AAI Internal Wiki

Immune-Engager FIH: Building an RP2D Evidence Chain When MTD Is Not Reached

閱讀時間約 18 分鐘

Immune-Engager FIH: Building an RP2D Evidence Chain When MTD Is Not Reached

免疫接合器 FIH：MTD 未達時建構 RP2D 證據鏈

English

The classical architecture of a phase 1 oncology dose-escalation study was built around a biological ceiling: the maximum tolerated dose (MTD). In the cytotoxic chemotherapy era, the reasoning was pragmatically sound — toxicity accumulates as dose increases, therapeutic benefit often tracks toxicity in parallel (because killing rapidly dividing cells is the mechanism), and the MTD identifies the dose just below unacceptable harm. From MTD, one step back gives you the recommended phase 2 dose. Clean, if imperfect.

Modern immune-engaging drugs break this architecture. T-cell engagers, NK cell engagers, checkpoint agonists, and costimulatory bispecifics are not primarily cytotoxic agents; their biological activity derives from activating immune cells, reshaping the tumor microenvironment, and enabling sustained immune surveillance. At the doses explored in phase 1, many of these agents never produce dose-limiting toxicities that converge on an MTD. Dose escalation reaches the maximum dose allowed by the protocol — defined by manufacturing constraints, PK projections, or practical safety limits — without finding a toxicity ceiling. At that point, the question “where is the MTD?” has no answer. The trial must instead answer a harder question: among doses that are all safe, which one is the right one to take forward?

Answering that question requires constructing what can be called an RP2D evidence chain — a multi-element biological argument that the selected dose is not merely safe, but operates at a level of immune engagement that is pharmacologically meaningful, practically achievable by monitoring, and consistent with the mechanism’s predicted activity range. Three recent FIH publications illustrate how this evidence chain is built, and what it looks like when it is incomplete.

AFM24: receptor occupancy as the biological ceiling when MTD is absent. AFM24 is a CD16A/EGFR bispecific innate cell engager designed to direct NK cells and macrophages to EGFR-expressing tumors. In a phase 1/2a FIH study (Clinical Cancer Research 2025), 35 patients received weekly fixed doses from 14 to 720 mg. Only one DLT occurred (grade 3 infusion-related reaction). PK was roughly dose-proportional. The key biological anchor for RP2D selection was CD16A receptor occupancy on NK cells: saturation was approached at 320–480 mg. Paired tumor biopsies showed dose-related activation of both innate and adaptive immune responses within the tumor microenvironment. Best objective response was stable disease in 10 of 35 patients (4 lasting 4.3–7.1 months), and 480 mg was selected as RP2D.

The pedagogical structure of this decision is important. The investigators did not select 480 mg because it was the highest dose tested or the most toxic tolerable. They selected it because it corresponded to receptor occupancy saturation — the point at which the drug had engaged its pharmacological mechanism to its maximum extent. Going to 720 mg after receptor saturation at 480 mg would produce higher drug concentrations without producing more CD16A engagement on NK cells; the dose-response relationship had plateaued at the pharmacodynamic endpoint that matters. This is the receptor occupancy ceiling substituting for the toxicity ceiling that MTD provides in classical dose-finding.

Cinrebafusp alfa (HER2/4-1BB bispecific): when RP2D requires active dose demarcation and PD biopsy evidence. Cinrebafusp alfa is designed to colocalize 4-1BB costimulation to HER2-positive tumors, avoiding the systemic immune toxicity of pan-4-1BB agonism. In the FIH study (Clinical Cancer Research 2025), doses from 0.0005 to 18 mg/kg were explored in HER2-positive advanced malignancies. The investigators made a critical conceptual move: they identified a threshold below which doses were non-active (producing no measurable PD changes) and above which doses entered the active pharmacological range. From 2.5–18 mg/kg, doses were classified as active based on dose-related changes in circulating 4-1BB levels, CD8 T-cell counts, Ki67, CD56, and granzyme B. Among 40 active-dose-evaluable patients, ORR was 12.5%, with responses concentrated at 8 and 18 mg/kg (confirmed ORR ~28.6% and 25.0% respectively). MTD was not reached.

The teaching point here is that RP2D for immune-engagers cannot be selected from within the non-active dose range. A dose that is safe but produces no biological activity is not a candidate for phase 2 — it will produce no clinical benefit. The evidence chain therefore requires explicit demarcation of which doses entered the pharmacologically active zone. PD biopsy data (showing immune cell changes in the tumor) corroborates that the drug’s mechanism is operating in the tissue that needs to respond, not just in peripheral blood. The question of whether 8 or 18 mg/kg represents the better RP2D requires additional evidence — ideally randomized dose comparison in expansion — because both produced objective responses and neither reached MTD.

W0180 anti-VISTA: the negative case teaches what evidence gaps look like. W0180 is a monoclonal antibody against VISTA (V-domain immunoglobulin suppressor of T-cell activation), a negative immune checkpoint in the tumor microenvironment. In a BMJ Oncology 2026 FIH study, 33 patients received weekly single-agent doses from 3.5 to 600 mg, with an additional arm combining W0180 with pembrolizumab. One DLT each of grade 2 cerebral infarction and grade 3 infusion-related reaction occurred. The trial was terminated early due to a sponsor business decision — MTD and the recommended expansion dose were not established. No objective responses were observed, though one patient each in the monotherapy and combination arms achieved prolonged stable disease. The biomarker analysis showed dose-dependent pharmacodynamic effects.

The critical pedagogical lesson from W0180 is the distinction between having a PD signal and having a complete RP2D evidence chain. “Dose-dependent pharmacodynamic effects” is the bottom rung of the evidence ladder — it means the drug did something measurable in the body. It does not mean the drug activated sufficient immune responses in the tumor microenvironment to produce objective responses. It does not mean the relevant dose range was fully explored. And because the trial ended early, the maximum dose that would have been feasible and safe was never established. A report that presents dose-dependent PD signals without objective responses and without MTD or expansion dose determination must clearly delineate what questions remain unanswered — not package incomplete evidence as sufficient clinical rationale.

The 2024 Regulatory Toxicology and Pharmacology paper on PK models for immune-activating FIH dose selection provides the mechanistic context for why these evidence chains are necessary from the very start of trial design. For CD3 constructs and other high-risk immune engagers, PK model assumptions have significant effects on predicted clinical exposure. The models must be pre-specified, their assumptions made transparent, and early PK data used to update model predictions and escalation decisions — rather than treating PK as a retrospective check on a dose selected by other means. This is the PK/PD-informed escalation philosophy applied to the specific case of immune-engagers, where the safety risk (CRS, ICANS, severe immune-related AE) is itself a pharmacodynamic consequence of the drug working.

For clinicians evaluating a FIH immune-engager publication, the evidence chain checklist should include five elements. First: is there a defined active dose range, distinguished from non-active doses by pre-specified PD criteria? Second: is there receptor occupancy or equivalent target engagement data showing that the proposed RP2D saturates or substantially engages the intended mechanism? Third: is there tumor microenvironment PD evidence (paired biopsies, immune phenotyping) confirming that immune activation is occurring in the tissue compartment where it needs to matter clinically? Fourth: is the toxicity profile — especially first-dose events like CRS or infusion-related reactions — characterized separately from target-dose events, so that the experience of the first patient differs from what maintenance patients will experience? Fifth: does the justification for RP2D explicitly acknowledge what remains unknown — specifically whether the selected dose is on the plateau of the dose-response curve or merely the highest dose explored without toxicity?

The last question matters most. An RP2D justified only by “we escalated to X mg and it was tolerable” is not an RP2D in the Project Optimus sense. It is a maximum tolerated dose without the dose-response evidence needed to claim it is the optimal dose. For immune-engagers where MTD is not the organizing principle, the rigor of the RP2D evidence chain is the primary measure of how seriously the trial team took dose optimization.

中文

第一期腫瘤劑量升量研究的古典架構圍繞一個生物學上限建立：最大耐受劑量（MTD）。在細胞毒性化療時代，這個推理在實用上是合理的——毒性隨劑量增加而累積，治療利益常與毒性平行（因為殺死快速分裂細胞是機轉），而 MTD 識別出剛好低於不可接受傷害的劑量。從 MTD 往後退一步，就得到推薦第二期劑量。雖不完美，但清晰。

現代免疫接合藥物打破了這個架構。T 細胞接合器、NK 細胞接合器、免疫檢查點促效劑和共刺激雙特異性藥物，並非主要的細胞毒性藥物；其生物活性來自活化免疫細胞、重塑腫瘤微環境，以及實現持續的免疫監視。在第一期探索的劑量下，許多這些藥物從未產生匯聚到 MTD 的劑量限制毒性。劑量升量達到試驗方案允許的最高劑量——由製造限制、PK 預測或實際安全限制定義——而沒有找到毒性上限。此時，「MTD 在哪裡？」這個問題沒有答案。試驗必須回答一個更難的問題：在所有都安全的劑量中，哪一個是正確的推進劑量？

回答這個問題需要構建可稱為 RP2D 證據鏈的東西——一個多元素的生物學論點，證明選定劑量不只是安全的，而且在藥理上有意義的免疫接合水平下運作、可通過監測實際實現，並與機轉的預測活性範圍一致。三個最近的 FIH 出版物說明了這個證據鏈如何構建，以及它不完整時是什麼樣子。

AFM24：MTD 缺失時受體佔有率作為生物學上限。 AFM24 是 CD16A/EGFR 雙特異性先天免疫細胞接合器，設計用於將 NK 細胞和巨噬細胞導向 EGFR 表現的腫瘤。在 Clinical Cancer Research 2025 的 phase 1/2a FIH 研究中，35 位病人接受了 14 至 720 mg 的每週固定劑量。只有一個 DLT（3 度輸注相關反應）。PK 大致與劑量成比例。RP2D 選擇的關鍵生物學錨點是 NK 細胞上的 CD16A 受體佔有率：在 320–480 mg 時接近飽和。成對的腫瘤切片顯示腫瘤微環境內先天和適應性免疫反應的劑量相關激活。最佳客觀反應是 10 位中 35 位病人的疾病穩定（4 位持續 4.3–7.1 個月），選擇 480 mg 作為 RP2D。

這個決策的教學結構很重要。研究者選擇 480 mg，不是因為它是測試的最高劑量或毒性最多可耐受的。而是因為它對應受體佔有率飽和——藥物已最大程度接合其藥理機轉的點。在 480 mg 受體飽和後繼續到 720 mg，將產生更高的藥物濃度，但不會在 NK 細胞上產生更多的 CD16A 接合；在重要的藥效學終點上，劑量—反應關係已達到平台。這是受體佔有率上限取代了古典劑量探索中 MTD 提供的毒性上限。

Cinrebafusp alfa（HER2/4-1BB 雙特異性）：RP2D 需要活性劑量劃定和 PD 切片證據時。 Cinrebafusp alfa 設計為將 4-1BB 共刺激局部化到 HER2 陽性腫瘤，避免泛 4-1BB 促效的全身免疫毒性。在 FIH 研究中，在 HER2 陽性晚期惡性腫瘤中探索了 0.0005 至 18 mg/kg 的劑量。研究者做了一個關鍵的概念性移動：他們識別了一個低於此水平劑量沒有活性（不產生可測量 PD 變化）的門檻，以及高於此水平劑量進入活性藥理範圍的門檻。從 2.5–18 mg/kg，劑量被基於循環 4-1BB 水平、CD8 T 細胞計數、Ki67、CD56 和顆粒溶素 B 的劑量相關變化分類為活性。在 40 位活性劑量可評估病人中，ORR 為 12.5%，反應集中在 8 和 18 mg/kg（確認的 ORR 分別約 28.6% 和 25.0%）。MTD 未達。

這裡的教學點是免疫接合器的 RP2D 不能從非活性劑量範圍內選擇。安全但不產生生物活性的劑量，不是第二期的候選——它不會產生臨床利益。因此，證據鏈需要明確劃定哪些劑量進入了藥理活性區。PD 切片資料（顯示腫瘤中的免疫細胞變化）佐證了藥物的機轉在需要臨床響應的組織中運作，而不只是在周邊血液中。8 或 18 mg/kg 哪個代表更好的 RP2D，需要額外證據——理想情況下是擴增中的隨機劑量比較——因為兩者都產生了客觀反應，且都未達 MTD。

W0180 抗 VISTA：陰性案例展示了證據缺口的樣子。 W0180 是針對腫瘤微環境中負性免疫調節檢查點 VISTA 的單株抗體。在 BMJ Oncology 2026 FIH 研究中，33 位病人接受了 3.5 至 600 mg 的每週單藥劑量，另有一組聯合 pembrolizumab。試驗因公司決策提前終止——MTD 和推薦擴增劑量未確立。沒有客觀反應，雖然單藥和聯合治療組各有一位病人達到延長的疾病穩定。生物標記分析顯示劑量相關的藥效學效應。

W0180 的關鍵教學教訓是 PD 訊號與完整 RP2D 證據鏈之間的區別。「劑量相關的藥效學效應」是證據梯子的最底層——它意味著藥物在體內做了可測量的事情。它不意味著藥物在腫瘤微環境中激活了足夠的免疫反應以產生客觀反應。它不意味著相關的劑量範圍被完全探索。因為試驗提前結束，可行且安全的最高劑量從未確立。呈現劑量相關 PD 訊號而沒有客觀反應、沒有 MTD 或擴增劑量確定的報告，必須清楚地描述哪些問題仍未得到回答——而不是將不完整的證據包裝成充分的臨床理由。

對評估 FIH 免疫接合器出版物的臨床醫師，證據鏈清單應包括五個要素。第一：是否有明確的活性劑量範圍，通過預先規定的 PD 標準與非活性劑量區分？第二：是否有受體佔有率或等效靶點接合資料，顯示建議的 RP2D 飽和或實質性接合了預期機轉？第三：是否有腫瘤微環境 PD 證據（成對切片、免疫表型分析），確認免疫活化正在臨床上需要響應的組織隔室中發生？第四：毒性特性——特別是像 CRS 或輸注相關反應的首劑事件——是否與靶點劑量事件分別描述？第五：RP2D 的理由是否明確承認什麼仍然未知——特別是選定劑量是否在劑量—反應曲線的平台上，還是只是探索的最高劑量而沒有毒性？

最後一個問題最重要。只以「我們升量到 X mg，可耐受」為理由的 RP2D，在 Project Optimus 的意義上不是 RP2D。它是沒有劑量—反應證據聲明最佳劑量的最大耐受劑量。對 MTD 不是組織原則的免疫接合器，RP2D 證據鏈的嚴謹性，是衡量試驗團隊認真對待劑量最佳化程度的主要標準。

Key Concepts | 核心概念

  • RP2D (Recommended Phase 2 Dose): the dose selected for further development in phase 2 trials. For immune-engagers, this must be justified by biological evidence of pharmacological engagement, not merely absence of MTD-level toxicity.
  • Evidence chain: the multi-element biological argument supporting RP2D selection — receptor occupancy, PD biopsy, active dose demarcation, PK proportionality, and early activity signal.
  • Receptor occupancy plateau: the dose at which increasing drug concentration no longer produces additional target receptor engagement. For AFM24 (CD16A), this plateau occurred at 320–480 mg and defined the upper bound of biologically meaningful escalation.
  • Active vs. non-active dose range: the pharmacological demarcation, established by pre-specified PD criteria, between doses that produce measurable immune activation and doses that do not. RP2D must come from within the active range.
  • Dose-dependent PD signal: the bottom rung of the RP2D evidence ladder — proof that the drug did something measurable. Necessary but not sufficient for RP2D selection.
  • First-dose vs. target-dose toxicity: immune-engagers often produce infusion-related reactions or CRS at first dose that are manageable and diminish with subsequent dosing. These should be reported separately from steady-state toxicity at the target dose.
  • Clinical utility gap: the distance between “biomarker shows dose-related change” and “dose selection based on biomarker improves patient outcomes.” W0180 illustrates a trial where the gap was not closed.

Related Pages | 相關頁面

關係圖譜

反向連結