Utility of mutational analysis for risk stratification of indeterminate thyroid nodules in a real-world setting

Objective American Thyroid Association (ATAn) 2015 guidelines recommend repeat fine-needle aspiration with molecular marker profiling (MMP) or diagnostic lobectomy in thyroid nodules yielding atypia of unknown significance/follicular lesion of unknown significance (AUS/FLUS) or follicular neoplasm/suspicious for follicular neoplasm (FN/SFN) cytology. Our objective is to describe the molecular profiles and histological correlates of these cytologically indeterminate nodules (CIN) to aid risk stratification. Design Retrospective chart review. Patients Adults with CIN that underwent MMP from 2017 to 2020. Measurements Pearsons' chi(2), Fisher's exact test, nonparametric testing and multiple regression analysis were performed. Results A total of 89 CIN underwent mutational analysis. Of 55% (n = 49) were Bethesda class III AUS/FLUS and 45% (n = 40) were Bethesda class IV FN/SFN. The US phenotype of a CIN was isoechoic (53%) or hypoechoic (32%) with well-defined margins (98%), absence of calcifications (75%) and mildly increased internal vascularity (70%). A total of 84% and 87% of nodules were classified as mild/moderate or low/intermediate risk per the Thyroid Imaging Reporting and Data System and ATA classifications, respectively. Based on the Thyroseq patient management resource, 6.7% (n = 6) of nodules had a high predicted probability of cancer (>= 95%), 41.6% (n = 37) were intermediate probability (40%-94%) and 51.7% (n = 46) were low probability (<40%). MMP revealed positive mutations in 45% (n = 40) of nodules, with 71% demonstrating RAS mutations. Of the nodules that underwent resection (n = 38), 39% (n = 15) had malignant pathology. Increasing the threshold to recommend surgical resection to a Thyroseq predicted probability of cancer to >= 50%, had a 100% sensitivity and 65% specificity for detecting malignant nodules (area under the ROC curve: 0.86). The positive predictive value was 37% and the negative predictive value was 100%. Conclusion US phenotypes of CIN nodules were variable and did not aid in differentiating malignant from benign nodules. Of the CIN nodules with a positive MMP, most were RAS and had a benign pathology. With the exception of high-risk genetic markers for malignancy, the threshold to recommend surgical resection should be raised for CIN. Further studies to improve risk stratification in these nodules are required.