Ki67 has been extensively assessed and reported as a prognostic and predictive marker in invasive breast cancer (BC) [
1‐
7]. High Ki67 expression in BC is associated with worse prognosis. In two meta-analyses published in 2007 and 2008, high Ki67 expression in both node-positive and node-negative invasive BC showed significantly worse overall and disease-free survival [
8,
9]. Additionally, results of a systematic review support the role of Ki67 as a prognostic marker [
10] and as an independent predictive factor for neoadjuvant chemotherapy in BC patients [
5,
6,
11]. Furthermore, the St. Gallen consensus panel has recommended Ki67 as a marker for the definition of intrinsic BC subtypes to differentiate between luminal A and luminal B subgroups [
12,
13].
In clinical practice, the evaluation of prognostic/predictive factors usually depends upon the stratification of the patients into distinct risk groups based on the status of such factor. The common approach is the choice of an optimal cut-off point for the prognostic/predictive factor, assessed as a continuous variable, e.g. percentage of cells stained, to define these groups. The optimal cut-point for Ki67 in BC is currently debatable despite the large number of published studies reporting significant results [
14,
15]. The recent report on the second phase of the Ki67 trial reported that there was a need to standardise the pre-analytical and analytical features for Ki67 immunohistochemistry, so that it can be incorporated to drive patient-care decisions in clinical practice [
16]. In 2009, the St. Gallen panel proposed that Ki67 expression should be stratified into three groups: low <15%, intermediate 16–30% and high >30% [
17]. This was based on univariate analysis carried out with different Ki67 expression cut-points to find those best stratifying the patients with lowest significant
p values according to survival using Ki67 immunoreactivity and standardised mitotic index [
18]. In 2011, St. Gallen recommended an alternative KI67 cut-point at 14% in order to separate Oestrogen Receptor (ER)-positive tumours into luminal A (<14%) and luminal B (≥14%) [
12]. This was derived from comparison with gene array data as a prognostic factor [
19]. In 2013, St Gallen revised their threshold to ≥20% for ‘high’ Ki67 status with the option to also use locally specified cut-points [
13]. Recently, at the 2015 St. Gallen Breast Cancer Conference, a median cut-off value of Ki67 within the range of 20–29% to differentiate ‘luminal B-like’ has been recommended [
20]. As shown by Urruticoechea et al. [
21], up to 17 studies that included more than 200 patients displayed statistically significant association between Ki67 and prognosis given that convincing evidence for a biological relationship. However, the cut-offs to discriminate a high from low level of Ki67 varied from 1 to 29%, consequently limiting its clinical utility. Furthermore, during the past decade, multiple research studies have additionally reported the assessment of Ki67 in BC using tissue microarrays (TMA) platform [
14] [
15], although it remains unclear as to their validity and comparison with assessment in whole tissue sections (WTS).
In this study, we aim to assess BC proliferative fraction using Ki67 assessment utilising matched cases prepared as TMA and WTS taking into account the optimal cut-off value for Ki67 assessed on TMA, the common method of proliferation assessment in the research setting on large cohorts. Herein we aimed at determining (1) to what extent Ki67 protein as well as transcriptome levels are matched between TMA and WTS; (2) whether the cut-point used for WTS is reproducible using TMA in different molecular classes. For the latter aim, the association between Ki67 expression cut-points assessed on TMAs and WTS and the standard clinicopathological variables and patient outcome was tested as endpoints.