A six-gene model for differentiating benign from malignant thyroid tumors on the basis of gene expression.

Abstract

Thyroid nodules are common; fine-needle aspirations commonly are read as indeterminate, necessitating surgery to exclude carcinoma. We developed a 6-gene array-based predictor model to diagnose benign versus malignant thyroid lesions. In this study, we verified whether quantitative reverse transcription-polymerase chain reaction (qRT-PCR) using this model reliably can differentiate benign from malignant thyroid nodules. Molecular profiles of benign (follicular adenomas, hyperplastic nodules) and malignant tumors (papillary thyroid carcinomas, follicular variants of papillary thyroid carcinomas) were analyzed using qRT-PCR from our 6-gene model (kit, Hs.296031, Hs.24183, LSM7, SYNGR2, C21orf4). The gold standard was standard pathologic criteria. A diagnosis-predictor model was built by using the training samples and was then used to predict the class of 10 additional samples analyzed as unknowns. Our predictor model using 47 training samples correctly predicted 9/10 unknowns. One sample diagnosed as benign by standard histologic criteria was diagnosed as malignant by our model (sensitivity 75%; specificity, 100%; positive predictive value, 100%; negative predictive value, 85.7%). Molecular diagnosis with our 6-gene model can differentiate between benign and malignant thyroid tumors with high sensitivity and specificity. In combination, these genetic markers may be a reliable test to preoperatively diagnose the malignant potential of thyroid nodules.