Two total intravenous anaesthesia techniques were compared in an open study of 80 ambulatory patients undergoing ENT endoscopic procedures randomly assigned to two groups: Group I midazolam-flumazenil n = 40, Group II propofol n = 40. The mean doses including induction were 0.75 +/- 0.31 mg kg-1 h-1 for midazolam and 171 +/- 64 micrograms kg-1 min-1 for propofol for 46.3 +/- 17.7 min and 50.3 +/- 24.8 min respectively. At the end of the procedure flumazenil 8.1 +/- 1.9 micrograms kg-1 was administered to Group I patients followed by a flumazenil continuous infusion at a minimal arousal rate (MAR) of 0.24 +/- 0.1 micrograms kg-1 min-1, and propofol discontinued in Group II patients. Baseline mean arterial pressure (MAP) and heart rate (HR) were similar in both groups and remained so during the procedure and recovery. In patients with cardiovascular disease, large variations (greater than or equal to 40% of baseline values) occurred more frequently in the propofol group whereas large variations in patients with no cardiovascular disease occurred more frequently in the midazolam group (P less than 0.05). Early recovery was more rapid after midazolam (P less than 0.05) whereas late criteria for recovery (maze and ambulation tests) were met more rapidly after propofol (P less than 0.05). It is concluded that with the midazolam-flumazenil sequence, early recovery is faster and haemodynamic stability better maintained in poor cardiovascular risk patients, whereas with propofol, street-fitness is more rapidly obtained, and haemodynamic stability better maintained in good risk patients.