The U-Shaped Association Between Remnant Cholesterol and Postoperative Survival in Hepatocellular Carcinoma:...

The prognostic role of remnant cholesterol (RC) in hepatocellular carcinoma (HCC) remains unexplored. This study aimed to investigate the association between RC and overall survival (OS) in HCC patients after hepatectomy and to develop a robust prognostic model.

439 HCC patients who underwent curative hepatectomy were retrospectively analyzed. RC was calculated as total cholesterol minus (HDL-c + LDL-c). To specifically evaluate the potential nonlinear relationship, the association between RC and OS was assessed using restricted cubic splines (RCS) in addition to Cox regression and subgroup analyses. A machine learning approach employing nine algorithms was used to develop a prognostic model, with model interpretability achieved using SHapley Additive exPlanations (SHAP). An online predictive tool was subsequently deployed.

A significant U-shaped relationship between RC and OS (P for non-linearity = 0.013) was identified, with the lowest risk observed at approximately 1.04 mmol/L. Both too low and too high RC levels were independent predictors of worse OS. Among the machine learning models, XGBoost demonstrated superior and consistent performance for predicting 1-, 3-, and 5-year OS. SHAP analysis confirmed RC as a key predictive feature, alongside TNM stage and tumor characteristics. An interactive web-based tool was successfully implemented for clinical use.

RC demonstrates a novel U-shaped association with HCC postoperative survival in an Asian HBV-endemic cohort, underscoring its role as a significant biomarker reflecting metabolic imbalance. The developed machine learning model, which integrates RC, provides accurate, interpretable, and individualized risk assessment, offering a valuable tool for clinical prognostication and potential guidance for personalized management strategies.