Severe alcoholic hepatitis is still a therapeutic challenge. It has been recently advocated that a 3-wk infusion with insulin and glucagon reduces its short-term mortality rate. A multicenter, randomized, single-blind, sequential trial was designed to compare this treatment with placebo. The triangular boundary was defined with alpha = 0.05, beta = 0.10 and estimated survival at 4 wk of 50% with placebo, 75% with treatment. Patients with biopsy-proven severe alcoholic hepatitis (presence of one or more of three criteria: encephalopathy, prothrombin activity less than or equal to 50%, bilirubinemia greater than or equal to 100 mumol/L) were randomized into two groups; one treatment group received an infusion (12 hr/day) of an association of insulin (30 IU) and glucagon (3 mg), and a control group received an infusion of glucose. Treatments were administered during a 3-wk period, and the mortality rate was noted at 4 wk. The decision to discontinue the trial was reached on the basis of results from the first 44 patients. Overall results were assessed in the 72 patients included at the time of this decision (treatment group: n = 37; control group: n = 35). Fifty-three patients had cirrhosis. No significant differences were noted between the two groups at inclusion on the basis of clinical, laboratory and histological criteria. The mortality rate was not significantly different in the two groups; 10 patients (27%) in the treatment group and 5 patients (14%) in the control group died. Causes of death were similar in the two groups and consisted primarily of gastrointestinal hemorrhage, hepatic failure and infectious events.(ABSTRACT TRUNCATED AT 250 WORDS)