A prognostic study based on 127 untreated patients with hepatocellular carcinoma was undertaken to evaluate their survival time and to find clinical and biologic criteria which allow the selection of patients with a survival time longer than 60 days who could enter a therapeutic trial. Twenty-eight clinical and biologic variables were assessed using univariate and multivariate semiparametric regression (Cox's) models. Ten variables were isolated by univariate analysis. Multivariate analysis found a negative relationship between a survival time longer than 60 days and five of these variables; these variables were in decreasing order: encephalopathy, alcohol consumption, aspartate amino transferase (AST), blood urea nitrogen, and total bilirubin. Prevalence, positive, and negative predictive values of encephalopathy were 20%, 27.5%, and 97% respectively. When three other criteria: ASAT greater than four times the upper limit of the normal (N), blood urea nitrogen greater than N, and total bilirubin greater than 2N were added, their prevalence, positive, and negative predictive values were 72%, 89.7%, and 57.1% respectively. These results suggest that in countries where incidence of hepatocellular carcinoma is low and recruitment of patients difficult, absence of encephalopathy must be the only criterion for selection of patients with hepatocellular carcinoma in therapeutic trials; whereas, in countries with a high incidence of hepatocellular carcinoma the other criteria may be added.