Developing physiologically-based toxicokinetic models to assess human exposure to brominated flame retardants and herbicides

Abstract

PBTK models are mathematical representations of chemical absorption, distribution, metabolism and excretion (ADME). Each model parameter describes a physiological, physicochemical or biochemical process that affects ADME. Distributions can be assigned to parameters that describe population variability and uncertainty. In this project, PBTK models were developed and calibrated using Bayesian analysis to assess human population exposure to herbicides (picloram and haloxyfop) and a brominated flame retardant (TBBPA). These models allowed assessment of internal exposure at the site of action and therefore assessment of the risk of toxicity. For picloram, a rat PBTK model was developed and calibrated. The difference between the predicted and observed toxicokinetic data was interpreted mechanistically leading to transporters being modelled at kidney and liver apical membranes. For haloxyfop mouse, dog and human PBTK models were built. There were two biologically-plausible approaches to fitting the same toxicokinetic data in the mouse PBTK model, and once extrapolated to humans had different implications for internal exposure assessment. A statistical model was developed to reconstruct operator exposure using experimentally determined LogP, mouse and human toxicokinetic data and human biomonitoring (HBM) data. Of the seven operators, one was calculated to be at potential risk of toxicity from his modelled external and internal exposure levels. A human PBTK model was constructed for TBBPA (perfusion-limited) and its major metabolite (permeability-limited). Parameter covariance was identified by calibrating the model to human toxicokinetic data. A biomonitoring equivalent (BE) for TBBPA was derived for a hypothetical adult population which was affected by parameter covariance. Comparison of the BE with HBM data suggested low risk of toxicity for individuals with measured TBBPA plasma concentrations. The findings add to the knowledgebase of PBTK models for use in chemical risk assessment, particularly with regard to the implementation of transporters and the effect of parameter covariance on model extrapolation and prediction.