A randomized trial was undertaken to determine efficacy of nitroglycerin when added to a vasopressin infusion in both reducing the complication rate and giving improved control of acute variceal hemorrhage. Seventy-two bleeding episodes in 57 patients were included, with vasopressin being used on 34 occasions and vasopressin plus nitroglycerin on 38 occasions, for an infusion period of 12 hr. At the end of the 12-hr period, hemorrhage had been controlled significantly more frequently in those receiving combined therapy (26 of 38; 68%) than in those given vasopressin alone (15 of 34; 44%, p less than 0.05), although this difference was not statistically significant if those patients in whom therapy was discontinued due to drug complications were excluded from the analysis [hemorrhage controlled in the combined group (68%) and vasopressin alone (48%); chi 2 = 2.4, p greater than 0.05]. Major complications requiring cessation of therapy were significantly less common in those given nitroglycerin--one occasion compared to seven in those given vasopressin alone (p less than 0.02). Thus, the addition of nitroglycerin to a vasopressin infusion results in a lower rate of complications and is more effective in controlling variceal hemorrhage.