Fifty-five patients with alcoholic hepatitis were studied in a 28- to 32-day randomized double blind treatment trial comparing prednisolone (40 mg per day) with placebo therapy. Of 31 placebo-treated patients, 4 died during the study interval and 2 more died within 5 days of study completion. Only 1 of 24 prednisolone-treated patients died during the same interval (Fisher exact test; P = 0.10). Stepwise discriminant analysis of laboratory factors associated with death revealed independently significant associations with prolongation of prothrombin time and height of serum bilirubin at the initiation of the study. When treatment was included as a variable in this discriminant analysis, it was found that corticosteroid therapy significantly decreased mortality (P less than 0.05). The corrected wedged hepatic venous presure decreased to a similar extent in the two groups. These studies suggest that corticosteroid therapy does decrease early mortality in patients with severe alcoholic hepatitis, but has no short term effect on the development of portal hypertension.