Quantitative imaging of immunocytochemical (PAP) estrogen receptor staining patterns in breast cancer sections.

"Receptogram Analysis" has been developed as a pattern-oriented approach for predicting endocrine response in breast cancer based upon quantification of the estrogen receptor immunocytochemical assay (ERICA), using a Quantimet Imaging System. Response prediction was evaluated in 58 stage III and IV patients receiving endocrine therapy (primarily Tamoxifen). The Receptogram is a composite of the univariate distributions of nuclear receptor content, IOD(S), and concentration (MOD), and their bivariate contour plot; where (S) is the calculated nuclear radius in section. MOD distributions were classified into four types based upon peak modality and kurtosis (I-IV), and contour plots were classified into four subtypes (A-D) based upon contour slope. Patients failing therapy were ERICA--or their receptogram revealed co-existent ER+ and ER- tumor cells (type II), highly skewed MOD distributions lacking defined peaks (type IV), or contours with nearly horizontal slope (type C). Response was realized in 9/16 type I patients, with a single positive MOD peak, and in 9/15 type III patients, with discrete, multimodal MOD peaks. In contrast, 0/8 type II, 0/12 type IV, and 0/10 type C patients were responders. Receptogram analysis was superior to cytosol assay (DCC) as a response discriminant: positive predictive value, 53% vs. 33%; negative predictive value, 100% vs. 75%; sensitivity, 100% vs. 83%; specificity, 68% vs. 23%; and accuracy, 78% vs. 41%, respectively. Alternately, patients were assigned to potentially responsive or non-responsive groups based upon thresholded mean receptor parameters: field MOD, mean nuclear MOD (NMOD), and mean NMOD(PF) where PF is the ER+ nuclear fraction. While these parameters correlated with DCC (r = .72, 0.69, and 0.69), they were only marginally better in predictive value.