Computer-Aided Quantitative Image Analysis of Multiple Human Epidermal Growth Factor Receptor 2 Immunohistochemistry Assays in Breast Carcinoma

Context.—: Quantitative image analysis (QIA) is increasingly applied for immunohistochemistry (IHC)-based biomarker assessment in pathology. With the introduction of drugs targeting breast carcinomas (BCs) with low human epidermal growth factor receptor 2 (HER2) expression levels, a need for more accurate and reproducible HER2 scoring is warranted. QIA has the potential to increase HER2 IHC scoring reproducibility. However, QIA reproducibility should be agnostic and uninfluenced by different IHC assays. Objective.—: To compare HER2 scoring accuracy and reproducibility by expert pathologists and QIA in BCs analyzed with different IHC assays. Design.—: Algorithm parameters of the QIA method were developed by using 183 slides from 6 Nordic immunohistochemical Quality Control (NordiQC) HER2 runs. The QIA method was then validated on 409 slides from 2 other runs. HER2 scores obtained by visual scoring, following 2023 European Society for Medical Oncology and American Society of Clinical Oncology/College of American Pathologists guidelines or by QIA, were compared. Results.—: Visual scoring and QIA reached 86% (κ = 0.81) overall scoring agreement for tissue cores in the final validation material. QIA showed 98% (95% CI, 97.8%-99.0%) sensitivity and 84% (95% CI, 80.3%-87.6%) specificity for the distinction between HER2 0 (no drug candidate) and HER2 1-3+ (drug candidates) scores. Ninety-nine percent (95% CI, 98.7%-99.9%) sensitivity and 100% (95% CI, 99.8%-100%) specificity were obtained for the separation between HER2 2-3+ in situ hybridization (ISH)-positive cores (HER2 classical overexpression) and HER2 0-2+ ISH-negative cores (HER2-low expression range). Conclusions.—: Our QIA method can be beneficial to use as a reliable HER2 scoring tool for HER2 classical overexpression and HER2-low in BCs. It was found to be agnostic and provided same accuracy irrespective of IHC assay applied.