Background: Classification of thyroid follicular neoplasms can be challenging for pathologists. Introduction of noninvasive follicular thyroid neoplasms with papillary-like nuclear features, the utilization of immunohistochemistry, and molecular analysis are all thought to be valuable diagnostic adjuncts. Our aim was to determine whether interobserver variability for follicular neoplasms has improved since the application of these adjuncts.
Methods: One representative section from a cohort of follicular neoplasms previously proven difficult for pathologists were examined independently by 7 pathologists and assigned to 1 of 3 diagnostic categories (benign, neoplasms with papillary-like nuclear features, or malignant). This process was carried out separately 3 times: (1) after viewing hematoxylin and eosin stain slides, (2) hematoxylin and eosin stain in conjunction with immunohistochemistry, and (3) hematoxylin and eosin stain/immunohistochemistry in conjunction with molecular analysis. The interobserver variability and overall agreement were then calculated using the free-marginal kappa coefficient.
Results: Agreement on hematoxylin and eosin stain was 57%, with a kappa coefficient of 0.36 (minimal agreement). The agreement improved slightly with the application of immunohistochemistry (kappa coefficient = 0.49 [weak agreement] and a percentage agreement 67%). The level of agreement decreased slightly after the addition of molecular analysis (kappa coefficient = 0.43 [weak agreement] and percentage agreement 62%).
Conclusion: Despite attempts to standardize the diagnostic criteria for neoplasms with papillary-like nuclear features and the utilization immunohistochemistry and molecular analysis, attaining pathologic consensus for difficult follicular neoplasms of the thyroid remains a challenge.
Copyright © 2020 Elsevier Inc. All rights reserved.