Molecular subtyping of male breast cancer using alternative definitions and its prognostic impact

Abstract

Background. Male breast cancer (MBC) is an uncommon disease and there is limited information on the prognostic impact of routinely used clinicopathological parameters. Material and methods. In a retrospective setting, we reviewed 197 MBC patients with accessible paraffin-embedded tumor tissue and clinicopathological data. Immunohistochemical (IHC) stainings were performed on tissue microarrays and histological grading on conventional slides. Cox proportional regression models were applied for uni- and multivariate analyses using breast cancer death as the event. Results. Estrogen receptor (ER) and progesterone receptor positivity were demonstrated in 93% and 77% of patients, respectively. Nottingham histologic grade (NHG) III was seen in 41% and HER2 positivity in 11%. Classification into molecular subtypes using IHC markers according to three alternative definitions revealed luminal A and luminal B in 81% vs. 11%; 48% vs. 44% and 41% vs. 42% of cases. Two cases of basal-like were identified, but no cases of HER2-like. Factors associated with an increased risk of breast cancer death were node positivity (HR 4.5; 95% CI 1.8–11.1), tumor size > 20 mm (HR 3.3; 95% CI 1.4–7.9) and ER negativity (HR 10.9; 95% CI 3.2–37.9). No difference in breast cancer death between the luminal subgroups was demonstrated, regardless of definition. Conclusion. MBC tumors were more often of high grade, whereas HER2 overexpression was as frequent as in FBC. Lymph nodes, tumor size and ER status were independent predictors of breast cancer death. The prognostic impact of molecular subtyping in MBC seems to differ from that previously established in FBC.

Male breast cancer (MBC) represents < 1% of all breast cancer. A debate exists deliberating whether there is a difference in prognosis compared with female breast cancer (FBC), with some studies indicating a better outcome, and others a poorer [1,2]. Today, MBC patients receive treatment according to guidelines in FBC and prognostic factors validated in FBC are important for treatment decisions. Established prognostic factors in FBC are: age, tumor size, lymph node status, human epidermal growth factor receptor 2 (HER2), histological grade and Ki-67, whereas hormone receptors are mainly considered to be predictive of endocrine responsiveness [3]. Transcriptional microarray profiling studies based on the “intrinsic” gene set have identified five distinct subtypes in FBC (luminal A/B, HER2-like, basal-like and normal-like) associated with significant differences in relapse-free survival [4]. Immunohistochemical (IHC) markers broadly matching the “intrinsic” gene expression subtypes have been identified and may be more feasible in clinical routine [5–7].

Knowledge about prognostic factors in MBC is limited and contradictory [8–10]. As in FBC, however, lymph node status seems to be an important prognostic factor [8,9]. Previous studies characterizing MBC tumors into “intrinsic” subtypes by using IHC markers as a proxy for gene expression profiles have indicated differences in the distribution of subtypes compared with FBC, but the prognostic impact is not fully understood [11–13]. Interestingly, differences in DNA aberrations and gene expression patterns have been demonstrated compared with FBC, implying that MBC may be a different disease and, hence, that molecular subtyping may also differ [14,15].

In the present study, the aim was to examine MBC tumors with regard to the prevalence and prognostic impact of clinicopathologic parameters and molecular subtypes with commonly used definitions.

Material and methods

Patients

The National Cancer Register was used to identify male patients diagnosed with breast cancer in two regions of Sweden, comprising a population of 1.70 (Lund region) and 1.96 million people (Uppsala-Örebro region). In the Lund region, 118 patients were identified between 1990 and 2005 and, in Uppsala-Örebro, 131 patients between 1990 and 2007. Patient charts were reviewed and paraffin-embedded tumor blocks collected. Tumor blocks from 113 patients in Uppsala-Örebro and from 92 patients in Lund were accessible. Patients with in situ cancer (Uppsala-Örebro n = 4, Lund n = 2) and cases with missing patient charts (Uppsala-Örebro n = 0, Lund n = 2) were excluded. Finally, 109 patients from Uppsala-Örebro and 88 patients from Lund were identified. Information on clinicopathological characteristics, treatment and recurrence status was collected from patients’ charts and updated information about the patients’ vital status and cause of death was retrieved from the National Population Register. The study was approved by the regional ethics committee in Uppsala, Sweden.

Tissue microarray construction and tumor grading

Two 1 mm cores from representative areas in each tumor block were punched and brought into recipient paraffin blocks to construct tissue microarrays (TMA). Four µm thick sections were cut and transferred to glass slides. Hematoxylin and eosin (HE) sections from the tumor blocks were re-evaluated and reclassified according to the Nottingham histological grade (NHG) by a board-certified breast pathologist (ST) using a Zeiss Axioscope 40X objective (and 10 × ocular) with a field of view 0.43 mm.

Immunohistochemistry

IHC stainings were performed on TMA-slides. A fully automated IHC staining machine (BenchMark Ultra, Ventana) was used for deparaffinization, cell-conditioning and staining. For antigen retrieval in ER, PR and cytokeratin (CK)5/6, a TRE-buffer, Cell Conditioner 1 (CC1), was used for 64 minutes (CC1 standard). For Ki-67 and HER 2, CC1 was used for 36 minutes (CC1 mild). Epidermal growth factor receptor (EGFR) was stained using DAKOs EGFRpharmDX. HER2 SISH, a fully automated silver in situ hybridization (SISH) from Ventana, was used for evaluation of HER2 amplification. For antibodies and dilutions used, see Table I.

Table I. Antibodies and dilutions used for IHC stainings.

Marker	Clone	Vendor	Dilution	Preparation
ER	SP1	Ventana	Ventanas Ready to Use	CC1 Standard (64 min)
PR	16	Leica	1:25	CC1 Standard (64 min)
Ki-67	MIB-1	DAKO	1:50	CC1 Mild (36 min)
CK5/6	D5/16B4	DAKO	1:25	CC1 Standard (64 min)
EGFR	1494	DAKO	EGFR pharmDx	Dakos wash buffer
HER2	4B5	Ventana	Ventanas Ready to Use	CC1 Mild (36 min)

CK, cytokeratin; EGFR, epidermal growth factor receptor; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; PR, progesterone receptor.

Evaluation of immunoreactivity scores

Estrogen receptor (ER) and Ki67 were analyzed by one investigator (CA), progesterone receptor (PR) by a second investigator (CN) and EGFR, CK5/6 by a third investigator (IJ). Investigators counted cells manually in high-power fields (40 × objective) using light microscopes. The percentage of positively stained cells was assessed by choosing the high-power field with the largest number of positively stained cells out of the two biopsies and dividing by the entire number of cells from the same high-power field. A minimum of 200 cells per tumor were counted. The method used has been validated in FBC [16].

Tumors were considered positive for ER and PR when more than 10% of nuclei were stained. To classify tumors as EGFR or CK5/6 positive, any degree of staining was acceptable. Ki-67 values ≥ 14% were considered to demonstrate high proliferation according to the guidelines from St Gallen consensus conference [3]. HER2 status determined by IHC staining was evaluated according to standard criteria with IHC staining of 3 + considered as positive and 2 + as equivocal. All cases were further evaluated by silver-enhanced in situ hybridization (SISH) and cases were scored as no amplification if five or fewer copies of the HER2 gene were present per nucleus in more than 50% of the tumor cells and based on the same principle as two-color FISH, i.e. amplification was considered to be present if the ratio between HER2 gene signals and chromosome 17 signals exceeded 2.2 [17]. Tumors were classified as HER2 positive with Herceptest 3 + or if amplified. All HER2/SISH evaluations were performed by a board-certified breast pathologist (R-M A). IHC markers were used to classify tumors into subgroups matching the “intrinsic” gene expression classification. Three different, previously used classifications for luminal tumors were evaluated; the “five biomarker classification” [18] in which the division into luminal tumors was based on HER2 status, “classification/NHG” [6] in which information on tumor grade was added and “classification/Ki-67” in which hormone receptor positive and HER2 negative tumors were divided according to high or low Ki-67 while hormone receptor positive and HER2 positive tumors were classified as a third luminal group (Table II) [7].

Table II. Summary of immunohistochemical criteria for defining breast cancer intrinsic subtypes according to three different classifications.

Classification	Subtype	ER and PR	HER2	NHG	Ki-67	CK5/6 and EGFR
Five biomarker classification	Luminal A	ER and/ or PR positive	negative	NA	NA	any
	Luminal B	ER and/ or PR positive	positive	NA	NA	any
	HER2-like	negative	positive	NA	NA	any
	Core-basal	negative	negative	NA	NA	CK5/6 and/or EGFR positive
	5NP	negative	negative	NA	NA	negative
Classification/NHG^a	Luminal A	ER and/ or PR positive	negative	I-II	NA	NA
	Luminal B	ER and/ or PR positive	HER2 positive and/or NHG III		NA	NA
Classification/Ki-67^a	Luminal A	ER and/ or PR positive	negative	NA	low	NA
	Luminal B	ER and/ or PR positive	negative	NA	high	NA
	Luminal HER2	ER and/ or PR positive	positive	NA	any	NA

CK, cytokeratin; EGFR, epidermal growth factor receptor; ER, estrogen receptor; 5NP, five marker negative phenotype; HER2, human epidermal growth factor receptor 2; NA, not applicable; NHG, Nottingham histological grade; PR, progesterone receptor.

^aOnly receptor-positive subtypes are described. The classification also includes HER2-like subtype (ER and PR negative, HER2 positive) and basal-like (ER, PR negative and HER2 negative).

Statistics

All statistical analyses were performed using the SPSS software package (version 17.0). The defined event was breast cancer death. χ² was used for the comparison of variables between cohorts and Cox proportional regression for the estimation of hazard ratios in uni- and multivariate analyses.

Results

Patient characteristics

Tables III and IV display clinicopathological and treatment characteristics. In patients with pT1 tumors, 37% were lymph node positive, with more than three positive nodes in 15%. ER positive cases (n = 183) were lymph-node positive in 39% (n = 71) and lymph-node negative in 45% (n = 82), in the remaining cases lymph node status was missing. Of those with ER positive, node-positive disease, 34% (n = 24) had died of breast cancer and in patients with ER positive, node-negative disease 6% (n = 5) had died of breast cancer.

Table III. Clinicopathological characteristics.

Variable	n	%
Age (years)
mean 70.1 (range 23–98)
Tumor size (mm)
median 20.0 (range 4–100)
Tumor side
right	85	43
left	108	55
TNM
pT1	91	46
pT2	60	31
pT3	6	3
T4	20	10
pT missing	20	10
pN0	88	45
pN1^a	47	24
pN2^b	17	9
pN3^c	14	7
pN missing	31	15
M0	178	90
M1	9	5
M missing	10	5
Stage
I	53	27
II	63	32
III	46	23
IV	9	5
stage missing	26	13
Histology
ductal	170	86
lobular	6	3
other histology	7	4
data missing	14	7
Grade
I (NHG 3–5)	15	8
II (NHG 6–7)	96	49
III (NHG 8–9)	81	41
not evaluable	5	3
ER
positive (> 10%)	183	93
negative (≤ 10%)	9	5
not evaluable	5	3
PR
positive (> 10%)	152	77
negative (≤ 10%)	38	19
not evaluable	7	4
Ki-67
high (> 14%)	92	47
low (≤ 14%)	99	50
not evaluable	6	3
HER2
IHC
0	122	62
1 +	41	21
2 +	19	10
3 +	3	2
not evaluable	12	6
SISH
not amplified	146	74
amplified	21	11
not evaluable	30	15
positive by IHC or SISH	21	11
EGFR
positive	7	4
negative	186	94
not evaluable	4	2
CK5/6
positive	2	1
negative	189	96
not evaluable	6	3

^a1–3 positive lymph nodes, ^b4–9 positive lymph nodes, ^c ≥ 10 positive lymph nodes.

Table IV. Treatment characteristics.

Treatment	n	%
Surgery
mastectomy	179	91
lumpectomy	12	6
no surgery	2	1
data missing	4	2
Axillary surgery
SN	10	5
axillary clearance	153	78
no axillary surgery	29	15
data missing	5	3
Postoperative radiotherapy
yes	81	41
no	98	50
N/A	18	9
Adjuvant chemotherapy
yes	21	11
no	158	80
N/A	18	9
Adjuvant endocrine therapy
yes	119	60
no	64	33
N/A	14	7

Among patients with ER negative disease (n = 9), three patients had metastatic disease at diagnosis and of the remaining six patients, four had lymph-node positive disease. Seven of nine ER negative patients had died of breast cancer.

In HER2 amplified tumors, IHC scoring 3 + was seen in three (14%), 2 + in 12 (57%) and 0–1 + in 6 (29%) patients.

All patients with HER2 positive breast cancer (n = 21) had localized disease at diagnosis; three patients received adjuvant chemotherapy which was combined with trastuzumab in two cases.

Mean follow-up was 54 months (range 0–180). Loco-regional recurrence was seen in 7% and distant recurrence in 23%. At the time of data collection, 123 of 197 patients had died, 41 (21%) patients of breast cancer and 82 (42%) of other causes. When comparing the two cohorts (Lund and Uppsala-Örebro), no difference regarding stage could be demonstrated. There were no differences in the proportions of patients undergoing adjuvant radiotherapy or endocrine treatment, but a higher percentage of patients in Uppsala-Örebro was treated with adjuvant chemotherapy (18 vs. 5%, p = 0.008).

Regression analyses

In univariate analyses, tumor size > 20 mm, positive nodal status, ER negativity (≤ 10%) and PR negativity (≤ 10%) were significantly associated with breast cancer death, whereas no difference could be demonstrated for age, HER2, grade or Ki-67. The prognostic impact of ER and PR was maintained when the variables were analyzed as continuous variables in univariate analyses (data not shown). The exclusion of HER2-amplified tumors with IHC staining 0–1 + did not change the above result. In a multivariate model including ER, nodal status and tumor size, all parameters were independently associated with breast cancer death with ER being the strongest factor (Table V). In a separate multivariate analysis including PR, nodal status and tumor size, only nodal status had independent prognostic information (data not shown).

Table V. Tumor markers in univariate and multivariate models by Cox Regression. Only variables with significant values in the univariate model were included in the multivariate model.

Variable	HR^a	95% CI	p-value	HR^b	95% CI	p-value
Tumor size; > 20 vs. ≤ 20 mm	2.9	1.5–5.7	< 0.01	3.3	1.4–7.9	< 0.01
Nodal status; pN1–3 vs. pN0	5.4	2.2–13.0	< 0.01	4.5	1.8–11.2	< 0.01
Tumor grade; III vs. I–II	1.5	0.8–2.8	0.16
ER; ≤ 10 vs. > 10%	6.2	2.7–14.2	< 0.01	10.9	3.2–37.9	< 0.01
PR; ≤ 10 vs. > 10%	2.4	1.2–4.7	0.01
Ki-67; ≥ 14 vs. < 14%	1.3	0.7–2.4	0.43
HER2; pos vs. neg	1.4	0.5–3.6	0.48
Age; ≤ 60 vs. > 60	1.1	0.6–2.2	0.69

CI, confidence interval; HR, hazard ratio.

^aUnivariate analysis; ^bMultivariate model including tumor size, nodal status and ER.

Molecular subtype analyses

With the “five biomarker classification”, the following subtypes were identified; luminal A in 81% (n = 160), luminal B in 11% (n = 21), core basal in 1% (n = 2) but no cases of HER2-like or 5NP (five marker negative phenotype). In “classification/ NHG” and “classification/ Ki-67”, hormone receptor positive tumors were almost equally divided into luminal A and luminal B. Seven percent of cases were considered uninterpretable due to loss of tissue cores during sectioning. Basal-like tumors were excluded from regression analysis because of their rarity. No prognostic difference between luminal tumors could be demonstrated by any classification (Table VI).

Table VI. Univariate analysis of different classifications into molecular subtypes. Cox proportional regression was used with breast cancer death as end-point.

		Prevalence
Classification	Subtype	n	%	OR	95% CI	p-value
Five biomarker classification^a	luminal A (Ref)	160	81.2	1.0
	luminal B	21	10.7	1.5	0.6–4.0	0.39
Classification/NHG^a	luminal A (Ref)	94	47.7
	luminal B	86	43.7	1.2	0.6–2.3	0.67
Classification/Ki-67^a	luminal A (Ref)	80	40.6	1.0
	luminal B	82	41.6	1.4	0.7–2.9	0.30
	luminal HER2	19	9.6	1.1	0.3–3.8	0.87

^aOnly receptor-positive subtypes are included in the regression analysis. In all classifications two cases (1%) with core-basal/basal-like subtype were identified but due to too small number of cases, these were not included in the regression analysis.

Discussion

The aim of this study was to further characterize MBC by describing the prevalence and prognostic value of clinicopathological factors and molecular subtypes often applied to FBC using IHC parameters.

We found that tumor size and nodal status were independent prognostic factors, which is in accordance with previously published results [8,9]. Furthermore, it seems that the prevalence of positive lymph-nodes is also high in smaller tumors (≤ 20 mm), which is in line with previous data [8].

ER status is firstly considered to be a predictive and not a prognostic factor in FBC. In our study, ER-negativity (≤ 10%) was a poor prognostic factor. To our knowledge this has previously not been shown, even though it has been indicated in a previous study [19]. A majority of patients received adjuvant endocrine treatment which could influence the results, but the prognostic impact of ER remained when adjustment for endocrine treatment were made (data not shown). Furthermore, different cut-off points defining ER negativity was analyzed and we found that also a cut-off of ER < 80% was associated with an increased risk of breast cancer death (HR 2.2; 95% CI 1.0–4.7, p = 0.045) in a multivariate analysis including tumor size and nodal status. Hence, the optimal cut-off point may be different in MBC. However, since only 10 cases had an ER expression between 10% and 80%, our material is too small to define the most optimal cut-off. More than 90% of MBC patients were ER-positive and the prevalence of lymph node positivity was high in comparison with FBC [20], which was also seen in small (≤ 20 mm) hormone receptor positive tumors. Consequently, ER-positive disease in men cannot automatically be considered indolent disease and it is important to take other prognostic factors into account.

The prevalence of PR positivity was 77%, which is higher compared with FBC [20]. PR status is an established prognostic factor in FBC [3], but data from MBC are contradictory [9,19]. We found that PR status did not have any prognostic value in the multivariate model, suggesting that PR status is of inferior prognostic value compared with ER status.

HER2-positivity has been observed in 11–25% of FBC and entails a poor prognosis, at least if anti-HER2 treatment is not administered [20,21]. In the present study, 11% of tumors were HER2 positive, but this was not associated with a poorer outcome, even though a majority of patients had not received adjuvant chemotherapy or trastuzumab. The latter observation needs be interpreted with caution due to the moderate power the study, but could indicate that the prognostic impact of HER2 positivity is different in MBC. Previous studies have demonstrated HER2 amplified tumors in 0–11% of patients but the prognostic impact was not evaluated due to lack of outcome data [22,23]. A study comparing gene expression profiles in MBC and FBC has suggested a reduced relevance of HER2 in MBC [24]. In contrast, Wang-Rodriguez et al. demonstrated a shorter disease-free survival in patients HER2-positive by IHC [10]. Molecular data in FBC suggest heterogeneity in HER2-positive tumors and gene expression profiles enabling stratification into good and poor outcome groups have been identified [25]. It would be of interest to evaluate this HER2 heterogeneity in MBC. A thought-provoking observation from our study was that only three of 21 amplified tumors were IHC 3 + and the majority of tumors were IHC 2 + (n = 12) or 0/1 + (n = 6). This differs from FBC, where a majority of HER2-amplified tumors are IHC 3 + and only a small percentage is amplified if IHC < 3 +. It has been hypothesized that tumors which are negative for HER2 by IHC, but amplified by in situ hybridization could be falsely IHC-negative due to epitope loss during fixation [26]. An alternative hypothesis could be that protein expression is not as highly correlated with gene amplification in MBC. Hence, the prognostic value of HER2 in MBC is still unclear and additional studies are warranted.

Tumor grade and Ki-67 are correlated to proliferation and a more aggressive tumor behavior in FBC. In our study, a high percentage of tumors were classified as high grade (NHG III) and/ or high Ki-67, but these parameters were not predictive of poorer breast cancer survival. Our result on Ki-67 differs from previous smaller studies that indicate a more aggressive disease in MBC patients with highly proliferating tumors [27,28].

Previous studies on the prognostic impact of histologic grade in MBC are contradictory [8,9]. However, it is known that histologic grading is difficult and inter-observer discrepancies are a problem. Possibly, consistent histologic grading is even more difficult in MBC since there is lack of normal breast tissue which is a prerequisite for assessing nuclear atypia correctly.

Gene expression profiling is considered to describe tumor heterogeneity and prognosis more accurately. IHC markers are often used as a surrogate for gene expression subtyping, and different combinations with similar prognostic information have been validated in FBC [5–7]. The different subtypes in FBC using IHC are luminal A in 44–66%, luminal B or luminal/HER2 in 6–19%, basal-like in 10–17% and HER2-like in 8–10% of cases. In MBC, the corresponding prevalences are luminal A in 75–98%, luminal B in 0–21%, basal-like in 0–4%, but no cases of HER2-like, keeping in mind the small sample sizes in these studies [11–13]. Shaaban et al. recently published a paper using HER2 status to separate the luminal subtypes. However, international guidelines on the evaluation of HER2 status were not adhered to: only IHC 3 + was considered HER2 positive and no FISH testing was performed. None of the tumors had IHC 3 + and hence, no tumors were classified as luminal B [13]. Male patients were compared with matched female patients and no differences in overall survival for luminal A tumors was shown between genders, but the authors also discussed tumor biological differences indicating that luminal A in males and females are not comparable. Kornegoor et al. used another definition by which hormone receptor positive tumors with high proliferation and/or HER2 positivity were classified as luminal B, identifying luminal A in 75% and luminal B in 21% of cases. In accordance with our study, a few cases with a basal-like subtype were identified, but no cases with HER2-like [12]. The prognostic relevance of the different subtypes was not evaluated due to lack of outcome data. In our study, we evaluated three models for classification into subtypes whereof one (“classification/ki-67”) is recommended by the St Gallen consensus conference [3]. Two tumors with core basal-like subtype were identified and the remaining cases were classified into the luminal subgroups. The prognostic impact of core basal-like tumors was not evaluated due to the small number of cases. In the “five biomarker classification”, in which the differentiation between the luminal subtypes was determined by HER2 status, no difference in the risk of breast cancer death between luminal groups could be demonstrated. This is in line with the results on HER2 in this material, which did not demonstrate any prognostic value. However, gene expression analysis in FBC has shown that luminal B tumors overexpress HER2 in 15–25% of cases and the remaining subset is classified as luminal B mainly due to the overexpression of proliferation genes, thus luminal/HER2 + does not correspond entirely to the luminal B gene expression profile [4]. In consequence, two alternative definitions, “classification/NHG” and “classification/Ki-67”, were evaluated but neither of them could reveal any difference in breast cancer death between luminal A and B. Interestingly, when the two latter classifications are applied, luminal A and B are almost equally common in MBC.

Non-significant results in this study should be interpreted with caution due to the moderate power of the study, but our results may imply that the molecular background for tumor evolution and behavior in MBC is different when compared with FBC. This hypothesis is supported by studies on gene expression and DNA aberrations showing differences between MBC and FBC. Johansson et al. compared global gene expression analyses in MBC tumors with FBC tumors and identified two novel subgroups in MBC tumors; luminal M1 and luminal M2. When evaluating the degree of similarity to the “intrinsic” subtypes in FBC tumors, more than half of the MBC tumors were left unclassified. In the MBC tumors, genes related to the immune system were of importance in distinguishing between the subgroups. A majority of patients were classified as luminal M1, a more aggressive phenotype with a poorer prognosis. Interestingly, even though the majority of MBC tumors were ER + by IHC, there was a significant difference in ER-related gene signaling between the groups, with luminal M1 displaying an inferior correlation to ER-signaling. This is different from results in FBC where only ER negative tumors are identified with similarly reduced ER-signaling scores [14]. The same MBC tumor material has also been evaluated on the DNA level by comparative genomic hybridization, revealing differences compared with FBC. Two subgroups were identified, one of which has not previously been described in FBC, and similar to the results from the gene expression study, a majority of patients were classified into the subgroup with a more aggressive tumor behavior [15]. All in all, data from the referred studies suggest that MBC displays differences in tumor biology compared with FBC, which in turn implies that the classification into molecular subtypes described in FBC may be of limited relevance. New classifications applicable in MBC need to be identified.

We are aware that our study has limitations, most importantly because of the retrospective setting and the relatively small patient material, and results should be interpreted with caution. However, compared with many other studies on MBC and additionally, considering the rareness of MBC, this is a comprehensive study. The main advantages include reliable outcome data and the reclassification of all tumors according to new IHC stainings and NHG.

In conclusion, this study demonstrates that ER-negativity, tumor size > 20 mm and positive lymph nodes are independent risk factors for breast cancer death in MBC. Age, HER2, grade and Ki-67 did not demonstrate any prognostic impact, but the moderate power of this study must be taken into account. Of interest, the classification using IHC markers into molecular subtypes does not seem to provide similar prognostic information as in FBC and other prognostic profiles need to be identified.

Acknowledgements

We should like to thank all involved Pathology Departments for providing tissue for the study. This study was supported by grants from the Regional Research Foundation in Uppsala-Örebro, the Lion´s Cancer Foundation, University Hospital, Uppsala and Västmanlands Research Foundation. We should also like to acknowledge Associate Professor Henry Letocha for linguistic support.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.