Comparison of five Ki67 antibodies regarding reproducibility and capacity to predict prognosis in breast cancer: Does the antibody matter?

Abstract

Although several antibodies are available for immunohistochemical (IHC) detection of Ki-67, even the most commonly used MIB-1 has not been validated yet. Our aim was to compare 5 commercially available antibodies for detection of Ki-67 in terms of agreement and their ability in predicting prognosis of breast cancer. Tissue microarrays were constructed from 378 breast cancer patients' representative formalin-fixed paraffin-embedded tumor blocks. Five antibodies were used to detect Ki-67 expression: MIB-1-using chromogenic detection and immunofluorescent labeled (IF) MIB1-, SP-6, 30-9, poly and B56. Semi-quantitative assessment was performed by two pathologists independently on digitized slides. To compare the 5 antibodies, intra-class correlation (ICC) and concordance correlation coefficient (CCC) were used. All the antibodies but MIB-1IF (at 20% and 30% thresholds, P=.993, P=.342, respectively) and B56 (at 30% threshold, P=.288) separated high and low risk patient groups. However, there was a significant difference (p values for all comparisons</=0.005) and moderate concordance (ICC=0.645) between their Ki67 labeling index (LI) scores. The highest concordance was found between MIB-1 and poly (CCC=0.785) antibodies. None of the antibodies except Ki67 LI as detected by poly (P=.031) at 20% threshold, and lymph node status (P<.001) were significantly linked to disease free survival in multivariate analysis. At 30% threshold, this was reduced to lymph node status (P<.001) alone. Our results showed that there are considerable differences between the different Ki67 antibodies in their capacity to detect proliferating tumor cells and to separate low and high-risk breast cancer patient groups.