Pharmacometrics of antiepileptic drugs: modeling and simulation-based studies of lamotrigine and carbamazepine pharmacokinetics

Abstract

The population pharmacokinetic modeling plays a pivotal role in quantitative learning about drugs from sparse data collected in clinical studies. It provides crucial information needed for individualization of dosage regimens especially in special population in which the intensive pharmacokinetic studies are of ethical concern. Lamotrigine and carbamazepine are antiepileptic drugs commonly used in elderly patients; however, dosing these drugs is based largely on studies from adult patients. Pharmacokinetic information of these drugs in elderly patients is limited. The aims of the current dissertation were to determine the population pharmacokinetic parameters of lamotrigine and carbamazepine in the community-dwelling elderly patients, and to quantitatively identify factors that have significant effects on these parameters. The further aim was to characterize the time course of carbamazepine deinduction by an enzyme turnover model. Due to the presence of collinearity between covariates during the covariate model development of lamotrigine, the effect of collinearity on power, bias, and precision of the parameter estimates in the population covariate model was further investigated by means of simulations. The population pharmacokinetic models of lamotrigine and carbamazepine were successfully developed and the models adequately described the data sets. The important information of drugs' pharmacokinetics was obtained and it can be beneficial in developing dosing strategies for elderly patients receiving these drugs. The time course of carbamazepine deinduction was well described by an enzyme turnover model. This model allowed the estimation of the half-life of the induced enzymes involving carbamazepine metabolism which is the important parameter for characterizing the time course of carbamazepine deinduction process. The power of selecting the true covariates depends on sample size of the data set, the magnitude of covariate coefficient, and the degrees of correlation. The investigation of the effect of collinearity between two true covariates revealed that an increasing collinearity between two true covariates not only decreases the power of selecting the true covariate model, but also leads to the biased and imprecise parameter estimates. The results from this study improve the understanding of how and to what extent the collinearity affects the parameter estimates in the covariate model building.