Issues in Performing Retrospective Exposure Assessment

Abstract

Historically, investigations of causality of chronic diseases and occupational exposures have relied upon employment within an industry and/or job as a surrogate for exposure. Similarly, for investigations of dose-response relationships, length of employment within these categories has been used. These surrogates, however, may result in large amounts of misclassification of subjects by exposure categories, which may severely affect risk estimates, particularly if the risks are low. Examples from the literature are provided to demonstrate that these surrogates for exposures may lower estimates of disease risks, obscure etiologic agents, create large confidence intervals (thereby reducing the likelihood of finding a statistical association), and affect dose-response relationships. Recently, more investigators have developed semiquantitative assessments, i.e., assigning jobs to low, medium, and high exposure categories. Although this approach is more satisfactory than the historical approach, it is less than satisfactory because the quantitative relationships among the categories are not known. Incorrect weighing of exposure categories can also result in misclassification of subjects when calculating measures such as cumulative exposure. Quantitative assessment, i.e., assigning a value in units used in industrial hygiene monitoring, is ideally the best approach. However, such an approach may be difficult, if not impossible, because monitoring data are rarely sufficient to allow calculation of measured exposure levels. Thus, assessments often require judgment in assigning exposure level, which can also lead to misclassification. Nevertheless, investigators should use the most quantitative procedure possible so as to develop exposure estimates that are reflective of dose. This approach will enhance the power of epidemiologic studies to detect and evaluate exposure-response associations.