A prospective randomized trial was conducted in unselected, consecutive patients with bleeding esophageal varices resulting from cirrhosis comparing (1) emergency portacaval shunt performed within 8 hr of initial contact (21 patients) with (2) emergency medical therapy (intravenous vasopressin and esophageal balloon tamponade) followed in 9 to 30 days by elective portacaval shunt in survivors (22 patients). All patients underwent the same diagnostic workup within 3 to 6 hr of initial contact, and received identical supportive therapy initially. All patients were followed up for at least 10 yr. The protocol contained no escape or cross-over provisions. There were no statistically significant differences between the two treatment groups in the incidence of any of the clinical variables, results of laboratory tests or degree of portal hypertension. Child's risk classes in the shunt group were A-2 patients, B-8 patients and C-11 patients, whereas in the medical group they were A-10 patients, B-5 patients, and C-7 patients, a significant difference (p < 0.01) that might have favored emergency medical treatment. Bleeding was controlled initially and permanently by emergency shunt in every patient, but by medical therapy in only 45% (p < 0.001). Mean requirement for blood transfusion was 7.1 +/- 2.6 units in the shunt group and 21.4 +/- 2.6 units in the medical group (p < 0.001). Eighty-one percent of the patients in the shunt group were discharged alive compared with 45% in the medical group (p = 0.027). Five- and 10-yr observed survival rates were 67% and 57%, respectively, after emergency shunt compared with 18% and 18%, respectively, after the combination of emergency medical therapy and elective shunt (p < 0.01). These survival rates produced by emergency shunt performed within 8 hr of initial contact confirm the effectiveness of this procedure observed in our previous unrandomized studies.