Genetic polymorphism and environment each influence individual variability in drug metabolism and disposition. It is preferable to predict such variability, which may affect drug efficacy and toxicity, before drug administration. We examined individual differences in the pharmacokinetics of atorvastatin by applying gas chromatography-mass spectrometry-based metabolic profiling to predose plasma samples from 48 healthy volunteers. We determined the level of atorvastatin in plasma using liquid chromatography-tandem mass spectrometry. With the endogenous molecules, which showed a good correlation with pharmacokinetic parameters, a refined partial least-squares model was calculated based on predose data from a training set of 36 individuals and exhibited good predictive capability for the other 12 individuals in the prediction set. In addition, the model was successfully used to predictively classify individual pharmacokinetic responses into subgroups. Metabolites such as tryptophan, alanine, arachidonic acid, 2-hydroxybutyric acid, cholesterol, and isoleucine were indicated as candidate markers for predicting by showing better predictive capability for explaining individual differences than a conventional physiological index. These results suggest that a pharmacometabonomic approach offers the potential to predict individual differences in pharmacokinetics and therefore to facilitate individualized drug therapy.
Keywords: atorvastatin; metabolomics; personalized medicine; pharmacokinetics; pharmacometabonomics; precision medicine; prediction.