Risk Threshold Algorithm for Thyroid Nodule Management Demonstrates Increased Specificity and Diagnostic Accuracy as...

We studied diagnostic performance of an algorithm guiding thyroid nodule management using a malignancy risk model as compared with extant management guidelines. Single-institution, retrospective study was performed with sequential cases from pathology registry from 2012 to 2015. Seventy-eight patients were enrolled, with benign and malignant groups defined by aspiration results. Risk Threshold Algorithm determined management based off of a logistic regression model and a risk threshold. American College of Radiology Thyroid Imaging, Reporting and Data System (ACR TI-RADS), Society of Radiologists in Ultrasound (SRU), and American Thyroid Association (ATA) guidelines were used in comparison. Sensitivity, specificity, positive/negative predictive values, receiver operating characteristic (ROC) values were derived, with significance assessed via McNemar and permutation tests. Forty-four benign nodules and 40 papillary thyroid carcinomas were included. Risk Threshold Algorithm area under the ROC curve was 0.80 versus 0.59 (ACR TI-RADS), 0.49 (SRU), and 0.44 (ATA); all areas under the ROC curve differences were statistically significant. Risk Threshold Algorithm demonstrates sensitivity, specificity, positive predictive value, and negative predictive values of 63%, 91%, 86%, and 73% at the risk threshold maximizing diagnostic performance, compared with 85%, 39%, 56%, and 74% (ACR TI-RADS); 85%, 18%, 50%, and 57% (SRU); and 89%, 11%, 50%, and 83% (ATA). Sensitivity and specificity were significantly different between all groups except SRU versus TI-RADS. The Risk Threshold Algorithm, based on a malignancy risk model, demonstrates increased overall diagnostic accuracy as compared with ACR TI-RADS, SRU, and ATA management guidelines. Through eliminating unnecessary biopsy, patient anxiety, and morbidity can be reduced.