Individualized Surgical Guidance in HCC-BDTT

This dual-center study recruited 151 HCC-BDTT patients who underwent radical hepatectomy (training cohort: 92; validation cohort: 59). An XGBoost-based R-learner (Rboost) algorithm was developed to estimate the individualized treatment effect (ITE) on 5-year overall survival (OS) between BDR and NBDR. Patients were stratified into ITE tertiles: BDR-recommended, ambiguous benefit, and NBDR-recommended. Model performance assessed with heterogeneity metrics (C-for-benefit, adjusted Qini index). Survival outcomes were compared within ITE strata between surgeries concordant and discordant with model recommendations (Kaplan-Meier/log-rank). Two-year disease-free survival (DFS) served as a treatment-proximal validity check.

Inclusion criteria

• Pathologically confirmed HCC with BDTT;
• Absence of extrahepatic malignancies at the time of HCC diagnosis;
• Undergone R0 curative hepatectomy;
• No prior antitumor therapies, such as chemotherapy or immunotherapy, before surgery;
• Availability of preoperative imaging (CT or MRI) with standardized tumor evaluation.

Exclusion criteria

• Pathological findings indicating concurrent cholangiocarcinoma and HCC;
• Loss to follow-up, preventing outcome evaluation, or insufficient clinical data for analysis.
• Significant comorbidities (e.g., severe cardiac or renal failure) that could affect surgical outcomes.