Flumazenil, a benzodiazepine antagonist, reliably reverses midazolam-induced sedation; however, its effect on respiratory depression has not been established completely. Twelve healthy volunteers received sufficient midazolam (0.13 +/- 0.01 mg.kg-1 mean +/- SE) to render them unresponsive to verbal command; they then received flumazenil 1.0 mg or placebo (flumazenil vehicle) in a randomized, double-blind fashion. Ventilatory drive was measured before and after administration of midazolam, as well as 3, 30, 60, and 120 min after administration of flumazenil or placebo. Seven to 30 days later, the study was repeated, with subjects receiving placebo or flumazenil (whichever they had not received during their first trial). Midazolam caused significant decreases in the slope of the CO2 response (-29 +/- 5%; P less than 0.005); minute ventilation (VE) at end-tidal CO2 tension (PETCO2) = 46 mmHg (-28 +/- 4%; P less than 0.001), and tidal volume at PETCO2 = 46 mmHg (-44 +/- 4%; P less than 0.005). Three minutes after intravenous administration of flumazenil 1.0 mg, VE46 and tidal volume increased to 108 +/- 6% and 105 +/- 6%, respectively, of their premidazolam values; at the same time after administration of placebo, VE46 and tidal volume remained significantly depressed (between groups, P less than 0.005 for each variable). Thirty minutes later, these variables did not differ between groups, probably because the effects of flumazenil and midazolam were diminishing.(ABSTRACT TRUNCATED AT 250 WORDS)