AAI Internal Wiki

ctDNA for Participant Selection and Prognosis in FIH Trials

閱讀時間約 14 分鐘

ctDNA for Participant Selection and Prognosis in FIH Trials

ctDNA 在 FIH 試驗的病人選擇與預後評估

English

When a patient with advanced solid tumor disease is referred to a phase 1 oncology trial, the standard eligibility checklist asks about ECOG performance status, organ function, prior lines of therapy, and the presence of a targetable molecular alteration. What it rarely asks is: at the pace this disease is currently moving, does this patient have enough time to complete the trial’s learning curve and potentially benefit? That question — the intersection of tumor kinetics, trial design burden, and patient goals — is where baseline ctDNA burden is beginning to contribute in ways that no ECOG score or biopsy can capture alone.

The most compelling data on ctDNA as a prognostic tool in the FIH trial referral context comes from the TARGET and TARGET National prospective cohort study (Journal of Immunotherapy and Precision Oncology 2025). Among 587 patients with advanced solid tumors who were referred for early phase clinical trial evaluation, investigators measured maximum variant allele frequency (maxVAF) — the highest allele frequency among all ctDNA-detected variants in a given sample — as a quantitative index of circulating tumor burden. A maxVAF threshold of 4% emerged as a prognostically meaningful cutoff: patients above this threshold had a median overall survival of approximately 5.9 months, compared with 12.1 months for those at or below 4%. Crucially, this relationship held in multivariate analysis adjusting for other baseline factors, and high maxVAF remained independently associated with inferior three-month landmark survival.

The clinical implication is not to use 4% maxVAF as an automatic exclusion criterion. That would be ethically premature given the cohort-specific nature of this finding and the absence of external validation across tumor types. The implication is more nuanced: a clinician who sees a high ctDNA burden in a referred patient should use that information to recalibrate the informed consent conversation. Phase 1 trials have real burdens — washout periods, mandatory on-treatment biopsies, frequent monitoring visits, potentially protracted dose escalation before the patient reaches a biologically relevant dose, and often limited single-agent activity in the escalation cohort itself. A patient whose tumor is rapidly evolving may not have the biological runway to complete those demands and reach a meaningful decision point for their own care. The ctDNA measurement, in this frame, is not a gatekeeper but a prognostic context-setter for shared decision-making.

This reframes ctDNA’s role in FIH trials into three functionally distinct questions that must not be collapsed together. The first is a selection question: does the patient carry the molecular target or alteration that defines the trial’s biological hypothesis? This is the companion diagnostic use — the test that says “this patient’s tumor has the right key for this drug’s lock.” The second is a pharmacodynamic / early response question: once on treatment, are ctDNA kinetics moving in the direction that suggests biological activity is occurring? This requires on-treatment serial measurements and is the question discussed in ctdna-informed-dose-finding. The third — the focus of this page — is a prognostic question: before the patient even starts, does the ctDNA burden suggest a trajectory of disease that may prevent meaningful participation and benefit?

Each of these three questions demands a different analytical approach, a different sampling time point, and a different interpretation framework. Conflating them produces dangerous errors. A high baseline ctDNA burden (prognostic signal) does not mean the drug won’t work. An absence of ctDNA at baseline (a selection challenge) does not mean the tumor lacks the target — it may mean the tumor is a low shedder. A ctDNA decline on treatment (early PD signal) does not automatically translate into improved overall survival. These distinctions must be taught explicitly, because the shorthand “liquid biopsy” conceals the heterogeneity of what ctDNA is actually being asked to do at each time point.

The Cancer Cell 2026 commentary by Tan and colleagues articulates the multi-utility architecture that this three-question framework implies. They propose teaching ctDNA’s roles across a 2-by-2 grid: detect versus monitor on one axis, tumor versus patient on the other. Detection of targetable alterations and prediction of therapeutic response sit in the eligibility and early response cells. Monitoring of tumor evolution and pharmacodynamic characterization sit in the on-treatment and resistance cells. Prognostic assessment of patient fitness for participation is an additional axis orthogonal to the drug’s pharmacology — it is about the patient’s trajectory, not the drug’s mechanism.

A critically important teaching point that emerges from both the TARGET cohort data and the broader ctDNA implementation literature is the false-negative problem. Patients with brain-predominant metastases, peritoneal carcinomatosis, or tumors that intrinsically shed little DNA into circulation may have very low or undetectable ctDNA at baseline despite a high disease burden. For these patients, a “negative” ctDNA result does not mean the tumor lacks the target or that the disease is minimal — it means the assay cannot see it. Any clinical protocol that uses ctDNA for participant selection or prognostic framing must define how it will handle low-shedding tumors: will these patients be excluded from the ctDNA analysis, treated as separate strata, or have ctDNA-informed decisions suspended in their case?

For the practicing oncologist evaluating a patient’s candidacy for a FIH trial, the maxVAF data supports a practical two-stage approach. First, assess conventional eligibility (performance status, organ function, molecular target, washout clearance). Then, if ctDNA is available, use it as a second layer of risk characterization: a patient with a maxVAF significantly above 4% in the context of a trial that will require 8–12 weeks before a meaningful efficacy signal is possible deserves an explicit conversation about disease velocity, the probability of reaching a decision point, and the alignment of the trial’s timeline with the patient’s goals. That conversation is not a denial of access — it is a higher-quality informed consent process.

中文

當晚期實體腫瘤病人被轉介參加第一期腫瘤臨床試驗時，標準收案資格清單詢問的是 ECOG 體能狀態、器官功能、先前治療線數，以及是否存在可治療的分子變異。但它很少問：以這個疾病目前的進展速度，這位病人是否有足夠的時間完成試驗的學習曲線並可能從中獲益？這個問題——腫瘤動態、試驗設計負擔與病人目標的交集——正是基線 ctDNA 負荷開始以 ECOG 評分或切片無法單獨捕捉的方式發揮作用的地方。

關於 ctDNA 作為 FIH 試驗轉介情境中預後工具最具說服力的資料，來自 TARGET 與 TARGET National 前瞻性世代研究（Journal of Immunotherapy and Precision Oncology 2025）。在 587 位被轉介接受早期試驗評估的晚期實體腫瘤病人中，研究者測量了最大變異等位基因頻率（maxVAF）——給定樣本中所有 ctDNA 偵測到的變異中頻率最高者——作為循環腫瘤負荷的量化指標。4% 的 maxVAF 門檻具有預後意義：高於此門檻的病人中位整體存活期約 5.9 個月，等於或低於 4% 者約 12.1 個月。在多變項分析中調整其他基線因素後，高 maxVAF 仍然與較差的三個月里程碑存活獨立相關。

這個發現的臨床意義，不是把 4% maxVAF 當作自動排除標準。考慮到此發現的世代特異性以及在不同癌種缺乏外部驗證，這樣做在倫理上過於草率。其意義更為細膩：當臨床醫師看到轉介病人有高 ctDNA 負荷時，應用這個資訊重新校準知情同意的對話。第一期試驗有真實的負擔——洗脫期、強制性治療中切片、頻繁監測就診，在升量世代中可能很長時間才到達生物學相關劑量，以及升量世代本身通常有限的單藥活性。腫瘤快速演化的病人，可能沒有足夠的生物學時間窗完成這些要求並達到有意義的決策點。ctDNA 測量在這個框架中，不是守門人，而是共同決策的預後情境設定者。

這把 ctDNA 在 FIH 試驗中的角色重新定義為三個功能上截然不同的問題，不能混為一談。第一個是選擇問題：病人是否攜帶定義試驗生物學假說的分子靶點或改變？這是伴隨診斷的用途——告訴我們這個病人的腫瘤是否有這個藥物的「對應鑰匙」。第二個是藥效學/早期反應問題：開始治療後，ctDNA 動態是否向提示生物活性正在發生的方向移動？這在 ctdna-informed-dose-finding 中有詳細討論。第三個——本頁的重點——是預後問題：在病人開始治療之前，ctDNA 負荷是否暗示疾病的進展軌跡可能妨礙有意義的參與和獲益？

這三個問題各自需要不同的分析方法、不同的採血時間點，以及不同的解讀框架。混淆它們會產生危險的錯誤。高基線 ctDNA 負荷（預後訊號）不代表藥物無效。基線無 ctDNA（選擇挑戰）不代表腫瘤缺乏靶點——可能意味腫瘤是低釋放者。治療後 ctDNA 下降（早期藥效學訊號）不會自動轉化為改善的整體存活。這些區別必須明確教授，因為「液態切片」這個簡稱掩蓋了 ctDNA 在每個時間點實際上被要求做的事情的異質性。

Tan 等人 2026 年 Cancer Cell 評論表達了這個三問題框架所蘊含的多功能架構。他們提出用二乘二矩陣教授 ctDNA 的角色：一個軸是偵測對監測，另一個軸是腫瘤對病人。可治療基因改變的偵測和治療反應的預測，位於收案和早期反應格。腫瘤演化的監測和藥效學特性描述，位於治療中和抗藥性格。病人參與適合度的預後評估，則是與藥物藥理學正交的另一個軸向——它關乎病人的疾病軌跡，而非藥物的作用機轉。

從 TARGET 世代資料和更廣泛的 ctDNA 應用文獻中浮現的一個關鍵教學點，是偽陰性問題。以腦轉移為主、腹膜轉移，或天生不向血液循環釋放大量 DNA 的腫瘤病人，儘管疾病負荷很高，基線 ctDNA 可能很低或無法偵測。對這些病人而言，「陰性」ctDNA 結果不代表腫瘤缺乏靶點或疾病程度很輕——它代表分析方法看不到。任何在病人選擇或預後框架中使用 ctDNA 的臨床試驗方案，都必須定義如何處理低釋放腫瘤：這些病人是否從 ctDNA 分析中排除、作為獨立分層，還是在他們的情況下暫停 ctDNA 導向決策？

對評估病人是否適合參與 FIH 試驗的執業腫瘤科醫師而言，maxVAF 資料支持實用的兩階段方法。首先，評估常規收案資格（體能狀態、器官功能、分子靶點、洗脫清除）。然後，若 ctDNA 可取得，將其作為風險特性描述的第二層：在需要 8 至 12 週才可能有有意義療效訊號的試驗背景下，maxVAF 明顯超過 4% 的病人值得明確討論疾病速度、達到決策點的可能性，以及試驗時程與病人目標的一致性。這個對話不是拒絕進入，而是更高品質的知情同意過程。

Key Concepts | 核心概念

  • maxVAF (maximum variant allele frequency): the highest allele frequency among all ctDNA-detected variants in a plasma sample; a proxy for circulating tumor burden. A 4% threshold was associated with distinct survival outcomes in the TARGET cohort.
  • Prognostic vs. predictive biomarker: prognostic = associated with outcome regardless of treatment; predictive = associated with differential response to a specific treatment. ctDNA burden is primarily prognostic; ctDNA kinetics under treatment may be predictive.
  • Low-shedding tumor: tumors that release little DNA into circulation, causing false-negative ctDNA results despite active disease. Common in brain metastases, mucinous tumors, and some low-grade malignancies.
  • Shared decision-making: a clinical communication process in which the clinician and patient jointly weigh treatment options, benefits, burdens, and goals. Baseline ctDNA burden can be one input into this conversation for FIH trial candidacy.
  • Three ctDNA questions in FIH: (1) selection — does the patient have the target? (2) pharmacodynamic — is the drug engaging the biology? (3) prognostic — does disease velocity allow meaningful participation?

Related Pages | 相關頁面

關係圖譜

反向連結