Population pharmacokinetic modeling to understand antineoplastic treatment

Abstract

<strong>Introduction:</strong> The pronounced variability in pharmacokinetics of antineoplastic drugs caused by known and unknown sources translates into variability in therapeutic outcome. This is a major concern for these drugs with a narrow therapeutic index because even small changes in concentrations may lead to untoward results. Tools to handle this problem and to predict optimal doses for patients remain to be established.<br /> <strong>Objectives:</strong> The main objective was to develop nonlinear mixed effects (NLME) models to understand the pharmacokinetics / pharmacodynamics of antineoplastic drugs, to quantify the associated variability and its sources and subsequently to make use of these models to predict drug exposure / toxicity under different therapeutic regimens.<br /> <strong>Methods:</strong> A standard diagnostic tool for NLME models called visual predictive checks (VPCs) was adapted to account for multimodality from unknown sources in distributions of pharmacokinetic parameters. Allocation of patients to a subpopulation was compared based on individual probability or on overall likelihood. This approach (mixture VPCs) was tested using both simulated data and real data. Empirical pharmacokinetic NLME models were developed for mitotane and methotrexate, whereas a semi-physiological pharmacokinetic-pharmacodynamic model was developed for 5-fluorouracil. Further evaluation was performed to identify covariates (e.g., demographic characteristics, organ function) influencing the pharmacokinetics and pharmacodynamics of studied drugs. Simulations were designed to understand and visualize drug exposure and toxicity under different dosing schedules in virtual subjects.<br /> <strong>Results:</strong> Mixture VPCs based on individual probability were found to be most useful to capture model misspecifications, which were not evident from the former classical approach. Mixture VPCs were further useful to diagnose bias associated with the allocation of individuals to subpopulations. An enzyme autoinduction model for mitotane supported concentration dependent metabolic enzyme induction, leading to change in drug clearance. Body mass index was related to mitotane volume of distribution. Simulations showed that a high dose regimen is most suitable to achieve appropriate exposure early, with a first therapeutic drug monitoring on day 16 of treatment. Myelosuppression under 5-fluorouracil was characterized by a linear relationship with the plasma concentrations. Body surface area influenced the pharmacokinetics of the parent drug and its metabolite, whereas cisplatin co-administration increased myelosuppression. Any covariate effects regarding methotrexate pharmacokinetics were clinically irrelevant due to marginal explanation of between-patient variability. Specifically, body surface area based methotrexate dosing was not found to be superior to flat dosing.<br /> <strong>Conclusions:</strong> The methods developed and the findings of the research work presented in this thesis are useful to assist the adjustment of doses and dosing schedules in antineoplastic drug treatment.