Removal of primary tumor improves survival in metastatic breast cancer. Does timing of surgery influence outcomes?

The Breast 20 (2011) 548e554

Contents lists available at ScienceDirect

The Breast journal homepage: www.elsevier.com/brst

Original article

Removal of primary tumor improves survival in metastatic breast cancer. Does timing of surgery influence outcomes? Jose Alejandro Pérez-Fidalgo a, e, *, Paola Pimentel a, e, Antonio Caballero b, Begoña Bermejo a, Juan Antonio Barrera a, Octavio Burgues c, F Martinez-Ruiz d, Isabel Chirivella a, Ana Bosch a, Angel Martínez-Agulló b, Ana Lluch a a

Department of Medical Oncology, Hospital Clinico Universitario, Avda Blasco Ibanez s/n, 46018 INCLIVA, Valencia, Spain Department of Surgery, Hospital Clinico Universitario, INCLIVA, Valencia, Spain Pathology Department, Hospital Clinico Universitario, Avda Blasco Ibanez s/n, 46018 INCLIVA, Valencia, Spain d Statistics Department, University of Valencia, Spain b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 27 September 2010 Received in revised form 2 May 2011 Accepted 26 June 2011

Background: Resection of intact primary tumor is controversial in metastatic breast cancer patients. The aim of this study is to review the impact of surgical resection of primary tumor on overall survival and to assess the role of timing of surgery on survival rates. Methods: 208 patients with metastatic breast cancer diagnosed between 1982 and 2005 in the Hospital Clinico of Valencia (Spain) were analysed. Exclusion criteria were age >80, PS 3e4, Charlson score 3 or follow-up < 90 days. 123 of these underwent surgery and 85 did not. In order to assess the role of timing, the "surgery" cohort was divided into two sub-groups: "before" (n ¼ 78) or "after" (n ¼ 45) diagnosis of disseminated disease. Results: In the surgery group, patients underwent mastectomy with axillary dissection (82.9%), without axillary dissection (8.9%) and conservative surgery (8.1%). After a median follow-up of 29.68 months, median OS in the "surgery" and the "non-surgery" groups were, 40.4 and 24.3 months. Removal of the primary tumor therefore had a significant positive impact on survival rates (p < 0.001). Benefits of surgery were observed mainly in patients with visceral disease (p ¼ 0.005); no statistical differences were found in those with bone disease (p ¼ 0.79). Univariate analysis for overall survival (OS) identified surgery, performance status, clinical T stage, hormone receptors and number and type of metastases as variables that impacted on survival. In the multivariate test, only resection of primary tumor and estrogen receptors maintained statistical significance, surgery having a protective effect with an HR 0.52 (95% CI 0.35e0.77). No differences in survival were found between the two sub-groups according to the timing of surgery: "before" vs "after"(p ¼ 0.996). Conclusions: Resection of primary tumor should be considered not only as a palliative or preventive strategy but also as an approach that possibly contributes to the control of the disease in selected patients. Ó 2011 Elsevier Ltd. All rights reserved.

Keywords: Primary tumor removal Advanced breast cancer Timing Debulking

Introduction Women with stage IV tumor at first diagnosis represent a minority of the total breast cancer population, in fact, only a 5% present metastasis at diagnosis.1 Disseminated breast cancer is an incurable disease and therapeutic approach at this stage is therefore mainly palliative. Aggressive strategies for local control have

* Corresponding author. Tel.: þ34 963862600. E-mail address: japfi[email protected] (J. A. Pérez-Fidalgo). e Both authors contributed equally to the development of this paper. 0960-9776/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.breast.2011.06.005

been traditionally reserved for either symptomatic control or prevention of primary tumor complications, such as ulceration, bleeding or pain. Preclinical studies in animal models have demonstrated that resection of primary tumors leads to secondary growths of metastatic disease.2 In clinical settings, however, aggressive local therapy has been found to improve survival in several metastatic tumors, such as renal cell3 and ovarian cancer.4 This apparent contradiction between preclinical and clinical models might be explained by the fact that therapeutic approaches in the latter are multidisciplinary, with local control being only one part of a global strategy.

J.A. Pérez-Fidalgo et al. / The Breast 20 (2011) 548e554

Over the past years, several retrospective studies have assessed the potential impact on survival of aggressive local treatment in cases of advanced breast cancer. Most of these studies have found a significant increase in survival when primary tumors have been resected.5 By contrast, other authors suggested that these apparently consistent results could be biased.6 One possible type of bias posing major limitations in any retrospective series comes from patient selection criteria. Another bias, as suggested by Bafford et al,7 is stage migration. In their series of 147 women (61 in the surgery arm versus 86 in the non surgery arm), surgery had a positive impact on survival rates after adjusting for other influencing variables with a p ¼ 0.003. However when the surgery group was divided into two sub-groups according to the timing of staging (i.e., patients operated before and after knowing presentation of stage IV), the effect on survival for the “after” group disappeared. Bafford et al thus suggested that surgery impact might be consequence of stage migration bias. Drawing upon this literature, we performed a retrospective study within a single institution (Hospital Clínico, Valencia, Spain) in order to explore both the impact of surgery and of its timing on survival rates in a cohort of patients diagnosed with stage IV breast cancer. Restrictive inclusion criteria were established in order to minimize selection bias, and timing of surgery was assessed in order to ascertain the potential of stage migration bias. Materials and methods Patients We conducted a retrospective analysis of our prospectively maintained database in the Hospital Clínico of Valencia (Spain) from 1982 to 2005. This included all patients diagnosed with confirmed carcinoma of the breast and with a stage IV in the TNM classification. At the time of their first assessment, 254 patients were identified as presenting a metastatic breast cancer. Patients’ data comprised a review of their complete medical history, including social demographic variables, comorbidities, clinical and pathological features, surgical modalities and cause of death. Social and demographic variables recorded were age; menopausal status and date of diagnosis (pre-1990, from 1990 to 1999 and post-1999). General condition was assessed via the ECOG performance status (PS) score, and data related to different comorbidities were classified following the Charlson comorbidity index.8 Histological diagnosis was classified as ductal carcinoma, lobular carcinoma, non-specified, and other. Clinical stage was established following the TNM stage system (AJCC 5th edition). Clinical stage of primary tumour was classified according to its size in <2 cm, ¼ 2 cm, or > 2 cm, and axillary lymph node involvement into cN0 or cNþ. HER2 overexpression was determined with herceptest, and thus classified into positive (herceptest was þþþ) or negative (herceptest was þ). In case of herceptest þþ a FISH amplification techniques was performed. Metastases were classified according to the number and location of metastic sites at diagnosis. Patients with only supraclavicular involvement were included in the analysis despite since the last revision of AJCC single supraclavicular metastasis is considered a stage IIIC. However the stage at the moment of deciding therapeutic approach was stage IV. As a consequence, this approach was always palliative and in fact some of these patients were not operated. Thus, inclusion criteria was based on the 5th edition of the AJCC classification valid at the time of diagnosis of all patients. Exclusion criteria affected patients older than 80 years, with PS higher than 2, Charlson score higher than 2, and with follow-up less

549

than 90 days from diagnosis. Patients with these features were not considered suitable for eventual surgery because of their bad general condition or their short life expectancy. Therefore, from an initial database of 254 patients, 10 patients were excluded for being older than 80 years, 8 patients for presenting a PS 3 or 4 at diagnosis, 26 patients for a follow-up period of less than 90 days and 2 patients for Charlson scores of 3. The final database comprised 208 patients of whom 123 (59.1%) underwent resection with the intention of completely removing their primary tumors (the “surgery” group), and 85 (40.9%) received no resection of the primary tumor other than biopsy (the “non-surgery” group). Firstly, a comparison between the “surgery” and the “nonsurgery” cohort was performed in order to assess resection impact on overall survival. Overall survival (OS) was defined as the period of time from diagnosis to either death or loss of follow-up. Secondly, potential prognostic or predictive factors for survival were assessed in the “surgery” cohort, such as type of surgery, pathological stage, affection of margins and radiotherapy. Primary tumor size was classified after pathological examination as < or ¼ 2 cms, between 2 and 5 cms or more than 5 cms. Radiotherapy was indicated, according to the guidelines approved for early breast cancer, in larger than 5 cm tumors, with either more than 3 affected axillary nodes or less than 10 examined nodes, or with invasion of resection margins. Patients included in the “surgery” cohort were classified into those who received radiotherapy according to guidelines for adjuvant therapy and those whose radiotherapy did not conform with the adjuvant recommendations. Finally, the role played by timing of surgery was assessed by classifying the “surgery” cohort into two groups: patients operated in a period of 3 months before knowing the presentation of stage IV cancer (the “before” subgroup) and patients operated on after diagnosis of systemic dissemination, as part of a global therapeutic strategy (the “after” subgroup). In order to explore the real confounded effect of a hypothetical stage migration bias, a second KaplaneMeier analysis of OS was performed with the three cohorts (“non-surgery” versus “surgery” and, within the later, “surgery before” versus “surgery after”). Follow-up was performed according to investigator criteria and depending on the symptomatic control, extend of disease and response to systemic therapies in the non-operated group, and on effectiveness of cytoreduction and extend of remaining disease in the operated cohorts. In patients with no evidence of disease follow-up was maintained more than 15 years. During the first 2e3 years at least patients were controlled every 3e4 months, except if symptom management needs, or therapeutic assessment recommended more frequent visits. Statistical analysis Patient features and survival curves were analysed using the SPSS for Windows statistical package. Patient features were compared using a ChieSquare test for qualitative variables (with Fisher statistical test for variables with only two values) and a Student’s-t test (2 independent samples) for quantitative variables. Survival curves for different cohorts were assessed using the KaplaneMeier method, which compares differences with the logrank test. An analysis of prognostic factors was performed using univariate Cox-regression tests, constructing a multivariate model for those prognostic factors that were statistically significant (p value inferior to 0.10) in the univariate model. All hypothesis contrasts were two-sided and considered significant with a level a ¼ 0.05. In case of multiple comparisons, Bonferroni correction was applied to p-values of associated tests.

550

J.A. Pérez-Fidalgo et al. / The Breast 20 (2011) 548e554

Table 1 Comparison of patient characteristics in both cohorts, surgery versus no surgery, and p values. SD: standard deviation, CT chemotherapy. % was calculated without considering not available data. Total (N ¼ 208)

No Surgery Surgery (N ¼ 85) (N ¼ 123)

p

57.2 (13.0)

59.2 (13.0) 55.9 (12.8)

0.071 0.342

Age Mean (SD) Menopausal status Pre-menopausal Postmenopausal Not available

57 (28.1%) 20 (24.1%) 146 (71.9%) 63 (75.9%) 5 2

37 (30.8%) 83 (69.2%) 3

Date of diagnosis <1990 1990e1999 > 1999

43 (20.7%) 17 (20.0%) 116 (55.8%) 51 (60.0%) 49 (23.6%) 17 (20.0%)

26 (21.1%) 65 (52.8%) 32 (26.0%)

145 (69.7%) 45 (52.9%) 38 (18.3%) 24 (28.2%) 25 (12.0%) 16 (18.8%)

100 (81.3%) 14 (11.4%) 9 (7.3%)

Charlson comorbidity score Charlson 0 Charlson 1 Charlson 2

162 (77.9%) 67 (78.8%) 33 (15.9%) 11 (12.9%) 13 (6.3%) 7 (8.2%)

95 (77.2%) 22 (17.9%) 6 (4.9%)

T stage (clinic) <¼ 2 cm. > 2 cm. Not available

20 (9.8%) 5 (5.9%) 185 (90.2%) 80 (94.1%) 3 0

15 (12.5%) 105 (87.5%) 3

N stage (clinic) N0 Nþ Not available

60 (29.4%) 25 (29.8%) 144 (70.6%) 59 (70.2%) 4 1

35 (29.2%) 85 (70.8%) 3

Histological subtype Ductal Lobular Other Carcinoma (Non-specified)

172 (82.7%) 19 (9.1%) 11 (5.2%) 6 (2.8%)

67 (78.8%) 7 (8.2%) 6 (7.1%) 5 (5.9%)

105 (85.4%) 12 (9.8%) 5 (4.1%) 1 (0.8%)

Histological grade Grade I Grade II Grade III Not available

15 (13.0%) 62 (53.9%) 38 (33.0%) 93

3 (10.0%) 20 (66.7%) 7 (23.3%) 55

12 (14.1%) 42 (49.4%) 31 (36.5%) 38

ER Negative Positive Not available

62 (42.8%) 83 (57.2%) 63

20 (37.7%) 33 (62.3%) 32

42 (45.7%) 50 (54.3%) 31

PR Negative Positive Not available

75 (51.4%) 70 (48.6%) 63

26 (48.1%) 27 (51.9%) 32

49 (53.3%) 43 (46.7%) 31

HER2 Negative Positive Not available

19 (48.7%) 20 (51.3%) 169

5 (38.5%) 8 (61.5%) 72

14 (53.8%) 12 (46.2%) 97

Metastatic sites 1 site 2 sites 3 or more sites

124(59.6%) 39 (45.9%) 48 (23.1%) 19 (22.4%) 36 (17.3%) 27 (31.8%)

85 (69.1%) 29 (23.6%) 9 (7.3%)

108(51.9%) 56 (26.9%) 30 (14.4%) 14 (6.7%)

52 44 17 10

Type of first line systemic treatment CT with anthracycline and/or taxanes CT other (nor anthracyclines nor taxanes) Endocrine therapy Not available

Total (N ¼ 208) Best response to first line systemic treatment Objective response (complete or partial) Stable disease Progressive disease Non- evaluable Not available

No Surgery Surgery (N ¼ 85) (N ¼ 123)

p 0.988

122 (59.5%) 49 (58.3%)

73 (60.3%)

44 (21.5%) 32 (15.6%) 7 (3.4%) 3

25 (20.7%) 19 (15.7%) 4 (3.3%) 2

19 (22.6%) 13 (15.5%) 3 (3.6%) 1

0.531

Performance status (PS) PS 0 PS 1 PS 2

Type of metastatic disease Visceral disease (any) Bone disease (only) Lymph node/soft tissue (only) Bone disease þ lymph node/soft tissue

Table 1 (continued)

Results <0.001

0.428

0.153

>0.999

0.368

0.263

0.387

0.604

0.501

<0.001

0.002 56 (65.9%) 12 (14.1%) 13 (15.3%) 4 (4.7%)

(42.3%) (35.8%) (13.8%) (8.1%) 0.285

141 (68.1%) 53 (62.4%)

88 (72.1%)

27 (13.0%)

12 (14.1%)

15 (12.3%)

39 (18.8%) 1

20 (23.5%) 0

19 (15.6%) 1

208 patients were included in the analysis, of whom 123 underwent resection of the primary tumor (“surgery” group) and 85 were not operated on (“non-surgery” group). Patient characteristics in both cohorts are shown in Table 1. PS and the number of metastatic sites were statistically different. The “surgery” cohort presented more favorable features, with better general condition (more patients with PS ¼ 0), less metastatic sites and a lower rate of visceral disease than the “non-surgery” group. However, no differences were found regarding the variables of age, menopausal status, Charlson comorbidity score, clinical stage (T and N), molecular prognostic and predictive factors or type, and response to systemic first line treatment. The most extended type of surgery in the entire cohort was mastectomy with axillary dissection (102 patients; 82.9%). Partial or total mastectomy without axillary dissection was performed in 11 patients (8.9%). Conservative surgery was performed in only 10 patients (8.1%) of whom 3 had no axillary dissection. Therefore, only 14 patients in our series underwent a primary tumor resection without axillary dissection. Resection margins were positive in 24.1% of patients in the surgery cohort, but no statistical differences were obtained between the “before” and “after” groups. Adjuvant radiotherapy after resection was administered to 57 patients (46.3%) and adjustment to guidelines in the adjuvant setting was assessed. Decisions about radiotherapy were made according to guidelines for adjuvant setting in 67 patients (54.5%). In 55 patients (45.5%), decisions were not adjusted to recommendations. 4 of this latter group received radiotherapy non-adjusted to adjuvant indications and 51 received no adjuvant treatment although this would be recommended in the adjuvant setting (mainly because of affected margins or low number of examined axillary nodes). After a median follow-up of 29.68 months, 165 of the patients had a breast cancer related death (79.3%), 7 patients presented a non-cancer related death (3.4%) and 15 patients were lost in follow-up (7.2%). 21 patients (17 in the surgery cohort and 4 in the non-surgery group) are still alive (10.1%), of whom 3 are long survivors e patients with more than 120 months of follow-up. KaplaneMeier survival curves comparing OS in each cohort showed a benefit of surgery in our series (see Fig. 1). The median OS in the “surgery” and the “non-surgery” group were, respectively, 40.4 and 24.3 months (mean OS was 60.6 and 32.9 months). The benefit of surgery was statistically significant (p < 0.001). Survival analysis of the subgroup of patients with bone disease (n ¼ 56, 12 “non-surgery”/44 “surgery”) failed to demonstrate any statistical differences (p ¼ 0.79). However in the subgroup of patients with visceral metastases (n ¼ 108, 56 “non-surgery”/52 “surgery”), differences were significant (p ¼ 0.005) (see Fig. 2). From 15 patients with single supraclavicular metastasis, in 11 patients removal of primary tumor was performed and the rest 4 patients were not operated.

J.A. Pérez-Fidalgo et al. / The Breast 20 (2011) 548e554

551

Univariate analysis for OS identified surgery, performance status, estrogen receptors and the number and type of metastatic sites as variables with impact on survival. T stage (p ¼ 0.058) and progesterone receptors (p ¼ 0.057) showed a non-significant trend to influenced survival and thus were considered for the multivariant analysis as potential influencing factors. In contrast, age, date of diagnosis and Charlson comorbidity score failed to attain statistical significance (see Table 2).

Table 2 Univariate and multivariate analysis for overall survival of potential prognostic factors. RR ¼ relative risk, CI ¼ confidence interval, ER ¼ estrogen receptors, PR ¼ progesterone receptors. Univariate Fig. 1. Overall survival curves in surgery (green line) versus non-surgery group (blue line). Differences are significant with a p < 0.001. ’For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article’. (i.e., if the figure is black and white in the printed version and colour in the web version).

A 1,0

Cum survival

0,8

0,6

0,4

0,2

0,0 0

12

24

36

48

60

72

84

96

108

120

Time (months)

B

1,0

Cum survival

0,8

0,6

0,4

0,2

0,0 0

12

24

36

48

60

72

84

96 108 120 132 144 156 168 180

Time (months) Fig. 2. Impact of removal of primary tumor by site of metastases. A: Bone disease only (N ¼ 56), no statistically significant differences in overall survival were found in patients with bone disease only (p ¼ 0.79). B: Visceral disease: In patients with at least one visceral metastasis (N ¼ 108) differences in overall survival were significant (p ¼ 0.005). Surgery cohort in both cases is represented with a green line and nonsurgery cohort is represented with a blue line. ’For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article’. (i.e., if the figure is black and white in the printed version and colour in the web version).

HR Surgery No Yes Age Menopausal status Premenopausal Postmenopausal Date of diagnosis <1990 1990e1999 > 1999 Performance status (PS) PS 0 PS 1 PS 2 Charlson comorbidity score Charlson 0 Charlson 1 Charlson 2 cT stage <¼ 2 cm >2 cm cN stage N0 Nþ Histological subtype Ductal Lobular Other Non-specified Histological grade Grade I Grade II Grade III ER Negative Positive PR Negative Positive HER2 Negative Positive Metastatic sites 1 site 2 sites 3 or more sites Type of metastatic disease Visceral metástasis (at least one) Bone disease only Soft tissue only Bone and soft tissue involvement

95% CI

Multivariate p-valor HR

95% CI

<0.001 (base line) 0.511 0.374e0.699

p-valor 0.001

(base line) 0.520 0.350e 0.773 e e

0.997 0.985e1.010 0.682 0.432 (base line) 1.150 0.811e1.632 0.316 (base line) 0.831 0.571e1.209 0.334 0.698 0.438e1.112 0.131 0.003 (base line) 1.861 1.255e2.759 0.002 1.620 1.001e2.623 0.050 0.303 (base line) 1.300 0.867e1.949 0.204 0.748 0.349e1.605 0.456 0.058 (base line) 1.702 0.982e2.950 0.385 (base line) 1.162 0.828e1.632 0.356 (base line) 0.971 0.577e1.633 0.911 1.656 0.892e3.076 0.110 0.588 0.145e2.380 0.457 0.144 (base line) 0.845 0.454e1.571 0.593 1.354 0.702e2.612 0.366 0.001 (base line) 0.541 0.372e0.787

e

0.120 0.041 0.855 e

0.289

e

e

0.001 (base line) 0.526 0.359e 0.773

0.057

0.777

0.883

e

(base line) 0.694 0.477e1.011 (base line) 1.060 0.489e2.299 0.047 (base line) 1.408 0.968e2.049 0.074 1.554 1.037e2.327 0.033 0.017

0.569 0.329 0.561 0.465

(base line) 0.584 0.405e0.844 0.004 0.812 0.516e1.277 0.368 0.511 0.257e1.016 0.055

0.974 0.121 0.289

552

J.A. Pérez-Fidalgo et al. / The Breast 20 (2011) 548e554

In the multivariate test, performance status, clinical T stage, progesterone receptors and number and type of metastatic sites lost statistical significance, while surgery and estrogen receptors maintained it, hence identifying the latter two variables as independent prognostic factors for OS. The importance of the protective effect of resection of the primary tumor was revealed in the multivariate analysis, with a hazard ratio of death of 0.52 (95% CI 0.35e0.77). These results support the role of surgery in stage IV breast cancer as an independent prognostic factor for survival. According to the timing of surgery, of the 123 patients in the “surgery” cohort, 78 patients were included in the “before” and 45 in the “after” group. All patients in the “before” group had undergone primary tumor resection as their first therapeutic approach within the first three months of having their histological diagnoses. All patients in the “after” group had been operated on after receiving at least 1 systemic treatment. Surgery in these patients was decided upon as part of the therapeutic approach for a stage IV breast cancer, and took place between 2.07 and 30.03 months after initiation of the first line of systemic therapy. KaplaneMeier curves comparing OS in the “non-surgery” and either the “surgery before” or the “surgery after” groups are shown in Fig. 3. No statistical differences in survival were found between the two sub-groups of the surgery cohort (p ¼ 0.996). However, the differences between the two surgery sub-groups were statistically significant when compared with the “non-surgery” cohort (p ¼ 0.012 for the “before” and the “no surgery” group, and p ¼ 0.003 for the “after” versus the “no surgery” groups). Discussion There is a lack of consensus about the real impact of surgery in metastatic breast cancer. Results supporting the role of surgery in this setting suggest that the benefit on survival rates might result either from elimination of tumor shedding from the primary tumor or from enhanced effectiveness of systemic therapy induced by a debulking process.9 Other explanation for this might be the reversion to the tumor-induced immunosuppresion observed in animal models after removal of primary tumors.10

1,0

Cum Survival

0,8

0,6

0,4

0,2

0,0 0

12

24

36

48

60

72

84

96

108 120 132 144 156 168 180

Time (months) Fig. 3. Overall survival for “non-surgery” cohort (blue line) versus “Surgery before” (green line) and “Surgery after”(red line). No statistical significant differences were found between surgery “before” versus “after”. ’For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article’. (i.e., if the figure is black and white in the printed version and colour in the web version).

The past ten years have seen an increasing number of publications of retrospective series that assess the role of surgery in stage IV cancer versus a control “non-surgery” arm. Since 2002, 13 retrospective series and 1 review of the literature have been published on this issue, 9 hospital-based11e18 and 4 population based.19e22 However, the disparity in these studies regarding inclusion criteria and several aspects of study-design continue to feed the controversy regarding the effect of surgery on survival rates. Studies based on population databases enable large size series to be examined, thereby increasing the statistical potency of their findings. However, single-institution studies provide more information about timing and type of surgery, as well as about complementary systemic therapy. In the present series, data showed a significant impact of surgery on OS rates. The benefit of surgery as an independent prognostic factor for survival was also demonstrated in the multivariate analysis. These data support the recommendation of performing primary tumor resection in metastatic breast cancer in selected patients. Surgery, however, was not shown to be statistically significant in patients with bone disease only, but no definitive conclusions can be drawn from this subgroup analysis due to small sample size. In contrast, patients with visceral involvement clearly benefited from aggressive local control, which suggests that primary tumor resection might be considered as a therapeutic approach in disseminated breast cancer with visceral metastases. One of the main criticisms of the results in retrospective series, particularly of those population-based studies, is the absence of restrictive inclusion criteria, which might increase the influence of the inherent selection bias factor. Older patients, with bad general condition, low life expectancy and severe comorbidity are less likely to undergo surgery, so control arms in any comparative studies should avoid inclusion of these patients. In fact, in most of the previous population and hospital-based retrospective studies, stage IV features as the single inclusion criteria (see Table 3), with only a few studies having included other restrictions, such as follow-up >90. Limited selection criteria for both cohorts of patients undergoing surgery or NS controls are a serious limitation of the data. In our study we established restrictive inclusion criteria (exclusion of patients with bad PS, old age, short followup and severe comorbidity) in an attempt to minimize selection bias. Timing of surgery has been previously explored as a potential factor for survival. Rao et al23 carried out a retrospective study of 75 patients undergoing resection of primary tumor that were classified by timing of surgery from diagnosis as follows: 0e2.9 months, 3e8.9 months and >8.9 months. Results showed a statistical benefit on progression-free survival for the 3e8.9 months group, with no difference in OS. These results, however, were based on data from a small number of events in each subgroup. Bafford et al7subsequently assessed the effect of surgery timing in two sub-groups compared with a cohort of 86 non-operated patients. Data showed a benefit of surgery only in the “before” group. No significant difference was found between patients in the “after” group and those in the “no surgery” cohort, suggesting that the observed benefit of surgery was due to a stage migration bias. In our study, the benefit of surgery on survival was maintained after assessing the effect on OS according to timing of surgery. Both surgical sub-groups (“before” and “after”) demonstrated statistical differences versus “non-surgery” cohort. Site and number of metastases were two variables that revealed statistical significance in the basal characteristics of the patients who underwent surgery versus those who did not. Thus, in our study we carried out a separate survival analysis by metastasis site

J.A. Pérez-Fidalgo et al. / The Breast 20 (2011) 548e554

553

Table 3 Retrospective studies published to date comparing overall survival of removal of intact primary (Yes) versus a control group of non-operated patients (No) in stage IV breast cancer and inclusion criteria of each study. SEER: Surveillance, Epidemiology and End Results. NCDB: National Cancer Database.

Hospital-Based Babiera et al. (2006) (Ref. 11) Blanchard et al. (2007) (Ref. 12)

Fields et al. (2007) (Ref. 13) Cady et al. (2008) (Ref 14)

Hazard et al. (2008) (Ref. 15) Leung et al. (2009) (Ref. 16) Bafford et al. (2009) (Ref. 7) McGuire et al. (2009) (Ref 17) Neuman et al. (2010) (Ref. 19) Current study

PopulationeBased Khan et al. (2002) (Ref. 19) Rapiti et al. (2006) (Ref. 22)

Gnerlich et al. (2007) (Ref. 20)

Number Primary of patients surgery

Period

224

1997e2002 USA (Houston)

395

409 622 (308 in matched analysis) 111 157 147 566 186 208

16023 300

9734

Ruiterkamp et al. (2009) 728 (Ref. 21)

Yes: 82 No: 142 Yes: 42 No: 153

Population

1973e1991 USA (Houston)

Median overall survival (OS) in months Whole group: 32.1 Yes: 27.1 No: 16.8

Yes: 187 No: 222 Yes: 234 No: 388

1996e2005 USA (Washington) Yes: 26.8 No: 12.6 1970e2002 USA (Boston) Yes: 33 No: 18

Yes: 47 No: 64 Yes: 52 No: 105 Yes: 61 No: 86 Yes: 154 No: 412 Yes: 69 No: 117 Yes: 123 No: 85

1995e2005 USA (Chicago) 1990e2000 USA (Richmond) 1998e2005 USA (Boston) 1990e2007 USA (Tampa) 2000e2004 USA(New York) 1982e2005 EuropeeSpain (Valencia)

Yes: 9162 1990e1993 USA (NCDB) No: 6861 Yes: 127 1977e1996 Europe No: 173 Switzerland

Yes: 26.3 No: 29.3 Yes: 25 No: 13 Yes: 42 No: 28 Yes: 33% at 5 ys No: 20% at 5 ys Yes: 40 No: 33 Yes: 40.4 No: 24.3

Yes: 4578 1988e2003 USA No: 5156 (SEER database)

Yes: 26.9/31.9 No: 19.3 Yes: positive margin 16% negative margin 27% No: 12% Yes: 36 and 18 No: 2 and 7

Yes: 288 No: 440

Yes: 31 No: 14

1993e2004 Europee Netherlands

in order to assess the importance of these variables on the effect of surgery. In this analysis, patients with bone disease did not benefit from surgery, but small sample size avoid to draw conclusions in this subgroup. However, in patients with visceral metastasis the effect of surgery was strongly significant. In conclusion, the effect on OS of surgery is demonstrated and maintains significance when considered alongside other influential variables. Timing of surgery has no influence on survival impact in our series. Conflict of interest None declared. References 1. Jemal A, Siegel R, Xu J, Ward E. Cancer statistics 2010. CA Cancer J Clin 2010;60:277e300. 2. Demicheli R, Retsky MW, Hruhesky WJM, Baum M, Gukas ID. The effects of surgery on tumor growth: a century of investigations. Ann Oncol 2008;19:1821e8. 3. Flanigan RC, Salmon SE, Blumenstein BA, Bearman SI, Roy V, McGrath PC, et al. Nephrectomy followed by interferon alfa-2b compared with interferon alfa-2b alone for metastatic renal-cell cancer. N Engl J Med 2001;345:1655e9.

P value Inclusion criteria (surgery vs no surgery) 0.091 (NS)

eStage IV

<0.0001

- Stage IV - None-mild-moderate comorbidity - Follow-up>90 days - Ductal or lobular carcinoma eStage IV

<0.0001 < 0.0001

0.52 (NS) 0.06 (NS) 0.093 (NS) 0.0015 0.004 0.005

- Stage IV - Matched pair anaylsis for metastatic site (n ¼ 100 for visceral disease and n ¼ 168 for bone disease) - Stage IV - Metastasis within 6 months of diagnosis eStage IV -Stage IV (diagnosis of stage IV within 30 days of surgery) - Stage IV -

Stage IV (excluded supraclavicular disease only) Surgery < or ¼ 30 months from initial diagnosis Stage IV Age < or ¼ 80 years Charlson score 0e2 PS 0e2 Follow-up>90 days

< 0.001

-Stage IV

0.0002

- Stage IV - Excluded patients registered the same day of death

<0.001

- Stage IV - Excluded patients with more than one record and if stage IV diagnosis was after the first primary diagnosis -Stage IV

<0.0001

4. Rafii A, Deval B, Geay JF, Chopin N, Paoletti X, Paraiso D, Pujade-Laurraine E. Treatment of FIGO stage IV ovarían carcinoma: result of primary surgery or interval surgery after neoadjuvant chemotherapy: a retrospective study. Int J Gynecol Cancer 2007;17:777e83. 5. Ruiterkamp J, Voogd A, Bosscha K, Tjan-Heijnen VC, Ernst MF. Impact of breast surgery on survival in patients with distant metastases at initial presentation: a systematic review of the literature. Breast Cancer Res Treat 2010;120:9e16. 6. Khan S. Primary tumor resection in stage IV breast cancer: consistent benefit or consistent bias? Ann Surg Oncol 2007;14:3285e7. 7. Bafford A, Burstein H, Barkley C, Smith BL, Lipsitz S, Iglehart JD, et al. Brest surgery in stage IV breast cancer: impact of staging and patient selection on overall survival. Breast Cancer Res Treat 2009;115:7e12. 8. Charlson ME, Pompei P, Ales K, McKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373e83. 9. Fields R, Margenthaler JA. Surgical resection of the primary tumor in stage IV breast cancer patients: is a randomized, controlled trial imperative or too costly? J Surg Oncol 2009;99:85e6. 10. Danna E, Sinha P, Gilbert M, Clemens VK, Pulaski BA, Ostrand-Rosenberg S. Surgical removal of primary tumor reverses tumor-induced immunosuppression despite the presence of metastatic disease. Cancer Res 2004;64:2205e11. 11. Babiera GV, Rao R, Feng L, Meric-Bernstam F, Kuerer HM, Singletary HE, et al. Effect of primary tumor extirpation in breast cancer patients who present with stage IV disease and an intact primary tumor. Ann Surg Oncol 2006;13:776e82. 12. Blanchard DK, Shetty PB, Hilsenbeck SG, Elledge RM. Association of surgery with improved survival in stage IV breast cancer patients. Ann Surg 2007;247:732e8. 13. Fields RC, Jeffe DB, Trinkaus K, Zhang Q, Arthur C, Aft R, et al. Surgical resection of the primary tumor is associated with increased long-term survival in

554

14.

15.

16.

17.

18.

J.A. Pérez-Fidalgo et al. / The Breast 20 (2011) 548e554 patients with stage IV breast cancer after controlling for site of metastasis. Ann Surg Oncol 2007;14:3345e51. Cady B, Nathan NR, Michaelson JS, Golshan M, Smith BL. Matched pair analyses of stage IV breast cancer with or without resection of primary breast site. Ann Surg Oncol 2008;15:3384e95. Hazard HW, Gorla SR, Scholtens D, Kiel K, Gradishar VJ, Khan SA. Surgical resection of the primary tumor, chest wall control, and survival in women with metastatic breast cancer. Cancer 2008;113:2011e9. Leung AM, Vu HN, Nguyen KA, Thacker LR, Bear HD. Effects of surgical excision on survival of patients with stage IV breast cancer. J Surg Res; 2009. doi:10.1016/j.jss.2008.12.030. McGuire KP, Eisen S, Rodriguez A, Meade T, Cox CE, Khakpour N. Factors associated with improved outcome after surgery in metastatic breast cancer patients. Am J Surg 2009;198:511e5. Neuman HB, Morrogh M, Gonen M, Van Zee KJ, Morrow M, King TA. Stage IV cancer in the era of targeted therapy, does surgery of the primary tumor matter? Cancer 2010;116:1226e33.

19. Khan SA, Stewart AK, Morrow M. Does aggressive local therapy improve survival in metastatic breast cancer. Surgery 2002;132:620e7. 20. Gnerlich J, Jeffe DB, Deshpande AD, Beers C, Znder C, Margenthaler JA. Surgical removal of the primary tumor increases overall survival in patients with metastatic breast cancer: analysis of the 1988e2003 SEER data. Ann Surg Oncol 2007;14:2187e94. 21. Ruiterkamp J, Ernst MF, van de Poll-Franse LV, Bosscha K, Tjan-Heijnen VC, Voogd AC. Surgical resection of the primary tumour is associated with improved survival in patients with distant metastatic breast cancer at diagnosis. Eur J Surg Oncol 2009;35:1146e51. 22. Rapiti E, Verkooijen HM, Vlastos G, Fioretta G, Neyroud-Caspar I, Sappino AP, et al. Complete surgical excision of primary breast tumor improves survival of patients with metastatic breast cancer at diagnosis. J Clin Oncol 2006;24: 2743e9. 23. Rao R, Feng L, Kuerer H, Singletary SE, Bedrosian I, Hunt KK. Timing of surgical intervention for the intact primary in stage IV breast cancer. Ann Surg Oncol 2008;15:1696e702.