Efficacy of BRAF Inhibitors in Combination With Stereotactic Radiosurgery for the Treatment of Melanoma Brain Metastases: A Systematic Review and Meta-Analysis

Background BRAF inhibitors have improved the outcome for patients with BRAF mutant metastatic melanoma and have shown intracranial responses in melanoma brain metastases. Stereotactic radiosurgery (SRS) is being used as a local treatment for melanoma brain metastasis (MBM) with better local control and survival. We searched for studies comparing the combination of two treatments with SRS alone to detect any clinical evidence of synergism. Materials and Methods PubMed, EMBASE, Medline, and Cochrane library were searched until May 2020 for studies with desired comparative outcomes. Outcomes of interest that were obtained for meta-analysis included survival as the primary, and local control as the secondary outcome. Results A total of eight studies involving 976 patients with MBM were selected. Survival was significantly improved for patients receiving BRAF inhibitor plus SRS in comparison to SRS alone as assessed from the time of SRS induction (SRS survival: hazard ratio [HR] 0.67 [0.58–0.79], p <0.00001), from the time of brain metastasis diagnosis (BM survival: HR 0.65 [0.54, 0.78], p < 0.00001), or from the time of primary diagnosis (PD survival: HR 0.74 [0.57–0.95], p = 0.02). Dual therapy was also associated with improved local control, indicating an additive effect of the two treatments (HR 0.53 [0.31–0.93], p=0.03). Intracranial hemorrhage was higher in patients receiving BRAF inhibitors plus SRS than in those receiving SRS alone (OR, 3.16 [1.43–6.96], p = 0.004). Conclusions BRAF inhibitors in conjunction with SRS as local treatment appear to be efficacious. Local brain control and survival improved in patients with MBM receiving dual therapy. Safety assessment would need to be elucidated further as the incidence of intracranial hemorrhage was increased.

Melanoma is the third most common cancer type (10%), after lung (50%) and breast cancers (20%), that spreads to the brain. Furthermore, patients with melanoma are at the highest risk of developing brain metastases (10%-44%) (17,18). Risk is increased to 75% in patients with metastatic melanoma (19). Autopsy series has revealed 80% central nervous system (CNS) involvement in patients with metastatic melanoma (19). Prognosis is poor for patients with melanoma brain metastases (MBM), and brain metastasis is the main contributor to mortality (up to 94.5%) in these patients (20). Management with surgery, chemotherapy, whole brain radiotherapy (WBRT), stereotactic radiosurgery (SRS), or their combinations has displayed a median survival of 3.8-7.69 months (19)(20)(21). MBM has been termed radioresistant. WBRT alone has been associated with limited local control and reduced median survival ranging from 2.86 to 3.86 months (19,21,22). SRS alone has shown better efficacy, with median survival times between 5.3 and 10.5 months, possibly due to better local control reported to be between 73% and 90% (23)(24)(25)(26)(27)(28)(29)(30)(31). The addition of WBRT to SRS has been inconclusive in this group of patients (29)(30)(31). As a result, a surge in SRS use was observed in patients with MBM from 2010 to 2015 due to radioresistance and late neurotoxicity associated with WBRT (32).
Outcomes for metastatic melanoma have improved impressively with targeted therapy and immunotherapy or their combinations (5)(6)(7). Studies have reported intracranial responses with targeted agents; however, their efficacy in melanoma patients with brain metastases has not been well established (33)(34)(35)(36). Vemurafenib's access to the brain was shown to be limited in preclinical studies involving ABCB1 and ABCG2 efflux pumps. Moreover, its hydrophobic and hydroscopic structure also suggests limited brain distribution (37)(38)(39)(40)(41). Nonetheless, vemurafenib has not only exhibited a protective effect against brain metastatic spread, but has also shown intracranial responses in several case reports, retrospective, and trial studies (33,34,(42)(43)(44)(45)(46)(47). Results of phase I/II trials have revealed that dabrafenib alone induced intracranial responses in melanoma patients with BRAF (V600E/G/L) mutations (35,48). Dabrafenib and trametinib (MEK inhibitor) combinations have demonstrated intracranial activity in BRAF (V600E/D/K/R)-mutated MBM patients with or without local therapy induction (36,49). Other BRAF inhibitor and MEK inhibitor combinations have also shown safety and intracranial activity, such as vemurafenib/cobimetinib, vemurafenib/trametinib, and encorafenib/binimetinib (50). Several studies have revealed the combination of SRS and BRAF/MEK inhibition to be safe and efficacious (51)(52)(53)(54). However, whether the addition of BRAF/MEK inhibitors to SRS is synergistic and better than SRS alone has yet to be determined. Several retrospective studies have revealed conflicting outcomes regarding the synergistic efficacy of BRAF/ MEK inhibitors plus SRS in the management of MBM (55)(56)(57)(58)(59)(60)(61)(62). Here, we attempt to address this issue by systematically reviewing the literature and performing meta-analysis of the outcomes for a better clinical perspective.

MATERIALS AND METHODS
Guidelines were followed according to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) (63). The protocol of this study is registered on PROSPERO: CRD42020185984.

Outcomes of Interest
The outcome of prime interest was overall survival (OS), whereas outcomes of secondary interest were brain control (local and distant control) and safety outcomes, including adverse events, intracranial hemorrhage, and radiation necrosis.

Search Strategy
Databases PubMed, EMBASE, Medline, and Cochrane library were searched until May 10, 2020. Various search terms relevant to the inclusion criteria were employed with language restricted to English. No restrictions were applied to the study design. Furthermore, relevant studies' references were examined for additional studies.

Study Selection
Endnote X9 Software was used to import the studies obtained from the databases. The studies were then organized and screened for duplication. After removal of duplicates, further screening for title and abstract was carried out. Eligible studies were scrutinized with full text reading. Two independent reviewers finally selected eligible studies for inclusion. The third reviewer was consulted in case of any disagreements.

Data Extraction
Data extraction was carried out using the modified form of "The Cochrane Collaboration Data Collection form-RCTs and non-RCTs" Extracted data included studies' attributes, design, first author, time period, publication year, number of participants, number of treated lesions, treatment regimens, and main efficacy and safety outcomes for the overall study group. Characteristics of the patients included age, sex, performance status (KPS), number of brain metastases, recursive partitioning analysis classes, diagnosis-specific graded prognostic assessment, and graded prognostic assessment class, if available. Furthermore, outcomes of interest (survival, brain local control, and safety) for treatment differences were extracted.

Assessment of Risk for Bias
Quality assessment was carried out using the modified checklist of Downs and Black aimed at assessing the methodological quality of non-randomized interventional studies (64). The checklist mainly covers four aspects of quality assessment: reporting, external validity, internal validity (bias and confounding), and statistical power. Twenty-seven questions are outlined, each carrying a score of one point, except for one question in the reporting section. Each section comprises of a different number of questions as follows: 10 questions in reporting, three questions in external validity, 13 questions in internal validity, and 1 question in statistical power. In this modified version, the statistical power question was also assigned a single point as opposed to the original, in which it carries five points. The modified version was used mainly for simplification and ambiguity avoidance (65). A grade was assigned according to the score obtained by each study as follows: excellent, if the score was between 24 and 28 points; good: 19-23 points; fair: 14-18 points, and poor: <14 points.

Measurement of Treatment Effect and Data Synthesis
Hazard ratios for the treatment effect (survival and local control) were extracted directly from papers if given. When hazard ratios were not published, they were extracted from the Kaplan-Meier curves using the Digital Equalizer and methods for incorporating summary time-to-event data into meta-analysis according to Tierney et al. (66). A similar approach was also applied for local control data. The acquired hazard ratios were pooled using the software "RevMan 5.3 software" (67, 68). An inverse variance statistical method was applied for pooling hazard ratios using the fixed effects analysis model. The significance level (P value) was set at <0.05. Heterogeneity was assessed using Chi 2 test and I 2 value. I 2 values of 25%, 50%, and 75% were considered low, moderate, and high, respectively (69). A random effects analysis model was used in case of moderate heterogeneity (50%).

RESULTS
Overall, eight retrospective studies, that met the inclusion criteria, were identified after a comprehensive research and selection process ( Figure 1) (55)(56)(57)(58)(59)(60)(61)(62). A total of 976 MBM patients had either received SRS alone (n = 728) or BRAF inhibitors plus SRS (n = 244) for management of their brain disease. Majorly, vemurafenib and minorly dabrafenib were used as the main choice of BRAF inhibitor. One study also used an MEK inhibitor in addition to a BRAF inhibitor (59). SRS was used as the main local brain radiation therapy, except in one study in which upfront WBRT and surgery were also applied. However, only survival outcomes were obtained regarding these participants. Other outcomes, such as local control, distant control, and side effects were obtained from patients only in the SRS recipients' subgroup (61). Two studies also included cohorts receiving immunotherapy; hence, specific comparative baseline characteristics were not available (58,59). The general characteristics of the studies are reported in Table 1.

Baseline Characteristics of Patients
There were some significant differences in patient characteristics between the two groups. BRAF mutant patients were younger than the patients with wild-type BRAF. These differences were reported in four studies (55,57,61,61). We used the data from four studies to perform meta-analysis of the age differences for BRAF mutant versus BRAF wild type as well as for BRAF inhibitor users versus non-users (57,60,61). Patients in the "BRAF mutant" and "BRAF inhibitor users" cohorts were comparatively younger (Figures S1, S4) ( Table 2). Mastorakose et al. (62), as well as Kotecha et al. (61) reported that patients with BRAF mutations were diagnosed with primary melanoma at a relatively younger age, 49 vs. 61 years and 60 vs. 64 year, respectively. This difference was maintained at the consequent BM diagnosis (58 vs. 66) (p<0.01) (61). Therefore, it could be speculated that BRAF mutations may expedite the process of oncological onset.
Similarly, male sex was also identified as the predominant sex in the BRAF wild-type cohorts in two studies (57,62). We performed a meta-analysis of six studies and the result revealed a significant predominance of male sex in BRAF wild type, but not amongst BRAF inhibitor non-users ( Figures  However, an open label, single arm, phase 2 trial of vemurafenib found no differences between the cohorts that were separated by the status of previous therapy for intracranial responses, progressionfree survival, and median survival (34). Similar results were also revealed for dabrafenib in a separate phase 2 trial (35). Kotecha et al. had allowed patients to receive upfront surgery, SRS, and WBRT. SRS was predominantly administered to patients receiving BRAF inhibitors. However, only survival outcomes were observed in this population. Other outcomes, such as local control and distant brain control, were extracted from the subgroup analysis of SRS recipients (n=119). BRAF inhibitor-receiving patients had better KPS scores than those receiving SRS alone (61). No significant differences in the patients' baseline characteristics between the groups were reported other than those mentioned above. The baseline characteristics and main outcomes of the studies are outlined in Table 1.
Subgroup analysis was performed to compare MBM patients receiving BRAF inhibitors with BRAF-mutant patients not receiving BRAF inhibitor and BRAF wild-type alone (57,58,60,62). Patients receiving BRAF inhibitors in addition to SRS were at a significant advantage in each comparison: BRAF mutant (HR 0. 66

Primary Diagnosis Survival
Meta-analysis of the survival difference between the two comparative treatments from the time of primary melanoma diagnosis revealed a significant overall survival advantage for BRAF inhibitor receivers (HR 0.74 [0.57, 0.95], p=0.02) ( Figure  4). The meta-analysis was based on the results of two studies providing four comparative outcomes (60,62

Progression Free Survival
Only one study assessed brain progression-free survival for the treatment difference, revealing longer brain progression-free survival (BPFS) for patients receiving a combination of the two treatments (p = 0.042) (62).

Safety Profile
The safety of the BRAF inhibitor-SRS combination was evaluated in several studies using various factors. These included, the rate of adverse radiation effects, adverse events, intracranial hemorrhage, and symptomatic and asymptomatic radiation necrosis.

Adverse Events
Several adverse events have been reported with BRAF inhibitors. Two studies reported the adverse events caused by vemurafenib and dabrafenib separately (60,62). We have outlined the events in Table 3

Radiation Necrosis
Considering the number of patients who developed radiation necrosis (RN), there was no difference between the treatments based on two studies (56,60

Publication Bias
Publication bias was assessed using a funnel plot for overall survival. All results were within the 95% CI indicating no evidence of publication bias in the SRS, BM, and primary diagnosis (PD) survival outcomes (Figures S8-S10).

DISCUSSION
Brain metastases are common in metastatic melanoma and are associated with poor prognosis (18)(19)(20)(21). Approximately 20% of patients with metastatic melanoma have brain metastasis at the time of diagnosis; over 50% develop these at some point during the course of the disease (18,19,70). Management of MBM includes surgery, SRS, WBRT, and cytotoxic chemotherapy (18). The addition of targeted agents and immunotherapy has improved the outcome significantly (5-7). SRS has been increasingly used as the treatment of choice for local therapy (32). In fact, the combination of radiation therapy and immune checkpoint blockers, such as ipilimumab and nivolumab has shown synergistic responses in various retrospective studies (53,58,59,(71)(72)(73)(74). BRAF inhibitors have also shown intracranial responses, suggesting that the two treatments could work synergistically (33)(34)(35)(36). Preclinical evidence suggests that the MAPK pathway, the pathway targeted by BRAF or MEK inhibitors, is activated following ionizing radiation, resulting in cell proliferation, differentiation, and survival. In vivo and in vitro inhibition of the MAPK signaling pathway was able to reverse these ionizing radiation effects (75,76). Ex vivo analysis of chromosomal breaks in patients treated with radiation plus BRAF inhibition showed increased radiosensitivity in patients treated with vemurafenib (P = 0.004) and vemurafenib switched to dabrafenib (P = 0.002). Dabrafenib was not shown to increase radiosensitivity in this study (77). The occurrence of skin toxicity (dermatitis) on previously irradiated skin in patients receiving vemurafenib was also suggestive of vemurafenib being a radiosensitizer (78). Thus, preclinical and clinical evidence suggests that combining the two treatments could lead to synergistic responses, thereby improving the survival outcome. Based on the evidence from eight studies, our results indicate that patients receiving SRS plus BRAF inhibitors had significantly better survival benefits. In a retrospective study, patients receiving BRAF inhibitors along with SRS also showed a similar surge in survival for MBM patients (18 months vs. 5 months, p = 0.009) (79). However, the patients had also used anti-CTLA-4 monoclonal antibodies, and the proportion of each drug was not specified, for which this study was excluded from our analysis. Only one study failed to report any survival advantage (55). The ICH rates for BRAF mutant and BRAF wild-type before treatment were compared (19/127 vs. 8/50, p=0.86). However, in a comparison of BRAF-mutant with BRAF inhibitors and non-BRAF inhibitor users, the rate was not specified. Evidently, hemorrhagic MBMs are associated with lower local control after SRS, and may lead to decreased survival for such patients (80). Fifteen of the 20 deaths attributed to CNS etiology were from ICH, indicating the ICH impact on survival analysis (55). In subgroup analysis, we observed a trend for patients with BRAF mutations receiving BRAF inhibitors to achieve far better survival than BRAF wild type and BRAF mutant without BRAF inhibitor use. The trend was maintained regardless of whether the survival assessment was from SRS or BM diagnosis. This is in contrast to studies associating worst survival with BRAF mutation status before the era of BRAF inhibitors (32,81,82). One reason could be the small number of patients in the comparative groups for BRAF mutants with/without BRAF inhibitors. In addition, current studies have been undertaken during the era of immune checkpoint inhibitors as immunotherapy, and these operate synergistically with SRS (53,58,59,(71)(72)(73)(74). Therefore, patients without BRAF inhibitors may have opted for such therapies, thereby improving the outcome (56,60,62). It has also been pointed out that BRAF mutation was associated with improved local control in patients with MBM, which may imply higher radiosensitivity, thereby eliciting better response than patients with BRAF wild-type (83). On the other hand, our finding is consistent with that of Menzies et al., who also revealed significant differences in 1-year OS rates for patients with BRAF-mutant BRAF inhibitor use (83%) compared to BRAF wildtype (37%), and BRAF mutant without BRAF inhibitor use (29%) (p<0.001) (84). Another important observation was the effect of induction timing of BRAF inhibitors with respect to SRS or BM development. Patients receiving BRAF inhibitors concurrently or after SRS were shown to have superior survival than patients receiving it before SRS. Similar observations were made in studies involving patients with renal cell carcinoma (RCC) and BM. In this study, patients receiving tyrosine kinase inhibitors (TKIs) after BM development had a significantly better survival advantage compared to patients developing BM while they were on TKIs (23.6 months vs. 2.08 months, p=0.0001) (85). This could reflect the higher sensitivity of patients to BRAF inhibitors receiving it for the first time. Added advantage could also come from the systemic disease control of these patients as well as better brain control.
Improved local control was also revealed based on data from four studies. Improvement in local control demonstrates that the survival benefit may be a result of synergism between the two treatments. Even though BRAF inhibitors have been shown to have limited brain penetration, the fact that SRS may focally disrupt the blood brain barrier by targeting the vasculature could possibly pave the way for targeted agents to reach the tumor (86,87). It is also hypothesized that targeting driver mutations with high specificity may lower the concentrations required for radiosensitization (83). Most of the studies revealed no difference in distant failure between the treatments. It could be hypothesized that SRS could only disrupt the blood brain barrier locally, thus leaving the distant control unaffected (86,87). Furthermore, acquired resistance to BRAF inhibitors could also lead to distant failure (88). Moreover, a significant delay in distant failure reported in one study may be a manifestation of a better response from the fact that BRAF inhibitors used after SRS were reported to have improved survival outcomes compared to concurrent use or use before SRS (57,61,62).
From a safety perspective, BRAF inhibitors have been associated with skin toxicity, ICH, and RN (33-36, 55, 56). Only one study elaborated the adverse events excluding RN in a comparative manner (60). No difference was revealed in that study between the treatments. Instead, a slower rate of adverse radiation effects was observed in patients receiving BRAF inhibitors. Intracranial hemorrhage was significantly increased in patients receiving BRAF inhibitor (56,60,62). Additionally, freedom from ICH was also reduced in the BRAF inhibitor cohort (55). MBMs are prone to intra-tumoral hemorrhages. Up to 50% of MBM become hemorrhagic (89)(90)(91). ICH rates of 0.9 to15.2% have been associated with SRS treatment as well (78,92). Hemorrhagic MBMs treated with SRS were found to be susceptible to local failure; hence, it may have an impact on the survival outcome (78). Surgery may be preferred in such patients, as surgery was shown to lessen the recurrence (89). BRAF inhibitors, both vemurafenib and dabrafenib, used alone have also been shown to cause intracranial hemorrhage (6%-7%) (33)(34)(35). Therefore, the combination of the two therapies may increase the odds of ICH in MBM patients. Ly et al. reported in their study that 15 of the 20 deaths attributed to CNS etiology were associated with ICH, suggesting that ICH rates may also have an impact on survival (55). Overall, these cases were managed with dose reduction, interruption, and in few cases, withdrawal.
There was no difference between the treatments in causing RN. BM patients receiving SRS treatment are at risk for RN (93,94). As both treatment groups had received SRS, it appears that the addition of BRAF inhibitors may not be associated with an increased risk of RN in these patients. Phase 2 trials using vemurafenib and dabrafenib alone without local treatment also did not show any evidence of RN (33)(34)(35). Patel et al., however, showed an increased 1-year cumulative incidence of SRN (symptomatic RN) in patients on BRAF inhibitors. Nonetheless, in this study, an extremely low number of patients were in the BRAF inhibitor plus SRS cohort (n=15) compared to SRS alone (n=72). As the 1-year cumulative incidence of RN with SRS is 5% to 10%, it may not have enough statistical power to detect such an association (95,96). In contrast, the 1-year cumulative incidence was significantly lower in the study by Kotecha et al. (61). Kim (52). In short, RN may only be associated with SRS. To confirm whether BRAF inhibitor may have a role in increasing the rate of RN, a larger study comprising such comparative groups should be undertaken.
Our study is limited by the fact that the included studies were retrospective in design. Retrospective studies are subject to confounders and tend to have selection bias, recall bias, and misclassification bias (98). Period coverage was longer for all the studies. In addition, a few studies had a small number of patients in comparative cohorts. Heterogeneity was observed in the local control outcome, and the random effects model was used for analysis.

FUTURE PERSPECTIVE
Management of MBM is ever expanding with the addition of several BRAF and MEK targeting agents as well as the success of immune checkpoint blockade agents (5)(6)(7). SRS is becoming a predominant local therapy, and its combination with immune checkpoint blockade agents, such as ipilimumab and nivolumab have been assessed in several retrospective studies revealing an improved outcome for patients with MBM (53,58,59,(71)(72)(73)(74). However, further class I evidence is needed to establish clinical guidelines. Likewise, BRAF and MEK inhibitors (alone or in combination) with SRS show promise based on the results of these retrospective studies, but further class I evidence is required (55)(56)(57)(58)(59)(60)(61)(62). Furthermore, the efficacy of targeted agents may be further enhanced by increasing the bioavailability of these drugs in the brain. The bioavailability of several anti-cancer targeted agents, including vemurafenib and dabrafenib, have been shown to be restricted by two members of the ATP-binding cassette (ABC) family of transporters, namely P-glycoprotein (P-gp; ABCB1) and breast cancer resistance protein (BCRP; ABCG2) (37)(38)(39)(40)(41)(99)(100)(101)(102). In fact, co-administration of elacridar, an ABCB1 and ABCG2 blocker, was demonstrated to improve the therapeutic efficacy of vemurafenib, especially for brain metastases located behind a functional blood-brain barrier (39). It is another area that could further enhance the effectiveness of these drugs in the brain with an improved outcome for MBM patients.

CONCLUSIONS
Our results suggest a survival benefit for patients with MBM receiving BRAF inhibitors in conjunction with SRS as local treatment in comparison to SRS alone. Patients receiving BRAF inhibitors after SRS may have a greater survival advantage. Improvement in local control for SRS plus BRAF inhibitors may suggest that the survival surge is a result of synergism between the two treatments. BRAF inhibitors in combination with SRS may increase the risk of intracranial hemorrhage in MBM patients and warrants further investigation. Other side effects were mild in nature. Our results provide a basis for a larger randomized controlled trial to be undertaken in order to establish class I evidence.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.