Flexible multivariate nonlinear models for bioequivalence problems

Abstract

Modelling the concentration of a drug in the bloodstream over time is usually done using compartment models. In pharmacokinetic data, they turn into highly nonlinear mixed-effects models (NLMEMs) when we take the heterogeneity between subjects into account. Fitting of NLMEMs can be difficult and may involve complex algorithms, with convergence critically depending on the initial values and maybe requiring data transformations. In this article, we propose a flexible alternative to the usual parametric compartment models, inspired by the Multivariate SuperImposition by Translation and Rotation (MSITAR) model but adapted to be applicable in this new field. A fully parametric one-compartment NLMEM is considered for comparison. We make use of a Bayesian approach and illustrate the method on a real dataset where the interest lies in contrasting the average and individual bioequivalence of a test and reference formulation of an anti-hypertensive drug.