ABSTRACT
Physiologically based pharmacokinetic (PBPK) models, like all computational models, are simplified representations of a more complex system. This chapter presents a brief introduction to application of statistical methods to PBPK models, with a particular focus on the concepts of identifiability and sensitivity as they relate to data, PBPK models, and their predictions. Rigorous statistical characterization of uncertainty and variability has not yet been routinely applied, as PBPK models and the data that underlie them present a number of challenges to traditional statistical inference. Ideally, PBPK model-based analyses should not only make “point estimate” predictions based on the data, but also characterize the likely range of predictions that are consistent with available data. Perchloroethylene (PERC) is a dry-cleaning solvent and a common environmental contaminant. Because it is presumed that metabolites, rather than the parent compound, are responsible for PERC toxicity, a number of groups have developed human PBPK models to estimate the amount of PERC metabolized by the body.
