The aim of this study was to assess, by a pragmatic controlled trial, the value of propranolol in the prevention of recurrent bleeding in patients with cirrhosis. From January 1982 to December 1984, 368 cirrhotic patients were admitted for bleeding episode due to portal hypertension. Ninety-nine patients were included in the trial 24 h after cessation of bleeding (5 days on the average after hospital admission) and were randomly assigned to 2 groups: 51 patients receiving propranolol at doses that reduced the resting heart rate by 25 p. 100 and 48 receiving neither treatment nor placebo. Ninety-one per cent presented a rupture of varices, and 9 p. 100 acute gastric erosions: 94 p. 100 had an alcoholic cirrhosis. The distribution according to Child Turcotte's classification was grade A: 45.5 p. 100, B: 45.5 p. 100, C: 100 p. 100. The two groups were not significantly different excepted for previous bleeding episodes. The cumulative percentages of recurrent bleeding were not significantly different: between the treated and untreated groups (60 p. 100 vs. 68 p. 100 at 18 months); between the compliant patients (n = 39) and untreated and not compliant patients altogether; between the treated and untreated patients of Child's grade A on the one hand, and patients grades B and C on the other hand. The curves of cumulative survival were not significantly different between treated and untreated patients (66 p. 100 vs. 78 p. 100 at 18 months).(ABSTRACT TRUNCATED AT 250 WORDS)