Breast conserving surgery versus mastectomy: overall and relative survival—a population based study by the Danish Breast Cancer Cooperative Group (DBCG)

Abstract

Background: Observational studies have pointed at a better survival after breast conserving surgery (BCS) compared with mastectomy. The aim of the present study was to evaluate whether this remains true when more extensive tumor characteristics and treatment data were included.

Methods: The cohort included patients registered after primary surgery for early invasive breast cancer in the database of the Danish Breast Cancer Cooperative Group, in the period 1995–2012. The cohort was divided into three groups: (i) patients who primarily had a mastectomy, (ii) patients treated by BCS, and (iii) patients who primarily had BCS and then mastectomy [intention to treat (ITT) by BCS]. The association between overall mortality and standard mortality ratio (SMR) and risk factors was analyzed in univariate and multivariate Poisson regression models.

Results: A total of 58,331 patients were included: 27,143 in the mastectomy group, 26,958 in the BCS group, and 4230 in the BCS-ITT group. After adjusting for patient and treatment characteristics, the relative risk (RR) was 1.20 (95% CI: 1.15–1.25) after mastectomy and 1.08 (95% CI: 1.01–1.15) after BCS first and then mastectomy, as compared to BCS. Statistically significant interactions were not observed for age, period of treatment, and nodal status, but patients with Charlson’s Comorbidity Index (CCI) score 2+ had no increased mortality after mastectomy, as opposed to patients with CCI 0–1. Loco-regional radiation therapy (RT) in node positive patients did not reduce the increased risk associated with mastectomy [RR = 1.28 (95% CI 1.19–1.38)].

Conclusion: Patients assigned to BCS have a better survival than patients assigned to mastectomy. Residual confounding after adjustment for registered characteristics presumably explained the different outcomes, thus consistent with selection bias. Diversities in RT did not appear to explain the observed difference in survival after BCS and mastectomy.

Introduction

The randomized trials conducted in the 1980-ies showed an equal outcome after breast conserving surgery (BCS) combined with radiation therapy (RT) and mastectomy [1–3]. Since then BCS has become the preferred method.

Recent results from observational studies [4–7] has pointed at a better survival after BCS compared with mastectomy. Accordingly, it has been argued that it may no longer be appropriate to offer women suitable for BCS the choice of mastectomy [8]. All of these studies have limitations, as they all more or less lacked significant prognostic variables characterizing tumor biology and treatment. The present material from the Danish Breast Cancer Cooperative Group (DBCG) was from a population-based cohort including variables comprehensively describing patient and tumor characteristics and the given treatment.

The aim of the present study is on a large population-based material to compare the outcome for Danish breast cancer patients treated by mastectomy and BCS. The comparison takes the available prognostic variables into a multivariate analysis and includes an intention to treat analysis.

Material and methods

Since the establishment in 1977, DBCG has provided standard diagnostic treatment algorithms for early breast cancer. Data on diagnosis, therapy, and follow-up on newly diagnosed breast cancer patients have been collected prospectively in the DBCG registry by the use of standardized forms [9].

The cohort used in this study, included patients who had surgery for invasive breast cancer in the period 1995–2012. Patients preoperatively diagnosed with distant metastatic, locally advanced breast cancer, or synchronous bilateral breast cancers were not included in the cohort. Furthermore, patients who had received neo-adjuvant therapy or patients with unknown nodal status were not included in the cohort.

The cohort was divided into three groups: (i) patients with assigned mastectomy as the first surgical procedure, (ii) patients with breast conservation as the definitive surgical procedure, and (iii) patients assigned with an initial BCS followed by mastectomy as a second or later procedure. The definitive surgical procedure is detailed registered in DBCG and we retrieved information on prior surgery from the Danish National Patients Registry and on prior examination of tissue from the Danish National Pathology Registry. Vital status and emigration was retrieved from the Danish Civil Registration System and mortality in the general Danish female population by age and calendar period was obtained from Statistics Denmark.

Comorbidity was described by Charlson’s Comorbidity Index (CCI) [10], based on data from hospital contacts using International Classification of Disease (ICD-8 and ICD-10) data up to 10 years prior to the breast cancer diagnosis date.

Statistics

Primary and secondary end-points were standardized mortality ratios (SMR) and overall survival. Overall survival was defined as time from surgery until death from any cause.

Estimated median follow-up time was quantified in term of Kaplan–Meier estimates [11]. SMR, was calculated as the ratio of the observed number of death among patients to the expected number of deaths. The expected number of deaths was estimated by multiplying the survival time accrued from the study cohort by the mortality rates of the general population of women matched by age (1 year groups) and calendar period (1 year groups). Estimates of SMR greater than unity indicates that breast cancer patients have a higher mortality rate than women in the general population. Kaplan–Meier estimates were used to estimate overall survival.

The association between mortality and risk factors was analyzed in univariate and multivariate Poisson regression models. In the multivariate analysis, we adjusted for: age at operation categorized in year intervals (18–44, 45–54, 55–64, 65–74, 75+), year of operation divided into the three periods (1995–2001, 2002–2007, 2008–2012), comorbidity described by CCI categories (0, 1, 2, 3+), tumor size categorized in intervals (0–10 mm, 11–20 mm, 21–30 mm, 31–50 mm, 51 + mm), number of positive lymph nodes in four nodal categories (0, 1–3, 4–9, ≥10), histological type (ductal, lobular, other), malignancy grade (I, II, III), percent of ER-positive tumor cells [negative (0–9%), positive (10–100%)], lymphovascular invasion (LVI) with two levels (yes versus no), multifocality with three levels (multifocal, unifocal, and unknown), chemotherapy, RT and endocrine therapy each with three levels (yes, no and unknown). Unknowns were included in most appropriate categories, unless otherwise specified. Furthermore, the results of the full multivariate model were compared with a corresponding model, where age, tumor size and number of positive lymph nodes were included as continuous variables. The results of these analyses did not alter the overall results noticeably. Likewise was the results not altered by including HER2 in the multivariate models. Hence, HER2 was not included and all variables were treated as categorical in the final multivariate model. Interactions between operation and age, year of operation, CCI, nodal status, and systemic therapy were investigated in separate models. An intention to treat analysis as well as an analysis based on the outcome operation was included.

p values <.05 were considered significant, and p values were two-tailed. The statistical analyses were performed in R version 3.2.3 [12].

Results

The population of surgical treated female breast cancer patients included in all 68,842 patients (Supplementary Figure S1). After exclusion of patients with locally advanced, distant metastatic, and bilateral breast cancer, and furthermore patients who were not surgical treated and patients who had neoadjuvant chemotherapy, the material consisted of 61,199 surgical treated patients. Lymph node status was not available in 2868 cases (4.7% of the total), and these patients were not included in the study. Thus, the material included 58,331 breast cancer patients, of which 31,373 were treated by mastectomy and 26,958 by BCS. In the mastectomy group, there were 4230 who initially had breast conservation, but within 3 months had a mastectomy as a second or third procedure.

Figure 1. Overall survival in the three groups: patients primarily assigned to mastectomy, patients primarily assigned to BCS, and patients primary assigned to BCS but who finally had mastectomy (BCS & Mast).

Patient characteristics are given in Table 1. BCS increased from 24.4% in the period 1995–2001 to 68.0% in the period 2008–2012. In the 75+ age group, 23.6% had BCS. Compared to the mastectomy groups, patients in the BCS group (all p < .0001) had smaller tumors (median 23 versus 15 mm, respectively), had fewer lymph node metastases (57.1 versus 34.5%), and were more often hormone receptor positive (73.7 versus 83.1%).

Table 1. Characteristics of 58,331 Danish breast cancer patients surgically treated in the period 1995–2002.

Characteristic	Mastectomy (final)		BCS (final)		BCS and mastectomy
	N	%	N	%	N	%	Total
	27,143	44.4	26,958	44.1	4230	6.9	58,331
Year of inclusion
1995–2001	12,735	47	4846	18	2249	53	19,830
2002–2007	9030	33	8907	33	1148	27	19,085
2007–2012	5378	20	13,205	49	833	20	19,416
Age at operation
18–44	2598	10	2631	10	672	16	5901
45–54	5227	20	6446	24	1239	29	12,912
55–64	6869	25	9497	35	1167	28	17,533
65–74	6656	25	6496	24	823	20	13,975
75+	5793	21	1888	7	329	8	8010
CCI
0	21,172	78	22,210	82	3572	84	46,954
1	3443	13	2907	11	375	9	6725
2	1565	6	1275	5	185	4	3025
3+	963	4	566	2	98	2	1627
Tumor size (mm)
0–10	2304	9	7063	26	1133	27	10,500
11–20	8842	33	13,893	52	1592	38	24,327
21–30	8371	31	4879	18	808	19	14,058
31–50	5471	20	949	4	443	11	6863
51+	1807	7	47	0	129	3	1983
Unknown	342	1	117	0	125	3	584
Positive lymph nodes
0	11,640	43	17,656	66	2409	57	31,705
1–3	8825	33	7333	27	1194	28	17,352
4–9	3936	15	1427	5	402	10	5765
10+	2742	10	542	2	225	5	3509
Histological type
Ductal	21,744	80	22,143	82	3204	76	47,091
Lobular	3487	13	2277	8	632	15	6396
Other	1699	6	2357	9	359	9	4415
Unknown	213	1	181	1	35	1	429
Grade (ductal and lobular)
I	6159	24	8776	36	990	256	15,925
II	10,624	42	9969	41	1532	40	22,125
III	6177	25	4861	20	761	20	11,799
Unknown	2271	9	814	3	553	14	3638
Estrogen receptor
Negative (0–9%)	6068	22	4023	15	905	21	10,996
Positive (10–100%)	19,999	74	22,412	83	3104	73	45,515
Unknown	1076	4	523	2	221	5	1820
HER2 receptor
Normal	7046	76	15,003	88	1063	78	23,112
Positive	2228	24	1999	12	302	22	4529
Unknown	17,869		9956		2865		30,690
Lymphovascular invasion
No	19,919	73	23,893	89	3162	75	46,974
Yes	5108	19	1952	7	488	12	7548
Unknown	2116	8	1113	4	580	14	3809
Multifocality
Multifocal	4302	16	741	3	832	20	5875
Unifocal	20,897	77	25,170	93	2878	68	48,945
Unknown	1944	7	1047	4	520	12	3511

BCS and mastectomy = patients initially assigned to BCS, but finally having mastectomy. Differences between groups were statistically significant for all variables with p values <.0001.

Survival is depicted in Figure 1. The estimated median potential follow-up time was 11.5 years. The 10-year survival rates were 57% after primary mastectomy and 82% in patients having final BCS. Patients who had initially BCS but eventually were mastectomized had a 10-year overall survival of 74%, thus closer to the result in the BCS group.

The differences in survival between the various surgical groups are further elucidated in Table 2 showing both univariate and multivariate risk estimates for overall mortality and SMR. As shown, the mortality was increased in patients treated by mastectomy, although adjusting for the various prognostic risk factors reduced the difference between the groups. Of further notice was an only modest difference between patients, by whom the intention was to do BCS, but later were converted to mastectomy, and those patients, in whom breast conservation succeeded: relative risk (RR) 1.08 (95% CI 1.01–1.15) in the adjusted model.

Table 2. Univariate and multivariate risk estimates for overall mortality and SMR.

	Unadjusted values				Adjusted values
	Overall mortality		SMR		Overall mortality		SMR
	RR	95% CI	RR	95% CI	RR	95% CI	RR	95% CI
M (ITT) versus BCS (ITT)	2.43	2.36–2.51	1.38	1.34–1.42	1.21	1.17–1.26	1.19	1.15–1.23
M (final) versus BCS (final)	2.46	2.38–2.54	1.44	1.39–1.48	1.23	1.18–1.28	1.20	1.15–1.25
BCS&M versus BCS (final)	1.51	1.43–1.60	1.28	1.21–1.36	1.10	1.03–1.17	1.08	1.01–1.15
M (ITT) versus BCS (final)	2.64	2.55–2.72	1.46	1.41–1.50	1.25	1.20–1.31	1.22	1.17–1.27

M = mastectomy; M (ITT) = patients primary allocated to mastectomy; M (final) = patents finally having mastectomy; BCS (ITT) = patients primary allocated to BCS; BCS (final) = patients finally having BCS; BCS&M  = patients initially allocated to BCS but finally having mastectomy.

The impact on mortality of the patient, tumor, and treatment variables included are given in Table 3, presenting the results from the multivariate Poisson model on the intention to treat populations. Patients primarily assigned to mastectomy had 21% increased risk in overall mortality and 19% increase in SMR compared to BCS.

Table 3. The impact on mortality of the patient, tumor, and treatment variables evaluated in a multivariate Poisson model on the intention to treat populations.

Variable	Overall mortality			SMR
	RR	95% CI	p value	RR	95% CI	p value
Operation			<.0001			<.0001
Mastectomy versus BCS	1.21	1.17–1.26		1.19	1.15–1.23
Age at operation			<.0001			<.0001
Age at operation 18–44 versus 55–64	0.71	0.67–0.76		4.13	3.88–4.40
Age at operation 45–54 versus 55–64	0.70	0.67–0.74		1.56	1.49–1.64
Age at operation 65–74 versus 55–64	1.56	1.50–1.62		0.62	0.60–0.65
Age at operation 75+ versus 55–64	2.45	2.34–2.57		0.34	0.33–0.36
Treatment program			<.0001			<.0001
Year of inclusion 2002–2007 versus 1995–2001	0.80	0.77–0.83		1.02	0.99–1.06
Year of inclusion 2008–2012 versus 1995–2001	0.56	0.53–0.59		0.90	0.86–0.95
CCI			<.0001			<.0001
CCI 1 versus 0	1.49	1.44–1.55		1.51	1.45–1.57
CCI 2 versus 0	1.66	1.58–1.75		1.72	1.63–1.81
CCI 3+ versus 0	2.33	2.18–2.48		2.53	2.37–2.69
Tumor size			<.0001			<.0001
Tumor size 0–10 versus 11–20^a mm	0.79	0.75–0.83		0.82	0.78–0.86
Tumor size 20–30 versus 11–20^a mm	1.19	1.15–1.23		1.19	1.15–1.23
Tumor size 30–50 versus 11–20^a mm	1.33	1.27–1.38		1.32	1.27–1.38
Tumor size 50+ versus 11–20^a mm	1.54	1.44–1.64		1.56	1.46–1.66
Positive lymph nodes			<.0001			<.0001
Positive lymph nodes 1–3 versus 0	1.38	1.33–1.43		1.35	1.30–1.40
Positive lymph nodes 4–9 versus 0	2.12	2.02–2.22		2.20	2.10–2.30
Positive lymph nodes 10+ versus 0	3.12	2.96–3.29		3.47	3.29–3.65
Estrogen receptor (ER) status			<.0001			<.0001
ER negative versus positive	1.20	1.14–1.25		1.19	1.14–1.25
ER unknown versus positive	1.06	0.99–1.13		1.07	1.00–1.15
Lymphovascular invasion			<.0001			<.0001
Lymphovascular invasion yes versus no^a	1.22	1.18–1.27		1.26	1.21–1.31
Histological type			.0001			<.0001
Lobular versus ductal	0.99	0.95–1.04		0.99	0.95–1.03
Other type^a versus ductal	0.89	0.84–0.94		0.87	0.83–0.92
Grade			<.0001			<.0001
Grade 2^a versus 1	1.14	1.10–1.19		1.14	1.10–1.18
Grade 3 versus 1	1.38	1.32–1.45		1.40	1.33–1.46
Focality			<.0001			.04
Multifocal versus unifocal	1.01	0.97–1.06		1.04	1.00–1.09
Unknown versus unifocal	1.17	1.11–1.22		1.05	1.00–1.11
RT			<.0001			<.0001
Unknown versus no	1.11	1.04–1.19		1.14	1.07–1.22
Yes versus no	0.83	0.80–0.87		0.93	0.89–0.97
Chemotherapy			<.0001			<.0001
Unknown versus no	1.65	1.53–1.78		1.94	1.79–2.09
Yes versus no	0.81	0.77–0.85		0.95	0.90–1.00
Endocrine therapy			<.0001			<.0001
Unknown versus no	1.70	1.58–1.83		1.78	1.65–1.91
Yes versus no	0.78	0.75–0.82		0.84	0.81–0.88

^a Unknown included in group.

Interactions between surgical treatment and selected predictive variables are depicted in Figure 2. Significant interactions were not observed for age, period of treatment, and nodal status. HER2 was primarily registered for patients included in the latter half of the study inclusion period, with the majority of patients with HER2-positive tumors receiving HER2-targeted treatment. Inclusion of HER2 in the multivariate models did not alter the overall results. However, there was a highly significant interaction between type of operation and CCI. Those with CCI-score 2+ had no increased mortality after mastectomy, as opposed to patients with CCI 0–1. Similar results were observed for overall survival and for the ITT groups (data not shown). Careful review of CCI groups reveled that chemotherapy and RT were less often given to patients with comorbidity (Supplementary Table S1).

Figure 2. Forrest plot showing interactions between selected risk factors and the risk of mortality (SMR) and the final surgical procedure.

Finally, the complex interaction between surgical type and both nodal status and RT was evaluated (Supplementary Table S2). No statistical significant interactions were observed (p = .93). Indication for loco-regional RT has varied over the years for the present study cohort. Comparing patients with macrometastases, loco-regional RT did not reduce the increased risk [RR = 1.28 (95% CI 1.19–1.38)], with an almost identical estimate for patients with macrometastases and no indication for loco-regional RT, i.e., mastectomy no RT versus BCS breast only RT, [RR = 1.27 (95% CI 1.14–1.42)]. Similar results were observed for overall survival and for the ITT groups (data not shown).

Discussion

The present study has shown a higher risk of mortality in breast cancer patients treated by mastectomy compared with patients having breast conserving therapy. Patients going through an attempt of breast conservation prior to mastectomy had a considerable better outcome than those primarily assigned to mastectomy. Adjusting for various prognostic risk factors decreased the excessive mortality after mastectomy to ∼20%. The increased risk seems independent of age and period of treatment. Differences in RT could not explain the different outcome between mastectomy and BCS. The observed differences seem to some degree to rely on residual confounding.

The present results confirm previous findings from register based studies of an up to 20–30% better survival in patients treated by BCS compared to mastectomy. The largest study based on data from the Surveillance, Epidemiology, and End Results (SEER) database [4], included more than 130,000 patients treated in the period 1998–2008 with tumor size up to 4 cm and ≤3 positive lymph nodes. In the multivariate analysis in that study, the hazard ratio (HR) for survival was 1.31 after BCS compared with mastectomy alone. The HR was even higher when comparing BCS with mastectomy combined with RT (1.47). The study had some major limitations, as it did not include information on systemic therapy, and the tumor characteristics were limited to hormone receptor status, grade, and lymph node status, with no information on LVI or HER2 status.

Similar results were reported from one Canadian [5], further two American [6,7], and two Norwegian studies [13,14]. These studies also all lacked some essential information on tumor biology and on treatment, especially on systemic treatment. Contrary to this, a study from the Netherlands Cancer Registry including around 37,000 patients treated in the period 2000–2004 had more prognostic factors included in their analysis. Thus, information on tumor size, nodal status, and grade were available for most patients, but data regarding hormone receptor status was incomplete and lacking for a large proportion of the patients. Also, the study did not report on HER2 status. The study population was restricted to subgroups T0–2, N0–1. Adjuvant systemic treatment was to only given to ∼50% of the patients included, and among patients undergoing mastectomy, all who had RT were excluded. Overall, the adjusted HR for 10-year survival was 0.81 in the mastectomy group, but looking at subgroups a significant benefit was only observed in the T1N0 group. Later, the Dutch group reported in a second study on patients with T1–2, N2 breast cancer [15]. The study included 3700 patients and found an overall adjusted HR at 0.88. Again significance was not found in all subgroups, and in this case it was restricted to the T2N2 group. The same group very recently published a third paper also based on data from the Netherlands Cancer Registry [16]. In that study, the population included 129,692 patients treated in the period 1999–2012, thus including the populations in the previous two studies. Interestingly, the last study report slightly diverging results compared to the previous results: no survival benefit after BCS in the T1–2N2 group while BCS was superior to mastectomy in all T1–2N0–1 subgroups. After stratification there was no difference in patients younger than 50 years and in patients with comorbidity.

The two first studies from the Netherlands, which were known to us when the present study was planned, included more prognostic risk factors in the multivariate analysis than in the aforementioned studies and gave less convincing indications for a better survival after BCS. Based on these observations, and as we were able to include more prognostic factors than presented in previous studies, we expected to find a less pronounced difference in outcome between BCS and mastectomy. But as shown, this was not the case. Even after adjustment for a wide and more complete range of prognostic factors, than in all other population-based studies so far, patients treated by BCS seemed to have a better outcome. It has been argued that it is the RT given to all BCS patients, as opposed to patients treated by mastectomy, which explains the observed differences [16,17], but the present results clearly reject that hypothesis.

Previous Danish studies have pointed at a higher recurrence risk and higher mortality in younger patients [18,19] and among these a worse outcome after BCS compared to mastectomy [18,20]. These studies were done on patients treated in the 1990-ties. The present results on a somewhat later treated cohort, where the proportion of patients treated by BCS increased over time, do not confirm that. The increased use of and improvement in systemic treatment and RT are probably the main explanation for this change in outcome results. The previous studies had longer observation periods and they showed a significant increased mortality even after 10 years observation and we cannot rule out that a longer observation period in the present study would have changed the picture somewhat, although we do not believe the general picture would have been markedly changed, as we consider our results robust.

It is obvious that confounding by indication is in play. Comorbidity has a strong impact with poorer survival [21], and we found that patients with more comorbidity were preferably treated by mastectomy, reducing the survival in this group. Also, the proportion of patients with higher risk tumors was increased. Adjusting for prognostic risk factors reduced the RR after mastectomy, but not completely. The observation that patients, who were initially assigned to BCS, but had a final mastectomy, had a significantly better outcome than patients primarily assigned to mastectomy and only had a slight increase in mortality relative to what was seen in the BCS group, strongly infers that there is some residual confounding, which is not accounted for in our study, and it seems more pronounced in patients with tumors belonging to the more favorable end of the spectrum and to patients without comorbidity. The minor differences observed between the two groups assigned to BCS might be related to recent findings showing that patients undergoing mastectomy secondary to BCS because of insufficient margins, had an increased risk of distant metastasis [22].

In theory, there are possible explanations for a worse prognosis among the group of patients treated by mastectomy, not accounted for in this study. Some patients selected for mastectomy have characteristics which give them a higher risk of recurrence, including dense breasts [23] and widespread DCIS in the surroundings [24]. Some studies [25–27] have also indicated that multifocal tumors have a more aggressive biology, although others have given diverging results [25,27–29]. It has been speculated that tumors detected by mammography screening preferably treated by BCS have a much more favorable prognosis and therefore should be considered a major confounding factor. Unfortunately, we did not have information on mammography screening that allowed us to include that in our study, but one of the previously cited studies from Norway looked specifically on the impact of detection mode on the difference in outcome between the surgical treatment groups and found a survival benefit after BCS independent of whether the tumors were screen detected or had presented clinically [14]. Multifocality/multicentricity was included in the present study, but the distribution among the groups was very skew, as multifocality has been considered a contraindication for BCS. Therefore, the adjustment did probably not fully adjust for this difference.

Finally, one cannot rule out that there are some adverse effects of mastectomy contrary to BCS. Mastectomy is a more extensive procedure and leads to more tissue damage and a more pronounced inflammatory response [30]. This could have a negative effect by suppressing the immune system and promoting growth of residual tumor cells and in the circulation for instance by angiogenesis of dormant avascular micrometastases and surgery-induced activity of single malignant cells [31,32].

Several issues should be considered when interpreting this study. First, we were able to identify patients in whom BCS was attempted but who went on to secondary mastectomy. Patients with repeat surgery had a significantly lower mortality compared to patients with a primary mastectomy, but still not as favorable as patients with BCS only. Second, detailed diagnostic and treatment characteristics were recorded prospectively, and adjusting for these factors partly explained the difference in outcome observed after BCS and mastectomy. Third, CCI could be estimated using administrative data and in patients with a CCI score of 0 or 1 mastectomy as first choice of surgery was associated with higher mortality and SMR while a CCI score of ≥2 was not. Fourth, our study comprises a large and population-based cohort allowing calculation of mortality relative to the Danish female population. Our study was however confounded and we are unable to determine whether the difference in outcome observed after controlling for repeat surgery and adjusting for other patient and treatment characteristics is a real effect or constitutes residual confounding. Our study had other limitations including lack of information on breast density, presence of DCIS, and HER2 status for a major part of the population.

In conclusion, patients assigned to BCS have a better survival than patients assigned to mastectomy. Residual confounding after adjustment for registered characteristics presumably explained the different outcomes, thus consistent with selection bias. Diversities in RT do not seem to explain the observed difference in survival after BCS and mastectomy.

Disclosure statement

No potential conflict of interest was reported by the authors.