Sites of distant recurrence and clinical outcomes in patients with metastatic triple-negative breast cancer

High incidence of central nervous system metastases




The purpose of the current study was to characterize the outcomes of patients with metastatic triple-negative breast cancers, including the risk and clinical consequences of central nervous system (CNS) recurrence.


Using pharmacy and pathology records, a study group of 116 patients who were treated for metastatic triple-negative breast cancer at Dana-Farber Cancer Institute between January 2000 and June 2006 was identified.


The median survival from time of metastatic diagnosis was 13.3 months. Sixteen patients (14%) were diagnosed with CNS involvement at the time of initial metastatic diagnosis; overall, 46% of patients were diagnosed with CNS metastases before death. The median survival after a diagnosis of CNS metastasis was 4.9 months. The age-adjusted and race-adjusted rate of death for patients whose first presentation included a CNS metastasis was 3.4 times (95% confidence interval, 1.9-6.1 times) that of patients without a CNS lesion at the time of first metastatic presentation. Of the 53 patients who developed brain metastases, only 3 patients were judged to have stable or responsive systemic disease in the face of progressive CNS disease at the last follow-up before death.


Triple-negative breast cancer is associated with poor survival after recurrence. CNS recurrence is common, but death as a direct consequence of CNS progression in the setting of controlled systemic disease is uncommon. Thus, it does not appear that the high rate of CNS involvement is because of a sanctuary effect, but rather is due to the lack of effective therapies in general for this aggressive subtype of breast cancer. New treatment strategies are needed. Cancer 2008. © 2008 American Cancer Society.

For years, clinicians have appreciated that breast cancer is not 1 disease, but a family of diseases, each with a distinct natural history and response to treatment. More recently, molecular interrogation of human breast cancers has led to the identification of at least 4 distinct tumor subtypes distinguished by their gene expression profiles: luminal A, luminal B, HER-2-positive, and basal-like.1–3

When patterns of gene expression are analyzed by hierarchical clustering, the basal-like subgroup consistently segregates as a distinct cluster, characterized by expression of cytokeratins 5/6 and 17, and by lack of hormone receptor and HER-2 overexpression.4 Although expression profiling is not routinely performed in a clinical context, approximately 80% of tumors that are estrogen receptor (ER)-negative, progesterone receptor (PR)-negative, and HER-2-negative (ie, ‘triple-negative’) demonstrate the basal-like gene expression pattern. Therefore, the use of these commonly performed markers can act as a reasonable surrogate of the basal-like phenotype.

Previous reports have indicated that patients with early-stage basal-like or triple-negative breast cancers experience reduced disease-free and overall survival compared with other breast cancer subtypes.2, 5, 6 In a retrospective study led by Haffty et al,6 the triple-negative subtype was found to be an independent predictor of distant recurrence and shortened cause-specific survival in patients with early-stage breast cancer. However, it is unclear whether the worse cause-specific survival in this cohort could be solely explained on the basis of the higher distant recurrence rate, or whether the inferior survival of women with triple-negative breast cancer relative to other breast cancer subtypes is also because of a shorter survival after the diagnosis of metastatic disease. In the Cancer and Leukemia Group B (CALGB) 9342 study, women with metastatic breast cancer were randomized to 1 of 3 doses of paclitaxel. Those women with triple-negative disease were found to have a shorter overall survival from study entry compared with all other patients.7

To our knowledge, there are limited data regarding specific sites of metastasis and clinical course over time of patients with triple-negative, metastatic breast cancer. One potential metastatic site of concern is the central nervous system (CNS). Tsuda et al8 compared the clinical features of 20 high-grade carcinomas with large, central acellular zones with those of 40 matched high-grade carcinomas without these zones. The presence of acellular zones appeared to be highly correlated with a myoepithelial immunophenotype, and these tumors were more likely to metastasize to the brain and lung. An apparent increased incidence of CNS metastases has also been observed in African-Americans and BRCA1 mutation carriers, 2 populations of patients who have a relatively higher incidence of triple-negative disease.9–13 Finally, in a study of 55 patients who developed brain metastases, the frequency of ER-negative, cytokeratin 5/6-positive, and epidermal growth factor receptor (EGFR)-positive tumors was higher than that observed in a comparison group of patients who did not have brain metastases.14 As systemic therapies have improved for patients with HER-2-positive metastatic breast cancer, neurologic progression has become a major source of morbidity and mortality.15–17 However, it is unknown whether the same is true of patients with triple-negative cancers, or whether the aggressive nature of their cancers dominates in both CNS and non-CNS sites.

The objective of this retrospective study was to characterize the sites of distant recurrence and clinical outcomes in a cohort of patients with metastatic, triple-negative breast cancer. We further described the clinical course of the subset of patients who developed CNS disease to explore the prognostic significance of CNS recurrence.


Approval for this study was obtained from the Institutional Review Board of the Dana-Farber/Harvard Cancer Center (DF/HCC). Using computerized order entry systems, patients with breast cancer who were treated at the Dana-Farber Cancer Institute (DFCI) between January 1, 2000, and June 30, 2006 and for whom at least 1 dose of the following chemotherapy drugs was ordered were identified: paclitaxel, docetaxel, liposomal doxorubicin, gemcitabine, vinorelbine, carboplatin, or bevacizumab. Patients whose only chemotherapy included doxorubicin plus cyclophosphamide (AC) or dose-dense AC followed by paclitaxel were excluded because these regimens are rarely administered in the metastatic setting. Patients who had received trastuzumab were also excluded. A total of 1042 patients were identified initially. Pathology records were reviewed to determine the ER, PR, and HER-2 status of the primary tumor and/or metastatic biopsy, if available. Patients were included if the primary tumor was negative for all 3 markers, or if biopsy of a metastatic lesion was negative for all 3 markers. HER-2 was considered negative if stained between 0 and 1+ by immunohistochemistry (IHC) and/or fluorescence in situ hybridization (FISH)-negative (defined as a HER-2:cen17 ratio of <2.0). Patients with any of 3 markers positive were excluded (n = 851), as were patients with missing marker data (n = 6). Patients were then excluded for the following reasons: 1) nonmetastatic breast cancer on further review of medical records (n = 43); 2) unusual breast cancer histology (n = 2, metaplastic carcinoma); 3) a history of second (nonbreast) malignancy (n = 7); or 4) inadequate follow-up (ie, lost to follow-up <6 months after the diagnosis of metastatic disease; n = 17). This left a total of 116 patients for analysis. Follow-up information was available through June 30, 2007.

Medical records from each patient included in this analysis were reviewed for the following information: date and stage of diagnosis of initial breast cancer, date of diagnosis of metastatic disease, initial staging workup at time of metastatic diagnosis, previous chemotherapy regimens, site(s) of initial and subsequent disease recurrence, status of metastatic disease at time of CNS diagnosis (if applicable), therapy for CNS disease, vital status, date of death (or date last seen by physician), status of CNS and non-CNS at time of death, and cause of death. For patients with bilateral breast cancer (n = 7), the date of initial diagnosis was defined as the date of primary breast cancer diagnosis closest to the date of distant disease recurrence. CNS disease was defined as radiographic evidence of parenchymal metastases and/or clinical or radiographic evidence of leptomeningeal metastases. Isolated intraspinal or epidural metastases were not included in the definition of CNS metastases because of the difficulty of distinguishing between true CNS involvement versus local extension from bony disease in a retrospective study. A line of CNS-directed therapy was defined as a discrete plan of therapy given within a consecutive period. For example, a patient who underwent craniotomy followed by whole-brain radiotherapy (WBRT) was classified as having received 1 line of CNS-directed therapy. In contrast, a patient who underwent WBRT followed several months later by stereotactic radiosurgery (SRS) for a progressive lesion was considered to have received 2 lines of CNS-directed therapy. Response was evaluated according to the interpretation of the primary oncologist from findings on physical examination and imaging studies, supplemented by review of radiology reports when available. Attempts were not made to verify response or progression according to Response Evaluation Criteria in Solid Tumors (RECIST). Patients were characterized as responding, having stable disease, or progressing at the time of diagnosis of CNS metastases. In addition, patients were categorized as having been treated at DFCI before the diagnosis of metastatic breast cancer or referred to DFCI after a diagnosis of metastatic breast cancer had been made.

The primary outcome variable was time to death as measured in years. Comparison of cases by descriptor variables was performed using a chi-square or Fisher exact test for discrete variables and a Student t test for continuous variables. Estimates of survival probabilities (with 95% confidence intervals [95% CIs]) were calculated using Kaplan-Meier product limit methodology and compared using a Wilcoxon log-rank test. Hazards rates were computed using a Cox proportional hazards model18 and were adjusted for age and race. All analyses were completed using the SAS statistical software package (SAS Institute, Inc, Cary, NC).19


Description of Study Population

Table 1 illustrates the clinical characteristics for the patient sample. The mean age at diagnosis was 47.5 years (standard deviation of 9.9 years). The majority of patients were white (85%). Approximately 60% of patients presented with stage I or II disease (determined according to the American Joint Committee on Cancer grading system); 11% of patients had evidence of distant disease at the time of initial presentation. As expected, the vast majority of patients had grade 3 invasive ductal carcinomas. Of patients without metastatic disease at the time of the initial presentation, 81% received adjuvant chemotherapy. The BRCA status was known in only 12 patients, 5 of whom were confirmed carriers of deleterious mutations. Approximately two-thirds of patients (75 of 116 patients) had been treated at DFCI before the diagnosis of metastatic disease. The remainder of patients were referred for care after a diagnosis of distant metastasis had been made.

Table 1. Patient and Tumor Characteristics
 Patients (N=116)
  1. A-based indicates anthracycline-based; T-based, taxane-based.

Age at initial diagnosis, y 
Stage at diagnosis 
Lymph node stage at diagnosis 
 Mixed ductal and lobular43
 Carcinoma, type not specified33
Nuclear grade 
BRCA testing performed 
BRCA1 mutation (of n=15 tested) 
 Result unknown3 
Initial surgery  
Types of prior adjuvant chemotherapy exposure 
 Both A-based and T-based4438
 Non-A based, non-T based1614
Adjuvant radiotherapy 

The median length of follow-up was 34.1 months from time of the initial breast cancer diagnosis. At the time of last follow‒up, 104 deaths were recorded, 11 patients were confirmed to be alive, and 1 patient was lost to follow-up. The median time from the initial breast cancer diagnosis to the diagnosis of metastatic disease was 19.9 months. Of those patients who initially presented with stage I through III disease, 75% of recurrences occurred within 3 years of the diagnosis of breast cancer (Fig. 1). Patients received a median of 3 lines of chemotherapy before death; the distribution of regimens is detailed in Table 2. Approximately 30% of patients received bevacizumab; the high degree of usage reflects the existence of two phase 2 trials including this drug that were open during the time period of this retrospective study and the greater use of bevacizumab after the results of the E2100 trial were presented in June 2005.20–22

Figure 1.

Time to metastatic disease by disease stage at the time of the initial presentation.

Table 2. Recurrence Characteristics
 Patients (N=116)
  • CNS indicates central nervous system.

  • *

    Two patients did not receive chemotherapy for metastatic disease: 1 patient whose disease recurred while receiving adjuvant chemotherapy and who refused further chemotherapy, and 1 patient who died before receiving the first planned cycle of chemotherapy.

Disease-free interval 
 Median19.9 mo
 Range0-134 7 mo
Site(s) of initial recurrence/metastasis 
 Breast/chest wall2824
 Soft tissue/distant lymph nodes2522
 Regional lymph nodes1715
Site(s) of initial and subsequent recurrence/metastasis 
 Breast/chest wall3833
 Soft tissue/distant lymph nodes3732
 Regional lymph nodes2219
No. of lines of metastatic chemotherapy 
Metastatic chemotherapy 

The most common sites of involvement at the time of diagnosis of metastatic disease were the lung (41% of patients), liver (29%), bone (24%), and breast or chest wall (22%) (Table 2). In terms of all sites of metastasis (initial + subsequent), lung and liver remained the 2 most common sites. Fifty-three (46%) patients were noted to have a CNS metastasis at some point, with 16 (14%) found to have a CNS metastasis at the time of first metastatic presentation and 4 with an isolated CNS metastasis at the time of the initial metastatic presentation.

Characteristics of Patients With CNS Disease

The majority of patients with CNS disease presented with parenchymal brain metastases; leptomeningeal involvement without parenchymal disease was less common (Table 3). Neurologic symptoms precipitated the initial diagnostic study in 77% of patients; the remainder of patients were found to have asymptomatic CNS metastases at the time of a screening study that was performed to fulfill the eligibility requirements of a clinical trial, before the use of bevacizumab, or as part of routine breast cancer staging. At the time of CNS diagnosis, systemic (ie, non-CNS) disease was stable or responding to therapy in only 9 (17%) patients. In 83% of patients, CNS metastases were diagnosed concurrently with new or progressive systemic metastases. The majority of patients were treated with WBRT, either alone or in conjunction with surgical resection, SRS, or intrathecal chemotherapy. Only 13 (24%) patients received more than 1 line of CNS-directed therapy.

Table 3. Characteristics of CNS Disease (n=53)
 Patients (N=53)
  • CNS indicates central nervous system; WBRT, whole-brain radiotherapy; SRS, stereotactic radiosurgery; IT, intrathecal.

  • *

    Systemic therapy includes hormonal therapy, chemotherapy, or investigational agents.

  • For example, radiation to the lumbar spine for symptomatic leptomeningeal metastases.

Site of CNS metastasis 
 Parenchymal only4075
 Leptomeningeal only815
 Both parenchymal and leptomeningeal59
Reason for diagnostic study 
 Neurologic symptoms4177
 Prior to clinical trial or bevacizumab treatment611
 Staging of breast cancer in asymptomatic patient611
Systemic disease status at time of CNS diagnosis 
No. of lines of treatment for CNS disease 
CNS-directed therapy 
 IT chemotherapy611
 Systemic therapy*48
 Other RT48

We attempted to ascertain the cause of death of patients with CNS metastases. Because of the retrospective nature of the current study, we were unable to assign a cause of death for 11 of 53 patients with CNS metastases who died during the study period. Of the remaining 42 patients, 12 died primarily of systemic disease progression, 11 died primarily of CNS disease progression, 17 died of both systemic and CNS progression, and 1 patient died of other causes (infection). We also characterized patients according to the status of CNS and non-CNS disease at the time of last follow-up before death. When analyzed in this way, only 3 patients were judged to have stable or responsive systemic disease in the face of progressive CNS disease.

Prognostic Factors for Overall Survival

The median overall survival from time of metastatic diagnosis was 13.3 months. Survival did not significantly differ by age, race, initial disease stage, or time of presentation to DFCI. Patients who were ever diagnosed with a CNS metastasis at any time did not differ from those without such lesions by age, race, stage of disease, time of presentation to DFCI, or overall survival. Patients whose first metastatic presentation included a CNS lesion also did not differ with respect to any of the demographic or disease characteristics described above, but they did have shorter survival time than persons whose first metastasis did not include a CNS lesion (P < .0001, log-rank test). The age-adjusted and race-adjusted rate of death for patients whose first presentation included a CNS metastasis was 3.4 times (95% CI, 1.9-6.1 times) that of patients without a CNS lesion at the time of first metastatic presentation. The 1-year, 2-year, and 3-year survival rates for patients without a CNS lesion at the time of initial presentation were 61.6% (standard error [SE] of 4.9%), 21.8% (SE of 4.3%), and 14.4% (SE of 3.8%), respectively, versus 18.8% (9.8%), 0%, and 0% for patients with CNS lesions at the time of presentation (Fig. 2). Among the 53 patients with CNS metastases, the median survival from time of diagnosis of first metastasis (CNS or non-CNS) was 11.6 months overall and 4.9 months from time of first CNS metastasis. Among those with a CNS lesion, survival did not vary by sex, race, stage of disease, or time of presentation to DFCI.

Figure 2.

Time to death by the presence/absence of central nervous system (CNS) metastasis at the time of first metastatic presentation (P < .001, log-rank test).


In this retrospective, single-institution study, we found that the poor outcome of patients with triple-negative breast cancer persists in the metastatic setting. The median overall survival was only 13.3 months from the time of metastatic diagnosis, which is far shorter than the median survival reported for all-comers with metastatic breast cancer in population-based studies.23 We also observed a strikingly high rate of CNS metastases in these patients. Patients with CNS involvement at the time of their initial metastatic diagnosis fared particularly poorly. In addition, among the 53 patients with CNS metastases at any time during their disease course, the median survival from the time of CNS diagnosis was only 4.9 months.

Historical series indicate a rate of CNS metastasis of approximately 10% to 16% in women with advanced breast cancer.9, 24, 25 Recently, Fulford et al26 published the results of a retrospective study examining the outcomes of patients with cytokeratin 14-positive tumors (a marker for myoepithelial/basal cells). Of the 443 tumors analyzed, 20% were cytokeratin 14-positive, and of these, 37 were associated with metastatic recurrences. When compared with patients with cytokeratin-negative tumors, patients with cytokeratin 14-positive tumors were more likely to develop brain metastases (hazards ratio [HR] of 1.92; P = .05). Our data support and extend these findings: patients with metastatic disease from triple-negative or basal tumors appear to be at much higher risk for CNS involvement than unselected historical controls.

To our knowledge to date, HER-2 positivity is the single risk factor that has been most prominently associated with an increase in CNS disease. We noted several similarities and differences in the natural history of patients with HER-2-positive brain metastases, as described in the literature, and in our study of patients with triple-negative tumors. First, as in patients with triple-negative breast cancer, CNS involvement is common among patients with advanced, HER-2-positive tumors, with approximately one-third of patients diagnosed with brain metastases before death.15, 27–29 This increase compared with unselected historical controls has been attributed to the combination of a biologic predisposition to CNS spread and a change in the natural history related to trastuzumab, which has afforded many women prolonged disease control outside the brain.30, 31 For example, in the series by Bendell et al,15 half of the women had stable or responsive systemic disease at the time of CNS diagnosis. In contrast, we found that in the triple-negative population, the diagnosis of CNS metastasis was typically made in the context of new and/or progressive systemic metastases and rarely occurred in the setting of stable systemic disease. Thus, it does not appear that the CNS involvement we observed is the result of a chemotherapy sanctuary effect. Rather, the data from the current study suggest that triple-negative tumors are particularly prone to CNS spread. The median survival after CNS diagnosis has been described in several retrospective studies as more favorable in HER-2-positive patients compared with HER-2-negative patients.32, 33 Many, though not all, studies have indicated a longer-than-expected median survival from CNS diagnosis, ranging from 13 months to 23.5 months, in patients with HER-2-positive breast cancer treated in the post-trastuzumab era.32–35 There is a clinical need for salvage therapies after initial radiotherapy in this patient population, and it has been estimated that as many as half of these patients may die primarily of refractory CNS disease. Unfortunately, the results of the current study, and of a recent study led by Nam et al,36 indicate that the prognosis for patients with CNS metastases from triple-negative breast cancer is almost uniformly poor. In addition, we found that death from isolated CNS progression was relatively uncommon. These data, if confirmed, point to the need for improvements in systemic therapy for triple-negative breast cancer in general. At the same time, close attention to the CNS will be required to avoid the problem of isolated CNS progression, which is now being observed in the HER-2-positive population.

The strengths of the current study are that we examined sites of both initial and subsequent metastasis and clinical outcomes in a large group of women with metastatic, triple-negative breast cancer who were treated within a relatively narrow time period with modern chemotherapy regimens. The majority of patients had received adjuvant chemotherapy, and approximately one-third of patients received bevacizumab-based therapy in the metastatic setting. We had access to detailed clinic records and radiology reports. In all but 1 patient, we were able to document the use of chest, abdominal, and pelvic computed tomography (CT) scans and/or positron emission tomography (PET)-CT as part of the initial staging; thus, the characterization of the initial sites of disease is likely to be accurate. Although there was an increasing use of ‘screening’ scans of the brain during this time period, largely reflecting the existence of clinical trials of angiogenesis inhibitors, the diagnosis of brain metastases still occurred as a result of neurologic symptoms in 77% of patients. All but 12 patients were followed until death.

The current study has several limitations. Patients were treated at a single academic medical center, thus raising the possibility of referral bias. However, we did not observe any interaction between time of presentation to DFCI and either CNS recurrence or overall survival. Next, consistent with clinical practice, the assignment of triple-negative phenotype was based on the characteristics of the primary tumor in the vast majority of cases. Recent reports have indicated that discordance in hormone receptor and HER-2 status between primary and metastatic lesions may occur.37, 38 Because the majority of patients in the current study did not undergo a biopsy of metastatic disease, we cannot rule out the possibility that some misclassification may have occurred. Finally, because patients were identified through pharmacy chemotherapy records, patients could enter the study population at any time during the course of their metastatic disease. As a result, this is not a true cohort study, which would require all patients to enter the cohort at the time of metastatic diagnosis. For this reason, we also did not include a comparison arm because there could be a bias introduced if the time interval during which chemotherapy was administered was held constant, given differences in the typical timing of chemotherapy initiation for metastatic disease based on tumor subtype. In addition, we were not able to calculate time to disease progression for each line or type of chemotherapy. To address some of these concerns, and based in part on results from this study, we are in the process of analyzing sites of metastasis and clinical outcomes in a much larger cohort of patients dichotomized by triple-negative status.

In contrast to the current dilemma with HER-2-positive disease, in which the CNS appears to be an important sanctuary site, in this study patients with triple-negative, metastatic breast cancer rarely experienced isolated CNS progression. In addition, death from CNS progression in the setting of controlled systemic disease was uncommon, and survival among patients with CNS involvement was uniformly short. These data underscore the need to develop better treatments for patients with triple-negative breast cancer, regardless of the presence or absence of brain metastases. Furthermore, any efforts to improve treatments for CNS disease in this patient population will need to consider the limited treatment options for extra-CNS disease and the finding that the majority of patients with CNS involvement also have disease progression at other sites.