 |
||
|
Exposure Research |  |
 |
||
|
|
|
|
|||||||||
|
|
Analysis of the National Human Exposure Assessment Survey (NHEXAS) Questionnaire and Measurement Data2003 Research Abstracts - Table of Contents National Exposure Research Laboratory - FY03 Research AbstractScientific Problem and Policy IssuesThe National Human Exposure Assessment Survey (NHEXAS) program was designed to address some of the limitations of single-chemical and single-media exposure studies. The purpose of NHEXAS is to evaluate comprehensive human exposure to multiple chemicals on a community and regional scale to identify predictors of exposure. Ultimately, the EPA anticipates that the information gained from NHEXAS will help individuals, communities, states, the EPA and other organizations understand the greatest health risks from various chemicals and decide whether steps to reduce those risks are needed. The NHEXAS pilot studies measured human exposures to several potentially high-risk chemicals for over 500 participants in three areas of the country: the state of Arizona; the greater Baltimore, Maryland area; and the EPA Region 5 states (IL, IN, MI, MN, OH, and WI). The studies measured people's exposures to environmental pollutants in selected environmental media, and "biomarkers" of exposure in their blood or urine. The participants also completed multiple questionnaires containing more than 200 questions to help identify possible sources of exposure and to characterize activities that might contribute to exposure. Following the pilot study, EPA's Science Advisory Board recommended that EPA develop a strategy to analyze the data in a way that makes optimal use of the data. To develop background information for this strategic plan, EPA convened a workshop to identify the types of scientific analyses needed to interpret the NHEXAS data. Input was obtained from scientific and policy experts from various EPA organizations, other federal agencies, states, academia, and private institutions. Based on the possible analysis projects identified, EPA developed a strategy for the analyses of the NHEXAS pilot study data. The work described in this research abstract reports on the outcome of the initial implementation of the data analysis strategy. Research ApproachFor the Arizona and Region 5 studies, the analysis approach included three phases: (1) data review and summarization, (2) questionnaire variable reduction, and (3) model-based analyses. The large number of questionnaire variables (more than 800, since some questions had multiple parts and others were asked daily), relative to the number of participants in each study, required the development of a systematic approach to reduce the number of variables used for analyses. Conceptual models which linked selected questions to concentration, exposure, and/or biomarker measurements were defined to allow testing of hypothesized relationships. Statistical methods were selected to evaluate the utility of specific questions to explain the variability in concentration, exposure, and/or biomarker measurements; to classify participants by their likely exposures; and to identify characteristics of highly exposed individuals. These included non-traditional statistical methods such as "Categorical Regression," for use with large numbers of questionnaire-type data (with "yes/no" or "pick one of the above" type answers), and "data mining" techniques used to select questions that help to group or classify participant's exposures. In the Maryland study, a smaller number of participants was selected as measurements were repeated up to six times during a year. The resulting data set was analyzed separately from the Arizona and Region 5 results due to the smaller number of subjects. The Maryland data were subjected to statistical methods which account for the relationship between measurements made on the same individual over time. Although the methods were different and only one compound was chosen from each chemical class, the analysis approach was similar: variable reduction followed by empirical model building in accordance with a conceptual source-to-dose framework. Here the emphasis was on identifying time-varying determinants of exposure to better understand exposure consistency over time and produce more accurate exposure estimates. Results and ImpactsThe analysis approach developed to identify the major predictors of exposure provided a systematic review of all the questions used in the NHEXAS studies. Most other analyses have focused on selected questions of interest. Questionnaire variables about the same topic area (e.g., smoking or attached garages), including those on different questionnaires, tended to group together. This indicated an internal consistency among responses to these questions, and indicated that only a few questions about each topic area may be needed in future exposure studies to obtain most of the information needed. The use of "Categorical regression" analysis allowed the original questionnaire variables to be analyzed to find relationships with the measurement results. When categorical data are used in traditional regression analysis, additional variables are created from the various questionnaire responses to represent the response categories. The inflated number of variables makes it difficult to conduct the analyses, given the small number of participants in each study. The questionnaire variables that were identified as major predictors in the analysis models seemed reasonable, i.e., they were consistent with results from other studies and our current understanding of human exposure. There were some instances where indirect relationships are suggested, or where more in-depth reviews of the relationship of question with others that were not selected for the model-based analyses are needed. The combination of innovative and traditional analysis techniques for these type of data allowed for the analysis of all questions used in the NHEXAS studies, and for analyzing cases without responses, which are often eliminated. The approach examines different ways these questions could be used to predict/explain concentrations, to classify participant exposures, and to identify characteristics of highly exposed individuals. Results from analyses for three target chemicals (lead, phenanthrene, and chlopyrifos) measured in the Maryland study included nearly four dozen different questionnaire items. Significant predictors included housing/demographic factors as well as intermittent activities. There was little overlap among the predictors for each chemical, suggesting that intervention strategies should be chemical class-specific. Including the explanatory variables in the models achieved a typical reduction in error (in predicting an individual observation) of 20 to 40%. Measurements in biological fluids and dust were generally more consistent over time than those in indoor air. Information from questionnaire items helped to reduce the number of measurements needed to achieve reliable year-long exposure estimates. However, for measurements where either the questionnaire items explained little within-person variance (e.g., dermal lead) or the proportion of within-person to total variance was low (e.g., chlorpyrifos in dust, and lead in blood) the desired number of measurements was not reduced. Overall, this analysis suggests that the degree to which questionnaire information can increase the precision in exposure estimates depends upon the compound of interest and the media in which it is measured. In summary, the NHEXAS studies have provided real-world data on individual exposures and activities to evaluate many multimedia and media-specific risk assessment and risk management issues, and to improve exposure factors and models. These data can also be used to identify major predictors of exposure in order to help identify how high exposures can be reduced, how activities contribute to exposures, and how available information on outdoor concentrations can be related to personal exposures. In addition, the systematic analysis of the questions used in NHEXAS relative to environmental concentration and exposure measurements, can help to minimize participant burden and costs for future exposure and health effects studies, including the planned National Children's Study. Research Collabortation and Research ProductsThis NHEXAS analysis project was coordinated by EPA's National Exposure Research Laboratory (NERL) in Las Vegas, with support from Anteon Corporation (Las Vegas, NV). Major contributions to the development of the analysis approach were made by Constella Health Sciences (Durham, NC) and Wilkes Technologies, Inc. (Bethesda, MD). Several other research organizations were consulted in the development, review, and implementation of the analysis approach. The data from the three NHEXAS studies, including questionnaire and diary responses, results of chemical measurements on the subject's blood, urine, and residential environment, and results from the measurement of quality control samples have been made publicly available through EPA's Human Exposure Database System at http://www.epa.gov/heds/. Analyses of the Maryland study results were conducted by NERL (Las Vegas), with support from the University of Nevada, Las Vegas, and Emory University, and are described in the following manuscript: Egeghy PP, Quackenboss JJ, Catlin S, Ryan PB. Determinants of Temporal Variability in NHEXAS-Maryland Environmental Concentrations, Exposures, and Biomarkers. Submitted to Journal of Exposure Analysis and Environmental Epidemiology, 2003. Future ResearchContinued analyses of the NHEXAS questionnaire and measurement results will be conducted to examine how models of exposure, based on different types of information, could be used to estimate/classify exposures in health effects studies, such as the National Children's Study.
Currently, other analyses of the NHEXAS data are being conducted based on the projects included in the Strategic Plan for Analysis of the NHEXAS Pilot Study (EPA 600/R-00/049), available at: Contacts for Additional Information
Questions and inquiries can be directed to the principal investigator: 2003 Research Abstracts - Table of Contents
|
|
|
|
|
||
|
|
