The aim of this study was to evaluate, using a double-blind technique, the efficacy of the association of transdermal nitroglycerin to vasopressin infusion for the treatment of variceal bleeding. Sixty-nine cirrhotic patients with active variceal bleeding were randomly allocated to receive vasopressin (0.4 to 0.8 unit per min until variceal bleeding has been controlled for 12 hr) associated with nitroglycerin administered transdermically in a slow-release preparation (10 mg in 24 hr) or placebo. An initial control of variceal hemorrhage was achieved in 83% of the patients receiving vasopressin-nitroglycerin and in 74% in the vasopressin-placebo group. Owing to a lower frequency of recurrent bleeding during therapy (18 vs. 42%, p = 0.11), vasopressin-nitroglycerin achieved a definitive control of bleeding in a higher proportion of patients than vasopressin-placebo (73 vs. 54%, p = 0.13). The group treated with the drug combination showed favorable results in relation to transfusion requirements (2.9 +/- 0.4 vs. 4.2 +/- 0.5 units, p = 0.05), total dose of vasopressin required (453 +/- 47 vs. 587 +/- 50 units, p less than 0.05), need of balloon tamponade (6 vs. 15, p less than 0.05) and requirement for emergency surgery (0 vs. 4, p = 0.07). There were no significant differences in the undesirable effects associated with treatment, observed in 37 and 49% of cases, respectively. Hospital mortality was similar (33 vs. 25%). This study demonstrates that transdermal nitroglycerin improves the effectiveness of vasopressin for controlling variceal hemorrhage.