Flumazenil, a specific benzodiazepine antagonist, was evaluated as adjunctive therapy in the management of benzodiazepine overdose. Thirteen emergency departments enrolled 326 patients in this double-blind, placebo-controlled trial; 162 patients were randomly allocated to receive flumazenil (maximum dose, 30 ml, providing 3 mg of flumazenil), and 164 were allocated to receive placebo (maximum dose, 30 ml). A successful response was the attainment of a score of 1 or 2 on the Clinical Global Impression Scale (CGIS), denoting a very much improved or much improved status, 10 minutes after the start of intravenous administration of the test drug. Among those patients whose drug screen revealed the presence of benzodiazepines, 75 (77%) of 97 patients given flumazenil and 13 (16%) of 83 given placebo attained such a response. The mean CGIS score at 10 minutes for benzodiazepine-positive patients treated with flumazenil was 1.95 versus 3.58 for those given placebo. As determined by the Neurobehavioral Assessment Scale, 61% of patients who initially responded became resedated; in these patients, the effect of flumazenil lasted a median of 90 minutes. At the investigator's discretion, patients who did not achieve a criterion response in the double-blind trial could receive open-label flumazenil, titrated as in the double-blind phase. Among the benzodiazepine-positive patients, 9 (53%) of 17 patients from the flumazenil group responded to the additional flumazenil, and 58 (81%) of patients previously given placebo responded. Safety was assessed in all 326 patients given the test drug. The most frequent adverse experiences after the administration of flumazenil were agitation (7%), vomiting (7%), abnormal crying (4%), and nausea (4%); these effects were observed with a lower frequency in the placebo group. Serious adverse experiences were reported in 4 patients; these included seizures and cardiac arrhythmias. Of the 3 patients with seizures, 2 had ingested large doses of cyclic antidepressants in addition to the benzodiazepine. The toxicology screen for 1 of the 2 showed 1900 ng/ml of amoxapine and 900 ng/ml of nortriptyline; the toxicology screen for the other, who also had ventricular tachycardia, showed 1928 ng/ml of loxapine and 301 ng/ml of amoxapine. The results of this study confirm published reports of the efficacy of flumazenil in reversing benzodiazepine-induced sedation in patients with benzodiazepine overdose. This was accomplished irrespective of the presence of coingested drugs. Flumazenil is not recommended for patients with serious cyclic antidepressant poisoning or those who use benzodiazepines therapeutically to control seizure disorders. When used as recommended, however, flumazenil has been shown to have an acceptable safety level.