Traditional exposure assessment for epidemiology combines the use of measurements and models to characterize the spatio-temporal field of environmental concentrations of a stressor with individual data on interactions of receptors (people) with their environment (for example, derived from questionnaires on time—activity patterns) to estimate personal exposures. In some studies, direct measurements of personal exposures (for example, film-badge measurements of occupational radiation exposures) and novel methods of tracking individual activities (for example, Global Positioning System (GPS) monitoring of locations and accelerometers for physical activity rates and videotaping of activities) have been used, either on an entire study population or on some sample for calibration or validation of model predictions. The emerging field of molecular epidemiology, based on the use of biomarkers of exposure (as well as of susceptibility or early signs of disease), offers potentially transformative advances in exposure science, particularly if combined with novel genomic, transcriptomic, metabolomic, and other “—omic” technologies and bioinformatic tools for organizing and integrating the massive, often disparate, data sets (see Chapter 5). Box 3-1 illustrates some of the complexities of exposure assessment for the National Children’s Study, with longitudinal measurements of a broad array of environmental and personal (external and internal) exposures and health outcomes.
Historically, comprehensive measurement of environmental exposures has not been possible, requiring statistical models to interpolate among relatively sparse measurements. The models can be purely statistical, such as geostatistical models for air pollution, or can be based on mathematical models for tracking agents from sources through intake by receptors (see NCRP 2010a for a discussion of the general principles of environmental dose reconstruction for radiation exposures and NCRP 2008, 2010b for recommended approaches to uncertainty analysis for external and internal exposures respectively). Box 3-2 provides an example of environmental pathway analysis applied to evaluate radionuclide exposures from the Hanford nuclear plant in Hanford, WA, and illustrates the value of involving the affected communities in all stages of the planning of an epidemiology project.
As novel sensing technologies, such as satellite imaging, become more widely available and more accurate, the need for models will remain, but the focus will shift from interpolation to exploitation of massive datasets. A key function of models is not just to provide point estimates of individual exposures but to quantify the uncertainty in exposure estimates, to understand measurement error in health analyses.
Environmental exposures typically occur over extended periods of time at varying intensities, requiring a shift in thinking from simple exposure-response to exposure-time-response relationships (Thomas 1988). These can be quite complex, involving modifying effects of age-at-exposure (for example, at particularly sensitive developmental stages), time-since-exposure, duration-of-exposure, or other time-related factors. In addition, for most conditions, little is known about whether short intense exposures have larger or smaller effects than