Bevacizumab for radiation necrosis following radiotherapy of brain metastatic disease: a systematic review & meta-analysis

Radiotherapy is the mainstay of brain metastasis (BM) management. Radiation necrosis (RN) is a serious complication of radiotherapy. Bevacizumab (BV), an anti-vascular endothelial growth factor monoclonal antibody, has been increasingly used for RN treatment. We systematically reviewed the medical literature for studies reporting the efficacy and safety of bevacizumab for treatment of RN in BM patients. PubMed, Medline, EMBASE, and Cochrane library were searched with various search keywords such as “bevacizumab” OR “anti-VEGF monoclonal antibody” AND “radiation necrosis” OR “radiation-induced brain necrosis” OR “RN” OR “RBN” AND “Brain metastases” OR “BM” until 1st Aug 2020. Studies reporting the efficacy and safety of BV treatment for BM patients with RN were retrieved. Study selection and data extraction were carried out by independent investigators. Open Meta Analyst software was used as a random effects model for meta-analysis to obtain mean reduction rates. Two prospective, seven retrospective, and three case report studies involving 89 patients with RN treated with BV were included in this systematic review and meta-analysis. In total, 83 (93%) patients had a recorded radiographic response to BV therapy, and six (6.7%) had experienced progressive disease. Seven studies (n = 73) reported mean volume reductions on gadolinium-enhanced T1 (mean: 47.03%, +/− 24.4) and T2-weighted fluid-attenuated inversion recovery (FLAIR) MRI images (mean: 61.9%, +/− 23.3). Pooling together the T1 and T2 MRI reduction rates by random effects model revealed a mean of 48.58 (95% CI: 38.32–58.85) for T1 reduction rate and 62.017 (95% CI: 52.235–71.799) for T2W imaging studies. Eighty-five patients presented with neurological symptoms. After BV treatment, nine (10%) had stable symptoms, 39 (48%) had improved, and 34 (40%) patients had complete resolution of their symptoms. Individual patient data was available for 54 patients. Dexamethasone discontinuation or reduction in dosage was observed in 30 (97%) of 31 patients who had recorded dosage before and after BV treatment. Side effects were mild. Bevacizumab presents a promising treatment strategy for patients with RN and brain metastatic disease. Radiographic response and clinical improvement was observed without any serious adverse events. Further class I evidence would be required to establish a bevacizumab recommendation in this group of patients.


Introduction
Brain metastasis (BM) is the most common adult intracranial disease, and it is diagnosed in approximately 20 to 30% of cancer patients [1][2][3]. The most common primary tumor metastasizing to the brain is lung cancer (up to 50%), followed by breast cancer (up to 25%), melanoma (up to 20%), and to a lesser extent, renal cell carcinoma, colorectal cancer, and others [1][2][3][4]. Nonetheless, the incidence and frequency of BM is growing as newer systemic and immunotherapeutic agents are entering the treatment paradigm of these primary cancers [5][6][7][8][9]. Patients are living longer and are more prone to experience BM in their lifetime.
Depending on various prognostic factors, management of BM may involve surgical resection and/or radiation therapy in the form of stereotactic radiosurgery (SRS), whole brain radiotherapy, or a combination of two [1,[10][11][12][13]. A surge has been witnessed in the use of radiosurgery in BM patients with the approval of various targeted and immunotherapeutic agents for the management of primary sites of systemic cancers [6][7][8][9]14]. Targeted agents after SRS for the brain have also been continued and have prolonged survival outcomes for patients with BM [6,7,9,15,16]. Radiation therapy has long been associated with the development of radiation necrosis (RN) in patients with intracranial disease [17][18][19][20][21]. The rate of RN following radiotherapy or radiosurgery has been estimated at 10-15% [17][18][19][20][21]. RN is considered as a dose-limiting toxicity for SRS [20,21]. An increased incidence of RN has also been reported with a combination of SRS and systemic agents [22,23]. In fact, the benefits of synergism from a combination of radiation and targeted agents are weighed against RN toxicity [7,22,23]. Hence, the management of RN takes a center stage in patients with intracranial disease.
Corticosteroids have long been the mainstay of RN treatment. It inhibits the pro-inflammatory response that promotes radiation necrosis and provides symptomatic relief via edema reduction, but long-term use is associated with serious side effects [19]. Surgery has also been used for resectable progressive RN, which can relieve mass effects and it also provides an opportunity to study tissue samples for diagnosis. However, persistent edema may need close monitoring for weeks [19,24]. Another treatment strategy employed is hyperbaric oxygen therapy (HBOT) [25]. It can increase oxygen concentration to stimulate angiogenesis, restore blood supply to necrotic lesions, and accelerate healing. It has also shown improvement in RN symptoms alone or in combination with Endostar (a recombinant endostatin product) [25]. Laser interstitial thermal therapy (LITT) has been demonstrated to relieve RN symptoms, reduce progression, and improve survival in patients with RN and brain metastases [26,27]. It has also been used to complement RN surgery [24]. Bevacizumab (BV) has also made it a treatment paradigm for RN [28][29][30]. Recent clinical trials have shown encouraging results [31][32][33].
Bevacizumab, an anti-VEGF monoclonal antibody, has been evaluated for RN treatment [28][29][30]. Its use in RN stems from the fact that RN tissues have elevated levels of VEGF [34,35]. Radiotherapy induces vasogenic edema and ischemia, resulting in hypoxia that leads to the induction of hypoxia-inducible factor 1α (HIF1α) [34][35][36][37][38]. HIF1α upregulates VEGF through astrocytes and endothelial cells [36,38]. White matter around necrotic areas has been identified as the main VEGF up-regulating site [36]. Immunohistochemistry (IHC) of RN surgical samples has confirmed increased levels of VEGF in reactive astrocytes surrounding the core of necrotic tissue [37]. VEGF is an important regulator of angiogenesis, leading to increased vascular permeability, damage to the bloodbrain barrier (BBB), and ensuing brain edema [39]. Bevacizumab reduces vascular permeability and alleviates blood-brain barrier damage and brain edema through its binding to VEGF [28,35,39].

Methods & materials
Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were rigorously followed [59].

Inclusion criteria Patients & study types
Studies reporting the efficacy of bevacizumab for radiation necrosis occurring in patients with brain metastases after undergoing radiotherapy for intracranial disease.

Outcomes of interest
Outcomes of prime interest were: radiographic response; edematous volume reductions on magnetic resonance imaging (MRI); and clinical improvement such as improvement/resolution of neurological symptoms and signs, increase in Karnofsky Performance Status (KPS) score, and decrease in dosage or discontinuation of dexamethasone. The secondary outcomes of interest were recurrence and safety outcomes, including adverse events.

Search strategy Databases
PubMed, Medline, EMBASE, and the Cochrane library were searched until 1st Aug 2020. Various search terms such as "bevacizumab" OR "Anti-VEGF monoclonal antibody" AND "Radiation necrosis" OR "Radiation induced brain necrosis" OR "RN" OR "RBN" AND "Brain metastases" OR "BM" etc., were employed. Language was restricted to English. Furthermore, references of the retrieved studies were also inspected for more relevant literature.

Study selection
Relevant studies obtained from databases were imported into Endnote X9.3 software for organization and screening. Duplicates were removed and titles and abstracts were thoroughly screened. Studies were selected according to the aforementioned inclusion criteria. In situations of discrepancies, other authors were consulted.

Data extraction
"The Cochrane Collaboration Data Collection form-RCTs and non-RCTs" was modified according to our requirements and used for recording data. The extracted data included general characteristics/attributes of the studies and participants and the main outcomes of interest. The characteristics of the studies recorded were the first author, publication year, period of recruitment, research design, institute of research, number of participants, and follow-up time. The recorded attributes of participants included age, sex, presenting symptoms, KPS, dexamethasone use, and adverse events.
Furthermore, outcomes of interest, including radiographic response, RN volume reduction on MRI images, clinical improvement, and safety. Scrutiny and examination of eligible studies was accomplished with full text reading by two independent reviewers (M.K. and Z.Z).

Assessment of risk for bias
Quality assessment was carried out using the Reporting Checklist for Authors developed by The Meta-analysis Of Observational Studies in Epidemiology (MOOSE) Group [60].

Statistical analysis
Descriptive statistics, including frequency, percentage, mean, median, range, and standard deviation, were calculated with Microsoft Excel for Mac 2019 v16.43. Mean reduction rates were directly extracted from the studies or indirectly via Engauge Digitizer. The weighted mean and standard deviation was estimated according to the methods described in the Cochrane Handbook for Systematic Reviews of Interventions version 6.0 [61,62]. Pooled estimates (weighted mean and confidence interval) was obtained with Open Meta Analyst software, which uses the R package "metafor" for meta-analysis [63][64][65]. The pooled mean was estimated using a continuous random effects model with the DerSimonian-Laird method [66]. Heterogeneity was assessed using the I 2 test. I 2 values of 25, 50, and > 50% were considered as low, moderate, and high heterogeneity [67]. P < 0.05 was considered statistically significant.

Measurement of MRI changes and calculation of reduction rate
Slight variations were noticed in methods for assessing the volume calculation and reduction rate on MRI images among the studies. Two studies estimated the area of lesion at the level of maximum bi-dimensional measurement according to McDonald's criteria, and the difference was expressed as percent change from the baseline MRI profiles [50,54,75]. In some studies, the hyperintense area was manually outlined, measured, and summed across slices and was multiplied by the layer thickness to calculate the total lesion area, but the reduction rate was estimated differently [33,[51][52][53]55]. Volume reduction was obtained by subtracting of posttreatment from pre-treatment volume, dividing posttreatment by pre-treatment volume, and the following formula: volume before BVvolume after BV / volume before BV [33,[51][52][53]55]. Zhuang et al. calculated the edema index as: EI = volume of (edema + necrosis)/volume of necrosis [33,55]. For T1 MRI, changes in the signals were measured in three different areas in the    strengthening region of necrosis and compared to the white matter signal value of the same MRI to obtain a ratio that was used to express the reduction rate as the difference between pre-and post-treatment [33,55]. We calculated the difference from the graphs available in their studies.

Recurrence
Only one study (n = 14) reported a recurrence rate [55]. The recurrence rate was very high: 10 of the 13 responding patients had RN recurrence. Sadraei et al. also           reported that four patients had RN recurrence, but the type of intracranial disease (primary brain tumor, NPC, or BM) was not identified [54]. A single patient in the study by Wang et al. also had recurrence with no evidence of intracranial disease type [50].

Adverse events
Overall, five studies (n = 63) reported adverse events occurring in 14 (22%) patients after bevacizumab treatment (Table 5) [33,[54][55][56]58]. A retrospective study reported grade 1 side effects in two (14%) patients. Adverse events  Fig. 2 Forest plot of meta-analysis of mean reduction rate on T1-contrast enhanced MRI (a) and T2W FLAIR MRI (b) after bevacizumab (BV) treatment for radiation necrosis (RN) in patients with brain metastases reported were mild allergy and hypertension [55]. Hypertension resolved spontaneously. Similar side effects (mild allergy, hypertension) in two (9.5%) patients were reported in a prospective clinical trial conducted by the same group [33]. Side effects reported for individual patients were available in the study by Sadraei et al. [54]. One patient with non-small cell lung carcinoma (NSCLC) reported grade 1 proteinuria, for which bevacizumab treatment was withheld. Similarly, the other NSCLC patient reported a grade 2 urinary tract infection that also required withholding one dose of BV treatment. Of the 17 patients with RN, five (29%) patients (two with NSCLC, one with melanoma, one with breast cancer, and one with NSTC) reported side effects after BV treatment. Grade 3 deep venous thrombosis (DVT) and pulmonary embolism (PE) were observed in melanoma patients. The patient with breast cancer reported grade 2 fatigue, and the NTSC patient experienced grade 2 hypertension. All the participants in the case series (n = 4) reported by Tanigawa et al. experienced side effects involving hypertension, edema, and proteinuria [56]. Only one patient had experienced side effects such arthralgia and dysgeusia in the study by Glitza, I. et al. [58]. Adverse events were not reported in the remaining studies [48][49][50][51][52][53]57].

Discussion
We retrieved studies evaluating the efficacy of BV in the management of RN in patients who had received radiation therapy for brain metastases [33,[48][49][50][51][52][53][54][55][56][57][58]. Most patients showed a reduction in the edema and RN volume by over 50% on MRI images until their last follow-up [33,[48][49][50][51][52][53][54][55][56]. In some studies, edema volume reduction was over 70% in patients with BM [52,53]. Radiographic   responses corresponded with improvements in clinical outlook. Neurological symptoms were stabilized, improved, or completely resolved upon BV induction (Table 3). Several studies have reported a similar efficacy data for BV in patients with primary brain tumors (gliomas and glioblastoma), and NPC [31,32,[45][46][47]. In a study by Wang et al., there were patients with other primary brain tumors who demonstrated a similar efficacy in reducing edema volume (T1 post-gd: 61%, T2 FLAIR: 57%), and showed improvement in neurological symptoms (100%) [50]. Fursue, et al. study, as well, had eight patients who had RN with primary brain tumors, other than the three BM patients [52]. A mean edema volume reduction rate of 45% was revealed for these patients. In addition to BM patients, seven other patients (five primary brain tumors and two arteriovenous malformations (AVM) patients) were also included in the study by Sadraei et al. [54].  [47]. All these studies, however, constitute a low-level clinical evidence for the efficacy of BV therapy [45-47, 50, 52, 54, 56-58]. Zhuang et al. conducted a prospective clinical trial involving 21 patients who had RN with brain metastatic disease [33]. All patients, except for one, showed radiographic improvement. There is class I evidence for patients with primary brain and NPC tumors [31,32]. Levin et al., in a randomized placebocontrolled trial, using a bevacizumab dose of 7.5 mg/kg every 3 weeks for seven patients with biopsy-proven RN with primary brain tumors, showed an average percentage change of 59 and 63% in RN volume on T1W and FLAIR images, respectively [31]. A recently concluded RCT involving 58 NPC patients treated with bevacizumab revealed a 65.5% (38/58) response rate [32]. The mean percentage change in RN volume observed on T1 post-gd and T2W FLAIR MRI were 25.5 and 51.8%, respectively. The mean change between before and after bevacizumab treatment was significant for both detected MRI images. Both these studies have reported significant differences in the radiographic responses and RN volume reduction rates observed on both MRI images between bevacizumab and placebo/corticosteroids, suggesting a better outcome for bevacizumab [31,32].
In our systematic review, one study reported a very high RN recurrence rate (77%) in BM patients [55]. Other studies have failed to report recurrence of such a magnitude. Other than the two studies mentioned in the Results section, there are few other studies that also have cases of RN recurrence [50,54]. Two patients in the RCT conducted by Levin et al. reported RN recurrence in glioma patients [31]. NPC patients from two other studies have also shown a moderate rate of recurrence [32,47]. A recurrence rate of 39.5% was observed in a retrospective study of 50 NPC patients [47]. A similar recurrence rate (36.8%) was also demonstrated in the RCT of 58 NPC patients conducted by Xu et al. [32]. The underlying mechanism has not been exclusively investigated in these patients. Apparently, all three kinds of intracranial diseases (primary brain tumors, metastatic, or NPC) have registered RN recurrence [31,32,47,55]. Recurrence was slightly higher in BM patients as reported, but the study had a low level of evidence. Hence, no conclusions could be drawn about the relationship between RN recurrence and the underlying intracranial disease type. Zhuang et al. identified a correlation between RN recurrence and duration after the initial BV withdrawal [55]. Further, Li et al. indicated that duration from induction of radiation therapy RN diagnosis and BV intervention as predictive factors for RN recurrence [47]. Further investigations are required to establish any underlying cause of RN recurrence. Another important aspect of RN recurrence is its diagnosis. Pathology is the standard for diagnosing RN or recurrence [76][77][78][79]. However, almost all of these studies relied on imaging criteria reported in previous studies for the diagnosis of RN and recurrence [31,32,47,55,[76][77][78][79]. For example, in a case report, re-enlargement of RBN after being on BV for 8 months was attributed to recurrence of lung cancer as resected specimen revealed necrotic areas with viable tumor cells [80]. Hence, an accurate recurrence rate could only be determined with pathology, which could be further examined by larger comprehensively organized trials.
In this systematic review, clinical improvement was observed in a majority of the patients; however, some patients did not show any clinical improvement or experienced symptomatic worsening and progression. Medical literature also reveals similar examples. In the study by Gronier et al., no clinical improvement was observed in all three participants with malignant brain tumors after BV therapy (10 mg/kg per month) [81]. One patient had experienced lymphopenia after one perfusion of bevacizumab; the other had developed a transient ischemic attack and a corneal ulcer. Side effects reported in our review were mild, and only one grade 3 pulmonary embolism was described [33,[54][55][56][57][58]. Several other investigations have also highlighted similar low-grade adverse events [31,32,46,50,53]. In the retrospective study of Torcuator, et al. (n = 6), only one patient experienced mild fatigue after BV treatment [46]. Grade 2 AEs, including hypertension, fatigue, and proteinuria, were observed in 18% (3/17) of participants of the study by Wang, et al. [50]. However, the patients' primary intracranial diseases were not identified. In the study by Yonezawa, et al., 33% (3/11) of participants had also shown grade 1 or 2 side effects such anemia, leukopenia, neutropenia, and lymphocytopenia [53]. More importantly, the class I evidence in this regard has shown the safety of BV therapy in primary brain tumors and NPC patients [31,32]. Levin et al. reported that six (55%) patients experienced side effects [31]. Three of these adverse events were considered serious, including aspiration pneumonia, pulmonary embolus secondary to DVT, and superior sagittal sinus thrombosis. The other three patients showed ischemic changes due to small vessel thrombosis [31]. Another RCT conducted by Xu et al. reported 40 grade 1 or 2 adverse events experienced by 58 patients with NPC [32]. Only one grade 3 adverse event of ischemic stroke was observed. Furthermore, a similar portfolio was revealed for the corticosteroid-treated group, suggesting that BV treatment may not increase the toxicity experienced by patients with RN [32].
From the literature, it appears that bevacizumab was able to elicit therapeutic efficacy at any prescribed dose or frequency [31][32][33][40][41][42][43][44][45][46][47][48][49][50][51][52][53][54][55][56]. The initial doses used were 5, 7.5, 10, and 15 mg/kg every 2 weeks to every 6 weeks. All doses were tolerated and were not associated with any increase in toxicity. It has been suggested that BV efficacy is associated with its anti-angiogenic effects rather than the dose [33]. In a case report, BV at a dose as low as 3 mg/kg was shown to be effective [48]. In a prospective clinical trial, patients were exposed to ultra-low doses of BV at 1 mg/kg [33]. Radiographic responses were observed in 20 of the 21 patients. Such a versatile dosing profile makes this treatment reachable to a broader population, as it is an expensive treatment. To date, exact cost-benefit relationship evaluation has not been adequately addressed for bevacizumab therapy [29]. It may cost around 4800 to 19,200 U.S. dollars (USD) for a single four to eight-week course of 5 to 10 mg/kg, administered every other week at a cost of 600 USD per 100 mg [82,83]. An increase of 2.4 months in survival, a 20% improvement in a patient's quality of life, or a linear combination of the two was required for bevacizumab treatment to be considered cost-effective according to a basic hypothetical calculation using 10,000 USD cost for a course of BV therapy and a quality-adjusted-life-year (QALY) threshold of 50,000 USD [84]. Hence, further studies are needed to establish a dose requirement for achieving the maximum benefit and to make the bevacizumab treatment cost-effective.
Several observations limit the results of our study. As a systematic review, the incorporated data comes from heterogeneous populations, diverse treatment centers, and a variety of research designs used for investigations. Moreover, the time period in which the case reports/ studies were undertaken also varied. We included case reports and some retrospective studies [48][49][50][51][52][53][54][55][56]. Retrospective studies are prone to selection bias, recall bias, or misclassification bias and are subject to confounding [85]. Most of these studies mainly constitute class III level evidence, except for two prospective studies [48][49][50][51][52][53][54][55][56]. The types of radiation also differed from patient to patient. Moreover, pathology reports are used as standard for the diagnosis of RN; however, these studies mostly used imaging studies for RN diagnosis [48][49][50][51][52][53][54][55][56]. Some of the studies reported global adverse events/recurrence rates without differentiating between tumor types; however, they also contained participants other than BM patients [50,53,54]. Nonetheless, we presented the recurrence rates in results and side effects in the Discussion section to construct a better recurrence rate/ adverse event profile for the readers. The follow-up for different studies also varied. The likelihood of only BVresponding patients being included in the study may also be prone to publication bias.

Conclusions
According to our results, bevacizumab can be considered safe and efficacious for BM patients diagnosed with RN. However, the level of evidence presented was low, making our bevacizumab efficacy results inconclusive. Furthermore, several dimensions of BV treatment for RN were less clarified and should be investigated in future trials. These include the diagnosis standard used for RN, impact of type/dose/fractionation of radiation therapy used on RN, patterns, and underlying mechanism of recurrence. The pending results of a phase II trial (NCT02490878) of BV plus corticosteroids versus corticosteroids plus placebo for radiation necrosis after radiosurgery for brain metastases will further define the role of bevacizumab in the management of radiation necrosis.