Malignant glioma including glioblastoma has poor prognosis, despite recent advances in molecular diagnostics and in modern individual treatment (1). The current standard of care comprises surgical resection followed by adjuvant radiotherapy and chemotherapy (2). Glioblastoma is characterized by overexpression of vascular endothelial growth factor (VEGF), a crucial factor of tumor-associated angiogenesis, and these tumors are composed of highly disorganized vessels (3). A promising alternative therapeutic approach for glioblastoma is the inhibition of angiogenesis through VEGF, which may slow down tumor growth and enhance the effects of radiotherapy and chemotherapy (4). Bevacizumab is a recombinant humanized monoclonal antibody targeting VEGF, which suppresses tumor angiogenesis and growth (5). In the AVAglio study, bevacizumab combined with standard treatment for patients with newly-diagnosed glioblastoma was associated with a 4.4-month increase in median progression-free survival (6). In the RTOG 0825 trial, first-line use of bevacizumab did not improve overall survival in patients with newly diagnosed glioblastoma, however, progression-free survival was longer in the bevacizumab group than in the placebo group (10.7 months vs. 7.3 months) (7).

Bevacizumab has some specific adverse effects, however, including proteinuria, hypertension, severe bleeding, protracted wound healing, and gastrointestinal perforation (5). In the RTOG 0825 trial, some adverse events were more prevalent in the bevacizumab group than in the placebo group, including hypertension (4.2% vs. 0.9%), thromboembolic disease (7.7% vs. 4.7%), and wound dehiscence (1.5% vs. 0.9%) (7). Administration of bevacizumab in patients with glioblastoma has been associated with an increased risk of intracerebral hemorrhage (ICH) (8). However, risk factors for the occurrence of ICH as a result of bevacizumab treatment are unclear. In this study, we therefore analyzed patients treated with bevacizumab for malignant glioma and investigated the characteristics of the patients in the cohort that developed ICH.

Bevacizumab has a specific toxicity profile related to its anti-angiogenic effect, and some bevacizumab-associated adverse events can be life-threatening (8). We investigated the overall incidence and risk of ICH among patients with high-grade glioma that underwent bevacizumab therapy. We found ICH to be associated with shorter overall survival. Also, tumor invasion into the corpus callosum before administration of bevacizumab was associated with an increased risk of developing ICH.

Among brain tumors, gliomas and metastasis tend to bleed most frequently (10). ICH occurs in 3.7-7.2% of gliomas, mainly in glioblastoma and oligodendroglioma (10). The most important etiopathogenesis of ICH in glioblastoma is considered to be the abnormalities of tumor vascularization (10). Glioblastomas are highly angiogenic and infiltrative tumors. Tumor cells invade along blood vessels to increase tumor growth, and glioblastomas displace the endfeet of astrocytes and affect pericyte stability, leading to perivascular niches and cell evasion (11). Microvascular proliferation with hyperplasia of endothelial cells following obliteration and the presence of numerous thin-walled, poorly formed, or dilated vessels may induce hemorrhages (10,11).

In the AVAglio study, the incidence of ICH was 2.5% in the placebo group and 4.3% in the bevacizumab group, suggesting that bevacizumab may contribute to development of ICH (6). A post-marketing surveillance study of bevacizumab for patients with newly diagnosed or recurrent malignant glioma in Japan reported an ICH incidence of 5.4% (12). In the current study, the hemorrhage rates were higher than in these reports. The AVAglio study reported that just 13% of patients in the bevacizumab-treated group underwent a biopsy (6), as compared with 58% in this study. We suggest the high hemorrhagic rate may be related to the patient population in this study, which includes a large portion of patients who underwent biopsy rather than tumor removal due to old age, poor preoperative performance status, and dissemination.

The pathophysiology underlying ICH in high-grade glioma with bevacizumab still requires elucidation. VEGF is thought to be important for endothelial cells to maintain the architecture and integrity of the microvasculature (5). In metastatic brain tumors, overexpression of VEGF and metalloproteinase, which degrade the extracellular matrix, may play a role in causing hemorrhages through rapid growth and breakdown of vessels around the tumor (13). Bevacizumab may cause hemorrhage by reducing the regenerative capacity of endothelial cells (14). However, dysfunction of endothelial cells alone cannot explain the ICH in patients that have received bevacizumab (13,14).

In our cohort, tumor invasion into the corpus callosum before administration of bevacizumab was associated with an increased risk of ICH. Gliomas invading the corpus callosum reportedly account for approximately 14% of gliomas (15,16). Tumor invasion into the corpus callosum has been said to be a more aggressive subtype of an already aggressive and incurable disease (16). In a recent retrospective cohort study, it was significantly associated with glioma WHO grade and PDGFRA mutation (17). Endocan, an endothelial-secreted proteoglycan reportedly activates PDGFRA and promotes a hypervascular phenotype of glioblastoma (18). The aggressive biological behavior and hypervascularization of the tumor might affect ICH after bevacizumab administration.

In relation to tumor-related ICH, a number of reports have discussed feeding arteries (11,19), but there has been little mention of the draining veins. Brain arteriovenous malformations with deep venous drainage have shown higher rates of hemorrhage, suggesting arteriovenous malformation pressurization due to draining veins (20). Inhibition of VEGF signaling suppresses the endothelial cell activation, predisposing to thromboembolic events (14). Even in malignant gliomas, the lesions invading the corpus callosum are associated with deep venous drainage, and we suggest that thrombosis of deep cerebral veins due to the use of bevacizumab may contribute to ICH.

This study has several limitations. Unlike a randomized study, there is a possibility of selection bias in the decision-making on treatment strategy. Second, the limited number of patients could be responsible for the absence of statistical power to detect differences between groups. The sample size of this study is insufficient to establish an association between invasion of corpus callosum and ICH after bevacizumab administration. Perhaps due to the small sample size, no risk factors other than corpus callosum invasion could be observed in this study. Further investigation in a larger population is required to examine other factors, such as antithrombotic medications, angiographical features, and microbleeds in the magnetic resonance imaging, which would contribute to a better understanding of bevacizumab-associated ICH in patients with malignant glioma.

In conclusion, in our cohort of patients with malignant glioma treated with bevacizumab, ICH was associated with shorter overall survival. It was suggested that invasion of the corpus callosum shown on magnetic resonance imaging might be a predictor of development of ICH in some patients that receive bevacizumab. Further studies in larger cohorts are required to confirm these results. Meticulous management, such as frequent follow-up and treatment of hypertension, may be necessary in patients at high risk for ICH after administration of bevacizumab.