Stereotactic irradiation of non-small cell lung cancer brain metastases: evaluation of local and cerebral control in a large series

Stereotactic radiotherapy (SRT) of brain metastases (BM) results are often reported in the heterogeneous primitive population. Here, we report our experience in consecutively treated patients who underwent SRT alone for BM from non-small cell lung cancer (NSCLC). This retrospective analysis included consecutive patients with no history of cerebral treatment who underwent Cyberknife™ SRT for BM from NSCLC in our institution from 2007 to 2016. One hundred patients were included in the study, with a median follow-up of 33 months (20–64). Mean age was 63 years (SD ± 10); 88% had Karnofsky Performance Status (KPS) > 70; 67% had unique BM; 18 patients received single-fraction SRT (20–25 Gy), and 82 received hypo-fractionated SRT (HSRT) (24–36 Gy in 3–5 fractions). We reported a complication rate of 17% (2% of G3-4). Median survival was 10.1 months [95% confidence interval (CI) 7.8–13.9]. At 1 year, local and cerebral control rates were respectively 78.7% (95% CI 70–86.5%) and 43% (95% CI 33.5–53%). Thirty patients underwent salvage treatment (whole brain radiation therapy, n = 13; SRT, n = 14; surgery, n = 3). Cyberknife™-based SRT is an effective treatment associated with high local control rate with low morbidity for patients with NSCLC’s BM. Close follow-up is necessary to perform salvage treatment.

www.nature.com/scientificreports/ survival for younger patients (< 50 years old), solitary BM, and GPA > 2 [6][7][8] . Adding WBRT to SRT reduces the development of distant BM, but it comes at the cost of cognitive toxicity 9,10 . There are only retrospective data evaluating the results of SRT in a homogenous population of BM from NSCLC 11,12 , showing a good local control at 1 year (80-90%) and distant brain control (60-80% ) rates after Gammaknife™ (GK) procedure. The goal of our study was to evaluate our practice for SRT with Cyberknife™ (CK) in this population.

Methods and materials
Data acquisition. Patients' agreement before using their data was required. Informed consent was obtained for all patients. The Institutional Committee on Human Research of Centre Oscar Lambret (Lille) has approved this retrospective study. All research was performed in accordance with relevant regulations.
Using the local database, we retrospectively studied the medical records of all patients aged 18 or over who underwent SRT by CK for the treatment of all of their NSCLC BM in our institution between 2007 and October 2016. All patients with previous history of cerebral treatment (surgery or radiotherapy) were excluded.
All lesions detected by magnetic resonance imaging (MRI) were irradiated by Cyberknife™. Baseline characteristics, treatment modalities, acute toxicities (CTCAE v4), clinical and radiological outcome every 3 months, and salvage therapy data were recorded for each patient. BM were considered metachronous if time from primary diagnosis was > 6 months.

Statistical analysis.
Clinical and treatment characteristics were described by median, range, mean, and standard error for continuous variables and by frequency and percentages for categorical variables. Treatment response was first estimated using the response evaluation criteria in solid tumors (RECIST) criteria, and then using the response assessment criteria for brain metastases (RANO-BM) criteria when clinical status was available. RANO-BM has been developed to avoid equating treatment effect with tumor progression. Indeed, it takes into account clinical assessment (neurological examination), corticosteroid dosing, and needs a 6-12 weeks imaging confirmation to assess progression.
Survival endpoints were computed from the date of the end of radiotherapy. Overall survival was estimated as time interval to death using Kaplan-Meier method, and patients alive were censored at the date of last news. Time to local progression (local control), time to cerebral progression (cerebral progression away from the treated site), time to neurological deterioration, and time to salvage treatment were estimated as time interval using the competing risk method (Kalbleisch and Prentice).
Associations between clinical or treatment factors and survival endpoints were analyzed using Cox proportional hazard regression models for overall survival, and using Fine and Gray models for time to cerebral progression, time to local progression, and time to neurological deterioration. Variables associated with survival with a significance level p < 0.10 in univariate analysis were included in multivariate regression models. A backward stepwise selection of variables was performed if the number of variables to include in the model was too large compared to the number of observed events (at least 10 events should be observed by variable). The significance level was set to p < 0.05. The statistical software used was Stata v13.1 (StataCorp, 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp, LP).

Results
patient and treatment characteristics. One hundred patients were included between 2007 and 2016.
The mean follow-up period was 33 months (range 20-64 months). Patients' demographics and clinical data are summarized in Table 1. Mean age was 63 years (SD ± 10 years); sex ratio was 3.3 men for 1 woman. KPS at baseline was 90-100, 70-80, and < 70 for 30%, 58%, and 12%, respectively. Forty-two percent of patients presented with extra-cranial metastasis. Forty-four percent were symptomatic for their BM. Sixty-seven percent had a solitary BM, 23% had 2 BM, and 10% had 3 or more BM. Mean long axis of BM was 16.6 mm (± 10.4 mm). Mean time from simulation and diagnosis MRI to treatment was, respectively, 18.1 ± 12.1 days and 27.7 ± 22.5 days. SRT was conducted with Cyberknife™. The gross tumor volume (GTV) was outlined on both enhanced contrast CT images and CT-MRI fused images. Planning target volumes (PTVs) were determined by expansion of the GTV with a 1 mm margin. The most frequent schedules were 27 Gy in 3 fractions (69%), 18-25 Gy in 1 fraction (18%), and 30 Gy in 5 fractions (9%), prescribed on isodose 80%.
There were no factors associated with cerebral control at 1 year using univariate and multivariate analysis. We observed a trend towards a worse cerebral control in tumors located in the posterior fossa (p = 0.06 in univariate and multivariate analysis).
Using univariate analysis, increasing number of BM (p = 0.004), metachronous BM (p = 0.041), and increasing time between simulation and treatment (p = 0.04) were associated with a lower local control. Salvage therapy. At 1 year, 22 and 56 local and cerebral progressions have been reported according to RECIST respectively. After initial SRT, a total of 30 patients received salvage therapy: 14 received salvage SRT, 13 received salvage WBRT, and 3 underwent salvage surgery as their first salvage treatment (Fig. 4). Sixty patients died without requiring a second brain treatment.     with EGFR-mutant adenocarcinoma and BM treated with icotinib exhibited prolonged survival, and a longer duration of cerebral control was observed with brain radiotherapy 13 . In a multi-institutional analysis Magnuson et al. demonstrated that the use of upfront EGFR-tirosyne kinase inhibitor (TKI), and deferral of radiotherapy, is associated with inferior overall survival in patients with EGFR-mutant NSCLC who develop brain metastases. Radiosurgery followed by EGFR-TKI resulted in the longest overall survival and allowed patients to avoid the potential neurocognitive sequelae of WBRT 14 . The optimal combination of SRT and the several TKI has not yet been determined. It would have been interesting for our population treated between 2007 and 2016 because some of them may have been treated with anti-EGFR or immunotherapy. However, genetic alterations were not determined in most of our patients due to the recruitment time.
Our overall survival, local and cerebral control rates, and acute toxicities are in agreement with the literature (Table 3). In the published studies, mainly retrospectives, median survival is approximately 10 months, 1-year local and cerebral control rates are, respectively, 70-90% and 40-60% 7, 11-16 . Only a few retrospective studies have evaluated SRT for homogenous NSCLC's BM populations. Zairi et al. reported better outcomes, but the population was strictly selected (82% of solitary BM, with limited size), as suggested by the one-year cerebral control rate of 79% versus 43% in our series 11 .
Several factors have been demonstrated to influence local tumor control achieved with SRT. These include prescribed radiation dose, lesion size, histology, and lesion morphology (tumor necrosis is associated with a poor response) 17 . A review of retrospective studies showed a local control rate > 70% for a BED 12 > 40 Gy (for α/β = 12 Gy) 18 . Rodrigues et al. published a predictive score for BM control after SRT, including dose and morphology, based on retrospective data 19 . This score should be validated prospectively. In our study, we calculated BED 12 of the SRT schedules: only 1 received less than 40 Gy (treated with 24 Gy in 4 fractions). We did not report BM morphology, but we found that adenocarcinoma histology was associated with lower neurological deterioration (p = 0.001), as well as solitary BM (p < 0.001), suggesting that these patients have the best clinical benefit from SRT procedures.
SRT, compared with WBRT, offers the possibility of salvage therapy 15 . In our study, more than half of patients who had a cerebral progression at 1 year (43%) underwent salvage therapy (25% (95% CI 17-33.8%) at 1 year), mainly by WBRT (13% (95% CI 7.3-20.4%) at 1 year) or SRT. However, most patients (n = 60) died without salvage therapy at the time of analysis. This is in agreement with literature. McTyre et al., in a retrospective analysis of 2,657 patients who underwent BM SRT, found a high rate of death without cerebral relapse 16 . Thus, some asymptomatic patients may not benefit from SRT procedures as they die rapidly from extra-cerebral disease. A careful selection of patients should be done before undergoing SRT.
Our series shows good local control after SRT with low acute toxicity; these patients require regular follow-up because of the risk of brain progression. Nevertheless, more than half of the patients did not require a second brain treatment. We have confirmed the prognostic role of the GPA score in multivariate analysis and patients with preserved health status and a limited number of metastases are the best candidates for SRT as recommended in the guidelines 20, 21 . www.nature.com/scientificreports/ conclusion While individualized oncologic treatments are becoming a common practice, it is necessary to evaluate BM management in a homogeneous primitive population. In patients with a limited number of BM from NSCLC, SRT is an effective treatment associated with high local control rate, low neurological deterioration, and with low morbidity. Close follow-up with MRI is mandatory to perform salvage treatment. Patients with BM from NSCLC should be included in randomized trials evaluating new systemic therapies, to monitor synergic effects (efficacy or toxicity), and to clarify the role of SRT in the global care of their BM.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.