Evaluation of a high-throughput in-vitro-to-in-vivo extrapolation (IVIVE) workflow for the prioritization of potential developmental toxicity of chemicals_suppl2

New approach methodologies (NAMs) are rapidly being developed to help improve the speed of risk assessment and reduce the use of animals. In vitro to in vivo extrapolation (IVIVE) is necessary to translate NAM data to human exposures. While past IVIVE efforts have demonstrated impressive success, several specialized scenarios exist where current IVIVE methods have not been tested, including pregnancy and developmental toxicity. To that end, this investigation proposes a preliminary IVIVE workflow for identification of potential developmental toxicants. Readily available in vitro and in vivo data with developmental toxicity endpoints were aggregated from the US Environmental Protection Agency’s CompTox Chemicals Dashboard. In vitro distribution models (i.e. Armitage model) and both generic (httk) and bespoke physiologically-based pharmacokinetic (PBPK) models were used to estimate exposures from blood concentrations associated with in vitro bioactivity (reverse dosimetry) and NAM-based bioactive doses were compared to in vivo endpoints (LOAELs) where available. Based on literature sources, this method identified chemicals as a high, medium, or low priority for follow-up as a developmental toxicant. Of the 23 chemicals with in vitro developmental toxicity assays, 7 had a NAM-based human oral equivalent dose (hOED) that was lower for developmental assays than for all available assays, indicating that the use of all in vitro data to derive a hOED would generally provide the most conservative approach. Potential data streams and refinements for improvement of the IVIVE workflow are also discussed. Plain language summaryTo reduce or even replace the use of animals in toxicology testing, a number of new approach methodologies (NAMs) are currently being developed. Developmental toxicity (DevTox), or toxicity to the developing embryo/fetus, is still in its early stages of NAM generation. Therefore, this study aimed to use public data to prioritize which chemicals may be more likely to cause DevTox. To do this, data on 23 example chemicals with known DevTox status were used in combination with computer models for estimating chemical concentrations in lab tests (in vitro distribution models) and for estimating chemical concentrations in the human body (pharmacokinetic models) to see if they could be prioritized as a potential DevTox chemical. Additionally, we used those models to estimate a daily dose of a chemical that may lead to DevTox. Ultimately, this study provides an initial step for DevTox NAMs, though a number of areas for improvement are identified.