Epidemiology, biology, and management of venous thromboembolism in gliomas: An interdisciplinary review

Abstract

Patients with diffuse glioma are at high risk of developing venous thromboembolism (VTE) over the course of the disease, with up to 30% incidence in patients with glioblastoma (GBM) and a lower but nonnegligible risk in lower-grade gliomas. Recent and ongoing efforts to identify clinical and laboratory biomarkers of patients at increased risk offer promise, but to date, there is no proven role for prophylaxis outside of the perioperative period. Emerging data suggest a higher risk of VTE in patients with isocitrate dehydrogenase (IDH) wild-type glioma and the potential mechanistic role of IDH mutation in the suppression of production of the procoagulants tissue factor and podoplanin. According to published guidelines, therapeutic anticoagulation with low molecular weight heparin (LMWH) or alternatively, direct oral anticoagulants (DOACs) in patients without increased risk of gastrointestinal or genitourinary bleeding is recommended for VTE treatment. Due to the elevated risk of intracranial hemorrhage (ICH) in GBM, anticoagulation treatment remains challenging and at times fraught. There are conflicting data on the risk of ICH with LMWH in patients with glioma; small retrospective studies suggest DOACs may convey lower ICH risk than LMWH. Investigational anticoagulants that prevent thrombosis without impairing hemostasis, such as factor XI inhibitors, may carry a better therapeutic index and are expected to enter clinical trials for cancer-associated thrombosis.

In 1823, the French physician Jean-Baptiste Bouillard was the first to report an association between malignancy and venous thrombosis.¹ Among systemic tumors, carcinomas of the pancreas, stomach, ovary, and lung, as well as hematologic malignancies, are at highest risk; melanoma, breast, and prostate cancer have much lower incidence.^1,2 Remarkably, only in the last 50 years has the remarkably high incidence of brain tumors, and in particular, glioblastoma, come to attention. The year 1975 witnessed the publication of an autopsy series demonstrating 31% of patients with glioblastoma had venous thromboses, compared to 17% of autopsies on non-malignancy controls.³ Several years later, Ruff and Posner reported that 29% of patients with high-grade glioma had confirmed venous thrombosis following diagnosis.⁴ In this manuscript, we focus on venous thrombosis in adult patients with diffuse gliomas, and in particular glioblastoma because of the strength of the association. We review epidemiology, our understanding of the biological underpinnings of the association, potential biomarkers, and attempts to stratify risk and prevention, and management.

Epidemiology and Clinical Risk Factors

Venous thromboembolism (VTE) is a common complication of glioblastoma (GBM), isocitrate dehydrogenase (IDH)wild-type with an estimated incidence of 20%–30%.^5–8 The risk is greatest during the first few months after the diagnostic surgery and remains higher than in other malignancies throughout the disease course.^9,10 In a large prospective multicenter study including 170 patients with newly diagnosed GBM, symptomatic VTE occurred in 24% of patients, particularly in the first 6 months after diagnosis.⁸ Patients with lower-grade glioma (LGG; WHO grade 2 and 3 gliomas) also have a higher VTE risk compared to the general population. A recent retrospective analysis of 256 patients with LGG demonstrated an overall cumulative incidence (CI) of VTE of 8.2% in patients with grade 2 and 9.2% in grade 3 gliomas,⁵ findings corroborated in a prospective observational study.¹¹ These findings suggest that even LGG are associated with a hypercoagulable state.^5,12

With the shift of glioma classification from purely histologic to a molecular paradigm, as reflected in the 2021 WHO classification, understanding the risk of VTE based on molecular glioma subtypes is also important. In a large retrospective study of 635 patients with glioma, the absence of IDH mutation in grade 2–4 glioma was associated with a 3-fold increase in VTE rate compared to patients with IDH mutation (hazard radio 3.06, 95% confidence interval 2.03–6.46).⁵ Among patients with GBM, there was no difference in VTE incidence according to O6-methyltransferase (MGMT) promoter methylation status and epidermal growth factor receptor (EGFR) overexpression.⁵ In contrast, a recent study in adult-type diffuse gliomas found that MGMT promoter methylation was associated with reduced VTE risk, independent of glioma molecular subtype.¹¹ The relationship between glioma type and VTE risk is discussed further in the next section. The impact of VTE development upon survival is uncertain: While one large cancer registry retrospective review of 9489 gliomas (of which almost 6000 were grade 3 and 4) showed a 30% increase in the 2-year risk of death, other retrospective¹³ and prospective^8,14 studies in grade 3 and 4 gliomas have shown no difference in survival according to VTE presence.

Multiple clinical risk factors for VTE with glioma have been described, principally based on retrospective studies. Some of these characteristics are not restricted to gliomas and rather represent well-described risk factors for VTE in general; for example, several studies have demonstrated a higher risk of VTE in high-grade glioma patients with prior history of VTE.^13,15 Increasing age seems to be related to VTE risk in a continuous fashion, with 2 separate studies finding a 3% increase in risk for each additional year of age.^16,17 Immobility is another general thrombosis risk factor that has been reproduced in the glioma population, who are at a particularly high risk of this complication due to the potential for neurological impairment from the tumor and its treatments. The presence of limb paresis is a classically described risk factor^4,18,19 that has been replicated in the prospective setting;²⁰ some of these studies have quantified the risk increase as over 2- to 3-fold of that in patients without motor impairment,^18,20 and others have also reported a higher likelihood of thrombosis developing in the paretic limb.^4,18 Perhaps as a surrogate of decreased mobility, some studies have also found poorer performance status to correlate with the development of VTE,^17,21 including one that controlled for the presence of hemiparesis.²¹ On a related note, longer surgical length has been associated with a higher risk of VTE in brain tumor resections in general,²² and the incidence of VTE in glioma peaks in the first 2 months following diagnosis.⁶

Non-O blood groups are also a recognized risk factor for thrombosis in the general population, likely mediated by increased levels of von Willebrand factor and factor VIII in these groups.^23,24 In gliomas, the evidence is inconsistent: Two retrospective studies have shown an increase in risk for patients with non-O blood types—groups A and AB in one study (B not reported),¹⁶ and B only in the other (A and AB not significant).²⁵ However, another prospective study of 107 high-grade glioma patients did not find any difference in VTE incidence according to ABO type.⁸ Similarly, increased platelet counts are associated with VTE in cancer patients,²⁶ but as is discussed later lower platelet counts were associated with a higher risk in a cohort of high-grade glioma patients.²⁷

In addition, there are some risk factors specific to gliomas, including tumor grade, size and location, and extent of surgery. The reported risk of VTE is consistently higher in patients with GBM compared to other tumor grades,^5,6,20 and one study found double the risk for those patients with tumors >5 cm.¹⁶ Moreover, greater volume of residual tumor may explain why patients who undergo a subtotal resection or biopsy appear to be at a higher risk of VTE than those who have their tumors totally resected, even after controlling for the confounding factor of hemiparesis.¹⁴ These observations could potentially be explained by a larger production of procoagulant substances—such as tissue factor (TF) or podoplanin (PDPN)—by larger, more aggressive tumors.^28,29 The risk may also be higher in tumors in a supratentorial location, for unclear reasons.³⁰

Lastly, treatment-related factors must also be considered. Besides the surgical factors already mentioned, the use of certain medications has been associated with a higher risk of developing VTE. An association between the use of chemotherapy and risk of VTE was reported in one relatively small retrospective study,¹⁸ but whether this continues to apply to more modern regimens after controlling for potential confounders—such as tumor grade or performance status—is unclear. On the other hand, the use of steroids has been reported as an independent risk factor in a multivariate analysis that included presence of paresis, Karnofsky performance status, and ambulatory status.³¹ There are conflicting data on whether the antiangiogenic agent bevacizumab, which is widely used in glioma management and is clearly associated with an increased risk of arterial thromboembolic events, increases the risk of VTE.^32,33 While instances of VTE have also been reported in clinical trials investigating bevacizumab in gliomas,^34,35 it is possible that this does not translate into a meaningful increase in the already elevated risk of VTE in GBM: After analyzing almost 1500 newly diagnosed GBM patients from 3 clinical trials, a meta-analysis found no significant difference in the risk of deep venous thrombosis (DVT) and pulmonary embolism (PE) in patients treated with bevacizumab compared to those in control arms, although there was a trend (p = .07) towards an increased risk of PE only.³⁶

Table 1 summarizes the most important risk factors for VTE that have been identified in glioma patients. Most of the studies included are retrospective, and there is great heterogeneity in patient selection, follow-up interval, and standard of care practices (for example, data from older studies predates the routine use of postoperative prophylactic heparin); therefore, while measures of association are provided when available, these are only for reference and cannot be directly compared. It is also important to note that, as previously mentioned, many of these risk factors may be interrelated (such as presence of paresis or larger tumor size with lower performance status), and to what extent each individual factor truly contributes to VTE risk is unclear; studies that controlled for potential confounders, suggesting a more definitive risk factor, are also summarized in Table 1. In a prospective observational study of VTE in 258 adult-type diffuse glioma patients, multivariable time-to-event analyses identified 7 factors associated with increased VTE risk (1) prior history VTE, (2) hypertension, (3) asthma, (4) white blood cell count, (5) WHO tumor grade, (6) patient age, and (7) body mass index.¹¹ Three were associated with decreased risk: (1) IDHmut, (2) hypothyroidism, and (3) MGMT promoter methylation. These intriguing results warrant further evaluation and prospective validation.

Table 1.

Risk Factors for VTE in Gliomas

General VTE risk factors	Prior history of VTE	OR = 4.3–7.113,15,a
	Increasing age	HR = 1.03 per additional year16,17,a
	Limb paresis	OR = 5.1–6.018,20,a
	Poor performance status	HR = 2.017a –3.021,b
	Non-O blood groups*	HR = 2.7 (group A)–9.4 (group AB)16,a, OR = 6.9 (group B)25,a
Factors specific to gliomas	Tumor grade (GBM vs. others)	OR = 2.0–6.85,6,20,a
	Tumor size >5 cm*	HR = 2.216,a
	Supratentorial location*	Risk difference = 6.3%30,**
	Lower platelet count	HR = 0.727,c
Treatment-related factors	Subtotal resection/biopsy	HR = 3.6 26,d
	Duration of surgery >4 h*	OR = 2.022,e
	Chemotherapy*	OR = 7.318,a
	Steroid use	HR = 2.231,f
	Bevacizumab	RR = 1.516,g, OR for PE = 5.136,h

HR = hazard ratio, OR = odds ratio.

^*Unconfirmed factor (reported in a single retrospective study, or there is conflicting evidence regarding this factor).

^**Incidence of VTE was 0% in infratentorial location, so other measures of association cannot be calculated.

^aUncontrolled for competing risk factors (or not explicitly mentioned).

^bControlled for age, sex, body mass index, hemiparesis, platelet count, D-dimer, and molecular characteristics (mutational status of IDH1, PTEN, P53, BRAF, TERT promoter; MGMT promoter methylation status, and EGFR amplification status).

^cControlled for type of neurosurgical intervention (resection vs. biopsy) and histological type of HGG.

^dControlled for age, sex, presence of paresis, and GBM histology; borderline statistical significance (P = .054).

^eControlled for age, functional status, impaired sensorium, presence of disseminated cancer, and postoperative complications (pneumonia, urinary tract infection, stroke, unplanned intubation, intubation ≥48 hours, coma >24 hours, sepsis, and septic shock).

^fControlled for age, sex, history of VTE, performance status, presence of paresis, ambulatory status, body mass index, white blood cell and absolute neutrophil count at start of adjuvant temozolomide, use of antihypertensives and use of gastric reflux medications.

^gRandomized controlled trial, randomization stratified by MGMT methylation status and other molecular characteristics.

Glioma Molecular Biology and its Relationship to VTE

As explained above, patients with IDH mutant (IDHmut) gliomas (which includes IDHmut astrocytoma grades 2–4 and 1p/19q codeleted oligodendrogliomas grades 2–3 in the current WHO classification) are at lower risk of VTE than IDHwt GBM.¹¹ Recent studies suggest a plausible mechanistic explanation. Within IDHmut gliomas, mutant IDH1 or IDH2 reduce α-ketoglutarate to D-2-hydroxyglutarate (D2HG) and consume NADPH instead of producing it.³⁷ D2HG inhibits enzymes that normally require α-ketoglutarate as a cosubstrate, including dioxygenases that demethylate DNA and histones.^38,39 Thus, D2HG-producing IDHmut tumors have more DNA and histone methylation compared to their IDHwt counterparts.^40–43 Importantly, TF and PDPN expression are higher in tumors with wild-type IDH1 compared to those containing a mutated IDH1.⁴⁴ Indeed, TF and PDPN were in the top 50 most differentially methylated and downregulated genes in proneural tumors.⁴¹ TF is a transmembrane protein that is expressed by many kinds of cancer and can initiate the clotting cascade by binding factor VII/VIIa.⁴⁴ PDPN is a sialomucin-like transmembrane glycoprotein that binds to C-type lectin receptor type 2 (CLEC-2) on platelets and induces aggregation. Both TF and PDPN are released from tumor cells in the form of extracellular vesicles (EVs), which are small membrane vesicles.⁴⁵ High levels of circulating TF and/or PDPN may increase VTE risk in a variety of cancers, possibly including gliomas (Figure 1),^{29,44,46–55} and reduction of TF production from IDHmut-mediated gene methylation may contribute to decreased VTE risk with IDHmut gliomas (Figure 2). While strong clinical association between expression of PDPN and VTE risk has been reported,⁴⁶ the relative importance of TF versus PDPN in the context of gliomas has not yet been established and their cooperation cannot be ruled out.⁴⁴ In one study that directly compared circulating preoperative levels of TF and PDPN with postoperative VTE, elevated TF showed a link with increased VTE risk, but PDPN did not.¹¹

Figure 1.

Pathways involved in venous thrombosis in glioma patients. Glioma cells, particularly glioblastoma, express podoplanin and tissue factor (TF) and release these membrane proteins on extracellular vesicles. Podoplanin activates platelets and TF activates the coagulation cascade, leading to venous thrombosis. This figure is created with Biorender.com.

Figure 2.

Effect of IDH mutational status on tissue factor (TF) expression

Prior to the 2021 WHO classification, the term “GBM” referred to both IDHwt gliomas and IDHmut astrocytomas that had necrosis and/or microvascular proliferation. Within that framework, 3 main RNA expression patterns categorized GBMs: Classical, mesenchymal, and proneural.⁵⁶ TF expression was higher in classical and mesenchymal GBMs than in proneural GBMs.⁵⁷ Moreover, IDHmut GBMs were nearly all within the proneural subset, as opposed to the classical or mesenchymal subsets.⁵⁶ In the current classification, wherein “GBM” only refers to IDHwt gliomas meeting the aforementioned histologic or specific molecular criteria, those IDHmut proneural GBMs are now regarded as central nervous system WHO grade 4 IDHmut astrocytomas. Thus, the association between high TF and classical/mesenchymal GBM subtypes is likely due, at least in part, to the lack of IDHmut astrocytomas in those groups. Among IDHwt GBMs, the presence of EGFR amplification, and/or constitutively active EGFRvIII (both of which are mutually exclusive with IDH mutation) are enriched in classical expression patterns and have been shown to directly upregulate TF expression.^49,57–59 Consonant with this is the more recent finding that EGFR amplification has been associated with increased VTE risk in IDHwt GBMs in some²¹ but not all^5,11 studies. A more granular analysis of GBM, however, may be required, as at the single-cell level the expression of coagulation effectors (PDPN, TF) is highly heterogeneous within individual lesions across GBM subgroups.⁴⁴ Thus, as in other cancers, oncogenic,⁶⁰ epigenetic,⁶¹ and microenvironmental factors⁶² likely interact to drive GBM-associated VTE risk.

Circulating and Tumoral Biomarkers of VTE Risk

Many studies have attempted to identify circulating biomarkers that predict VTE in glioma. The most comprehensive study²⁷ examined 141 high-grade glioma patients, 24 (17%) of whom developed a VTE in the 12-month follow-up period. Circulating biomarkers that were associated with VTE and/or were used in a risk assessment model²⁷ as well as important tumoral biomarkers are listed in Table 2. The best risk assessment model included soluble P-selectin (≥ 75 percentile) and platelet count (<25 percentile of the study population). The cumulative VTE probability was 9.7% for score 0, 18.8% for score 1, and 83.3% for score 2. These results suggested that glioma-activated platelets led to an increase in soluble P-selectin and a decrease in platelet count.

Table 2.

Circulating and Tumoral Biomarkers that are Associated With Elevated Risk of VTE in Brain Cancer Patients

Circulating Biomarkers

• High soluble P-selectin levels

• High D-dimer levels

• High white blood cell count (most still within the normal range)

• Low platelet count (most still within the normal range)

• Elevated coagulation factor VIII activity

Tumoral biomarkers

• Brain tumor histological subtype (highest risk in glioblastoma)

• Podoplanin expression

• IDH1 mutation status (high risk in patients with wild-type IDH1)

A subsequent study tested the hypothesis that primary brain tumors express PDPN.⁴⁶ PDPN/CLEC-2 interactions were also found to be responsible for intratumoral platelet thrombi formation in a murine GBM model.⁶³ PDPN expression was evaluated in 213 patients with brain tumors of whom 29 (13.6%) developed VTE in the 2-year follow-up period. Importantly, high PDPN expression in the tumor was associated with increased risk of VTE. In addition, PDPN expression was associated with a lower platelet count.

How might PDPN expression by brain tumors cause VTE? The most likely answer is that brain tumors release PDPN-positive EVs that circulate in the blood and trigger VTE. One study showed that GBM cells released EV-associated, bioactive PDPN and that patients had elevated levels of PDPN in plasma.⁴⁴ However, it should be noted that no experiments were performed to show that PDPN in human plasma was associated with EVs. Mice bearing human glioma xenografts expressing PDPN had increased levels of circulating PDPN, increased levels of the platelet activation marker PF4, and a decreased platelet count.⁴⁴ Similar effects were induced in mice upon intravascular injection of PDPN-carrying EVs from glioma cells, but not TF-carrying EVs.⁴⁴ Another study found that mice bearing a PDPN-expressing human melanoma tumor had increased venous thrombosis that was reduced by inhibition of PDPN.⁶⁴ Taken together, these studies suggest that one major pathway by which brain tumors increase the risk of VTE is via expression of PDPN and release of PDPN-positive EVs into the circulation with the subsequent activation of platelets via CLEC-2.

As previously discussed, another protein that might contribute to VTE in brain cancer is TF, a highly procoagulant transmembrane protein that is released from brain tumors on EVs.⁶⁵ Unruh and colleagues⁶¹ found that GBM patients with wild-type IDH1 had a higher rate of VTE (30/117 patients, 26%) compared to patients with a mutant IDH1 (0/45 patients, 0%). This is expected from the high levels of both TF and PDPN in tumors with wild-type IDH1. Importantly, GBM patients with tumors expressing wild-type IDH1 had significantly higher levels of TF expression in the tumor compared to GBM patients with tumors expressing a mutant form of IDH1. Interestingly, TF expression in tumor specimens in malignant brain tumor patients, most of whom had GBM, was not associated with VTE in 96 patients (15 of whom had a VTE).⁶⁶ Other studies have measured levels of circulating EV-TF activity in brain tumor patients. As expected, a significantly higher level of EV-TF activity was observed in GBM patients with wild-type IDH1 compared with GBM patients with a mutant form of IDH1.⁶¹ In a prospective cohort study of 119 patients with unspecified malignant brain tumors, 19 (15.9%) of whom developed VTE,⁵⁴ no association between EV-TF activity and VTE was identified. However, it is possible that brain tumor patients at highest risk of VTE will have both PDPN-positive and TF-positive EVs.

In summary, a small number of studies have measured biomarkers that are associated with VTE in patients with glioma. However, further studies are needed to confirm and extend these initial studies and determine if any of the biomarkers will be clinically useful in identifying patients with gliomas who are at high risk of VTE.

Diagnosis of Venous Thromboembolism in Patients with Glioma

The high incidence of VTE in patients with glioma warrants consideration of this entity with each clinical interaction. Classic symptoms of DVT include leg pain, swelling, redness, warmth, or engorged superficial veins.⁶⁷ Based on a simplified clinical model, predictive factors for DVT include: Presence of active cancer, paralysis/paresis or recently bedridden for 3 days or more or major surgery in the previous 12 weeks, localized tenderness along the distribution of deep venous system, entire leg swelling, calf swelling at least 3 cm larger that the asymptomatic leg, pitting edema confined to the symptomatic leg, collateral superficial veins, and previously documented DVT.⁶⁸ However, patients with glioma often develop limb paresis or plegia and receive steroid therapy producing leg swelling,³¹ confounding the clinical picture of DVT. In a large prospective multicenter study that sought to identify the risk factors for symptomatic VTE in newly diagnosed patients with grades 3 and 4 gliomas, 26 of 107 (24%) developed symptoms of VTE in the first 6 months after initial diagnosis.⁸ Leg swelling or cramping, dyspnea, or fatigue were mentioned as indications for VTE testing, although the specific VTE signs and symptoms and the corresponding frequencies were not reported in the study.⁸

PE represents a life-threatening form of VTE. Typical clinical signs of PE include dyspnea, chest pain, tachycardia, apprehension, tachypnea, syncope, and hypoxia.⁶⁷ The Wells score is commonly used as a bedside evaluation to assess the likelihood of PE. It incorporates 7 clinical features: Clinical symptoms of DVT, heart rate >100, immobilization 3 or more days or surgery in previous 4 weeks, previous DVT/PE, hemoptysis, malignancy, and likelihood of alternative diagnosis.⁶⁹ The utility of the Wells score in patients with cancer is unclear, as only a minority of patients with cancer were included in the study.^67,69 In patients with extra-CNS cancer, the Khorana scale is most often utilized for VTE risk assessment. The score assigns points to 5 clinical and pre-chemotherapy laboratory parameters: Site of cancer, platelet count of 350 × 10⁹/L or more, hemoglobin concentration of 100 g/L or lower or use of erythropoiesis-stimulating agents, leukocyte count of 11 × 10⁹/L or higher, and body mass index of 35 kg/m² or higher.^70–72 A total score of 3 or more is considered high-risk, 1–2 intermediate-risk, and 0 low risk for VTE. However, the Khorana scale was undiscriminating as a VTE predictor in patients with GBM.³¹ Perhaps this is due to the fact that all patients with GBM were assigned to the very high-risk category cancer site (2 points) and only required 1 additional risk factor to reach a total high-risk score of 3 or more.³¹ Moreover, as noted earlier, paradoxically a low platelet count is associated with VTE in GBM. Based on a population-based study assessing the incidence of symptomatic VTE in 115 patients with GBM, a simple scoring system that includes Karnofsky performance status, age, smoking, and hypertension has been proposed.⁷³ With a maximum score of 6, patients scoring higher than 3 points were 5 times more likely to develop VTE than those scoring below. Finally, the 10 risk factors identified (Supplementary table 1) in a prospective study¹¹ have been integrated into a web-based app that calculates estimated VTE risk in patients with newly diagnosed adult-type diffuse glioma patients (https://kbellburdett.shinyapps.io/GliomaPredictVTE/). Prospective validation is necessary to determine the utility of such tools.

Venous ultrasound (US), which combines vein compression (B-mode imaging) and pulsed Doppler spectrum analysis, is the imaging of choice for evaluation of suspected DVT with a sensitivity and specificity of approximately 94% for proximal DVT.^67,74 When venous US evaluation is inconclusive, other imaging modalities such as conventional venography, computed tomography (CT), and magnetic resonance are alternatives.⁷⁵ CT pulmonary angiography is the diagnostic imaging modality of choice in diagnosing PE due to its sensitivity and specificity.^67,75 D-dimer, a nonspecific marker of ongoing thrombin generation and fibrinolysis, is typically elevated in patients with VTE. However, its specificity is low and it can be elevated in other acute or inflammatory conditions.⁶⁷ In patients with grade 3 glioma and GBM, D-dimer levels were similar between patients who did and did not develop symptomatic VTE.⁸

Perioperative Prophylaxis

VTE is a major cause of complication in patients undergoing craniotomy for GBM, with an incidence as high as 21% in the first 3 months after craniotomy.^13,76 GBM patients have one of the highest risks of perioperative VTE and, compared to patients undergoing a craniotomy for non-neoplastic disease, reported rates of postoperative VTE are twice as high.⁷⁷

In addition to the tumor and patient-related risk factors for VTE, the perioperative phase adds additional ones. These include venous stasis from perioperative immobility, endothelial injury, and the inflammatory response related to the craniotomy.⁷⁸ Perioperatively, most patients receive dexamethasone. Even though the precise effect of dexamethasone remains to be elucidated, a recent study found that preoperative steroid usage was associated with post-discharge DVT.⁷⁹ Other factors contributing to the risk of (perioperative) VTE are the size of the tumor (patients with tumors greater than 5 cm have a higher risk of VTE), extent of resection (subtotal resections resulting in a higher risk of VTE than gross total resections),¹⁴ longer operative times (predictive for VTE during hospitalization, but not for VTE occurring after discharge), and admission 2 or more days before craniotomy (predictive for DVT, but not for PE).⁷⁹

Given the increased risk of perioperative VTE, most patients who undergo surgical resection of a GBM receive prophylaxis perioperatively. Whereas heparin was previously the most used anticoagulant for treatment of VTE, since 1999 low molecular weight heparins are the most commonly used prophylaxis.⁸⁰ While data specific to GBM are lacking, guidelines suggest the use of pre and postoperative compression stockings and/or intermittent pneumatic compression, with the addition of low molecular weight heparin low molecular weight heparin (LMWH) postoperatively, for DVT prevention.⁸¹

Compression stockings that apply continuous circumferential pressure to the leg in a graduated fashion have been shown to reduce the risk of perioperative VTE in glioblastoma patients by increasing venous clearance and preventing venous stasis. Even though use of intermittent pneumatic compression (in combination with pharmacological prophylaxis) has been suggested to reduce VTE in high-risk patients⁸² other studies in GBM patients have not shown additional benefit of these devices, which can be cumbersome and require assembly once fitted.⁸³

There are few studies on the length of VTE prophylaxis, and there is considerable practice variation. Lengths of administration range from VTE prophylaxis throughout hospitalization, to prophylaxis until the patient is mobile or 7–10 days after surgery, to timing based on the neurosurgeon’s preference, or to prophylaxis based on the patient’s risk profile.

The risk of intracerebral hemorrhage (ICH), the most feared complication of VTE prophylaxis after a craniotomy, causes many neurosurgeons to favor a conservative strategy. The incidence of ICH is highest in the first days after craniotomy. In contrast, the increased risk of VTE experienced by glioblastoma patients extends beyond the period of hospitalization, especially for PE.⁷⁹

A study comparing a cohort of patients receiving VTE prophylaxis during hospitalization to a cohort of patients receiving VTE prophylaxis up to 21 days, showed that prolonging prophylaxis beyond discharge did not reduce the risk of VTE, but was associated with an increased risk of ICH.⁸⁴ Therefore, there is currently insufficient evidence to justify prolongation of prophylaxis beyond the length of stay in the hospital, and perioperative (prophylactic) management of VTE remains a matter of delicate balance in the perioperative period.

Primary Prophylaxis of Venous Thromboembolism in GBM Beyond the Perioperative Period

As summarized above, the incidence of perioperative DVT and PE is high after surgery for high-grade gliomas and is mitigated with the use of both mechanical and pharmacologic prophylaxis. Beyond the perioperative period, the incidence of VTE in these patients continues to be high, and is estimated to be about 1.5%–2.0% per month of survival.⁹ Risk factors such as muscle weakness, disease recurrence, mobility challenges, and prothrombotic therapies likely play a role in the continued risk of clotting. In a cohort study of over 1400 GBM patients, 15.7% had a diagnosis of a VTE event, half of which occurred in the first 1–2 months after surgery.¹⁵

In the antithrombotic era just prior to the introduction of direct oral anticoagulants (DOACs) a large primary prophylaxis trial randomized patients to LMWH versus placebo daily for up to a year.⁸⁵ The study was terminated early for logistical reasons. Accrual was slower than anticipated, likely due to the placebo injections. There was a trend toward reduction in objectively confirmed DVT/PE but also a trend toward increased bleeding in the LMWH-treated group.

Since the emergence of DOACs, with their oral administration and good safety profiles, several clinical trials have tested the use of pharmacological prophylaxis in ambulatory patients with cancer types at high risk of VTE; however, few included many, if any, patients with glioblastoma. A post hoc analysis of the AVERT trial studied 24 patients with high-grade glioma, out of the 574 cancer patients in the overall study.⁸⁶ The numbers are very small but there were no intracranial bleeding events in patients randomized to apixaban. A further truncated n = 10 trial of apixaban in newly diagnosed patients used apixaban 2.5 mg twice daily starting 2–21 days after craniotomy and continued for up to 6 months.⁸⁷ There were no bleeding or VTE events in this study.

A Cochrane review found that in ambulatory cancer patients, primary prophylaxis with DOACs may reduce the incidence of VTE but probably also increases the risk of bleeding.⁸⁸ Given these findings, the routine use of DOACs such as apixaban is not recommended for ambulatory patients with brain tumors.

Going forward, there is a need for disease-specific randomized trials testing the newer oral anticoagulants versus placebo in newly diagnosed GBM. DOACs appear relatively safe and it should be relatively easy to accrue to such studies. Ideally, these need to be targeted to patients at higher risk of VTE. Factors such as IDH mutational status, circulating biomarkers associated with thrombotic risk, and clinical factors such as performance status and clinical thrombosis scores can be used for study enrichment in patients with glioma.

Treatment of Established Venous Thromboembolism

Heparin and Low Molecular Weight Heparin

Initial treatment of VTE includes LMWH, DOACs, or unfractionated heparin (UFH). In patients with primary or metastatic brain tumors, current guidelines recommend anticoagulation as for other patients with cancer, although uncertainty remains as to the choice of agent.^89,90 Current clinical practice guidelines recommend LMWH for the initial treatment of VTE in patients with cancer, while DOACs are alternative initial anticoagulants in those who do not have a high risk for gastrointestinal or genitourinary bleeding.^89,90 For patients initiating treatment with parenteral anticoagulation, LMWH is preferred over UFH for the initial 5 to 10 days in patients without renal impairment.⁹⁰ UFH is generally reserved for high-risk patients, those with symptomatic pulmonary embolism, and patients with renal impairment due to its shorter half-life and reversibility with protamine.¹⁰

For long-term treatment, LMWH is preferred over warfarin due to its increased effectiveness in reducing the risk of recurrent VTE, lack of significant drug interaction, and convenience of not needing frequent blood monitoring.¹⁰ Randomized studies directly comparing LMWH to warfarin specifically for patients with brain tumors are lacking. In the CLOT trial, which randomized patients with cancer to LMWH or warfarin, LMWH was more effective in reducing recurrent VTE risk without increasing the risk of bleeding compared to warfarin. However, this study population of 673 cancer patients included only 34 patients with primary brain tumors and an unknown number of patients with brain metastases.⁹¹

The principal concern with anticoagulation in patients with primary or secondary brain tumors is the potential increased risk of ICH. In the CLOT trial, ICH was reported in 1 of the 14 patients with primary brain tumors who received dalteparin. However, this trial was not designed to assess the risk of ICH.⁹¹ Recent retrospective data on the risk of ICH in patients with glioma on therapeutic anticoagulation show conflicting results. A large retrospective matched cohort study of 220 patients with grade 3 and 4 gliomas reported a higher, but not statistically significant, 1-year CI of ICH of 17% in patients with VTE treated with LMWH versus 9% in those who were not treated with anticoagulation (p = .36).⁹² Among patients without VTE, the 1-year CI of ICH was 13%.⁹² Nor did anticoagulation with a variety of agents for atrial fibrillation (AF) in patients with GBM significantly increase the risk of ICH compared to patients with GBM without AF or patients without GBM anticoagulated for AF.⁹³ Moreover, another retrospective study including 133 patients with grade 3 and 4 gliomas also did not find a significant difference in the 1-year CI of any and measurable ICH in patients who received LMWH for VTE compared to patients who did not have VTE.⁹⁴ However, this study reported a 3-fold increased risk of major ICH in patients treated with LMWH (14.7% vs. 2.5%, p = .036).⁹⁴ Other retrospective studies likewise reported a significantly higher incidence of ICH among patients with VTE treated with either heparin, LMWH, or UFH compared to those who were not on anticoagulation (16% vs. 2%).^95,96 ICH risk did not significantly differ with age, hypertension, thrombocytopenia, concurrent use of aspirin, and treatment with antiangiogenic agents.^93,94 Contradictory data of ICH risk in patients with glioma could be due to the lack of standardized definition of ICH risk (measurable, symptomatic, or major), retrospective nature of all the studies, and the small number of patients and events in these studies.⁹²

Oral Anticoagulants

Since the first approval of a DOAC by the FDA in 2010, these agents have been increasingly preferred over other anticoagulants by both clinicians and patients in light of the oral administration and lack of need for monitoring of drug levels or clotting times. In adult patients with cancer-associated VTE, large randomized trials have established DOACs as non-inferior to LMWH without an increased risk of major bleeding.^97–100 However, patients with glioma and other primary brain tumors were severely underrepresented or excluded from these trials due to risk of ICH. Thus, the limited available data regarding use of DOACs in patients with glioma are derived from retrospective studies (some of which also included patients with brain metastasis, whose risk of ICH may be different).¹⁰¹

One study analyzed 172 patients with primary (n = 67) or metastatic brain tumors (n = 105) with VTE who were anticoagulated with either a DOAC or LMWH.¹⁰² In the primary brain tumor cohort, comprised entirely of patients with glioma, 20 patients received a DOAC and 47 patients received enoxaparin. The CI of any ICH at 12 months was 0% in patients receiving a DOAC compared with 36.8% in those receiving enoxaparin (P = .007). A second study included 125 patients with primary and metastatic brain tumors diagnosed with acute VTE (n = 104 patients with primary brain tumors; n = 78 gliomas).¹⁰³ When evaluating only the patients with primary brain tumors, rates of major bleeding were lower in the DOAC group at 4 of 44 patients (9.1%) compared with the LMWH group at 13 of 47 patients (28%) (P = .02). The rate of ICH also remained lower in the DOAC group at 2 of 44 patients (4.5%) than in the LWMH group at 7 of 47 patients (15%) (P = .10). A third study focused on patients with GBM with postoperative PE.¹⁰⁴ Of 46 patients with PE, 14 (30%) received a DOAC and 32 (70%) received LMWH. There was no significant difference in major ICH or recurrent VTE between the 2 cohorts. Most recently, a fourth study analyzed 121 patients with GBM who were diagnosed with acute VTE and treated with either a DOAC (n = 33) or LMWH (n = 88).¹⁰⁵ The incidence of clinically relevant ICH at 30 days was 0% in the DOAC group and 9% in the LMWH group (P = .11), and the CI of clinically relevant ICH at 6 months was 0% in the DOAC group and 24% in the LMWH group (P = .001), with 4 fatal ICH events in the LMWH group.

Taken together, these data, although retrospective and limited by relatively small sample sizes, suggest that DOACs are unlikely to be associated with higher risk of ICH compared to LMWH and represent a reasonable choice for patients with glioma and VTE who desire oral therapy. Selection of a DOAC should take into account the possibility of drug–drug interactions affecting DOAC metabolism. Rivaroxaban and, to a lesser extent, apixaban, are CYP3A4 substrates. These agents should be avoided in patients taking strong enzyme-inducing antiepileptic drugs including phenytoin, carbamazepine, and phenobarbital; caution should be taken with any DOAC in the setting of a strong hepatic microsomal enzyme inducer.¹⁰⁶ In the event of ICH associated with use of a DOAC, a reversal agent should be considered, such as andexanet alfa for ICH associated with a direct factor Xa inhibitor (eg, apixaban, edoxaban, or rivaroxaban). Oral activated charcoal to reduce gastrointestinal absorption may be given if the patient can take oral medications and the last dose of the anticoagulant was within 4 hours.¹⁰⁷

Warfarin remains a reasonable oral option for anticoagulation in select brain tumor patients with VTE. For example, in patients with severe renal insufficiency (creatinine clearance < 30mL/minute) who desire oral therapy, warfarin may be used in place of dose-adjusted LMWH or DOAC therapy. Studies of warfarin in patients with primary brain tumors indicate that the risk of ICH is not significantly increased if the degree of anticoagulation is carefully monitored.^19,108,109 However, caution must be exercised to avoid critical drug–drug interactions, including those with antiepileptics such as valproic acid.¹¹⁰Table 3 provides a summary of anticoagulant options and features.

Table 3.

Treatment of Venous Thromboembolism in Patients With Cancer^89,90

Agents	Initial Treatment	Long-Term Treatment	Comments
UFH and Warfarin	UFH: 80 U/kg IV bolus, then 18 U/kg per hour IV, dose-adjusted based on aPTT	Warfarin: adjusted to maintain INR 2–3	• Preferred over LMWH in patients with severe renal impairment (creatine clearance <30 mL/min), symptomatic PE, and high-risk patients
LMWH
Dalteparin	200 IU/kg once daily or 100 IU/kg every 12 h (mo 1)	150 IU/kg once daily	• Preferred parenteral AC in patients without renal impairment • More effective than warfarin in reducing VTE risk91 • Conflicting data exist on risk of ICH
Enoxaparin	1.5 mg/kg once daily or 1 mg/kg ever 12 h	40 mg daily, 1.5 mg/kg once daily, or 1 mg/kg every 12 h
Tinzaparin	175 U/kg once daily	175 U/kg once daily
DOAC
Edoxaban	At least 5 d of parenteral AC, typically LMWH	60 mg once daily or 30 mg once daily**	• Use with caution in patients with increased risk for GI and GU bleeding • Possible drug–drug interaction with CYP3A4 inducers, such as phenytoin, carbamazepine, and phenobarbital
Rivaroxaban	15 mg every 12 h for 21 d (doses with food)	20 mg once daily (doses with food)
Apixaban	10 mg twice daily for 7 d	5 mg twice daily

UFH = unfractionated heparin; aPTT = activated partial thromboplastin time; LMWH = Low molecular weight heparin; PE = pulmonary embolism; DOAC = Direct oral anticoagulants; AC = anticoagulation; ICH = intracranial hemorrhage; GI = gastrointestinal tract; GU = genitourinary.

^**For patients weighing ≤60 kg or having creatinine clearance of 30–50 ml/min or those receiving treatment with potent P-glycoprotein inhibitors.

Investigational Antithrombotic Therapies

Currently approved anticoagulants utilized for VTE treatment increase the risk of ICH in patients with brain tumors. In a systematic review and meta-analysis of randomized controlled trials or observational studies, the incidence of ICH in primary brain cancer patients was 6.4% overall, with a rate of 12.5% (95% CI, 8.0–18.8) in those treated with anticoagulation versus 4.4% (95% CI, 2.5–7.7) in the absence of anticoagulation.¹¹¹ This compares to an annual ICH rate of 0.1%–0.2% in patients without brain tumors on long-term DOACs.¹¹² Furthermore, although DOACs seem to be associated with a lower risk of ICH than LMWH in the relatively small cohorts that have been directly compared, the need for yet safer agents with a wider therapeutic window in this patient population is perhaps more urgent than in any other clinical scenario.

To this end, there has been much interest in the development of therapeutic agents that target factor XI (FXI) and/or FXIa. TF and TF-EVs trigger thrombin formation, which in turn activates FXI, suggesting FXI is an interesting target in GBM hypercoagulability. As a component of the intrinsic pathway of coagulation, FXI was not traditionally considered a candidate target molecule for anticoagulant therapy. However, epidemiologic evidence has now established that individuals who are congenitally deficient in FXI rarely experience spontaneous bleeding and have a reduced lifetime risk of VTE and stroke (although interestingly, not myocardial infarction).¹¹³ Animal models also support that genetic deletion or pharmacologic inactivation of FXI affords protection against both venous and arterial thrombosis.¹¹⁴ These studies have led to the development of inhibitors of FXI and/or FXIa, including a] small molecules that directly inhibit FXIa; b] inhibitory monoclonal antibodies that block FXI activation or FXIa activity; c] antisense oligonucleotides that reduce hepatic synthesis of FXI; and d] aptamers that bind FXI or FXIa. Thus far, clinical trials have addressed prevention of VTE in the setting of total knee arthroplasty, or prevention of stroke in patients with AF. In the former category, a meta-analysis of 6 phase II trials demonstrated a 41% reduction of (mostly asymptomatic) VTE with FXI inhibitors compared to prophylactic dose LMWH, and a 59% reduction in bleeding risk.¹¹⁵ While these data in the primary prophylaxis of VTE are certainly encouraging, it remains to be seen whether FXI/XIa inhibition is equally successful in the secondary prophylaxis (ie, short- and long-term treatment) of established VTE.

Inferior Vena Cava Filters

Ascertainment of what constitutes an absolute contraindication to anticoagulation in patients with glioma is based on expert opinion. Acute or recent intratumoral bleeding is an incontrovertible contraindication and has led to the suggestion of a non-contrast head CT prior to initiation of anticoagulation.¹¹⁶ Coagulopathy and thrombocytopenia are other accepted contraindications. Neither the presence of intratumoral microhemorrhages nor the use of bevacizumab preclude anticoagulation.¹¹⁷

In the general population with VTE, it is recognized that inferior vena cava (IVC) filters reduce the risk of PE, increase the risk of DVT, and have little impact on mortality.¹¹⁸ Nor is there good evidence that the combination of IVC filters with anticoagulation in particularly high-risk patients is more effective than anticoagulation alone. Consequently, the strongest indication for filter placement is in patients with VTE and an absolute contraindication to anticoagulation. Even in this setting, evidence supporting benefit from filters is not compelling.¹¹⁸ A second potential indication is for patients experiencing proven recurrent VTE events while on therapeutic doses of anticoagulation. In such cases it is recommended to first check compliance with anticoagulation, exclude drug–drug interactions, and consider increasing the dose or switching to an alternative agent prior to placing a filter.¹¹⁹ When filters are utilized, retrievable filters are recommended, and it is advised to remove the filter once the absolute contraindication to anticoagulation is alleviated or when the risk of clinically significant PE is acceptably low, as the risk of filter-associated DVT increases over time. Indications for cancer patients, in general, are similar: The American Society for Clinical Oncology guidelines state that filters may be offered in the setting of an absolute anticoagulation contraindication in the acute treatment setting (VTE diagnosis in the previous 4 weeks) for a potentially life-threatening clot burden.⁹⁰

The limited literature on malignant brain tumors suggests that IVC filters may be less effective and more prone to complications than in the general population. A study of permanent IVC filters in a population of patients with malignant brain tumors identified PE in 12%, with 57% of patients experiencing recurrent DVT, post-phlebitic syndrome, or filter thrombosis.¹²⁰ A recent study reported that 60 (41%) of 145 of patients with GBM and VTE underwent IVC filter placement, 39 filter alone, and 21 with anticoagulation. Twenty patients were diagnosed with recurrent VTE, including 14/39 without and 6/21 with anticoagulation.¹²¹ Thus, the American Society For Clinical Oncology guidelines are pertinent to gliomas, with an understanding that most patients with gliomas are anticoagulation candidates, and that there should be a plan to remove retrievable filters when employed.

Secondary Prevention of VTE

While in most cases, glioma patients with VTE should receive therapeutic anticoagulation in the acute phase, the optimal duration of treatment is unclear. General guidelines for cancer patients recommend continuation of anticoagulation beyond 6 months—effectively transitioning from acute treatment to secondary prophylaxis—in patients with active cancer, with either LMWH or DOACs as the agent of choice.^90,122 However, there are no specific guidelines on how to apply these recommendations to the glioma population, where the high risk of recurrent VTE must be balanced with the risk of ICH. A retrospective study in 145 glioblastoma patients showed a recurrent VTE rate of 26.9%, with most cases occurring in patients, not on therapeutic anticoagulation.¹²¹ On the other hand, most of the studies investigating the risk of ICH in anticoagulated glioma patients with VTE have focused on the period following the thrombotic event, and thus there is very little data regarding ICH risk beyond the 6-month mark. A recent retrospective matched cohort analysis investigating ICH risk in patients with high-grade glioma and VTE treated with LMWH reported no significant difference in 1- and 5-year CI of ICH between anticoagulated patients and matched controls with VTE but no anticoagulation;⁹² however, broader data and information on other anticoagulants such as DOACs are not available.

Even in the absence of more targeted evidence, the general guidelines for cancer patients can be applied to recommend indefinite anticoagulation for GBM patients with VTE,¹²³ since the definition of active cancer includes malignancies diagnosed or receiving treatment in the preceding 6 months as well as recurrent or regionally advanced tumors,¹²⁴ and most GBM patients will fall under this designation for the majority of their disease course. The decision is less straightforward for LGG patients, in whom the risk of either recurrent VTE or ICH is not defined; although the argument can be made that these tumors are always active in as much as they are incurable, it is reasonable to carefully consider the individual risk and benefits before committing a LGG patient with VTE to lifelong anticoagulation, particularly for patients early in their disease course who may have years before a recurrence makes the disease truly “active” again.

Conclusion

Compelling evidence supports the inclusion of GBM with cancers having a markedly elevated risk of VTE, with LGGs a lesser albeit still increased propensity. The biological underpinnings related to TF, PDPN, and epigenetic factors are starting to emerge. For most patients with established VTE, we recommend anticoagulation; while there are more data for LMWH, DOACs are a reasonable and more patient-friendly option. Management of patients who cannot receive anticoagulants remains unsatisfactory. Identification of high-risk patients combined with confirmation that prophylactic anticoagulation in this population is safe and effective remains an important unmet need. We hope that the recent advances in our understanding of the molecular biology of VTE in gliomas will spur much-needed prospective studies identifying risk factors and optimal management strategies for this common and life-threatening problem.

Funding

C.H. was supported by NIH grants R01NS102669, R01NS117104, R01NS118039, the Northwestern University SPORE in Brain Cancer P50CA221747, and the Lou and Jean Malnati Brain Tumor Institute. N.M. was supported by the NIH grant R35HL155657. We appreciate the assistance of Dr Yohei Hisada for help preparing the figure.

Conflicts of Interest

The authors report no conflicts of interest relevant to the manuscript.