Case study application of high-throughput new approach methodologies for exposure to the interpretation of matched biomarker and indoor media measurements.

backgroundThe U.S. Environmental Protection Agency, under its exposure forecasting (ExpoCast) project, has developed new approach methodologies for understanding complex exposures, including approaches that infer chemical intake from metabolite measurements, high-throughput models that predict multi-pathway exposures, and machine-learning (ML) methods for filling data gaps using chemical structure and information.

objectiveNew and existing models were used to evaluate the relationship between measured indoor media and urine concentrations in an indoor exposure study to understand exposure pathways for potential endocrine-disrupting compounds (EDCs).

methodsIndoor air, dust, and urine samples from 120 females (aged 60 to 80 years) were previously analyzed for the presence of EDCs, including pesticides, flame retardants, and consumer product chemicals or their metabolites. A Bayesian air/dust partitioning model was applied to infer concentrations below the LOD. These estimates were used to train ML models to predict air and dust concentrations for other chemicals from use and structure descriptors; the models were validated with literature data. The ML models were used to develop media concentration estimates for metabolite parents that were unmeasured in air or dust. Measured or predicted concentrations were then used to estimate exposures for 49 metabolite parent chemicals using a high-throughput (HT) exposure model. These exposure estimates were combined with existing HT exposure predictions for food contact pathways and compared with the inferred parent exposures derived from the urine metabolite measurements.

resultsSuccessful ML models for media concentrations could be built (R IMPACT STATEMENT: Understanding exposure to chemicals indoors is complex, as chemicals may have multiple sources and pathways. Exposure studies can identify substances present indoors, but it can be difficult to relate measured biomarkers with corresponding measured media concentrations and other potential sources; and biomarkers such as urine metabolites may have multiple parents and measurements may often be censored. Here, high-throughput modeling methods from the US EPA's Exposure Forecasting (ExpoCast) project are used to relate air and dust concentrations with measured urine metabolite concentrations obtained in an indoor exposure study of potential endocrine-disrupting chemicals.