Ninety-three patients with cirrhosis and esophageal varices participated in a controlled prospective investigation of the prophylactic portacaval shunt procedure for the prevention of bleeding varices. Surgery sharply reduced the risk of esophageal bleeding but failed to improve longevity. Shunted patients have a greater prevalence of hepatocellular failure and are significantly more likely to develop hepatic encephalopathy. Abstinence from alcohol had a beneficial effect in preventing bleeding varices and ascites in unshunted patients but did not significantly improve mortality; continued drinking among surgical patients did not alter any of these variables. Attempts to estimate survival or to delineate clinical factors at selection leading to variceal hemorrhage were unsuccessful. In the management of cirrhosis of the alcoholic in lower socioeconomic groups the prophylactic portacaval shunt should be abandoned.