In a prospective randomized trial of 76 patients at high risk with bleeding esophageal varices, transection of the esophagus with the EEA [electroencephalic audiometry] stapling apparatus was compared with injection sclerotherapy in the management of patients with Child''s class B and C liver status. Thirty-nine patients underwent transection and 37 patients, sclerotherapy with a total of 92 injection procedures (2.4/patient). The perioperative morality (< 30 days) was 28.9% overall; 33.3% for esophageal transection and 24.3% for injection sclerotherapy (.chi.2 = 0.375, P > 0.05). Gross ascites, severe encephalopathy and emergency operations were associated with a high mortality in the transection group, but other risk factors such as age and hypersplenism did not influence the outcome in either group. Only patients in Child''s class C died after transection, but patients who died in the sclerotherapy group (mainly from recurrent bleeding) included patients from both Child''s class B and C. Early recurrence of nonfatal bleeding affected one of 39 patients (2.5%) after transection but was evident in 18 of 37 patients (48.6%) after sclerotherapy (.chi.2 = 19.12, P > 0.0005) and 6 patients died. Hemorrhage did not recur after transection during a follow-up period of 2 yr, but a further 22 episodes of bleeding were recorded in 13 patients receiving sclerotherapy with 5 deaths. Postoperative complications and long term morbidity were similar in the 2 groups. Including readmissions for bleeding and repeat procedures, the mean hospital stay per patient was shorter for transection (14.5 vs. 19.1 days) and the requirements for blood were less (1.9 U/patient vs. 3.6 U/patient) than for sclerotherapy. Apparently, esophageal transection effectively protects against short term recurrence of bleeding. Preoperative control of gross ascites will further reduce the mortality and comatose patients should be excluded from operation. Sclerotherapy provides little if any protection against recurrent bleeding and its use in the management of variceal hemorrhage in patients with advanced liver disease remains questionable. It is recomended as a temporary measure in patients at high risk until such time that more effective surgical treatment can be performed.