Development of central nervous system metastases as a first site of metastatic disease in breast cancer patients treated in the neoadjuvant trials GeparQuinto and GeparSixto

Background The incidence of central nervous system (CNS) metastases in breast cancer patients is rising and has become a major clinical challenge. Only few data are published concerning risk factors for the development of CNS metastases as a first site of metastatic disease in breast cancer patients. Moreover, the incidence of CNS metastases after modern neoadjuvant treatment is not clear. Methods We analyzed clinical factors associated with the occurrence of CNS metastases as the first site of metastatic disease in breast cancer patients after neoadjuvant treatment in the trials GeparQuinto and GeparSixto (n = 3160) where patients received targeted treatment in addition to taxane and anthracycline-based chemotherapy. Results After a median follow-up of 61 months, 108 (3%) of a total of 3160 patients developed CNS metastases as the first site of recurrence and 411 (13%) patients had metastatic disease outside the CNS. Thirty-six patients (1%) developed both CNS metastases and other distant metastases as the first site of metastatic disease. Regarding subtypes of the primary tumor, 1% of luminal A-like (11/954), 2% of luminal B-like (7/381), 4% of HER2-positive (34/809), and 6% of triple-negative patients (56/1008) developed CNS metastases as the first site of metastatic disease. In multivariate analysis, risk factors for the development of CNS metastases were larger tumor size (cT3–4; HR 1.63, 95% CI 1.08–2.46, p = 0.021), node-positive disease (HR 2.57, 95% CI 1.64–4.04, p < 0.001), no pCR after neoadjuvant chemotherapy (HR 2.29, 95% CI 1.32–3.97, p = 0.003), and HER2-positive (HR 3.80, 95% CI 1.89–7.64, p < 0.001) or triple-negative subtype (HR 6.38, 95% CI 3.28–12.44, p < 0.001). Conclusions Especially patients with HER2-positive and triple-negative tumors are at risk of developing CNS metastases despite effective systemic treatment. A better understanding of the underlying mechanisms is required in order to develop potential preventive strategies.


Background
Central nervous system (CNS) metastases in breast cancer patients are a clinically relevant problem. CNS metastases are associated with shorter survival and impaired quality of life compared with extracranial metastases. Despite the use of neurosurgery and radiotherapy, only the minority of patients survives longer than 1 year [1].
Depending on tumor subtype, survival times of 3.7-15 months after the occurrence of CNS metastases have been described [2]. As in the primary tumor setting, patients with triple-negative breast cancer (TNBC) have the worst prognosis. In a retrospective study by Niikura et al. with 1256 patients diagnosed with brain metastases, the median overall survival of TNBC patients after the CNS metastasis diagnosis was 4.9 months and that of human epidermal growth factor receptor 2 (HER2)-positive patients was 11.5 months [3] In the treatment of non-metastatic breast cancer, relevant improvement has been achieved in the last years resulting in prolonged survival. Neoadjuvant systemic treatment has become a standard procedure for the treatment of primary breast cancer. Pathologic complete response (pCR) rates after neoadjuvant chemotherapy are increasing with current treatment standards and are reflected by an improved patient outcome. pCR rates of 58% for HER2 positive and 37% for TNBC could be achieved [4]. A 5-year breast cancer-specific survival rate of 96% and 75% was observed in an analysis of Boughey et al. for HER2-positive and TNBC patients, respectively [5].
However, about 4% of HER2-positive and 7% of TNBC patients develop CNS metastases after the adjuvant treatment (5-year cumulative incidence of CNS metastases) in historic cohorts [6], and it remains unclear which patients are at high risk for the development of CNS metastases after neoadjuvant treatment with the current standard.
So far, limited insight is available into the biology of CNS metastases in breast cancer patients. It could be assumed that patients who develop CNS metastases as a first site of metastatic disease have distinct clinical or tumor features and that these might help to better understand factors that are a predisposition for CNS metastases. However, only limited studies are published concerning risk factors for the development of CNS metastases. Moreover, most studies on this topic evaluated cohorts with only a small number of patients that did not receive treatment regimens as of the current standard. This is of relevance since modern targeted therapy could potentially influence the pattern of metastatic spread. We therefore investigated the clinical factors associated with the occurrence of CNS metastases as the first site of metastatic disease in 3160 breast cancer patients after modern neoadjuvant systemic therapy.
The preliminary results of our evaluation were presented at the San Antonio Breast Cancer Symposium in 2017 [7].

Methods
The clinical data for the analysis were derived from the neoadjuvant trials GeparQuinto and GeparSixto of the German Breast Group, consisting of 3160 patients, who have been treated with neoadjuvant therapy for early breast cancer.
In GeparSixto, patients with previously untreated, non-metastatic, TNBC (n = 315) and HER2-positive (n = 273) breast cancer were treated with paclitaxel and non-pegylated liposomal doxorubicin. Patients with triple-negative breast cancer received simultaneous bevacizumab. Patients with HER2-positive disease received simultaneous trastuzumab and lapatinib. Patients were randomly assigned to receive, at the same time as the backbone regimens, either carboplatin or no carboplatin [11].
Written informed consent was obtained from all patients before enrolment in the GeparQuinto and GeparSixto trials.

Statistics
The primary objective of our analysis was to assess the influence of the baseline characteristics and pCR on the occurrence of CNS metastases as the first site of metastatic disease (with or without simultaneous occurrence of other metastases) in both trials mentioned above. It was done by assessing time (in months, from randomization in the study) to the occurrence of the CNS metastasis as the first site of metastatic disease according to the competing risk model of Fine and Gray [12], and cumulative incidence function was presented graphically; other distant metastases, contralateral breast cancer, secondary malignancies, or death before any event were considered competing events.
The following covariates were included into a Fine-Gray model: age in years (continuous), tumor stage (cT1-2 vs. cT3-4), nodal status (cN0 vs. cN+), primary tumor subtype (TNBC, HER2 positive/negative, luminal (hormone receptor-positive, HER2-negative) A-like (grades 1-2) and B-like (grade 3)), and study (to adjust for possible heterogeneity). Additionally, a multivariate Fine-Gray model including covariates above and pCR was performed to explore what portion of the effect is mediated by pCR; to avoid guarantee-time bias [13], a landmark of 24 weeks for GeparQuinto patients and 18 weeks for GeparSixto patients was used.
Secondary endpoints were as follows: To assess time (in months, from randomization in the study) to the occurrence of the non-CNS metastasis as the first site of relapse (and to compare to the time to occurrence of the CNS metastases as the first site of metastatic disease). CNS metastases, contralateral breast cancer, secondary malignancies, or death before any event were considered competing events. The cumulative incidence functions for CNS and non-CNS metastases were presented graphically. To describe the first metastatic site at the first distant relapse. The data concerning the first metastatic site was collected in the context of follow-up investigations.
The following categories were assessed: If a patient had CNS metastasis as a first event (yes/ no), if CNS metastases were the only localization (yes/no), or if CNS metastases were diagnosed simultaneously with other metastases (skin, bone, liver, lung/pleura, non-locoregional lymph node, or other localization) If a patient had non-CNS metastases as a first event.
The following categories of the metastasis localization could be documented: skin, bone, liver, lung/pleura, non-locoregional lymph node, or other localization The statistical analyses were performed with SAS 9.2 under SAS Enterprise Guide 4.3 and with SAS 9.4 (for competing risk analyses).

Development of metastases (CNS metastases and non-CNS metastases)
After a median follow-up of 61 months (IQR 45-73), 108 (3%) of a total of 3160 patients developed CNS metastases as the first site of metastatic disease, 411 (13%) patients had distant metastases outside the brain as the first site of metastatic disease, and 2641 (84%) of the patients had no metastatic disease ( Fig. 1: cumulative incidence of metastases). A total of 72 patients (2%) had CNS metastases as the only localization of the metastatic disease, and 36 patients (1%) had other distant metastases in addition to CNS metastases.
CNS metastases as the first site of metastatic disease occurred less frequently than other metastases (5-year cumulative incidence of CNS metastases was 3.8%, and cumulative incidence of the metastases outside the brain was 14.6% according to the competing risk model, Fig. 1: cumulative incidence of metastases).
The median age of patients with CNS metastases as the first site of metastatic disease was 49 years (range 27-75). Seventy-nine percent of the patients (n = 85) were initially treated within GeparQuinto, and 21% (n = 23) within the GeparSixto study.
Patients with non-CNS metastases as the first site of metastatic disease had a median age of 48 years (range 21-76). Eighty-six percent of the patients (n = 352) were initially treated within GeparQuinto, and 14% (n = 59) within the GeparSixto study.
Sixty-three percent (n = 259) of the patients with non-CNS metastases as the first site of metastatic disease had a cT1-2 tumor at the time of diagnosis, 60% of the patients had node-positive disease (cN1-cN3, n = 244), and 52% had a high-grade tumor (n = 213 with a G3 tumor). Ninety-two percent of the patients in this group had no pCR after the neoadjuvant chemotherapy For the detailed summary concerning patient's characteristics, see Table 2.
Factors associated with the occurrence of CNS metastases as the first site of metastatic disease In univariate analysis, larger tumor size, node-positive disease, negative hormone receptor status (both estrogen and progesterone negative), poor tumor differentiation (G3 tumor, high Ki67), HER2-positive or triple-negative tumor subtype, and no pCR after neoadjuvant chemotherapy were significantly associated with the occurrence of CNS metastases as the first site of metastatic disease.
Thus, the 5-year cumulative incidence for the occurrence of CNS metastases was 5.92% for patients with a larger initial tumor size (cT3-4) compared to 3.07% for patients with a smaller initial tumor size (cT1-2) (hazard ratio (HR) 2.19, 95% confidence interval (CI) 1.50-3.22, p < 0.0001). Patients with a node-positive disease had a 5-year cumulative incidence for the occurrence of CNS metastases of 5.74%. The risk of node-negative patients was 2.07% (HR 3.05, CI 1.98-4.71, p < 0.0001). Concerning tumor subtype, a 5-year cumulative incidence for CNS metastases was 4.77% in HER2-positive patients and 6.27% in triple-negative patients. Thus, HER2 positivity was associated with a fourfold and triple negativity with a fivefold higher risk for CNS metastases (HR 4.1, CI 2.06-8.12, p < 0.0001 and HR 5.43, CI 2.81-10.47, p < 0.0001) compared to luminal A-like tumor subtype.
We additionally evaluated the incidence of CNS metastases among patients with HER2-positive hormone receptorpositive and HER2-positive hormone receptor-negative patients. No significant difference concerning the incidence of CNS metastases could be shown in our cohort of patients between these groups of patients (HR 0.8, 95% CI 0.5-1.22, p = 0.268).
The 5-year cumulative risk for development of CNS metastases in patients without achieving a pCR after neoadjuvant chemotherapy was 4.18% vs. 2.49% for patients with a pCR (HR 1.85, CI 1.06-3.24, p = 0.0316).
The BRCA status did not show a statistically significant association with the occurrence of CNS metastases as the first site of metastatic disease in our cohort. The data was available for a total of 36 of 108 patients with brain metastases (n = 9 BRCA positive, n = 27 BRCA negative). Of these patients with known BRCA status, those with a positive BRCA status had a 5-year CNS metastasis incidence rate of 7.84% compared to 4.68% by patients with a negative BRCA status (HR 1.67, 95% CI 0.78-3.57, p = 0.189).
We additionally analyzed the incidence of CNS metastases (relating to subtype) among patients without a pCR. A 5-year cumulative CNS metastasis incidence rate in the cohort of patients without a pCR was 1.3% for luminal A-like subtype, 2.53% for luminal B-like, 5% for HER2-positive, and 8.46% for triple-negative patients. The risk of development of the CNS metastases as the Fig. 1 Cumulative incidence of metastases first site of metastatic disease was statistically higher in patients with HER2-positive breast cancer as well as for triple-negative patients (HR 3.66, 95% CI 1.78-7.55, p = 0.0004 and HR 6.50, 95% CI 3.32-12.71, p < 0.0001) compared to patients with luminal A-like breast cancer.
In the cohort of patients with pCR, the median survival was not achieved to the time of evaluation. The 5-year cumulative incidence rates for the patients with pCR after the neoadjuvant chemotherapy were 2.4% for CNS metastases and 5.7% for non-CNS metastases.

Discussion
CNS metastases have become a major challenge in the management of patients with metastatic breast cancer, as the occurrence of CNS metastases is associated with a particularly bad prognosis. The incidence of CNS metastases in patients that have received neoadjuvant treatment according to current standards has not been reported so far. Only a few reports have examined risk factors for the development of CNS metastases as the first site of metastatic disease in breast cancer patients.
Identification of risk factors for developing CNS metastases would identify a patient group that might benefit from preventive management strategies of CNS metastases, e.g., cranial irradiation or CNS metastasis screening, e.g., by means of magnetic resonance imaging. Both options are not part of the management of patients with breast cancer in clinical routine today. However, the low incidence of CNS metastases as the first metastatic site observed in our cohort and the retrospective design of the study would not justify a screening approach during follow-up. The development of a calculator for the risk assessment could be an aim of the further evaluation of the collected data. On the basis of our analysis, we can recommend to take particular account into neurological complaints of patients with until now not metastasized breast cancer with initially large tumor size, node-positive disease, no-pCR status after systemic therapy, and HER2-positive or triple-negative tumor biology. We suggest carrying out clinical studies investigating the role of screening options (e.g., brain magnetic resonance imaging) for metastatic breast cancer patients who are at high risk for developing CNS metastases. In our cohort of patients after neoadjuvant chemotherapy, 16% of patients developed metastatic disease after a median follow-up of 5 years. Three percent of patients developed CNS metastases as the first metastatic site. Factors significantly associated with the occurrence of CNS metastases as the first site of metastatic disease were higher tumor stage before therapy (cT3-4), node-positive disease, HER2-positive or triple-negative subtype, and no pCR after neoadjuvant chemotherapy. We could show that tumor subtype remained the highest risk factor for CNS metastases. Our results are predominantly in line with already published analyses of smaller patient cohorts not receiving current treatment regimens.
Gonzales-Angulo et al. [14] determined the incidence of CNS metastases and examined associated disease characteristics in patients with locally advanced breast cancer or inflammatory breast cancer after a neoadjuvant systemic therapy. Five percent of patients (38/768 patients) developed CNS metastases as the first site of metastatic disease. Negative hormone receptor status, nuclear grade 3, positive nodal status, and higher stage of disease were significantly associated with the time to CNS metastasis in this patient cohort.
Other researchers evaluated, in contrary to our design, patient cohorts with either adjuvant systemic treatment or mixed patient cohorts with adjuvant and neoadjuvant approach. Thus, Dawood et al. showed an association of CNS metastases as the first site of metastatic disease and higher breast cancer stage for patients with initially non-metastatic triple-negative breast cancer (n = 115) [15]. Patients in this cohort received either adjuvant or neoadjuvant systemic treatment. Atahan et al. observed a correlation between CNS metastases as the first site of metastatic disease and the initial size of the primary tumor and nodal status in 32 patients. These patients received an adjuvant chemotherapy [16].
A different probability of breast cancer subtypes to metastasize into the brain was also described by others [17][18][19][20]. Soni et al. and Kennecke et al. detected an association between HER2-positive and TNBC subtype and the development of CNS metastases. Martin et al. evaluated 968 patients with brain metastases at the time of diagnosis of breast cancer and detected that incidence proportions were highest among patients with hormone receptor-negative HER2-positive and triple-negative subtypes [21]. Sihto et al. [19] and Smid et al. [20] showed an association between basal-type cancer and CNS metastases. Lin et al. detected that triple-negative tumors were associated with a greater risk of brain metastases relative to hormone receptor-positive/HER2-negative tumors [22]. Vaz-Luis et al. evaluated 3394 patients with HER2-positive breast cancer and could show that hormone receptor-negative HER2-positive patients were more likely to have recurrence in the brain compared to hormone receptor-positive HER2-positive patients [23].
A noteworthy finding in our study relates to pCR rates in the analyzed subgroups. The pCR rate in patients with CNS metastases as the first site of metastatic disease was higher (15%) compared to patients with other distant metastases as the first site of metastatic disease (9%). Thus, a good response rate to a systemic treatment seems not to have a deciding role in the prevention of CNS metastases. Possibly, more aggressive disseminated tumor cells, which also did not respond to a systemic treatment, gain a propensity to overcome the blood-brain barrier and survive in the central nervous system. Most systemic treatment approaches are assumed not to cross the intact blood-brain barrier; this fact might explain why tumor cells that already reached the CNS cannot be affected. Probably, the initially more aggressive tumor is also a significant risk factor in the development of CNS metastases. In our cohort, 16 patients developed CNS metastases as the first site of metastatic disease despite achieving a pCR. All of them had either HER2-positive (n = 9) or TNBC (n = 7) primary tumor.
The strength of our analysis is a large cohort of patients who received up-to-date treatment with anthracyclines and taxanes as well as up-to-date HER2-directed therapy. However, given the relatively low incidence of CNS metastases, we cannot identify differences between the incidence of CNS metastases in different treatment groups of the trials, e.g., between lapatinib-and trastuzumab-treated patients with HER2-positive breast cancer or in the cohort of triple-negative patients with carboplatin therapy compared to patients without carboplatin therapy. A small number of patients with an available BRCA status in our cohort also do not allow a confident statistical statement concerning the correlation with the development of CNS metastases and may be regarded solely as trend. A further limitation of our analysis is that the incidence of CNS metastases after the occurrence of metastatic disease was not documented, as in many other trials. This point must be taken into consideration by the interpretation of Fig. 1. The flattening of the curve after 24 months does not prove that CNS metastases do not appear after this period of time as the curve does not include patients that developed CNS metastases after diagnosis of other distant metastases.

Conclusions
In conclusion, our analysis showed that especially patients with HER2-positive and triple-negative tumors are at risk of developing CNS metastases despite active systemic treatment, especially when no pCR was reached. However, also patients with pCR are at risk of developing CNS metastases. A better understanding of the underlying mechanisms is required in order to develop potential preventive strategies and should help to identify molecular mechanisms of resistance to modern therapies and clonal selection of tumor cells with special brain tropism. Further studies should aim to identify a group of patients that might have a benefit of screening for brain metastases. Further research on biomarkers for CNS metastasis development should be performed. A translational research project in our cohort is ongoing.