The CNS effects resulting from the combined administration of midazolam and flumazenil were studied in 8 healthy volunteers to develop a model of the pharmacokinetic-pharmacodynamic interaction. Electroencephalograms (EEG) were recorded between Fp1-M1 and Fp2-M2. The EEG parameter total number of waves between 12 and 30 Hz (TNW12-30) derived by aperiodic analysis was used to quantify the effect. Following a 15 min baseline EEG recording, infusion of placebo or flumazenil was started. Infusion regimens for flumazenil were designed so that 'steady-state' concentrations of 10 and 20 micrograms/L were obtained. Doses of midazolam 15, 30 and 60 mg over 5 min were given 30 min after the start of placebo infusion (session A) or flumazenil infusion to 10 micrograms/L (session B) or 20 micrograms/L (session C), respectively. Venous blood samples were taken until 8 h after the start of the flumazenil or placebo infusion. A sigmoid maximum effect (Emax) model was used to characterise the relationship between the plasma concentration of midazolam which is in equilibrium with the effect compartment concentration (Cem) [Cem/Kp] and TNW12-30. Within 2 to 5 min of starting the midazolam infusion all subjects fell asleep, with loss of eyelid reflex. They awoke between 25 and 82 min later in all 3 sessions. The mean (+/- SD) plasma drug concentrations of midazolam corresponding to half the maximum increase in TNW12-30 (EC50) were 276 +/- 64, 624 +/- 187 and 1086 +/- 379 micrograms/L in sessions A, B and C, respectively. The half-lives reflecting equilibration between plasma concentration and effect (t1/2ke0), estimated by a nonparametric method, were 2.2 +/- 1.2, 3.3 +/- 3.3 and 2.9 +/- 1.2 min for the 3 different sessions. Emax and N were not affected by flumazenil. In each subject the plot of the average measured steady-state plasma flumazenil concentration versus the EC50 of midazolam showed a linear relationship. The plasma concentration of flumazenil that doubled the EC50 of midazolam (Cf,2) was 6.5 +/- 1.0 micrograms/L. The observed interaction is consistent with the competitive nature of the antagonism of midazolam by flumazenil.