Prognostic value of different cut-off levels of Ki-67 in breast cancer: a systematic review and meta-analysis of 64,196 patients

A proliferative marker, expressed as the percentage of cells in a cell cycle, has been developed and used as a discriminant of more aggressive malignant phenotypes in early breast cancer (BC). The marker is usually expressed by the immunohistochemical staining of the cell cycle antigen Ki-67. It has not, however, yet been definitely evaluated, due to methodological concerns, which specific Ki-67 cut-off provide the strongest prognostic information in resected BC. We conducted a meta-analysis to explore the prognostic value of different cut-off levels of Ki-67 in terms of overall survival (OS) and disease-free survival (DFS) in early BC. The databases of PubMed, the ISI Web of Science, EMBASE, SCOPUS, the Cochrane Central Register of Controlled Trials, and CINHAL were used to identify the relevant literature. Data from studies reporting a hazard ratio (HR) and a 95 % confidence interval (CI) calculated as a multivariate analysis were pooled in a meta-analysis, with metaregression used to test for trends in predefined subgroups. All the statistical tests were 2-sided. Forty-one studies encompassing 64,196 BC patients were included in the analysis. Overall, n = 25 studies were available for the OS analysis. The pooled HR for high versus low Ki-67 was 1.57 (95 % CI 1.33-1.87, P < 0.00001). Twenty-nine studies were available for the DFS analysis. The pooled HR for high versus low Ki-67 was 1.50 (95 % CI 1.34-1.69, P < 0.00001). When a cut-off of Ki-67 staining a parts per thousand yen 25 % was used, the pooled HR for OS was 2.05 (95 % CI 1.66-2.53, P < 0.00001), which was significantly different to studies where the cut-offs chosen were < 25 %. In ER+ tumors, the HR for high versus low Ki-67 was similar and significant (HR = 1.51, 95 % CI 1.25-1.81, P < 0.0001). We conclude that Ki-67 has an independent prognostic value in terms of OS in BC patients. The Ki-67 threshold with the greatest prognostic significance is as yet unknown, but a cut-off > 25 % is associated with a greater risk of death compared with lower expression rates.