Quantification of estrogen receptors on paraffin-embedded tumors by image analysis

Emphasis on early detection of breast carcinomas has increased the number of instances in which an insufficient amount of tissue is available for biochemical estrogen receptor (ER) assay. Image analysis, used together with immunohistochemistry, introduces the possibility of quantifying molecules, such as ER, in routinely processed tissues. To explore that possibility, sections from 40 formalin-fixed, paraffin-embedded breast carcinomas with biochemically determined ER values were reacted with H222 anti-ER antibody (ER-ICA) and quantified on a CAS 200 image analysis system. A minimum of ten fields comprising at least 15,000 microns 2 of nuclear area were analyzed in each case. If the antigen distribution were not homogeneous, proportional sampling of the different tumor areas was carried out. Of the 31 ER-positive (10 to 344 fmol/mg) tumors, 27 (87%) were immunoreactive by the ER-ICA assay (greater than 5% positive nuclear area), which represents an improvement over simple microscopic evaluation (68%). Bivariate analyses showed statistically significant, albeit only modest, correlation between ER values and the percentages of positive area (r = 0.556) and positive stain (r = 0.518). The following obstacles were found to interfere with a proper correlation: (a) antigen loss during fixation and processing; (b) intratumoral antigenic heterogeneity; and directly associated with it, (c) interobserver variability. Uniform tissue handling or, alternatively, the use of internal controls to compensate for fixation-induced differences, together with thorough assessment of the tissue, should reduce these obstacles and facilitate accurate and reproducible quantification.