Abstract 3762: Automated analysis of breast cancer tissue microarrays to investigate etiologic heterogeneity in the Breast Cancer Association Consortium

Abstract

Abstract Background: Data suggest that breast cancer risk factor associations vary by tumor marker expression. However, individual studies often lack the power required to assess these relationships and pooled analyses are limited by non-standardized scoring of tumor markers. To address these limitations, the Breast Cancer Association Consortium is assessing whether automated image analysis of immunohistochemically stained tissue microarrays (TMAs) can permit standardized scoring of tumor markers. Methods: TMA sections prepared in 10 case-control studies containing up to 9,298 breast cancers stained for estrogen receptor (ER); progesterone receptor (PR); human epidermal growth factor receptor 2 (HER 2); epidermal growth factor receptor (EGFR) and cytokeratin 5/6 (CK 5/6) were scanned as digital images. Automated algorithms were used to score markers in tumor cells using the Ariol system. We are analyzing consistency of automated scores within individual studies; between studies; and against visual reads. Results: Preliminarily, distributions of automated scores were generally similar across different TMA blocks within a study; however isolated blocks yielded disparate results related to low or high intensity staining for a marker or the counterstain across the entire slide. Distributions of scores varied between studies, suggesting the need for study-specific cutpoints for negative vs. positive. Analysis of receiver operator characteristic curves comparing automated analysis to independently recorded visual reads in two studies demonstrated areas under the curves of 0.90 and 0.93 for tumors classified visually as ER+/PR+ and 0.91 and 0.88 for tumors scored visually as ER-/PR-. Cores scored as negative for EGFR or CK 5/6 by automated methods agreed well with visual scores, but many tumors scored as positive by automated analysis were negative by visual review, irrespective of ER / PR status. Thus, restricting analyses to ER-/PR- tumors (a subset of interest) reduced the number of discrepancies. Ongoing analyses are comparing automated and visual scores for all markers / studies and are assessing risk factor associations for breast cancer by tumor marker expression (negative vs. positive and strength of expression) based on automated scores. Conclusions: Automated analysis of immunohistochemically stained TMAs of breast cancers may be useful in standardizing high throughout characterization of cancer markers in multicenter studies and in consortia, although additional methodologic work is needed. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 3762.