An exact test for comparing two predictive values in small-size clinical trials.

Abstract

Positive and negative predictive values describe the performance of a diagnostic test. There are several methods to test the equality of predictive values in paired designs. However, these methods were premised on large sample theory, and they may not be suitable for small-size clinical trials because of inflation of the type1 error rate. In this study, we propose an exact test to control the type1 error rate strictly for conducting a small-size clinical trial that investigates the equality of predictive values in paired designs. In addition, we execute simulation studies to evaluate the performance of the proposed exact test and existing methods in small-size clinical trials. The proposed test can calculate the exact P value, and as a result of simulations, the empirical type 1 error rate for the proposed test did not exceed the significance level regardless of the setting, and the empirical power for the proposed test is not much different from the other methods based on large-sample theory. Therefore, it is considered that the proposed exact test is useful when the type1 error rate needs to be controlled strictly.