A Bayesian population pharmacokinetic meta-analysis of individual participant data

The effect of a drug is directly related to drug concentration in blood. Therefore, predicting and describing the time course of drug concentration in the body, i.e. the study of pharmacokinetics, is crucial in establishing and optimizing drug therapy. The immunosuppressant mycophenolate mofetil (with active ingredient mycophenolic acid), is extensively used in renal transplantation in order to prevent acute allograft rejection. Large inter-subject and inter-occasion variability in the exposure of mycophenolic acid in combination with one standard dosing regimen for all renal transplant recipients makes it uncertain if the therapeutic window is reached in every patient. With the pharmacokinetic data of six historical studies, it is tried to find the underlying causes responsible for the inter-subject variability in order to optimize the mycophenolic acid exposure in every patient using Bayesian statistics. As the statistical methods are essential in answering such a research...Show moreThe effect of a drug is directly related to drug concentration in blood. Therefore, predicting and describing the time course of drug concentration in the body, i.e. the study of pharmacokinetics, is crucial in establishing and optimizing drug therapy. The immunosuppressant mycophenolate mofetil (with active ingredient mycophenolic acid), is extensively used in renal transplantation in order to prevent acute allograft rejection. Large inter-subject and inter-occasion variability in the exposure of mycophenolic acid in combination with one standard dosing regimen for all renal transplant recipients makes it uncertain if the therapeutic window is reached in every patient. With the pharmacokinetic data of six historical studies, it is tried to find the underlying causes responsible for the inter-subject variability in order to optimize the mycophenolic acid exposure in every patient using Bayesian statistics. As the statistical methods are essential in answering such a research question and these data is already analysed using using traditional, maximum likelihood, methods by Van Hest et al., emphasis is also put on the comparison of our and their results. The data were described with nonlinear mixed effects models using WinBUGS and its interface PKBUGS. Analysing the data of the individual studies resulted in identifiability problems, i.e. the data of most studies were insufficient in estimating all model parameters with non-informative priors. However, advantages of Bayesian methods emerged, as prior information was increased to obtain accurate posterior summary measures. Combining the data of all studies, i.e. the individual participant data meta-analyses, improved the quality of the data, the analyses and the reliability of the results. Subsequently, the identifiability problems diminished and using close to non-informative prior information resulted in reliable, precise posterior summary measures on almost all model parameters and some covariates were found to explain the intersubject variability. Comparing our results with the results obtained from the analysis with non-Bayesian methods turned out to be hard. More precise and very different estimates of the model parameters were obtained with Bayesian methods. Probably caused by differences between the used models, e.g. in contrast to the model used for non-Bayesian methods, we did not consider covariates with missing values and time-dependent covariates but did correct for inter-study variability. Nevertheless, the contribution of prior information is a major advantage in the analysis of pharmacokinetic data, especially because pharmacokinetic data is hard to collect and highly reliable historical pharmacokinetic studies are available for each drug on the market.Show less

Leiden University Student Repository

Documents

In Collections

A Bayesian population pharmacokinetic meta-analysis of individual participant data