Malignancy risk and reproducibility associated with atypia of undetermined significance on thyroid cytology

Background The Bethesda System for Reporting Thyroid Cytopathology (TBSRTC) describes several subcategories within atypia of undetermined significance (AUS), including (1) presence of focal nuclear atypia (AUS-N), (2) focal microfollicular proliferation (AUS-F), (3) focal Hürthle cell proliferation (AUS-HC), and (4) other (AUS-O). Several publications suggest that 5–15% is an underestimate of the malignancy risk for AUS, and that the underestimation is owing to the similarity between AUS-N and suspicious for malignancy (SFM). Thus, we investigated the AUS subcategories during morphologic re-review at a tertiary care center and their associated malignancy risk. Methods Of 4,827 fine-needle aspiration specimens were sent between January 2009 and August 2013 for morphologic re-review, 806 were categorized as AUS. Comparison of AUS subcategory diagnoses were made between outside and re-review results. The malignancy risk was also determined for 255 nodules with available surgical pathology. Result The outside diagnoses of the 806 cases read as AUS on second review were as follows: 5 insufficient (0.1%), 149 benign (19%), 463 AUS (57%), 124 SFN or suspicious for follicular or Hürthle cell neoplasm (15%), 56 SFM (7%), and 9 malignant (1%). Of the 463 cases in which both the outside and re-review diagnosis was AUS, the distribution of the subcategories in order of increasing frequency was 53 AUS-HC (11%), 74 AUS-O (16%), 79 AUS-F (17%), and 257 AUS-N (56%). Of the 255 resected nodules, 99 (39%) were malignant. Subcategory malignancy rates were: AUS-HC, 19% (9/47); AUS-O, 26% (14/54); AUS-F, 39% (19/49); and AUS-N, 54% (57/105). Cases in which both the referring institution and re-review agreed about the AUS-N subcategory had an even greater risk of malignancy (68%; 17/25). Conclusion Disagreement about the diagnosis of AUS between institutions is frequent. The malignancy risk for AUS is higher than originally proposed by TBSRTC and attributable to the high risk of AUS-N. Furthermore, agreement on AUS-N after re-review portends a malignancy risk that borders on that of SFM. This suggests that AUS-N may have discrete features that can provide specific morphologic predictors and enable the consolidation of AUS-N into SFM.