In a prospective, randomized clinical trial, immediate injection sclerotherapy was compared with treatment by a combined infusion of vasopressin (0.4 unit per min) and nitroglycerin (40 to 400 micrograms per min) in 50 consecutive patients with 64 episodes of endoscopy-proven variceal hemorrhage. Control of bleeding was assessed over a 12-hr period following entry into the trial. Patients in the vasopressin + nitroglycerin group were then treated by sclerotherapy, as were those in the sclerotherapy group who continued to bleed. At 12 hr, bleeding was controlled in 29 (88%) of the 33 episodes treated by sclerotherapy compared with 20 (65%) of 31 episodes treated by vasopressin + nitroglycerin (p less than 0.05). Recurrence of variceal bleeding occurred at the same frequency (31%). Although admission mortality was less in those initially treated by sclerotherapy compared to those managed by vasopressin + nitroglycerin, this did not reach statistical significance (27 and 39%, respectively, p greater than 0.20). Sclerotherapy carried out as the first treatment of the active variceal hemorrhage proved both safe and effective, even in the presence of major hemorrhage, and as compared to combined vasopressin and nitroglycerin it proved superior.