Comparing correlated areas under the ROC curves of two diagnostic tests in the presence of verification bias.

To assess relative accuracies of two diagnostic tests, we often compare the areas under the receiver operating characteristic (ROC) curves of these two tests in a paired design. Standard methods for analyzing data from a paired design require that every patient tested has the known disease status. In practice, however, some of the patients with test results may not have verified disease status. Any analysis using only verified cases may result in verification bias. In this paper, we propose a verification bias correction procedure for comparing areas under ROC curves under the missing-at-random (MAR) assumption. We also develop an approach for testing the validity of the MAR assumption. Finally, we use the proposed method to compare the relative accuracies of MRI and CT in evaluation of pancreatic cancer.