Value of sonographic features in predicting malignancy in thyroid nodules diagnosed as follicular neoplasm on cytology

Summary Background The cytological diagnosis of follicular neoplasm (Thy3F) remains a diagnostic challenge. The main aim of this study was to stratify the risk of malignancy in thyroid nodules diagnosed as Thy3F on cytology (Thy3F) using thyroid imaging reporting and data system (TIRADS). Methods A database of thyroid nodules with Thy3F cytological results from ultrasound‐guided FNA (US‐FNA) between January 2007 and March 2014 was studied retrospectively. Information on patient demographics, ultrasound characteristics and final histology of the nodules was collated. The number of suspicious US features of each thyroid nodule was counted based on TIRADS. The malignancy rate of each of the TIRADS category was also calculated based on the final histological outcomes of the nodules and compared to that calculated using a recently proposed thyroid malignancy risk prediction model. Results The overall malignancy rate of Thy3F cytology was 24·3%. There were significantly higher percentages of malignant nodules with irregular margins (20·0% vs 0%, P = 0·000), hypo‐echogenicity (74·3% vs 51·4%, P = 0·013) and taller‐than‐wide morphology (17·1% vs 0·9%, P = 0·001) when compared to benign nodules. The risk of malignancy increased with advancing TIRADS score: TIRADS 4A (14·3%), TIRADS 4B (23·1%), TIRADS 4C (87·5%) and TIRADS 5 (100%). The malignancy rate calculated using the prediction model similarly increased with advancing TIRADS score: TIRADS 4A (6·2%), TIRADS 4B (32·5%), TIRADS 4C (79·9%) and TIRADS 5 (90%). Conclusion Thyroid nodules with TIRADS scores 4C and 5 should be considered for single definitive surgery in view of the high malignant rate.