Reevaluation of Human Epidermal Growth Factor Receptor 2 (HER2) Expression in Breast Cancer Samples Previously Scored As Immunohistochemistry 0 or 1+ to Characterize HER2-Ultralow Prevalence.

Context.—:Prevalence data of human epidermal growth factor receptor 2 (HER2)-ultralow (immunohistochemistry [IHC] 0 with membrane staining in ≤10% of cells; HER2 IHC 0+) breast cancer are limited. Objective.—:To evaluate the proportion of HER2-ultralow among metastatic breast cancer samples originally scored HER2 IHC 0 (with or absent membrane staining) and 1+ and to evaluate interobserver concordance between pathologists' rescores and original scores. Design.—:Whole slide images (WSIs) from metastatic breast cancer biopsy specimens originally scored HER2 IHC 0 (with or absent membrane staining) and 1+ were rescored as HER2 IHC 0 absent membrane staining, IHC 0+ (with membrane staining), IHC 1+, or IHC 2+ by 3 pathologists. Each pathologist performed 2 blinded readings (washout period of ≥2 weeks). Results.—:Of the 246 IHC 0 WSIs, 78 (31.7%), 75 (30.5%), and 86 (35.0%) were rescored as HER2 IHC 0 absent membrane staining, 0+, and 1+, respectively. Of the 138 IHC 1+ WSIs, 3 (2.2%), 10 (7.2%), 117 (84.8%), and 6 (4.3%) were rescored as IHC 0 absent membrane staining, 0+, 1+, and 2+, respectively. Interobserver concordance was substantial (overall percentage agreement [OPA; 95% CI], 86.7% [82.9%-90.0%]; Cohen κ [95% CI], 0.781 [0.726-0.835]) to moderate (OPA [95% CI], 62.0% [59.9%-66.9%]; Cohen κ [95% CI], 0.441 [0.379-0.502]). Agreement between pathologists' consensus scores and original scores was moderate (OPA [95% CI], 72.1% [67.3%-76.5%]; Cohen κ [95% CI], 0.461 [0.381-0.541]). Conclusions.—:Scoring discrepancies highlight the potential need/benefits of training for evaluating HER2-low/HER2-ultralow cases in metastatic breast cancer, which could improve consistency in HER2 scoring by pathologists.