Predictive performance of two software packages (USC ∗ PACK PC and Abbott PKS system) for the individualization of amikacin dosage in intensive care unit patients

Abstract

Many dosing methods (nomogram, pharmacokinetic methods, Bayesian methods) can be used for the individualization of amikacin dosing. Among these methods, it is now well known that the Bayesian method provides a rapid and accurate means for individualizing dosage requirements for patients with diverse pharmacokinetic profiles. However, one problem has not been fully resolved. Should we use population-based parameters reflecting the patient population being monitored or should we used general population parameters? The aim of this study was to answer this question using two widely used software programs (USC ∗ PACK PC and Abbott PKS system) and two different population parameters sets. Predictive performance of these methods was assessed with respect to the prediction of amikacin serum concentrations in intensive care unit (ICU) patients. Our results show that the differences between predicted and measured concentrations were unbiased when the population parameters used were adequate. Precision values were comparable with previously reported values. The predictive performance of the two tested software programs are very comparable in ICU patients. In addition, we demonstrated that performance can be enhanced when using population-based parameters which reflect the patient population being monitored. It is therefore advisable for each user to properly characterize each particular patient population.