Sensitivity Analysis of the Inputs for Bioactivity-Exposure Ratio Calculations in a NAM-Based Systemic Safety Toolbox

To support regulatory decision-making without animal testing, Next-Generation Risk Assessment (NGRA) frameworks leverage New Approach Methodologies (NAM). A widely used strategy in such frameworks is to calculate Bioactivity-Exposure Ratios (BERs), which compare NAM assay-derived points of departure (PODs) and estimated human internal exposures. However, key methodological choices in NGRA, such as toxicokinetic modeling software, POD sources, and adjustments for in vivo and in vitro free concentrations, may introduce uncertainty. To determine the robustness of our previously reported NGRA workflow, we conducted a comprehensive sensitivity analysis of these variables as applied to 35 chemicals covering a range of end uses including consumer, pesticide and pharmaceutical. Physiologically based kinetic (PBK) modeling using both commercial (GastroPlus) and open-source (httk) software was evaluated across three levels of parameterization. The httk model produced individual and population C_max estimates that are comparable to GastroPlus when the chemical-specific input parameters are aligned. BERs were calculated using C_max estimates, previously reported in vitro PODs from human-relevant cell-based assays, ToxCast, and fixed values based on internal threshold of toxicological concern (iTTCs), with and without adjustment for unbound fractions in plasma and media. We found that BERs from human-relevant cell-based PODs and nominal concentrations best matched risk classifications, with improved predictive performance over iTTC or ToxCast-derived PODs. Consideration of population variability in PBK had limited impact on risk prediction. Adjustments for free concentrations did not enhance classification or in vitro/in vivo concordance. These findings support the potential of human-relevant in vitro PODs and nominal dose metrics in NGRA, and demonstrate that accessible, well-parameterized PBK tools can enhance efficiency and reproducibility.