Investigating the uncertainty of prediction accuracy for the application of physiologically based pharmacokinetic models to animal-free risk assessment of cosmetic ingredients

Abstract

Physiologically based pharmacokinetic (PBPK) models are considered useful tools in animal-free risk assessment. To utilize PBPK models for risk assessment, it is necessary to compare their reliability with in vivo data. However, obtaining in vivo pharmacokinetics data for cosmetic ingredients is difficult, complicating the utilization of PBPK models for risk assessment. In this study, to utilize PBPK models for risk assessment without accuracy evaluation, we proposed a novel concept—the modeling uncertainty factor (MUF). By calculating the prediction accuracy for 150 compounds, we established that using in vitro data for metabolism-related parameters and limiting the applicability domain increase the prediction accuracy of a PBPK model. Based on the 97.5th percentile of prediction accuracy, MUF was defined at 10 for the area under the plasma concentration curve and 6 for Cmax. A case study on animal-free risk assessment was conducted for bisphenol A using these MUFs. As this study was conducted mainly on pharmaceuticals, further investigation using cosmetic ingredients is pivotal. However, since internal exposure is essential in realizing animal-free risk assessment, our concept will serve as a useful tool to predict plasma concentrations without using in vivo data.