Tumor proliferative activity represents one of the most thoroughly investigated cellular functions in breast cancer (BC) for its association with tumor behavior [¹]. Assessment of proliferation rates has been shown to provide useful information on prognosis and aggressiveness of individual cancers and can potentially be used to guide treatment protocols in clinical practice. Recently, a meta-analysis of publicly available breast cancer gene-expression signatures has identified proliferation as the key biologic driver in all nine prognostic signatures included in the study [²]. It was also demonstrated that the assignment of basal-like and ERBB2⁺ (HER2⁺) breast cancers to the poor-prognosis groups by first-generation gene signatures is determined mainly by high expressions of proliferation-related genes [²,³]. Moreover, it has been reported that expression of genes involved in cell proliferation is the most heavily weighted component in calculating the recurrence score [⁴] and is the basis of genomic grading [⁵,⁶].

Various techniques have been developed to quantify proliferation rates, including, mitotic-count estimates, measurement of DNA synthesis, and flow cytometry [¹,^7-9]. Newer techniques include detection of antigens closely associated with proliferation by using immunohistochemistry (IHC). In theory, the latter methods are quicker, cheaper, and easier to use than flow cytometry and autoradiography and are more reproducible than is mitotic figure counting [⁷]. Although most studies of different proliferation assays displayed significant agreement in outcome predictions for individual patients, no consensus exists on the best proliferation assay [¹⁰]. Mitotic-count scores (MSs) are used in routine BC histologic grading, and the prognostic significance is well established [¹¹,¹²]. Immunohistochemical expression of Ki-67 is now widely used as an IHC measure of proliferation [^13-15]. Although Ki-67 is not yet incorporated into routine pathology practice, its use in proliferation assays in translational studies has yielded promising results [^16-18]. Several studies have sub-classified luminal tumors (that is, estrogen receptor (ER)-positive, HER2 negative) based on differential expression of proliferation-associated genes "proliferation signature," gene-expression profiling (GEP), or by Ki-67 in IHC studies [¹⁶,¹⁷,¹⁹]. However, some technical and validation issues are to be addressed before Ki-67 is used in BC routine practice. It remains to be determined whether Ki-67 IHC outperforms mitotic counting as a predictor of outcome. Moreover, evidence of the prognostic value of using MS in an algorithm for defining molecular BC classes is yet to be elucidated.

Consequently, this study was conducted on a large, well-characterized series of early (stage I through III) invasive BC with long-term follow-up to study the prognostic value of MS and Ki-67LI, as methods of proliferation assays, in different BC molecular classes. We used a split-sample development and validation approach to determine the optimal Ki-67LI cut-offs and demonstrated its prognostic relevance in the validation cohort in luminal/HR⁺, HER2⁺, and triple-negative tumors.

In the current study, informative results for Ki-67LI, MS, and follow-up data were available for 1,457 cases (94% of the whole series). To assess the optimal number of FFSs sufficient to report Ki-67LI reliably, four FFSs were cut from four different paraffin blocks representative of 25 invasive BC cases. These showed high levels of concordance between sections when Ki-67LI was analyzed as continuous variable (P < 0.0001; Table 2) and when categorized by using different cut-off points (kappa value, 0.834; 95% CI, 0.76 to 0.92), indicating that one FFS is sufficient for reliable Ki-67LI assessment.

Significant association was observed between Ki-67LI and MKI67 gene transcript, as defined by microarray-based gene-expression profiling (r² = 0.24; P < 0.0001), and between Ki-67LI and MS (P < 0.0001). High Ki-67LI and MS were associated with premenopausal status, larger tumor size, definite vascular invasion, and lymph node involvement (P < 0.0001). Significant differences appeared between BC molecular classes regarding their Ki-67LI (ANOVA, F = 167; P < 0.0001) (Figure 1). Bonferroni post hoc testing of Ki-67LI revealed significant differences between luminal (lowest Ki-67LI) and HER2⁺ (intermediate Ki-67LI) and between HER2⁺ and triple-negative (highest Ki-67LI) classes. However, the difference between BLBC and TNnon-B was not significant (P = 0.263). Similar results were found when MS was used to define proliferation. When the prognostic results of Ki-67LI were expressed as a continuous variable in the class of luminal ER⁺/HER2^-tumors, this showed association with outcome (χ² = 80; P < 0.0001; HR = 1.1; 95% CI, 1.0 to 1.2). Moreover, Ki-67LI cut-offs at 10% increments within the same class showed statistically significant differences between the resulting patients' subsets (LR = 142.64; P < 0.0001; HR = 1.3; 95% CI = 1.2 to 1.3; Figure 2).

Although the prognostic significance of proliferation in BC has been documented and validated in several independent studies, recent global gene-expression profiling studies have reemphasised its biologic and prognostic importance. Several studies have identified the proliferation signature as a key element in the molecular classification of BC and in the composition of different prognostic and predictive gene signatures [²]. Therefore, some recent studies have raised the issue of using an immunohistochemical marker for assessment of proliferation (for example, Ki-67) to be used in combination with other IHC surrogate panels used for BC molecular classification: ER, PR, and HER2 [¹⁶,¹⁷]. These studies have demonstrated that Ki-67LI is associated with outcome and, when used in combination with other markers mentioned previously, provides valuable prognostic information and can subclassify luminal (HR⁺ tumors) into prognostically distinct subclasses. Therefore, in the current study, we used a well-characterized series of operable invasive BC to assess the prognostic significance of Ki-67LI with regard to BC molecular classes compared with that provided by the routinely assessed MS and addressed some technical issues for use in routine practice in assessing tumor proliferation.

The distributions of Ki-67LI values in this series (mean and median) were consistent with those found in previous studies [²⁹,³³], and have a significant correlation with MKI67 gene transcript and MS. Consis-tent with the results of GEP studies [²], the majority of HER2⁺ and TN tumors showed high proliferative activity in terms of high Ki-67LI and high MS, whereas up to half of luminal BCs were of low MS and showed low Ki-67LI.

In luminal BC, proliferation assays using MS and Ki-67LI identified subdivisions with statistically different patient outcomes. In this group, both MS and Ki-67LI retained their significant and independent association with outcome and identified three subclasses; one (low-proliferative group) associated with excellent prognosis where adjuvant chemotherapy is unlikely to provide benefit and could potentially be withheld, and one class (high-proliferative group) with a worse prognosis akin to HER2⁺ tumors trastuzumab (Herceptin)-naïve patients) that may be an appropriate subclass likely to benefit from chemotherapy. These findings confirm the prognostic relevance of routinely assessed MS in the luminal class, as previously reported in BC [¹,⁷,⁸]. It is, however, important to recognize that our results relating to MS are based on optimized and standardized methods for tissue handling, fixation, and preparation [²¹]. Suboptimal tissue fixation has been demonstrated to affect adversely the ability to assess mitotic frequency, resulting in a systematic downgrading of cases [^34-36]. Critical evaluation of these issues with recommendations for good practice has been provided by professional organizations. Significant improvements in the consistency of assessment of mitotic counts and hence histologic grading have been observed on a national basis in the United Kingdom through publication of guidelines with linked educational activity and associated external quality assurance [³⁴]. However, it is worth mentioning that the prognostic information obtained by Ki-67LI in luminal tumors is equivalent to or even greater (in terms of HR) than that provided by MS alone and than that provided by histologic grade or HER2 or PR status.

Different authorities have highlighted the prognostic significance of Ki-67LI in luminal BC, with emphasis on its ability to stratify patients into low-risk and high-risk populations. Although, in these studies, luminal BCs were subdivided into distinct subgroups, the optimal cut-off point remains nonstandardized, with most studies using a single, usually the median, cut-off point; dividing these tumors into two subgroups [¹⁶]. In this study, two cut-offs for Ki-67LI (10% and 50%) were identified as derived from associations with patient outcome in luminal tumors, which is the largest class of BC; three subdivisions with significantly different outcomes with adequate numbers of cases within each subgroup. In this study, the proliferative activity assessed by using Ki-67LI was compared with mitotic counts that were scored in routine practice by using full-face sections (one to four sections per tumor). Ki-67 LI was assessed by using full-face sections to avoid missing the hot spot and to provide data that can be used in routine practice rather than TMA, which is currently a research tool. Although some research studies of Ki-67LI have reported correlation between Ki-67LI assessed on TMA or needle-core biopsy [³⁷] and that assessed on full-face sections, these correlations are not absolute, and discrepant cases exist. The results of the current study support other studies that demonstrated that Ki-67LI should be assessed in the most active areas (hot spots), a method that has also been used and validated in the assessment of mitotic counts [²¹].

The majority of HER2⁺ and TN tumors are known to be poorly differentiated and highly proliferative [²]. No association between MS and outcome could be identified in either HER2⁺ or TN classes. None of the previously reported Ki-67LI cut-off points was able to stratify HER2⁺ or TN tumors into clinically relevant subclasses. Although the high Ki-67LI cut-off generated in HER2⁺ and TN training sets appeared to stratify these tumors into proliferative subgroups based on outcome, no association with survival was identified in the validation sets. These findings could be explained by the small number of cases in these molecular classes after splitting them into training and validation sets or by their high proliferation rate, which limits the ability of a proliferation marker to identify clinically distinct subclasses.

The number of sections from each case required to assess Ki-67LI reliably in a clinical laboratory setting also was addressed. The results obtained from each of four sections assessed per tumor showed a high level of concordance, indicating that using a single FFS/case appears appropriate and representative.

It is important, however, to mention that in the current study, outcome was assessed in a context in which the treatment given may not be homogeneous, and indeed, in some cases, the parameter being addressed (mitotic score as a grading component) may have affected the original systemic treatment decision. This may limit the ability of the prognostic analyses in making assertions that either mitotic counts or Ki-67LI can identify a group in which chemotherapy could be withheld.

In conclusion, proliferation assessment by using Ki-67LI and MS can distinguish subgroups of patients with luminal/HR⁺ BC with significantly different clinical outcomes. Overall, both showed comparable results, with Ki-67LI having a marginal advantage in terms of patient-cohort separation. Neither MS nor Ki-67LI has additional prognostic value in HER2⁺ and TNBC. This study emphasises the importance of determination of appropriate clinically relevant cut points for Ki-67LI and demonstrates that sufficient IHC assessment of Ki-67LI can be achieved by using a single FFS. This method could provide a cost-effective method for prognostic subclassification of luminal/HR⁺ BC in routine clinical practice.

CISH: chromomeric in situ hybridization; ER: estrogen receptor; FFPE: formalin-fixed paraffin-embedded; HER2: human epidermal growth factor receptor 2; IHC: immunohistochemistry; Ki-67LI: Ki-67 labeling index; MSs: mitotic frequency scores; PR: progesterone receptor; TN: triple negative; TNnon-B: triple-negative non-basal.

The authors declare that they have no competing interests.

MA performed the practical work, performed the statistical analysis, and drafted the manuscript. EAR conceived and designed the project, participated in the statistics and in writing the manuscript. AAB, FB, and ARG participated in the practical work and contributed to the manuscript. IOE and JSR-F contributed to the project design and to the manuscript. All authors read and approved the final manuscript.

We acknowledge the support of the University of Nottingham and Nottingham University Hospitals Trust in funding EA Rakha and IO Ellis. We thank the Ministry of Higher Education (Egypt) for funding MA Aleskandarany, and the Breast Cancer Campaign for funding AR Green.