Pharmacokinetic Model Selection for Target Controlled Infusions of Propofol

Background Computer-assisted target controlled infusions (TCI) result in prediction errors that are influenced by pharmacokinetic variability among and within patients. It is uncertain whether the selection of a propofol pharmacokinetic parameter set significantly influences drug concentrations and clinical acceptability. Methods Thirty patients received similar propofol TCI regimens after being randomly allocated to one of three parameter sets. Arterial and venous concentrations were measured and prediction errors calculated from pooled and intrasubject data. Results Arterial propofol concentrations in the Dyck group revealed greater bias (mean 43%) than did those in the Marsh (-1%) and Tackley (-3%) groups. The Dyck group also showed greater inaccuracy (mean:47%) than the Marsh (29%) and Tackley (24%) groups. There was little tendency for measured concentrations to vary from targeted values over time (divergence). Variability about an observed mean in individual patients (wobble) was low. Venous propofol concentrations were initially much less than arterial concentrations, but this difference decreased over time. Conclusions Although it may be preferable to administer propofol TCI by using a locally derived parameter set, it is acceptable to use a model from elsewhere. The Marsh and Tackley models produced equally good performance and are appropriate for propofol TCI within the range of 3-6 micrograms/ml. The Dyck model was less accurate at maintaining anesthetic concentrations, possibly because it was derived from low concentrations. Concentrations in blood, the most sensitive indicators of performance, demonstrated differences among the parameter sets. Clinically, TCI worked well, and by clinical criteria, the choice of pharmacokinetic model did not appear to make a difference.