Evaluation of the effect of residual error covariance for a parent‐metabolite population pharmacokinetic model

• Published in: Clinical pharmacology and therapeutics 2005-02, Vol.77 (2), p.P91-P91
• Main Authors: Houk, B. E., Gastonguay, M. R., Amantea, M.
• Format: Article
• Language: English
• ISSN: 0009-9236; 1532-6535
• DOI: 10.1016/j.clpt.2004.12.242

Background The residual error of parent and metabolite concentrations in pharmacokinetic (PK) samples can be correlated due to issues such as mislabeled timing, sample extraction, and misspecification, although it is not typically estimated in population modeling. We investigated the impact of excluding this covariance on population estimates and clinical trial simulations. Methods Multivariate data from a single‐dose oral study in Japanese and Caucasian male healthy volunteers (n=25) were described with a parent‐metabolite population PK model using nonlinear mixed‐effects modeling. Weight effects on clearances and volumes of distribution were included. Covariance between parent and metabolite residual errors was estimated or excluded. A multiple‐dose, study was then simulated using both models. Results Both models described the observed data well. Parameter estimates for clearances and volumes were similar between models. The correlation between proportional and additive residual errors was estimated at 61 and 100%, respectively. Multiple‐dose simulations of 100 trials displayed that excluding the covariance term influenced predicted covariate effect on exposure by −4 to +17%. Conclusions The exclusion of covariance in residual error may have only minor impacts on resulting population parameter estimates, however the distribution of simulated multiple‐dose trial outcomes may become exaggerated, which may influence the predicted exposure in subpopulations. Clinical Pharmacology & Therapeutics (2005) 77, P91–P91; doi: 10.1016/j.clpt.2004.12.242