Exclusion of Eligible Indeterminate Thyroid Nodules in Estimates of Negative Predictive Value for the Gene Expression Classifier

To the Editor Al-Qurayshi et al calculated the negative predictive value (NPV) of the Gene Expression Classifier (GEC) in 154 indeterminate thyroid nodules (ITNs). We reviewed 12 articles that evaluated the diagnostic accuracy of the GEC. In those studies, as with that of Al-Quyrashi et al, estimates for test performance were unreliable, owing to serious methodologic flaws in the authors' approach. The most impactful of these flaws was the exclusion of unexcised GEC-benign ITNs from the analysis, resulting in the calculation of NPV on a nonrepresentative sample GEC-benign ITNs selected for excision. A better method would have been to include GEC-benign ITNs that do not develop ultrasound findings suspicious for malignant neoplasm as true-negative results in the calculation of NPV (Table). We can assume the 22 unexcised GEC-benigns did not have suspicious findings on ultrasound, given that the authors imply that GEC-benign ITNs that "…developed indications for surgery…" underwent excision. Using previously published estimates of the diagnostic accuracy of the TIRADS in ultrasound reporting, we can estimate that about 3 (15%) unexcised 22 GEC-benign ITNs may have also been malignant. Therefore, the observed NPV of the GEC in this study can be estimated to be 44 of 58 (76%).

http://ift.tt/2oNjAjq