We report here the final analysis of a multicentre randomized, single-blind trial of propranolol for prophylaxis of first bleeding in cirrhosis. One hundred and seventy-four consecutive patients with large esophageal varices were randomly assigned to either propranolol in doses reducing the resting heart rate by 25% (85 patients) or to a placebo (vitamin K: 89 patients). Three were lost to follow-up and 26 had to be withdrawn from propranolol because of side effects (n = 23) or low compliance (n = 3). The cumulative value over 42 months of patients free of bleeding was 74% (95% confidence interval = 85%-63%) in the propranolol and 59% (95% CI = 79%-43%) in the control group and the corresponding survival figures were 51% (95% CI = 63%-39%) and 59% (95% CI = 75%-43%): neither of the differences was significant. A retrospective analysis according to the presence of ascites at randomization showed that in the subset without ascites the proportion of patients free of bleeding was significantly higher in the propranolol group than in the control group (83% vs. 61%; 95% CI = 97%-69% and 78%-44%, respectively; P = 0.028); this difference was even more evident in the ascites-free period (94% vs. 58%; 95% CI = 100%-86% and 76%-40%, respectively; P = 0.002). No differences were found in patients with ascites at randomization. Survival was not significantly affected by treatment in any subgroup, although it was shorter in the ascitic patients given propranolol than in controls (33% vs. 49%; 95% CI = 51%-15% and 71%-27%, respectively; P = 0.07).(ABSTRACT TRUNCATED AT 250 WORDS)