Application of Immunocytochemistry on Cell Block Sections for the Investigation of Thyroid Lesions

Abstract

Objective: To investigate the potential of Classification and Regression Trees (CARTs) for the diagnosis of thyroid lesions based on cell block immunocytochemistry and cytological outcome. Study Design: A total of 956 histologically confirmed cases (673 benign and 283 malignant) from patients with thyroid nodules were prepared via liquid-based cytology and evaluated; 4 additional slides were stained for cytokeratin 19 (CK-19), galectin 3 (Gal-3), Hector Battifora mesothelial cell 1 (HBME-1), and thyroglobulin. On the basis of immunocytochemistry and the cytological diagnosis, a CART algorithm was constructed and used for evaluation. Results: The major important factors contributing to the diagnostic CART model were: cytological outcome, CK-19, Gal-3, and HBME-1. The sensitivity and specificity of the cytological diagnosis were 96.27% and 88.26%, respectively (cut-off: category 3 of The Bethesda System [TBS-3]). The introduction of immunocytochemistry and the CART model increased the sensitivity and specificity to 98.88% and 99.11%, respectively. CK-19 presented the best performance for discriminating papillary thyroid carcinomas, followed by HBME-1 and Gal-3. In the TBS-2 cases, CK-19 and, subsequently, Gal-3 were important immunocytochemistry markers. Ultimately, CK-19 and HBME-1 on TBS-5 or TBS-6 cases demonstrated the best results. Conclusions: The hierarchical structure of the CART model provides a diagnostic algorithm linked with the risk of malignancy at every step of the procedure. It also provides guidance on the use of ancillary examinations as it goes by simple, human understandable rules.