TOPLINE

RET fusion positivity was identified as an independent risk factor in occult lymph node metastasis in clinically lymph node negative (cN0) papillary thyroid cancer (PTC). In addition, BRAF mutation, a known risk factor for PTC, was also independently associated with occult lymph node metastasis. These molecular risk factors, along with other clinical and ultrasonographic factors, showed predictive accuracy in a new machine learning model for the prediction of occult lymph node metastasis in cN0 PTC. The authors caution, however, that further validation of the model in larger studies is needed.

METHODOLOGY

• Occult lymph node metastasis in PTC is potentially linked to poorer outcomes, and preoperative detection rates are low; therefore, the identification of risk factors is important. Researchers developed a machine learning model to predict risk for occult lymph node metastasis in cN0 PTC patients.
• A cohort of 961 cN0 PTC patients was randomly divided into two groups: a training group and a test group; patients with tumors sized ≤ 1 cm were further stratified for internal validation.
• Using multivariate analysis, the authors identified independent risk factors, including molecular biomarkers, for occult lymph node metastasis.
• Eight machine learning models were then trained using clinical, ultrasonographic, and molecular features.

TAKEAWAY

• Multivariate analysis showed that patients who were RET fusion-positive had a significantly higher rate of occult lymph node metastasis, with an odds ratio (OR) of 3.34 compared with RET fusion-negative patients (P = .010).
• BRAF mutation positivity, known to occur in 25%-82.3% of PTC cases but with a debated role in PTC metastasis, was also significantly associated (OR, 1.71; P = .003). 
• Other risk factors significantly associated with occult lymph node metastasis and used in the model included: younger age (OR, 0.97; P < .001), male sex (OR, 1.40; P = .036), larger tumor size (OR, 1.07; P < .001), higher Thyroid Imaging Reporting and Data System score (OR, 1.14; P = .001), presence of clustered punctate echogenic foci (OR, 1.70; P = .002), and tumor wider-than-tall (OR, 1.73; P = .010).
• Of the eight models tested, the random forest model showed optimal performance with an area under the curve (AUC) of 0.906 in the training set and 0.733 in the test set. Further analysis of tumors sized ≤ 1 cm internally validated the model’s robustness with an AUC of 0.719.

IN PRACTICE

“RET fusion testing may be considered in select cases but cannot be endorsed for routine preoperative screening at this time,” the authors noted. However, the identification of BRAF mutation positivity as an independent risk factor for occult lymph node metastasis “has substantial implications for understanding the biological behavior of PTC and guiding clinical decision-making,” they added. 

SOURCE

The study was published in The Journal of Clinical Endocrinology and Metabolism.

A preliminary version of the model is available online; however, the authors caution that the model is still for research purposes only, pending external validation.

LIMITATIONS

The study was retrospective; therefore, some biases may exist. In addition, the authors noted that the random forest model indicated a performance gap between the training and test sets and demonstrated suboptimal recall on the test set, leading to a risk of missing true occult lymph node metastasis cases. The prevalence of RET fusion in the study was just 2.6%, rendering the evidence preliminary, despite the observed association.

DISCLOSURES

The authors had no disclosures to report.