The purpose of the study was to determine whether corticosteroids affect short-term mortality from alcoholic hepatitis. A metanalysis was conducted using studies identified through a MEDLINE computer search from 1966 to 1989 and extensive manual searches of associated bibliographies. Eleven randomized studies that assessed mortality in hospitalized patients diagnosed with alcoholic hepatitis and treated with corticosteroids were evaluated. Overall, the protective efficacy of corticosteroids was 37% (95% confidence interval 20% to 50%). Protective efficacy was higher among those trials with higher quality scores and in trials that excluded subjects with active gastrointestinal bleeding in patients with hepatic encephalopathy, protective efficacy was 34% overall (confidence interval 15% to 48%). In subjects without hepatic encephalopathy, corticosteroids were not believed to have a protective effect; this lack of efficacy was noted across all trial subgroups. Results of the metanalysis suggest that corticosteroids reduce shortterm mortality in patients with acute alcoholic hepatitis who have hepatic encephalopathy. The protective effect is dependent on exclusion of patients with acute gastrointestinal bleeding.