Interobserver consistency and diagnostic challenges in HER2-ultralow breast cancer: a multicenter study

Abstract

Background

Recent advancements in novel antibody–drug conjugates (ADCs) have demonstrated efficacy in patients with human epidermal growth factor receptor 2 (HER2)-ultralow breast cancer (BC), expanding the eligibility for anti-HER2 targeted therapy to include some patients previously categorized as HER2 immunohistochemistry (IHC) 0. This expansion underscores the need for pathologists to accurately differentiate HER2-null and HER2-ultralow.

Materials and methods

Thirty-six pathologists from four centers nationwide conducted microscopic visual assessments on HER2 IHC slides from 50 consecutive BC surgical specimens, all previously diagnosed as HER2 IHC 0.

Results

The interobserver consistency in differentiating HER2-null from HER2-ultralow, measured by Fleiss κ, was only 0.230—lower than the consistency for combined HER2 IHC 0 cases (Fleiss κ = 0.344) and binary classification (HER2-null versus HER2-non-null; Fleiss κ = 0.292). High agreement for HER2-null versus HER2-ultralow differentiation was achieved in only 4% of cases, while combining them into HER2 IHC 0 raised high agreement cases to 32%, higher than the 18% seen in the binary classification. Consensus among the 36 pathologists aligned with historical scores in 72% of cases; however, when subdividing HER2 IHC 0 into HER2-null and HER2-ultralow, the consistency dropped to 54%.

Conclusions

The low consistency among pathologists in distinguishing HER2-null, -ultralow, and 1+ cases may impact patient eligibility for new ADC therapies. To address this challenge, there is a need for improved detection methods, artificial intelligence-assisted quantitative assessments, and larger clinical datasets to refine the definition of HER2-ultralow.