Comparison between digital image analysis and visual assessment of immunohistochemical HER2 expression in breast cancer

Background: Assessment of HER2 status is considered standard of care in the histopathologic workup of breast cancer and conveys prognostic and predictive information used to guide treatment decisions. The assessment is often carried out in a two-step approach where immunohistochemical expression of HER2 protein is first evaluated by conventional microscopy and equivocal cases are further analyzed by in-situ hybridization techniques to assess gene amplification status. In this study we compared conventional manual assessment of immunohistochemical HER2 expression with digital image analysis (DIA) and consensus manual assessment by a panel of three pathologists. Methods: From our archive we retrieved sections of 109 invasive breast carcinomas stained for HER2 with corresponding HER2 score from the original pathology report. The glass slides were assessed by three pathologists to reach a consensus score. Next, the slides were scanned into whole slide images and DIA was performed using Aperio Imagescope. The scoring results were then compared with gene amplification status evaluated by dual in-situ hybridization (DISH). Results: Comparing manual assessment with consensus assessment and DIA, good agreement was obtained with weighted kappa coefficients of 0.79 (manual vs. consensus) and 0.67 (manual vs. DIA). When compared with gene status assessment by DISH, agreement analysis yielded weighted kappa coefficients of 0.52 (manual vs. DISH), 0.58 (consensus vs. DISH) and 0.78 (DIA vs. DISH). There were no false negatives by any of the three methods and false positives ranging from 0.9 to 2.8%. The proportion of equivocal cases by each method was 44% (manual), 33.3% (consensus) and 14.7% (DIA). Application of DIA reduced the number of equivocal cases by 67% without increasing the proportion of false negatives. Conclusion: We conclude that DIA is an accurate method to reduce the number of HER2 equivocal cases without affecting the sensitivity of the HER2 assessment.