Background
The 2025 ATA Thyroid Cancer guidelines subdivided the 2015 intermediate risk of recurrence group into low-intermediate (L-I) and intermediate-high (I-H) groups and incorporated histologic specific categories. Thus, cancers in a 2015 risk category may fall into a different 2025 category. This study sought to evaluate demographic and genomic differences in cancers recategorized from 2015 to 2025.
Methods
A retrospective, single-center study of thyroid cancer cases that underwent Afirma Genomic Sequencing Classifier testing as part of routine clinical care. Each case was assigned 2015 and 2025 ATA risk of recurrence labels. Cases that shifted categories were compared with those that remained stable. Most shifts were from the 2015 ATA low risk category; therefore, it was used as the reference for demographic and molecular differences. Expressed molecular variants and fusions and differences in Afirma Genomic Resource for Intelligent Discovery (GRID) data were evaluated.
Results
Seventy-two cases were classified using ATA 2025 criteria: 26 low, 19 L-I, 16 I-H, and 11 high. Of the 3 cancers that were 2015 ATA high risk, all 3 remained high risk by 2025 classification. However, of 45 ATA 2015 low risk cancers, only 25 remained low risk in 2025, while 15 shifted to L-I, 3 to I-H, and 2 to high risk. ATA 2025 I-H and high samples were combined for analysis. Average tumor size increased across 2025 risk strata: from low (1.7 cm) to L-I (2.05 cm) to I-H/high (3.02 cm). No significant differences between groups were seen in age, sex, or Bethesda category. Additionally, no significant differences were seen in expressed variants and fusions, including BRAFp.V600E, KRASp.Q61R, SPOPp.P94R, NRAS variants, and the ALK::EML4 fusion. Finally, analysis of GRID molecular hallmarks demonstrated a trend for linear decreases in Immunomodulatory signature signaling and Apical surface hallmark while there was a linear increase in Angiogenesis hallmark signaling from 2025 ATA low to ATA I-H/high groups (p =0.1).
Discussion/Conclusion
When reassessed using the 2025 ATA criteria, a substantial proportion of cases originally classified as low risk in 2015 were reassigned to higher risk categories. This shift was not explained by differences in commonly expressed variants or gene fusions, suggesting genomic signals identified by GRID may inform a molecular basis for reclassification. These findings provide biologic support for the 2025 ATA risk classification by showing that their refined clinical categories likely correspond to underlying genomic differences. Our results suggest that the 2025 ATA criteria, although developed from clinicopathologic data, may reflect distinct molecular profiles that were previously unrecognized. Larger, multi-institutional studies are warranted to increase sample size and power to validate these observations and define their prognostic utility.