A retrospective analysis of the performance of the RosettaGX Reveal™ thyroid miRNA and the Afirma Gene Expression Classifiers in a cohort of cytologically indeterminate thyroid nodules.

Background: Molecular tests are increasingly used to triage cytologically indeterminate thyroid nodules for surgery and/or follow-up. We retrospectively compared the performance of the Afirma Gene Expression Classifier (AGEC) with that of the more recently developed RosettaGX Reveal™ miRNA Classifier (Reveal) in a cohort of Bethesda III-V thyroid FNAs with surgical follow-up.

Design: Eighty-one samples (54 Bethesda III, 26 Bethesda IV, 1 Bethesda V) with available AGEC (74 AGEC-SUSP and 7 AGEC-BENIGN) and surgical pathology results were studied from three academic centers. Reveal was performed in a blinded fashion.

Results: The final diagnoses were benign/NIFTP (n = 63) and malignant (n = 18). The overall "correct" rate was 64.2% for Reveal and 28.4% for AGEC (P = 1.4e-6). The specificity of Reveal was 60.3%, compared with 9.5% for AGEC (P = 2.1e-9). Among the 18 malignant cases, 77.8% and 94.4% were correctly classified as suspicious by Reveal and AGEC, respectively (P = 0.2). In the FLUS and the FN group, the specificity of AGEC was lower than the specificity of Reveal. Whether the 7 NIFTP in our study were considered benign or malignant, specificity and PPV of Reveal were higher than those of AGEC. Reveal also outperformed AGEC in correctly classifying the 26 benign Hürthle lesions studied (P = 7.6e-5).

Conclusion: Reveal outperformed AGEC in this cohort, whether NIFTP is considered benign or malignant, and in Hürthle lesions. Reveal has the potential to reduce the number of unnecessary resections in patients with indeterminate thyroid cytology. Based on our findings and the practical advantages offered by Reveal methodology, large prospective studies are warranted. Diagn. Cytopathol.