Comparison of thyrotropin assays by relative operating characteristic analysis

We compared two versions of a commercial double-site immunoradiometric assay for human thyrotropin, one of them claimed to show superior ability in apportioning euthyroid and hyperthyroid populations. Relative operating characteristic (ROC) analysis permits the calculation of the area under the ROC curve as an unbiased unit measure of diagnostic validity, and also yields an optimal decision point for separating the two populations, based on relative prevalences and locally perceived utilities of test outcomes. We examined 48 hyperthyroid and 50 euthyroid subjects with the original assay and with a later version from the same manufacturer and conclude that the latter demonstrates a small but significant superiority in diagnostic performance throughout the clinical range of interest. Furthermore, peak diagnostic efficacy is attained when the boundary between the two populations is set at 0.3 mIU/L. At this decision point, sensitivity is 58%, specificity is 96%, and the predictive value of a positive test is 75%.