We conducted a prospective, randomized single-blind trial of propranolol for the prevention of recurrent variceal bleeding. Seventy-nine patients shown to have variceal hemorrhage at endoscopy were included in the study within 72 hr following diagnosis. Fifty-seven patients had alcoholic cirrhosis, 10 cryptogenic cirrhosis, 6 posthepatitic cirrhosis, 4 biliary cirrhosis, 1 portal vein thrombosis without cirrhosis and 1 idiopathic portal hypertension. The severity of liver disease at inclusion was assessed according to the Pugh modification of the Child-Turcotte classification: 9 (11%) had Class A; 41 (52%) Class B and 29 (37%) Class C disease. Patients were randomly assigned by sealed envelope to the propranolol group (42 patients) or the placebo group (37 patients). Propranolol dosage was titrated in order to produce plasma concentrations of propranolol of 50 to 150 ng per ml. β-blockade was also confirmed by isoproterenol testing. The cumulative percentages of patients free of rebleeding 1 and 2 years after inclusion were 31 and 21% in the propranolol group, and 25 and 17% in the placebo group; both differences were not significant. Cumulative 1 and 2 years survival were also comparable: 64 and 54% in the propranolol group vs. 70 and 63% in the placebo group. There was no evidence for a therapeutic effect of propranolol after adjusting for potential confounding variables by multiple logistic regression. We conclude that propranolol is not effective for the prevention of variceal rebleeding, when administered early following the initial bleed, in cirrhotics unselected with respect to the severity of the liver disease.