Applicability of next-generation risk assessment generic PBK models and their performance within the chemical domain

Physiologically based kinetic (PBK) models are a cornerstone of in vitro to in vivo extrapolation and are therefore central to next-generation risk assessment (NGRA). This study assesses the performance of a generic, high-throughput PBK model for 26 organic chemicals spanning a wide range of physicochemical properties. Model predictions based on standard in silico and in vitro input parameters were benchmarked against published, chemical-specific PBK models and in vivo kinetic data. Predicted plasma maximum concentrations (Cmax) were within tenfold of in vivo values for 50% of the chemicals and within threefold for 31%. Published PBK models generally showed closer agreement with in vivo Cmax, likely because they incorporate in vivo-derived kinetic parameters or chemical-specific kinetic processes. An analysis of the chemical space revealed a lipophilicity-dependent bias, where Cmax for lipophilic compounds tended to be underpredicted, whereas hydrophilic compounds were overpredicted. No other consistent trends were observed with respect to physicochemical or ADME descriptors, possibly reflecting the limited dataset size. Overall, while the generic PBK model workflow appears suitable for NGRA applications, the observed lipophilicity-dependent bias indicates that chemical-specific corrections or refinements may be required for more accurate predictions.