Comparison of Ki-67 labeling index measurements using digital image analysis and scoring by pathologists

Background Routine analysis of Ki-67 is not widely recommended for clinical decision-making because of poor reproducibility. Furthermore, counting numerous cells can be laborious for pathologists. Digital image analysis for immunohistochemical analysis was recently developed; however, the clinical efficacy of the Ki-67 index obtained using image analysis is unknown. Methods We retrospectively identified female patients with breast cancer with immunohistochemical Ki-67 and survival data using the pathology database at the Tokai University, Japan. Ki-67 expression was scored by three pathologists. Slides were scanned and converted to virtual slides; Ki-67-positive cells were counted using image analysis. Ki-67 indices obtained by the pathologist’s scoring and image analysis were evaluated by 2 × 2 analysis. Relationships between Ki-67 index and survival outcomes were evaluated using the Kaplan-Meier method and compared using the log-rank test. Results Based on the 2 × 2 analysis, Ki-67 index obtained using image analysis was moderately correlated with the pathologist’s scoring for all patients (κ 0.41; sensitivity, 0.573; specificity, 0.878). Poorer relapse-free survival was associated with high Ki-67 index than with low Ki-67 index for estrogen receptor-positive, human epidermal growth factor receptor 2-negative, and stage I or II patients scored by pathologists (p < 0.001) and obtained using image analysis (p = 0.031). Conclusions The Ki-67 indices obtained using image analysis were moderately correlated with those scored by pathologists. Digital image analysis can be effective for measuring Ki-67 values, because they are associated with relapse-free survival in estrogen receptor-positive, human epidermal growth factor receptor 2-negative, and patients at stage I or II.