Abstract
BACKGROUND: With the increasing prevalence of obesity and type 2 diabetes, the number of patients with hepatitis B virus (HBV)-related hepatocellular carcinoma (HCC) coexisting with hepatic steatosis is steadily rising. However, the impact of hepatic steatosis on tumor recurrence following radical resection remains unclear. METHODS: We retrospectively analyzed a cohort of 733 HBV-infected patients diagnosed with HCC who underwent curative liver resection. Propensity score matching (PSM) was performed at a 1:2 ratio using 12 covariates to reduce selection bias and explore the association between preoperative hepatic steatosis and recurrence-free survival (RFS). Furthermore, we constructed a postoperative recurrence prediction model based on hepatic steatosis and other clinicopathological factors using 101 combinations of machine learning algorithms. The optimal model was identified through comprehensive evaluation and validation. RESULTS: After PSM, survival analysis revealed that patients without hepatic steatosis had significantly better RFS compared to those with steatosis. Multivariate Cox regression analysis confirmed that preoperative hepatic steatosis was an independent risk factor for recurrence following radical resection ( P = 0.006, HR:1.564, 95% CI:1.137-2.150). A recurrence prediction model was developed using hepatic steatosis and additional clinicopathological features through machine learning. Among the 101 models tested, the Random Survival Forest (RSF) model exhibited the best predictive performance, achieving a C-index of 0.719 in the training cohort. The model demonstrated high predictive accuracy for 1-, 2-, and 3-year recurrence, with AUC of 0.782, 0.856, and 0.898, respectively. Compared to conventional staging systems such as BCLC and CNLC, our model achieved superior performance, and decision curve analysis (DCA) demonstrated favorable clinical utility. CONCLUSION: Preoperative hepatic steatosis is an independent predictor of recurrence after radical resection in patients with HBV-related HCC. The RSF-based machine learning model incorporating hepatic steatosis and other clinicopathological factors effectively predicts postoperative recurrence risk and may facilitate personalized clinical decision-making in this patient population.