Analysis of Novel Immunohistochemical Markers with a Cluster Analysis Approach To Define an Optimal Panel for the Differential Diagnosis of Renal Epithelial Neoplasms with Eosinophilic Cytoplasm
MJ Wasco, JC Carvalho, J Siddiqui, LP Kunju, DG Thomas, RB Shah. University of Michigan, Ann Arbor
Background: Renal epithelial neoplasms with eosinophilic cytoplasm often share overlapping morphologic features which pose difficulties in their classification, specifically in limited samples.
Design: To analyze and compare a panel of immunohistochemical markers useful in the differential diagnosis of eosinophilic renal tumors (S100A1, CD10, AMACR, RCC, Vimentin, CAIX, CK7, MOC-31, EPCam, Claudin 7 and 8, PR, C-kit, E-cadherin, and Parvalbumin), we stained a tissue microarray of 75 renal tumors with eosinophilic cytoplasm (18 chromophobe renal cell carcinomas (ChRCCs), 20 oncocytomas (ROs), 20 Papillary RCC (PRCCs), 17 clear cell RCC (CRCCs)), and analyzed the data with cluster analysis. Expression was categorized as none, weak/focal, or strong (0-2 scale). For CK7, a score of 2 was considered positive, while for other antibodies, a score of 1 or 2 was considered positive.
Results: Cluster analysis sorted tumors in five groups, with like tumors clustering together. ChRCC clustered separately as a distinct group from other entities. RO were a distinctly different group, but clustered closer to CRCC and PRCC. Some PRCC clustered together, but others demonstrated overlap with the CRCC group. Expression pattern of EPCam, MOC-31, and Claudin-7 clustered together, as did C-kit and E-cadherin, CD10 and AMACR, and RCC and Vimentin. S100A1 expression excluded ChRCC with 89% sensitivity and 84% specificity. A panel of CK7 (negative/focal) and S100A1 (positive, score 1 or 2) distinguishes RO from ChRCC with 90% sensitivity and 94% specificity. A panel of vimentin (negative) and C-kit (positive) distinguishes RO from CRCC and PRCC with 85% sensitivity and 86% specificity. Staining results in figure 1. PR and Claudin-8 had minimal utility on initial analysis and were excluded.
Conclusions: Our results show that hierarchical cluster analysis is an effective approach to analyze high volume immunohistochemical data to generate an optimal useful panel. A panel of Vimentin, C-kit, CK7, and S100A1 appears to be the most effective panel for difficult cases or for small biopsies.
Category: Genitourinary (including renal tumors)
Monday, March 9, 2009 1:00 PM
Poster Session II # 118, Monday Afternoon