What is the risk of intracranial bleeding during anti-VEGF therapy?

Abstract

Vascular endothelial growth factor (VEGF) is a key mediator of physiological and pathological angiogenesis. All solid tumors are dependent on pathological angiogenesis, and anti-VEGF therapy has demonstrated clinical benefit in breast, colorectal, non-small-cell lung, and renal carcinomas. Central nervous system metastases are common in many of these tumor types. An increased risk of bleeding has been reported with anti-VEGF therapy, but the risk of intracranial bleeding is unknown with this type of therapy. We reviewed the available data to investigate the risk of intracranial bleeding with anti-VEGF therapy in the presence and absence of CNS metastases. The PubMed and Medline databases and the Proceedings of the American Society of Clinical Oncology (ASCO) annual meetings were searched for articles, abstracts, and presentations of clinical trials. We identified 57 trials examining the safety and efficacy of anti-VEGF therapy in a total of 10,598 patients. Four trials examined the use of anti-VEGF therapy in treating patients with brain metastases. The presence of CNS metastases was a stated exclusion criterion in 76% of trials. The rate of intracranial bleeding was negligible. We conclude that there is no trial evidence that anti-VEGF therapy confers an increased risk of intracranial bleeding, even in the presence of CNS metastases. Future trials of anti-VEGF therapy should not exclude patients with controlled CNS metastases at enrollment.

Vascular endothelial growth factor (VEGF) is an important mediator of new blood vessel formation (angiogenesis) in both physiological and pathological settings.¹ All solid tumors are dependent on pathological angiogenesis for growth to a clinically detectable size.^2–4 Anti-VEGF therapy with the monoclonal antibody bevacizumab and the tyrosine kinase inhibitors sorafenib and sunitinib has demonstrated clinical benefit in randomized clinical trials involving patients with a number of solid tumor types. All three agents have demonstrated efficacy as single agents compared with standard care in renal cell carcinoma^5–7 and, for bevacizumab, in combination with cytotoxic chemotherapy in breast, colorectal, and non-small-cell lung carcinoma (NSCLC).^8–12 The reason for the efficacy of anti-VEGF therapy in combination treatment regimens is unknown, but it has been suggested that such therapy “normalizes” disorganized tumor neovasculature, allowing better penetration of cytotoxic drugs.¹³

An increased risk of bleeding was noted with bevacizumab therapy in the earliest clinical studies reported.^5,11,12,14 The reason for this is unknown but probably relates to alterations in endothelial function. Two patterns of bleeding have been described. The first pattern is minor mucocutaneous bleeding; for example, grade 1 or 2 epistaxis occurred in approximately 30% of patients with advanced or recurrent NSCLC treated with bevacizumab in combination with carboplatin and paclitaxel. This compared to 6% of patients treated with carboplatin and paclitaxel alone.¹²

Second, in the same study, life-threatening bleeding (described as either hematemesis or hemoptysis) occurred in 6 of 66 patients (10%) treated with bevacizumab; 4 of these patients died as a result. All six patients had centrally located tumors near major blood vessels, five of six patients had tumor cavitation or necrosis, and four patients had squamous histology. In light of these data, patients with squamous histology have been excluded from subsequent clinical trials of bevacizumab in advanced NSCLC.

Central nervous system metastases are common in many tumor types such as breast, lung, and renal carcinomas and in melanoma.¹⁵ Bleeding into brain metastases is a devastating complication associated with significant morbidity and mortality. The risk of bleeding into CNS metastases varies with the histology of the tumor, with fewer than l% and 5% of lung and breast brain metastases, respectively, showing evidence of spontaneous bleeding, compared with significantly higher rates with thyroid cancer, melanoma (40%–50%), renal cell cancer (70%), and choriocarcinoma.

Patients with brain metastases are often excluded specifically from clinical trials of systemic therapy, partly because such patients generally have a poor prognosis and also because many chemotherapeutic agents are thought not to cross the blood-brain barrier.¹⁶ A further consideration that may have led to the exclusion of patients with brain metastases from trials of anti-VEGF therapy is the possibility of intracranial hemorrhage as a result of treatment. Patients are generally screened for CNS metastases before enrollment into trials either clinically or with radiological imaging. It is likely that clinical screening misses a small but significant number of patients who have asymptomatic brain metastases, but radiological imaging will diagnose all but very small brain metastases.

This review aims to evaluate clinical trials of anti-VEGF therapy in the treatment of cancer to address whether such treatment increases the risk of intracranial bleeding both in the presence and absence of CNS metastases. This has significant implications for the therapy of patients with common solid tumors, given the relatively high incidence of brain metastases and the increasing use of anti-VEGF therapy.

Materials and Methods

Data for this review were compiled using the PubMed databases and the Proceedings of the American Society of Clinical Oncology (ASCO) annual meetings to search for articles, abstracts, and presentations of clinical trials up to June 30, 2007.

We examined the published literature to assess whether there was a significant increase in bleeding from anti-VEGF therapies. Only articles reporting safety data and published in English-language peer-reviewed journals were considered. We searched using the generic and drug development names of the agents considered and limited the results to publications involving clinical trials. The search terms included axitinib (AG-013736), bevacizumab, cediranib (AZD2171), semaxanib (SU5416), sorafenib (BAY 43-9006), sunitinib (SU011248), vande-tanib (ZD6474), and vatalanib (PTK787/ZK222584).

Because of the potential confounding factor of increased bleeding risk with significant thrombocytopenia, we recorded the proportion of greater than or equal to grade 3 thrombocytopenia, as defined by the Common Terminolology Criteria for Adverse Events (CTCAE) version 3 as <50,000 platelets.¹⁷ We recorded the incidence of severe bleeding for each of the studies, recorded the reported incidence of CNS bleeding, and noted whether it was recorded that bleeding occurred in the presence of metastases.

Results

We identified 57 published trials reporting the use of anti-VEGF agents in cancer patients. The reports included patients receiving bevacizumab (22), sorafenib (12), sunitinib (5), and a variety of other agents (18) (Tables 1–3). Eleven of these papers were randomized studies comparing anti-VEGF therapies with placebo or interferon (Table 3). We also identified four presentations of data examining the effect of anti-VEGF therapies on brain metastases, two of which reported efficacy data (Table 4). We identified five papers or meeting abstracts on the use of anti-VEGF agents in patients with high-grade gliomas (Table 5).

Table 1.

Phase I and II studies of anti-VEGF agents that excluded brain metastases

Drug	Patient Number	Grade 3 Thrombocytopenia	Rate of CNS Bleeding	⩾Grade 3 Bleed Outside CNS
Bevacizumab29–38	669	13 (2%)	1 (0.2%)	30 (5%)
Sorafenib39–47	499	15 (3%)	1 (0.2%)	9 (2%)
Semaxanib48–54	99	8 (8%)	0	4 (4%)
Sunitinib55–57	197	12 (6%)	0	0
Vatalanib58, 59	70	0	0	0
Axitinib60	36	0	0	1 (3%)
Vandenitib61–63	141	1 (0.7%)	0	0
Total	1,711	49 (3%)	2 (<1%)	44 (3%)

Table 3.

Rates of CNS bleeding in randomized trials of anti-VEGF therapy

First Author	Tumor	Study Agent	Patient Number	Control	Brain Metastases	⩾Grade 3 Thrombocytopaenia		Rate of CNS Bleeding		⩾Grade 3 Bleeding Outside CNS
						Study	Control	Study	Contro	Study	Control
Kabbinavar74	CRC	Bevacizumab+ FL	104	FL	Excluded	0	0	0	0	3%	2%
Hurwitz8	CRC	Bevacizumab+ IFL	813	IFL	Excluded	0	0	0	0	3%	2%
Giantonio75	CRC	Bevacizumab+ FOLFOX4	829	FOLFOX, or Bevacizumab	Included	0	0	1 (<1%)	1 (<1%)	3%	2%
Hurwitz76	CRC	Bevacizumab+ FL	210	IFL	Excluded	0	0	0	0	7%	1%
Miller9	Breast	Bevacizumab	462	Capecitabine	Excluded	1%	1%	0	0	1%	1%
Demetri77	GIST	Sunitinib	312	Placebo	Included	5%	0	0	0	0	0
Johnson12	NSCLC	Bevacizumab+ CP	99	CP	Excluded	1	0	0	0	9%	0
Sandler10	NSCLC	Bevacizumab+ CP	878	CP	Excluded	0	0	1%	0	4%	1%
Yang5	RCC	Bevacizumab	116	Placebo	Excluded	0	0	0	0	0	0
Escudier6	RCC	Sorafenib	903	Placebo	Excluded	0	0	0	0	3%	2%
Motzer7	RCC	Sunitinib	750	Interferon-α	Excluded	8%	0	0	0	1%	0%

Abbreviations: CRC, colorectal cancer; FL, 5-fluorouracil and leucovorin; IFL, irinotecan, 5-fluorouracil, and leucovorin; FOLFOX4, 5-fluorouracil and oxaliplatin; GIST, gastrointestinal stromal tumor; NSCLC, non-small-cell lung cancer; CP, carboplatin and paclitaxel; RCC, renal cell carcimoma.

Table 4.

Trials of anti-VEGF therapy targeted at or including brain metastases

First Author	Agent	Cancer	Patient Number	Number of Patients with BrainMetastase	Grade 3 Thrombo cytopenia	Rate of CNS Bleeding	⩾Grade 3 Bleed Outside CNS
Tamaskar78	Sunitinib	Renal cell carcinoma	35	5	0	0	0
Heymach79	ZD6474 ± docetaxel	Lung	127	13	0	0	2 (2%)
Amaravadi80	Sorafenib with temozolamide	Melanoma	38	38	6 (16%)	0	0
Henderson81	Sorafenib	RCC	2,488	65	0	0	0

Table 5.

Studies of anti-VEGF agents in patients with high-grade gliomas

Drug	Patient Number	Grade 3 Thrombocytopenia	Rate of CNS Bleeding	⩾Grade 3 Bleed Outside CNS
Bevacizumab + irinotecan18	32	0	0	0
Bevacizumab + irinotecan19	22	2	0	0
Vatalinib20	55	0	0	0
Vatalanib + temozolomide or lomustine21	60	0	0	0
Cediranib22	30	0	0	0
Total	199	2 (1%)	0	0

Exclusion of Brain Metastases

Brain metastases were excluded in 35 out of 46 phase I and II studies (76%), and in 9 out of 11 randomized studies (82%). The two randomized studies that permitted brain metastases involved gastrointestinal stromal tumors (GIST) and colorectal cancer.

Bleeding Risk in Patients Treated for Systemic Cancers

Significant bleeding was reported in a very low proportion of phase I and II studies that excluded brain metastases (Table 1), with only two episodes of CNS bleeding reported in 1,755 patients treated with anti-VEGF therapy (<1%). In phase I and II studies that enrolled patients with intracerebral metastases, there was only one episode of intracranial bleeding in the 524 patients treated with anti-VEGF therapy (<1%). Rates of significant hemorrhage outside the CNS were also very low in studies that did or did not exclude cerebral metastases, with systemic bleeding rates of 3% and 2%, respectively.

We identified 11 phase III trials using anti-VEGF therapies in a variety of clinical contexts and using a variety of agents (Table 3). Eight studies examined the use of bevacizumab in 3,511 patients with breast, colorectal, lung, renal, and pancreatic cancers. Seven studies involved concurrent chemotherapy. Two studies examined sunitinib in a total of 1,062 patients with GIST or renal cancer, and 1 study examined sorafenib in 930 patients with renal cancer. Overall, 5,476 patients were enrolled in the 11 studies.

Grade 3 thrombocytopenia was recorded in a minority of patients, with the highest figure being 8% in the renal cell study comparing sunitinib and interferon-α. Only five episodes of CNS bleeding were recorded in all the phase III studies, representing <0.1% of patients. Three episodes were in the lung cancer trial, in which CNS metastases were excluded, and all occurred in the study arm (bevacizumab, carboplatin, and paclitaxel). In the study by Giantonio et al.⁷⁵ (folinic acid, fluorouracil, and oxaliplatin [FOLFOX] with or without bevacizumab or bevacizumab alone in colorectal cancer), there was one CNS bleed in each of the bevacizumab-containing arms. This study did not exclude brain metastases.

Rates of significant bleeding outside the CNS in these studies were significantly higher in the Hurwitz et al.,⁸ Johnson et al.,¹² and Sandler et al.¹⁰ studies, with subgroups defined retrospectively as being of higher risks, such as patients with squamous NSCLC, cavitating lesions, or lesions eroding blood vessels.

Bleeding Risk in Patients with Known Brain Metastases

We identified four studies of anti-VGFR therapy specifically designed to include or treat patients with active brain metastases (Table 4). In total, 2,688 patients were enrolled in these studies, of whom 121 had brain metastases, and there were no episodes of intracranial hemorrhage.

Bleeding Risk in Patients Treated for High-Grade Gliomas

Only two studies of bevacizumab have been published examining the use of anti-VEGF agents in high-grade gliomas.^18,19 In addition, at least two abstracts have presented data on the use of vatalanib in glioblastoma multiforme, as well as one abstract on cediranib.^20–22 Table 5 shows data from these three studies of anti-VEGF therapy used in the treatment of high-grade gliomas. There were no reported instances of intracerebral bleeding or grade 3 or 4 bleeding at other sites.

Discussion

New treatment strategies for cancer are evolving. In particular, increasing understanding of tumor angiogenesis has allowed the development of novel antiangiogenics targeting a variety of molecular mechanisms. Anti-VEGF agents such as bevacizumab, sunitinib, and sorafenib have become standard-of-care treatments in a range of cancers.^5,7,23

The prevalence of CNS metastases in patients with metastatic cancers is 7% in breast cancer, 17% in kidney cancer, 16% in lung cancer, 5% in colon cancer, and up to 55% in melanoma. Autopsy series indicate the presence of brain metastases in 25% of cases.²⁴ Surgical excision should be considered for resectable lesions, and radiotherapy may control the disease for a limited period of time.^25–27 Chemotherapy has only modest efficacy in disease control, and response rates are likely to be considered sufficiently low for alternative, low-toxicity targeted agents to be an attractive therapeutic option.

An infrequent but concerning complication of anti-VEGF therapy is hemorrhage at the tumor site or at distant sites. The mechanism by which bleeding may occur is not well understood. Most likely the mechanism involves localized tumor necrosis at a critical site, in association with diminished regeneration of endothelial cells and increased vascular fragility as a result of VEGF blockade.² The risk of bleeding may be compounded by thrombocytopenia, generally associated with the concurrent use of myelosuppressive cytotoxic agents.

We were interested to review the bleeding risk associated with anti-VEGF therapy with particular reference to intracerebral haemorrhage. We examined all phase I, II, and III studies reporting anti-VEGF therapy and reviewed the incidence of hemorrhage. Given that one might expect an increased risk of bleeding in patients with cerebral metastases or high-grade gliomas, we specifically assessed toxicity data in these patient groups.

Our review of the current literature demonstrates three things. First, in the setting of exclusion of brain metastases, clinically significant bleeding in the CNS is very rare and probably not above baseline. Clinically significant thrombocytopenia also seems very rare with these agents, even in combination with chemotherapy. Second, even in the presence of known brain metastases, anti-VEGF therapy appears to be safe, with no recorded episodes of intracerebral hemorrhage. Third, based on the limited data currently available, treatment of high-grade gliomas with anti-VEGF therapy is not associated with an increased risk of intracerebral bleeding.

Our study is limited by a number of factors. First, a comprehensive review of intracerebral bleeding events is not the primary focus of the studies analyzed. For example, details of bleeding site, treatment, outcome, and the presence of potentially contributing intercurrent hypertension are rarely reported in the published studies. Second, a source of uncertainty is the process of exclusion of brain metastases, and this is often not specified within studies, and, when specified, it is not uniform among studies. One might expect a significant discrepancy in detection of brain metastases, depending on whether they are screened for by history, CT scan, or MRI scan. Third, the nature of this retrospective analysis and the relatively small numbers of patients treated with metastases or gliomas may not reflect real clinical outcomes.

Of interest, in other settings of potential high risk for bleeding, for example, in the setting of venous thromboembolism in patients with high-grade glioma, treatment with anticoagulants causes no increase in the rate of bleeding. In a retrospective comparison, Ruff found no difference in the rate of intracranial bleeding in 103 anticoagulated patients (1.9%), compared with 272 glioma patients not on anticoagulation (2.2%).²⁸

Our review supports the inclusion of patients with brain metastases and high-grade gliomas in anti-VEGF therapy trials. Clinicians should carefully discuss the risk-benefit ratio in such situations, recognizing there may still be some doubts about the safety because of the small numbers of patients with known brain metastases and gliomas treated with such agents.

Table 2.

Phase I and II studies of anti-VEGF agents that included brain metastases

Drug	Patient Number	Grade 3 Thrombocytopenia	Rate of CNS Bleeding	⩾Grade 3 Bleed Outside CNS
Bevacizumab64–67	187	2 (1%)	0	7 (4%)
Sorafenib23,68	174	0	1 (1%)	0
Semaxanib69–73	163	1 (1%)	0	1 (1%)
Sunitinib	0	0	0	0
Vatalanib	0	0	0	0
Total	524	3 (<1%)	1 (<1%)	8 (2%)