|Home | About | Journals | Submit | Contact Us | Français|
We investigate the variance in patterns of failure after Gamma Knife™ radiosurgery (GKRS) for patients with brain metastases based on the subtype of the primary breast cancer. Between 2000 and 2010, 154 breast cancer patients were treated with GKRS for brain metastases. Tumor subtypes were approximated based on hormone receptor (HR) and HER2 status of the primary cancer: Luminal A/B (HR+/HER2(−)); HER2 (HER2+/ HR(−)); Luminal HER2 (HR+/HER2+), Basal (HR(−)/ HER2(−)), and then based on HER2 status alone. The median follow-up period was 54 months. Kaplan–Meier method was used to estimate survival times. Multivariable analysis was performed using Cox regression models. Median number of lesions treated was two (range 1–15) with a median dose of 20 Gy (range 9–24 Gy). Median overall survival (OS) was 7, 9, 11 and 22 months for Basal, Luminal A/B, HER2, and Luminal HER2, respectively (p = 0.001), and was 17 and 8 months for HER2+ and HER(−) patients, respectively (p <0.001). Breast cancer subtype did not predict time to local failure (p = 0.554), but did predict distant brain failure rate (76, 47, 47, 36 % at 1 year for Basal, Luminal A/B, HER2, and Luminal HER2 respectively, p <0.001). An increased proportion of HER2+ patients experienced neurologic death (46 vs 31 %, p = 0.066). Multivariate analysis revealed that HER2+ patients (p = 0.007) independently predicted for improved survival. Women with basal subtype have high rates of distant brain failure and worsened survival. Our data suggest that differences in biologic behavior of brain metastasis occur across breast cancer subtypes.
Survival for patients who develop brain metastases has improved significantly over time due to earlier detection , better brain-directed therapies , and improved systemic therapies . Advances in central nervous system (CNS)-directed therapies for metastatic brain disease, including CNS-penetrating systemic agents, surgery, radiosurgery and stereotactic radiotherapy have improved control of brain disease to the degree that many patients with brain metastases will ultimately die from their systemic disease [4, 5]. Because of improving systemic therapies and the relative responsiveness of some breast cancers to systemic therapies, breast cancer patients comprise a large proportion of the long term survivors of brain metastases .
Recent advances in molecular profiling have demonstrated that breast cancers are a genetically heterogeneous population and can be classified into subtypes that yield prognostic value . While widespread molecular classification is not currently feasible in the clinical setting, subtypes derived from immunohistologic criteria have allowed for individualized and targeted therapies, and have led to improved survival in patients with certain profiles [8, 9]. Up to one-third of patients with advanced HER2 positive breast cancers develop CNS disease [4, 5], while approximately 25–46 % of basal subtype patients (estrogen and progesterone receptor (−), HER2(−)) develop brain metastases .
Although there is greater concern for the occurrence of brain metastases in women with HER2 positive breast cancer and basal subtype, there may be improved survival when the event occurs in HER2 positive breast cancers . Therefore, a question has emerged as to whether the biologic behavior of brain metastasis in the different breast cancer subtypes differs. To address this question, we present a single institution retrospective review evaluating the clinical outcomes of patients with brain metastases from breast cancer after Gamma Knife™ stereotactic radiosurgery (GKRS), emphasizing how the subtype of the primary breast cancer affects these outcomes. We evaluated patterns of failure based on breast cancer subtype, as well as factors predicting for overall survival (OS), patterns of failure, and the lifetime likelihood of neurologic death among these cohorts.
This study was approved by the Institutional Review Board at our institution. Data was reviewed and collected on 154 women with histologically-confirmed breast cancer who underwent GKRS for brain metastases between January 2000 and September 2010. Electronic medical records were reviewed to determine clinicopathologic characteristics including age, sex, race, date of diagnosis, prior whole brain radiotherapy (WBRT), hormonal receptor (HR) status, HER2 positive status, date of first brain metastasis, date of GKRS, number of lesions treated with radiosurgery, marginal dose and systemic therapy at presentation of brain metastases, specifically trastuzumab and lapatinib. RPA class was determined as described by Gaspar et al. . Subtypes of patients’ primary cancers were approximated into the following subtypes as previously described by Arvold : Basal ((HR) negative/HER2(−)), Luminal A/B (HR+/HER2(−)); HER2 (HER2+/HR(−)); Luminal HER2 (HR+, HER2+), and then based on HER2 status alone. Patient characteristics are summarized in Table 1.
Patients were monitored with serial MRI every three months after initial radiosurgery. Local and distant brain failures were based upon imaging evidence of intracranial recurrence. Local brain failure was defined as either a pathologically-proven recurrence within the GKRS treatment field, or a combination of imaging and clinical characteristics of local treatment failure. Imaging characteristics of treatment failure included an increase in size of enhancement by 25 % and/ or serial increases in size of enhancement with corresponding increased perfusion on perfusion-weighted imaging. Patients with suspected local treatment failure were generally followed initially with short interval imaging and treated conservatively prior to determination of a treatment failure in order to rule out radiation necrosis. Distant brain failure was defined as a new brain metastasis that was not within the GKRS treatment volume.
Patients who developed further brain metastases after GKRS were generally treated with further GKRS, while WBRT was generally reserved for salvage of 4 or more total brain metastases over time or short-interval distant failures if it was not given previously. The decision to treat with WBRT in the upfront setting was left to the discretion of their primary treating physician, which was based upon a combination of patient performance status, number of metastatic lesions, tumor volume, and if the patient had already previously received WBRT. In general, WBRT was offered as part of upfront therapy if patients had greater than 4 metastases at time of diagnosis. Median time from completion of prior WBRT to GKRS treatment was 7 months (range 1–54 months). Seven patients received GKRS for initial management of brain metastases in conjunction with WBRT (Table 1).
Neurologic death was defined as previously reported by Patchell et al. . Patients were deemed to have died of neurologic causes if they had stable systemic disease, and had evidence of worsening neurologic dysfunction. In the event that a patient experienced progressive neurologic dysfunction in the setting of intercurrent illness or progressive systemic disease, they were also considered to have died of neurologic causes. If cause of death was unknown, the patient was excluded from analysis of neurologic death.
Prior to radiosurgery, each patient underwent a high-resolution contrast-enhanced stereotactic magnetic resonance imaging (MRI) study of the brain. Treatment planning was performed using the GammaPlan Treatment Planning System (Elekta AB, Stockholm, Sweden). GKRS was performed using either the Leksell Model B (years 2000–2004), Model C (years 2004–2009) or Perfexion (years 2009–2010) Gamma Knife units (Elekta AB, Stockholm, Sweden). Median prescribed dose of 20 Gy (range 9–24 Gy) was delivered to tumor margin with doses generally prescribed to the 50 % isodose line. Dose prescription was determined based on size and volume of each metastasis generally following the guidelines published by Shaw et al. for single fraction radiosurgical treatment of brain metastases .
Chi-square or exact tests were used to assess tumor subtype and HER2 positivity differences in categorical variables and analysis of variance on raw or ranked data was used to assess differences in continuous variables. The Kaplan–Meier method was used to estimate the time to event distributions and log-rank tests were used to assess unadjusted differences in these outcomes between tumor subgroups. Cox proportional hazards regression models were used to determine which covariates were associated with the risk of the various events and to assess the effect of HER2 positivity after adjusting for covariates. Covariates included age at diagnosis, stage at initial presentation (I–IV), lymph node status (0, 1–3, 4+), HER2 status (− or +), prior WBRT (no or yes), RPA classification (I–III), Karnovsky Performance Status (KPS) (>70 or <70), number of metastatic lesions (1, 2–4, >4), status of systemic disease at time of GKRS treatment (stable, progressive), and marginal dose (<16 Gy, ≥16 Gy). Primary endpoints included time to local brain failure after GKRS, time to distant brain failure, incidence of neurologic death, and OS. The length of time to recurrence of the original lesion treated was calculated from the date of GKRS to the date of radiographic evidence of recurrence demonstrated by MRI. All analyses were done using SAS version 9.2.
The Kaplan–Meier estimates of OS stratified by primary tumor subtype are depicted in Fig. 1. Overall survival at 1 year for patients with Basal, Luminal A/B, HER2, and Luminal HER2 subtypes was 34, 39, 49, and 67 %, respectively; the median survival time was 7.1, 9.0, 11.3 and 22.0 months for the four subtypes, respectively (p = 0.001). Median survival time was 16.7 months for HER2+ patients versus 8.4 months for HER2 negative patients (p <0.001). Cause of death was available for 113 of 130 patients who had died by time of our analysis; 53 of the 113 patients with known cause of death died of neurologic causes. Time to neurologic death by HER2 positivity is summarized in Table 2. 18 of the 59 HER2–patients with known neurologic status died of neurologic causes versus 35 of 76 HER2+ patients (31 vs 46 %, p = 0.066). Median time to neurologic death in the HER–and HER+ groups was 21.7 and 28.2 months, respectively (p = 0.651).
Freedom from local and distant brain failure after radiosurgery, stratified by tumor subtype, are plotted in Figs. 2 and and3.3. Freedom from local failure at 1 year was 77, 76, 78, and 82 % for Basal, Luminal A/B, HER2, and Luminal HER2 subtypes, respectively (p = 0.554). Freedom from distant failure at 1 year was 24, 53, 53, and 64 % for Basal, Luminal A/B, HER2, and Luminal HER2 subtypes, respectively (p <0.001).
Freedom from local and distant brain failure after radiosurgery, stratified by HER2 positivity, are summarized in Table 2 and plotted in Fig. 4. Freedom from local failure for HER2− and HER2+ tumors at one-year were 77 and 79 %, respectively (p = 0.767). Freedom from distant failure for HER2− and HER2+ tumors at one-year were 40 and 58 %, respectively (p = 0.034).
Ten patients (7 %) ultimately developed leptomeningeal failure with the following subtypes: Luminal A/B (three patients); HER2 (four patients); Luminal HER2 (one patient), and Basal subtypes (two patients). To treat leptomeningeal disease, two of these patients received WBRT as salvage therapy, two patients received intrathecal chemotherapy, and a single patient received systemic chemotherapy with prolonged durable response. The remaining patients opted to terminate care at time of leptomeningeal failure.
A separate analysis was performed for patients who had never received prior WBRT to determine if tumor subtype affected the need for salvage WBRT. A total of 81 patients in our series had never received WBRT prior to GKRS. Of these patients, 23 ultimately went on to receive WBRT in the salvage setting. Freedom from WBRT at 1 year was 70 and 84 % for HER(−) and HER2+ patients, respectively (p = 0.180).
An analysis was performed on HER2+ patients to determine if patients receiving lapatinib experienced differences in times to the various events. Kaplan–Meier estimates of OS, time to distant brain failure, and freedom from neurologic death were calculated in HER2+ patients receiving and not receiving lapatinib. Overall survival at one year in HER2+ patients was 54 and 59 % (p = 0.730), freedom from distant failure at one year was 55 and 58 % (p = 0.715), and freedom from neurologic death at one year was 65 and 75 % (p = 0.788) in patients who did and did not receive lapatinib, respectively.
Multivariate analysis using Cox regression was used to determine factors that were associated with OS, neurologic death, or local or distant failure. HER2(+) receptor status (HR = 0.51, p <0.001) and no prior WBRT (HR = 0.67, p = 0.023) predicted for improved survival, whereas worse KPS (HR = 4.18, p = 0.017) and systemic progression (HR = 1.53, p = 0.033) at the time of GKRS predicted for worse outcomes on univariate analysis (UVA). Both HER2 receptor status (p = 0.007) and no prior WBRT (p = 0.02) remained significant on multivariable analysis. No statistically significant prognostic factors were identified for local failure. Multivariable analysis revealed that prior WBRT (HR = 0.45, p = 0.024) predicted for decreased distant brain failure, while >4 lesions (HR = 3.28, p = 0.006) had inferior control. Previous WBRT (HR = 2.31, p = 0.017) was the only predictor that was both univariably and multivariably associated with increased risk of neurologic death.
As breast cancer patients have been found to be a heterogeneous population, there has been recent interest in how breast cancer subtype may affect outcomes of patients with brain metastases. In the current analysis, we were able to demonstrate significant survival differences with greatest median OS in patients with Luminal HER2 subtype (22.0 months) and the poorest survival seen in those with Basal subtype (7.1 months). Multivariate analysis confirmed that HER2 status was predictive of improved OS (HR = 0.51, p = 0.007). Recently, Sperduto et al. published a Graded Prognostic Assessment of 400 breast cancer patients based on prognostic factors including Age, KPS, ER/PR status, and HER2 status . Similar to the findings in the current analysis, survival was dependent upon breast cancer subtype, lending credence to the argument that tumor subtype belongs in a breast cancer-specific prognostic index.
In our analysis, it appeared that both the breast cancer subtype (p <0.001) and the HER2 status (p = 0.034) affect the distant brain failure rate. Distant brain failure is presumed to be from the presence of microscopic disease at the time of radiosurgery, or secondary to tumor re-seeding from uncontrolled systemic disease . Basal tumors had the highest rate of distant brain failure (76 % at 1 year) of all the subtypes. Previous analyses of brain metastases have suggested that predictors for a lower rate of distant brain failure after radiosurgery include <4 brain metastases, absence of extracranial disease and non-melanoma histology . Dyer and colleagues recently identified progressive extracranial disease and basal subtype as key predictors for OS . Although progressive extracranial disease predicted for worse OS only by UVA in this current dataset, which included patients that received prior WBRT, it is likely that the inherent aggressiveness of basal subtype breast cancers and the general inability to control systemic disease in these patients has led to the higher rate of distant brain failure.
The benefit of WBRT at the time of initial presentation is unclear, particularly because it cannot prevent subsequent re-seeding of the brain from extracranial disease. Three published randomized trials have compared SRS with adjuvant WBRT against SRS alone in the primary management of brain metastases showing no worse survival in patients in which WBRT was withheld in the upfront setting [20–22]. In all of these trials, WBRT decreased the likelihood of distant brain failure and in two of the studies, the likelihood of neurologic death was decreased with WBRT. This benefit must be weighed against the potential toxicity of WBRT which can be detected within 6–8 weeks of treatment , and does not have a plateau in terms of the likelihood or severity of post-treatment cognitive decline [24, 25]. As treatment failure has been associated with worsening decline, the proper management may be to use upfront WBRT in patients who experience rapid distant brain failure . The basal subtype may represent such a population given the shortened life expectancy and the shortened time to distant brain failure. In our analysis of patients who received GKRS without any prior WBRT, there was a trend towards earlier need for WBRT in patients with HER2(−) status (log rank p = 0.180).
Patients with brain metastases generally experience a poorer prognosis related to progressive neurological progression and death . In the current series, time to neurologic death did not differ by breast cancer subtype; however, a greater proportion of HER2+ patients (46 vs 31 %, p = 0.066) experienced neurologic death. This finding may be explained by more active systemic treatment options for HER2 positive patients and the fact that the blood brain barrier prevents CNS penetration of many such agents. As a result, HER2(−) patients have a greater competing risk of death from systemic metastases. Given the high rate of neurologic death in HER2+ patients, improved brain-directed therapies are necessary. Such improvements may come from either targeted systemic agents that penetrate the CNS, improved detection of occult metastases, and use of surgical or radiosurgical techniques. In our current analysis, there was no improvement in survival, local control or distant brain control in HER2+ patients who were treated with lapatinib versus those who never received lapatinib. However, in a recent Phase II multicenter prospective trial of 240 women with progressive CNS disease, lapatinib demonstrated a ≥ 20 % volumetric reduction in brain metastases in approximately 21 % on exploratory analysis .
There are several limitations to our current study. The retrospective nature of the review does not allow for control of variables, resulting in heterogeneity in surgical, radiotherapeutic, and chemotherapeutic interventions, introducing potential selection bias. In addition, patients were treated at a single academic institution, raising concern for referral bias. The current series found on multivariate analysis that prior WBRT had a greater likelihood of neurologic death as well as worsened overall survival. This may be due to confounding factors in our patient population. Approximately half of our cohort received GKRS, some of whom having failed previous WBRT and may have exhausted treatment options or have more aggressive disease. Finally, recent reports suggest that it is possible that tumors may occasionally change subtype from time of diagnosis to time of development of brain metastases . Few patients underwent craniotomy, so pathologic confirmation that the subtype remained the same was impossible.
The present study was able to demonstrate differences in survival and distant brain failure after GKRS based on breast cancer subtypes. HER2+ patients have a lower risk of death and distant failure. HER2(−) patients tend to fail distantly in the brain and require WBRT more quickly, suggesting that an inability to control systemic disease is a major factor in predicting brain failure in these patients.
Presented in part at the 53th Annual ASTRO Meeting, September 2011.
Presented in part at the 16th Annual SNO Meeting, November 2011.
Conflict of interest None of the authors of this manuscript have an actual or potential conflict of interest in the preparation or publication of this manuscript.
Tamara Z. Vern-Gross, Department of Radiation Oncology, Comprehensive Cancer Center, Baptist Medical Center, Wake Forest University, Medical Center Boulevard, Winston-Salem, NC 27157, USA.
Julia A. Lawrence, Department of Medicine (Hematology & Oncology), Wake Forest University, Winston-Salem, NC, USA.
L. Douglas Case, Division of Public Health Sciences, Wake Forest University, Winston-Salem, NC, USA.
Kevin P. McMullen, Department of Radiation Oncology, Comprehensive Cancer Center, Baptist Medical Center, Wake Forest University, Medical Center Boulevard, Winston-Salem, NC 27157, USA.
J. Daniel Bourland, Department of Radiation Oncology, Comprehensive Cancer Center, Baptist Medical Center, Wake Forest University, Medical Center Boulevard, Winston-Salem, NC 27157, USA.
Linda J. Metheny-Barlow, Department of Radiation Oncology, Comprehensive Cancer Center, Baptist Medical Center, Wake Forest University, Medical Center Boulevard, Winston-Salem, NC 27157, USA.
Thomas L. Ellis, Department of Neurosurgery, Wake Forest University, Winston-Salem, NC, USA.
Stephen B. Tatter, Department of Neurosurgery, Wake Forest University, Winston-Salem, NC, USA.
Edward G. Shaw, Department of Radiation Oncology, Comprehensive Cancer Center, Baptist Medical Center, Wake Forest University, Medical Center Boulevard, Winston-Salem, NC 27157, USA.
James J. Urbanic, Department of Radiation Oncology, Comprehensive Cancer Center, Baptist Medical Center, Wake Forest University, Medical Center Boulevard, Winston-Salem, NC 27157, USA.
Michael D. Chan, Department of Radiation Oncology, Comprehensive Cancer Center, Baptist Medical Center, Wake Forest University, Medical Center Boulevard, Winston-Salem, NC 27157, USA.