Between 1988 and 1990 an unblinded, randomized trial of terlipressin or vasopressin plus transdermal nitroglycerin, as part of a treatment strategy including emergency sclerotherapy for actively bleeding varices, was conducted during 165 admissions in 137 patients with cirrhosis and upper digestive bleeding. Eighty-four patient admissions were assigned to terlipressin (2 mg every 6 h) and 81 to vasopressin (0.4 to 0.8 unit per min) plus transdermal nitroglycerin (20 to 80 mg). The two groups were comparable for relevant clinical data, but there were slightly more patients with hepatocellular carcinoma or terminal conditions in the terlipressin group. After the 24-h study period, failure to control bleeding was 20/84 (25%) in the vasopressin and 14/81 (17%) in the terlipressin group (p = 0.19). Corresponding figures for patients bleeding from varices (emergency sclerotherapy in 43 and 45, respectively) were 13/55 (24%) and 5/56 (9%; p = 0.035), from other sources 5/16 (31%) and 2/15 (13%; p = 0.23), from undefined sources 2/10 (20%) and 7/13 (54%; p = 0.1). In a logistic multivariate regression model the odds ratio for terlipressin adjusted for prognostic factors was 0.45 (p = 0.07). There were seven major side effects requiring treatment discontinuation in the vasopressin and one in the terlipressin group. These results suggest that terlipressin alone is as effective as vasopressin plus transdermal nitroglycerin, with less severe side effects, in 24-h control of upper gastrointestinal bleeding in patients with cirrhosis.