基于RCB与ctDNA-MRD的早期三阴性 乳腺癌围手术期风险适配治疗： 从免疫骨架到ADC强化

Risk-Adapted Perioperative Management of Early Triple-Negative Breast Cancer Using Residual Cancer Burden and ctDNA-MRD: From Immunotherapy Backbone to ADC Escalation

摘要: 早期三阴性乳腺癌进入围手术期免疫治疗时代，但残余病灶仍主导复发风险，pCR二分法难以支撑术后治疗的精确方案。本文以RCB量化残余肿瘤负荷，概述ctDNA-MRD的平台、采血窗口与假阴性或假阳性限定，强调ctDNA清除较单点阳性更能反映反应深度。综合证据提出RCB × ctDNA的联合分层，用于界定分子极高危并指导新辅助治疗后阶段强化，并在免疫骨架下总结卡培他滨与奥拉帕利的使用证据，评述TROP2-ADC及ADC × IO前移的依据与挑战。最后概括关键试验终点与真实世界实施要点。

Abstract: Early-stage triple-negative breast cancer (TNBC) has entered the era of perioperative immunotherapy; however, residual disease after neoadjuvant therapy remains the dominant source of relapse, and a binary pathological complete response (pCR) endpoint is insufficient to support risk-adapted escalation or de-escalation. This review positions residual cancer burden (RCB) as a quantitative pathological anchor and critically summarizes circulating tumor DNA-based minimal residual disease (ctDNA-MRD), focusing on assay strategies (tumor-informed vs tumor-naïve), sampling windows across the perioperative timeline, and interpretability pitfalls including false negatives and false positives. Evidence from serial ctDNA studies suggests that ctDNA clearance better captures depth of response than a single time-point result and can further stratify risk within non-pCR populations. Integrating these dimensions, we propose a pragmatic framework (RCB low/high × ctDNA negative/positive) to define a “molecular ultra-high-risk” subgroup for post-neoadjuvant intensification. Within the current immunotherapy backbone, we outline the evidence boundaries for capecitabine and olaparib and discuss the rationale and emerging data supporting the forward shift of antibody–drug conjugates (ADCs), including TROP2-directed agents and ADC-immunotherapy combinations. Finally, we highlight trial-design considerations that keep event-free and overall survival as primary endpoints while using ctDNA clearance as a mechanistic/early signal endpoint, and we summarize workflow requirements for real-world implementation.