Among women presenting with de novo stage IV breast cancer, 35% to 60% undergo local therapy, presumably to avoid uncontrolled chest wall disease. Several studies suggest that resection of the primary tumor may prolong survival, but chest wall outcome data are notably lacking. The authors reviewed chest wall status, time to first progression (TTFP), and overall survival (OS) in this group of women.
Women presenting at the Lynn Sage Breast Center (1995-2005) with an intact primary tumor and stage IV breast cancer or postoperative diagnosis of distant metastases were identified. Logistic regression and Cox proportional hazards models, adjusted for relevant covariates, were used to examine associations between surgical treatment and chest wall status, TTFP, and OS.
Of 111 eligible women, 47 (42%) underwent early resection of the primary tumor. Chest wall status was available for 103 women. Local control was maintained in 36 of 44 (82%) patients in the surgical group versus 20 of 59 (34%) patients without surgery (P = .001). TTFP was prolonged in the surgical group (adjusted hazards ratio [HR], 0.493; P = .015). The adjusted HR for OS in the surgical group was 0.798 (P = .520). Chest wall control was associated with improved OS regardless of whether surgical resection of the tumor was performed (HR, 0.415; P < .0002).
The Surveillance Epidemiology and End Results database has reported a 5% incidence of stage IV breast cancer for white women, and a 9% incidence for African American women during the period 1996 to 2003.1 Therefore, approximately 11,000 women presented with stage IV disease in 2007. The traditional approach to de novo metastatic breast cancer is based on the rationale that survival depends upon the metastatic disease burden and local therapy will not impact overall survival (OS). Thus the therapeutic mainstay has been systemic therapy, with locoregional treatment reserved for palliation of symptomatic disease. In 2002, we published an analysis of data from the National Cancer Database of the American College of Surgeons, intended to examine patterns of care and outcomes in this group of women.2 Among 16,024 women presenting with stage IV disease between 1990 and 1993, surgical resection of the primary tumor occurred in 58% of women, and was associated with prolonged survival, with a hazards ratio (HR) of 0.61 (95% confidence interval [CI], 0.58-0.65) if negative surgical margins were achieved. These findings, although subject to the inherent biases of a retrospective review, challenged traditional dictums regarding the initial treatment of women presenting with stage IV breast cancer. Recent retrospective studies from single institutions and population-based databases not only support these findings, but also confirm that 35% to 60% of women presenting with metastasis undergo locoregional therapy.3–6 Presumably, the widespread use of surgical resection is driven by the desire to avoid uncontrolled chest wall disease and improve quality of life, because the survival benefit of surgical resection of the primary tumor in stage IV disease remains unproven. However, data supporting improved chest wall control with surgical resection are lacking; a single retrospective study examining local control in the setting of metastatic disease included 20 women.7
To investigate the role of surgical resection in the control of chest wall disease among women with stage IV breast cancer, we performed a retrospective study of patients treated at Northwestern Memorial Hospital (NMH) between 1995 and 2005. We sought to determine whether surgical intervention affected chest wall outcome, time to progression, and OS. In addition, because any effect of surgical therapy on OS would logically be mediated by improved chest wall control, we examined the impact of locoregional control on OS.
MATERIALS AND METHODS
After institutional review board approval, the Lynn Sage Database and the NMH Tumor Registry were queried for women presenting with stage IV breast cancer and an intact primary tumor or women diagnosed with breast cancer and found to have metastasis within 6 months of diagnosis (radiographic or biopsy-proven evidence of disease outside the breast and axilla). Preoperative staging in clinical stage I-II disease consisted of routine laboratory tests and a chest radiograph. Bone scintigraphy and computed tomography surveillance was performed postoperatively when a high axillary lymph node burden was documented pathologically. The presence of symptoms related to distant sites of involvement was not a selection criterion for the surgical and nonsurgical groups. There were patients with distant symptoms at presentation in both groups. However, we were not convinced that symptom information was reliably captured in all patients, and therefore did not include metastatic symptoms in the analysis.
Medical records were reviewed for the age at diagnosis, clinical and pathologic size of the primary tumor, lymph node status, hormone receptor status, HER-2/neu overexpression, location of metastasis(es), mode of surgical treatment, margin status, use of radiotherapy, mode of systemic therapy, time to first progression (TTFP), and chest wall status at the time of death or last contact. TTFP was defined as the interval between initiation of systemic therapy and the first change in therapy because of disease progression, either local or distant. The chest wall status was designated “free” if chest wall/breast evaluation at last follow-up or death revealed no evidence of disease; “asymptomatic” if an intact breast tumor was present on evaluation but was asymptomatic (no pain or skin involvement); and “symptomatic” if a fungating growth was present, or if the skin was involved. Twenty-eight women who met the eligibility criteria were excluded: 14 died within 120 days of diagnosis, 5 refused all treatment of their disease, 2 were concomitantly diagnosed with another primary cancer, 1 had metastatic disease consistent with breast cancer without an identified primary breast lesion, and no follow-up information was available for 6 patients. Patients in the surgical group underwent resection of the primary tumor at diagnosis, or following significant clinical response to systemic treatment. Those who did not undergo resection of the primary tumor, or underwent palliative (toilette) mastectomy because of later progression of local disease while on systemic therapy, were included in the nonsurgical group.
Baseline characteristics between patients who received surgery and those who did not were compared using Student t test with pooled variance for continuous variables and Fisher exact test for categoric variables. Variables that were found to differ between the 2 groups at 2-sided P value <.05 were adjusted for in multivariate models for chest wall, TTFP, and OS outcomes.
Logistic regression was used to evaluate the association between surgical treatment and chest wall status. Univariate logistic regression models were used to examine associations between potential covariates other than surgical treatment and chest wall outcome, and any predictors demonstrating statistical significance at 2-sided P value <.05 were adjusted for in a multivariate logistic regression model in which surgical treatment status was the main predictor of interest. In addition, hormone receptor status, number of metastatic locations, bone versus visceral metastasis, and age were controlled for in multivariate analyses because of their previously noted association with survival.
Cox proportional hazards models were used to evaluate the association between surgical treatment and both TTFP and OS outcomes. Univariate Cox models were used to examine associations between potential covariates other than surgical treatment and TTFP and OS time, and any predictors demonstrating statistical significance at 2-sided P value <.05 were adjusted for in a multivariate Cox model in which surgical treatment status was the main predictor of interest. In addition, hormone receptor status, number of metastatic locations, and bone versus visceral metastasis were controlled for in multivariate analyses because of their previously noted association with TTFP and OS time.8, 9 The proportionality assumption for these models was confirmed by testing for interactions with time for the covariates used in the model. Kaplan-Meier plots were used to demonstrate the difference in first progression and survival times stratified by surgical treatment status.
Of the 139 women presenting to NMH between 1995 and 2005 with stage IV disease and an intact primary tumor, or metastatic disease diagnosed postoperatively, 111 met inclusion and exclusion criteria. The clinical characteristics of this population are shown in Table 1. The surgical group was younger (P = .033), had a higher proportion of hormone-receptor–negative tumors (P = .0005), and were more likely to undergo local radiotherapy (P < .001) than the nonsurgical group. There was a trend toward fewer metastatic sites in those undergoing surgery (P = .057). Clinical tumor size, lymph node involvement, histology, grade, HER-2/neu overexpression, and metastatic disease type (osseous/soft tissue vs visceral) were similar between the 2 groups.
Table 1. Patient Characteristics
All Patients (N=111)
Nonsurgical Group (n=64)
Surgical Group (n=47)
P Comparing Nonsurgical and Surgical Groups
SD indicates standard deviation.
Mean age (SD), y
Clinical tumor size, cm
Positive lymph nodes
Hormone receptor positive
Site of metastasis
No. of metastatic sites
All women received systemic therapy; first-line therapy consisted of an endocrine agent in 34 women, an anthracycline-based regimen in 18 women, an anthracycline plus taxane-based regimen in 45 women, a high-dose regimen with autologous bone-marrow support in 8 women, and miscellaneous combinations in 6 women. Trastuzumab was added to taxane-based regimens in 7 women, when indicated based on HER-2 status during the latter part of the study period.
Surgical resection of the primary tumor was performed in 47 of 111 (42%) women. Of these, 26 women were diagnosed with stage IV disease postoperatively, almost all of whom (23 patients) underwent metastatic surveys because of 4 or more pathologically positive lymph nodes in the surgical specimen (mean number of positive lymph nodes, 13.8; range, 4 lymph nodes-32 lymph nodes). In the remaining 3 women, metastases were confirmed 1 to 5 months postoperatively (lung, liver, and brain, respectively). In the surgical group, systemic therapy was delivered preoperatively in 24 (51%) women. Seventeen (36%) women had partial mastectomy, 30 (64%) women had total mastectomy, and 31 (66%) women had axillary lymph node dissection. Of the 31 women undergoing axillary lymph node dissection, this was performed as part of a surgical procedure before the diagnosis of metastases in 26 women; in 2 women it was performed as part of surgery after systemic therapy, and in 3 women in the nonsurgical group it was performed as part of delayed surgical treatment for palliation. Free surgical margins were achieved in 32 of 45 (71%) women, and 2 patients had unknown surgical margins. Patients were more likely to have negative margins with mastectomy (24 of 28 women; 86%) than partial mastectomy (8 of 17 women; 47%) (P = .0083). Locoregional radiotherapy was used in 30 of 45 (67%) of the surgical patients and 18 of 63 (29%) of the nonsurgical patients.
Of the 64 nonsurgical patients, 10 (16%) patients had palliative surgery for chest wall progression. In these 10 patients, the interval between diagnosis and surgery ranged from 7 to 70 months. In the remaining 54 nonsurgical patients, radiotherapy was used to control local progression in 13 (21%) women. Thus, some form of local therapy was necessary for chest wall progression despite systemic therapy in 23 (36%) women. In this subgroup, there does not appear to be a relation between delayed surgery and OS, TTFP, or chest wall status. This is very likely because of the small sample size (10 with delayed surgery, and 54 without).
Chest Wall Disease
Chest wall data were available for 103 of 111 patients (93%); date were unknown in 5 patients in the nonsurgical group and 3 patients in the surgical group. The chest wall/breast was free in 36 of 44 (82%) women who had surgery compared with 20 of 59 (34%) women who did not (adjusted P = .002). Of the nonsurgical group, 10 of 59 (17%) women were maintained with asymptomatic, intact tumors in the breast throughout their course. Symptomatic chest wall disease was less frequent in the surgical group than in the nonsurgical group (8 of 44 [18%] women vs 29 of 59 [49%] women; P = .002). We examined the effect of surgery on chest wall outcome, in which asymptomatic and free chest wall status was combined, and compared with symptomatic chest wall disease in a logistic regression analysis. There was significant protection against the development of symptomatic chest wall disease in the surgical group (adjusted odds ratio [OR], 0.101; 95% CI, 0.03-0.41 [P = .001]). The results of adjustments made for hormone receptor status, location of metastatic disease as well as number of metastatic locations, use of locoregional radiotherapy, and age are shown in Table 2. No other variables were significant in the univariate analyses.
Table 2. Factors Determining the Odds of Uncontrolled Chest Wall Disease: Multivariate Analysis
95% CI indicates 95% confidence interval; ER, estrogen receptor; +, positive; PR, progesterone receptor; −, negative.
Surgery (vs no surgery)
ER+ and/or PR+ (vs ER− and PR−)
Age (per 10 y)
Radiotherapy (vs none)
Bony metastases (vs viscera)
2 metastatic sites (vs 1)
3 metastatic sites (vs 1)
Chest Wall Disease and OS
Women with free chest wall/breast evaluations and those with asymptomatic disease were grouped together and compared with those with symptomatic chest wall disease. Of the 103 women with known chest wall status, 66 (64%) had either no evaluable disease or asymptomatic in-breast tumors at last contact. Of these, 36 patients received surgical resection of the primary tumor as part of the therapeutic plan (surgical group), and 30 patients did not have surgery. Symptomatic chest wall disease was identified in 37 of 103 (36%) women, 8 in the surgical group and 29 in the nonsurgical group. In Cox proportional hazards models, a controlled chest wall, regardless of surgical intervention, was associated with an OS advantage (HR, 0.415; 95% CI, 0.260-0.662 [P = .0002]). The Kaplan-Meier curves seen in Figure 1 are presented by chest wall status rather than surgical intervention. When patients with evaluable disease, both symptomatic and asymptomatic, are grouped together and compared with those with no evaluable breast or chest wall disease, we find that the differences in OS by chest wall status are even larger (data not shown).
Time to First Progression
Data on TTFP were available for 107 of 111 (96%) patients. At last contact, 10 (9%) patients had no disease progression, 9 in the surgical group (9 of 44 patients; 20%) and 1 in the nonsurgical group (1 of 63 patients; 1.6%). In adjusted Cox proportional hazards models, there was a difference between the surgical and nonsurgical groups in TTFP with an HR of 0.493 (95% CI, 0.28-0.87). In this model, hormone receptor status, age, site of metastatic disease, and number of metastatic sites were not significantly associated with TTFP, whereas surgical therapy remained an independently significant predictor of TTFP (Table 3).
Table 3. Overall Survival and Time to First Progression: Cox Proportional Hazards Model
Multivariate hazards ratio adjusted for hormone receptor status, age, radiotherapy use, type of metastatic disease (soft tissue/bone vs visceral), and number of metastatic sites (1 vs 2 and 1 vs 3). No other variables were found to be significant on univariate analyses, and therefore were omitted from this model.
In the surgical group, the estimated median TTFP was 11 months (95% CI, 9 months-25 months), and the estimated probability of experiencing first progression within the first 3 years was 0.73 (95% CI, 0.56-0.83). In the nonsurgical group, the estimated median TTFP was 9 months (95% CI, 7 months-11 months), and the estimated probability of experiencing first progression within the first 3 years was 0.94 (95% CI, 0.84-0.97). The Kaplan-Meier curves seen in Figure 2a illustrate early overlap of the curves with later divergence.
The chest wall was included in the assessment of TTFP. The chest wall was the site of first progression in 7 (15%) women in the surgical group and 15 (23%) women in the nonsurgical group. This was not significantly different (P = .265).
The median follow-up time was 26.9 months (range, 2.5 months-138 months), and 30 patients (27%) were alive at last contact. Of the 81 patients with recorded death outcomes, 30 (37%) were in the surgical group, and 51 (63%) were in the nonsurgical group. In an adjusted Cox proportional hazards model, the HR of surgical versus nonsurgical groups was not significantly different (0.798; 95% CI, 0.40-1.52 [P = .52]). As in the analysis of TTFP, the association between surgery and OS persisted after adjustment for hormone receptor status, age, site of metastasis, and the number of metastatic sites (Table 3).
In the surgical group, the estimated median survival time was 26.3 months (95% CI, 23.4 months-100.7 months), and the 3-year survival was 0.43 (95% CI, 0.31-0.60). In the nonsurgical group, the estimated median survival time was 29.2 months (95% CI, 25.0 months-41.8 months), and the estimated 3-year survival was 0.37 (95% CI, 0.27-0.54). The Kaplan-Meier curves for the surgical and nonsurgical groups are seen in Figure 2b.
Comparison of Surgical Subsets
We looked for differences in outcomes within the surgical group, comparing the subsets of women who were diagnosed with metastases postoperatively to those diagnosed preoperatively. There were no significant differences between these 2 subsets in terms of tumor size (P = .363), OS (P = .108) or TTFP (P = .208). Chest wall outcome was compared using Fisher exact test, and there was not a statistically significant difference between the 2 groups with a P value of .436; 5 of 20 preoperatively diagnosed patients and 3 of 24 postoperatively diagnosed patients had symptomatic chest wall disease at last contact or death.
Despite the standard recommendation to treat stage IV breast cancer primarily with systemic therapy, there is widespread use of surgical resection for the treatment of the primary tumor in women with metastatic breast cancer,2, 4, 5 presumably with the goal of preventing uncontrolled chest wall disease. However, there is a surprising dearth of published data as to the frequency of uncontrolled chest wall disease, a parameter that cannot be examined in large database reviews. In addition, to our knowledge, none of the single-institution retrospective analyses demonstrating an association of primary tumor resection and improved survival in women with metastatic disease have addressed this issue.3, 10–12 The importance of chest wall control data clearly stems from the negative impact of an uncontrolled chest wall on quality of life, regardless of any survival considerations. In our original study,2 we demonstrated that the association of surgical resection and prolonged survival was strongest when the resection margins were tumor free, consistent with the idea that surgery may in fact be beneficial in the setting of distant disease. In the present study, we take this logic a step further and hypothesize that the possible benefit afforded by surgical resection of the primary tumor is related to a greater likelihood of maintaining a disease-free chest wall in the surgical group. To this end, we have assembled the largest series of chest wall outcomes in women with de novo stage IV breast cancer reported to date, with detailed examination of chest wall outcome relative to time to progression and survival.
We used a conservative definition of “uncontrolled chest wall disease” and assigned women with asymptomatic local disease (ie, a silent tumor in an intact breast without skin invasion) to the “chest wall controlled” group, rationalizing that these were women whose tumor was controlled to some extent by systemic therapy. Analysis of the data stratified by chest wall status rather than surgical therapy indeed supports this hypothesis, with a very significant decrease in the hazard of death among women with a controlled chest wall regardless of whether surgical therapy was used (HR, 0.415; 95% CI, 0.26-0.66). We submit that these findings strengthen the overall concept that adequate local control is beneficial, even in the setting of metastatic disease, and again raise the question of whether or not metastatic foci are perpetuated by continual shedding of tumor stem cells from an uncontrolled primary site. We were unable to examine the impact of surgical margins on outcomes in the present study, because only 13 women underwent resection with involved margins.
We found that the frequency of symptomatic chest wall disease was significantly lower in women who underwent resection of their primary tumor early in their clinical course than in those who did not receive early surgical therapy for their primary tumor, with an adjusted OR of 0.139 (95% CI, 0.039-0.491). We also observed a significant improvement in TTFP with surgical resection of the primary tumor, with an adjusted HR of 0.500 (95% CI, 0.3-0.8), similar to another single-institution study,3 and found that a lower proportion of women in the surgical group experienced disease progression during the first 3 years (73%) than in the nonsurgical group (94%). The improvement in OS in the surgical group was not statistically significant, and our adjusted HR for survival of 0.798 was higher than in other studies.2, 4, 5, 10 This difference from published retrospective studies is very likely related to the smaller size of the present study, but may also be caused by different selection patterns for the use of surgical therapy in women with stage IV breast cancer. We were unable to examine breast-cancer–specific mortality, but in a metastatic disease setting it is likely that most (if not all) deaths are caused by breast cancer; comorbidity is not a strong determinant of survival, and most women die of their disease.10, 13 The examination of all-causes mortality is also a more unbiased way of examining survival data.
Approximately half of the surgical group (26 women) were diagnosed with metastases postoperatively, similar to previous studies.3, 4, 7 These women almost certainly had metastatic disease at presentation, but were diagnosed postoperatively by metastatic surveys prompted by a finding of ≥4 pathologically involved lymph nodes. Although the inclusion of these women introduces the possibility that they have an inherently better prognosis than those presenting with metastases preoperatively, there were no significant differences between patients diagnosed with metastases preoperatively or postoperatively with regard to OS, TTFP, chest wall status, or tumor size. This group has also been included in previous retrospective analyses of de novo stage IV breast cancer.3, 4, 10
We attempted to examine the impact of radiotherapy on local control, but found that the use of radiotherapy was extremely varied, being used before surgery, after surgery, and in the absence of surgery. Although we did not note any significant differences in outcome with the use of radiotherapy, our study lacked the power to address this issue. This remains an important question; if local control provides benefit to women with metastatic breast cancer, it is likely that such a benefit would be consolidated by the addition of radiotherapy. Similarly, the number of women not undergoing axillary dissection in the surgical group was only 16, so that the effect of this on outcomes could not be assessed.
In the nonsurgical group, 36% of women demonstrated local progression. To our knowledge there are no published data regarding this outcome in women with de novo stage IV disease, but in the analogous situation of in-breast recurrence with synchronous metastases, the reported fraction of those progressing locally in the absence of surgical resection is very similar (32%).14
There are inherent biases to retrospective studies that may not be fully overcome with adjusted analyses.15, 16 We found no significant differences between our surgical and nonsurgical groups in most tumor characteristics, site and number of metastatic lesions, and the use of systemic therapy. Similar to other studies, our surgical group was younger,4, 5, 10 was more likely to have hormone-receptor–negative disease and to have received locoregional radiation, and displayed a trend toward fewer metastatic sites. Our study groups differed with regard to estrogen receptor/progesterone receptor status, number of metastatic sites, use of radiation therapy, and age. We adjusted all the analyses for these variables, and observed differences in outcomes by surgical group that were statistically independent. However, it is not possible to ascertain all sources of selection bias in a retrospective study, and there are likely other differences that we were unable to measure that may have contributed to differences in outcome. Nevertheless, the data presented here add to the considerable mass of retrospective data that have been reported in recent years pointing to the possible salutary effect of local therapy in women presenting with stage IV breast cancer.2–5, 10 We now extend these observations by demonstrating a logical (and very plausible) link between local control and survival outcome in metastatic disease, which appears to be of at least similar magnitude to that observed in earlier stages of breast cancer (ie, the prevention of local recurrences translates into long-term survival in a subset of women).17 To restate the hypothesis we have presented previously,2 it is time to question the assumption that local therapy is important for survival only in women with locoregional disease, and is irrelevant in women with detectable metastatic disease. The alternative proposition, that the course of breast cancer may in fact represent a continuum, where local disease, when present, is a hazard for survival in women with both detectable and undetectable metastases, deserves serious consideration. To our knowledge, there are now at least 5 different studies demonstrating an association between resection of local disease and improved survival in women with metastatic breast cancer16; however, all of the reported data are retrospective, and selection bias cannot be ruled out as an explanation for these findings.15 To our knowledge, none of the published studies have addressed the role of radiotherapy or axillary lymph node dissection in the local management of the primary tumor, which in fact cannot be addressed using retrospective data, because the biases that go into a 3-part decision (local tumor resection, axillary lymph node dissection, radiotherapy) will be impossible to adjust for meaningfully in a retrospective setting, no matter how large the dataset.
Our findings demonstrate that the possible impact of local therapy on survival is mediated through the achievement of improved local control, and provide a link between data indicating that local therapy may be helpful in women with metastatic breast cancer, and the well-documented concept that local control translates into improved survival.17 Because local therapy is the most reliable means of achieving local control, we hope that our findings will further support the need for a randomized trial examining the benefit of early local therapy of the primary breast lesion in women presenting with metastatic disease, a question that is also pertinent to the treatment of an in-breast recurrence presenting with synchronous distant metastases.14 A trial randomizing women who present with metastatic breast cancer and an in-breast tumor to early locoregional therapy (after response to primary systemic therapy) versus delayed locoregional therapy (in those who progress locally on systemic therapy) is required to settle this question. An alternative design approach would be to randomize women with de novo stage IV breast cancer who have not undergone tumor resection to first-line local therapy versus systemic therapy. However, we believe that the current standard of care (primary systemic therapy for patients with distant metastases) needs to be adhered to in the design of a randomized trial. The experimental treatment (local therapy) should be added to the standard of care once it is clear that the distant disease is stable or responsive to systemic therapy. If a randomization to systemic therapy or local therapy is performed at the outset, the initiation of the standard of care treatment (systemic therapy) would be delayed. Also, it does not appear plausible to us that women with resistant distant disease could benefit from local therapy in the presence of metastases, and therefore these women should logically be excluded from such a trial.
Such a trial will address the pressing question of whether women with clinically detectable metastases should receive full locoregional therapy; and will also provide data regarding appropriate selection of patients for such therapy.