Two single-arm studies evaluated the use of bevacizumab and irinotecan in 67 patients with recurrent high-grade glioma and compared the outcomes to historic controls. Radiographic response was observed in 60% of study subjects, with 6-month progression-free survival (PFS-6) of 38–46% [34, 35]. These results were in contrast to other contemporary salvage regimens that demonstrated a radiographic response of 5–10% and PFS-6 of 9–25% [36].

Two subsequent prospective phase 2 studies were conducted which led to accelerated FDA approval of bevacizumab as monotherapy for recurrent glioblastoma in 2009 [6]. In the phase 2 BRAIN study, patients were randomized to bevacizumab or bevacizumab plus irinotecan. The overall response rates (ORR) were 28.2 and 37.8%, respectively, with PFS-6 of 42.6 and 50.3%, respectively [37]. Median overall survival (OS) was 9.2 months versus 8.7 months. Older radiographic response criteria were used to assess radiographic response, and the study was not a superiority trial and allowed for crossover from single-agent bevacizumab to combination therapy. In a single-arm study of bevacizumab in 48 patients, the ORR and PFS-6 were 35 and 29%, respectively, with a median OS of 7.75 months [38]. While the FDA approved the use of bevacizumab in recurrent GBM based on these trials, the European Medicines Agency declined approval due to the lack of a non-bevacizumab control arm, modest improvement in OS, and challenges with radiographic response assessment [39]. This has led to a difference in standard of care for recurrent glioblastoma in the U.S. versus Europe.

Subsequent phase 2 trials have evaluated bevacizumab in various combinations with irinotecan, cetuximab, carboplatin, etoposide, fotemustine, sorafenib, temozolomide, erlotinib, panobinostat, and temsirolimus [37, 40–59]. There have also been trials evaluating bevacizumab and re-irradiation [60–62].

The only trial to show a survival benefit of combination therapy over bevacizumab alone was the BELOB study, a randomized phase 2 study of 148 patients with recurrent glioblastoma randomized to lomustine, bevacizumab, or both. Combination therapy resulted in a PFS-6 of 41% compared with 11 and 18% with OS at 9 months of 59% compared to 43 and 38% for lomustine and bevacizumab alone, respectively [63]. Based on these results, a phase 3 study of lomustine versus lomustine plus bevacizumab in patients with recurrent glioblastoma (EORTC 26101) was conducted, demonstrating no difference in OS with a median of 9.1 months for combination therapy versus 8.6 for lomustine alone [64]. However, there was a benefit in PFS of 4.2 months for combination therapy compared to 1.5 months for lomustine monotherapy. 35.5% of patients in the control arm of this study did cross over to receive bevacizumab.

Resistance to bevacizumab inevitably develops with resulting rapid clinical deterioration. Retrospective data has suggested that continuing bevacizumab beyond initial progression may modestly improve outcome [65]. The ongoing phase 3 TAMIGA trial (NCT01860638) aims to evaluate whether adding bevacizumab to lomustine as second-line therapy followed by standard of care for third line therapy with bevacizumab improves survival compared to lomustine alone followed by standard of care third line therapy with placebo.

In summary, clinical data to date offer only limited support for combining bevacizumab with chemotherapy in the setting of recurrent glioblastoma.

Several single-arm phase 2 studies of bevacizumab in combination with temozolomide and radiation showed near doubling of median PFS to 13–14 months compared to historic benchmarks. However, only a modest improvement in median OS to 19–21 months was observed in these studies [66–68]. The GLARIUS study was a phase 2 trial that compared the combination of bevacizumab and radiotherapy with either irinotecan or temozolomide in newly diagnosed glioblastoma patients whose tumors expressed the DNA repair enzyme O6-methyl guanine DNA methyltransferase (MGMT). Loss of MGMT function through methylation of the gene promotor in GBM has been shown to confer increased sensitivity to therapy with the DNA alkylating agent temozolomide [69]. The GLARIUS study demonstrated a significant prolongation of PFS but no difference in OS in the bevacizumab containing arm [70]. PFS was 5.99 months in the control arm compared to 9.7 months in the bevacizumab/irinotecan arm, and median OS was 17.5 months in the control arm compared to 16.6 months in the bevacizumab/irinotecan arm. Neither therapy regimen was superior in delaying the time to deterioration in any of the pre-specified dimensions of quality of life.

Two randomized, placebo-controlled, phase 3 trials, AVAglio and RTOG 0825, investigated the addition of bevacizumab to the standard of care treatment regimen consisting of surgery and chemoradiation with temozolomide in patients with newly diagnosed glioblastoma. Both studies failed to show an improvement in OS.

In the AVAglio study, newly diagnosed glioblastoma patients were randomized to bevacizumab versus placebo in combination with standard chemoradiation. The median PFS for standard therapy plus bevacizumab was 10.6 months versus 6.2 months for standard therapy with placebo [71]. The predefined OS endpoint, however, was not met with OS of 16.8 months in the bevacizumab arm compared to 16.7 months in the placebo arm. The RTOG 0825 trial also compared bevacizumab to placebo in combination with standard therapy and demonstrated an improved PFS of 10.7 months versus 7.3 months with placebo [72]. This increase in PFS did not meet the predefined significance level of P = 0.004. Similar to the AVAglio trial, there was no difference in OS with a median survival of 15.7 months in the bevacizumab arm compared to 16.1 months in the placebo arm. In both studies, approximately 30–50% of controls crossed over and received bevacizumab at progression, potentially confounding the true impact on OS. While AVAglio used the revised RANO criteria to assess disease progression, RTOG 0825 used the traditional Macdonald criteria, which only evaluates enhancing disease [73, 74]. The differences in the radiographic assessments used in the AVAglio and RTOG0825 trials are laid out in detail in a publication by Chinot et al. [75].

Both trials also attempted to assess other measures of net clinical benefit, including Karnofsky performance status, corticosteroid requirement, and quality of life measures. Interestingly, the European-led AVAglio and the US-led RTOG 0825 studies showed conflicting quality of life outcomes. In the AVAglio trial, bevacizumab prolonged maintenance of Karnofsky performance status and decreased steroid utilization. Moreover, time to deterioration was prolonged in 5 pre-specified domains: global health status, physical and social functioning, motor dysfunction, and communication deficit. In contrast, the RTOG 0825 study found that bevacizumab consistently led to decreased objective and perceived cognitive function (as assessed by formal neurocognitive testing), as well as motor dysfunction and communication deficits compared to controls. The cause of the differences in quality of life outcomes is unclear, but possible reasons include different radiographic response criteria used, substantial dropout among RTOG participants, and different methods of statistical modeling.

In addition to the VEGF neutralizing antibody bevacizumab, other inhibitors of the VEGF pathway as well as inhibitors of other angiogenic growth factors have also failed to show an overall survival benefit in glioblastoma. Aflibercept, a recombinant fusion protein that binds VEGF and placental growth factor (PIGF), improved survival in preclinical studies of glioblastoma but failed to meet its primary endpoint of PFS-6 in a single-arm phase 2 study of patients with recurrent glioblastoma [76, 77]. Receptor tyrosine kinase inhibitors (TKIs) that inhibit the VEGF and other angiogenic pathways have also been evaluated in glioblastoma, including cediranib, sunitinib, pazopanib, vandetanib, and sorafenib [78–83]. The only agent to reach phase 3 of clinical development was cediranib. A phase 2 study evaluating single-agent cediranib in patients with recurrent high-grade glioma showed a 27% radiographic response rate with a 6 month PFS of 26% [78]. However, a randomized, placebo-controlled, phase 3 trial of cediranib monotherapy and cediranib in combination with lomustine compared to lomustine alone failed to reach its primary endpoint of PFS prolongation [79].

Other approaches have included VEGF receptor blockade and targeting upstream pathways that lead to increased VEGF expression. A phase 2 trial of CT-322, a pegylated protein that binds and blocks VEGFR2, was terminated early due to insufficient efficacy [84]. An open label phase 2 study of the anti-VEGFR2 monoclonal antibody ramucirumab versus the anti-PDGFR monoclonal antibody IMC-3G3 has completed accrual with results pending. Enzastaurin is an oral serine/threonine kinase inhibitor that targets the proangiogenic protein kinase C and PI3K/AKT pathways. A randomized, phase 3 trial of enzastaurin versus lomustine in recurrent glioblastoma showed no difference in PFS and OS [85].

Despite promising preclinical data with antiangiogenic therapy in GBM animal models and increased PFS in patients, VEGF pathway inhibition has yet to demonstrate a survival benefit in GBM patients. Given the multiple pathways involved in tumor angiogenesis, it is perhaps not surprising that VEGF pathway inhibition alone does not durably or completely block tumor angiogenesis. There are many proposed adaptive mechanisms of resistance to anti-VEGF therapy. First, the hypoxic tumor microenvironment may trigger the release of alternative proangiogenic factors such as HGF, FGF, ANG-2, SDF1α, and interleukin-8 [29, 86–89]. Preclinical studies in GBM animal models demonstrate that dual targeting of VEGF and the alternative proangiogenic factor Ang-2 may overcome resistance to anti-VEGF monotherapy [24, 90, 91]. Second, antiangiogenic therapy may foster other modes of vessel recruitment, such as vessel co-option, intussusception, vascular mimicry, recruitment of endothelial progenitor cells, and differentiation of cancer stem-like cells into endothelial cells [92–94]. The process of vessel co-option, whereby tumors utilize native brain vessels to recruit blood supply, is under increasing investigation as an escape mechanism to antiangiogenic therapy. The molecular mechanisms of vessel co-option are still poorly understood and will yield novel approaches once the pathways involved have been identified. Third, there may be inherent insensitivity to VEGF inhibition among different tumor blood vessel subtypes with decreased anti-VEGF sensitivity in pericyte-covered tumor vessels [23, 95]. Furthermore, recruitment of bone marrow-derived cells such as monocytes and M2-skewed macrophages may rescue tumor angiogenesis through production of pro-angiogenic factors [9, 27, 96]. Lastly, studies in animal models have also shown that VEGF inhibition may induce transformation from a proneural to a more invasive mesenchymal phenotype [97, 98].

Retrospective data in high grade gliomas and prospective data in other cancers have suggested potential benefit of continuing antiangiogenic therapy past progression, suggesting that resistance is potentially epigenetically based and reversible [65, 99–101]. This becomes an important consideration in clinical trials assessing subsequent-line therapies.

Unlike other targeted therapies, there are currently no established biomarkers that can be utilized to select patients who are more likely to respond to antiangiogenic agents. Biomarkers that have been assessed as possible predictors of increased efficacy include a 10-gene panel identified by RTOG 0825 [102], proneural transcriptional glioblastoma subtype [103], VEGF expression [104], epidermal growth factor receptor (EGFR), platelet-derived growth factor receptor a (PDGFR-a), and c-KIT [105]. However, none of these markers have been validated and none are intended for clinical practice. Other candidate biomarkers include circulating cytokines such as VEGF and sVEGFR2 [71, 106], SDH1α [107], and matrix metalloproteinases [27, 108].

Efforts have also been directed at accurately defining tumor response and progression as well as using novel imaging techniques to predict response. Unlike the Macdonald criteria, the RANO criteria account for the possible effect of antiangiogenic treatment on reducing tumor enhancement when determining disease progression [73]. Antiangiogenic therapy decreases vessel permeability, leading to a usually transient phenomenon of decreased enhancement known as “pseudoresponse”. Possible imaging markers to predict tumor response include apparent diffusion coefficient [109], restriction spectrum imaging [110], dynamic contrast enhanced (DCE) and dynamic susceptibility-contrast (DSC) techniques [111, 112], vessel architectural imaging [113], and dopamine and positron emission tomography [114, 115]. Cerebral blood flow has also been investigated, with increased tumor perfusion correlating with an increase in overall survival in newly diagnosed and recurrent glioblastoma patients treated with cediranib, a pan-VEGF kinase inhibitor [105, 116].

Responses to immunotherapy across a spectrum of cancers has led to interest in this strategy in glioblastomas despite the historical view that the central nervous system is immune-privileged due to the blood brain barrier. Preclinical data in extracranial tumors have suggested that antiangiogenic therapies increase tumor delivery of activated T cells, making the tumor more susceptible to immune attack [20]. Moreover, vascular normalization may promote an “immunosupportive tumor microenvironment” [18, 24, 90], thereby enhancing the effects of immunotherapy. There are a number of clinical trials evaluating the use of antiangiogenic therapy in combination with immune checkpoint inhibitors (pembrolizumab, MEDI4736), immune stimulants (Plerixafor), and vaccines (SL-701, rindopepimut, heat shock protein peptide complexes) in patients with glioblastoma. Preliminary results from a phase 2 trial of standard of care plus bevacizumab in combination with a dendritic cell vaccine showed improved OS in the combination group compared to the vaccine or bevacizumab alone [117]. Preliminary results from the phase 2 ReACT study of patients with EGFRvIII mutant recurrent glioblastomas found that the combination of the vaccine rindopepimut with bevacizumab prolonged median OS to 12 months versus 8.8 months for the control arm (bevacizumab plus keyhole limpet hemocyanin) [118]. PFS-6 was also significantly increased from 11% in the control arm to 27% in the experimental arm. Additionally, animal models have shown promising results in the use of antiangiogenic therapy in combination with adoptive cell transfer [119].

Multiple clinical trials of antiangiogenic agents in newly diagnosed and recurrent glioblastoma have failed to show an overall survival benefit. Given the existence of multiple, redundant, pro-angiogenic signal transduction pathways and the propensity of glioblastoma to develop resistance to therapeutics targeting one pathway, combination strategies are the logical next step in the development of anti-angiogenic agents. These may include combinations of multiple anti-angiogenic agents or the combination of anti-angiogenic drugs with other classes of therapeutics like immunotherapy. To better assess which patients are most likely to derive clinical benefit from antiangiogenic agents, further research on imaging and biologic markers is essential. Advances in genetically modified mouse models (GEMMs), patient-derived and stem-like cell models of glioma, and importantly, human tumor-bearing humanized mouse models will allow for more translatable preclinical studies. Additional studies are also needed to clarify the conflicting data on the benefits of anti-VEGF therapy on quality of life, an area of particular importance to patients with glioblastoma. Given retrospective data suggesting that treatment of patients with high-grade glioma with low doses of bevacizumab (5 mg/kg per week) may be superior to standard dosing, further work is needed to clarify the optimal dose and administration schedule [120]. Lastly, a better understanding of the mechanisms of resistance to antiangiogenic therapy will facilitate the development of more effective therapeutic targets and treatment strategies.