Development and validation of mRNA expression-based classifiers to predict low-risk thyroid tumors.

Area of Science:

• Oncology
• Genomics
• Molecular Diagnostics

Background:

• Thyroid nodule molecular variants offer population-level prognostic data but show variable clinical behavior.
• Exome-enriched gene expression analysis may surpass limitations of point mutation or fusion-based models.
• Accurate preoperative prediction of thyroid cancer invasion and metastasis is crucial for treatment planning.

Purpose of the Study:

• To develop and validate mRNA-based classifiers for preoperative risk assessment of thyroid tumor invasion.
• To create classifiers capable of ruling out lymph node metastasis in thyroid cancer patients.
• To improve personalized surgical approaches and reduce overtreatment in thyroid cancer management.

Main Methods:

• Retrospective cohort study using histopathology reports from thyroid cancer patients.
• Development of mRNA expression-based classifiers using literature-derived signatures and differentially expressed genes.
• Machine learning algorithms employed for classifier development and validation on independent cohorts.

Main Results:

• Classifiers demonstrated high negative predictive values (NPVs) for ruling out invasion (97.6% development, 99% validation) and lymph node metastasis (98.6% development, 100% validation).
• Significant percentages of patients were classified as low-risk for invasion and metastasis, enabling potential surgical de-escalation.
• Validation cohort confirmed the high accuracy of the mRNA classifiers in a real-world clinical setting.

Conclusions:

• Gene expression-based classifiers can confidently rule out thyroid tumor invasion and lymph node metastasis preoperatively.
• These classifiers support personalized surgical strategies, potentially decreasing overtreatment and associated complications.
• Reduced surgical intervention and improved patient outcomes, including fewer instances of postoperative hypothyroidism, are anticipated.