AAI Internal Wiki

ADC FIH: Clinical Pharmacology Framework

閱讀時間約 12 分鐘

ADC FIH: Clinical Pharmacology Framework

ADC 首次人體試驗：臨床藥理框架

English

An antibody-drug conjugate is not a single molecule. Understanding this sentence is the prerequisite for reading any ADC first-in-human trial properly. A conventional cytotoxic drug has one concentration you measure and one curve to interpret. An ADC is architecturally layered: there is the intact conjugated ADC, the total antibody (conjugated plus unconjugated), the unconjugated free payload released in circulation, and sometimes relevant metabolites. Each layer has its own pharmacokinetic behavior, its own relationship to efficacy, and its own relationship to toxicity. The FDA made this explicit in its 2024 final guidance on ADC clinical pharmacology, which requires sponsors to characterize “the whole molecule and its relevant constituent parts.” If a trial only reports total antibody concentrations, it is flying with one instrument in a storm that requires three.

The clinical pharmacology of an ADC is shaped by five structural decisions made during drug design: the target antigen, the antibody backbone, the linker chemistry, the payload class, and the drug-to-antibody ratio (DAR). Each of these creates predictable risk signatures. A topoisomerase I inhibitor payload — exemplified by deruxtecan, the warhead shared by trastuzumab deruxtecan, datopotamab deruxtecan, and ifinatamab deruxtecan — carries class-specific risks of myelosuppression, nausea, and interstitial lung disease. A microtubule inhibitor payload like monomethyl auristatin E (MMAE) tends to produce peripheral neuropathy and neutropenia. The DAR determines how much payload hangs on each antibody molecule; higher DAR can increase potency but may also accelerate antibody clearance and increase off-target payload exposure. The linker stability determines how much payload is released in circulation before the ADC reaches its target. Reading an ADC FIH paper without knowing these five parameters is like evaluating a surgical complication without knowing what operation was performed.

The FDA 2024 guidance codified six clinical pharmacology domains that reviewers scrutinize across all ADC programs: bioanalytical methods, exposure-response relationships, intrinsic factors (hepatic function, body weight, age, race), extrinsic factors (drug interactions), QTc assessment, and immunogenicity. A 2025 review of all 13 FDA-approved ADCs found that deficiencies in these areas — particularly exposure-response for safety signals, hepatic impairment data, and validated immunogenicity assays — frequently generated post-marketing requirements. This means that a beautiful ORR in phase 1 cannot compensate for an absent pharmacology evidence package: the reckoning simply arrives later, after approval.

The exposure-response framework for ADCs is inherently more complex than for small molecules or monoclonal antibodies, because efficacy and toxicity may each be driven by different analytes. In several approved ADCs, tumor response correlates best with intact ADC or conjugated antibody exposure, while payload-related toxicities such as peripheral neuropathy and myelosuppression correlate with unconjugated payload concentrations. This creates a situation where the same dosing decision simultaneously increases the probability of response and the probability of a specific toxicity, with the two curves not necessarily parallel. The dose optimization problem is therefore not “find the highest tolerable dose” but rather “find the exposure window where the efficacy curve exceeds the toxicity curve by enough margin to allow durable treatment.”

Body size and dosing metric matter more than they are usually taught. The FIH dose-expansion study of sigvotatug vedotin (an integrin beta-6-targeting ADC) compared three regimens directly: 1.25 mg/kg total body weight on days 1 and 8 of a 21-day cycle; 1.5 mg/kg total body weight every 2 weeks; and 1.8 mg/kg adjusted ideal body weight every 2 weeks. In 117 patients with advanced NSCLC, the adjusted ideal body weight regimen showed lower pharmacokinetic variability than the total body weight regimens, and the grade 3 or higher adverse event rate was 35% versus 48% overall. This illustrates that dosing metric — which “kg” you put in the denominator — is itself a dose optimization variable, not a pharmacy dispensing detail.

Immunogenicity and drug-drug interactions round out the pharmacology profile. Anti-drug antibodies can neutralize the ADC, accelerate its clearance, or alter the payload exposure profile, which is particularly dangerous if a trial interprets apparent PK variability as inter-patient biology rather than immune response. Payloads that are CYP3A4 substrates — which many MMAE-based ADCs are — create clinically significant interactions with azole antifungals and certain HIV medications, both commonly used in oncology patients. QT risk assessment for ADCs is best done with fit-for-purpose concentration-QTc analyses integrated into the FIH trial rather than with traditional thorough QT studies in healthy volunteers, because cytotoxic payloads are inappropriate for healthy subjects and cancer patients have baseline cardiovascular comorbidities.

The practical teaching message is this: when you read an ADC FIH paper, do not stop at the DLT table and the ORR headline. Ask which analytes were measured. Ask whether exposure was connected separately to efficacy and to toxicity. Ask whether the recommended phase 2 dose (RP2D) was chosen because of a coherent exposure-response argument or simply because it was the highest dose that did not hit a DLT. Ask whether hepatic impairment patients were studied, because payloads often have hepatic metabolism. Ask about immunogenicity. If these answers are missing from the paper, the dose is not optimized — it is merely untested at a lower level.

中文

抗體藥物複合體（ADC）不是單一分子。理解這句話，是正確閱讀任何 ADC 首次人體試驗（FIH）的前提。傳統細胞毒性藥物只有一條濃度曲線需要解讀；ADC 的結構卻是分層的——包含完整偶聯 ADC、總抗體（偶聯加未偶聯）、循環中釋放的游離 payload，有時還有代謝物。每一層有自己的藥動學行為，各自對療效和毒性的貢獻也不同。FDA 在 2024 年 ADC 臨床藥理最終指引中明確要求：試驗必須同時描述「整體分子與相關組成部分」的 PK/PD。如果一篇試驗只報告總抗體濃度，等於在需要三種儀器的暴風中只帶一種飛。

ADC 的臨床藥理由五個藥物設計決策決定：標靶抗原、抗體骨架、連接子化學性質、payload 類別，以及藥物抗體比（DAR）。每一個都創造可預測的風險特徵。Topoisomerase I inhibitor payload（如 deruxtecan 系列）帶來骨髓抑制、噁心和間質性肺病的類別風險；MMAE 類的微管抑制劑則傾向造成周邊神經病變和嗜中性球低下。DAR 決定每個抗體分子攜帶多少 payload；較高 DAR 可能提升效力，但也可能加速抗體清除並增加 off-target 暴露。連接子穩定性決定 payload 在到達腫瘤前有多少在血液中提早釋放。不知道這五個參數就去讀 ADC FIH 試驗，就像不知道做了什麼手術就去評估術後併發症。

FDA 2024 指引將 ADC 臨床藥理劃分為六個核心領域：生物分析方法、暴露—反應關係、內在因子（肝功能、體重、年齡、種族）、外在因子（藥物交互作用）、QTc 評估，以及免疫原性。2025 年對所有 13 個 FDA 核准 ADC 的回顧研究發現，這些領域的缺失——特別是安全性暴露—反應分析、肝功能不全資料，以及免疫原性驗證——頻繁導致上市後要求。換言之，一期試驗漂亮的 ORR 無法彌補臨床藥理證據的缺失；帳單只是推到核准後再算。

ADC 的暴露—反應關係天生比小分子或單株抗體複雜，因為療效和毒性可能各自受到不同分析物驅動。在幾個核准的 ADC 中，腫瘤反應與完整偶聯 ADC 或偶聯抗體的暴露最相關；而 payload 相關毒性——如周邊神經病變和骨髓抑制——則與游離 payload 濃度相關。這造成同一個劑量決策同時影響反應機率和毒性機率，且兩條曲線不一定平行。劑量最佳化的問題因此不是「找最高可耐受劑量」，而是「找到療效曲線足夠超越毒性曲線、允許長期給藥的暴露窗口」。

體型和給藥指標的影響比通常認為的更大。Sigvotatug vedotin（整合素 beta-6 標靶 ADC）的 FIH 擴增研究直接比較三種方案：1.25 mg/kg 總體重每 21 天第 1 和第 8 天、1.5 mg/kg 總體重每 2 週，以及 1.8 mg/kg 調整理想體重每 2 週。在 117 位晚期非小細胞肺癌病人中，調整理想體重方案的藥動學變異低於總體重方案，3 級以上不良事件發生率為 35%，而整體為 48%。這說明給藥指標——分母用哪個「公斤」——本身就是劑量最佳化變數，不只是藥局換算細節。

免疫原性和藥物交互作用構成藥理輪廓的最後一塊。抗藥抗體可以中和 ADC、加速清除或改變 payload 暴露，若試驗把 PK 變異歸因於個體間差異而非免疫反應，危險就在此。以 CYP3A4 為底物的 payload（許多 MMAE 類 ADC 如此）與唑類抗黴菌藥和某些 HIV 藥物產生臨床顯著交互作用——兩類都是腫瘤科病人常用藥物。ADC 的 QT 風險評估最好在 FIH 試驗中整合濃度—QTc 分析，而非在健康受試者做傳統的全面 QT 研究，因為細胞毒性 payload 不適合健康人，而腫瘤病人本身就有心血管共病。

實務教學的核心訊息是這樣的：讀 ADC FIH 論文時，不要停在 DLT 表和 ORR 標題。問量了哪些分析物；問暴露是否分別連到療效和毒性；問建議第二期劑量（RP2D）是基於清晰的暴露—反應論點，還是只因為它是沒碰到 DLT 的最高劑量；問是否研究了肝功能不全病人；問免疫原性。如果這些答案都不在論文裡，劑量沒有被最佳化——只是在較低劑量上未被測試。

Key Concepts | 核心概念

  • ADC analyte hierarchy | ADC 分析物層次: intact ADC → total antibody → unconjugated payload → metabolites；每層各有其 PK 意義
  • Five ADC design parameters | 五個設計參數: target antigen, antibody, linker, payload, DAR — 各自決定不同毒性風險
  • Six FDA pharmacology domains | FDA 六個藥理領域: bioanalysis, exposure-response, intrinsic factors, extrinsic factors, QTc, immunogenicity
  • Dosing metric matters | 給藥指標的重要性: total body weight vs. adjusted ideal body weight 影響 PK variability 和毒性
  • Payload class toxicity | Payload 類別毒性: topoisomerase I inhibitor → ILD/myelosuppression; MMAE → peripheral neuropathy

Related Pages | 相關頁面

關係圖譜

反向連結