Prognostic relevance of Ki67 expression in primary male breast cancer: determination of cut-off points by different evaluation methods and statistical examinations

Abstract

Purpose

1% of all breast cancer cases occur in men. There are significant differences regarding clinical behaviour and genetic profiles between female (FBC) and male breast cancer (MBC). Parameters for decision-making on treatment and prognosis are derived from FBC. Ki67 has a high value as a prognostic and predictive factor in FBC, but accurate Ki67 cut-off points for MBC are missing. In this study, we aimed to evaluate adequate examination methods and reliable cut-off points for Ki67 to assess the highest prognostic value for patient’s overall survival (OS).

Methods

In this multicentric retrospective study, histological specimens were obtained from 104 male patients who were diagnosed and treated for primary invasive breast cancer. We applied three methods of Ki67 analysis: Tumor average scoring (TA), tumor border scoring (TB) and hot-spot scoring (HS). Calculated Ki67 cut-off points for each method were assessed as a threshold for patients’ overall survival (OS).

Results

Ki67 cut-off points were 13.5 for the TA group, 22.5 for the HS group and 17.5 for the TB group. Only Ki67 TA cut-off calculations demonstrated statistical significance (p = 0.04). Ki67 expression analysis of TA showed that more than 90% of patients with low Ki67 levels (< 13.5) were alive after 5-year follow-up.

Conclusion

Our findings demonstrate that determination of Ki67 expression in TA is the most reliable to define a cut-off point with high prognostic value. A Ki67 cut-off point of 13.5 shows highest statistical power to define luminal A subgroup and OS.

Introduction

Only 1% of all breast cancer cases occur in men (Siegel et al. 2020). Recently increasing incidence was noted. According to the Association of Population-based Cancer Registries in Germany (GEKID) there were 437 (0.9%) new registered cases in 2003 and 710 new cases (1.1%) in 2016 (Krebsregisterdaten 2020). Unlike female breast cancer (FBC), male breast cancer (MBC) is an under researched topic. Only few studies focus on male patients leading to a lack of data regarding risk factors, specific tumor biology, prognostic values and treatment options for male patients (Leon-Ferre et al. 2018).

Even though FBC and MBC share similarities, there are significant differences regarding risk factors and prognosis (Cardoso et al. 2018). In FBC, prognostic values are well defined. Histopathological parameters including hormone receptor (HR) status, HER2 expression, Ki67 or BRCA status do not only have a prognostic value, they are also critical determining relevant therapy regimes (Cheang et al. 2009).

Unlike FBC, MBC histopathologically appears to be more homogenous with only minor differences in HR status and HER2 distribution. The vast majority of MBC tumors are highly HR positive and HER2 negative (Nilsson et al. 2013). Therefore, luminal A and B subtypes are predominant (Cardoso et al. 2018).

Clearly there is the necessity to have parameters that can distinguish between subtypes for MBC, to make an adequate recommendation regarding therapy options. At St. Gallen’s International consensus in 2013 experts agreed on Ki67 discriminating between luminal A and luminal B tumors in FBC (Goldhirsch et al. 2013). Also, research was able to prove that beside its prognostic value Ki67 has predictive value to forecast female patient’s response to chemotherapy (von Minckwitz et al. 2013). Even though Ki67 is an established biomarker for FBC, no standardized evaluation methods for male patients are established.

Firstly, there is no consensus on which area of the tumor is most likely to evaluate Ki67 and therefore allowing a more informed decision-making process for patients’ diagnosis and potential treatment (Dowsett et al. 2011). This unresolved topic prompted us to examine whether it makes a difference to determine Ki67 at the tumor average (TA), at Ki67 tumor hot spots (HS) or at the tumor border (TB). Analysing TA is the most straightforward approach. It is, nevertheless, conceivable that at TB, the area where tumor cells infiltrate healthy tissue and the tumor HS represent key pathophysiological compartments that carry information about the tumors’ aggressiveness.

Secondly, Ki67 cut-off levels remain uncertain in FBC as well as in MBC. In FBC, Denkert et al. was able to show that a cut-off level of 15% has a predictive value (Denkert et al. 2013) whereas Bustreo et al. presented a higher cut-off level at 20% (Bustreo et al. 2016).

The goal of this study was to clarify which area of the tumor carries the strongest value in calculating Ki67 cut-off points for MBC patients. Furthermore, we aimed to define reliable cut-off points and to investigate which method has the highest prognostic value for patient’s overall survival (OS).

Material and methods

Patient selection

In this multicentric retrospective study, histological specimens were obtained from 104 male patients who were diagnosed and treated for primary invasive breast cancer from 1995 to 2018 at Breast Cancer Units in Bergisch Gladbach, Chemnitz and Zwickau, Germany. The patient data, histopathological findings as well as treatment reports and follow-up data were collected with the approval of the appropriate institutional review boards from the hospital archives. The study was approved by the institutional local ethics committee.

Histopathological analysis

Tumor specimens were fixed in formalin and embedded in paraffin. Immunohistological staining (IHC) was performed using an automatic immunostaining system (Ventana Benchmark GX, Roche Diagnostics, Mannheim, Germany) according to the manufacturer’s instruction. All specimens were counterstained with hematoxylin. Pathohistological evaluation was performed within 2 weeks after immunohistological staining.

The slides were microscopically reviewed by two examiners independently (M.E.L. and S.B.). Histopathological carcinoma classification was recorded in accordance with current World Health Organization staging (Hoon Tan et al. 2020). Pathological staging was classified according to the International Union Against Cancer (Wittekind 2012). HR status was evaluated using the Remmele and Stegner score (Remmele and Stegner 1987), which is an immune reactivity score (IRS). It quantifies the HR by multiplying their staining intensity with the percentage of stained cells, resulting in a negative (0–2), weak positive (3–4), moderate (6–8) or high (9–12) IRS. HER2 protein expression was determined with HercepTest™ (DAKO Corp., Hamburg, Germany). HER2 membrane staining intensity and pattern were analyzed according to FDA-approved criteria: 3+ immune staining was considered positive and 0/1+ immune staining was defined to be negative. 2+ staining was categorized as equivocal. In this case, HER2 positivity was defined by its gene amplification status (FISH) using standardized protocols (Piccart-Gebhart et al. 2005). For Ki67 immunostaining, clone 30-9 (Roche Diagnostics) was used. Ki67 score is the percentage of positively nuclear stained cells among the total number of malignant cells in the investigated area. The intensity of staining was not decisive. We applied three methods of Ki67 analysis: Tumor average scoring (TA), tumor border scoring (TB) and hot-spot scoring (HS). For all cases, the slides were screened at a high-power field (40× objective). Evaluation of Ki67 expression was applied according to scoring protocol of the Ki67-QC international working group (Nielsen et al. 2020). Briefly, evaluation was performed in four steps. First, examination of the whole section using low-power magnification and estimation of the percentage of invasive tumor and content of Ki67 positive cells. Secondly, selection of the fields that can be used for high-power evaluation. Representative tumor areas were selected for each evaluation method. Thirdly, count of invasive tumor nuclei from the top of the selected scoring field until either 100 invasive tumor nuclei have been counted or all invasive tumor nuclei in the entire scoring field have been counted. Counting procedure was repeated until ten different tumor areas were scored. Finally, the Ki67 score was calculated. For Ki67 tumor border scoring, ten fields from the tumor border were selected. The hot spot method requires to visually select 1 high-power field with the highest apparent staining rate and, within that area only, count up to 500 invasive tumor nuclei in a ‘typewriter’ pattern (Leung et al. 2019).

Statistical analysis

The analysis of standard deviation, variance and range are among the measures of dispersion and were evaluated. Using the F-test, we have ensured that our counted values correspond to a statistically correct homogenous series of measurements (Fig. 1).

Fig. 1.

Frequency distribution of Ki67 values for each Ki67 evaluation method

To determine Ki67 subgroups (high versus low) within our cohort, we aimed to define a Ki67 cut-off value based on OS data. Cut-off values for Ki67 were assessed as follows: receiver operating characteristic (ROC) curve analysis and area under the curve (AUC) calculations for each Ki67 analyzing method were applied. Subsequently the calculated Ki67 cut-off point for each method was assessed as a threshold for best sensitivity and 1-specifity regarding patients’ OS and its’ statistical significance. For each subgroup, overall survival benefit was calculated according to log-rank, Kaplan–Meier-Curve (KMC) survival analysis and Cox regression analysis. Odds ratio and 95% confidence intervals were calculated. We tested for significance using Wilcoxon-Mann–Whitney-test (Utest) when comparing two groups and Kruskal–Wallis-test (H-test) when comparing more than two groups. All results were considered significant at p values < 0.05. All analysis was performed using SPSS software package version 26 (SPSS Inc., Chicago, IL, USA, and Microsoft^® Excel^® 2010, version 14 for Windows (Microsoft Corp., Redmond, WA, USA).

Results

Data as well as tumor specimens were available from 104 primary MBC patients. Patient age ranged from 34 to 88 years (median 66 years). Median follow-up time was 69 months (1–169 months), median OS 48.5 months (range 1–207 months). 42.3% (n = 44) of patients deceased.

Due to the retrospective characteristic of this study, data regarding treatment was incomplete. In 92.3% (n = 96) of cases patients underwent modified radical mastectomy, 6.7% (n = 7) of patients received a breast preserving surgery. 51.9% (n = 54) were treated adjuvant with tamoxifen. Seven patients (6.7%) received aromatase inhibitors. Adjuvant chemotherapy was administered in 34 cases (32.7%): 14 received anthracycline-containing regimens, 12 patients cyclophosphamide/methotrexate/fluorouracil (CMF) and 8 patients received taxanes in combination with anthracyclines.

Most tumors were non-specific-type carcinoma (98.0%, n = 102) and had intermediate differentiation (G2, 76.0%, n = 79). 9.7% (n = 10) were well (G1) and 1 4.4% (n = 15) poorly differentiated. 62.5% (n = 65) of patients were diagnosed at pT1 (n = 36, 34.6%) or pT2 (n = 29, 27.9%) stage and more than one third were pT4 staged (34.4%, n = 34). 38.5% (n = 40) of patients already had axillar lymph node metastasis at diagnosis.

In our cohort, 102 (98.1%) tumors expressed estrogen receptors (ER) (high intensity 93.3%, n = 97; intermediate intensity1.9%, n = 2; low intensity 2.9%, n = 3) and 96 (94.2%) progesterone receptors (PR) (high intensity 60.6% n = 63; intermediate intensity 17.3%, n = 18; low intensity16.3%, n = 17). Only four patients (3.8%) showed HER2 positivity. All of those were HR positive. Three patients (2.9%) were triple negative.

Ki67 immunostaining was successful in 103 out of 104 tumors. Due to high standard deviation in HS (23.7) and TB (13.4) compared to TA (9.91) as well as the F-test results, it is obvious that the descriptive statistic recommends TA analysis for the evaluation of Ki67 values. In our study, the F-test revealed that Ki67 data examined by the TA method represent the most eligible measurement series.

To discriminate between high and low Ki67 expression, cut-off points for all methods of analysis were based on patient’s OS as a result of threshold calculations (Fig. 2). Based on OS data, calculated Ki67 cut-off points were 13.5 for the TA group, 22.5 for the HS group and 17.5 for the TB group (Table 1). Only Ki67 expression in TA analysis demonstrated statistical significance (p = 0.04) regarding ROC correlations (Fig. 2). Ki67 expression was correlated with different tumorbiological parameters (Table 3). High Ki67 expression showed positive correlation between tumor grading and all three Ki67 evaluations (Table 1). However, there was no correlation regarding tumor size and nodal stage. Tumor stage and nodal status were correlated positive between overall survival (tumor stage: p < 0.001 and nodal status: p < 0.05).

Fig. 2.

Ki67 cut-off calculations using ROC-curves and AUC correlations for three different Ki67 examinations

Table 1.

Ki67 cut-off calculation for each evaluation method

Ki67 cut-off	Sensitivity	1 − specificity
(a) TA Ki67 expression
1.00	1.00	1.00
2.00	0.98	0.98
4.00	0.96	0.98
7.00	0.93	0.81
11.00	0.74	0.52
13a	0.74	0.48
17.00	0.46	0.33
22.00	0.24	0.16
27.00	0.13	0.16
32.00	0.07	0.07
37.00	0.04	0.07
45.00	0.04	0.03
65.00	0.00	0.02
81.00	0.00	0.00
(b) HS Ki67 expression
1.00	1.00	1.00
2.00	1.00	0.98
4.00	0.98	0.98
7.00	0.98	0.97
12.00	0.91	0.79
17.00	0.80	0.64
22a	0.60	0.43
27.00	0.51	0.34
32.00	0.36	0.21
37.00	0.27	0.19
45.00	0.16	0.19
55.00	0.02	0.07
65.00	0.02	0.05
75.00	0.00	3.00
85.00	0.00	0.02
91,00	0,00	0.00
(c) TB Ki67 expression
1.00	1.00	1.00
2.00	1.00	0.98
4.00	0.98	0.98
7.50	0.96	0.91
12.50	0.80	0.71
17.5a	0.71	0.48
22.50	0.56	0.33
27.50	0.38	0.24
32.50	0.22	0.16
37.50	0.11	0.14
45.00	0.07	0.07
55.00	0.02	0.03
65.00	0.02	0.02
80.00	0.00	0.02
9.00	0.00	0.00

TA tumor average, HS hot spot, TB tumor border

^aCalculated cut-off point

Table 3.

Prognostic factors in primary MBC

	Own data		Literature (Giordano et al. 2004; Cardoso et al. 2018)
	Median OS (month)	Hazard ratio (95% CI)	Median OS (month)	Hazard ratio (95% CI)
pT1	145.00		104.00
pT2	85.00	1.871 (0.81–4.28)	57.00	1.46 (1.38–1.55)
pT3–4	45.00	1.746 (1.34–2.26)	35.00	1.26 (1.12–1.14)
pN0	145.00		126.00
pN+	70.00	1.83 (1.12–3.46)	73.00	1.41 (1.28–1.57)
		95% CI		95% CI
Luminal A	160.00	138.38–181.11	114.00	85.2–115.2
Luminal B (HER2 neg.)	77.00	51.10–102.89	80.00	69.6–94.8
Lumunal B (HER2 pos.)	76.00	53.48–98.51	120.00	68.6-n.e
Grade 1	112.00	45.2–179.9	153.00	123.6–184.78
Grade 2	102.00	83.71–121.5	123.00	100.8–135.6
Grade 3	108.00	87.25–119.75	108.00	78.0–154.8

Using different cut-off points and evaluation methods, KMC revealed significant overall survival benefit for low Ki67 expression (TA: p < 0.001, TB: p < 0.001, HS: p = 0.005, Fig. 3a–c). It is of interest that Ki67 expression analysis of TA and TB showed that more than 90% of all patients with low Ki67 levels were alive after a 5-year follow-up. In multivariate analysis, there was only a positive correlation between tumor stage (p < 0.001) and Ki67 with overall survival (TA: p = 0.002; TB: p = 0.004, HS: p = 0.03). In luminal A subtype defined by Ki67 TA positivity and/or HER2 positive hormone receptor positive tumors, a 5-year OS of 91.8% could be demonstrated.

Fig. 3.

Kaplan–Meier-curves for overall survival regarding three Ki67 evaluations: Ki67 expression in tumor average (a), Ki67 expression at tumor border (b) and Ki67 expression in tumor hot spots (c)

Discussion

In clinical practice, MBC is often perceived and treated as a gender variant of breast cancer in the female neoplasia context (Losurdo et al. 2017). Therefore, adjuvant treatment recommendations are mostly based on guidelines for its female counterpart (Hassett et al. 2020). However, there are significant differences regarding clinical behaviour and genetic profiles (Masci et al. 2015; Rudlowski et al. 2006). Due to its rare occurrence, it is mandatory to base treatment decisions on male specific tumor biology. Molecular and immunohistochemical approaches have been initiated to distinguish MBC subtypes (Vermeulen et al. 2017).

In accordance with the literature, our data demonstrate that MBC is dominated by luminal A and B subtype tumors (Cardoso et al. 2018; Jylling et al. 2020; Nilsson et al. 2013). Luminal B subtype in MBC is predominantly defined by high Ki67 expression. In our study, HER2 positivity was less than 4% which is in line with recent published data (Rudlowski et al. 2004). This is of importance for treatment decisions in the adjuvant setting. Actually, recommendations for Ki67 staining and analysis in MBC are derived from guidelines for FBC (Dowsett et al. 2011). This includes Ki67 cut-off points (Coates et al. 2015). To avoid high-cost approaches like multigene expression panels for evaluating treatment options, it is essential to determine a cut-off value to differentiate between luminal A-like and B-like tumors (Maranta et al. 2020). In breast cancer, the most widely practiced method for comparing proliferation between tumor samples involves the immunohistochemical assessment of Ki67 antigen. Ki67 is expressed in all phases of the cell cycle other than G0 phase. This is an important advantage compared to other alternatives like counting mitotic numbers or calculating the mitotic index (International Ki-67 in Breast Cancer Working Group 2021).

The aim of this study was to define a specific Ki67 cut-off point for primary MBC as well as to examine its prognostic value for clinical practice (Table 2).

Table 2.

Ki67 evaluation methods and corresponding cut-off points in correlation with tumor characteristics

	Tumor size (n = 84)			Lymph nodes (n = 84)			UICC-stage (n  = 86)			Grading (n = 104)
	p	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p	OR	95% CI
Ki67 average (< 13.5 > 13.5)	0.03*	1.7	1.03–2.9	0.59	0.99	0.99–1.00	0.77	0.9	0.57–1.52	0.02	3.8	1.3–11.7
Ki67 border (< 17.5 > 17.5)	0.13	1.4	0.89–2.4	0.67	0.9	0.99–1.00	0.85	0.9	0.58–0.155	0.02	3.7	1.28–11.5
Ki67 hot spots (< 22.5 > 22.5)	0.94	1.4	0.93–2.35	0.86	1.00	0.99–1.00	0.96	0.9	0.59–1.55	0.03	3.1	1.11–8.74

Therefore, we analysed Ki67 expression in different areas within the tumor and calculated cut-off points for each area regarding OS. We aim to reflect varying tumor biology and clonal differences by examination of the average Ki67 expression in the tumor, Ki67 expression in tumor hot spot areas and at tumor borders. Our findings demonstrate that determination of Ki67 expression in TA is the most reliable to define a cut-off point with high prognostic value regarding MBC patient’s outcome. A Ki67 cut-off point of 13.5 shows highest prognostic power for overall survival which correlates to examinations in FBC (Coates et al. 2015).

Despite of its accordance with established cut-off points known from FBC our results demonstrate that Ki67 positivity in MBC needs to be calculated accurate and concise. Therefore, we have carefully used a standardized histopathological analysis paradigm combined with a well characterised patient group as well as an adequate follow-up. By focussing on three different loci for Ki67 analysis, we aimed to estimate the role of Ki67 in primary MBC. Our data indicate a close relationship between Ki67 positivity and a more advanced tumor grading. Histopathological grading is a prognostic factor for FBC (Rakha et al. 2010). However, recent studies did not identify any association with grading and prognosis in MBC (Vermeulen et al. 2017). Cardoso et al. (2018) showed no statistical correlation between grading, Ki67 expression and OS in a large retrospective cohort of MBC patients. The prognostic factors in primary MBC calculated in our study were compared with data from the literature and summarized in Table 3.

In this study, we found a significant association with Ki67 positivity (≥ 13.5 TA) and overall survival by multivariate analysis. In contrast to Cardoso et al., our patient group revealed a higher percentage of Ki67 positivity associated with more advanced tumor grading (Cardoso et al. 2018). Our data indicate that it is important to calculate an individual Ki67 cut-off point for each patient group and to use central pathology.

For clinical routine, it appears of substantial importance to determine subtypes in MBC by a simple and standardized method to decide on adjuvant treatment options. Our study provides data how to use Ki67 expression analysis for clinical use. If using Ki67 TA method, evaluations can be performed both in core biopsies and large tumor samples.

A limitation of our present study may be given by the fact that the data are based on a retrospective study with a limited sample size. Furthermore, we did not use complementary molecular subtyping. In this context, it may be worth mentioning that the application of molecular testing for example by Oncotype DX™ is an useful tool for primary MBC patients (Williams et al. 2020) especially if Ki67 expression hits the upper margins or shows a discrepancy to other clinicopathological parameters within the luminal B setting. It will be a future task to cross-validate the present results in (a) a prospective study setting and (b) with parallel molecular profiling.

Conclusions

Our present study illustrates that a precise evaluation of Ki67 is important as a prognostic marker for patient overall survival in MBC.

Funding

This work was supported by the German Cancer Foundation (Project Number: 70-3157).

Availability of data and material

Data and material are available on request.

Declarations

Conflict of interest

All authors disclose any financial and personal relationships with other people or organisations that could inappropriately influence (bias) our work. They have no conflict of interest.

Ethical approval

The patient data, histopathological findings as well as treatment reports and follow-up data were collected with the approval of the appropriate institutional review boards from the hospital archives. The study was approved by the institutional local ethics committee. Data were blinded for statistical evaluations.