A prospective randomized controlled clinical trial was performed in 43 consecutive histologically proved cirrhotic patients with endoscopically proved actively bleeding esophageal varices. Twenty-two were randomly selected to have esophageal tamponade with the Sengstaken-Blakemore tube, and 21 were selected to have endoscopic sclerosis of the esophageal wall. The two groups were similar in demographic, clinical and laboratory data. Bleeding was controlled by the Sengstaken-Blakemore tube in 16 of 22 patients (73%) and by endoscopic sclerosis in 20 of 21 (95%). Among those controlled by the Sengstaken-Blakemore tube, seven (44%) rebled and three (43%) were again controlled by the Sengstaken-Blakemore tube; in the endoscopic sclerosis group, four (20%) rebled and three (75%) were controlled. Thus, hemorrhage was definitively controlled in 52% of patients and 66% of bleeding episodes in the Sengstaken-Blakemore tube group and in 90% of patients and 92% of bleeding episodes in the endoscopic sclerosis group. The definite control of hemorrhage was significantly better in the endoscopic sclerosis group (p less than 0.01). The Sengstaken-Blakemore tube patients received no definitive therapy after bleeding had been controlled. Within 30 days, six patients (27%) in the Sengstaken-Blakemore tube group had died compared to 2 (10%) in the endoscopic sclerosis group which is statistically significant (p less than 0.01) in favor of endoscopic sclerosis. The frequency of complications was similar in the two groups. Endoscopic sclerosis patients received serial endoscopic sclerosis after bleeding had been stopped during the whole period of follow-up.(ABSTRACT TRUNCATED AT 250 WORDS)