Misclassification bias arising from random error in exposure measurement: implications for dual measurement strategies.

Despite the resulting loss of information, continuous exposure variables are often categorized in epidemiologic data analysis. It has been shown that nondifferential measurement error of exposure can lead to differential misclassification under these circumstances. This paper extends these findings to the misclassification bias that results from nondifferential measurement error if dual measurement strategies are employed with dichotomized exposures. Different scenarios of random measurement error are presented in the context of case-control studies. Although differences in the misclassification rates between cases and controls are generally small, the bias toward the null is usually much weaker than expected under truly nondifferential exposure misclassification even with single exposure measurements. In most cases, the bias is further reduced by the use of dual measurements. Caution is needed, however, if the analysis is restricted to individuals with concordant exposure classifications (both measurements above or both measurements below the cutpoint). In this case, the true exposure-disease association is often overestimated.