Molecular classification of thyroid nodules using high-dimensionality genomic data.

Area of Science:

• Molecular diagnostics
• Oncology
• Genomics

Background:

• Thyroid nodules are common, but distinguishing benign from malignant ones can be challenging.
• Fine-needle aspiration (FNA) is a primary diagnostic tool, but cytopathology can be indeterminate.
• Accurate molecular tests are needed to improve thyroid nodule diagnosis and patient management.

Purpose of the Study:

• To develop and validate a molecular test for classifying thyroid nodules as benign or malignant using FNA samples.
• To overcome limitations of previous molecular tests that performed poorly on FNA specimens.

Main Methods:

• mRNA expression analysis of over 247,186 transcripts in 315 thyroid nodules (surgical tissues and FNA samples).
• Training and independent validation of two classifiers: one on surgical tissues and another on FNA samples.
• Evaluation on 48 prospective FNA samples, including those with indeterminate cytopathology and RNA degradation.

Main Results:

• A classifier trained on FNA samples demonstrated improved performance compared to a tissue-trained classifier.
• The FNA-trained classifier achieved 96% negative predictive value and 84% specificity on an independent test set.
• The classifier showed high accuracy even with 80% dilution and substantial RNA degradation, indicating robustness.

Conclusions:

• The FNA-trained molecular test is effective in classifying thyroid nodules, even with compromised sample quality.
• Its high tolerance to dilution and degradation makes it suitable for routine clinical use, addressing sampling error concerns.
• Further validation in a large, multicenter clinical trial is ongoing to confirm performance on prospective thyroid FNA samples.