Efficacy and Tolerance of Post-operative Hypo-Fractionated Stereotactic Radiotherapy in a Large Series of Patients With Brain Metastases

Purpose: The aim of this study was to assess, in a large series, the efficacy and tolerance of post-operative adjuvant hypofractionated stereotactic radiation therapy (HFSRT) for brain metastases (BMs). Materials and Methods: Between July 2012 and January 2017, 160 patients from 2 centers were operated for BM and treated by HFSRT. Patients had between 1 and 3 BMs, no brainstem lesions or carcinomatous meningitis. The primary endpoint was local control. Secondary endpoints were distant brain control, overall survival (OS) and tolerance to HFSRT. Results: 73 patients (46%) presented with non-small cell lung cancer (NSCLC), 23 (14%) had melanoma and 21 (13%) breast cancer. Median age was 58 years (range, 22–83 years). BMs were synchronous in 50% of the cases. The most frequent prescription regimens were 24 Gy in 3 fractions (n = 52, 33%) and 30 Gy in 5 fractions (n = 37, 23%). Local control rates at 1 and 2 years were 88% [95%CI, 81–93%] and 81% [95%CI, 70–88%], respectively. Distant control rate at 1 year was 48% [95%CI, 81–93%]. In multivariate analysis, primary NSCLC was associated with a significant reduction in the risk of death compared to other primary sites (HR = 0.57, p = 0.007), the number of extra-cerebral metastatic sites (HR = 1.26, p = 0.003) and planning target volumes (HR = 1.15, p = 0.012) were associated with a lower OS. There was no prognostic factor of time to local progression. Median OS was 15.2 months [95%CI, 12.0–17.9 months] and the OS rate at 1 year was 58% [95% CI, 50–65%]. Salvage radiotherapy was administered to 72 patients (45%), of which 49 received new HFSRT. Ten (7%) patients presented late grade 2 and 4 (3%) patients late grade 3 toxicities. Thirteen (8.9%) patients developed radiation necrosis. Conclusions: This large multicenter retrospective study shows that HFSRT allows for good local control of metastasectomy tumor beds and that this technique is well-tolerated by patients.


INTRODUCTION
Brain metastases (BMs) are the most frequent brain tumors, and, throughout disease course, 20-40% of cancer patients will develop a BM (1). In subjects in good general health and presenting with a single BM, surgical resection has been shown to improve survival (2,3). After surgery, adjuvant wholebrain radiotherapy (WBRT) allows to significantly reduce local and brain recurrence rates, as well as the risk of death from neurological cause (4, 5). Nevertheless, WBRT has not been shown to be beneficial in terms of overall survival (4-6) and the length of time in which patients remain functionally independent (4, 5). In addition it contributes, in the short term, to a poorer quality of life in patients (6) and causes acute toxicities including asthenia, alopecia, nausea, and a decline in learning and memory functions (7). Stereotactic radiosurgery (SRS) allows for good local control of the disease while avoiding the neurocognitive decline triggered by WBRT (8). Consequently, after resection of a BM, SRS and Hypofractionated stereotactic radiation therapy (HFSRT) are increasingly being used and could be considered as an alternative treatment standard to WBRT allowing to limit toxicity (7,8).
To date, there is no consensus on the optimal dose, fractionation, or prescription regimens of HFSRT on the surgical cavity. Several prescription patterns are described in the literature, including schemas of 3 fractions with doses ranging from 7.7 to 11 Gy, (9)(10)(11)(12) or schemas of 5 fractions (13,14). Such heterogeneity in prescription doses prevents any direct comparison between studies. The largest phase III randomized study, comparing SRS to WBRT published by Brown et al. showed a longer cognitive-deterioration-free survival in patients assigned to SRS (median 37 months) than in patients assigned to WBRT (median 30 months) (p < 0.0001) (8). Overall survival (OS) was identical in the 2 arms, but local and distant brain control were lower in the SRS arm.
The aim of this study was to evaluate the efficacy and safety of post-operative HFSRT in resection cavity of secondary brain lesions in a large cohort of patients.

Population
Between July 2012 and January 2017, patients treated with postoperative HFSRT to the resection cavity in two French centers were included. Data were retrospectively collected. Inclusion criteria were: adult patients, with 1 to 3 BMs, no previous radiotherapy treatment to the brain, treated by surgery for BM of a solid tumor and with anatomical pathology data, no brainstem lesion or carcinomatous meningitis, eligible to be treated by HFSRT as decided in a multidisciplinary meeting, with a life expectancy of more than 3 months, and not opposed to the use of their medical data for research and educational purposes.

HFSRT Technique
Patients were immobilized using a thermoplastic mask system. Computed tomography (CT) scan and gadolinium contrastenhanced magnetic resonance imaging (MRI) were used for treatment planning. Imaging was performed using millimetric slices and rigid registration. Target volumes and organs at risk were contoured on MRI and concordance with CT was controlled. Contouring software's used were Oncentra (version 4.3.0) and Multiplan (Accuray, version 3.2.0).
Target volumes were contoured using the surgical and anatomical pathology assessment of resection specimens. The clinical target volume (CTV) included the surgical cavity, contrast enhancement of tumor border and a 1-2 mm margin which delineates it on the CT scan and planning MRI. In the case of metastasis in contact with dura, the CTV included a larger margin (5-10 mm) beyond the area where there was contact before surgery. The gross tumor volume (GTV) was defined if macroscopic disease could be identified by nodular contrast enhancement by T1-Gadolinum MRI imaging. The planning target volume (PTV) was defined as CTV + 1 mm.
HFSRT treatment was delivered using a CyberKnife R -type robotic accelerator (Accuray, Inc., Sunnyvale, CA), using 6 MeV photon beams, in Centre Oscar Lambret in Lille and Centre François Baclesse in Caen. Dose was prescribed at the 80% isodose and patients were treated every 2 days.

Follow-Up
Follow-up of patients included collection of clinical data and brain perfusion MRI at 2 months and then every 3 months after the end of irradiation during the first year, and every 4 to 6 months thereafter. Local recurrence was defined as the appearance or growth of nodules in the surgical cavity visible on a T1-gadolinium MRI sequence. OS was defined as the time from HFSRT until death from any cause. Time-to distant brain control (DBC) was defined as to the time from HFSRT until progression in the brain outside of the surgical cavity. Radiation necrosis was diagnosed based on clinical, morphological, and metabolic criteria, and was validated by experts. MR spectroscopic imaging and 18F-DOPA PET (Positron emission tomography) were used to support the diagnosis if needed.

Statistical Analysis
Patient and disease characteristics were analyzed using descriptive statistics. Quantitative variables were expressed as median and range. Survival curves were estimated using the Kaplan Meier method. For time to progression, patients were censored at the date of last news or date of death from any cause. Time interval for overall survival was calculated from the date of HFSRT to the date of death from any cause. Patients alive were censored at the date of last news. Patients were censored at day 1 in case of missing information on the event.
After having checked the proportional hazard assumption (Schoenfeld residuals), prognostic factors of survival were identified using a univariate Cox regression model. Hazard Ratios (HR) and the 95% CI as well as the calculated probability (p-value) were presented for each model. In cases of non-proportional hazards, the "restricted mean survival time difference" (RMSTD) was used (15). Significant variables at p = 0.10 in the univariate model were included in the multivariate Cox stepwise backward model analysis. The following factors were analyzed: sex, age, primary disease, primary histology, RPA (recursive partitioning analysis) score, DS-GPA (Diagnostic-Specific Graded Prognostic Assessment) score, controlled primary tumor, location of BM, extracranial metastasis status, number of BM, time between primary tumor and BM diagnosis, partial resection and gross total resection, interval time between surgery and HFSRT, dose of HFSRT, salvage WBRT, SRS or HFSRT, pre-and post-operative volumes, conformity index, and homogeneity index.
The association of the radiation necrosis with the different factors was analyzed with the Fisher exact test for qualitative variables and with the Wilcoxon Mann-Whitney test for quantitative variables.
Statistical analyses were performed using Stata 13.1 (StataCorp. 2013 Stata Statistical Software: Release 11. College Station, TX: StataCorp LP) and significance level was set at a p-value of 0.05 for all analyses.

Population
This was a retrospective study involving 160 patients and 167 surgical cavities. Patients were 76 women (47.5%) and 84 men. The median age at diagnosis of BM was 58 years (range, 22-83 years) ( Table 1). Seventy-three patients (46%) presented with primary lung cancer, 23 patients (15%) with melanoma and 21 patients (13%) with breast cancer. The median time interval between the primary tumor diagnosis and BM surgery was 8.4 months (range, 0-148.6 months).

Description of BM and Treatment by HFSRT
At the time of diagnosis, 115 patients (72%) had a single BM; 77 patients (50%) had symptoms of intracranial hypertension  Figure 1A). No factor appears to be prognostic of local control.  (Figure 2). In the univariate analysis, different prognostic factors appears to be associated with overall survival. Lung primary tumor was associated with a significant reduction of the risk of death  Table 3). The number of BM did not achieve significance with a HR = 1.16 (p = 0.055).

Tolerance
According to the Common Terminology Criteria for Adverse Events (CTCAE) grading system, version 4.03, 5 patients (3.4%) presented an acute grade 2 toxicity and 1 patient (0.7%) presented an acute brain hemorrhage of grade 3. Ten patients (7.2%) developed late toxicity of grade 2 and 4 patients (2.7%) a late toxicity of grade 3 (two brain necroses, one seizure and one stroke). Radiation necrosis during follow-up occurred in 13 patients (8.9%). The stereotactic re-irradiation or WBRT was the only factor associated with an increased risk of developing a radiation necrosis (p < 0.001, Fisher exact test). Among patients that received HFSRT exclusively, the rate of radiation necrosis at the end of follow-up was 6.9% and among the 7 patients treated  with stereotactic re-irradiation in the surgical cavity, 4 (57%) presented with a radiation necrosis.

DISCUSSION
This large series, the second largest after Keller et al.'s to our knowledge (12), shows that post-operative HFSRT to the surgical cavity of BM allows for a good local control with acceptable acute and late toxicity profiles.

Local Control
Local control rates achieved in our series with HFSRT to the surgical cavity are comparable to those of the larger retrospective series. Keller

Early Local Recurrence
In our study, 46 (30%) patients presented a nodular contrast enhancement by planning MRI, even though resection was macroscopically complete in 92% of the cases. This diagnosis is difficult, and because RANO and RECIST 1.1 criteria are not adapted in the post-surgery setting, radiologists used heterogeneous methods for diagnosis (21,22). In the study from Jarvis et al. before post-operative radiosurgery, 12% of the patients presented a local recurrence at 1 month. The early local recurrence rate was 37.5% at 1 month in patients with a subtotal resection (23). In these studies, the median delay between surgery and radiotherapy was 4-7 weeks (11,13,18), but could range from 18 days (14,24) to 4.5 months (25).

Distant Brain Control
In our series, DBC rates are comparable to those reported in the literature after SRT (12). The phase III studies from Mahajan et al. and de Kocher et al. reported similar distant brain control rates, 43% at 1 year and 58% at 2 years, respectively in the observation arms (4, 16). In this study, leptomeningeal disease seems more frequent (5,12,26 (12). In Atalar's study, 161 brain metastasis resection cavities treated from 1998 to 2011 with post-operative SRS were retrospectively reviewed. One and 2 year rates of leptomeningeal disease were 13%, but until 34% at 1 year for breast cancer (26). In our study, the rate of leptomeningeal disease may have been overestimated as we also reported very moderate leptomeningeal disease on MRI.

Overall Survival
Median OS of patients in our study was 15.2 months [95% CI, 12.0-17.9 months] and it was comparable to that observed in other studies. In the meta-analysis by Gans  In the meta-analysis by Gans et al. a higher prevalence of single metastases in the cohort was the only factor associated with higher OS (p < 0.02) (17). The study by Keller et al. reported in multivariate analysis, that a RPA score of 3 (p = 0.02), piecemeal resection (p = 0.017) and an increased number of BMs (p < 0.001) were independent prognostic factors for a lower OS. Patients with multiple BMs had a risk of death 2.4 times greater than patients with a solitary BM (p < 0.001) (12). Kocher et al. randomized phase III study evaluating the interest of adjuvant WBRT found in a multivariate analysis that the only factors with a significant impact on survival with Performance Status (PS) ≤ 2 were the initial PS (0 vs. 2, p = 0.004) and the presence of macroscopic tumor outside the brain (absent vs. present p = 0.001) (4).
Based on 7 randomized studies of the RTOG and 2,350 patients treated for BMs, Barnholtz-Sloan et al. developed a nomogram to estimate OS in patients with BM (27). The model revealed that the primary site was predictive of OS, with breast cancer and lung adenocarcinoma being associated with improved survival. Contrary to previous studies, in our series, the survival of patients with NSCLC could be improved with treatments including immunotherapy, targeted therapy and third generation tyrosine kinase inhibitors (28).

Safety and Radiation Necrosis
The tolerance was acceptable with 2.7 and 0.7% of patients presenting acute grade 2 and grade 3 toxicities, and 7.2 and 2.7% late grade 2 and 3 toxicities, respectively. These results are in line with the 10% toxicity rate after HFSRT revealed by the meta-analysis by Gans et al. (17). Risk of radiation necrosis has been shown to decrease with lower doses, greater number of fractions and smaller volume of the treated surgical cavity (29,30). Eaton et al. demonstrated that for cavities bigger than 3 cm, treatment by radiosurgery was associated with a greater rate of radiation necrosis, with a HR = 3.81 [95% CI, 1.04-13.93], (p = 0.043) compared to treatment by HFSRT (29). The risk of radiation necrosis at 1 year was of 10.3% with HFSRT and of 19.2% after radiosurgery. In our study, 8.9% of patients presented a radiation necrosis during follow up and only re-irradiation was found to be predictive of radiation necrosis, possibly due to a lack of statistical power related to a low number of events. Median volumes of brain that received doses of 10 Gy (V 10 Gy ), 12 Gy (V 12 Gy ), and 21 Gy (V 21 Gy ) were not found to influence radiation necrosis. In Minniti et al.'s study using a schema of 3 × 9 Gy, the rate of radiation necrosis was 7% at 1 year and 16% at 2 years (31). This study showed that V 24 Gy was the only factor associated with radiation necrosis with a 16.8 mL threshold (p = 0.03).
In order to standardize practice, Soliman et al. recently published CTV contouring guidelines for SRS of completely resected cavity BM defined by 10 experts based on 10 clinical cases (32). Our delineation practices are in line with these guidelines. Improvements in local control can be achieved by adding a 2 mm margin around the resection cavity (33). The choice of 1 mm to define PTV is arguable. In our study, this PTV allowed to compensate for repositioning errors. In other studies a margin between 0 and 4 mm was more frequently used. However, in Gans et al.'s meta-analysis the use of a margin to define PTV did not allow to improve local control or OS (17).

Limitations
This is a retrospective study and several irradiation schemas were used. Nevertheless, the prescription regimen at the 80% isodose was homogeneous. For 55% of patients a dose of 8, 9 or 10 Gy in 3 fractions was prescribed.

CONCLUSION
This large retrospective multi-center study shows that, in our population of patients operated for BM, adjuvant treatment by HFSRT allows for good local control in the surgical cavity. This non-invasive technique was well-tolerated by patients. HFSRT is an efficient treatment option for patients with operated BM. The rate of distant recurrence and in particular leptomeningeal disease seems higher than the rate observed after WBRT. A close follow-up by MRI is necessary in patients with a high risk of intra-cerebral recurrence.

DATA AVAILABILITY
All datasets generated for this study are included in the manuscript and/or the supplementary files.

ETHICS STATEMENT
This study was carried out in accordance with the recommendations of local ethics committee and conducted in accordance with the Helsinki declaration with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Centre O. Lambret ethics committee.

AUTHOR CONTRIBUTIONS
DP and GM contributed to the conception, design, and manuscript preparation. GM performed data acquisition. GM, JG, DS, EB, ER, NR, EE, SM-M, RM-A, LB, XM, EL, and DP performed data analysis and interpretation. EB carried out statistical analysis.