You are here:
Using publically available data and quantitative models of uncertainty to characterize composition of consumer products in a simulation model of chemical exposure
Citation:
Price, P., K. Dionisio, K. Isaacs, AND K. Phillips. Using publically available data and quantitative models of uncertainty to characterize composition of consumer products in a simulation model of chemical exposure. 2018 ACS National Meeting & Expo, Boston, MA, August 19 - 23, 2018.
Impact/Purpose:
This presentation will give EPA an opportuinty to share its research approach for collecting and organizing chemical composition data in consumer products.
Description:
EPA is developing a probabilistic software tool (the Human Exposure Model, HEM) to determine human exposures to chemicals from the use of certain consumer products. A challenge for this project is that composition of many products are considered confidential. To address this, EPA has established: 1, a set of 241 Product Use Categories (PUCs) that are based on the purpose the product fulfills and the products’ exposure-related characteristics; and 2, a database (Chemical Product Database or CPDat) of publically available composition data organized by PUC. The composition data comes from two sources, material safety data sheets (MSDSs) and lists of ingredients disclosed by the manufacturer. Each source reports useful, but incomplete composition data. MSDSs often report ingredients’ weight fractions (WFs) as ranges. Ingredient lists generally do not report WFs but ingredients are typically listed in descending order of WF. When modeling use of products by individuals, HEM assigns corresponding product compositions using a two-stage process. First, a composition record of a product from the same PUC is selected from CPDat. Second, probabilistic models estimate distributions of WF values for each reported chemical in the product that are consistent with the available data and have an equal probability of occurring. When ranges of WFs for a product are reported in an MSDS, the model assumes a uniform distribution across these ranges. WFs for products with ingredients list data are determined using a predictive approach based on the length of the list and the rank of each chemical. HEM samples from these distributions to assign a WF for each chemical in each product. The values of WF are then used with the chemicals’ physicochemical properties and product-specific exposure scenarios to estimate exposure for each product ingredient.
