The real-world performance of ThyroSeqV.2 to diagnose thyroid "neoplasm requiring surgery"

Fine-needle biopsy (FNB) predicts benign or malignant thyroid nodules. For indeterminate (ITN) FNBs, commercial molecular tests may improve the diagnostic accuracy and reduce the number of operations. These tests have had limited independent implementation studies in routine clinical practice. This is a prospective observational study. At Boston Medical Center, the 1,316 consecutive FNBs were classified to one of the six categories in the Bethesda classification system. Those ITN samples were submitted for ThyroSeqV.2 next generation sequencing panel analysis. The performance of ThyroSeqV.2 to predict "neoplasm requiring surgery" (NRS) was evaluated. ThyroSeqV.2 assay was performed in 398 FNBs on 384 cytologically ITN nodules (308 Bethesda III, 47 Bethesda IV and 29 Bethesda V). The first evaluable ThyroSeq result for each nodule was used for final analysis. Seventy-seven (72.0%) of 107 patients with a high risk molecular test underwent thyroid surgery resulting in 41 NRS (53.2%) and 36 benign nodules (46.8%). Of the 249 patients with a low risk or negative molecular analysis, 51 (20.5%) had surgery revealing 47 benign nodules (92.2%) and 4 NRS (7.8%). Based on surgical outcome of 128 ITN with evaluable ThyroSeq results, this molecular test had a sensitivity of 91% (95% CI: 79%-98%), specificity of 56% (45%-67%), positive predictive value (PPV) of 53% (42%-65%), negative predictive value (NPV) of 92% (81%-98%), and an overall accuracy of 69% (55%-85%) with a prevalence of NRS of 35% (27%-44%). ThyroSeqV.2 in this clinical use study in ITN nodules provided a similar NPV but a lower PPV than expected compared to published studies due to the detection of an array of mutations in benign nodules. The NPV of 92.0% for ITN cytology confirmed its utility as a "rule-out" test to exclude NRS.