Interobserver agreement in scoring HER2-negative and HER2-low immunohistochemistry in breast cancer: Reasons for discordance and impact of a single training session.

Abstract

In the DESTINY-Breast04 trial, human epidermal growth factor receptor 2 (HER2) low (HER2-L) is defined as an HER2 immunohistochemistry (IHC) score of 1+ or 2+ with negative fluorescence in situ hybridization. Limited data report poor agreement in scoring HER2-0 vs 1+. We aim to investigate the HER2 IHC concordance rate, with focus on HER2-0 vs HER2-L groups. We evaluate the impact of a single training session on concordance. Nine breast pathologists from the same institution reviewed 60 HER2 IHC stains on breast biopsy specimens in 2 rounds of 30 cases each. An educational slide review session was provided between the 2 rounds. Interobserver Cohen κ values were computed and compared. Overall complete agreement for HER2 IHC was noted in 37 (62%) of 60 cases, with similar agreement in the first round (19/30 [63%]) and second round (18/30 [60%]). For all 60 cases, κ values ranged from .517 to .895, with 92% of κ values in substantial agreement or better range. For combined HER2-0 and HER2-L cases, κ values ranged from .298 to .826, with only 45% of κ values in substantial agreement or better range. In HER2-L only cases, 50% of the scorer pairs of κ values were 0 or less (no agreement), and only 14% of pairs showed substantial or better agreement. The educational session did not improve the κ values. Faint and heterogeneous HER2 expression, cytoplasmic blush, dislodged cells, and in situ component led to poor concordance in HER2-0 vs HER2-L. Poor concordance on HER2-0 vs HER2-L and lack of improvement after a training session likely suggest ineffective HER2 IHC expression range in HER2 low spectrum.