IMAGE-ANALYSIS FOR QUANTITATION OF ESTROGEN-RECEPTOR IN FORMALIN-FIXED PARAFFIN-EMBEDDED SECTIONS OF BREAST-CARCINOMA

Monoclonal antibody to human estrogen receptor (ER) provides a useful immunohistochemical tool for the evaluation of ER content in breast carcinoma, but visual interpretation is subjective. Computer-assisted image analysis had proved effective in immunohistochemical quantitation of ER in fresh tumor imprints and cryostat section. We examined the usefulness of this technique in 5-mu-m thick formalin-fixed paraffin-embedded tissue sections of 66 cases of primary breast carcinoma previously assayed by dextran-coated charcoal (DCC) analysis. Immunohistochemistry was automated and performed on a Code-on slide stainer (Instrumentation Laboratories, Lexington, MA) using Pronase predigestion, a monoclonal antibody (ER-ICA; Abbott, Chicago, IL), and a biotin-labeled secondary antibody. Detection was achieved with an avidin-alkaline phosphatase conjugate and nitrobule tetrazolium (NBT) bromochloroindoyl phosphate (BCIP) substrate. The immunohistochemical ER staining was analyzed visually and with the CAS/200 image analyzer (Elmhurst, IL). The visual semiquantitiative histologic scores (HSCORE), the automated quantitative assays including the percentage of positive nuclear areas (PNA), and the quantitative immunocytochemical scores (QIC SCORE = of positive stain/10) were compared with the corresponding PNA x % of positive stain/10) were compared with the corresponding DCC results. Linear correlations were demonstrated between all immunohistochemical assays and the logarithm of DCC, the strongest correlation seen with PNA (r = 0.91). Threshold points for positive HSCORE, QIC SCORE, and PNA assays were extrapolated using DCC as the reference. ER immunodetection by PNA as compared with visual examination alone was enhanced by 18% (up to 88%) in sensitivity and 34% (up to 94%) in specificity, and the DCC concordance rate increased by 26% (up to 91%). A comparative chart extrapolating DCC from PNA was thus established. Increasing the positivity threshold to attain a specificity of 100% did not lower the sensitivity of the PNA assay but demonstrated variable DCC concordance rates, decreasing from 100% to 78% with increasing DCC levels. Computer-assisted image analysis for ER quantitation on paraffin sections of breast carcinoma can be effectively used as an objective alternative to DCC.