ECE2016 Guided Posters Thyroid Cancer (1) (10 abstracts)
Thyroid nodules are very common and benign in most cases. Thus, malignancy detection avoiding overtreatment is a challenge. Nodule evaluation mainly supports on US and fine needle aspiration cytology (FNAC). The Bethesda classification (BC) for reporting thyroid cytopathology is now currently used for the interpretation of results but it does not enable classifying 30% of samples.
The objective was to identify, by transcriptome analysis, a molecular signature to improve the accuracy of preoperative diagnosis of nodules. We built a combined Bethesda-Molecular predictor that takes into account the prevalence of the disease.
Methods: In this prospective study, 722 patients with a solid thyroid nodule more than one centimetre diameter underwent FNAC. The molecular test, a transcriptomic array of 20 genes selected from a previous study, was performed on FNA material in operated patients. The optimal set of genes was identified using a logistic regression model to discriminate malignant and non-malignant nodules and was constituted of 7 genes. The performances of a combined predictor (molecular test in addition to BC) were compared to that of BC alone using the area under the ROC curve (AUC) for different levels of malignancy prevalence.
Results: Among the 225 operated nodules, 128 underwent the molecular test. In these patients, with a prevalence of malignancy of 36%, the combined predictor presented a 95% specificity and a 76% sensitivity. The AUC was 93.5%; significantly higher than the AUC of BC alone (P=0.004). In a general population of unselected nodules with an estimated prevalence of malignancy of 7%, the specificity of the test would be optimal (100%), but sensitivity would be lower (47.8%).
Conclusion: This very specific molecular test improves the detection of thyroid cancer in addition to standard cytological analysis and may be particularly useful in case of indeterminate result at cytological analysis.