We compared two procedures for the emergency treatment of bleeding esophageal varices in patients who did not respond to blood transfusion and vasoactive drugs. We randomly assigned 101 patients with cirrhosis of the liver and bleeding esophageal varices to undergo either emergency sclerotherapy (n = 50) or staple transection of the esophagus (n = 51). Four patients assigned to sclerotherapy and 12 assigned to staple transection did not actually undergo those procedures, but all analyses were made on an intention-to-treat basis. Total mortality did not differ significantly between the two groups; the relative risk of death for staple transection as compared with sclerotherapy was 0.88 (95 percent confidence interval, 0.51 to 1.54). Mortality at six weeks was 44 percent among those assigned to sclerotherapy and 35 percent among those assigned to staple transection. Complication rates were similar for the two groups. An interval of five days without bleeding was achieved in 88 percent of those assigned to staple transection and in 62 percent of those assigned to sclerotherapy after a single injection (P less than 0.01) and 82 percent after three injections. In only 2 of the 11 patients who received a third sclerotherapy injection was bleeding controlled for more than five days, and 9 died. We conclude that staple transection of the esophagus is as safe as sclerotherapy for the emergency treatment of bleeding esophageal varices and that it is more effective than a single sclerotherapy procedure. We currently recommend surgery after two injection treatments have failed.