Predictive accuracy of continuous alfentanil infusion in volunteers: variability of different pharmacokinetic sets.

To evaluate the variability of the predictive accuracy of alfentanil by using different pharmacokinetic data sets, eight healthy young male adult volunteers were given the same alfentanil infusion for 4 h. Nineteen venous blood samples were taken from each volunteer, and alfentanil concentrations were titrated by radioimmunoassay. For each volunteer, the pharmacokinetic variables of a two-compartment model were calculated, averaged, and considered as a reference set. Based on the infusion profile given to the volunteers, central compartment concentrations were calculated by using the reference set and nine previously published pharmacokinetic sets of alfentanil concentrations in healthy adults. The bias, inaccuracy, and dispersion of each data set were assessed by determining the median performance error, the median absolute performance error (MDAPE) and the 10th and 90th percentiles, respectively. By using the pharmacokinetic variables of the volunteers, the predictive accuracy was excellent (MDAPE, 7.25%). Among the 10 averaged pharmacokinetic sets, there was a significant correlation between their bias and clearance (R2 = 0.996). The reference set had the best predictive accuracy (MDAPE, 23.6%). Five sets from the literature also showed a reliable predictive accuracy but four other sets with a clearance more than 5 mL.kg-1.min-1 and derived from a large bolus injection were inaccurate (MDAPE > 50%) as they underestimated the alfentanil concentrations. We conclude that pharmacokinetic sets derived from large bolus should not be selected to accurately predict alfentanil infusion.