Background: Lesions of the thyroid gland composed of Hurthle cells encompass pathologic entities ranging from hyperplastic nodules with Hurthle cell metaplasia to Hurthle cell carcinomas. The cytologic distinction between these entities can be diagnostically challenging. Many cytologic features of Hurthle cell lesions that distinguish neoplastic Hurthle cell lesions requiring surgery from those that are benign and nonneoplastic have been described, but with variable usefulness. This is due, in part, to the small numbers of cases examined in previous studies and the limited application of statistical analysis. A morphologic study was made of 139 Hurthle cell lesions of the thyroid gland and statistical analysis applied to identify a set of cytomorphologic features that distinguish benign Hurthle cell lesions (BHCL) from Hurthle cell neoplasms (HCN).
Methods: Fine-needle aspiration biopsies (FNABs) of thyroid nodules with a predominant Hurthle cell component and corresponding histologic followup were included in the study. Cases were divided into BHCL and HCN groups on the basis of the histologic diagnosis. All cases were reviewed to assess the following 14 cytologic features: overall cellularity, cytoarchitecture, percentage of Hurthle cells, percentage of single cells, percentage of follicular cells observed as naked Hurthle cell nuclei, background colloid, chronic inflammation, cystic change, transgressing blood vessels (TBV), intracytoplasmic lumina, presence of multinucleated Hurthle cells, nuclear to cytoplasmic ratio, nuclear pleomorphism/atypia, and nucleolar prominence. The results were evaluated by using univariate and stepwise logistic regression (SLR) analysis; statistical significance was achieved at P-values < 0.05.
Results: One hundred thirty-nine FNAB specimens, corresponding to 56 HCN and 83 BHCL, fulfilled the study criteria. Six of the 14 cytologic features evaluated were shown by univariate analysis to be statistically significant in predicting HCN: nonmacrofollicular architecture (P < 0.001), absence of background colloid (P < 0.001), absence of chronic inflammation (P < 0.001), presence of TBV (P < 0.001), > 90% Hurthle cells (P < 0.001), and >10% single Hurthle cells (P = 0.014). The first four of these features were also shown to be statistically significant in the SLR analysis (P = 0.005, 0.010, 0.016, and 0.045, respectively), and when all four of these features were present HCN was correctly identified 86% of the time.
Conclusions: In the current study of 139 FNAB specimens of thyroid Hurthle cell nodules, 14 cytologic features were examined and 6 were found to be statistically significant in identifying HCN. The following four features, when found in combination, were found to be highly predictive of HCN: nonmacrofollicular architecture, absence of colloid, absence of inflammation, and presence of TBV.
Copyright 2006 American Cancer Society.