Statistical analysis of the concordance of immunohistochemical stains with the final diagnosis in spitzoid neoplasms.
INTRODUCTION: The classification of spitzoid melanocytic tumors can be difficult, and pathologists rely on both histological features and clinical information to arrive at a diagnosis. We proposed that an immunohistochemical panel could be useful in classifying these neoplasms and designed a study to test the independent contribution of the panel to the final diagnosis. METHODS: We identified 121 cases previously signed out either as (1) Spitz nevus, (2) atypical spitzoid neoplasm, favor Spitz nevus, (3) atypical spitzoid neoplasm of uncertain malignant potential, (4) atypical spitzoid neoplasm, favor melanoma, and (5) spitzoid melanoma. The slides were reveiwed in random order by 4 pathologists. For the first review, the pathologists received only hematoxylin and eosin sections and patient age. Subsequently, the same pathologists interpreted the immunohistochemically stained slides (S-100A6, HMB-45, and MIB-1) on the same cases in randomized order without the benefit of either hematoxylin and eosin sections or patient age. The original diagnosis (based on a combination of clinical information, hematoxylin and eosin-stained sections and immunohistochemical stains) was the gold standard used for statistical analysis. The primary aim of the study was to determine the level of agreement between interpretions based on hematoxylin and eosin sections and age, the immunostains alone, and the gold standard, thus providing a measurement of the degree to which each of these elements contributes to the final diagnosis. The agreement between the gold standard and external review was also determined for those cases sent for external review. RESULTS: The generalized kappa statistic was 0.95 for both the hematoxylin and eosin-stained slides alone and the immunohistochemical stains alone, implying a high level of agreement among the 4 pathologists. The combined weighted kappa statistic for the comparison of hematoxylin and eosin sections and patient age to the gold standard was 0.49, and for the immunohistochemically stained slides to the gold standard 0.48, indicating that a diagnosis based on hematoxylin and eosin sections alone or immunostains alone show only a moderate and similar level of agreement with the gold standard diagnosis. Only the most controversial cases were sent for external review. The weighted kappa statistic estimate was 0.30 for the gold standard diagnosis on those cases and the external review. CONCLUSIONS: Spitzoid neoplasms remain a difficult area in dermatopathology and experts frequently disagree on the most challenging cases. An immunohistochemical panel contributes to the diagnosis of spitzoid tumors, and the contribution is statistically similar to that of hematoxylin and eosin sections and age. Interpretation remains subjective, as evidenced by the comparison of the gold standard and external review.
Puri, PK; Ferringer, TC; Tyler, WB; Wilson, ML; Kirchner, HL; Elston, DM
Volume / Issue
Start / End Page
Electronic International Standard Serial Number (EISSN)
Digital Object Identifier (DOI)