The use of digital images improves reproducibility of the ki-67 labeling index as a proliferation marker in breast cancer.

The proportion of Ki-67 immunostained cells (Ki-67 labeling index, LI) is one of the most commonly used histology methods for estimating proliferation of breast carcinomas. Although the Ki-67 LI is used in treatment decision making, its reproducibility shows variation in different studies, and is generally less then optimal. The aim of the present study was to investigate how the use of a standardized, partially digitalized counting method could affect reproducibility of determining the Ki-67 LI. Thirty breast cancer core-biopsy samples were stained with B-56, SP-6 and MIB-1 monoclonal Ki-67 antibodies. Each sample was represented by a single digital photograph taken with a x20 objective. Four investigators determined the Ki-67 LI on these digital images by estimation, then by counting with the help of a grid overlaid on the same images. Altogether 720 evaluations were made by 4 independent pathologists. Good to excellent correlation was found between estimations and calculations of each observer. Kappa values >0.6 suggest substantial inter-observer agreement when classifying the cases into a 15 % and 30 % cut-off determined three-tiered or a 4-quarter-based four-tiered categorization, which is better than the fair reproducibility gained on the real slides in a previous study. The results also suggest that the type of the antibody may also impact on the consistency of both estimating and calculating the Ki-67 LIs. The results indicate that counting on digital images may significantly improve reproducibility of determining the KI-67 LI. Interestingly, estimation on the same images is not worse, but is obviously faster and more convenient.