Pharmacokinetic model-controlled infusion of midazolam. A prospective evaluation during general anaesthesia

The accuracy of a computer-controlled infusion of midazolam, based on previously published pharmacokinetic parameters, was tested prospectively in 12 adult female patients undergoing general anaesthesia. Anaesthesia consisted of an initial bolus followed by an exponentially decreasing infusion of midazolam given according to body weight, fentanyl, nitrous oxide and vecuronium. Venous blood samples were taken at 15 min-intervals throughout the procedures and for 1-2 h postoperatively. The bias of the model was -5.1% (95% CI -11.3 to 1.2%) and precision 24.8% (95% CI 20.9 to 28.6%). Least squares regression analysis decreased the bias to -2.8% but did not alter precision. Retrospective fitting of an alternative set of published parameters for midazolam resulted in significant deterioration of the model. The precision was similar to that found in past studies of intravenous anaesthetic agents. Further improvement in the accuracy of midazolam infusion awaits improved understanding of the causes of pharmacokinetic variability.