Institut national de l’environnement industriel et des risques (INERIS), Unité de toxicologie expérimentale, Parc technologique ALATA, BP 2, 60550 Verneuil-en-Halatte
Risk assessment of chemicals requires toxicokinetic studies to determine the association between exposure and the quantity of toxin that reaches target tissues or cells. To characterize this link, experimental data about the agent’s spatiotemporal distribution in the body (e.g., blood concentrations at various times) can be collected and analyzed with parametric models (called toxicokinetic or TK models). These models are generally compartmental and can be based on physiology (PBTK models). Appropriate statistical treatment allows fitting TK/PBTK models to the data. Bayesian analysis takes into account and estimates the uncertainty and variability inherent in TK data. It can integrate prior information on parameter values into the estimation process, thus limiting the need for experimental exposures. In this paper, we detail the Bayesian process of TK analysis, including the estimation (or calibration) of parameters and the checking and validation of models. The process can be completed by choosing between competing models or optimizing the design of experimental protocols. To illustrate this process, we analyze and model the toxicokinetics of 1,3-butadiene, a potential human carcinogen.