OBJECTIVES: To determine the factors predicting mortality from bleeding esophageal varices and to examine the possibility of an association between the development of adult respiratory distress syndrome (ARDS) and the use of ethanolamine oleate as an esophageal variceal sclerosant. DESIGN: Retrospective review. SETTING: ICU in a teaching hospital. PATIENTS: A total of 101 patients with endoscopically confirmed bleeding esophageal varices were admitted on 124 occasions from 1985 to 1990. Mean age was 50 +/- 13.5 (SD) yrs. There were 62 males and 39 females. Using the Child-Pugh classification, 21.8% patients were class A, 38.6% class B, and 39.6% class C. Mean ICU and hospital lengths of stay were 5.4 +/- 5.1 and 19.6 +/- 16.1 days, respectively. Mean Acute Physiology and Chronic Health Evaluation (APACHE II) score on admission was 16.5 +/- 7.6. INTERVENTIONS: Endoscopic variceal sclerotherapy was performed in 99 (79.8%) of 124 ICU admissions in the 101 patients. Esophageal balloon tamponade was performed in 64 (51.6%) and a vasopressin infusion was administered in 47 (37.9%) of the 124 ICU admissions. A variety of factors was studied to find predictors of mortality and the development of ARDS. RESULTS: Forty-eight (48.5%) of the 101 patients died during the hospital stay. Independent predictors of mortality (by stepdown logistic regression) were total volume of ethanolamine oleate injected during sclerotherapy, multiple blood transfusions, Glasgow Coma Scale score, International normalized ratio for prothrombin test, and the presence of circulatory shock on ICU admission. Age, sex, Child-Pugh score, APACHE II score, serum bilirubin, albumin, and creatinine concentrations, use of esophageal balloon tamponade or vasopressin infusion, sepsis, pneumonia, congestive cardiac failure, aspiration, and ARDS were not statistically independent predictors of outcome. There was no difference in the mortality rates for the various causes of liver disease. Pulmonary complications occurred in 44 (43.6%) patients; sepsis occurred in 31 (25%) patients. ARDS developed in 14 patients (11.3% admissions, 13.9% patients). Statistically independent predictors of ARDS were sepsis, low plasma albumin concentration, use of esophageal balloon tamponade, and more than one sclerotherapy session. The volume and type of sclerosant used were not statistically independent predictors. CONCLUSIONS: Outcome is poor for patients with bleeding esophageal varices requiring ICU admission and is related to the severity of liver failure, the degree of blood loss, and failure of therapy to stop the bleeding. The findings do not support an association between the use of the sclerosant ethanolamine and the development of ARDS.