Stereotactic Radiation Therapy for Brain Metastases: Factors Affecting Outcomes and Radiation Necrosis

Simple Summary Brain metastases constitute a severe event in many patients affected by solid tumors. Indeed, even in those cases in which the original disease is sensitive to a systemic treatment, the particular vascularization of the brain may limit its efficacy in the site. Stereotactic radiation therapy (SRT) plays a major role in the multidisciplinary management of oncological patients with brain metastases (BMs). SRT is generally delivered in single or multiple (3–5) fractions. Data from 87 analyzed patients treated at our institution suggest that this technique is characterized by a good effectiveness in local control and patients with stable extracranial disease benefit most from SRT. Tumor histology does not affect local control. Radiation necrosis (RN) occurrence was registered in 16% of treated sites, and it appeared to be related to left location and adenocarcinoma histology, while chemotherapy reduced the risk. When RN occurs, prompt recognition is needed to establish a treatment. Abstract Stereotactic radiation therapy (SRT) is a proven effective treatment for brain metastases (BM); however, symptomatic radiation necrosis (RN) is a late effect that may impact on patient’s quality of life. The aim of our study was to retrospectively evaluate survival outcomes and characterize the occurrence of RN in a cohort of BM patients treated with ablative SRT at Federico II University Hospital. Clinical and dosimetric factors of 87 patients bearing a total of 220 BMs treated with SRT from 2016 to 2022 were analyzed. Among them, 46 patients with 127 BMs having clinical and MRI follow-up (FUP) ≥ 6 months were selected for RN evaluation. Dosimetric parameters of the uninvolved brain (brain without GTV) were extracted. The crude local control was 91% with neither clinical factors nor prescription dose correlating with local failure (LF). At a median FUP of 9 (1–68) months, the estimated median overall survival (OS), progression-free survival (PFS), and brain progression-free survival (bPFS) were 16, 6, and 9 months, respectively. The estimated OS rates at 1 and 3 years were 59.8% and 18.3%, respectively; bPFS at 1 and 3 years was 29.9% and 13.5%, respectively; PFS at 1 and 3 years was 15.7% and 0%, respectively; and local failure-free survival (LFFS) at 1 and 3 years was 87.2% and 83.8%, respectively. Extracranial disease status was an independent factor related to OS. Fourteen (30%) patients manifested RN. At multivariate analysis, adenocarcinoma histology, left location, and absence of chemotherapy were confirmed as independent risk factors for any-grade RN. Nine (20%) patients developed symptomatic (G2) RN, which improved or stabilized after 1–16 months of steroid therapy. With prompt recognition and, when necessary, medical therapy, RN radiological and clinical amelioration can be obtained.


Introduction
The incidence of brain metastases (BMs) is increasing, due to the improvement of both diagnostic tools and oncological treatment for the primary tumor [1]. The most recent guidelines for the management of patients with BMs provided by the European Association of Neuro-Oncology and the European Association of Medical Oncology (EANO-ESMO) from 2021 [2] remind that, for optimal therapeutic strategy planning, age, performance score, histotype, and cranial and extracranial disease status should be examined [3]. Surgery should be considered when there is doubt about the neoplastic nature, when the primary is unknown or the primary rarely generates BMs, when the change in molecular profile can affect the decision making [1], or when there are acute symptoms of increased intracranial pressure [4][5][6]. Stereotactic radiation therapy (SRT) is recommended for patients with a limited number (1 to 9) and size of BMs (typically a cumulative volume lower than 15 cc) [2] and typically with a Karnofsky performance status (KPS) ≥ 70 and stable extracranial disease [7,8]; SRT is also recommended after surgery for improving local control [9]. However, radiation necrosis (RN) is a serious late complication, with a 5-25% reported incidence [10,11]. The pathophysiological responsible mechanisms include changes in blood vessel fibrinoids, coagulative necrosis, demyelination, and gliosis [12,13]; disruption of the blood-brain barrier is in part mediated by VEGF, released in response to hypoxia [14]. RN typically develops in the brain parenchyma adjacent to the tumor site-typically the uninvolved brain parenchyma receiving the highest dose [15]. Clinical or magnetic resonance imaging (MRI) features may help in the differential diagnosis between RN and disease relapse/recurrence, but a biopsy may be needed for a definitive diagnosis, particularly in patients who are symptomatic and have worsening imaging findings over time [11]. RN could be asymptomatic, with evidence only at imaging, or symptomatic and requiring treatment [16][17][18][19]. This single academic center study aims to retrospectively evaluate survival and RN outcomes in patients treated for BM with ablative SRT. Clinical and dosimetric factors associated with patient outcomes were investigated.

Population Selection and Data Collection
Between June 2016 and November 2022, 113 consecutive patients and 257 BMs were treated with ablative SRT at Federico II University Hospital, Naples, Italy. In the present study, all patients with available follow-up (FUP) information were included. Patient-, BM-, and treatment-related characteristics were collected. Patients with at least 6 months of clinical and MRI FUP were evaluated for RN occurrence.

Simulation, Planning, and Treatment
Thermoplastic masks for IMRT treatments with reinforcing bands and 3.2 mm thickness were used for the CT simulation. Thermoplastic pillows were additionally used in some patients to increase the degree of immobilization. CT scan started from the vertex to the second cervical vertebra, setting FOV L 360 and image reconstruction thickness at 2 mm. CT images were transferred to MIM Maestro ® contouring software version 6.6.7 (MIM Software, Cleveland, OH, USA) and then to Pinnacle PHILIPS TPS software version 9.10 (Philips, Amsterdam, The Netherlands). The GTV was contoured after rigid registration on T1-weighted MRI sequences. To obtain the PTV, the GTV (or CTV in postsurgery) was given an isotropic 2-5 mm expansion, depending on the lesion size (lower volumes required greater expansion to reach ≥1 cc volume) [20]. For planning purposes, the following organs-at-risk (OARs) were contoured: optic pathway, lens, eyes, brainstem, and brain. Patients were treated with single-fraction SRT (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)  prescription isodose surface was chosen such that 95% of PTV received a minimum of 95% of the prescription dose. All the plans were developed with Volumetric Modulated Arc Therapy (VMAT) technique with noncoplanar approach when necessary [21]. The treatments were delivered by Varian TrueBeam STx version 2.0. The institutional online IGRT protocol consisted of a prefraction cone beam CT. Dexamethasone 4 mg was generally prescribed for 3-5 days.

Follow-Up
At discharge, the patients were given indications for antiedema therapy continuation and oncological treatment restart if a temporary discontinuation during radiation treatment had occurred. Patients were followed with serial MRI and clinical re-evaluation after SRT. Telematic clinical monitoring was also offered. Post-treatment MRI was performed 6 to 8 weeks following SRT and every 3 months thereafter, unless a closer follow-up was required, possibly with advanced MRI technique integration. The tumor response was evaluated according to the Response Assessment in Neuro-Oncology Brain Metastases (RANO-BM). CT or 18 FDG PET/CT periodic extracranial disease re-evaluation was performed, and the following information was then updated: any changes in the KPS and any progression/switching to a next oncological therapeutic line.

Outcome Measure
Overall survival (OS) was evaluated from the end of SRT. Local failure-free survival (LFFS) was defined as time from treatment and the event of local failure (LF) in the treated field once pseudo-progression had been excluded; brain progression-free survival (bPFS) was defined as time from treatment and the event of LF or new metachronous BM appearance; extracranial progression-free survival (ePFS) was defined as time from treatment and progression in any site but brain; and progression-free survival (PFS) was defined as time from treatment and the event of intracranial or extracranial progression. All survival outcomes were evaluated by patient, except for factors possibly related to LF, evaluated by single BM.

Radiation Necrosis Evaluation
RN was evaluated both by patient and by BMs and scored according to the Common Terminology Criteria for Adverse Event (CTCAE) version 5.0. Grade 1 necrosis was defined by growth of a previously treated lesion on MRI with strong radiographic features of necrosis of the surrounding brain parenchyma, asymptomatic, and with intervention not indicated. Grade 2 necrosis was defined by associated moderate symptoms and corticosteroids indicated. Grade 3 was defined by associated severe symptoms and medical intervention indicated. Grade 4 was defined as life threatening, and Grade 5 as death.

Dosimetric Analysis
For each patient, the dose-volume histogram (DVH) of the uninvolved brain contour, obtained as brain with GTV subtraction, was extracted. To consider the different SRT fractionation schemes, all physical doses were voxel-wise converted using MIM Maestro into 2 Gy equivalent dose (EQD2) with α/β = 2 for OARs [22] and 20 [23] for target volumes. When a retreatment was administered, plan summation was obtained after rigid image registration considering previous RT treatments, whole brain, or SRT [24].
The following dosimetric parameters were extracted from accumulated DVHs: uninvolved brain volume receiving more than X Gy (Vx) in increments of 2 Gy and the maximum, minimum, and mean doses (Dmax, Dmin, and Dmean). The dosimetric analysis was performed by single lesion.

Statistical Analysis
Survival outcomes were estimated with the Kaplan-Meier method with log-rank test for subgroup analysis. Cox proportional hazards regression model was used for multivariate analyses.
For single-time-point analysis, the relationships between candidate prognostic (clinical and dosimetric) factors and binary LF, any-grade RN (≥G1), were tested by χ2-test and by Mann-Whitney U test, when appropriate. Of note, the analysis was performed for single BM. All parameters showing a significant correlation (p < 0.05) at univariate analysis were included into a multivariate analysis. Due to the exploratory nature of this analysis, no corrections were made for multiple comparisons. In order to avoid a collinearity problem, a preselection of dosimetric variables was performed removing redundant variables highly correlated with each other (Pearson's correlation coefficient), and only those variables most highly correlated with RN were included in the subsequent analysis. Statistical analysis was performed with SPSS version 28.

Study Population and Treatments
With a median FUP of 9 (1-68) months, 87 patients and 220 BMs met the inclusion criteria. Clinical and treatment characteristics of the analyzed patients are reported in Table 1.
Seven patients, with a lower KPS, after a median time of 12 (2-41) months, underwent whole brain treatment, receiving the dose of 30 Gy in 10 fractions; four died 0.25-18 months later (median, 8 months).
Twenty patients underwent further SRT after a median time of 9 (2-38) months; 11 patients (55%) kept the first assigned KPS unchanged, 4 patients (20%) were given a higher score, and 5 patients (25%) presented a worsened PS. Patients died after a median of 18.5 (3-54) months. A patient with breast primary underwent further SRT at a 4-month follow-up and 5 months later again a WB treatment due to further cranial progression; she then died 9 months later. Two patients, after the first treatment, underwent two further SRTs: one patient is still alive with stable extracranial disease; the other patient, with melanoma primary, was previously treated with whole brain and then underwent three SRTs with death occurring two months after the fourth brain treatment due to both intraand extracranial progression.
Two patients, 10-11 months later, due to local recurrence, underwent reirradiation with a single-fraction treatment delivering 10 Gy, and a fractionated one, delivering the total dose of 18 Gy. In one patient, local control was achieved (he died 20 months later due to extracranial progression); in the other patient, the treated lesion slowly progressed, and he died 17 months later.
Twenty patients underwent further SRT after a median time of patients (55%) kept the first assigned KPS unchanged, 4 patients (20%) score, and 5 patients (25%) presented a worsened PS. Patients died af (3-54) months. A patient with breast primary underwent further SRT a up and 5 months later again a WB treatment due to further cranial pr died 9 months later. Two patients, after the first treatment, underwen one patient is still alive with stable extracranial disease; the other patie primary, was previously treated with whole brain and then underw death occurring two months after the fourth brain treatment due extracranial progression.
Two patients, 10-11 months later, due to local recurrence, unde with a single-fraction treatment delivering 10 Gy, and a fractionated total dose of 18 Gy. In one patient, local control was achieved (he died to extracranial progression); in the other patient, the treated lesion slow he died 17 months later.

Factors Affecting Outcomes
KPS ≥ 70 and stable extracranial disease at time of SRT (log-ran 0.015, respectively) were significantly related to better survival. No sta differences in OS were found between patients treated for single or m 0.82), by age (p = 0.6), or time elapsed from diagnosis to brain metastas survival time between patients undergoing SRT with controlled or disease was 22 ± 3.2 (95% CI, 15.7-28.9) and 12 ± 2.4 (95% CI, respectively (Figure 2). At multivariate Cox regression, extracranial independently related to OS (HR 1.8; 95% CI, 1.02-3.14; p = 0.043).   Univariate analysis did not reveal any factor, clinical (primary histotype), morphological (size and location), or dosimetric (prescription dose), significantly related to the 19 local failure events.

Radiation Necrosis Adverse Events
Forty-six (127 BMs) out of eighty-seven patients with clinical and MRI FUP ≥ 6 months were evaluated for late adverse events. RN occurred in 14 (30%) patients and 20 (16%) BMs, with a median time to onset of 6 (1-46) months. Eight out of twenty (40%) radionecrotic lesions did not cause associated symptoms (G1), while the other twelve (60%) were classified as G2.
In five patients, asymptomatic RN was observed (median time to the imaging appearance 5 (1-19) months); for four patients, radiological stabilization was obtained spontaneously, and they are all still alive except one who died from COVID-19 complications; one patient died shortly after due to leptomeningeal progression.
Nine patients with neurological RN-associated deficits were treated with steroids; all recovered after a variable time (1-16 months) of steroid dependence (Figure 3). One patient developed a G1-RN that became symptomatic 17 months after. Another patient had a new RN progression after 18 months, and steroids were reintroduced. No cases required surgical decompression (G3). Six patients (67%) are still alive. The other three patients (33%) died after a median time of 28 (26-58) months: one patient for cranial progression, another patient for extracranial progression without brain progression, and for the other patient, the details are unknown.
Univariate analysis did not reveal any factor, clinical (primary hist morphological (size and location), or dosimetric (prescription dose), significantly to the 19 local failure events.

Radiation Necrosis Adverse Events
Forty-six (127 BMs) out of eighty-seven patients with clinical and MRI FU months were evaluated for late adverse events. RN occurred in 14 (30%) patients (16%) BMs, with a median time to onset of 6 (1-46) months. Eight out of twenty radionecrotic lesions did not cause associated symptoms (G1), while the other (60%) were classified as G2.
In five patients, asymptomatic RN was observed (median time to the im appearance 5 (1-19) months); for four patients, radiological stabilization was ob spontaneously, and they are all still alive except one who died from CO complications; one patient died shortly after due to leptomeningeal progression.
Nine patients with neurological RN-associated deficits were treated with stero recovered after a variable time (1-16 months) of steroid dependence ( Figure 3 patient developed a G1-RN that became symptomatic 17 months after. Another had a new RN progression after 18 months, and steroids were reintroduced. N required surgical decompression (G3). Six patients (67%) are still alive. The othe patients (33%) died after a median time of 28 (26-58) months: one patient for progression, another patient for extracranial progression without brain progressio for the other patient, the details are unknown.  . Serial axial gadolinium-enhanced T1-w (a) and T2-w (b) sequences in follow-up M ages of a 74-year-old man. The brain metastasis was centrally located, infratentorial, at grey matter interface, in the left occipital lobe. Symptomatic RN arose during the 9th month. months of steroid dependence, clinical and radiological stabilization were obtained. The pat now been steroid-free for 7 months and is in excellent general health. However, a close follo still required.

Factors Affecting Radiation Necrosis
Results from the univariate analysis for candidate factors affecting any-grade R vs. G1-G2) occurrence are reported in Table 2.    year-old man. The brain metastasis was centrally located, infratentorial, at grey-white matter interface, in the left occipital lobe. Symptomatic RN arose during the 9th month. After 9 months of steroid dependence, clinical and radiological stabilization were obtained. The patient has now been steroid-free for 7 months and is in excellent general health. However, a close follow-up is still required.

Factors Affecting Radiation Necrosis
Results from the univariate analysis for candidate factors affecting any-grade RN (G0 vs. G1-G2) occurrence are reported in Table 2.

Discussion
Our findings confirm the effectiveness, in local control, of the ablative radiation treatment, for a limited number of brain metastases, regardless of the presumed histotype's radiosensitivity. The estimated local control at 1 year in our series was 87.2%. Dose, fractionation, and outcomes are in line with those reported in the literature, in which the need for a higher dose is reported only for melanoma metastases. In Redmond et al., from the HyTEC group [23], a model of Tumor Control Probability for the ablative treatment of brain metastases was developed reporting, based on tumor size, local control rates at 1 year ranging from 69 to 95%.
The historical Recursive Partitioning Analysis (RPA) or the more recent Graded Prognostic Assessment (GPA), considers several factors; in our experience, among them, extracranial disease status influenced survival more than the others (age, single or multiple BMs, and performance score) [23].
A third of patients treated at our institution faced retreatments for metachronous lesions or local relapses. Two patients underwent reirradiation: a single-fraction treatment delivering 10 Gy, and a fractionated one, delivering the total dose of 18 Gy, with a median survival of 8 months and further local recurrence in one. In a recent meta-analysis by Loi et al. [25], out of 389 reirradiated lesions, a median dose of 19 (15.5-26.5) Gy was delivered at the time of the second SRT; treatment was delivered using a single-fraction and a multifraction regimen in 72% and 28% of patients, respectively. The local failure rate was 24% at 1 year, suggesting that local control rates after reirradiation do not dramatically differ from those reported on the first SRT, with a median survival time of 14 months.
In our study, any-grade brain RN late events occurred in 30% of patients (16% BMs). Symptomatic RN in our series occurred in 20% of patients (9% BMs). This finding is in agreement with that reported in the literature [11].
The main risk factor for RN has been reported to be lesion size [22]; in our series, we considered PTV instead of GTV size for its better impact on clinical practice. It resulted in a significant risk factor at univariate analysis, but not at multivariate. Of note, the two postsurgery targets developed symptomatic radiation necrosis; this data should not be surprising as tumor bed targets are the biggest, and radiation necrosis affects not the BMs but the surrounding healthy brain parenchyma. In their meta-analysis, Leher et al. [26] suggest the use of multiple fraction treatments for large lesions.
Our findings suggest that adenocarcinoma histology, left location, and absence of chemotherapy course are independent risk factors for RN.
Few authors have evaluated the predictive value of histology on the development of radiation necrosis after SRT for brain metastasis, so the potential impact of tumor histology remains unclear. In some series [27,28], the cancer type was not a significant variable, while in others, renal [29] and lung adenocarcinoma histology have been identified as risk factors [30]; in particular, ALK+ and EGFR+ lesions were associated with higher rates [31].
Few studies suggest that a relationship between RN and BM location is possible. Minniti et al. [32] evaluated 206 patients and 310 BMs treated with SRT as the initial treatment for 1-3 brain metastases. Brain necrosis occurred in 24% of the treated lesions, being symptomatic in 10% and asymptomatic in 14%. A univariate analysis showed that parietal location was a significant variable for any-grade brain necrosis. Also, a deep location, particularly within deep white matter, could influence RN development, as showed by Choi [34].
Many studies have evaluated the relationship between brain necrosis and oncological systemic treatment course. According to Colaco et al. [35], patients who received chemotherapy were found to have a reduced risk of developing radiation necrosis. If chemotherapy has a cellular suppressive effect, which may reduce the inflammatory response to high-dose radiation that causes radiation necrosis, the inflammatory response could instead be exasperated by immune response enhancers. In the study by Martin et al. [36], symptomatic necrosis occurred in 23 out of 115 patients who received immunotherapy (ipilimumab or PD-1 inhibitor) and in 25 out of 365 patients who did not. Tallet et al. [37] also showed an increased RN rate in patients undergoing immunotherapy treatment or BRAF inhibitors. Controversial is the risk brought by T-DM1, an antibody-drug conjugate (trastuzumab and emtansine) that plays a role in the inflammatory response characterized by increased levels of cytokines including tumor necrosis factor. A significantly higher RN has been reported in a series of patients treated with concomitant T-DM1 for HER2+ breast cancer brain metastases [38,39]. Significantly increased risks of post-SRT necrosis were observed also with concomitant use of VEGFR and EGRF tyrosine kinase inhibitors. The use of ALK inhibitors, on the other hand, does not seem to increase the risk of RN [22].
Previous whole-brain treatment does not seem to increase the risk of RN in our and other series; the risk of RN is instead influenced by previous focal ablative treatments [22]. However, in our series, the accumulated Dmax to the uninvolved brain, which considers all possible previous brain treatments, was significantly related to RN, at least at the univariate analysis. It should be underlined that there is still no uniformity regarding the definition of the brain as an organ at risk; some consider the total brain, others consider the brain minus the gross tumor volume. According to Milano et al. [22], for a singlefraction SRS of brain metastases, brain tissue volumes (including target volumes) that receive 12 Gy of 5 cm 3 , 10 cm 3 , or >15 cm 3 are associated with the risk of symptomatic brain necrosis of about 10%, 15%, and 20%, respectively. For 3-fraction FSRT, normal brain tissue V18 < 30 cm 3 and V23 < 7 cm 3 are associated with a <10% risk of necrosis [22]. In the study by Dohm et al. [40], brain necrosis occurred in 4 of 39 larger, unresectable lesions, treated with two-staged radiosurgery separated by one month; brain V20 to 87.8 values were analyzed as factors potentially related to necrosis, with significant p values at V44.5 to 87.8 Gy. Radiation necrosis is thus certainly related to the delivered dose, but a standardized dosimetric reporting is needed.
There are some limitations of the current study. Due to a retrospective and single institution source, patient selection bias is possible. In addition, the statistical power is limited by the small sample size. For the above issues, this study has an exploratory aim, and our study findings are hypothesis-generating. Further studies on large populations are needed to develop and validate a robust RN prediction model.

Conclusions
Our experience confirms that ablative radiation treatment of brain metastases is effective, with excellent local control. Patient prognosis remains poor and depends mostly on extracranial disease status. Radiation necrosis is a late event that can affect the quality of life in long-surviving patients; therefore, it must be recognized promptly and, if necessary, treated. Radiation necrosis seems to depend on histology and laterality. Chemotherapy seems to decrease the risk. The literature data also suggest paying attention to bigger targets, dose, deep location, or when concomitant immunotherapy or target therapy are used.