A cooperative study was conducted to determine the efficacy of 30 days of treatment with either a glucocorticosteroid (prednisolone) or an anabolic steroid (oxandrolone) in moderate or severe alcoholic hepatitis. One hundred thirty-two patients with moderate disease and 131 with severe disease were randomly assigned to one of three treatments: prednisolone, oxandrolone, or placebo. During the 30 days, mortality in the groups receiving steroid therapy was not significantly different from mortality in the placebo group. Thirteen per cent of the moderately ill patients and 29 per cent of the severely ill patients died. Although neither steroid improved short-term survival, oxandrolone therapy was associated with a beneficial effect on long-term survival. This was especially true in patients with moderate disease: among those who survived for one or two months after the start of treatment the conditional six-month death rate was 3.5 per cent after oxandrolone and 19 to 20 per cent after placebo (P = 0.02). No consistent long-term effect was associated with prednisolone therapy.