Methodological issues in case-control studies. III--The effect of joint misclassification of risk factors and confounding factors upon estimation and power.

The effect of joint misclassification of polytomous risk factors and polytomous covariates in case-control studies upon estimation and power has been examined using computer simulation and mathematical models. It is concluded that non-differential misclassification of covariates may be as important as non-differential misclassification of exposure variables, and that the resultant bias of the relative risk estimate may be either toward or away from the null. It is pointed out that when a confounding variable is misclassified the calculated p value from the particular study will be incorrect and may either be too large or too small depending on the nature of the confounding.