Abstract
Misclassification of historically reconstructed exposures is a common occurrence in occupational epidemiology studies, especially in situations where exposure ratings form the basis of assessment. It is well recognized that nondifferential misclassification of exposure, which is generally the most reasonably assumed form of misclassification in epidemiologic research, will bias effect estimates toward the null value. Moreover, observed exposure-response gradients will be attenuated when exposure misclassification occurs. This paper reviews some indirect methods for examining the possible extent of bias in effect estimates that can result from nondifferential misclassification of exposure. In addition, misclassification of a confounder is considered, where the confounder may be another occupational exposure that is highly correlated with the main exposure under study. Equations are presented to make adjustments of observed cohort data, cases, and person-years to explore the magnitude of distortion of effect estimates resulting from such bias.