Performance of automated scoring of ER, PR, HER2, CK5/6 and EGFR in breast cancer tissue microarrays in the Breast Cancer Association Consortium

Abstract

Breast cancer risk factors and clinical outcomes vary by tumour marker expression. However, individual studies often lack the power required to assess these relationships, and large‐scale analyses are limited by the need for high throughput, standardized scoring methods. To address these limitations, we assessed whether automated image analysis of immunohistochemically stained tissue microarrays can permit rapid, standardized scoring of tumour markers from multiple studies. Tissue microarray sections prepared in nine studies containing 20 263 cores from 8267 breast cancers stained for two nuclear (oestrogen receptor, progesterone receptor), two membranous (human epidermal growth factor receptor 2 and epidermal growth factor receptor) and one cytoplasmic (cytokeratin 5/6) marker were scanned as digital images. Automated algorithms were used to score markers in tumour cells using the Ariol system. We compared automated scores against visual reads, and their associations with breast cancer survival. Approximately 65–70% of tissue microarray cores were satisfactory for scoring. Among satisfactory cores, agreement between dichotomous automated and visual scores was highest for oestrogen receptor (Kappa = 0.76), followed by human epidermal growth factor receptor 2 (Kappa = 0.69) and progesterone receptor (Kappa = 0.67). Automated quantitative scores for these markers were associated with hazard ratios for breast cancer mortality in a dose‐response manner. Considering visual scores of epidermal growth factor receptor or cytokeratin 5/6 as the reference, automated scoring achieved excellent negative predictive value (96–98%), but yielded many false positives (positive predictive value = 30–32%). For all markers, we observed substantial heterogeneity in automated scoring performance across tissue microarrays. Automated analysis is a potentially useful tool for large‐scale, quantitative scoring of immunohistochemically stained tissue microarrays available in consortia. However, continued optimization, rigorous marker‐specific quality control measures and standardization of tissue microarray designs, staining and scoring protocols is needed to enhance results.

Performance of Three-Biomarker Immunohistochemistry for Intrinsic Breast Cancer Subtyping in the AMBER Consortium

“…The advent of automated methods to analyze digital pathology data has begun to support high-throughput IHC-based breast cancer subtyping in large epidemiologic studies [11]. Simultaneously, RNA-based methods have become more readily available for application in formalin-fixed paraffin-embedded (FFPE) tissues [12].…”

Crowdsourcing for translational research: analysis of biomarker expression using cancer microarrays

“…However, this proof of principle was performed in the most common cancer and marker available, which can be analysed accurately using automated methods (e.g., Turbin et al , 2008; Bouzin et al , 2015; Howat et al , 2015). Here, we tested analytically challenging cancer types as well as immunohistochemical stains for which algorithms are either scarce or require considerable involvement from experts.…”

“…Here, we tested analytically challenging cancer types as well as immunohistochemical stains for which algorithms are either scarce or require considerable involvement from experts. By testing the crowdsourcing approach across a breadth of samples, we have shown this method to be flexible and widely applicable, including in sample types where algorithms struggle (Howat et al , 2015). Although both sample types we used for IHC scoring achieved high correlations with experts, the higher level of accuracy for bladder/p53 samples compared with lung/EGFR is most likely caused by the fact that the former is a nuclear marker whereas the latter is membranous.…”

“…Many of the solutions proposed to address this deficit can only be realised in the long-term, whereas more resource is required immediately to ensure an ongoing contribution of tissue sample interrogation to translational research. Machine learning promises to automate many routine evaluations (Bolton et al , 2010; Wilbur, 2014; Bouzin et al , 2015; Howat et al , 2015), but commonly requires large, validated data sets for its development. Crowdsourcing can provide such data sets in addition to solving an immediate need for analytical resource.…”

“…Tissue microarrays facilitate high-throughput molecular analysis of tissue samples to investigate associations between tumour-specific protein expression and clinical outcomes (Giltnane and Rimm, 2004). Although automated analysis of TMAs has proven to be effective for specific screening protocols, particularly in breast cancer (Turbin et al , 2008; Bolton et al , 2010; Konsti et al , 2011; Howat et al , 2015), it was also observed that algorithms underperform on less well-established markers such as cytokeratin (CK) 5/6 and epidermal growth factor receptor 1 (EGFR/HER1; Howat et al , 2015). In the same study, 20–25% of samples had to be manually excluded from the analysis.…”