Evaluation of exposure-response relationships is a crucial component in assessment of causes of cancer. Quantitative exposure assessment is therefore an essential element in high-quality epidemiologic investigations. Occupational and Environmental Epidemiology Branch (OEEB) investigators devote considerable effort to improving exposure assessment techniques and evaluating the reliability and validity of procedures used in cohort and case-control studies of occupational and environmental exposures. Below are a few examples.
Epidemiologic investigations of environment-cancer relationships rely on accurate, quantifiable exposure measurements. Such studies may use Geographic Information Systems (GIS) and regulatory and other environmental monitoring data to complement retrospective surveys, which are sometimes limited by participants’ knowledge of exposures in their surrounding environments. Cancer studies have benefited from the increasing availability of historical air and water monitoring data, satellite imagery, census, and other geographic datasets that allow for reconstruction of residence- and other location-based exposures over a substantial portion of a person’s lifetime.
OEEB implements GIS-based exposure assessments using georeferenced historical data resources and residential histories collected in our studies of environmental hazards and cancer risk. Our approaches include using GIS and spatial-analytic methods to characterize exposure to environmental risk factors, incorporating space-time-activity information in exposure assessments, and employing biological and environmental measurements for exposure validation. This work addresses important epidemiologic considerations in geography-based environmental exposure assessments, including residential mobility, positional error, and the challenges of extrapolation over space and time.
OEEB investigators have developed state-of-the-art quantitative exposure assessment methods that maximized the available measurements and exposure determinant information to predict historical exposure levels in several studies including diesel exhaust in the Diesel Exhaust in Miners Study, benzene in a cohort study Benzene-Exposed Workers in China, and specific pesticides within the Agricultural Health Study.
Job Exposure Matrices (JEMs) are an efficient way to assign exposure estimates in population-based studies and may be the only exposure assessment option when only job and industry are available for study participants. DCEG investigators develop methods to improve JEMs by incorporating databases of exposure measurements to calibrate JEMs across time and across jobs and industries. For example, in the Shanghai Women's Health Study, investigators developed a novel framework to systematically combine a JEM with historical measurements to calibrate JEM exposure ratings used to estimate historical benzene and lead exposure.
Occupation- and industry-specific modules ask detailed questions about work activities and exposures within its population-based case-control studies to better capture within-occupation differences in exposure. Usually these module responses are reviewed job-by-job by an exposure assessor to assign exposure estimates. OEEB investigators develop methods to more efficiently and transparently assign exposure in studies that use modules and examine the validity and reliability of these methods. For example, programmable decision rules based on questionnaire response patterns were developed to estimate occupational diesel exhaust exposure for the New England Bladder Cancer Study that had moderately-high agreement with estimates obtained from expert reviews of each job. We have also developed a method to extract patterns in the questionnaire responses that predict an expert’s exposure assignments using classification and regression tree (CART) models. The extracted decision rules can be used to improve the transparency and efficiency of applying the exposure decisions to other study subjects.
To assist exposure assessment efforts, OEEB conducts comprehensive literature reviews for specific agents to identify when, where, and how much exposure to that agent was likely to occur. Exposure concentrations identified in these reviews are extracted into exposure databases that can be used in statistical models to characterize exposure determinants, such as time trends. We demonstrate the utility of meta-regression models to account for both the number of measurements and the exposure variability within each summary statistic.
For more information, contact Dr. Melissa Friesen.