Large esophageal varices predict hepatocellular carcinoma in newly diagnosed cirrhosis: insights from machine learning and propensity score analysis

Introduction and Objectives Hepatocellular carcinoma (HCC) is a major complication of chronic liver disease (CLD) requiring improved risk stratification. Esophageal varices (EVs), as manifestations of severe portal hypertension, may promote HCC development through vascular remodeling and hypoxia-related pathways. We aimed to evaluate whether variceal size independently predicts HCC risk and whether specific biomarker profiles can be identified. Patients and Methods In this 12-year prospective cohort study, 545 patients with newly diagnosed compensated advanced chronic liver disease (cACLD) were enrolled at a tertiary hepatology center and followed for a median of 81 months. Baseline evaluations included endoscopy, elastography, hepatic venous pressure gradient measurement, imaging, and laboratory tests. EVs were classified as absent, small, or large according to Baveno VI nomenclature; large EVs were the primary exposure. Propensity score methods and Cox models assessed their impact on HCC risk, and random survival forests evaluated predictive value. Results During follow-up, 92 patients (16.9%) developed HCC. Large EVs at baseline emerged as the strongest independent predictor of HCC onset, surpassing alpha-fetoprotein, hepatic venous pressure gradient, and liver stiffness. Hazard ratios ranged from 7.2 to 14.7 (all p < 0.001). After correction for multiple testing, HA, CA9, I-FABP, and TGF-β1 remained associated with large varices, suggesting extracellular matrix remodeling, hypoxia, intestinal barrier injury, and profibrogenic signaling. Conclusions Large esophageal varices independently predict HCC in cACLD, reflecting severe portal hypertension and a pro-tumorigenic milieu, and should inform improved surveillance and preventive strategies targeting vascular and fibrotic pathways.