The Genomic Landscape of Medullary Thyroid Carcinoma Identified by the Afirma RNA-Sequencing Classifier: Insights from a Large Real-World Cohort

Introduction

The Afirma RNA-sequencing Medullary Thyroid Carcinoma (MTC) classifier, a component of the Afirma Genomic Sequencing Classifier (GSC), was previously validated as a reliable tool for preoperative identification of MTC from fine-needle aspiration (FNA) specimens.^1,2

• We aimed to characterize the genomic landscape of MTC identified using this classifier in a large real-world cohort of FNA samples.

Methods

We retrospectively analyzed MTC positive samples by the Afirma MTC classifier in the Veracyte CLIA laboratory between January 2018 and June 2024.

• Genomic variants identified by the Afirma Xpression Atlas (XA) were characterized.³

Results

• 73% of MTC-positive thyroid nodules had at least one pathogenic variant identified by XA, most commonly in RET (53%) and RAS (19%) with no significant difference across Bethesda categories (Table 2).
• Most RET alterations were in codons 918 (19%) and 634 (10%), while HRAS was the most frequently altered RAS isoform (15%).
• Mutually exclusive oncogenic fusions were identified in 8 non-RET/RAS SNV mutated samples [EML4::ALK (n=1), MKRN1::BRAF (n=6), and SPECC1L::RET (n=1)].
• Additional isolated altered variants identified in non-RET/RAS samples included AKT1, DICER1, PIK3CA, and TP53.

• Among 252,510 FNA samples tested, 732 (0.3%) were classified as MTC.
• Median patient age was 63.2 years (IQR, 52.1-72.1) and median nodule size was 2.1 cm (IQR, 1.6-3.0).
• On cytology, 71% (520/732) of nodules were Bethesda III or IV, 18% were Bethesda V and 11% were Bethesda VI (Table 1).

Conclusion

• This large study reinforces the known genomic landscape of MTC.
• Additionally, our findings highlight that a subset of thyroid nodules sent for Afirma GSC testing are unrecognized MTCs, underscoring the value of molecular testing in improving preoperative diagnostic accuracy.
• Gene expression analysis showed that ERK activity was slightly lower in RET-driven than in RAS-driven MTC (p=0.02), but significantly higher in both groups compared with non-RET/ non-RAS samples (p<10^-10).
• Further study is needed to molecularly define the nearly 30% of specimens that lacked detectable alterations.