Inter-observer concordance of Ki-67 labeling index in breast cancer: Japan Breast Cancer Research Group (JBCRG) Ki-67 Ring Study.

Background: The standardized assessment of Ki-67 labeling index (LI) plays pivotal roles in identifying the patients (pts) with primary breast cancer who could benefit from systemic chemotherapy, in particular among pts with estrogen receptor(ER)-positive cancers. Therefore, in this study, we evaluated the inter-observer concordance of the assessment of Ki-67 LI in archival materials. Methods: Six surgical pathologists specializing in breast pathology from different Japanese institutions participated in this study. All slides were prepared from archival tissues of breast cancer fixed in 10% buffered formalin for 24 hours in a single institution (KU). Three independent studies were conducted. Study 1) Six consecutive slides were prepared from 5 cases. A slide from each case was stained with MIB-1 (DAKO, Denmark) in each institution according to their routine methods. Total of 30 stained slides were assessed for Ki-67 LI by each pathologist using two different modes of assessment. One is the scoring system in which the rate of positive cells was scored from 1 (0-9 %) to 10 (90-100%) without counting the cell number. The second one is the counting system in which approximately 1000 cells in total were counted in the hot spots and the positive rate was calculated. Study 2) Twenty tumors with Ki-67 LI ranging from 5 to 25 (15 ± 10) %, stained in a single institution (KU) were assessed by each pathologist by the counting system. Study 3) In order to avoid variations by assessment in varied microscopic fields and to further evaluate the variation of threshold of immunointensity interpreted as positive by different pathologists, fifteen printed photographs of Ki-67-stained slides were sent and assessed for Ki-67 LI by each participating breast pathologist. Results: Study 1) The counting system demonstrated a better correlation of Ki-67 LI among six pathologists than the scoring system {the intraclass correlation coefficient (ICC), 0.66 (95% confidence interval 0.52-0.78) for the counting system, 0.57 (0.42-0.72) for the scoring system}. The two assessment systems showed a moderate correlation {ICC, 0.68 (0.60-0.75)}. Study 2) The assessment of Ki-67 LI in 20 slides with Ki-67 LI of 5 to 25 % demonstrated a correlation similar to that in the specimens with an unrestricted range of Ki-67 LI in the study 1 {ICC, 0.68 (0.50-0.81) for the study 2, 0.66 (0.52-0.78) for the study 1}. Study 3) The assessment of Ki-67 LI in the same photographs yielded a considerably significant concordance among six pathologists {ICC, 0.94 (0.88-0.97)}. Conclusion: The counting system turned out better than the scoring system in terms of the inter-observer agreement of the Ki-67 LI assessment. The degree of concordance was by no means influenced by the range of Ki-67 LI. The concordance of the Ki-67 LI assessment among six participating pathologists was significantly high when the assessed field was fixed using the same photographs for evaluation, suggesting that the selection of the fields for evaluation is critical. These results suggest that identification of hot spots for evaluation is pivotal for obtaining the accurate Ki-67 LI of breast cancer and still images of these hot spots could provide reproducible results. [ABSTRACT FROM AUTHOR]