Assessing improvements in the dexamethasone suppression test using receiver operating characteristics analysis.

Abstract

This article demonstrates methods by which receiver operating characteristics (ROC) analysis may be used to evaluate and improve the Dexamethasone Suppression Test (DST). ROC indices provide a description of nonbinary diagnostic tests that is independent of any particular cutoff and that allows investigators to use rigorous mathematical methods for selecting cutoffs and assessing overall performance. In this study, ROC analyses of previously published data suggest that (1) the accuracy of the DST might be improved by using a 0.5-mg oral steroid dose; (2) incorporation of dexamethasone serum levels does not improve the performance of the DST; and (3) the serum cortisol level used to define "no suppression" (or DST "positivity") should be varied to take into account diagnostic alternatives, prior probability of affective disorder, and the clinical goal for performing the test (e.g., maximizing utility, increasing information).