AUTOMATED QUANTITATION OF IMMUNOCYTOCHEMICALLY LOCALIZED ESTROGEN-RECEPTORS IN HUMAN BREAST-CANCER

Abstract

Frozen sections of breast tumor tissue have been stained using an immunoperoxidase [estrogen receptor (ER)-immunocytochemistry] kit incorporating a monoclonal antiserum [H222] to visualize nuclear human ERs. Quantitation of specific staining has been performed by manual procedures using optical microscopy and by a computer-assisted image analysis system (CAS 100). Initial investigations with a test panel of ER-immunocytochemistry-positive tumors revealed a good qualitative agreement between CAS and manual assessments. Reduced variance was, however, observed between quantified ER-immunocytochemistry results from four experienced investigators using the CAS analysis. An extended study confirmed the relationships between CAS and manual methods of assessment. These findings were evident when studies were scored either by assessment of the percentage of positively stained cells (n = 92; r = 0.919; P < 0.01) or by H-score calculations (n = 92; r = 0.913; P < 0.01). A good correlation was also found between CAS quantification and the results of an ER enzyme immunoassay of 48 primary breast cancer specimens (r = 0.715; P < 0.05). In 49 cases it was possible to relate CAS-defined ER status and levels to the subsequent response of patients to endocrine therapy. ER was assessed on specimens obtained prior to commencement of treatments for recurrent breast cancer. Presuming the presence of ER to be a prerequisite for successful therapy, very good correlations between response and both status and levels of positivity were recorded. None of 16 patients with CAS-ER-negative tumors responded to treatment, while 16 of 33 (48.4%) CAS-ER-positive patients achieved an objective response according to International Union Against Cancer criteria. A relationship between response and the degree of CAS-ER positivity was obtained when the CAS score divisions of 0, 1-100, and >100 (response rates, 0, 41, and 64%, respectively) were used. These data demonstrate that automated image analysis offers a reliable, reproducible procedure for quantifying ER in immunocytochemically stained sections. It has potential advantages over manual procedures, providing less opportunity for subjective influences in scoring sections. Future advances in software design should further reduce elements of subjectivity and increase both the speed and reliability of results. We anticipate image analysis becoming a valuable tool in investigations concerning, for example, the influence of heterogeneity of steroid receptor distribution on the rate of recurrence of breast cancer after mastectomy and in the clinical course of the disease.