We evaluated the clinical usefulness of flumazenil (formerly Ro15-1788), a benzodiazepine antagonist, in combination with midazolam in upper gastrointestinal endoscopy. Thirty outpatients were randomized into two groups: those receiving flumazenil and those receiving placebo after endoscopy. For sedation, only midazolam was used. Performances pre-sedation and post-sedation (at 30 and 60 min) were analyzed using the Trieger test, Number Connection test, and Digit Symbol test. Patients receiving flumazenil were fully alert and able to ambulate 5 min after injection with this medication. Performances at 30 min in the Trieger, Number Connection, and Digit Symbol tests were significantly better in the group receiving flumazenil, p less than 0.005, p less than 0.025, and p less than 0.01, respectively. No phlebitis, nausea, vomiting, or anxiety were noted. No resedation events were documented. We conclude that flumazenil can dramatically shorten the recovery period following sedation with midazolam in upper gastrointestinal endoscopy, and its use is not associated with major side effects.