Performance of a Multigene Genomic Classifier in Thyroid Nodules with Suspicious for Malignancy Cytology

Molecular testing is increasingly used to refine the probability of cancer and assess recurrence risk in thyroid nodules with Bethesda III/IV fine needle aspiration (FNA) cytology. However, limited data exist for Bethesda V (suspicious for malignancy [SFM]) samples. This study evaluated the performance of ThyroSeq v3 (TSv3) in thyroid nodules with SFM cytology. In this single-institution retrospective cohort study, consecutive thyroid FNA samples diagnosed as SFM with TSv3 testing and known surgical outcome were identified. Clinical, pathology, and molecular findings were reviewed. The TSv3 Cancer Risk Classifier was used to determine molecular risk groups (MRGs). For test-negative cases diagnosed as cancer/noninvasive follicular thyroid neoplasm with papillary-like nuclear features, TSv3 was performed on the resected tumors. Among 128 SFM samples studied, 100 (78.1%) were TSv3 positive, and 28 (21.9%) were negative. The cancer prevalence on surgery was 82.8%. Among test-positive samples, 95% were malignant and 5% benign. Among test-negative samples, 17 (60.7%) were benign and 11 (39.3%) malignant. Overall, TSv3 had a sensitivity of 89.6% (confidence interval; CI 82.4-94.1) and a specificity of 77.3% (CI 56.6-89.9). For a cancer prevalence of 50-75% expected in SFM cytology by the Bethesda system, the negative predictive value was expected to range from 71.2% to 88.1% and the positive predictive value from 79.8% to 92.2%. Among test-positive nodules, 20% were MRG-Low (mostly RAS-like alterations), 66% MRG-Intermediate (mostly BRAF-like alterations), and 14% MRG-High. Among patients with cancer, 65 (61.3%) were American Thyroid Association low risk, 25 (23.6%) intermediate risk, and 6 (5.7%) high risk. During the mean follow-up of 51.2 months (range: <1 to 470 months), 12 (13.0%) patients had disease recurrence, which was more common in MRG-High (54.6%) compared with MRG-Intermediate (9.5%) and MRG-Low (0%) cancers (p < 0.001). Upon reexamining tumors with false-negative results, half of evaluable cases had alterations likely missed due to limiting FNA sampling, and the remainder represented low-risk tumors. Potentially targetable alterations were identified in 10 samples. In this large series of SFM thyroid nodules, TSv3 further improved cancer prediction and detected RAS-like, BRAF-like, high-risk, and potentially targetable alterations, all of which may inform more optimal patient management. MRGs were associated with recurrence-free survival, offering potential preoperative cancer risk stratification.