Quality assurance for thyroid FNAs: A more useful way to compare performance

We read with interest the recent publication by Elsheikh et al.1 concerning quality assurance in thyroid fine-needle aspirates. As the authors point out, there are many different rates that can be measured, including the rates of atypical cells of undetermined significance (AUS), AUS/malignancy rates, molecular testing results, abnormal molecular results, and malignancy rates. The authors demonstrate correlations between some of these results and propose various reasons why individual cytologists may have different AUS rates. Although the data allow comparison with their fellow cytologists, they do not clearly show what the optimal AUS rate might be and the impact the AUS rate has on their patients. In the past, an elevated AUS rate led to a relatively large percentage of patients undergoing unnecessary surgery, and recommendations for a very low AUS rate that were in previous versions of The Bethesda System for Reporting Thyroid Cytopathology made sense. With the advent of molecular testing, a much smaller percentage of patients with AUS go to surgery, raising the possibility that it might be beneficial to have a higher AUS rate if this was associated with detecting more malignancies. In this article, it is possible to compare both the relative yield of malignancies in groups of cytologists with different AUS rates and the relative number of patients who underwent surgery for benign disease. To do this, we grouped together the cytologists who had the highest and lowest AUS rates to improve the statistical power, estimated the total number of malignancies detected assuming that all suspicious for malignancy and malignant cases were in fact malignant on resection, and used the AUS data only for patients who had molecular testing. This indicates that the cytologists with lower AUS rates (mean, 12.9%; range, 7.2%–17.2%) sent almost 2% fewer of all patients to resection for benign disease than the cytologists with higher AUS rates (mean, 30.4%; range, 18.8%–40.6%; 1.0% vs. 2.9%, respectively; p < .001), but they also detected 2.2% fewer malignancies (6.6% vs. 8.8%, respectively; p < .001). For pathologists with an AUS rate <17%, this results in a relative loss of sensitivity for malignancy of 23% in a test with a sensitivity well above 95% in most studies. There are limits to this analysis. No Bethesda category IV cases were included, follow-up for patients with AUS diagnoses without molecular testing were not included, we had to estimate the malignancy rate for suspicious and positive cases, and the risk of malignancy for abnormal molecular testing is relatively high. We tried using different assumptions for the risk of malignancy for suspicious/positive aspirates (50%/75%) and different cutoff points for AUS rates without any significant change in the result. Nevertheless, there also may be limitations of which we are unaware in how these data are presented, and a more complete analysis of the data in the current study may be of value. However, we believe the data we present here can be useful for cytologists who wish to improve their performance of thyroid aspiration. They can decide whether it is worth missing 23% of malignancies in their patients to achieve an AUS rate <17% and reducing the number of patients who undergo unnecessary surgery by 2%. The author declares no conflicts of interest.