Between September 1982 and April 1988, 60 cirrhotic patients with prior variceal hemorrhage were randomized to undergo the placement of an elective shunt (distal splenorenal: 26; nonselective: 4) or long-term endoscopic sclerotherapy (n = 30). Eighty-six percent of patients had alcoholic cirrhosis, and 33% were classified as Child's class C. After a mean follow-up of 87 months, 60% of patients undergoing sclerotherapy and 17% of shunt patients experienced rebleeding (p < 0.001). Shunt patients have survived longer than those who had sclerotherapy (6-year survival rates of 53% and 26%, respectively; p < 0.05). In part because of the wide geographic distribution of patients, only 4 of 13 patients in whom sclerotherapy failed (31%) could undergo salvage by shunt surgery. Although hepatic portal perfusion was better maintained after sclerotherapy, there were no major differences between the groups in terms of post-therapy hepatic or psychoneurologic function. In a predominantly alcoholic cirrhotic patient population (half non-urban), the results of elective shunt surgery were superior to those of chronic endoscopic sclerotherapy with respect to the prevention of recurrent variceal hemorrhage and survival.