You are here:
High Throughput PBTK: Evaluating EPA’s Open-Source Data and Tools for Dosimetry and Exposure Reconstruction
Citation:
Wambaugh, J., R. Pearce, C. Ring, Woodrow Setzer, AND B. Wetmore. High Throughput PBTK: Evaluating EPA’s Open-Source Data and Tools for Dosimetry and Exposure Reconstruction. 2016 ISES Annual Meeting, Utrecht, NETHERLANDS, October 10 - 13, 2016.
Impact/Purpose:
This talk provided an update to an international audience about the state of science being conducted within the EPA’s Office of Research and Development to develop and refine approaches that estimate internal chemical concentrations following a given exposure, known as toxicokinetics. Experimental data across hundreds of chemicals have been generated and have now been incorporated into an open-source platform along with pharmacokinetic modeling tools to allow scientists to make use of the data for their own chemical safety assessments. Recent efforts that refine related modeling tools to allow their broader applicability to environmental chemicals were also described.
Description:
Thousands of chemicals have been profiled by high-throughput screening (HTS) programs such as ToxCast and Tox21; these chemicals are tested in part because most of them have limited or no data on hazard, exposure, or toxicokinetics (TK). While HTS generates in vitro bioactivity data for characterizing potential chemical hazards, TK models are needed to inform in vitro to in vivo extrapolation (IVIVE) to real world situations. The U.S. Environmental Protection Agency has created a new tool (R package “httk”) for building, simulating, and evaluating TK and physiologically-based TK (PBTK) models for both IVIVE and exposure inference from biomonitoring data (i.e., reverse dosimetry). We are now able to rapidly parameterize generic PBPK models using in vitro data to allow IVIVE for 543 chemicals. Our high throughput toxicokinetics (HTTK) tools were implemented in the R statistical platform in part to allow statistical analysis of both IVIVE and our TK models. We have statistically evaluated our TK predictions using in vivo measurements of human steady-state serum concentrations, rat serum concentrations, rat tissue partition coefficients, and human volumes of distribution. We find that for many chemical classes our methods and models perform reasonably, and that we can begin to identify chemical classes for which our methods perform poorly. Our PBTK models are parameterized with not only chemical-specific parameters derived from in vitro measurements and predicted from chemical structure; but also with physiological parameters for a virtual population. We simulate population physiological parameters based on data from the most recent U.S. Centers for Disease Control National Health and Nutrition Examination Survey (NHANES), which describe distributions of demographic and anthropometric quantities in the modern U.S. population. A Monte Carlo approach, accounting for the correlation structure in physiological parameters, can be used to estimate margins between IVIVE predicted bioactive doses and estimates of exposure for the most sensitive portion of the population. While these new models are expected to have limited accuracy due to their simplicity and generalization of assumptions, the confidence in the predictions can be in part assessed using our comparison to TK in vivo data. Ultimately, we are working to identify the chemicals for which these new tools may be used with confidence, and to identify those chemicals where alternative approaches are needed.
