Testing the Assumption of Non-differential Misclassification in Case-Control Studies

Part of this paper is based on the result obtained in the Master degree thesis of the second author (QH) (Hui, 2011), which was written under the supervision of the first author (TL). The work in Figure 1 and Tables 3b, 4b and 5b is attributed to QH; the remaining work is attributed to TL. QH is grateful to TL who provided him with an invaluable guidance and constant encouragement while he was writing his thesis. One of the not yet solved issues regarding the misclassification in case-control studies is whether the misclassification rates are the same for both cases and controls. Currently, a common practice is to assume that the rates are the same, that is, the non-differential misclassification assumption. However, it has been suspected that this assumption may not be valid in practical applications. Unfortunately, no test is available so far to test the validity of the non-differential misclassification assumption. A method is presented to test the validity of non-differential misclassification assumption in case-control studies with 2 × 2 tables when validation data are not available. First, a theory of exposure operating characteristic curve is developed. Next, two non-parametric methods are presented to test the assumption of non-differential misclassification. Three real-data sets taken from practical applications are used as examples to illustrate the methods.