The combination of four molecular markers improves thyroid cancer cytologic diagnosis and patient management

摘要: Papillary thyroid cancer is the most common endocrine malignancy. The sensitive and specific diagnostic tool for nodule diagnosis fine-needle aspiration (FNA) biopsy with cytological evaluation. Nevertheless, FNA not always decisive leading to “indeterminate” or “suspicious” diagnoses in 10 %–30 % of cases. BRAF V600E detection currently used as molecular test improve nodules, yet it lacks sensitivity. aim present study was identify novel markers/computational models discrimination between benign malignant lesions. We collected 118 pre-operative samples. All samples were characterized presence mutation (exon15) by pyrosequencing further assessed mRNA expression four genes (KIT, TC1, miR-222, miR-146b) quantitative polymerase chain reaction. Computational (Bayesian Neural Network Classifier, discriminant analysis) built, their ability discriminate tumors tested. Receiver operating characteristic (ROC) analysis performed principal component visualization purposes. In total, 36/70 carried mutation, while all 48 wild type exon15. Bayesian neural network (BNN) analysis, including showed a very strong predictive value (94.12 % 92.16 %, respectively) discriminating from patients. correct classification 100 % group, 95 % BNN. KIT miR-146b highest accuracy ROC curve, area under curve values 0.973 0.931 miR-146b. model proposed this proved be highly discriminative status compared assessment alone. Its implementation clinical practice can help identifying malignant/benign nodules that would otherwise remain suspicious.