Thirty-nine patients admitted during a 16-month period for acute bleeding from varices confirmed by emergency endoscopy were randomized to receive either continuous intravenous infusions of vasopressin alone (0.66 units per min) (Group I: 19 patients) or vasopressin plus sublingual nitroglycerin (0.6 mg every 30 min for 6 hr) (Group II: 20 patients). The two groups of patients were similar in the type and severity of their cirrhosis. Bleeding was controlled initially in 47% (9/19) of the patients in Group I and 55% (11/20) of the patients in Group II after 6 hr of infusion (not statistically significant). Complete control of bleeding during 24 hr of infusion was achieved in only 4 of 19 patients in Group I (21%) but in 9 of 20 in Group II (45%). This difference is not statistically significant. The total number of patients with complications during infusions were significantly different statistically in the vasopressin and vasopressin-nitroglycerin groups, respectively (17/19 vs. 7/20, p less than 0.001). Major complications requiring immediate cessation of infusions were observed in 6 of 19 of the patients in Group I (32%) and in 2 of 20 in Group II (10%) (p less than 0.05). Mortality (58% in Group I, 55% in Group II) and transfusion requirements were similar in the two groups. This study shows that the addition of sublingual nitroglycerin to intravenous vasopressin does not alter the efficacy of vasopressin alone in controlling hemorrhage from esophageal varices, but it does significantly reduce the complications.