Introduction

Breast cancer is the most common cancer among women worldwide and the second leading cause of brain metastases (BrM) [1-3]. Despite recent treatment advances, the prognosis of patients with breast cancer BrM remains poor [4-12], particularly among those with HER2− disease. Although the cause of death among patients with breast cancer BrM is challenging to ascertain, approximately 50% of patients with HER2+ BrM are thought to die of central nervous system (CNS) disease involvement [5]. The proportion of patients with triple-negative breast cancer (TNBC) or hormone receptor (HR)+/HER2− metastatic breast cancer (MBC) who succumb to CNS disease is not well understood, but more data will emerge as results of prospective BrM screening trials become available.

Patients with breast cancer BrM typically receive a combination of local treatments consisting of surgery alone, surgery followed by postoperative radiation, or radiation alone in addition to systemic therapy. The decision regarding which treatments to offer can be challenging and often depends on clinical factors such as the number and size of BrM, presence of neurologic symptoms, patient performance status, and degree of extracranial disease control [13-21]. Unfortunately, patients with BrM perform poorly on quality of life indicators owing to neurologic symptoms secondary to their intracranial metastatic disease and treatment-related toxicity [22-32]. As the incidence of BrM among women living with MBC continues to increase over time, likely due to improvements in early detection and systemic disease control, there is an increasing need for the implementation of evidence-based interventions that prolong survival with minimal side effects and, in particular, preservation of neurocognitive function.

Recently, stereotactic radiosurgery (SRS) has emerged as an excellent nonsurgical option with high precision and limited toxicity [19, 20]. Owing to its superior local control rates and sparing of neurocognitive function, SRS has displaced whole brain radiotherapy as the preferred radiotherapeutic option for patients with limited-volume BrM [19, 20]. In this study, we assessed the treatment patterns and outcomes of women with breast cancer BrM at an academic cancer center in the modern era of SRS. SRS became standard of care for the treatment of patients with MBC in 2009 with the publication by Chang et al. that whole brain radiotherapy (WBRT) is associated with neurocognitive decline [33]. Uptake of SRS gradually increased, initially for selected patients with excellent performance status, controlled extracranial disease, and 1–4 BrM.

Materials and Methods Study Design and Population

This is a single-institution retrospective study of patients (≥18 years of age) with a diagnosis of MBC treated for BrM with initial surgery, WBRT, or SRS at Sunnybrook Odette Cancer Centre, Toronto, Canada, between 2008 and 2018. This study period was selected to allow for 2-year follow-up among all study participants. This study was approved by our institutional ethics review board. Males, patients with a history of other primary malignancies, and those with an uncertain date of BrM diagnosis were excluded from the analysis. Data on patient demographic, pathological, and clinical outcomes were collected from the electronic patient record.

Study Outcomes

Overall survival (OS) was defined from the date of BrM diagnosis to the date of death due to any cause. Brain-specific progression-free survival (bsPFS) was defined from the date of BrM diagnosis until the date of radiographic disease progression in the brain specifically (patients with systemic disease progression were not censored by definition). The dates of second-line local treatment to the brain were used as surrogates for progression in cases in which the date of progression was not documented.

Statistical Analysis

Descriptive statistics were used to summarize patient and treatment characteristics as continuous (median, range) or categorical variables (frequency). Chi-square tests were used to compare categorical variables (i.e., treatment patterns), whereas the Kaplan-Meier method was used to estimate OS. Univariable and multivariable analyses (UVAs and MVAs, respectively) of OS were performed with Cox proportional hazards models. From the UVA, only clinically relevant variables with a p value <.05 were included in the final MVA. All statistical analysis were conducted with R Version 3.6.1 on RStudio. A p value of <.05 was considered to be statistically significant.

Results Patient Characteristics

A total of 683 patients with breast cancer BrM were included in the study (Fig.1). The median age at BrM diagnosis was 56 years (range 24–93), and the median time between the diagnosis of MBC and development of BrM was 8.1 months (interquartile range [IQR] 0–24.4). The distribution of breast cancer subtypes included the following frequencies: 153 (22.4%) TNBC, 188 (27.5%) HER2+, 246 (36.0%) HR+/HER2−, and 61 (13.3%) of patients had an unknown subtype. Of 646 patients for whom data regarding symptoms were available, the majority (77.5%) had neurological symptoms at their initial presentation with BrM. Asymptomatic BrM were detected as part of an evaluation for clinical trial participation or via practice patterns of select physicians, who referred patients with screen-detected BrM to our cancer center for radiation therapy. In addition, the cumulative incidence of leptomeningeal disease (LMD) as diagnosed either clinically or via imaging was 27.3% (n = 174), including 11.1% (n = 76) who presented with LMD at the time of initial BrM diagnosis. In patients with subsequent LMD progression, the median time from BrM to LMD was 5.3 months (IQR 2.3–13.4). The most common sites of extracranial metastatic disease at the time of first BrM diagnosis were bone (67.6%), followed by lymph nodes (61.3%), lung (55.9%), and liver (53.0%). Details regarding systemic therapy received prior to first treatment with radiotherapy for BrM is included in supplemental online Table 1. A complete description of patients' clinical characteristics is presented in Table 1.

Constitution of the study population.

Abbreviations: BrM, brain metastases; CNS, central nervous system; HR, hormone receptor; MBC, metastatic breast cancer; SRS, stereotactic radiosurgery; WBRT, whole brain radiation therapy.

Table 1. Patient characteristics Characteristics Total (n = 683) HR+/HER2−(n = 246) HER2+ (n = 188) TNBC(n = 153) Unknown (n = 96) Median age at BCBM diagnosis, yr 56 57 55 55 62 Median time between MBC diagnosis and BCBM diagnosis, mo 8.1 12.2 8.1 2.6 7.1 Neurological symptoms at BCBM diagnosis, n (%) Yes 529 (77.5) 192 (78.0) 142 (75.5) 125 (81.7) 67 (73.6) No 117 (17.1) 48 (19.5) 33 (17.6) 22 (14.4) 12 (13.2) Unknown 37 (5.4) 6 (2.4) 13 (6.9) 6 (3.9) 12 (13.2) Leptomeningeal disease, n (%) Yes 174 (27.3) 72 (29.3) 45 (23.9) 37 (24.2) 20 (22.0) No 485 (76.0) 169 (68.7) 139 (73.9) 112 (73.2) 60 (65.9) Unknown 24 (3.8) 5 (2.0) 4 (2.1) 4 (2.6) 11 (12.1) Lymph node metastasis, n (%) Yes 419 (61.3) 155 (63.0) 117 (62.2) 106 (69.3) 38 (41.8) No 234 (34.3) 88 (35.8) 65 (34.6) 45 (29.4) 35 (38.5) Unknown 30 (4.4) 3 (1.2) 6 (3.2) 2 (1.3) 18 (19.8) Lung metastasis, n (%) Yes 382 (55.9) 141 (57.3) 107 (56.9) 98 (64.0) 33 (36.3) No 271 (39.7) 96 (39.0) 76 (40.4) 52 (34.0) 45 (49.5) Unknown 30 (4.4) 9 (3.7) 5 (2.7) 3 (2.0) 13 (14.3) Liver metastasis, n (%) Yes 362 (53.0) 152 (61.8) 99 (52.7) 62 (40.5) 46 (50.5) No 287 (42.0) 90 (36.6) 81 (43.1) 83 (54.2) 31 (34.1) Unknown 34 (5.0) 4 (1.6) 8 (4.3) 8 (5.2) 14 (15.4) Bone metastasis, n (%) Yes 462 (67.6) 198 (80.5) 113 (60.1) 91 (59.5) 57 (62.6) No 197 (28.8) 44 (17.9) 71 (37.8) 57 (37.3) 23 (25.3) Unknown 24 (3.5) 4 (1.6) 4 (2.1) 5 (3.3) 11 (12.1) Abbreviations: BCBM, breast cancer brain metastasis; HR, hormone receptor; MBC, metastatic breast cancer; TNBC, triple-negative breast cancer. BrM Treatment Patterns and Tumor Burden

The most common first-line treatment modality for BrM was WBRT (n = 459, 67.2%), followed by SRS (n = 126, 18.4%) and surgical resection (n = 69, 10.1%; Table 2). The proportion of patients requiring radiotherapy who were treated with SRS increased over time during the course of our study period from 0% in 2008, to 4% in 2009, to 66% in 2018 (p < .01; supplemental online Fig. 1). Patients who presented with neurological symptoms at the time of BrM diagnosis (n = 529, 77.5%) were more likely to require surgical resection (Risk Ratio (RR) = 2.74) and were less likely to be treated with SRS alone compared with those with asymptomatic disease (19.1% vs. 32.7%, RR = 0.53). Among the 117 asymptomatic patients, 107 (91.4%) received radiotherapy, whereas 5 (4.3%) received surgery as a first-line local therapy (Table 3). In patients with LMD, 12 women received SRS treatment for BrM before the diagnosis of LMD with a median time from first SRS treatment to diagnosis of LMD of 9.0 months (IQR 4.7–13.6). Patients with symptomatic versus asymptomatic BrM had similar likelihood of presenting with a solitary lesion as opposed to multiple lesions (17.3% vs. 19.2%, p = .81).

Table 2. Local treatments for breast cancer brain metastases by subtype Treatment modality Total (n = 683) HR+/HER2− (n = 246) HER2+ (n = 188) TNBC (n = 153) First-line local therapy, n (%) Radiotherapy-based 587 (85.9) 218 (88.6) 153 (81.4) 128 (83.7) SRS only 126 (18.4) 43 (17.5) 45 (23.9) 21 (13.7) WBRT only 459 (67.2) 175 (71.1) 106 (56.4) 107 (69.9) SRS + WBRT 2 (0.2) 0 (0) 2 (1.1) 0 (0) Surgery-based 69 (10.1) 17 (6.9) 28 (14.8) 21 (13.7) Surgery alone 33 (4.8) 7 (2.8) 14 (7.4) 10 (6.5) Surgery + radiotherapy 36 (5.3) 10 (4.1) 14 (7.4) 11 (7.2) Treatment 656 (96.0) 235 (95.5) 181 (96.3) 149 (97.4) No treatment 27 (4) 11 (4.5) 7 (3.7) 4 (2.6) Retreatment for BrM progression Retreatment, n (%) 182 (27.7)a 51 (21.7)a 83 (45.9)a 33 (22.1)a Median time from treatment for first BrM to retreatment, mo 9 7.4 10 6.4 a Percentages of patients requiring retreatment out of all patients who received first-line local therapy. Abbreviations: BrM, brain metastasis; HR, hormone receptor; MBC, metastatic breast cancer; SRS, stereotactic radiosurgery; TNBC, triple-negative breast cancer; WBRT, whole brain radiation therapy. Table 3. Local treatments by presence versus absence of neurological symptoms at time of brain metastases diagnosis Neurological symptom at BrM diagnosis Absent (n = 117) Present (n = 529) Radiotherapy-based 107 (91.4) 446 (84.3) SRS only 35 (29.9) 85 (16.1) WBRT only 72 (61.5) 359 (67.9) SRS + WBRT 0 (0) 2 (0.4) Surgery-based 5 (4.3) 62 (11.7) Surgery 3 (2.6) 29 (5.5) Surgery + radiotherapy 2 (1.7) 33 (6.2) Treatment 112 (95.7) 508 (96) No treatment 5 (4.3) 21 (4) Data are presented as n (%). Abbreviations: BrM, brain metastasis; SRS, stereotactic radiosurgery; WBRT, whole brain radiation therapy.

Overall, 182 patients (26.6%) received more than one line of local treatment for BrM. Among re-treated patients, the median time to retreatment was 7.9 months (IQR 4.2–13.7). However, patients initially treated with WBRT had lower rates of retreatment (15% vs. 45%) as well as longer median times to retreatment (9 months vs. 6.8 months) compared with patients who were initially treated with SRS. Interestingly, although the first-line local treatment modality did not differ by breast cancer subtype (chi-square test, p = .068), patients with HER2+ BrM were more likely to have received a second-line local therapy (n = 83, 44.1%) compared with patients with HR+/HER2− (n = 51, 21.7%) disease and TNBC (n = 33, 22.1%).

bsPFS and Overall Survival Outcomes

The median bsPFS in our patient population was 4.1 months (IQR 1.0–9.6). On univariate analysis, only age >60 (hazard ratio, 1.29; 95% confidence interval [CI], 1.03–1.61) was prognostic for shorter bsPFS. After adjusting for the effect of age, breast cancer subtype, and first local treatment modality, presence of neurological symptoms at BrM diagnosis was independently prognostic for shorter bsPFS (hazard ratio, 1.35; 95% CI, 1.02–1.78) (Table 4).

Table 4. Factors associated with brain-specific progression-free survival in patients with BCBM Variable Hazard Ratioa 95% CIa p valuea Univariate analysis Age at BrM diagnosis, yr ≤60 1 >60 1.29 1.03–1.61 .028 Site of metastasis at BCBM diagnosis Lung No 1 Yes 1.09 0.87–1.37 .72 Liver No 1 Yes 1.11 0.89–1.38 .43 Bone No 1 Yes 0.98 0.77–1.24 .93 Node No 1 Yes 0.91 0.72–1.14 .50 Neurological symptom at time of BrM diagnosis No 1 Yes 1.34 1.02–1.64 .082 Subtype .09 HR+/HER2− 1 HER2+ 0.87 0.68–1.11 .26 Triple negative 1.19 0.90–1.58 .22 First local treatment modality .07 WBRT 1 Surgery 0.76 0.55–1.05 .10 SRS 0.74 0.57–0.96 .023 No treatment 0.71 0.36–1.38 .31 Multivariate analysis Age at BrM diagnosis, yr ≤60 1 >60 1.34 1.07–1.68 .011 Neurological symptom at time of BrM diagnosis No 1 Yes 1.35 1.02–1.78 .037 First local treatment modality WBRT 1 Surgery 0.74 0.53–1.02 .070 SRS 0.73 0.56–0.96 .022 No treatment 0.70 0.36–1.38 .31 Subtype HR+/HER2− 1 HER2+ 0.92 0.71–1.18 .51 Triple negative 1.22 0.92–1.62 .17 a Patients with unknown breast cancer subtype were excluded from this analysis; 587 patients with HR+/HER2-, HER2+ or triple negative breast cancer were included. Abbreviations: BCBM, breast cancer brain metastasis; BrM, brain metastasis; HR, hormone receptor; MBC, metastatic breast cancer; SRS, stereotactic radiosurgery; TNBC, triple-negative breast cancer; WBRT, whole brain radiation therapy.

The median OS of our overall patient population was 5.1 months (IQR 2.0–11.7). In our MVA, OS was significantly associated with the type of first-line local therapy, as well as the subtype of breast cancer. For example, compared with patients treated with WBRT in the first-line setting, those treated with up-front surgery (hazard ratio, 0.20; 95% CI, 0.08–0.49) or SRS (hazard ratio, 0.37; 95% CI, 0.22–0.63) lived longer (Table 5; Fig. 2). In addition, patients with HER2+ MBC had superior outcomes compared with those with TNBC and HR+/HER2− disease, respectively, in terms of median OS (12.2 vs. 3.6 vs. 4.4 months), median bsPFS (7.0 vs. 3.1 vs. 3.8 months), time to radiation retreatment (9 vs. 6.4 vs. 7.4 months), and time from retreatment to death (10.5 vs. 2.4 vs. 3.5 months). Additional prognostic factors associated with shorter OS in our MVA included age >60 years (hazard ratio, 1.94; 95% CI, 1.32–2.86; p < .001) and the presence of neurological symptoms at BrM diagnosis (hazard ratio, 1.84; 95% CI, 1.06–3.18; p < .05) (Table 5).

Table 5. Factors associated with overall survival of patients with BCBM