This study was designed to assess postoperatively the time course of respiratory depression due to fentanyl (F) or sufentanil (S), as well as the plasma concentrations. Seventy patients scheduled for orthopaedic surgery lasting more than 3 hours were randomly assigned to two groups, F (n = 8) or S (n = 9). Anaesthesia was induced with etomidate (0.3 mg.kg-1), droperidol (0.15 mg.kg-1), vercuronium (0.1 mg.kg-1), a loading dose of either F (10 micrograms.kg-1) or S (1 microgram.kg-1), and maintained with 60% nitrous oxide in oxygen, and an infusion of F (6 micrograms.kg-1.h-1) or S (0.6 microgram.kg-1.h-1). Mechanical ventilation was maintained postoperatively in the recovery room until the patient could be extubated. PetCO2, SpO2, fR and F and S plasma concentrations were assessed at the end of the opioid infusion, at extubation, every hour for the first 6 hours, and thereafter every 2 h for a further 10 and 18 h. Time to extubation was the same in both groups (301 +/- 141 and 307 +/- 148 min). At the time, plasma concentrations of F and S were 1.35 +/- 0.9 ng.ml-1 and 0.14 +/- 0.07 ng.ml-1 respectively. Secondary peaks in plasma concentration (78% mean increase in comparison to the previous figure) were observed in 6 patients in group F. No similar peaks occurred in group S. Mean elimination half-life was shorter with sufentanil (457 +/- 130 min) than with fentanyl (325 +/- 132 min) (not significant). The results of this study suggest that sufentanil results less frequently in postoperative secondary peaks than fentanyl.