A lack of consistent methods to evaluate Ki67 expression is problematic in terms of accurately predicting prognosis in breast cancer. Accordingly, this study aimed to identify the causes of discrepancies in Ki67 labeling index measurements by different observers under different conditions using breast cancer samples.
This Japanese study group compared and assessed immunohistochemical (IHC) analysis of the Ki67 labeling index when measured by different pathologists. Six pathologists (pathologists A-F) in Japan participated in this ring study. One hundred and ten surgical cases of estrogen receptor-positive and human epidermal growth factor receptor 2-negative invasive breast cancer treated in 2007 were identified from the breast cancer database of Tokai University Hospital and were included in this study.
For all 6 pathologists, the Ki67 labeling index were significantly different between grade 3 and grade 1 cases and between grade 3 and grade 2 cases, whereas the index tended to be different between grade 1 and grade 2 cases. Further, the Ki67 labeling indexes measured by the 6 pathologists were strongly correlated (rho: 0.73-0.88). The IHC scores recorded by pathologist A were in moderate to good agreement with those recorded by the others in patients with a Ki67 labeling index of < 13.25 % and in those with a Ki67 labeling index of > 13.25 % (kappa = 0.429-0.660). The Ki67 low and high concordance rates between pathologist A and the others were 0.452-0.778 and 0.862-0.979, respectively. The most pertinent reason for discrepancy in scores seemed to be the selection of the area for counting and the quality of nuclear staining.
The Ki67 labeling index measured by 6 pathologists without method standardization was in fair to good agreement. We plan to undertake a second ring study, pending recommendations by the international Ki67 panel.