Central Nervous System Metastases in Patients with HER2-Positive Metastatic Breast Cancer: Incidence, Treatment, and Survival in Patients from registHER

For women with advanced breast cancer, 10% to 16% develop central nervous system (CNS) metastases (1, 2). However, because most CNS metastases are diagnosed in response to clinical symptoms rather than by routine staging, the total incidence of CNS metastases is likely higher. In a study that required prescreening for CNS lesions, 14.8% of patients had occult lesions (3) and autopsy data have revealed that up to 30% of patients with advanced metastatic breast cancer (MBC) have CNS metastases (4). Several risk factors have been associated with the development of CNS lesions in patients with MBC, including younger age (<50 year), having more than 2 metastatic sites at MBC diagnosis, and HER2-positive disease (5). Patients with HER2-positive MBC tumors are 2 to 4 times more likely to develop CNS tumors than patients with HER2-negative disease (2, 6, 7). The recent apparent increase in the incidence of CNS metastases likely results from advances in detection of CNS lesions and from improvements in systemic treatment that allow more patients to live long enough to develop CNS metastases (2); this has been observed in HER2-positive MBC patients treated with trastuzumab (1, 8).

In general, survival for breast cancer patients with CNS metastases is poor, with 1-year survival of approximately 20% (1). In one study of patients with CNS metastases from breast cancer (n = 121), survival rate was 8.0% at 2 years, 6.9% at 3 years, and 1.3% at 5 years (9). Although surgery can be an effective treatment option for patients with limited CNS disease burden (10), steroids and radiotherapy have been the mainstays of treatment of CNS metastases (11–14). For patients with fewer lesions, stereotactic radiosurgery is associated with longer survival (14, 15). Because an intact blood–brain barrier (BBB) prevents most chemotherapeutic agents from entering the CNS, chemotherapy is not used routinely to treat CNS metastases. For patients whose tumors overexpress HER2, there is an increased risk of developing CNS disease and a decrease in overall survival (OS; refs. 1, 8, 16). Treatment of HER2-positive patients who have CNS metastases with trastuzumab (Herceptin; Genentech, Inc.), a monoclonal antibody directed against HER2, seems to prolong survival, even compared with patients with HER2-negative disease (1, 8, 17).

As a multicenter, prospective, observational study of 1,023 patients, registHER is the largest study population of patients with recently diagnosed HER2-positive MBC. This study offers a unique opportunity to assess the incidence, potential risk factors, and outcomes for patients with HER2-positive MBC who developed CNS metastases.

registHER is a multicenter, prospective observational U.S.-based cohort study of patients from community and academic settings who have newly diagnosed HER2-positive MBC. The objectives of registHER were to describe the natural history of disease and treatment patterns for patients with HER2-positive MBC and to explore associations between specific therapies and patient outcomes. The study was approved by all local institutional review boards, and all enrolled patients provided informed consent.

Patient recruitment was conducted from December 2003 to February 2006. Eligible patients included women (n = 1,013) and men (n = 10) with breast cancer. HER2 status was determined according to institutional guidelines. Patients were eligible regardless of treatment before enrollment or during study follow-up. To minimize patient selection bias, investigators were encouraged to recruit all eligible patients at their practice.

Patients received care according to their physicians' standard practice without study-specified evaluations. Patient information [including demographics, tumor characteristics, initial metastasis sites and progressions, systemic treatment received (with start/stop dates), and response to treatment] were recorded at enrollment and updated every 3 months thereafter. Participating physicians or designated staff entered patient data into electronic case report forms (eCRF), using a Web-based electronic data capture system. Formal, prespecified, and scheduled assessments for tumor response were not required, and tumor progression was reported by physicians according to their standard practice.

Starting in January 2008, investigators following patients identified as having CNS metastases were asked to complete an additional retrospective questionnaire to gather data on diagnosis, leptomeningeal involvement, treatment, and response of CNS metastases to treatment. At that time, the eCRF was modified to prospectively capture these CNS metastases–specific data.

Prior to the approval of lapatinib in March 2007, data on use of lapatinib and other investigational drugs were incompletely captured. The study's eCRF was amended at the time of lapatinib approval to allow lapatinib use to be reported.

Analyses incorporate follow-up data on registHER patients as of June 15, 2009 (database lock). Enrollment of 83 patients whose MBC diagnosis was more than 6 months (up to 9 months) prior to enrollment was permitted and these patients are included in all analyses.

Patients with CNS metastases were defined as those with either leptomeningeal metastases (LM) and/or metastases in the brain parenchyma. Time to CNS metastases was defined as the time from the date of MBC diagnosis to date of diagnosis of CNS metastases. Survival after CNS metastases was defined as the time from diagnosis of first CNS metastases to date of death from any cause or last follow-up. Overall survival was defined as the time between a patient's initial metastatic diagnosis and death from any cause or last follow-up.

Patient demographics and tumor characteristics were tabulated by CNS metastases status and comparisons between groups used the χ² test, as appropriate. Time to CNS metastases and survival after CNS metastases were estimated using the Kaplan–Meier product-limit method. Data from patients with no event recorded were censored at the date of last follow-up. Kaplan–Meier curves and the log-rank test were used to compare unadjusted survival after CNS metastases in different treatment groups.

Patients were classified as receiving trastuzumab if they had received 21 total days or more (i.e., equivalent to ≥2 doses) of drug following their first CNS event. Patients were classified as receiving chemotherapy for treatment of CNS metastases if they had chemotherapy started after a CNS event or if chemotherapy started before a CNS event continued for 21 days or more after a CNS event. If radiation therapy or surgery was recorded either on the eCRF or on the CNS questionnaire, that patient was counted as having received radiation therapy or surgery, respectively. Radiation therapy or surgery specifically directed to sites other than the brain was not included.

Cox proportional hazards models were used to evaluate the association between survival and various treatments, adjusting for relevant baseline covariates that were considered a priori to be important factors for the physician's treatment decision or to be prognostic for survival. Several factors were evaluated, but the final multivariate model presented here includes as covariates age, Eastern Cooperative Oncology Group performance status (ECOG PS) at MBC diagnosis, hormone receptor status, cancer stage at diagnosis, CNS involvement at MBC diagnosis, and the main treatment options after CNS metastases (trastuzumab, chemotherapy, radiotherapy, and surgery). Although we describe the use of lapatinib as it was observed in this study, given the timing of the approval of lapatinib after completion of patient recruitment in registHER, no adjustment for the use of lapatinib was made.

Genentech, Inc., participated in design of the study; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the manuscript for publication.

Of the 1,023 patients enrolled at 240 sites in the United States, 1,012 (99%) had a diagnosis of confirmed, HER2-positive tumors and enrolled within 12 months of their MBC diagnosis, permitting them to be included in this analysis. The median follow-up time from metastatic diagnosis was 29 months. Thirty-seven percent (377 of 1,012) of patients with MBC had a diagnosis of CNS metastases, 7% (75 of 1,012) at the time of their initial MBC diagnosis and 30% (302 of 1,012) as a subsequent site of disease progression (Fig. 1A).

For 106 (28.1%) patients, CNS metastases were their only site of first progression after metastatic diagnosis (Fig. 1A). For 15 (14.2% of this group, or 4.0% of all patients with CNS metastases), the CNS was the only site of metastatic disease. Patients who developed CNS metastases were younger (<50 years old, 43% vs. 37%), more likely to have hormone receptor–negative disease (51% vs. 39%), and had a higher disease burden (62% vs. 49% had ≥2 metastatic sites) than patients without CNS metastases (Table 1).

CNS-specific data were recorded for 83% (313 of 377) of patients with CNS metastases. For the 308 patients for whom neurologic symptom data were available, 81% had neurologic symptoms at the time CNS metastases were diagnosed (Table 2). Of the patients with CNS metastases and who had LM data available, 9.6% (30 of 311) were described as having LM, with or without parenchymal disease (Table 2).

For patients with any CNS events (n = 377), the median OS from time of MBC diagnosis was 26.3 months (73.7% of patients died) compared with 44.6 months for patients with no CNS events (n = 635; 42.4% of patients died; Fig. 1B). Patients with CNS metastases at the time of MBC diagnosis had a median OS of 20.3 months; for the 15 patients for whom CNS was the only site of metastasis, median OS was 26.3 months. Patients who developed CNS metastases after their MBC diagnosis (n = 302) had a median OS of 27.0 months. Patients with LM disease had a median OS of 22.9 months.

For all patients with CNS metastases (n = 377), the median time to first CNS event was 10.8 months (Fig. 2A). For the subset of 302 (80%) patients who developed CNS metastases during treatment of MBC, the median time to first CNS event was 13.3 months (range, 0.6–50.1 months). Of these patients, 8.3% (25 of 302) were receiving first-line therapy for MBC at the time; 42.1% (127 of 302) were receiving second-line therapy, and 47.7% (144 of 302) were receiving subsequent lines of therapy (Fig. 2B). In the subset of patients (n = 106) who had CNS metastases as their only site of first progression after metastatic diagnosis, the median time to first CNS event was 10.5 months (range, 0.6–43.5 months).

Prior to development of CNS metastases, trastuzumab was the most commonly received systemic therapy; 93% of patients had received trastuzumab before their first CNS event. For patients who received trastuzumab before their first CNS event (n = 280), the median time to diagnosis of CNS metastases was 13.6 months (range, 1.7–50.1 months). For the cohort (n = 22) who did not receive trastuzumab before CNS metastases, median time to CNS metastases was 7.1 months (range, 0.6–25.8 months).

Of the 377 patients with CNS disease, 71.4% received radiation therapy, 69.5% received chemotherapy, 68.4% received trastuzumab, 24.1% received lapatinib (78% of whom also received trastuzumab), and 7.7% had surgery as part of treatment of CNS disease (Table 2). Of the 269 patients who received radiotherapy, 216 (80.3%) received whole-brain radiotherapy (WBRT), 27 (10.0%) received stereotactic radiotherapy, and 13 (4.8%) received both types (Table 3). Twenty-four (6.4%) patients did not receive any of these treatment options. A sequence or combination of radiation therapy, chemotherapy, and trastuzumab was the most common treatment after CNS metastases diagnosis, which was received by 26.5% of patients (Table 2).

For the subset of patients (n = 302) who developed CNS metastases after their MBC diagnosis, 71.5% received radiotherapy, 65.9% received chemotherapy, 63.3% received trastuzumab, 25.2% received lapatinib, and 5.6% had surgery. Of the 75 patients who had CNS metastases at the time of initial MBC diagnosis, 70.7% received radiotherapy, 84.0% received chemotherapy, 89.3% received trastuzumab, 20% received lapatinib, and 16% underwent surgery.

The median survival after first diagnosis of CNS metastases for all patients was 13.0 months (range, 0.1–55.5 months; Fig. 3A). For the patients who presented de novo with CNS metastases (n = 75), the median survival was 20.3 months (range, 1.0–55.5 months). For those diagnosed with CNS metastases subsequent to their MBC diagnosis (n = 302), median survival following diagnosis of CNS metastases was 9.6 months (range, 0.1–54.5 months; Fig. 3A). In all patients with CNS metastases (n = 377), chemotherapy and trastuzumab treatment after CNS diagnosis were each strongly associated with improved survival (log-rank test P < 0.001 for both comparisons; unadjusted; Fig. 3B and C). For patients who received trastuzumab following diagnosis of CNS disease (n = 258, 68.4%), the median survival was 17.5 months, compared with 3.7 months for patients who did not (n = 119; Fig. 3B). For patients who received chemotherapy following diagnosis of CNS disease (n = 262, 69.5%), the median survival was 16.4 months, compared with 3.7 months for patients who did not receive chemotherapy (Fig. 3C). For patients who received surgery, the median survival was also longer (20.3 vs. 11.3 month, P = 0.013). Radiation therapy seemed to increase survival (13.9 vs. 8.4 month), but the difference was not significant (P = 0.134; Fig. 3D). For patients with reported LM disease (n = 30), the median survival from the time of CNS diagnosis was 5.8 months (range, 0.6–55.5 months).

After adjusting for the clinical characteristics of age, ECOG PS at time of MBC diagnosis, hormone receptor status, cancer stage at initial diagnosis, and CNS involvement at metastatic diagnosis in a multivariable proportional hazards model, we found that treatment with trastuzumab (HR = 0.33; 95% CI: 0.25–0.46; P < 0.001) or chemotherapy (HR = 0.64; 95% CI: 0.48–0.85; P = 0.002) were independently associated with decreased hazard of death after CNS metastases, whereas CNS-directed radiotherapy was not (P = 0.898; Table 3). Although not statistically significant, there was a trend to suggest that surgery is associated with decreased hazard of death after CNS metastases (HR = 0.63, 95% CI: 0.39–1.02; P = 0.062). The interactions between ECOG PS and chemotherapy or trastuzumab treatment were investigated but found to be not significant. In additional sensitivity analyses, treatment with trastuzumab continued to be associated with decreased hazard of death after CNS metastases in the multivariable proportional hazards model (HR = 0.34; 95% CI: 0.24–0.48), even after excluding potential palliative care patients [i.e., patients who survived <21 days (n = 8)] or who did not receive any treatment after their CNS diagnosis (n = 47) from the nontrastuzumab comparison group (see Supplementary Table).

As the largest prospectively followed cohort of patients with HER2-positive MBC, registHER affords a unique opportunity to study the natural history of CNS metastases in this patient population in the context of contemporary therapies. Consistent with the incidence of CNS metastases found in retrospective analyses, approximately one third of the 1,012 patients analyzed in registHER had CNS metastases (2, 18). Potential predictive factors for the development of CNS metastases were similar to those previously noted for breast cancer patients, including younger age (<65 years), hormone receptor–negative disease, and greater disease burden. Seven percent of registHER patients had CNS metastases at the time of their initial MBC diagnosis. However, because not all patients underwent CNS screening at the time of MBC diagnosis, and not all patients with CNS metastases had neurologic symptoms, the actual incidence of CNS disease at MBC diagnosis is probably higher.

Although HER2-positive breast cancer is a relatively aggressive subtype, and it was considered that patients in registHER might have an increased incidence of LM (a particularly aggressive form of CNS disease associated with poor prognosis), we found that the incidence of LM among the HER2-positive patients in registHER (9.6% in patients with CNS disease and 2.9% overall) was comparable with previous reports (4, 13, 18, 19).

In registHER, patients who had CNS metastases at the time of MBC diagnosis survived twice as long from the time CNS disease was diagnosed (20.3 vs. 9.6 months) as did patients whose CNS disease was diagnosed after their MBC diagnosis. This raises the question of whether earlier detection of CNS lesions leads to prolonged survival post–CNS metastasis because CNS-specific therapies were administered earlier in the course of treatment (e.g., radiation therapy, or surgical resection for smaller, solitary lesions). However, the difference in survival postdetection of CNS observed in this study is likely a reflection of the difference in time to diagnosis of CNS metastases (i.e., median time to first CNS metastases was 10.8 months) rather than due to a true improvement in survival. Of note, the improved survival postdetection of CNS metastases did not correspond to improvement in OS from time of MBC diagnosis—the median OS from time of initial metastatic diagnosis was shorter in patients whose CNS metastases were noted at initial diagnosis than those whose CNS metastases were diagnosed subsequently (20.3 vs. 27.0 months). This is consistent with the possibility that patients with MBC who present with CNS metastases have a more aggressive disease.

An inherent limitation of a nonrandomized observational study results from “confounding by indication,” whereby certain treatments are preferentially given to sicker patients and other treatments to healthier patients (e.g., patients who are candidates for surgical resection of their CNS tumors generally have a lower disease burden, a more favorable location of metastasis, overall better control of systemic extracranial disease, and may represent a patient population with a relatively better prognosis). Likewise, patients who received no therapy may represent a more debilitated group of patients. These results are consistent with several retrospective analyses (1, 2) and were corroborated by the multivariable analysis we conducted to adjust for clinically recognized prognostic factors and potential differences between treatment groups. In registHER, the reduction in death observed after treatment with trastuzumab, chemotherapy, and possibly surgery persisted after controlling for the key prognostic factors of age, ECOG PS, cancer stage at initial diagnosis, hormone receptor status, and CNS disease at metastatic diagnosis.

There also remains the potential for residual confounding by clinical factors that were not collected or sufficiently captured. For example, ECOG PS was missing for 52% of patients and represented ECOG PS at the time of the patient's MBC diagnosis, and not necessarily ECOG PS at the time CNS disease developed, which may have occurred months later. Therefore, the model did not fully account for this confounder. In addition, as data regarding post–CNS treatment were collected via questionnaire, those data may be incomplete. This could in part explain why patients treated with radiotherapy trended toward longer survival, yet the difference was not statistically significant. Alternatively, it may be that at the time registHER was conducted, it was not yet common practice to treat isolated CNS metastases with both WBRT and stereotactic radiotherapy, the practice now known to lead to longest survival (20). Finally, because patients may have a diagnosis of MBC up to 12 months prior to enrollment and patients with longer survival may be more likely to enroll, the OS estimates from time of MBC diagnosis in registHER may be slightly higher than expected in a general MBC patient population.

Results from registHER are consistent with recent data; for patients with MBC and CNS disease, median survival after CNS metastases was 6.1 months for patients with HER2-positive disease who received no trastuzumab (n = 32) and 11.6 months for those with HER2-positive disease who did receive trastuzumab (8). Because most (78%) of the 91 patients who received lapatinib after CNS metastases also received trastuzumab, we were unable to accurately assess the independent effect of lapatinib as anti-HER2 therapy on survival after CNS metastases.

The finding that chemotherapy and trastuzumab treatment are associated with improved survival even after adjusting for relevant variables in patients after the development of CNS metastases is interesting. Because most chemotherapy agents used in the treatment of breast cancer either do not cross an intact BBB, or are pumped out of the CNS by phosphoglycoprotein in the endothelial cells comprising the BBB, they may not reach sufficient therapeutic levels to treat CNS metastases (21). Trastuzumab does not readily cross an intact BBB; in patients without CNS metastases (i.e., with an intact BBB), the ratio of trastuzumab in blood plasma to trastuzumab in the cerebral spinal fluid is greater than 300:1 (22, 23). However, in the setting of CNS lesions, the permeability of the BBB is likely increased via physical perturbation by tumors (24) and by structural changes in tumor microvessels that increase vascular permeability (25–27), including changes mediated by tumor-secreted factors such as VEGF (28). Recently, it has been shown in preclinical models that although the blood–tumor barrier (BTB) may show observable permeability changes in brain metastases, the BTB function in these models was only partly compromised and retained significant ability to impede chemotherapeutic uptake (29). Radiation therapy, which was received by 71% of the patients in registHER with CNS disease, has also been shown to increase BBB permeability (23). In addition, the finding of improved survival in patients receiving chemotherapy and trastuzumab treatment after the development of CNS metastases is a reflection of the importance of extracranial disease control with anti-HER2 therapy and chemotherapy, as a substantial proportion of patients with CNS metastases would likely die from their extracranial disease without adequate therapy (8).

A prospective study to determine which factors most affect treatment efficacy for CNS metastases of HER2-positive MBC would be of great interest.

P.A. Kaufman has received research grant support from Genentech, Inc., and has performed consulting services for Genentech, Inc. H.S. Rugo received other commercial research support from Genentech to UCSF. D. Tripathy received honoraria from speakers bureau from Genentech. I.C. Tudor, D.A. Yardley, and D. Tripathy are consultant/advisory board members of Genentech.

Stephanie Kareht, PhD, of Genentech, Inc., was the medical writer and assisted in writing the manuscript.

Genentech, Inc., provided funding for this study and support for the writing of the manuscript.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.