Prediction of Composition and Emission Characteristics of Articles in Support of Exposure Assessment

Abstract

The risk to humans from chemicals in consumer products is dependent on both hazard and exposure. The prediction and quantification of chemical exposure from household articles such as furniture and building materials is an ongoing effort. As opposed to, for example, cosmetic formulations which are regulated by the FDA, the chemical composition of articles are less clear, as are the resulting chemical emission characteristics and exposures. We have developed a modeling methodology to predict the weight fraction of chemicals in a polymeric substrate and corresponding emission characteristics based on chemical and substrate structure. We constructed a database of reported and measured chemical concentrations in articles from publicly available sources including Health Product Declarations (HPDs). This database was then used to train a random forest algorithm which predicts a chemical weight fraction bin based on chemical structure and properties. From the predicted weight fractions and chemical properties, we applied a group contribution method, UNIversal quasi-chemical Functional-group Activity Coefficients-Free Volume (UNIFAC-FV), to approximate steady-state gas phase concentrations (y0) at the substrate surface. The model was compared to published experimental y0 data from chamber experiments. The resulting estimates of y0 can be used to parameterize existing high-throughput exposure models for substrate-chemical combinations found in consumer articles. Thus, from only the “first-principles” of chemical and substrate molecular structure, we can generate an estimate of chemical exposure for use in screening-level risk evaluation. This abstract does not reflect EPA policy.