The aim of this study was to identify high-risk patients for hepatocellular carcinoma (HCC). Among 151 patients with histologically proven cirrhosis hospitalized from 1987 to 1990 and prospectively followed-up until June 1994, 31 developed HCC. We assessed the predictive value of 22 variables recorded at enrollment for HCC occurrence by the log rank test and the Cox proportional hazards model. Six clinical and biological variables summarized predictive information of HCC: age > or = 50 years (P = .01), male (P = .01), large esophageal varices (EV) (P = .03), prothrombin activity < 70% (P = .04), serum alpha-fetoprotein (AFP) > or = 15 ng/L (P = .06), and anti-hepatitis C virus antibodies (P = .08). A clinicobiological predictive score identified two groups of patients at low (n = 67; 3-year cumulative incidence, 0%) and high risk for HCC (n = 84; 3-year cumulative incidence, 24%). The predictive value of this score was confirmed using an independent population of 49 patients with cirrhosis. Furthermore, liver large-cell dysplasia (LCD) had an additional predictive value in high-risk patients (P = 10(-4), which thus helped to define a subgroup at very high risk for HCC (n = 12; 3-year cumulative incidence, 72%). In Western patients with cirrhosis, a limited number of usual variables can identify a group of patients at high risk for HCC. Among these patients, liver biopsy allows for the determination a subgroup of patients at very high risk for HCC requiring intensive screening or preventive measures.