OBJECTIVE To determine whether flumazenil, a drug used to reverse benzodiazepine-induced respiratory depression and approved only for i.v. use, is effective by alternative routes. METHODS A randomized, controlled, nonblinded, crossover canine trial was performed to evaluate reversal of midazolam-induced respiratory depression by flumazenil when administered by alternative routes. Mongrel dogs were sedated with thiopental 19 mg/kg i.v., then tracheally intubated. With the dogs spontaneously breathing, tidal volume, end-tidal CO2, and O2 saturation were observed until a stable baseline was achieved. Incremental doses of midazolam were administered until respiratory depression (30% decline in tidal volume, 10% decrease in O2 saturation, and 15% increase in end-tidal CO2) occurred. Flumazenil was administered by a randomly selected route [0.2 mg followed 1 minute later by 0.3 mg i.v., sublingual (s.l.) or intramuscular (i.m.); or 1 mg followed 1 minute later by 1.5 mg per rectum (PR)]. Time to return to baseline respiratory functions was recorded ("time to reversal"). Each of 10 dogs was studied using all 4 routes of flumazenil administration with a washout period of at least 7 days. An additional dog served as a control (no flumazenil). RESULTS The control time to reversal was 1,620 seconds. The i.v. route was significantly faster (mean 120 +/- 24.5 sec) than the other 3 routes (p < 0.005). The SL route was the second fastest (mean 262 +/- 94.5 sec), the IM route was the third fastest (mean 310 +/- 133.7 sec) and the PR route was the s;owest (mean 342 +/- 84.4 sec). The SL, IM, and PR routes did not differ significantly from one another. CONCLUSIONS Flumazenil administered by all 4 routes reversed midazolam-induced respiratory depression in a dog model. For the selected dosages used, the i.v. route was significantly faster than all 3 other routes, and SL was the second fastest.