Previous reports have suggested that "endogenous" benzodiazepines could contribute to neural inhibition in hepatic encephalopathy. RO 15-1788 (flumazenil), a specific antagonist of brain benzodiazepine receptors, could thus reverse the neurological symptoms of hepatic encephalopathy. To test this possibility, we conducted a double-blind, placebo-controlled crossover trial of the efficacy of flumazenil in cirrhotic patients in hepatic coma. Seventy-seven cirrhotic patients in hepatic coma were evaluated. Fifty-six were excluded from the trial because of multiorgan failure or because coma was precipitated by prior use of benzodiazepines, and 21 patients were randomly assigned to the flumazenil group (11 patients) or the placebo group (10 patients). Treatment was administered intravenously as a 20-ml solution (placebo or 2 mg flumazenil); seven patients were crossed over. Clinical status was assessed blindly by two observers, using a modified Glasgow scale, every 15 min for 6 hr. Electroencephalogram tracings obtained before and after drug administration were evaluated blindly by two independent observers. Serum concentrations of benzodiazepines before treatment were measured by means of a fluorescence polarization immunoassay. Improvement in neurological symptoms was observed in six patients treated with flumazenil, whereas none in the placebo group showed improvement (p < 0.05; Fisher's exact test). Improvements in electroencephalogram tracings were demonstrated in four patients treated with flumazenil, compared with two patients in the placebo group (p = NS). Benzodiazepines were found in the serum of four patients treated with flumazenil (two responders and two nonresponders); all of these patients had received pharmaceutical benzodiazepines 4 to 6 days before the trial.(ABSTRACT TRUNCATED AT 250 WORDS)