Using a randomized, double-blind study design, the intra- and postoperative sedative effects of three intravenous central nervous system depressants were compared in 64 patients during regional anesthesia. After establishing regional anesthesia, methohexital 59 +/- 29 mg IV, etomidate 12.6 +/- 8 mg IV, or midazolam 3.7 +/- 1.5 mg IV (mean dose +/- SD) were infused in a titrated fashion over 5-15 minutes to achieve similar end points of sedation. This level of sedation was then maintained with a variable-rate infusion of either methohexital 180 +/- 65 mg/hr, etomidate 32 +/- 12 mg/hr, or midazolam 7.5 +/- 4 mg/hr (mean dose +/- SD) To maintain a stable level of sedation, changes in the maintenance infusion rate were required more frequently with methohexital (4.6 +/- 3 times) than with etomidate (2.6 +/- 2 times) or midazolam (2.7 +/- 3 times). Decreases in oxygen saturation below 95% were more frequent with midazolam than with methohexital and etomidate, while recall of intraoperative events was less frequent after midazolam. Observer evaluations found less sedation in the methohexital and etomidate groups than in the midazolam group during the early postoperative period. However, the patients' evaluation of residual sedative effects (e.g., postoperative sedation visual analogue scores) and discharge times from the recovery room were similar in all three groups. Nevertheless, patients given midazolam had significantly greater impairment of performance on the digit-symbol substitution test in the early postoperative period than did patients given either methohexital or etomidate.