Assessing the influence of inter-individual variability on the pharmacokinetics of environmental chemicals using PBPK modeling

Overview & Synopsis

Physiologically based pharmacokinetic (PBPK) models' ambition is to describe precisely the mechanisms of absorption, distribution, metabolism and elimination of chemicals in and out of the body. Their relative success at inter-species or inter-route extrapolation and prediction of inter-individual susceptibility is the main reason of their fame. Obviously, there is no set limit to their complexity and Occam's razor gets duller when only computing power shapes cutting edge research. The appeal of fast throughput has led to the development of large models with complex structure, probably too complex for most substances, but versatile. The PBPK models on the market are now able to screen many substances without having to make a costly investment in human intelligence and parsimony. The increasing complexity of PBPK models is not so much due to changes in their fundamental structure (they are still compartmental models) than to the addition of multiple ancillary models such as quantitative structureproperty relationships (QSPR) and parameter databases. QSPR models render them "substance-independent" and databases "subject-independent": A significant part of inter-individual variability and susceptibility to toxicity can be ascribed to known differences in pharmacokinetic parameters between sexes, ages, ethnic origins, physiological or pathological states. Monte Carlo sampling of these databases simulates mixed populations in which "susceptible" individuals can be studied to identify the particular parameter configurations at the origin of their singular pharmacokinetic behaviour (Jamei et al. 2009). That is an a priori approach, going forward from "first principles". It is useful, however, to be able to identify factors leading to susceptibility from population studies, as in the population pharmacokinetic approach. Such a posteriori observation-driven studies require much more sophisticated statistical analyses. So far, Bayesian numerical tools, such as Markov chain Monte Carlo simulations, are the best way to integrate the data collected together with the un-ignorable prior knowledge on human physiology in the context of complex models (Mezzetti et al., 2003). A current challenge, both in PBPK model structure and parameterisation, is to estimate and predict the part of susceptibility due to pharmacokinetic, metabolic or pharmacodynamic interactions between the substances we are exposed to through food, medical treatment, work, or lifestyle ... Traditional modelling approaches and concepts meet their match here, and we may need to turn to the new arsenal of systems biology (Bois, 2010).