Steroids are recommended in severe alcohol-induced hepatitis, but some data suggest that artificial nutrition could also be effective. We conducted a randomized trial comparing the short-and long-term effects of total enteral nutrition or steroids in these patients. A total of 71 patients (80% cirrhotic) were randomized to receive 40 mg/d prednisolone (n ؍ 36) or enteral tube feeding (2,000 kcal/d) for 28 days (n ؍ 35), and were followed for 1 year or until death. Side effects of treatment occurred in 5 patients on steroids and 10 on enteral nutrition (not significant). Eight enterally fed patients were prematurely withdrawn from the trial. Mortality during treatment was similar in both groups (9 of 36 vs. 11 of 35, intention-to-treat) but occurred earlier with enteral feeding (median 7 vs. 23 days; P ؍ .025). Mortality during follow-up was higher with steroids (10 of 27 vs. 2 of 24 intention-to-treat; P ؍ .04). Seven steroid patients died within the first 1.5 months of follow-up. In contrast to total enteral nutrition (TEN), infections accounted for 9 of 10 follow-up deaths in the steroid group. In conclusion, enteral feeding does not seem to be worse than steroids in the short-term treatment of severe alcoholinduced hepatitis, although death occurs earlier with enteral nutrition. However, steroid therapy is associated with a higher mortality rate in the immediate weeks after treatment, mainly because of infections. A possible synergistic effect of both treatments should be investigated. (HEPA-TOLOGY 2000;32:36-42.)The outcome of alcohol-induced hepatitis (AH) varies with its severity. Whereas patients with mild cases easily recover by stopping alcohol intake, the course of patients with severe disease is much poorer despite becoming abstinent, with a short-term mortality rate higher than 40%. 1-3 Moreover, dietary intake in patients with mild AH is usually adequate, whereas severe patients remain anorectic for several weeks.A number of pharmacologic approaches to the treatment of severe AH have been attempted in the last 3 decades, most of them with disappointing results. Corticosteroids have been extensively evaluated in this setting. Two of the 3 published meta-analyses of the placebo-controlled randomized trials showed a beneficial effect of steroids, 4,5 whereas the third found no differences. 6 Nevertheless, rather than rejecting a therapeutic effect of steroids, this latter meta-analysis emphasizes