An intra- and interobserver reproducibility analysis of the Ki-67 proliferation marker assessment on core biopsies of breast cancer patients and its potential clinical implications.

OBJECTIVE The Ki-67 proliferation index has received an important role in treatment tailoring and molecular classification of estrogen receptor-positive breast carcinomas. The aim was to analyze the reproducibility of assessing proliferation on the basis of Ki-67 immunohistochemistry. METHODS Thirty core biopsy samples of breast cancer patients were analyzed after immunostaining with B56, SP6 and MIB-1 monoclonal Ki-67 antibodies. All samples were evaluated twice and independently by 3 pathologists, with each observer performing his daily routine practice. The ratio of Ki-67-positive cells was estimated with 5% accuracy. Correlation was calculated for the results of each investigator for all pairs of antibodies and for the results of each antibody for all pairs of investigators. Ki-67 scores were divided into categories of either 4 quarters or into 3 groups reflecting the St. Gallen consensus recommendations with 15 and 30% as cutoff values. The reproducibility of classifying the tumors into these categories was assessed with ĸ statistics. RESULTS Altogether, 540 evaluations were made. Good to excellent correlation (Spearman's and Pearson's coefficient range 0.74-0.92 and 0.73-0.93, respectively) was noted for the pairwise comparison of antibodies by observer and of observers by antibody. The inter- and intraobserver reproducibility of the Ki-67 score classification into equal quarters (1-25, 26-50, 51-75 and 76-100%) or into 3 categories with cutoffs at 15 and 30% was fair to poor in the middle categories, but moderate to substantial in the low and high ranges. Interobserver differences in practice potentially impacted on less consistent classification. CONCLUSION Our results indicate that the three different Ki-67 antibodies tested do not substantially influence the reproducibility of the estimated proliferation rates. Although reproducibility is better in the clinically more relevant distinction of high versus low proliferation, without standardization, the current practice of Ki-67 assessment in many laboratories does not allow proper and consistent therapeutic decision-making.