Endocrine Abstracts

Objectives: Fine needle aspiration (FNA) cytology of highly prevalent thyroid nodules results in an indeterminate cytopathologic diagnosis in approximately 20% of cases. To better triage patients for diagnostic vs definitive thyroid surgery or to avoid diagnostic surgery, low-cost gene panels can be used as a rule-in molecular test. However, there remains a need to pre-surgically diagnose indeterminate thyroid nodules where somatic mutations are not identified. The extent of this need is dependent on the comprehensiveness of the respective gene panel. In this study, direct digital counting of miRNA expression was used to identify miRNAs that are significantly differentially expressed between benign and malignant thyroid tumors. Discriminator and stably expressed miRNAs were selected to create a real time PCR (TaqMan) panel. Indeterminate FNAs were analyzed on the TaqMan panel, and classifiers were trained and validated to risk stratify for malignancy.

Methods: A retrospective cohort study was conducted. A discovery cohort of thyroid nodule FNAs and FFPEs with no common somatic mutations was analyzed for 798 miRNAs on the Human v3 miRNA panel (nCounter). A panel of differentially expressed and stably expressed miRNAs was selected from the discovery cohort and the literature. Air-dried smear and liquid cytology FNAs with no common somatic mutations were analyzed on the custom TaqMan panel, and miRNA expression levels were used to train and validate malignancy risk classifiers in both sample types.

Results: Based on the discovery panel and the literature, a 16-miRNA panel was developed and validated in 127 air-dried smear indeterminate thyroid FNA cytology specimens and 157 liquid thyroid FNA cytology specimens. A gradient boosting classifier achieved sensitivity of 100% and specificity of 83% in the air-dried smear sample type. However, miRNA expression was completely overlapping between benign and malignant tumors in the liquid FNA cytology specimens.

Conclusions: This study presents a novel, highly accurate miRNA expression classifier that could be used to incrementally risk stratify indeterminate air-dried smear FNA cytologies where common somatic mutations have been ruled out. This study also presents novel data indicating methanol-based preservation of thyroid liquid FNA cytology may hinder use of miRNA expression levels for molecular diagnostics, this finding requires further investigation.