Portacaval anastomosis has proved to be effective in avoiding active and recurrent hemorrhage from gastroesophageal varices in liver cirrhosis. However, hepatic encephalopathy is the most common and serious complication of this procedure. The aim of this study was to investigate by multivariate analysis the predictive factors of development of hepatic encephalopathy in 50 Child's A and B cirrhotic patients whose variceal bleeding was treated with emergency (n = 17) or elective (n = 33) portacaval anastomosis. The etiology of the cirrhosis was alcoholic in 74% of cases. The mean follow-up was 22.7 +/- 16.6 months (range 1-60 months). The 2-yr probability of suffering from at least one episode of hepatic encephalopathy in the overall group was 43%. The multivariate analyses (Cox's regression method) of 37 variables based upon clinical history, physical examination, and laboratory data disclosed that only five of these variables had independent predictive value: need for diuretic treatment in the days prior to surgery, absence of hepatomegaly, and serum levels of total bilirubin, gamma-globulin, and hemoglobin. According to the contribution of each one of these factors to the final model, a prognostic index was obtained which allowed the division of patients in two different groups of risk for developing hepatic encephalopathy (20% and 74%, respectively, after 2 yr of surgery; p = 0.0002). This index may help to better choose those candidates for portacaval anastomosis.