Treatment mechanism and research progress of bevacizumab for glioblastoma

Abstract

Hypervascularization is a notable pathological hallmark of glioblastoma (GBM). Bevacizumab (Bev) remains the sole antiangiogenic agent approved by the U.S. Food and Drug Administration (FDA) for GBM treatment. The approval for this indication was supported by several phase II studies demonstrating that Bev significantly improved progression-free survival and the best imaging response in patients with recurrent GBM. Three large phase III randomized controlled trials reported that Bev did not significantly extend overall survival (OS). Nevertheless, Bev has been shown to delay the deterioration of patients’ quality of life by postponing tumor progression. This review synthesizes findings from recent investigations exploring Bev in combination with targeted therapies, immunotherapy, or reirradiation. Additionally, this review discusses dosing regimens, administration, treatment failure patterns, third-line therapeutic applications, and prognostic markers of Bev. By synthesizing current evidence, this review aims to inform clinical decision-making for neuro-oncology clinicians.

Introduction

Glioblastoma (GBM) is characterized by elevated incidence and recurrence rates alongside substantial mortality. GBM accounts for 50.9% of adult malignant central nervous system tumors, exhibiting an increasing prevalence with a 6-month progression-free survival (PFS) rate of 15% and a 5-year survival rate of 6.7% [1]. Furthermore, GBM patients demonstrate severe deterioration in health-related quality of life (QoL), characterized by progressive neurocognitive decline attributed to advanced age (median age: 65), rapid disease progression, or toxicity from treatment drugs. These factors underscore critical unmet therapeutic demands for GBM.

Bevacizumab (Bev; Avastin^®), a monoclonal IgG1 antibody that targets vascular endothelial growth factor (VEGF), has opened a new chapter in GBM therapeutics. Building on efficacy observed in metastatic colorectal cancer, Stark-Vance et al. pioneered the application of Bev for glioma therapy in 2005 [2,3]. A study enrolling 21 patients with recurrent malignant glioma (10/21 with high-grade glioma) demonstrated a 42.9% objective response rate (ORR) with Bev-irinotecan (Iri) combination therapy [2]. Phase II clinical studies subsequently validated Bev’s efficacy as monotherapy for recurrent GBM (rGBM), yielding ORR and 6-month progression-free survival (PFS6) rates of 28.2%-57% and 29%-46%, respectively [4-6]. This evidence led to the U.S. FDA’s accelerated approval of Bev for rGBM in 2009 [7].

The dual objectives of GBM therapy are survival prolongation and preservation or enhancement of QoL. GBM management has transitioned from traditional surgical resection combined with the Stupp protocol to multimodal approaches incorporating tumor-treating fields, targeted agents, and immunotherapies. Bev-based clinical strategies for rGBM initially focused on monotherapy and chemotherapeutic combinations (e.g., alkylating agents, Iri, and carboplatin), whereas recent investigations explore its integration into frontline settings, including administration during or prior to concurrent chemoradiotherapy or as a presurgical intervention. Three pivotal phase III randomized controlled trials (RCTs) failed to establish significant overall survival (OS) benefits with Bev [8-10]. However, Bev may mitigate QoL deterioration through delayed tumor progression [5,11,12]. This review critically evaluates emerging combinatorial strategies integrating Bev with targeted therapies, immunotherapies, or reirradiation. Furthermore, this review discusses dosing regimens, administration routes, failure patterns, third-line applications, and prognostic markers of bev, which have great value in clinical practice (Figure 1).

Figure 1.

Graphical abstract of the review article. Hypoxia-induced dysregulation of angiogenic factors promotes pathological vascular remodeling and the formation of an immunosuppressive TME in GBM. Bev, a VEGF-A-targeting monoclonal antibody, is utilized in clinical practice as part of combination therapies to disrupt these pathways. This review further examines four critical dimensions of Bev in clinical practice. SCLGC, stem cell-like glioma cell; VEGF, vascular endothelial growth factor; bFGF, basic fibroblast growth factor; TGF-β, transforming growth factor-β; TME, tumor microenvironment; TSP-1, thrombin-sensitive protein-1; INF-alpha, interferon-alpha; PDGF, platelet-derived growth factor; EGF, epidermal growth factor; PlGF, placental growth factor; VEGFR, VEGF receptor. Created in BioRender. Yang, T. (2025) https://BioRender.com/avq3e42.

This synthesis of pivotal Bev-associated clinical trials in GBM seeks to inform evidence-based decision-making for neuro-oncology practitioners.

Mechanism

Hypervascularization activity, characterized by irregular vascular morphology, inefficient transport, and pericyte deficiency, represents a critical pathological hallmark of GBM [13]. Hypoxia-induced dysregulation of angiogenesis-related factors in the tumor microenvironment (TME) manifests as upregulated pro-angiogenic factors - including vascular endothelial growth factors [VEGFs], basic fibroblast growth factors [bFGFs], and transforming growth factor-β [TGF-β]) - coupled with diminished expression of anti-angiogenic factors such as thrombin-sensitive protein-1 (TSP-1) and interferon (INF)-alpha, thereby perpetuating the activation of the tumor angiogenesis switch [14].

VEGF plays a pivotal role in both physiological and pathophysiological processes: it maintains endothelial cell survival, proliferation, and migration; promotes glioma cell growth via autocrine signaling; enhances vascular permeability, also known as vascular permeability factor; and contributes to the immunosuppressive TME [15,16].

Bev exerts therapeutic effects by neutralizing VEGFs (primarily VEGF-A), thereby blocking its interaction with receptors (Flt-1 and KDR) and inhibiting downstream signaling [17]. Bev mainly targets genomically stable endothelial cells, indirectly suppressing tumor progression rather than directly killing tumor cells, which confers a lower propensity for drug resistance. Furthermore, Bev disrupts the proangiogenic effect of stem cell-like glioma cells (SCLGCs), inhibiting transplanted tumor growth in experimental models [18].

Both endothelial cells and glioma cells can express VEGF and its receptors. In GBM, VEGF expression is regulated by two primary pathways: hypoxia-dependent mechanisms mediated by hypoxia-inducible factor-1 (HIF-1) [19] and hypoxia-independent mechanisms involving epidermal growth factor receptor (EGFR) activation [20,21]. SCLGCs continuously secrete substantial VEGF in vitro, amplifying the proangiogenic function of endothelial cells - a process exacerbated by hypoxia [18,22]. Notably, VEGFR expression has been detected in GBM cells. Joensuu et al. reported that 39% of GBM patients exhibit amplification of VEGFR2 [23]. Quantitative analyses across 12 brain tumor types revealed significantly elevated VEGFR levels in GBM [24]. Preclinical studies confirm that VEGF-VEGFR pathway blockade suppresses glioma cell growth in immunodeficient rat models [25].

Progressive and recurrent glioblastoma (rGBM)

Bev monotherapy

Bev has been recommended as the preferred treatment for rGBM by the National Comprehensive Cancer Network (NCCN) guidelines in the U.S. [26]. Table 1 summarizes key clinical trial outcomes. The noncomparative phase II AVAREG trial evaluated Bev against fotemustine (FTM) in rGBM patients, randomizing participants at a 2:1 ratio (Bev vs. FTM), with the primary endpoint being the 6-month OS rate [12]. No significant differences were observed in the 6-month OS rate or median OS (mOS) between the two groups [12]. A similar conclusion was reached in the Japanese JO22506 study [27].

Table 1.

Bev alone or in combination with cytotoxic chemotherapy for the treatment of rGBM

Ref.	Study Design	No. of patients	Study arms	Bev Dose (mg/kg)	ORR (%)	mOS (months)	mPFS (months)	PFS6 (%)
Stark-Vance 2005 [2]	NA	GBM: 11	Bev + Iri	5	43	NA	NA	NA
		Other: 10
Vredenburgh 2007 [6]	II	35	Bev + Iri	10	57	9.2	4.2	46
			Bev + Iri	15
Bokstein 2008 [97]	NA	GBM: 17	Bev + Iri	5	47.3	7	4.2	25
	2005-2007	AO: 2
		AOA: 1
Kreisl 2009 NCI 06-C-0064E [5]	II, comparative	48	Bev→Bev + Iri	10	Levin: 71	7.75	4	29
	2006-2007				Macdonald: 35
Friedman 2009 BRAIN [4]	II, multicenter, open-label, randomized, noncomparative	167	Bev + Iri	10	28.2	8.7	5.6	NA
	2006-2007		Bev			9.2	4.2
Nghiemphu 2009 [11]	Retrospective	44	Bev	5	NA	9.01*	4.25*	41
	2005-2006	79	Other			6.11	1.82	18
Raizer 2010 [96]	II	GBM: 50	Bev	15	24.5	6.5	NA	25
Taal 2014 BELOB [36]	II, multicenter, randomized	47	LOM	10	5	8.0	1.0	13
	2009-2011	51	LOM90 + Bev		34	12.0	4.0	41
		47	Bev		38	8.0	3.0	16
Field 2015 CABARET [38]	II, multicenter, randomized, open-label	60	Bev + Carboplatin	10	14	6.9	3.5	15
	2010-2012	62	Bev		6	7.5	3.5	18
Brandes 2016 AVAREG [12]	II, Noncomparative, multicenter, randomized	91	Bev	10	29	7.3	3.38	19.6
	2011-2012		FTM		9	8.7	3.45	10.7
Wick 2017 EORTC 26101 [10]	III	288	LOM + Bev	10	13.9	9.1	4.2	28.4
	2011-2014	149	LOM		41.5	8.6	1.5	16.8
Gilbert 2017 RTOG 0625 [146]	II, randomized	60	Bev + dose-dense TMZ	10	19	9.4	4.7	39
	2007	57	Bev + Iri		28	7.7	4.1	38.6
Desjardins 2019 [147]	Retrospective, single-institution, real-world study	74	Bev	10	NA	11.1	6.4	24.3▼
	2009-2012
Cloughesy 2020 TOCA 5 [148]	II/III, open-label, multicenter, randomized	201	Toca 511 + Toca FC	10	NA	11.10	NA	NA
	rHGG	202	SOC (Bev/TMZ/LOM)			12.22
	2015-2018
Detti 2021 [149]	Retrospective, rHGG	92 (GBM: 71)	Bev	10/15	36.1	9	6.9	58
	2009-2019
Friedman 2023 [33]	II, multicenter, open-label, Noncomparative		Bev	10	28.2	9.2	4.2	42.6
			Bev + Iri		37.8	8.7	5.6	50.3
Smolenschi 2023 [150]	Retrospective, real-world study, single-institution	202	Bev	10/15	42	23.7	6.8	NA
	rGBM
	2006-2016
Witte 2024 [151]	Retrospective	134	HCQ + BEV + aRCT	NA	NA	23.92	NA	NA
	rGBM		BEV + aRCT			9.63*
	2006-2016
Lee 2024 [152]	Nationwide population-based study	450	Bev	NA	NA	22.60	NA	NA
	rGBM	396	Bev + Iri			20.44
	2008-2021

^▼PFS12;

AOA, Anaplastic oligoastrocytoma; AO, Anaplastic oligodendroglioma; PFS6, 6-month progression-free survival; PFS, progression-free survival; OS, overall survival; GBM, glioblastoma multiforme; q, every; TMZ, temozolomide; Carbo, carboplatin; LOM/CNNU, carmustine; FTM, fotemustine; NA, not available; SOC, standard of care; Toca 511, vocimagene amiretrorepvec; Toca FC, flucytosine; HCQ, chloroquine; aRCT, adjuvant-radiochemotherapy;

^* P < 0.05.

However, the use of Bev in rGBM remains contentious. Unlike the FDA, the European Medicines Agency has not approved Bev for rGBM treatment [28].

First, although Bev improves the ORR and median PFS (mPFS) in rGBM patients, it fails to extend OS [4-6]. This discrepancy suggests that surrogate endpoint improvements do not translate into an OS benefit. Ballman et al. reported a 90% concordance between PFS6 and 12-month OS in their analysis of phase II trials [29]. However, their analysis did not include clinical trials related to Bev, leaving its generalizability to Bev-treated cohorts uncertain.

Second, Bev complicates neuro-oncological response assessment. Traditional imaging assessments rely mainly on contrast-enhanced MRI based on the MacDonald criteria, but these can be influenced by various factors, such as radiation damage, MRI equipment parameters, and steroid use [30]. Bev’s ability to stabilize the blood-brain barrier (BBB) and reduce contrast leakage may interfere with comparisons of pre- and posttreatment enhanced MR images. As a result, a reduction in contrast enhancement may mask an actual increase in tumor burden, inducing “pseudoresponse”. To address this issue, the Response Assessment in Neuro-Oncology (RANO) criteria were developed, which incorporate T2-weighted imaging/fluid-attenuated inversion recovery (T2/FLAIR) to evaluate nonenhancing lesions for a more accurate and comprehensive evaluation of patient efficacy [31]. Additionally, the timing of assessments is crucial. Early MRI scans post-Bev initiation (e.g., within 24 hours) often show reduced enhancement unrelated to antitumor efficacy [5]. Interestingly, patients with “partial response” on day 4 or 28 scans exhibit prolonged PFS [5]. These patients may have greater sensitivity to Bev treatment.

Finally, several clinical trials have design flaws. The NCI 06-C-0064E study relied on historical controls for their interpretation [5], while the AVF3708g study employed a noncomparative design despite randomization [4].

In conclusion, despite the aforementioned controversies, two points are clear: First, owing to differing mechanisms of action, the two drugs exhibit different toxicity profiles, with Bev primarily causing adverse events such as bleeding, thrombosis, hypertension, and proteinuria, whereas alkylating agents (such as FTM, lomustine [LOM], and TMZ) mainly lead to hematological toxicity, such as thrombocytopenia and neutropenia [12,32]. A recent study showed that thrombocytopenia can lead to dose adjustments or discontinuation of CCNU, thereby decreasing survival in rGBM patients [32]. Second, Bev can significantly improve QoL and effectively decrease the use of corticosteroids by reducing BBB permeability and alleviating cerebral edema [5,11,12,33]. Therefore, in a context where treatment options are limited, Bev may provide an alternative treatment option for rGBM patients.

With chemotherapy (CT)

Theoretically, Bev combined with CT may exhibit a synergistic effect [34]. Bev induces vascular normalization, which enhances drug distribution uniformity and increases intratumoral drug perfusion. Furthermore, unlike TMZ, the first-line CT agent for GBM, CT agents such as Iri and carboplatin, do not have cross-resistance and demonstrate proven antitumor efficacy in other malignancies [35].

Main clinical trial outcomes are summarized in Table 1. In 2014, the first randomized controlled trial of Bev for rGBM - the phase II BELOB study - was conducted by Taal et al., in which 145 rGBM patients were randomized to three arms: CCNU monotherapy, Bev monotherapy, or CCNU+Bev combination therapy [36]. The 9-month OS rates were reported as 43%, 38%, and 59%, respectively, with the combination group showing superior survival compared to the Bev alone group [36]. However, Bev monotherapy achieved a higher ORR (38% vs. 5%) and longer mPFS (3 vs. 1 month) than CCNU monotherapy. No crossover Bev administration occurred among the 3 groups [36].

To validate the BELOB findings, the phase III EORTC 26101 trial, led by Wick et al., enrolled 437 rGBM patients randomized 2:1 to receive Bev+CCNU or CCNU alone [10]. Although combination therapy significantly prolonged PFS (4.2 vs. 1.5 months), no mOS benefit was observed. The addition of Bev exhibited higher rates of grade 3-5 adverse events (AEs) (63.6% vs. 38.1%), and it did not improve patients’ neurocognitive function; in fact, their overall health status scores were worse [10]. Consequently, the BELOB results were not replicated in this larger trial. Furthermore, a cost-utility analysis by Chen et al. further concluded that the Bev+CCNU combination regimen lacked cost effectiveness in the EORTC 26101 study [37].

The randomized phase II CABARET study compared Bev+carboplatin to Bev monotherapy in rGBM, revealing no significant differences in mPFS, mOS, or ORR [38]. Notably, the ORR (6%-14%) was markedly lower than historical controls, potentially attributable to the mRANO criteria (incorporating T2/FLAIR assessments) and approximately one-third of the enrolled patients with ≥2 prior recurrences [38].

In summary, while Bev+CT combinations prolong PFS, they concurrently increase toxicity without conferring OS benefits. Large phase III trials remain scarce due to the low incidence of GBM. Additionally, heterogeneity in trial design, eligibility criteria, prior therapies, and response assessment protocols limits cross-trial comparability.

With targeted therapy

The limited efficacy of Bev monotherapy in rGBM may be attributed to redundant angiogenesis signaling pathways or compensatory dysregulation of downstream signaling molecules of VEGF/VEGFR signaling molecules, such as epidermal growth factor receptor (EGFR) overexpression or gene amplification (observed in ~50% of cases), platelet-derived growth factor receptor (PDGFR) overexpression (75%), and aberrant mesenchymal-epithelial transition factor pathways activity [39,40]. Consequently, recent therapeutic strategies have focused on combining Bev with multiple targeted agents to overcome resistance.

Several phase I/II trials have explored the effects of Bev combined with targeted therapies for rGBM, including three randomized controlled phase II studies (see Table 2). Among these studies, only one study demonstrated a survival benefit [41]. This single-arm phase II study reported that TVB-2640 (a fatty acid synthase inhibitor) combined with Bev significantly improved mPFS (4.6 vs. 3 months) and PFS6 (31.4% vs. 16%) compared to Bev monotherapy in the historical BELOB cohort [41]. However, no significant OS advantage was observed.

Table 2.

Bev in combination with targeted therapy for the treatment of rGBM

Ref.	Study Design	NO. patients	Study arms	ORR (%)	mOS (months)	mPFS (months)	PFS6 (%)
Puduvalli 2020 [153]	II, multicenter, randomized, bayesian adaptiv	49	Vorinostat + Bev	NA	7.8	3.7	25
		41	Bev		9.3	3.9	28
Lee 2020 NRG/RTOG 1122 [154]	II, randomized, double-blinded, placebo-controlled study	57	Trebananib + Bev	4.2	7.5	4.8	22.6
		58	Trebananib + Pla	5.9	11.5	4.2*	41.4
Kelly 2023 [41]	II, single-institution, open-label	25	TVB-2640 + Bev	56	8.9	4.6	31.4►►,*
McCrea 2021 [155]	I	13	Cetuximab + Bev	30	7.2	NA	NA
	2013-2018
Brenner 2021 [136]	II, open-label, Single group	33	Evofosfamide + Bev	9	4.6	53 days	31‡
	2015-2017
Galanis 2022 N1174 [156]	I/II	52	TRC105 + Bev	13	9.7	2.9	25
		49	Bev	16	7.4	3.2	30
Zhao 2024 [42]	II, Single group, rHGG	18	Anlotinib	28	15	4.2	NA
	2020-2022	7	Anlotinib + Bev†	43	9.8	8
Cloughesy 2017 GO27819 [39]	II, double-blinded, randomized controlled, multicenter	64	Onartuzumab + Bev	22.2	8.8	3.9	29.7
		65	Pla + Bev	23.7	12.6	2.9	26.2
Cardona 2021 [157]	Retrospective, EGFR amplification and EGFRvIII mutation rGBM	15	osimertinib + Bev	13.3	9.0	5.1	46.7

Note: Pla, placebo;

^‡4-month PFS rate;

^►►control from the Bev monotherapy group in the BELOB study;

^†When peritumoral edema develops, patients will receive interim treatment with Bev;

TRC105, anti-CD105 antibody;

^* P < 0.05.

Additionally, a phase II trial involving 25 recurrent high-grade glioma (rHGG) patients evaluated anlotinib (a multitarget tyrosine kinase inhibitor targeting VEGFR and PDGFR, etc.) alone or combined with Bev (initiated upon peritumoral edema occurred) [42]. The combination failed to enhance efficacy but significantly increased treatment-related adverse events (TRAEs) [42].

In summary, as molecular profiling of GBM advances, targeted therapy has emerged as a hot topic of intensive research. While double-blind phase II RCTs provide level II evidence [43], no published phase III studies have yet validated the effectiveness of Bev-targeted therapy combinations. Thus, the use of Bev combined with targeted agents for rGBM remains investigational and is not recommended outside clinical trials.

With immunotherapy (IT)

Patients with rGBM often derive limited benefit from IT alone, partially due in part to the VEGF-driven immunosuppressive TME [44]. Bev can reduce the use of corticosteroids, which decrease effector immune cell levels [45]. In addition, Bev synergistically enhances the antitumor activity of IT, as evidenced by preclinical and clinical data from other cancer types [46].

Currently, the application of Bev-IT combinations in rGBM remains exploratory. The findings from the main studies are compiled in Table 3. The phase III CheckMate 143 study compared nivolumab (Nivo, a PD-1 monoclonal antibody) monotherapy to Bev monotherapy in rGBM, revealing superior PFS and ORR for Bev, though OS and serious adverse events rates did not differ significantly [47]. In a phase II study Nayak et al., 80 rGBM patients randomized to pembrolizumab (Pemb, a PD-1 monoclonal antibody) ± Bev demonstrated improved mPFS (4.1 vs. 1.4 months) and ORR (20% vs. 0%) with combination therapy, yet no OS advantage (8.8 vs. 10.3 months) [48]. Similarly, durvalumab (anti-PD-L1) ± Bev failed to meet efficacy expectations in another phase II trial (mOS: 6.7-9.3 months) [45].

Table 3.

Summary of published clinical studies of Bev in combination with immunotherapy for the treatment of rGBM

Ref.	Study Design	NO. patients	Study arms	ORR (%)	mOS (mos)	mPFS (mos)	PFS6 (%)
Wang 2024 [51]	IIa, single center, single-arm, open-label	32	Tislelizumab + low-dose Bev	25	22.3	4.0	NA
	2021-2023
Guo 2024 [50]	II, single center, open-label	32	Tislelizumab + low-dose Bev	56.3	14.3	8.2	NA
	2022-2023
Chiu 2023 [49]	I, open-label, nonrandomized	5	Avelumab + Bev	NA	13	3	16.5
	2018-2019	7	LITT + avelumab + Bev		13.5	2.5
Bota 2022 [158]	Prospective	21	SITOIGANAP + GM-CSF + cyclophosphamide + Bev + Nivo or Pemb	NA	19.63	9.14	76.19
	2018-2021
Sahebjam 2021 [72]	I, single-arm, open-label	24 (Bev-naïve cohort)	HFSRT + Pemb + Bev	83	13.4	7.92	66.67
	2015-2019	8 (Bev-resistant cohort)		62	9.3	6.54
Nayak 2021 [48]	II, randomized, multicenter, open-label, Bev-naïve	50	Pemb + Bev	20	8.8	4.1	26.0
	2015-2016	30	Pemb	0	10.3	1.43	6.7
Ahluwalia 2021 [159]	II, randomized	90	Nivo + standard-dose Bev (10 mg/kg)	NA	41.1	NA	NA
	2018-2020		Nivo + low-dose Bev (3 mg/kg)		37.7▼
Brenner 2020 [54]	I/II,	72 (all)	VB-111 dose escalation	63	315 days	55	NA
	rGBM		VB-111 monotherapy	53	223 days*	60*
	2011-2015		VB-111→(VB-111 + Bev)	21	414 days	90
			VB-111 + Bev	20	141.5 days**	63
Cloughesy 2020 (GLOBE) [53]	III, randomized 1:1	128	VB-111 + Bev	27.3	6.8	3.4	22.7
	2015-2017	128	Bev	21.9	7.9	3.7	28.9
Reardon 2020 (CheckMate 143) [47]	III, randomized 1:1	184	Nivo vs. Bev	7.8	9.8	1.5	15.7
	2014-2015	185		23.1	10.0	3.5**	29.6
Nayak 2022 [45]	II, multicenter, nonrandomized	A: 40 (unmethylated MGMT promotor)	Durvalumab + standard RT	10.3	15.1	4.6	41.1
	2015-2017	B: 31 (bevacizumab naïve)	Durvalumab	12.9	6.7	3.0	19.4
		B2: 33 (bevacizumab naïve)	Durvalumab + t standard-dose Bev	9.1	8.7	3.7	15.2
		B3: 33 (bevacizumab naïve)	Durvalumab + low-dose Bev	9.1	9.3	3.7	17.2
		C: 22 (bevacizumab refractory)	Durvalumab + t standard-dose Bev	0	4.5	1.9	0
Wang 2024 [51]	IIa, single center	32	Tislelizumab + Bev	25	22.3	4	NA
	rGBM
	2021-2023

Note: LITT, laser interstitial thermal therapy; Nivo, nivolumab; Pemb, pembrolizumab; SITOIGANAP, tumor cells and lysates; RT, radiation therapy;

^▼1-year OS rate;

^* P < 0.05;

^** P < 0.001;

Tislelizumab, an anti-PD-1 antibody.

Contrastingly, recent phase I/II studies have indicated that Bev combined with avelumab (a PD-L1 monoclonal antibody) or toripalimab (a PD-1 monoclonal antibody) can improve survival, with a median OS of 13-22.3 months [49-51]. Meanwhile, Yang et al. reported a case of rGBM with lung metastasis achieving 11-month PFS and 27-month OS. This patient received Bev combined with Pemb due to PD-L1 overexpression [52]. In addition to the limited number of case reports and differences in local drug penetration concentrations, further research is warranted to investigate whether extracranial metastatic sites and intracranial recurrence sites exhibit distinct TMEs and tumor cell biological characteristics [52].

Novel approaches, such as viruses or vaccines, have also been explored. The phase III GLOBE trial evaluated VB-111 (an anticancer viral therapy that induces endothelial cell apoptosis and activates antitumor immunity) ± Bev, showing no mOS improvement but increased grade 3-5 adverse events with combination therapy (67% vs. 40%) [53]. A study by Brenner et al. yielded similar results in the same year [54]. Additionally, Brenner et al. found a significant prolongation of survival in the subgroup of VB-111 monotherapy upon continued in combination with Bev after recurrence [54], suggesting timing-dependent immune activation. Similarly, neoadjuvant/adjuvant Pemb improves survival versus adjuvant-only regimens, potentially via enhanced tumor-specific T cell priming [55]. Reardon et al.’s phase II trial of rindopepimut (a vaccine targeting EGFRvIII) ± Bev reported comparable PFS6 but a higher 24-month OS rate than the control group did (20% vs. 3%) and corticosteroids discontinued rates (33% vs. 0%) in the combination arm [56].

In summary, the conclusions drawn from various studies differ, which may be related to the characteristics of the drugs studied, the baseline characteristics of the enrolled patients, and so on. Currently, there is no conclusive high-level evidence supporting the use of Bev combined with IT drugs in the treatment of rGBM patients. However, this combination therapy shows great promise and has become a hot area of current research (see Table 4).

Table 4.

Summary of ongoing clinical studies of Bev in combination with immunotherapy for the treatment of rGBM (data from https://clinicaltrials.gov)

NCT	Study Start	Enrollment	Immunological drug	Phase	Interventions	Primary endpoint
NCT04116658	2020-07-13	100	EO2401, Nivo	I/II	EO240	OS
					EO2401 + Nivo
					EO2401 + Nivo + Bev
NCT06061809	2024-08-07	20	N-803, PD-L1 t-haNK	II	N-803 + PD-L1 t-haNK + Bev	TEAEs, SAEs
NCT04277221	2019-09-19	118	ADCTA	III	ADCTA vs. Bev	OS
NCT06047379	2023-11-01	134	Ipi, Pemb and Nivo	I/II	NEO212 + Ipi	MTD, PFS6
					NEO212 + Pemb
					NEO212 + Nivo
					NEO212 + regorafenib NEO212 + carboplatin + paclitaxel
					NEO212+ FOLFIRI + Bev
					NEO212
NCT05502991	2022-12-11	60	Sintilimab	II	Sintilimab + Bev	12/18-month OS rate

Note: TEAEs, treatment-emergent adverse events; SAEs, serious adverse events; ADCTA, autologous dendritic cell/tumor antigen; NEO212, TMZ + perillyl alcohol; Ipi, ipilimumab; FOLFIRI, 5-fluorouracil + leucovorin + Iri.

With reirradiation

Reirradiation (re-RT) is an important therapeutic modality for rGBM. The mechanisms underlying the combination of Bev with re-RT are multifactorial. Preclinical studies have demonstrated that ionizing radiation markedly elevates VEGF levels in tumors, while Bev enhances intratumoral oxygenation and disrupts the vascular niche harboring cancer stem cells (CSCs) [57]. Clinical evidence further indicates that Bev mitigates the risk of radiation necrosis (RN) and cerebral edema [58].

When Bev is combined with re-RT techniques - including HSRT, SRS, and GK - mOS ranges from 10.1 to 13.9 months, and mPFS spans 5-11 months (Table 5). The NRG Oncology/RTOG1205 trial, the first multicenter phase II study to utilize modern RT techniques in rGBM, evaluated the efficacy of integrating re-RT with Bev [59]. Between 2012 and 2016, 170 patients were randomized 1:1 to Bev+re-RT or Bev monotherapy. No significant difference in mOS was observed (10.1 vs. 9.7 months); however, the combination group exhibited superior mPFS (7.1 vs. 3.8 months) and PFS6 rates (54.3% vs. 29.1%) [59]. Acute grade 3+ TRAEs occurred in only 4.8% of the combination group [59]. Notably, baseline imbalances were observed: the combination group included more patients with ≥2 more recurrences (25.6% vs. 13.1%), a poor prognostic factor, but also a higher proportion with O6-methylguanine DNA methyltransferase (MGMT) promoter methylation (20.9% vs. 14.3%), a favorable prognostic factor [59]. Prolonged PFS may delay declines in QoL and neurocognitive function [59]. Multivariate analysis identified Karnofsky performance status (KPS) - not age - as a significant OS prognosticator, supporting this regimen’s applicability to elderly patients with KPS scores ≥ 70%. A retrospective study also reported that SRS+Bev is a positive prognostic marker for HGG [60]. The ongoing phase II HSCK-005 trial (NCT05611645) is investigating whether HFSRT combined with Bev enhances PFS6 in rGBM.

Table 5.

Bev in combination with re-RT for the treatment of rGBM

Ref.	Study design	No. patients	Treatment	mOS (months)	mPFS (months)	PFS6 (%)	Median GTV (cc) (Range)	RT dose (Gy/Fractions)
Tsien 2023 [59]	II, andomized	170	BEV + re-RT	10.1	7.1	54.3	18 (0.5-208)	35 Gy/10 f
	2012-2016		BEV	9.7	3.8*	29.1*
She 2022 [160]	Retrospective	26	Bev + re-RT	13.6	8.0	65.4	Median PTV size (ml) (range): 114.8 (11.9-360.1)	30-35 Gy/10 f (for the small GTV)
	2019-2021							40-45 Gy/20-27 f (for the large GTV)
Arvold 2017 [65]	Retrospective	67	re-RT	10.7	4.8	NA	NA	30 Gy/5 f (36%), 35 Gy/10 f (21%), 40 Gy/15 f (15%), 18-20 Gy/1 f (15%)
	2010-2014	117	Bev	NA	NA
Palmer 2018 [61]	Retrospective	118 (GBM: 87; AA: 31):	Bev → HFSRT	13.3	NA	NA	44.8 cm3	35 Gy/10 f
	2006-2013	50	FSRT → Bev	13.9			17.04 cm3
		68
Zhang 2024 [161]	Retrospective	19	Bev	6.5	5.0	NA	15.9 ± 11.5	NA
	2012-2022	57	Bev + GK	11.5	7.7		7.6 ± 11.6
		19	GK	7.9*	4.9*		13.2 ± 16.1
Mantica 2023 [162]	II	16	Bev + SRS	11.73	NA	65.2	NA	NA
	2015-2017		SRS	8.74*		33.2*
Morris 2019 [163]	Retrospective	45	Bev + GK	13.3	5.2	NA	2.2 cm3 (0.1-25.2 cm3)	17.0 Gy
	2009-2015
Abbassy 2018 [164]	I	9	Bev + SRS	13	7.5	NA	4.4 cm3 (2.1-8.8)	18 Gy
							5.13 cm3 (4.38-6.47)	20 Gy
							4.64 cm3 (2.54-6.62)	22 Gy

Note: HFSRT, hypofractionated stereotactic radiotherapy; SRS, stereotactic radiosurgery; GK, Gamma Knife;

^* P < 0.05.

Bev administration timing relative to re-RT - neoadjuvant (neoBev), concurrent (concBev), or adjuvant (adjBev) - has been explored. A large retrospective study by Palmer et al. found no survival difference in rHGG patients receiving Bev before or after re-RT [61]. Conversely, Cuneo et al. reported that adjBev post-SRS significantly improved mOS (5.2 vs. 2.1 months) and 1-year OS compared to no Bev [62]. A retrospective multicenter study of 482 rHGG patients (1997-2023), presented at the 2024 American Society of Clinical Oncology annual meeting, found adjBev unrelated to survival outcomes, whereas concBev correlated with poorer OS but lower neurotoxicity [63]. Thus, optimal Bev-re-RT sequencing remains undefined.

Consensus is similarly lacking on optimal re-RT target volume and dosing with Bev. In re-RT monotherapy, comparable survival was observed with 25 Gy/5 f and 35 Gy/5 f [64]. There was no difference in OS or PFS when comparing local re-RT to systemic treatment with Bev [65]. However, Kulinich et al. identified Bev as a survival enhancer in HFSRT-treated patients [66]. Recently, the GLIAA study prospectively compared 18F-fluoroethyltyrosine (FET) positron emission tomography (PET) and T1Gd MRI-guided re-RT in rGBM, revealing no differences in survival, local control rates, or RN rates [67]. Additionally, a large-scale retrospective study reported that large-volume re-RT (median PTV: 135 cm³) has been deemed feasible for refractory rGBM [68].

Radiation-related toxicity remains a key concern. As GBM typically recurs near surgical margins, re-RT target often overlap initial RT fields. RN - the most severe late radiation complication - is pathologically characterized by brain tissue necrosis. The exact mechanisms underlying RN are still unclear, but they may be related to the loss of pericytes and vascular smooth muscle cells [69]. Clinically, RN usually presents with the re-emergence of symptoms and a decline in neurological function, while imaging reveals enhancing lesions indistinguishable from tumor progression. A recent retrospective analysis revealed that apparent diffusion coefficient values exhibit significant discriminative capability [70]. Additionally, most clinical studies do not report the incidence of RN. The phase II CHROME trial protocol recently proposed chlorophyllin for RN management in diffuse gliomas [71], though its efficacy versus corticosteroids, Bev, or surgery warrants validation.

In summary, the ideal Bev-re-RT regimen remains uncertain, with most evidence derived from retrospective studies prone to selection bias. Incorporating IT into the Bev-re-RT strategies represents a promising frontier. A phase I trial of HFSRT combined with pembrolizumab (Pemb) and Bev in 32 rHGG patients demonstrated safety profile but no survival benefits [72]. Ongoing trials (e.g., NCT03743662, NCT06160206) aim to address these gaps.

Newly diagnosed glioblastoma (nGBM)

In rGBM, Bev demonstrates potent tumor-shrinking activity and significantly prolongs PFS, while its combination with RT or TMZ has been demonstrated to be safe. However, the inclusion of Bev in standard first-line therapy for nGBM remains controversial.

With the Stupp regimen

Several phase II studies have reported that the mOS for standard therapy combined with Bev is approximately 19.6-23 months, with a mPFS of approximately 13-14.2 months (Table 6). While PFS was significantly prolonged compared to historical or contemporary controls, OS improvements were observed only relative to historical controls [73]. These findings suggest the survival status of nGBM patients treated with the same first-line regimen has improved compared with that in the EORTC-NCIC era, potentially attributable to advancements in patient management [9,74]. However, cross-trial comparisons are confounded by variability in OS definitions - for instance, Vredenburgh et al. calculated OS from enrollment [73], whereas Lai et al. used the date of diagnosis as the starting point [74].

Table 6.

Bev in combination with chemoradiotherapy for the treatment of nGBM

Ref.	Study design	No. of patients	Study arms	mOS (months)	mPFS (months)
Vredenburgh 2011 [73]	II, single group	75	BEV + RT + TMZ→TMZ + Bev + Iri	21.2	14.2
	2006-2008
Lai 2011 [74]	II, single group	70	I: BEV + RT + TMZ→TMZ + Bev	19.6	13.6
	2006-2008		C1 (UCLA): RT + TMZ→TMZ	21.1	7.6
			C2 (EORTC-NCIC): RT + TMZ→TMZ	14.6	6.9
Chinot 2014 AVAglio [8]	III, randomized	921	I: Bev + RT + TMZ→TMZ + Bev	16.8	10.6
	2009-2011		C: RT + TMZ→TMZ	16.7	6.2*
Gilbert 2014 RTOG 0825 [9]	III, randomized	637	I: BEV + RT + TMZ→TMZ + Bev	15.7	10.7
	2009-2011		C: RT + TMZ + placebo→TMZ	16.1	7.3
Herrlinger 2016 GLARIUS [165]	II, open-label, randomized, MGMTmet	122	I: BEV + RT→Bev + Iri	16.6	5.99
		60	C: RT + TMZ→TMZ	17.5	9.7*
Reyes-Botero 2018 ATAG [85]	II, single group, older than 70 years, KPS < 70%	66	BEV + TMZ	5.98	3.83
Nagane 2022 [166]	II, multicenter	90	BEV + RT + TMZ→TMZ + Bev→Bev	25	14.9
Omuro 2014 MSKCC 08-126 [90]	II,	40	HFSRT (36 Gy/6 f) + TMZ + Bev→TMZ + Bev	19	10
Ney 2015 [89]	II, 2010-2013	30	Hypo-IMRT (60 Gy/10 f) + Bev + TMZ	16.3	14.3
Carlson 2015 [93]	II, comparative	26	Hypo-IMRT (60 Gy/10 f) + TMZ	16.3	9.4
		30	Hypo-IMRT (60 Gy/10 f) + TMZ + Bev	16.3	12.8
Wirsching 2018 ARTE [76]	II, noncomparative, multicenter, older than 65 years	50	Arm A: Hypo-IMRT (40 Gy/15 f) + Bev	12.1	7.4
	2013-2015	25	Arm B: Hypo-IMRT (40 Gy/15)	12.2	4.8*
Matsuda 2018 [167]	Retrospective, older than 75 years	18	HFRT (45 Gy/15 f) + TMZ/Bev	20	2.5
	2014-2017
Ohno 2019 [88]	Retrospective, older than 75 years	20	HFRT (45 Gy/15 f) + TMZ	12.9	8.5
	2007-2018	10	HFRT (45 Gy/15 f) + TMZ/Bev	14.6	10.4
Kanamori 2025 [168]	II, 2015-2018	49	CNNU + RT + TMZ + Bec	24.8	11.8

Note: UCLA, University of California, Los Angeles; HFSRT, hypofractionated stereotactic radiotherapy; hypo-IMRT, hypofractionated-intensity modulated radiotherapy; HFRT, hypofractionated radiotherapy;

^* P < 0.05.

Two phase III studies (AVAglio [8], RTOG 0825 [9]) subsequently evaluated Bev’s addition to first-line standard therapy, yielding divergent outcomes. Although both studies failed to demonstrate OS benefits, AVAglio reported prolonged PFS (3-4 months) and delayed deterioration in health-related QoL, neurocognitive function, and corticosteroid dependence [8,9]. Conversely, the RTOG 0825 study found accelerated QoL and neurocognitive decline in the Bev arm [75], a trend echoed in the ARTE trial [76]. These discrepancies may reflect differences in response criteria (e.g., the AVAglio study’s modified MacDonald criteria [8]), crossover effects (48% of control-group patients received post-progression Bev), or imbalances in treatment cycles (17-27% higher 6-cycles TMZ completion rates in Bev arms) [8,9]. The incidence of adverse events was similar in both studies, but the percentage of patients who discontinued treatment due to toxicity or complications was approximately 15% greater in the experimental group than in the control group [8,9]. Notably, subgroup analysis from the AVAglio study hinted at OS/PFS benefits in patients not receiving post-progression therapy [77], though clinical implementation remains impractical due to challenges in pretreatment identification.

On the basis of the results of the AVAglio study, Japan approved the use of Bev for treating nGBM in 2013 [78]. Motoo et al. subsequently conducted postmarketing surveillance of Bev in the Japanese GBM population [8,78]. The analysis replicated the AVAglio trial’s safety profile (≥ grade 3 AE rates: 15.1%) and reported a 1-year OS rate of 78% at 18-month follow-up, while the mOS had not yet been reached [78]. However, can the reported survival improvement be entirely attributed to Bev? Can the results from this population be effectively extrapolated to other ethnicities?

Recently, some researchers have suggested the use of Bev before chemoradiotherapy or even before surgery (Table 7). The theoretical basis is that Bev can shrink tumors and prune blood vessels, thereby reducing surgical difficulty and improving the R0 resection rate; Bev can alleviate patients’ brain edema, improve their physical condition, and enhance surgical tolerance [79]. Several small-sample phase II studies have shown that this approach has no advantage in terms of ORR, PFS, or OS and may negatively affect the completion of subsequent treatments due to toxicity [79-81]. The TEMAVIR study further revealed lower concurrent chemoradiation completion rates in the Bev group (58% vs. 82%) [81]. Thus, for large, unresectable nGBM, the benefits of neoadjuvant Bev are limited.

Table 7.

Bev before chemoradiotherapy for the treatment of nGBM

Ref.	Study design	No. of patients	Study arms	ORR (%)	mOS (months)	mPFS (months)
Chauffert 2014 TEMAVIR [81]	II, noncomparative, randomized, unresectable	60	BEV + Iri→RT + TMZ + Bev→Bev + Iri	NA	11.1	7.1
	2009-2011	60	RT + TMZ→TMZ		11.1	5.2
van Linde 2015 [79]	Single group	19	BEV + RT + TMZ→TMZ	NA	16	9.6
Balana 2016 GENOM 009 [80]	II, randomized, unresectable	44	BEV + TMZ→RT + TMZ + Bev→TMZ	22.9	10.6	4.8
	2009-2013	43	TMZ→RT + TMZ→TMZ	6.7*	7.7	2.2
Tanaka 2024 [169]	I/II	15	Bev + TMZ→Surgery→RT + TMZ→TMZ	NA	16.5	9.5
	2017-2021

^* P < 0.05;

NA, not available.

In conclusion, the use of Bev in nGBM patients is not recommended due to its high incremental cost-utility ratio. One of the future strategies is to identify a beneficial population, such as the proneuronal subtype linked to OS benefits in AVAglio post hoc analyses [82].

With hypofractionated radiotherapy (HFRT)

HFRT represents a viable therapeutic option for elderly or frail nGBM patients [83]. A systematic review indicates comparable OS between HFRT and standard fractionated in HGG, particularly for GBM patients aged ≥60 years [84]. Bev has also demonstrated potential survival benefits in this population [5,11,74,85]. Notably, TMZ monotherapy or TMZ+HFRT remains preferred for elderly GBM patients with MGMT promoter methylation [86,87].

Compared with conventional fractionation, HFRT offers advantages such as shortened treatment duration, improved patient compliance, and mitigation of tumor cell repopulation via reducing the repair of sublethal damage. However, its application is limited by the elevated risk of RN and delayed neurotoxicity. Bev may theoretically reduce RN and brain edema. Therefore, Bev is expected to improve the tolerance of nGBM patients to hypofractionated radiotherapy while increasing the biological effective dose (BED).

Clinical investigations of HFRT+Bev in elderly GBM patients are summarized in Table 6. A phase II ARTE study compared the HFST ± Bev in elderly or frail nGBM patients with unmethylated MGMT promoter [76]. The results indicated that the addition of Bev extended PFS by approximately 3 months but did not improve OS, a finding similar to that in younger adult nGBM patient populations [76]. Additionally, a retrospective study compared HFRT combined with TMZ or TMZ+Bev in nGBM patients aged ≥75 years and reported that adding Bev increased a nonsignificant 2-month OS [88]. Interestingly, subgroup analysis in this study showed MGMT status showed no prognostic impact [88], contradicting the conclusions of the CAN-NCIC-CE6 study [87]. However, all three studies included surgically treated patients with acceptable tolerance, leaving the utility of these regimens in inoperable elderly GBM patients unclear.

In younger adult patients with nGBM, several studies have investigated Bev+HFRT (Table 6). This section focuses on RN and recurrence patterns reported in these studies. First, higher PTV volumes and escalated radiation doses in HFRT are associated with an increased risk of RN. A phase II study with a median PTV1 volume of 131.1 cm³ (BED = 96) reported RN in 50% of patients, prompting early study termination [89]. Conversely, the MSKCC 08-126 study, which restricted tumor volume to ≤60 cc (BED = 57.6), observed no RN cases [90]. Second, RN has been linked to extended survival outcomes, with patients experiencing RN demonstrating a median OS advantage of approximately 3 months compared to non-RN cohorts [89,91,92]. However, long-term survivors with RN exhibited diminished QoL. Third, Bev seems to have a limited role in RN prevention [89,93], and its use correlates with a shift in failure patterns from local to distant intracranial recurrence (see below) [91-93].

In summary, Bev+HFST did not achieve the expected survival outcomes, and its efficacy in elderly GBM patients is similar to that in younger adult patients. Prospective studies contradict retrospective analyses, as the former do not support Bev’s clinical utility. Current evidence remains confined to phase II studies.

Optimization of Bev dosing regimens and administration strategies

The optimal dosing regimen for Bev in GBM remains undefined. Preclinical glioma mouse models demonstrate that both low and high Bev doses induce vascular regression, with the latter also inhibiting the activity and growth of glioma cells [22]. However, high doses of Bev may exacerbate hypoxia and promote immunosuppressive myeloid cell infiltration.

Interestingly, clinical studies suggest no dose-dependent antitumor efficacy for Bev [94,95]. Most trials have adopted regimens of 5-15 mg/kg every 2-3 weeks. A phase II single-arm study first explored 15 mg/kg Bev every 3 weeks, achieving a PFS6 rate of 25% and mOS of 25.6 weeks [96]. Bokstein et al. reported comparable efficacy with reduced toxicity using low-dose Bev (5 mg/kg biweekly) in HGG patients with poor performance status (mean KPS = 65%) [97]. An early meta-analysis of rGBM studies found no dose-response effect, with similar outcomes between high- (10-15 mg/kg) and low-dose (5 mg/kg) regimens [97,98]. However, it remains unclear whether higher doses lead to a faster onset of action [98]. The phase II VAMANA study recently reported an mOS of 6.1 months for ultralow-dose Bev (1.5 mg/kg triweekly) combined with CCNU in rGBM [99]. In fact, nonrandomized controlled phase II trials were classified as Class III evidence [43].

Enhancing Bev’s intracranial delivery is critical. Rubenstein et al. proposed that antiangiogenic drugs target abnormal blood vessels and are therefore not limited by the blood-brain barrier (BBB) [100]. Indeed, intravenous administration often fails to achieve therapeutic intracranial concentrations in GBM patients due to the BBB and systemic dose-limiting toxicity. Selective intra-arterial drug infusion - direct drug delivery to tumor-feeding arteries - elevates local drug exposure by 3-5.5-fold [101,102]. Patel et al. combined hyperosmotic BBB disruption with superselective intra-arterial brain infusion (SIACI)-administered Bev in nGBM patients, achieving a mPFS of 11.5 months (95% CI 7.7-25.9 months) and mOS of 23.1 months, with no Grade ≥3 TRAEs [103]. Admittedly, this new treatment technique is more precise and shows improved efficacy [103]. However, the study has several limitations, including the absence of a control group and a small sample size (n = 31) [103]. SIACI faces challenges related to complex hemodynamics and transient drug retention [104], and the technique relies on the operator’s skill level, which has a long training period, making widespread adoption difficult.

Notably, the application of novel nanotechnology-based carriers in in vitro models has been reported to suppress angiogenesis and reduce tumor volume, demonstrating promising therapeutic efficacy [105,106]. Moreover, the nanocapsules can be administered intranasally [107]. Recently, an exosome-based Bev delivery platform engineered by Chu et al. demonstrated enhanced BBB penetrability [108].

In conclusion, the distinct pharmacokinetic and pharmacodynamic profiles of targeted therapies compared to traditional chemotherapy necessitate comprehensive dose-response studies for Bev in GBM, providing important evidence for the use of low-dose Bev in patients with poor performance status or in cost-constrained therapeutic settings. The development of advanced drug delivery systems specifically targeting GBM remains insufficient, whereas novel biomimetic materials persist as a focal point in current research efforts.

Treatment failure patterns associated with Bev

The duration of response to Bev in GBM is short, with 40-60% of patients experiencing recurrence within six months, predominantly exhibiting local progression [109,110]. Treatment failure may be driven by phosphofructokinase-1, muscle isoform, mesenchymal transition, and aberrant activation of the cell adhesion molecule pathway [111-113]. Through microarray analysis, DeLay et al. classified Bev-resistant GBM (BRG) MRI enhancement patterns into two subtypes: enhanced (62%) and nonenhanced type (34.2%) [114,115]. BRG patients have an mOS of 2.5 months (range: 1-4.5 months) [96,116].

The impact of Bev on recurrence patterns - specifically, diffuse vs. nondiffuse invasion and multifocal vs. focal growth - remains controversial. Lucio and Rubenstein reported that long-term use of Bev enhances GBM invasiveness, characterized by multifocal satellite lesions and finger-like tumor cell formations [100,117]. The formation of these structures helps hijack existing host blood vessels [100]. Norden et al. observed higher rates of diffuse and metastatic progression in the Bev-treated patients compared to controls (30% vs. 21%) [109,110], potentially mediated by CXCL12/CXCR4 axis activation and MET/VEGFR2 complex formation [118,119]. Conversely, post hoc analyses of the GLARIUS and AVAglio trials found no association between Bev and multifocal or diffuse recurrence [120,121]. Furthermore, the prognostic significance of diffuse progression is unclear. Pope et al. reported similar survival outcomes for patients with local-to-diffuse versus local-to-local progression [122]. Another study indicated that T2 diffuse progression is associated with prolonged survival [123]. Prognostic factors in Bev-treated GBM will be discussed in detail later.

Bev as a third-line therapeutic option for Bev-naïve or resistant rGBM

Currently, there is no standard third-line therapy established for rGBM. Given the limited treatment options, whether Bev can be used to delay or continue treatment for secondary or later relapses remains exploratory. Franceschi et al. retrospectively analyzed 168 GBM patients receiving third-line treatment, revealing a 2-month survival benefit with Bev compared to chemotherapy (P = 0.014) [124]. In most third-line Bev studies, first recurrence (second-line) treatments primarily include CT (such as nitrosoureas) or Bev.

Post-progression options after second-line Bev include continued Bev, CT, and re-RT. First, two randomized phase II trials - CABARET and TAMIGA - evaluated whether Bev continuation after progression during Bev treatment improved outcomes, with no significant survival or QoL benefits compared to non-Bev therapies [116,125]. Schaub et al. reported similar PFS between third- and second-line Bev treatment [126]. Second, retrospective studies noted that Bev combined with nitrosourea increased toxicity without survival benefits [127]. Finally, re-RT combined with Bev demonstrated good tolerability and superior survival (mOS: 4.8-8.8 months) compared to historical controls in rHGG after second-line Bev failure [128-132]. These findings suggest that Bev+re-RT may outperform BEV-based CT in survival and toxicity. Can the addition of IT to re-RT combined with Bev further reverse Bev resistance? A recent phase II study reported improved survival with re-RT plus Pemb and Bev in Bev-refractory patients [133].

On the other hand, two retrospective studies support single-agent Bev feasibility in GBM patients progressing after second-line nitrosoureas (such as CCNU or FTM), with mOS of 6-7.5 months and PFS6 rates of 13-21.5% [134,135].

In summary, for patients progressing after second-line CT or Bev treatment, single-agent Bev or Bev plus re-RT may be viable options. However, due to short survival, many patients only receive first- or second-line treatments, driving current research to optimize interventions for newly diagnosed patients or first recurrence GBM. As OS improves, identifying an optimal subsequent therapy plan will become increasingly important. Recent efforts explore novel agents, such as evofosfamide [136], carotuximab [137], and base excision repair inhibitor TRC102 [138], for Bev-refractory GBM.

Prognostic markers for Bev

In the context of limited survival benefits in the general GBM population, the identification of patient subgroups likely to benefit from Bev is critical.

Genetic, molecular, imaging, and clinical markers associated with Bev response in GBM have been identified in recent studies (Table 8). For instance, Takei et al. demonstrated that reduced FOXM1 expression predicts prolonged survival in Bev-treated GBM patients [139]. Clinical factors such as hypertension have also been linked to favorable responses [140], while Hiller-Vallina et al. identified sexualbiased necroinflammation as a novel positive predictor [141]. Conversely, negative correlations have been reported with soluble CD146 secretion [142], elevated antibody-dependent cellular cytotoxicity (ADCC) activity [143], and c-Met/VEGFR2 overexpression [144], underscoring the multifactorial nature of resistance mechanisms in Bev-treated GBM.

Table 8.

Prognostic biomarkers of BEV

Ref.	Prognostic factors	Positive/Negative
Fu 2021 [82,112]	Low ITGAM expression, proneural subtype	Positive
Takei 2022 [139]	FOXM1 low-expression	Positive
Scheer 2023 [140]	Hypertension	Positive
Strauss 2025 [170]
Kessler 2023 [171]	Neurofibromin 1 mutations	Positive
Kim and Breda-Yepes 2023 [172,173]	vessel size index and relative cerebral blood volume↑	Negative
Hiller-Vallina 2024 [141]	Sexual-biased necroinflammation	Positive
Joshkon 2022 [142]	high soluble CD146 secretion	Negative
Lallemand 2020 [143]	ADCC activity↑	Negative
Carvalho 2021 [144]	c-Met/VEGFR2 overexpression	Negative
Ellingson 2023 [174]	time to tumor regrowth, depth of response	Positive
Nagane 2022 [166]	macrophage or microglia activation	Positive
Jiguet-Jiglaire 2022 [175]	matrix metalloproteinase 9	Negative

However, the clinical translation of predictive biomarkers is hindered by several challenges. First, study design biases, high examination costs, invasive sample collection, and stringent regulatory approvals limit their practical application [145]. Second, the extent to which predictive biomarkers demonstrate reproducible and independent predictive utility across diverse cohorts and clinical contexts represents a pivotal consideration. Finally, future research should prioritize the integration of technological advancements, including multi-omics approaches and artificial intelligence, alongside fostering interdisciplinary collaborations and standardized protocols to expedite the translation of biomarkers from experimental research to clinical practice.

In brief, the development of personalized treatment paradigms guided by prognostic markers is recognized as an inevitable trajectory in the era of precision medicine.

Summary and outlook

GBM is the most common primary malignant brain tumor, known for its rapid progression and poor prognosis, with limited effective treatment options. Bev is one of only three options approved by the FDA for GBM treatment in the past two decades, leading to a rapid increase in related research. This paper provides a comprehensive review of recent advancements in BEV research, with a focused discussion on topics pertinent to clinical decision-making.

However, the clinical outcomes of Bev in GBM have been disappointing, with several unresolved challenges. First, the mechanisms underlying Bev’s limited efficacy are not fully understood. Second, improvements in short-term efficacy metrics - such as the ORR, PFS6 rates, and PFS - have not translated into OS benefits. Additionally, the optimal sequencing or combination of Bev with other therapies for second- or third-line therapy remains unclear. Finally, clinically applicable prognostic markers to guide Bev use in GBM are still lacking.

Recent advances, such as regorafenib’s survival benefits in rGBM, highlight the potential of antiangiogenic therapies. The exploration of novel agents or multimodal treatment approaches, such as multitarget antiangiogenic agents, phytochemicals, or combination strategies, warrants further investigation. Given the constraints imposed by BBB, enhancing drug delivery efficiency to the brain tumor has emerged as a critical focus in advancing GBM therapeutics development. Concurrently, the identification of patient subgroups likely to benefit from Bev through prognostic biomarker-driven stratification has emerged as a key area of investigation in current research. The Global Brain Tumor Adaptive Clinical Trial System (GBM AGILE, NCT03970447) has pioneered an innovative framework for clinical trial design, significantly streamlining the assessment of emerging therapies.

Disclosure of conflict of interest

None.