During a 7-yr period (1967-1974), 89 patients with alcoholic liver disease and at least one severe upper gastrointestinal hemorrhage thought to be from esophageal varices entered a randomized, controlled trial of medical therapy vs. end-to-side portacaval shunt. Follow-up continued to September, 1979, so that all surviving patients had at least 5 yr observation after randomization. Among 45 patients randomized to surgical therapy, 4 did not receive portacaval shunt, for various reasons. Among shunted patients there were 11 episodes of upper gastrointestinal bleeding, none fatal and none thought to be from esophageal varices. Thirty-seven percent of eligible patients have had moderate or severe hepatic encephalopathy ascribed to the shunt. Of 44 patients randomized to medical therapy, 7 eventually received portacaval shunt after multiple bleeding episodes. Since randomization there have been 190 episodes of bleeding requiring 589 transfusions and resulting in 23 deaths from bleeding or hepatic failure precipitated by bleeding. THere are 12 survivors in the surgically treated group and 8 in the group treated medically. Life-table analysis shows a small increase in survival in the surgically treated group throughout the study, which is not statistically significant. From our data, we could not identify risk factors that would improve the selection of patients for medical or surgical therapy.