Risk of intracranial hemorrhage with direct oral anticoagulants vs low molecular weight heparin in glioblastoma: A retrospective cohort study

Abstract

Background

Glioblastoma (GBM) is associated with a high incidence of venous thromboembolism (VTE), but there are little data to guide anticoagulation in patients with GBM, in whom the risks of VTE must be balanced against the risk of intracranial hemorrhage (ICH).

Methods

We performed a single-institution retrospective cohort study of patients with GBM diagnosed with VTE from 2014 to 2021 who were treated with low molecular weight heparin (LMWH) or a direct oral anticoagulant (DOAC). The incidence of ICH was compared between the LMWH and DOAC groups. The primary outcome was clinically relevant ICH within the first 30 days of anticoagulation, defined as any ICH that was fatal, symptomatic, required surgical intervention, and/or led to cessation of anticoagulation. Secondary outcomes included clinically relevant ICH within 6 months, fatal ICH within 30 days and 6 months, and any bleeding within 30 days and 6 months.

Results

One hundred twenty-one patients were identified in the cohort for 30-day outcome analyses (DOAC, n = 33; LMWH, n = 88). For 6-month outcome analyses, the cohort included only patients who were maintained on their initial anticoagulant (DOAC, n = 32; LMWH, n = 75). The incidence of clinically relevant ICH at 30 days was 0% in the DOAC group and 9% in the LMWH group (P = .11). The cumulative incidence of clinically relevant ICH at 6 months was 0% in the DOAC group and 24% in the LMWH group (P = .001), with 4 fatal ICHs in the LMWH group.

Conclusions

DOACs are associated with a lower incidence of clinically relevant ICH in patients with GBM-associated VTE compared to LMWH.

Key Points.

• Intracranial hemorrhage complicates treatment of venous thromboembolism in GBM.

• The safety of DOACs for treatment of GBM-related venous thromboembolism is unknown.

• In this study, intracranial hemorrhage was less common with DOACs compared to LMWH.

Importance of the Study.

Glioblastoma (GBM) is associated with a high incidence of venous thromboembolism (VTE), but anticoagulation is challenging in patients with GBM due to an increased risk of life-threatening intracranial hemorrhage (ICH) and an absence of evidence-based guidelines. While direct oral anticoagulants (DOACs) are increasingly being used to treat cancer-associated VTE, there are little safety data to guide their use in GBM. This study provides compelling evidence for the preferential use of DOACs in the treatment of GBM-associated VTE due to a significantly reduced incidence of anticoagulation-related ICH in patients receiving DOACs compared with patients receiving low molecular weight heparin (LMWH). To the best of our knowledge, this represents the largest body of evidence supporting the use of DOACs for VTE in patients with GBM.

Glioblastoma (GBM) carries one of the highest risks of venous thromboembolism (VTE) of any cancer.^1,2 An estimated 15%-30% of patients with GBM will develop deep vein thrombosis (DVT) or pulmonary embolism (PE) over the course of their disease.^3–6 This elevated risk is due to a combination of factors, including tumor-associated hypercoagulability, patient factors such as immobility and paresis, and heightened clotting risk associated with anti-vascular endothelial growth factor (VEGF) therapy, which is commonly used in the treatment of GBM.⁷ Despite this, little data exist to guide therapeutic anticoagulation in this population, in whom effective treatment of VTE must be balanced against a heightened risk of life-threatening intracranial bleeding. Even without anticoagulation, patients with GBM are at risk of intracranial hemorrhage (ICH) due to the presence of angiogenic factors, like VEGF and matrix metalloproteinases, and malformed intratumoral blood vessels.⁸ The incidence of spontaneous ICH in GBM is estimated at 6%-13%; with therapeutic anticoagulation, the incidence of ICH may increase as much as 3-fold, though there are conflicting data on the magnitude of this increased risk.^9–12

For years, low molecular weight heparins (LMWH) were the recommended first-line anticoagulants for the treatment of cancer-associated VTE.¹³ More recently, large, randomized trials have established direct oral anticoagulants (DOACs) as non-inferior to LMWH in the cancer population, and their ease-of-use means that DOACs are often preferred by both patients and clinicians.^14–17 However, patients with brain tumors have been underrepresented in these trials or excluded entirely due to their increased risk of ICH. As a result, there are no guidelines to dictate anticoagulant selection in the treatment of GBM-associated VTE. In fact, most of the existing evidence base for anticoagulation in brain cancer is extrapolated from data on patients with brain metastases, rather than primary brain tumors. There is evidence that the incidence of anticoagulation-associated ICH is higher for patients with primary brain tumors than brain metastases, which underscores the need for safety data specific to the GBM population.¹⁸ Limited retrospective data does suggest that DOACs may have a favorable safety profile among patients with both primary and metastatic brain tumors, but there are no randomized prospective trials in either population.^19–21 Moreover, to the best of our knowledge, a head-to-head comparison of DOACs vs LMWH for treatment of VTE in patients with GBM has not been previously reported, including retrospective analyses. To address this clinically significant gap in the neuro-oncology literature, we performed a retrospective cohort study evaluating the risk of ICH in patients with GBM and VTE treated with DOACs vs LMWH.

Patients and Methods

Patient Cohort

This was a single-center, retrospective cohort study performed at the Hospital of the University of Pennsylvania. The study period was between January 1, 2014 and December 31, 2021 and included patients diagnosed with GBM between January 1, 2014 and June 30, 2021. The start of this date range was selected based on the Food and Drug Administration’s approval of apixaban for VTE in 2014; the end of this date range was selected to allow at least 6 months of follow-up time in all patients. Cases were identified using International Classification of Diseases, Tenth Revision (ICD-10) codes for (a) “malignant neoplasm of brain” (C71) AND (b) “pulmonary embolism” (I26) OR “deep venous thrombosis” (I82.4). Charts of identified cases were then manually reviewed for confirmation that the patient was (a) an adult 18 years or older, (b) had a histopathologically confirmed diagnosis of GBM, isocitrate dehydrogenase (IDH) wild type, WHO grade IV,²² and (c) had a radiographically confirmed DVT and/or PE diagnosed at any time between the patient’s initial surgery for GBM and the patient’s death. Demographic and clinical data (including treatment details, concomitant use of antiplatelet agents, performance status, and renal function) were collected on all patients in the cohort. Data were stored in a deidentified fashion on a HIPAA-compliant web-based application for data collection (REDCap v11.4.3, Vanderbilt University). This study was approved by the University of Pennsylvania’s Institutional Review Board.

Study Design and Endpoints

The primary exposure variable was choice of anticoagulant: DOAC vs LMWH. A composite endpoint of clinically relevant ICH was created, defined as ICH meeting at least one of three criteria: (1) being fatal or symptomatic, (2) requiring intervention with either surgery or anticoagulant reversal, or (3) leading to cessation of anticoagulation. The primary endpoint was clinically relevant ICH within the first 30 days of anticoagulation. This time frame was chosen because the risk of bleeding is highest within the first weeks of anticoagulation therapy.^23,24 Secondary endpoints analyzed were incidence of clinically relevant ICH at 6 months, incidence of recurrent VTE at 6 months, and incidences of any ICH, fatal ICH, and any bleeding (including intra- and extracranial bleeding) at both time points.

Anticoagulation

Manual chart review was performed to determine each patient’s method of anticoagulation, with careful attention to the fact that some patients transitioned between different anticoagulants during their treatment. For example, anticoagulation might start with an unfractionated heparin drip while inpatient, transition to LMWH at discharge, then eventually switch to a DOAC weeks or months later as an outpatient. To avoid the confounding effect of anticoagulant switching, all patients who transitioned between anticoagulants (ie, LMWH to DOAC or DOAC to LMWH) during the first 30 days following VTE and during the first 6 months following VTE were excluded from the 30-day and 6-month analyses, respectively (Figure 1). This exclusion was not applied to patients who received unfractionated heparin as the initial anticoagulant immediately upon diagnosis of VTE prior to transitioning to DOAC or LMWH as the longer-term anticoagulant.

Fig. 1.

CONSORT flow diagram representing patient selection and inclusion in the 30-day and 6-month cohorts.

Intracranial Hemorrhage

Radiology reports for all available intracranial imaging (including both magnetic resonance imaging of the brain and computed tomography scans of the head) were reviewed to identify instances of ICH (a) within the first 30 days following the diagnosis of VTE (primary analysis) and (b) within 6 months of anticoagulation following a VTE (secondary analysis). When an ICH was identified, the patient’s chart was reviewed to ascertain the clinical context of each scan, and the ICH was classified as either incidental (ie, noted on a scheduled scan) or symptomatic (ie, imaging was obtained in response to symptoms, or symptoms attributable to the ICH ultimately developed after the incidental finding). Based on ICH dimensions calculated in the radiology reports, instances of ICH were further characterized as trace, measurable, or major. Using criteria established in prior literature, a trace hemorrhage was defined as being <1 mL in volume, or too small to measure (petechial ICH detectable only by susceptibility weighted imaging (SWI) or gradient echo (GRE) sequences was not included in this definition); a measurable ICH was any bleed ≥1 mL in volume; and a major ICH was defined as any bleed >10 mL in volume, requiring surgical intervention, or leading to clinical symptoms.^11,25 Any intracranial bleeds that were fatal were recorded as such. Instances of ICH are additionally characterized as being clinically relevant or not clinically relevant, using the criteria detailed above.

Statistical Analysis

Baseline characteristics and treatment information were compared between the two study groups using Student’s t tests for continuous variables and Fisher’s exact tests for categorical variables. Cumulative incidence for each outcome of interest was calculated by summing the number of events of interest (eg, clinically relevant ICH events) and dividing by the total number of patients in the cohort at risk of that event for the specified time interval (eg, 30 days). For all outcomes of interest, Fisher’s exact test was used to compare cumulative incidences between the DOAC and LMWH groups. For both the primary endpoint analysis (30-day clinically relevant ICH) and the secondary endpoint analysis of clinically relevant ICH within 6 months, we performed competing-risk analyses in which death was considered as a competing risk. Cumulative incidence of clinically relevant ICH was compared between the DOAC and LMWH groups using the Gray’s test to account for death as a competing risk.²⁶ All statistical tests were two-sided and completed using Stata version 14 (StataCorp LP, College Station, TX, USA).

Results

Patient Characteristics

A preliminary query of the electronic medical record system using diagnostic codes identified 169 patients with GBM diagnosed with VTE during the study period. Of those 169 patients, 20 were excluded because they did not receive therapeutic anticoagulation, due to either unacceptably high bleeding risk or diagnosis of VTE near the end of life. A further 14 patients were excluded whose primary anticoagulant was warfarin (n = 8) or were never treated with any anticoagulant other than an initial course of unfractionated heparin (n = 6). Of the remaining patients, 14 switched anticoagulants within the first 30 days and were thus excluded from our analysis, leaving a total of 121 patients in the primary 30-day analysis cohort: 88 in the LMWH group and 33 in the DOAC group. Of those patients, 59.5% (72/121) were initially treated with unfractionated heparin drip in the hospital prior to starting their longer-term anticoagulant (45.5% in the DOAC group, 64.8% in the LMWH group). Within the DOAC group, 84.8% (28/33) received apixaban, 12.1% (4/33) received rivaroxaban, and 3.0% (1/33) received edoxaban. In terms of dosing, 53.5% (15/28) of apixaban patients received a loading dose of 10 mg twice daily for 7 days before transitioning to the maintenance dose of 5 mg twice daily while 46.4% (13/28) were started at the maintenance dose. All four rivaroxaban patients received a loading dose of 15 mg twice daily and then transitioned to a maintenance dose of 20 mg daily after 21 days. The edoxaban patient received 60 mg daily. All patients in the LMWH group received enoxaparin; 96.6% (85/88) received 1 mg/kg twice daily while 3.4% (3/88) received 1.5 mg/kg daily. Over the 6-month period of interest, an additional 14 patients switched anticoagulants (occurring at a median of 43 days after VTE, IQR 34-45 days) and were therefore excluded from 6-month analyses. All 14 of these patients received a DOAC for the majority of the 6-month period of interest, and there were no clinically relevant ICH events in this group. The cohort of patients who had been anticoagulated exclusively with either a DOAC or LMWH during the 6-month follow-up period included 32 patients and 75 patients, respectively (Figure 1). None of these 107 patients had transient breaks in anticoagulation (for a procedure, for example) during the 6-month follow-up period.

Key demographic and clinical characteristics according to anticoagulation treatment group (DOAC vs LMWH) are displayed in Table 1. The cohort was predominantly male (68.6%) and the median age at the time of VTE was 63 (IQR 54-68, range 25-80). The breakdown of VTE by type was: 52.9% (64/121) with a DVT alone, 28.9% (35/121) with a PE, and 18.2% (22/121) diagnosed with DVT and PE simultaneously. At the time of VTE diagnosis, 20.7% of patients were on bevacizumab and 40.5% were on cytotoxic chemotherapy. Baseline characteristics were similar between the DOAC and LMWH groups, with the exception of use of antiplatelet agents and duration of time between most recent surgery and VTE. A total of 12 patients in the LMWH group (13.6%) were on an antiplatelet agent at the time of the VTE compared to 1 (3.0%) in the DOAC group (P = .093). The majority of those patients (11 out of 13) were continued on their antiplatelet agent when they started anticoagulation. Patients in the LMWH group were diagnosed with VTE closer to the date of most recent intracranial surgery compared to patients in the DOAC group (mean of 149 vs 269 days, P = .019).

Table 1.

Patient Characteristics at the Time of VTE Diagnosis

	LMWH	DOAC	P-value
	n = 88	n = 33
Sex
Male	59 (67.0%)	24 (72.7%)	.549
Female	29 (33.0%)	9 (27.3%)
Age (mean ± SD)	60.1 (±12.4)	60.5 (±12.7)	.88
KPS range
80%-100%	47 (20.5%)	11 (33.3%)	.33
50%-70%	25 (63.6%)	18 (54.5%)
10%-40%	16 (15.9%)	4 (12.1%)
Type of VTE
DVT	47 (53.4%)	17 (51.5%)	.977
PE	25 (28.4%)	10 (30.3%)
DVT and PE	16 (18.2%)	6 (18.2%)
Platelet count (mean ± SD)	195.2 (±114.2)	186.4 (±102.4)	.7
Creatinine (mean ± SD)	0.75 (±0.2)	0.83 (±0.3)	.09
Recent bleeding event
Yesa	5 (5.7%)	4 (12.1%)	.229
No	83 (94.3%)	29 (87.9%)
Time since most recent surgery in days (mean ± SD)	149.3 (±178.3)	269.6 (±366.7)	.019
On antiplatelet agent	12 (13.6%)	1 (3.0%)	.093
On bevacizumab	17 (19.3%)	8 (24.2%)	.551

Abbreviations: DOAC, direct oral anticoagulant; DVT, deep vein thrombus; KPS, Karnofsky Performance Scale; LMWH, low molecular weight heparin; PE, pulmonary embolism; SD, standard deviation; VTE, venous thromboembolism.

^aRecent bleeding events by type: postoperative ICH, 6 (67%); other ICH, 2 (22%); GI bleed, 1 (11%).

Primary and Secondary Outcomes

In our primary analysis, the cumulative incidence of clinically relevant ICH within the first 30 days of anticoagulation was 9.1% (8/88) in the LMWH group (95% CI 0.04-0.17) compared with a cumulative incidence of 0% (0/33) in the DOAC group (one-sided 97.5% CI 0-0.11) (P = .11, Table 2). In secondary 30-day analyses, the cumulative incidence of fatal ICH within 30 days was 2.3% (2/88) among LMWH patients and 0% (0/33) among DOAC patients (P = 1.0), and the cumulative incidence of any bleeding within 30 days (including ICH of any severity, gastrointestinal (GI) bleeding, and any other bleeds requiring medical attention) was 19.3% (17/88) in the LMWH group vs 3.0% (1/33) in the DOAC group (P = .024).

Table 2.

Primary and Secondary Outcomes

	LMWH (n = 88 at 30 days; 75 at 6 months)	DOAC (n = 33 at 30 days; 32 at 6 months)	P-value
Primary outcome—no. (%)a
Clinically relevant ICH within 30 days	8 (9)	0 (0)	0.11
Secondary outcomes—no. (%)a
Any ICH within 30 days	9 (11)	1 (3)	0.28
Fatal ICH within 30 days	2 (2)	0 (0)	1.0
Any bleeding within 30 days	17 (19)	1 (3)	0.024
Clinically relevant ICH within 6 months	18 (24)	0 (0)	0.001
Any ICH within 6 months	20 (27)	1 (3)	0.004
Fatal ICH within 6 months	4 (5)	0 (0)	0.32
Any bleeding within 6 months	31 (41)	3 (9)	0.001
Recurrent VTE in 6 months	3 (4)	0 (0)	0.55

Abbreviations: DOAC, direct oral anticoagulant; ICH, intracranial hemorrhage; LMWH, low molecular weight heparin; VTE, venous thromboembolism.

^aValues displayed are cumulative incidences.

At 6 months, the cumulative incidence of clinically relevant ICH was 24% (18/75) in the LMWH group (95% CI 0.15-0.37) and 0% (0/32) in the DOAC group (one-sided 97.5% CI 0-0.11) (P = .001). The cumulative incidence of fatal ICH within the first 6 months of anticoagulation was 5.3% (4/75) in the LMWH group and 0% (0/32) in the DOAC group (P = .32). The cumulative incidence of any bleeding in the first 6 months was 41.3% (31/75) in the LMWH group compared with 9.4% (3/32) in the DOAC group (P = .001). Within the LMWH group, the breakdown of the total bleeding events was 64.5% (20/31) ICH (60% measurable, 20% major, 20% fatal), 3.2% (1/31) GI bleeds, and 32.3% (10/31) other bleeds; within the DOAC group, the breakdown was 33.3% (1/3) trace ICH and 66.7% (2/3) GI bleeds. The cumulative incidence of recurrent VTE while on anticoagulation within 6 months was 4% (3/75) in the LMWH group and 0% (0/32) in the DOAC group (P = .55).

Cumulative incidence curves for 30-day clinically relevant ICH and 6-month clinically relevant ICH are displayed in Figure 2A and B, respectively. Competing-risk analysis accounting for competing risk of death (ie, Gray’s test) did not significantly alter the P-values when comparing the cumulative risk of clinically relevant ICH in the DOAC vs LMWH groups (30-day clinically relevant ICH, P = .08; 6-month clinically relevant ICH, P = .004).

Fig. 2.

Cumulative incidence curves displaying the cumulative incidence of clinically relevant ICH at 30 days (A) and 6 months (B). Gray’s test was used to account for the competing risk of death. Abbreviation: ICH, intracranial hemorrhage.

Discussion

In this retrospective cohort of 121 patients with GBM, DOACs were associated with an overall lower cumulative incidence of clinically relevant ICH compared to LMWH for the treatment of cancer-associated VTE. In fact, there were zero clinically relevant ICHs among DOAC patients at both the 30-day and 6-month time points following diagnosis of VTE. Although the differing incidence of clinically relevant ICH in DOAC- vs LMWH-treated patients did not reach statistical significance at the 30-day endpoint, at 6 months the DOAC patients had a 0% cumulative incidence of clinically relevant ICH compared with 23% in the LMWH group (P = .001). In addition, there were 4 fatal ICHs in the LMWH group (5% of patients at 6 months) compared with 0 in the DOAC group. We also found that there was a lower cumulative incidence of any bleeding at 6 months among patients on DOACs (9%) compared to LMWH (41%) (P = .001) and no difference in the cumulative incidence of recurrent VTE. Taken together, these results suggest that DOACs are safe to use in patients with GBM and may be preferable over LMWH.

VTE is common among patients with GBM, but anticoagulation is challenging in these patients due to their risk of intracranial bleeding, which is elevated even in the absence of anticoagulation. While DOACs have become a preferred class of anticoagulant for many indications, there are limited safety data to guide their use in patients with GBM, who were underrepresented in or excluded entirely from randomized trials of DOACs for cancer-associated VTE.^13–15 Our findings add to a small but growing body of retrospective evidence suggesting that DOACs have a favorable safety profile in CNS malignancies. Carney et al and Swartz et al found that there was a lower incidence of ICH among patients with both primary and metastatic brain tumors who were treated with a DOAC compared to LMWH.^19,21 Dubinski et al found no major difference in major ICH among patients with GBM with PEs who were anticoagulated with either DOAC or LMWH.²⁰ To the best of our knowledge, our study represents the largest head-to-head comparison of DOACs and LMWH for the treatment of VTE in patients with GBM. Our findings strengthen and expand upon the existing literature by differentiating between clinically relevant and irrelevant ICH in our analysis, as well as specifically accounting for fatal bleeding events.

The primary limitation of this study is its retrospective design. Bleeding events may have been underestimated, and it was not possible to assess true compliance to anticoagulation. It is also possible that underlying differences between the DOAC and LMWH groups accounted for their differential bleeding outcomes. Although the DOAC and LMWH groups were similar in terms of age, performance status, bevacizumab use, pre-VTE bleeding history, and time since most recent surgery, arguing against such confounding, we cannot completely rule out differences in medical comorbidities or other unmeasured confounders between the two groups. Moreover, since there were no ICH events in patients treated with DOACs in our cohort, we were statistically unable to perform multivariate logistic regression to adjust for other variables that could affect the choice of anticoagulant and/or incidence of ICH. Of note, there was a higher incidence of concurrent usage of antiplatelet agents in the LMWH group (13% vs 3% among DOAC patients). The preferential use of LMWH in this group with higher baseline bleeding risk could be related to the absence of a DOAC reversal agent at the beginning of the study period. Regardless, it did not seem to affect outcomes, as there were only 2 clinically significant ICHs among the LMWH patients also on an antiplatelet agent over the full 6-month study period. That represents a lower incidence than among the LMWH patients not on an antiplatelet agent (16% vs 27%). The other statistically significant difference between our study groups was the mean duration of time between VTE and most recent intracranial surgery (149 days in the LMWH group compared to 269 in the DOAC group). We suspect this difference may also be attributable to the preferential use of an anticoagulant with a reversal agent in patients perceived to be at higher risk of bleeding. Because VTEs in both groups occurred months out from surgery, it is unlikely that this disparity confounds our results.

Another limitation is that all our data are drawn from a single institution in which anticoagulation decisions were typically made by a patient’s oncologist. In our initial sample, 20 patients diagnosed with VTE did not receive therapeutic anticoagulation, typically because they were determined to be at unacceptably high bleeding risk. It is possible that these patients would have been anticoagulated elsewhere, a selection bias that would lower the overall bleeding rate in our sample compared with similar patients cared for at other institutions. An additional limitation is the lack of radiology review of intracranial imaging, though it is unlikely that such a review would have changed the cumulative incidence of clinically relevant ICH. Despite these limitations, which we attempted to mitigate as much as possible, the overall extremely low incidence of ICH in patients treated with DOACs compared to LMWH provides some of the strongest available evidence to date for the safety of DOACs in patients with GBM-associated VTE.

In summary, our data provide compelling evidence for the preferential use of DOACs in the treatment of GBM-associated VTE due to a significantly reduced incidence of anticoagulation-related ICH. While a randomized controlled trial would be needed to definitively answer this question, it is unlikely that such a trial will ever be performed due to strong patient and provider preferences around choice of anticoagulant, lack of enthusiasm from pharmaceutical companies, and insurance limitations. Instead, larger, multicenter retrospective trials are needed to confirm our findings. Until that point, this study represents the largest body of evidence to the best of our knowledge supporting the use of DOACs for VTE in patients with GBM.

Funding

None.

Conflict of interest statement. None.

Authorship statement. Experimental design: L.R.G. and S.J.B. Data collection: L.R.G. Data analysis and interpretation: L.R.G., D.C., and S.J.B. Manuscript drafting and revisions: L.R.G., A.S.D., R.E.P., D.C., K.A., M.O., S.B., D.M.O., N.M.A., and S.J.B.