"人工智能分型指导CDK4/6抑制剂耐药HR+/HER2-晚期乳腺癌精准治疗-LINUX研究(多中心、开放标签、随机对照Ⅱ期平台试验)"方案解读OA

Background and purpose:The LINUX trial is a multicenter,open-label,randomized,controlled phase Ⅱ platform trial designed to continuously enroll patients with hormone receptor(HR)-positive/human epidermal growth factor receptor 2(HER2)-negative advanced breast cancer who have developed resistance to CDK4/6 inhibitors(CDK4/6i).This article analyzed its clinical trial protocol to provide a reference for clinicians conducting precision therapy platform trials.Methods:After enrollment,patients are subtyped using a previously constructed artificial intelligence model applied to HE-stained pathological slide images of metastatic lesion specimens(if metastatic biopsy specimens are unavailable,digital pathological subtyping is performed on primary tumor samples).Patients are classified into four subtypes(SNF1,SNF2,SNF3,and SNF4)and then randomized 2 to 1 to a precision therapy arm tailored to the molecular characteristics of each subtype or a control arm(investigator's choice of chemotherapy).The primary endpoint is objective response rate(ORR).Secondary endpoints include clinical benefit rate(CBR),investigator-assessed progression-free survival(PFS),and overall survival(OS).Safety assessment includes evaluation of the incidence and severity of treatment-related adverse events(TRAEs).Throughout the exploratory process,a Bayesian adaptive design is used to monitor the efficacy of different precision therapy arms within each SNF subtype to predict the likelihood of success in larger-scale phase Ⅲ clinical trials.Results:There are three core design features of the LINUX protocol.① A platform architecture supporting parallel evaluation of multiple subtypes and dynamic incorporation of new regimens.② Artificial intelligence(AI)-based digital pathology replacing multi-omics sequencing to reduce costs and improve accessibility.③ A Bayesian design enabling"stop for futility,advance for efficacy."Results from the first cohort of 105 patients showed that precision therapy in SNF2 and SNF4 met pre-specified success criteria,with ORRs of 65%vs 30%and 70%vs 20%,respectively,with significantly prolonged PFS and manageable safety;these arms have advanced to phase Ⅲ confirmatory studies.The protocol also specifies dose modification rules,concomitant therapy management,and stratified follow-up procedures.The LINUX protocol establishes a complete framework of"AI subtyping-precision therapy-Bayesian monitoring"with core innovations in achieving clinical translation of molecular subtyping through low-cost digital pathology,accommodating multi-regimen evaluation via platform design,and optimizing resource allocation through Bayesian statistics.Conclusion:This protocol suggests that when designing precision oncology trials,clinicians should emphasize the accessibility of subtyping tools,statistical flexibility,and comprehensiveness of the endpoint system,thereby providing methodological support for the translation from"omics discovery"to"clinical decision-making."