Interobserver variation among cytopathologists in the evaluation of oncocytic cell‐predominant thyroid FNA cytology: A validation study of 173 cases with surgical follow‐up and correlation with molecular testing
Tarik M. Elsheikh, Erika E. Doxtader, Marko Golusin, Juan Xing, James F. Bena, Lame BalikaniAbstract
Background
Interpretation of oncocytic cell‐predominant cytologies (OCP) is very challenging as a majority of those cases are diagnosed as indeterminate, including atypia of undetermined significance (AUS) and follicular neoplasm (FN). The authors previously reported a combination of a four‐risk factor model highly predictive of neoplasia and malignancy that included four cytologic features and nodule size. However, that model was based on a single cytopathologist (CP) review. The current study assessed the reproducibility of previous findings among several CPs in a larger separate patient cohort, correlated with molecular studies.
Design
Four CPs independently evaluated cytologic features of 173 surgically confirmed OCP (24 AUS, 149 FN). Nodule size was recorded separately and incorporated into risk factor scores. A subset of 80 cases underwent molecular testing with Afirma or ThyroSeq. Univariable and multivariable logistic regression and ROC analyses were performed.
Results
Interobserver agreements were strong for high cellularity, moderate for isolated single cells, and poor for absent colloid and anisonucleosis. Risk factor scores demonstrated strong agreement, high specificity (77%–92%) and low sensitivity (6%–22%). Pathologist agreement did not better predict neoplasm or malignancy. Afirma and ThyroSeq molecular tests had similar performances with overall sensitivity, specificity, and false‐negative rates of 86%, 35%, and 13%–17%, respectively.
Conclusion
Cytologic evaluation of OCP was limited by wide interobserver variability and inconsistent associations with outcomes. Agreement of CPs on presence of specific cytologic features did not enhance prediction of neoplasm or malignancy. Molecular testing provided more objective performance displaying high sensitivity; however, specificity was low and FNR was elevated. Clinicians should be made aware of the greater limitations associated with cytology and molecular test results of OCP compared to those reported for their nononcocytic counterparts.