To assess the effectiveness of propranolol in the prevention of initial variceal hemorrhage, a double-blind, randomized trial was carried out in three centers. Patients with cirrhosis (78% alcoholic), hepatic venous pressure gradients greater than 12 mm Hg and endoscopically proven esophageal varices were randomly assigned to propranolol (51 patients) or placebo (51 patients). Of the 102 patients, 58% were Child's class A, 34% were Child's class B and 8% were Child's class C. Daily dosage was determined by the administration of progressively increasing doses of propranolol with the hepatic vein catheter in place to achieve a 25% decrease in hepatic venous pressure gradient, a decrease in hepatic venous pressure gradient to less than 12 mm Hg or a decrease in resting heart rate to less than 55 beats/min. During a mean follow-up period of 16.3 mo, 11 patients in the placebo group (22%) bled from esophageal varices compared with 2 in the propranolol group (4%) during a mean period of 17.1 mo (p less than 0.01). Three additional patients (6%) in the placebo group bled from portal hypertensive gastropathy compared with none in the propranolol group. Propranolol appeared effective in preventing bleeding from large varices. Eleven deaths (22%) occurred in the placebo group compared with eight deaths (16%) in the propranolol group (NS). The mean dose of propranolol was 132 mg/day, and the median dose was 80 mg/day. Using a compliance index (pill count, clinic attendance, alcohol and propranolol levels and alcohol history), 81% of the propranolol patients and 77% of the placebo patients were considered compliant. Complications severe enough to require cessation of therapy occurred in eight patients (16%) in the propranolol group and four in the placebo group (8%) (NS). We conclude that propranolol effectively prevents the first variceal hemorrhage in patients with alcoholic cirrhosis and large esophageal varices but does not improve survival.