Nectin-4 相關晚期實體腫瘤病人中直接比較三種時程。測試 2.5、5.0、7.5 mg/m² 每週；7.5 mg/m² 第 1 天和第 8 天每 21 天；以及 7.5 或 10.0 mg/m² 每 2 週。MTD 確認為每 2 週 7.5 mg/m²。但建議第二期劑量（RP2D）選擇了每週 5.0 mg/m² 和第 1/8 天 7.5 mg/m²。這正是 Project Optimus 要求的區分：MTD 和 RP2D 是不同的臨床概念，RP2D 需要持久耐受性的積極證據，而不僅僅是沒有急性 DLT。整體 ORR 為 24%，尿路上皮癌亞組為 38%，提供足夠的療效訊號來支持時程比較設計。

當 MTD 未達到時，劑量選擇問題變得更難而非更容易。SHR-A1904，一個靶向 CLDN18.2 的 ADC（payload 為拓樸異構酶 I 抑制劑），納入 95 位已接受治療的 CLDN18.2 陽性晚期胃癌或胃食道交界癌病人。劑量從 0.6 到 8.0 mg/kg 的升量在 4.8 mg/kg（發熱性嗜中性球低下、膽紅素升高）和 6.0 mg/kg（胃黏膜損傷）出現 DLT，但協議定義的 MTD 從未達到。研究者隨後在 6.0 和 8.0 mg/kg 開設擴增 cohort。確認 ORR 分別約 24% 和 25%——幾乎相同——而 3 級以上藥物相關不良事件發生率為 62.1%。這個案例強迫的教學問題是：當兩個劑量產生相等療效、其中一個產生更多毒性時，進入二期需要什麼額外證據？不是直覺，不是試驗慣性，而是明確的暴露—反應資料、劑量修改率、治療持續時間和病人自報耐受性。

模型輔助藥物開發（MIDD）已成為在不運行無限平行 cohort 的情況下填補這些證據缺口的合法工具。2026 年 datopotamab deruxtecan（Dato-DXd）在 Clinical and Translational Science 的分析，建立了連接 PK 暴露、腫瘤大小動態和 PFS 結果的腫瘤生長抑制—無惡化存活模型，應用於荷爾蒙受體陽性/HER2 陰性乳癌。虛擬試驗模擬支持每 3 週 6 mg/kg 優於較低劑量。這裡重要的教學時刻是模型能做和不能做的事：它可以從有限臨床資料延伸推斷，讓頭對頭隨機化不實際時的劑量比較顯式化。它不能做的是替代真實的耐受性資料——顯示真實病人能否在多個週期保持每 3 週 6 mg/kg，而不因口腔炎、噁心、眼毒性或肺炎而減量或停藥。MIDD 放大證據；它不能取代證據。

從 FIH 到 FDA 核准的路徑比 MTD 故事所暗示的更長。Telisotuzumab vedotin 提供了完整的教學弧線：該藥是 c-Met 標靶 ADC（payload 為 MMAE），2025 年 5 月 FDA 加速核准，適應症為先前接受治療的局部晚期或轉移性非鱗狀 NSCLC，且腫瘤有高 c-Met 蛋白過表達（至少 50% 腫瘤細胞呈強 3+ 染色，需伴隨診斷）。核准劑量每 2 週 1.9 mg/kg（體重超過 100 kg 者上限 190 mg），來自一期 FIH 升量、高 c-Met cohort 的二期療效資料（確認 ORR 約 35%，中位反應持續時間約 7.2 個月），以及 2026 年整合兩期資料的群體 PK/暴露—反應分析。該分析發現較高 ADC 偶聯暴露與改善的 ORR、PFS 和 OS 相關——但也與 2 級以上周邊神經病變和角膜上皮病變相關。劑量反映一個平衡：高到足以產生有意義療效，低到允許長期給藥。伴隨診斷要求則反映了生物標記的教訓：並非所有 c-Met 表達病人都等同，劑量最佳化和病人選擇是不可分割的問題。

FOR46 案例增加了另一個維度：FIH 中的轉譯終點。這個靶向 CD46 的 ADC（payload 為 MMAE）用於轉移性去勢抗性攝護腺癌，建立了調整體重 2.7 mg/kg 的 MTD，主要 3 級以上不良事件為嗜中性球低下。可評估腺癌亞組的 PSA50 反應率 36%，確認 ORR 20%。但研究者還進行了全血質量細胞儀分析，發現有反應者在治療期間呈現更高的循環效應 CD8 T 細胞。這標誌著一個成熟的期望：ADC FIH 試驗不只是附帶療效資料的毒理學實驗；它們能且應該產生與機轉相關的藥效學證據，為劑量選擇理由和後續試驗的富集策略提供依據。

Key Concepts | 核心概念

  • MTD ≠ RP2D: Schedule comparison and durable tolerability are required to justify the recommended phase 2 dose
  • When MTD is not reached | MTD 未達時: Equal ORR at two doses means lower dose should be favored unless explicit evidence supports higher
  • MIDD role | MIDD 的角色: Extends dose evidence without infinite cohorts, but cannot substitute for actual tolerability observation
  • Companion diagnostics | 伴隨診斷: Patient selection and dose optimization are inseparable (telisotuzumab vedotin c-Met 3+ example)
  • Translational endpoints in FIH | FIH 中的轉譯終點: Immune cell phenotype changes can inform mechanism and enrichment strategy

Related Pages | 相關頁面

關係圖譜

反向連結