Three Modern FIH Platforms: ADC, T-cell Engager, and Radiopharmaceutical Compared

三種現代首次人體試驗平台比較：ADC、T 細胞接合器與放射性藥物治療

English

The most common error in teaching first-in-human oncology trials is treating them as a generic category. A student who has learned the 3+3 design, knows what a DLT is, and can define BOIN versus CRM will still make systematic errors when they encounter an ADC FIH if they apply the small-molecule mental model. They will make different but equally systematic errors when they encounter a T-cell engager trial. And they will be largely lost when faced with an alpha-emitter radiopharmaceutical trial. The three platforms are not minor variations on a shared theme; they have different primary failure modes, different optimization targets, and different evidence architectures. Teaching them together within a shared comparative framework is more efficient and clinically more useful than teaching each independently.

What “dose” means across the three platforms

For a conventional cytotoxic drug or a small-molecule targeted agent, “dose” means milligrams administered at a point in time, and the primary pharmacokinetic question is what plasma concentration that dose produces over time. For an ADC, “dose” means milligrams per kilogram, but immediately requires disambiguation: which kilogram (total body weight, adjusted ideal body weight, adjusted body weight), which analyte (intact ADC, total antibody, unconjugated payload), and which time window matters for efficacy versus toxicity. For a T-cell engager, the mg/kg concept is largely preserved, but the relevant pharmacodynamic quantity is not plasma drug concentration per se — it is the degree of T-cell activation that the drug concentration produces, which depends on local tumor cell density, effector-to-target ratio, and immune microenvironment status. For a radiopharmaceutical, “dose” splits into administered activity (MBq), mass dose (µg), and organ-specific absorbed dose (Gy) — three separate numbers that cannot be collapsed into one.

Primary failure modes during dose escalation

Each platform’s characteristic failure mode during FIH is distinct. For ADCs, the most common failure mode is payload-related cumulative toxicity that is not visible in a single-cycle DLT window: peripheral neuropathy from MMAE accumulates over months; ILD from deruxtecan-class ADCs can emerge after the second or third cycle and be severe. The FIH design challenge is to capture these delayed signals during dose escalation, which requires longer follow-up per cohort and must pre-specify monitoring protocols for class-specific toxicities. For T-cell engagers, the most common early failure is acute CRS in the first one to three doses, driven by the immune synapse the drug was designed to create. The failure mode here is speed: CRS can escalate from grade 1 to grade 3 within hours if not monitored and treated promptly. The FIH design challenge is temporal — building a slow enough ramp-up to allow immune calibration while not condemning patients to months of subtherapeutic dosing. For radiopharmaceuticals, the failure mode is cumulative organ absorbed dose that exceeds the radiation tolerance of kidney, bone marrow, or salivary glands across multiple cycles. The DLT in cycle 1 is acute hematological toxicity, but the dose-limiting event for the entire treatment course may be late renal or marrow effects that emerge after 3-4 cycles.

The optimization target in each platform

When Project Optimus asks “what is the optimal dose,” the answer has different dimensions for each platform. For an ADC, optimization means finding the schedule (weekly, biweekly, days 1/8 of a q21d cycle), the dosing metric (TBW vs. AIBw), and the mg/kg that produces acceptable cumulative payload exposure while maintaining durable efficacy across cycles. The optimization is both a toxicity optimization (minimize neutropenia, ILD, neuropathy) and a pharmacokinetic optimization (minimize variability to make the dose reliably effective). For a T-cell engager, optimization means finding the step-up path that brings patients to the target dose without grade 3+ CRS, plus finding the target dose itself that produces the best response rate with acceptable on-treatment infection, hypogammaglobulinemia, and cytopenia burden. The optimization is temporal and immunological, not purely pharmacokinetic. For a radiopharmaceutical, optimization means finding the administered activity and cycle schedule that delivers the highest achievable tumor absorbed dose while keeping kidney, marrow, and other at-risk organ doses below radiation injury thresholds, across all cycles combined. The optimization is dosimetric and radiobiological, operating on a fundamentally different physical scale.

What “MTD not reached” means in each platform

All three platforms commonly fail to reach a protocol-defined MTD, and all three require different interpretive frameworks when this happens. For an ADC where MTD is not reached, the clinician should ask: Were the class-specific toxicities (ILD, neuropathy, myelosuppression) monitored with adequate follow-up? Are there exposure-response data connecting drug concentration to the relevant toxicities? Is the ORR at the tested doses consistent enough to justify a specific dose selection, or does it require a randomized dose comparison? For a T-cell engager where MTD is not reached, the clinician should ask: Was CRS documented by dose period (step-up vs. target dose)? Is there receptor occupancy or pharmacodynamic data supporting that the highest tested dose produces saturating biological effects? Did the trial investigate step-up regimen variants that might further reduce CRS risk? For a radiopharmaceutical where administered activity limits have not been reached, the clinician should ask: Were dosimetry measurements performed to estimate organ absorbed dose? Is the absence of DLT in a single-cycle window convincing evidence of multi-cycle safety? Are there delayed toxicity signals in the patients who have received the most cycles?

Evidence architecture for recommended phase 2 dose

The minimum evidence requirements for a defensible RP2D are platform-specific. For an ADC, a defensible RP2D requires: the proposed dose’s exposure-response relationship for both efficacy and key toxicities; a schedule rationale (why this regimen over alternatives); dosing metric justification (if adjusted body weight is used, PK variability data); organ impairment subgroup data for the critical metabolic pathway; and immunogenicity monitoring data if there is any ADA signal. For a T-cell engager, a defensible RP2D requires: starting dose rationale (MABEL or modified MABEL with assay justification); step-up regimen rationale with at least pre-specified simulation or comparison data; target dose CRS characterization by dose period; PK/PD data connecting drug concentration to T-cell activation and cytokine kinetics; and real-world tolerability data about the feasibility of the monitoring schedule. For a radiopharmaceutical, a defensible RP2D requires: dosimetry data from at least a subset of patients showing tumor absorbed dose and organ-at-risk absorbed dose; a fractionation rationale; delayed toxicity follow-up data beyond the standard DLT window; and companion imaging data supporting the patient selection criterion (target expression confirmed by diagnostic imaging, not assumed).

Teaching synthesis

The hardest but most valuable teaching moment in modern FIH oncology education is when students recognize that the same question — “what dose do we recommend for the next trial?” — has completely different answers depending on what drug class is being studied, because the dose-limiting mechanisms are physically and biologically distinct. A student who asks “was the MTD reached?” has asked a first-order question. A student who asks “what kind of dose-limiting event is most likely for this drug class, and does this trial design have the capacity to detect it?” has asked a second-order question. The second question is the one that clinical judgment requires.

The practical implication for FIH trial reading is a platform-specific checklist rather than a universal one. When reading an ADC FIH, anchor on: payload class, analytes measured, exposure-response, ILD/neuropathy monitoring protocol, schedule variants, body-size metric. When reading a T-cell engager FIH, anchor on: MABEL rationale, step-up design and simulation, CRS by dose period, tocilizumab/steroid use, infection data over time, proof-of-mechanism at the target dose. When reading a radiopharmaceutical FIH, anchor on: target confirmation by imaging, administered activity and mass dose distinction, organ dosimetry, cycle schedule rationale, and late toxicity monitoring design.

---

中文

在教授首次人體腫瘤試驗時最常見的錯誤是將它們視為通用類別。一個學了 3+3 設計、知道什麼是 DLT、能定義 BOIN 與 CRM 的學員，如果用小分子的心智模型去看 ADC FIH，仍然會犯系統性錯誤。面對 T 細胞接合器試驗時會犯不同但同樣系統性的錯誤。面對 alpha 發射體放射性藥物試驗時則幾乎會完全迷失。這三個平台不是共同主題的小變體；它們有不同的主要失敗模式、不同的最佳化目標和不同的證據架構。在共同的比較框架中一起教授它們，比分別獨立教授更有效率且臨床上更有用。

三個平台中「劑量」的含義

對傳統細胞毒性藥物或小分子標靶藥物，「劑量」意味著在某個時間點給予的毫克數，主要藥動學問題是該劑量隨時間產生什麼血漿濃度。對 ADC，「劑量」意味著每公斤毫克數，但立即需要澄清：哪個公斤（總體重、調整理想體重、調整體重），哪個分析物（完整 ADC、總抗體、游離 payload），以及哪個時間窗口對療效與毒性重要。對 T 細胞接合器，mg/kg 概念基本保留，但相關藥效學數量不是血漿藥物濃度本身——而是藥物濃度產生的 T 細胞活化程度，這取決於局部腫瘤細胞密度、效應細胞與靶細胞比和免疫微環境狀態。對放射性藥物，「劑量」分裂為給予活性（MBq）、質量劑量（µg）和器官特異吸收劑量（Gy）——三個不同的數字，不能合併為一個。

劑量升量期間的主要失敗模式

每個平台在 FIH 中的特徵性失敗模式是不同的。對 ADC，最常見的失敗模式是 payload 相關累積毒性，在單一週期 DLT 窗口中不可見：來自 MMAE 的周邊神經病變在數月後累積；來自 deruxtecan 類 ADC 的 ILD 可能在第二或第三週期後出現且嚴重。FIH 設計挑戰是在劑量升量期間捕捉這些延遲訊號，這需要每個 cohort 更長的隨訪，且必須預先指定類別特異毒性的監測方案。對 T 細胞接合器，最常見的早期失敗是前一到三劑的急性 CRS，由藥物設計的免疫突觸驅動。這裡的失敗模式是速度：如果不及時監測和治療，CRS 可以在數小時內從 1 級升級到 3 級。FIH 設計挑戰是時間性的——建立足夠慢的加速以允許免疫校準，同時不將病人置於數月的亞治療劑量暴露。對放射性藥物，失敗模式是累積器官吸收劑量在多個週期超過腎臟、骨髓或唾液腺的放射耐受性。第 1 週期的 DLT 是急性血液毒性，但整個治療過程的劑量限制事件可能是在 3-4 個週期後出現的遲發腎臟或骨髓效應。

每個平台的最佳化目標

當 Project Optimus 問「什麼是最佳劑量」時，答案對每個平台有不同的維度。對 ADC，最佳化意味著找到在多個週期維持持久療效同時產生可接受累積 payload 暴露的時程（每週、每兩週、q21d 第 1/8 天）、給藥指標（總體重 vs. 調整理想體重）和 mg/kg。最佳化同時是毒性最佳化（最小化嗜中性球低下、ILD、神經病變）和藥動學最佳化（最小化變異以使劑量可靠有效）。對 T 細胞接合器，最佳化意味著找到在沒有 3 級以上 CRS 的情況下將病人帶到目標劑量的逐步升量路徑，加上找到目標劑量本身，以在可接受的治療期感染、低免疫球蛋白血症和血球低下負擔下產生最佳反應率。最佳化是時間性和免疫學的，不是純粹藥動學的。對放射性藥物，最佳化意味著找到在所有週期合計保持腎臟、骨髓和其他危及器官劑量低於放射損傷閾值的同時，遞送最高可實現腫瘤吸收劑量的給予活性和週期時程。最佳化是劑量測定的和放射生物學的，在從根本上不同的物理尺度上運行。

教學整合

現代 FIH 腫瘤教育中最難但最有價值的教學時刻，是當學員認識到同一個問題——「我們向下一個試驗推薦什麼劑量？」——依所研究的藥物類別有完全不同的答案，因為劑量限制機轉在物理和生物學上是不同的。問「MTD 是否達到？」的學員問了一個一階問題。問「這個藥物類別最可能的劑量限制事件是什麼，這個試驗設計是否有能力偵測它？」的學員問了一個二階問題。臨床判斷需要第二個問題。

對 FIH 試驗閱讀的實際含義是平台特異清單而非通用清單。閱讀 ADC FIH 時，錨定於：payload 類別、量測的分析物、暴露—反應、ILD/神經病變監測方案、時程變體、體型指標。閱讀 T 細胞接合器 FIH 時，錨定於：MABEL 理由、逐步升量設計和模擬、按劑量期的 CRS、托珠單抗/類固醇使用、隨時間的感染資料、目標劑量的機轉證明。閱讀放射性藥物 FIH 時，錨定於：影像確認的靶點確認、給予活性和質量劑量的區別、器官劑量測定、週期時程理由和遲發毒性監測設計。

Key Concepts | 核心概念

平台	主要失敗模式	最佳化目標	MTD 未達時的關鍵問題
ADC	Payload 累積毒性（ILD、神經病變，常在 DLT 窗口後才出現）	Schedule + 給藥指標 + PK variability + 暴露—反應	類別毒性是否有足夠隨訪？暴露—反應資料是否存在？
T cell engager	急性 CRS（小時至天內出現）	Step-up 路徑 + 目標劑量 + 感染和低免疫球蛋白的長期管理	CRS 是否按劑量期報告？是否有 PK/PD 支持飽和效果？
放射性藥物	累積器官吸收劑量（腎臟、骨髓，跨多個週期）	腫瘤吸收劑量 vs. 危及器官吸收劑量 + 週期分次	是否有劑量測定資料？標準 DLT 窗口是否足以捕捉延遲毒性？

• Platform-specific failure mode recognition is the core skill for modern FIH reading
• MTD not reached has different implications across platforms — cannot use a single interpretive template
• Dosimetry (RPT), step-up design (TCE), and schedule comparison (ADC) are platform-specific dose optimization evidence requirements

Related Pages | 相關頁面

• adc-fih-clinical-pharmacology — ADC platform detail
• adc-dose-selection-regimen-optimization — ADC dose selection framework
• adc-fih-case-studies-b7h3-trop2 — Concrete ADC cases
• t-cell-engager-fih-crs-step-up-dosing — TCE CRS and step-up pharmacology
• t-cell-engager-clinical-workflow-and-cases — TCE clinical workflow
• radiopharmaceutical-therapy-fih — RPT dosimetry and dose optimization