Fifty episodes of bleeding from esophageal or gastric varices in 33 patients with cirrhosis were randomized to treatment with either intravenous terlipressin (2 mg initially and 1 mg every 4 hr for 24 hr together with bolus injection and continuous infusion of placebo) or with somatostatin (250 micrograms as a bolus and continuous infusion of 250 micrograms/hr somatostatin for 24 hr and placebo injections). Standard therapy with transfusions, fluid and electrolyte correction and lactulose was administered in both groups. In the terlipressin group, 22 of 25 bleeding episodes (88%) were initially stopped by the vasoactive drugs, and in the somatostatin group 19 of 25 bleeding episodes (76%) were initially stopped by the vasoactive drugs. Two of the three bleeding episodes not arrested by terlipressin and five of the six bleeding episodes not arrested by somatostatin were controlled by balloon tamponade. In one patient in each group variceal bleeding initially could not be stopped, and the patients died. The failure rate of the vasoactive treatment alone, including rebleeding episodes within the study period, was 20% in the terlipressin group and 32% in the somatostatin group. The control rate, including balloon tamponade, was 96% in both groups. The hospital mortality rate was 16% (4 of 25) in the terlipressin group and 24% (6 of 25) in the somatostatin group. Blood transfusions, use of balloon tamponade and duration of bleeding did not differ significantly.(ABSTRACT TRUNCATED AT 250 WORDS)