You are here:
RECONSTRUCTING EXPOSURE SCENARIOS USING DOSE BIOMARKERS - AN APPLICATION OF BAYESIAN UNCERTAINTY ANALYSIS
Citation:
Sohn, M. D., T. E. McKone, M L. Rigas, J N. Blancato, F. W. Power, AND A. M. Tsang. RECONSTRUCTING EXPOSURE SCENARIOS USING DOSE BIOMARKERS - AN APPLICATION OF BAYESIAN UNCERTAINTY ANALYSIS. Presented at International Society of Exposure Analysis Annual Meeting, Charleston, SC, November 4-8, 2001.
Impact/Purpose:
Research will be conducted to develop and apply integrated microenvironmental, and physiologically-based pharmacokinetic (PBPK) exposure-dose models and methods (that account for all media, routes, pathways and endpoints). Specific efforts will focus on the following areas:
1) Develop the Exposure Related Dose Estimating Model (ERDEM) System.
Includes: Updating the subsystems and compartments of the ERDEM models with those features needed for modeling chemicals of interest to risk assessors;
Designing and implementing the graphical user interface for added features.
Refining the exposure interface to handle various sources of exposure information;
Providing tools for post processing as well as for uncertainty and variability analyses;
Research on numerical and symbolic mathematical/statistical solution methods and computational algorithms/software for deterministic and stochastic systems analysis.
2) Apply ERDEM and other quantitative models to understand pharmacokinetics (PK) and significantly reduce the uncertainty in the dosimetry of specific compounds of regulatory interest.
Examples of the applications are:
exposure of children to pesticides
study design
route-to-route extrapolation
species extrapolation
experimental data analysis
relationship between parametric uncertainty and the distribution of model results
validity of scaling methods within species
validity of scaling methods from one species to another species
reduction of uncertainty factors for risk assessment
Description:
We use Bayesian uncertainty analysis to explore how to estimate pollutant exposures from biomarker concentrations. The growing number of national databases with exposure data makes such an analysis possible. They contain datasets of pharmacokinetic biomarkers for many pollutants (e.g., CDC National Health and Nutrition Examination Survey (NHANES)) and detailed information about human activity and consumption patterns (e.g., the EPA National Human Exposure Assessment Survey (NHEXAS), the EPA National Human Activity Pattern Survey(NHAPS), and the USDA national survey of food and water consumption). However, variability and uncertainty in the assessed information, and in estimated (or predicted) pollutant concentrations in foods, air, water, and soils, complicates the determination of unique links between biomarkers and exposures. Moreover, biomarkers only provide snap-shots of exposure events and may not describe their temporal history. Thus, reconstructing exposure scenarios, and assessing their quality, must include each component of the exposure-to-biomarker concentration relationship. We explore these issues using a Bayesian uncertainty analysis framework. A physiologically-based pharmacokinetic (PBPK) model is linked with an indoor air model, and variability (or uncertainty) in each is evaluated. The Bayesian framework allows us flexibility for evaluating multiple exposure scenarios and alternative datasets, simultaneously. Using surrogates for volatile and semi-volatile chemicals, we demonstrate how reconstructed exposure scenarios can contain great uncertainties, even with well-documented human activity and biomarker information. We also present preliminary findings of our ongoing work to determine: (i) what exposure assessment information is useful in the reconstructions; (ii) what information, other than those commonly gathered in exposure assessments, could improve them; and (iii) how quality and quantity of data affects the reconstructions, thereby assisting future data gathering. For example, we find that uncertainty in describing metabolism has great impact on estimates of exposure to volatile organic compounds, like trichloroethylene. We also find that since body weight and breathing rate variability is well-described in the surveys, additional information will not significantly improve exposure estimates; in contrast, additional information about still highly uncertain exposure factor like dermal contact duration may significantly improve them.
This work has been funded in part by the United States Environmental Protection Agency under Interagency Agreement No. DW-89938190 with Lawrence Berkeley National Laboratory. It has been subjected to Agency review and approved for publication.