Abstract

There is dearth of studies that provide a practical working formulation of breast cancer gene expression analysis for the surgical pathologist. ER, PR, HER2 were used as surrogate markers to classify 205 breast carcinomas into molecular classes. Ki-67 labeling index was calculated using an image analysis system. The data was analyzed for molecular class prevalence, and inter-relationships amongst morphologic parameters, Ki-67 index, and molecular classes. Of the 205 tumors, 113 (55%) were classified as luminal A (strong ER+, HER2 negative), 34 (17%) as luminal B (weak to moderate ER+, HER2 negative), 32 (15%) as triple negative (negative for ER/PR and HER2), 8 (4%) as ERBB2 (negative for ER/PR but HER2+), 10 (5%) as luminal A-HER2 hybrid (strong ER+ and HER2+), and 8 (4%) as luminal B-HER2 hybrid (weak to moderate ER+ and HER2+). The average Ki-67 index was lowest in luminal A (15.8%), intermediate for ERBB2 (27.8%) and highest for triple negative tumors (>50%). Multivariate logistic regression analyses found the following associations: ERBB2 tumors with apocrine differentiation (p=0.0031); Triple negative tumors with high Ki-67 index (p<0.0001) and CK5 positivity (p<0.0001); HER2 negative-low receptor positive tumors (luminal B) with increased lymph node involvement (p=0.0141). The immunohistologic criteria were validated on a different set of 359 cases treated with neoadjuvant chemotherapy, which showed a pathologic complete response predominantly in ERBB2 and triple negative tumors. Immunohistochemistry is a reliable surrogate tool to classify breast carcinoma according to the gene expression profile classification.