Vitamin K Antagonists and Risk of Subdural Hematoma
Meta-Analysis of Randomized Clinical Trials
Background and Purpose—Subdural hematomas are an important bleeding complication of anticoagulation. We quantify the risk of subdural hematoma associated with anticoagulation with vitamin K antagonists (VKAs) compared with other oral antithrombotic therapies.
Methods—Randomized trials were identified from the Cochrane Central Register of Controlled Trials and were included if published since 1980 and compared oral VKAs with antiplatelet therapy or with direct-acting oral anticoagulants. Two reviewers independently extracted data with differences resolved by joint review.
Results—Nineteen randomized trials were included that involved 92 156 patients and 275 subdural hematomas. By meta-analysis, VKAs were associated with a significantly increased risk of subdural hematoma (odds ratios, 3.0; 95% confidence interval, 1.5–6.1) compared with antiplatelet therapy (9 trials, 11 603 participants). The risk of subdural hematoma was also significantly higher with VKAs versus factor Xa inhibitors (meta-analysis odds ratios, 2.9; 95% confidence interval, 2.1–4.1; 5 trials, 49 687 patients) and direct thrombin inhibitors (meta-analysis odds ratios, 1.8; 95% confidence interval, 1.2–2.7; 5 trials, 30 866 patients) versus VKAs. The absolute rate of subdural hematoma among 24 485 patients with atrial fibrillation treated with VKAs pooled from 6 trials testing direct-acting oral anticoagulants was 2.9 (95% confidence interval, 2.5–3.5) per 1000 patient-years.
Conclusions—VKA use significantly increases the risk of subdural hematoma by ≈3-fold relative to antiplatelet therapy. Direct-acting oral anticoagulants are associated with a significantly reduced risk of subdural hematomas versus VKAs. Based on indirect comparisons to VKAs, the risks of subdural hematoma are similar with antiplatelet monotherapies and factor Xa inhibitors.
Intracranial hemorrhage is the most serious complication of warfarin and other oral vitamin K antagonists (VKAs).1 About one third of intracranial hemorrhages during VKA therapy are subdural hematomas,2–4 collections of blood between the dura mater and the surface of the brain that result from bleeding from the bridging leptomeningeal veins. Subdural hematomas (the other intracranial hemorrhage) have received scant attention relative to the more devastating primary intracerebral hemorrhages, but they have health consequences comparable to ischemic strokes and carry an overall 20% mortality.1,5,6 The risk of subdural hematoma associated with the use of VKAs has not been well defined.
We undertook systematic review and meta-analyses of randomized clinical trials testing VKAs compared with antiplatelet therapy and with direct-acting oral anticoagulants to quantify the risk of subdural hematoma associated with VKA therapy.
Randomized clinical trials testing oral VKAs were sought by a computerized search of the Cochrane Central Register of Controlled Trials from 1980 to June 2013 (the era of modern neuroimaging required to diagnose subdural hematomas reliably), not restricted by language. In addition, reference lists from systematic reviews of randomized trials of VKAs in patients with atrial fibrillation7 and VKAs in coronary artery disease8 were reviewed. We screened the abstracts of 1667 articles and identified 110 articles that seemed relevant for full review (Figure 1). We included all trials reporting randomized comparisons of oral VKAs with antiplatelet monotherapy or with direct-acting oral anticoagulants, that included >1-month treatment, and that reported the occurrence of ≥1 subdural hematoma (trials reporting no subdural hematomas were excluded). Additionally, we contacted the investigators of trials that did not report subdural hematomas if the trials were published after 1998 and either reported ≥5 intracranial hemorrhages, had >1000 participants, or reported ≥20 major bleeds. All subdural hematomas, acute and chronic and traumatic and atraumatic, were included. Trials were excluded if they tested low-dose VKA anticoagulation in which the mean intensity of anticoagulation was not prolonged beyond the normal range, if all patients initially received a VKA and were later randomized to continue or stop (ie, randomized to different durations of anticoagulation), or if the comparator was dual antiplatelet agents. Trials in which patients were initially treated with heparin/low-molecular-weight heparins were included if the duration of treatment was <15% of the exposure to a VKA.
Inclusion was restricted to trials testing the following VKAs: warfarin, dicoumarol, phenprocoumon, fluindione, phenindione, acenocoumarol, and ethyl biscoumacetate and to the following antiplatelet agents: aspirin (any dosage), ticlopidine, clopidogrel, prasugrel, dipyridamole, triflusal, and cilostazol. Trials comparing dual antiplatelet therapy to VKAs were not considered (only 1 was identified). Direct-acting oral anticoagulants (ie, oral anticoagulants that interact directly with their protein target to inhibit clotting) included dabigatran etexilate, ximelagatran, apixaban, rivaroxaban, betrixaban, darexaban, and edoxaban, but parenteral agents (eg, idraparinux) were excluded. Three separate meta-analyses were planned: VKAs versus antiplatelet monotherapies, VKAs versus oral direct thrombin inhibitors, and VKAs versus oral factor Xa inhibitors.
Two reviewers (B.J.C., R.G.H.) independently extracted data from published sources using a standardized assessment form and calculated the Jadad score,9 with disagreements resolved by joint review. Data were extracted regarding trial design, number of patients treated, anticoagulant interventions (ie, target intensity), mean achieved international normalized ratio (INR) or prothrombin time ratio, total follow-up exposure, and number of subdural hematomas (traumatic and atraumatic) by treatment assignment. The diagnosis of subdural hematoma was accepted based on that reported by the investigators; specific information about computed tomography or MRI confirmation was not available. If the years of exposure for each treatment arm were not provided, they were estimated from the number of primary events divided by the annualized event rate; if the annualized primary event rate was not provided, then the mean follow-up multiplied by the number of participants was used.
Intention-to-treat results were used for the analyses when available (and footnoted when not). Meta-analyses of the trial results are presented as odds ratios (OR) computed assuming a random effects model with the assumption of statistical homogeneity of the treatment effect (across trials) tested using the QL statistic for the relative odds scale. The fixed effects model was chosen if rarity of events,10 and if the count in ≥1 of the cells for a trial was 0, then 0.5 was added to each of the 4 cells. ORs for individual trials are reported if the total number of subdural hematomas was ≥20. Heterogeneity across trials was also evaluated using the I2 index (percentage of the total variability in a set of effect sizes because of between-studies variability). Rates of occurrence and confidence intervals (CIs) across trials were computed by dividing the total number of subdural hematomas by the total patient-years of exposure and assuming a Poisson distribution. All CIs and P values are 2-sided; α of 0.05 was accepted as statistically significant. MedCalc for Windows, version 12.7.7 (MedCalc Software, Mariakerke, Belgium) and SPSS, version 20 were the software used.
Nineteen randomized trials involving 92 156 patients were included that compared VKAs with antiplatelet therapy (9 trials, 11 603 participants) or direct-acting oral anticoagulants (10 trials, 80 553 participants) (Table 1).11–29 In 11 trials involving 46 764 participants, subdural hematomas were not reported, and unpublished data were obtained from the study investigators. The 9 trials involving patients with atrial fibrillation contributed most patients (70%, n=64 609), with 5 large trials testing direct-acting oral anticoagulants in atrial fibrillation accounting for 66% of patients (n=60 694) (Table 1).
VKA Versus Antiplatelet Therapy
Data regarding subdural hematoma were available from 9 randomized trials (including unpublished data from 6 trials) with 11 603 participants and 34 subdural hematomas (Tables 1 and 2). Trial participants included patients with atrial fibrillation (3 trials), noncardioembolic stroke (4 trials), and heart failure/reduced left ventricular ejection fraction (2 trials). Mean participant age was 65 years (range, 61–82 years), and the mean achieved INRs ranged from 2.1 to 3.2 (Table 2). Antiplatelet therapy was aspirin (dose range, 30–1300 mg/d) except 1 trial each that tested triflusal 600 mg/d16 and clopidogrel 75 mg/d.18 Compared with antiplatelet therapy, assignment to VKAs was associated with a significantly increased risk of subdural hematoma (meta-analysis OR, 3.0; 95% CI, 1.5, 6.1; I2 index, 0%; P=0.8 for heterogeneity). When restricted to the 8 trials in which the mean achieved INR was between 2 and 3, the meta-analysis OR associated with VKA use was 2.8 (95% CI, 1.3–5.8). When restricted to 8 comparisons of VKA with aspirin, the meta-analysis OR was 2.3 (95% CI, 1.1, 5.0).
VKA Versus Oral Direct Factor Xa Inhibitors
Five trials involving 49 687 patients with a mean age of 68 years and 146 subdural hematomas compared warfarin (target INR range, 1.6–3.0) with rivaroxaban, apixaban, or edoxaban in patients with atrial fibrillation (3 trials) or venous thromboembolism (2 trials) (Tables 1 and 3). In 3 of these trials, per-protocol/modified intention-to-treat analyses excluded <1% of randomized patients and reported only on-therapy data for subdural hematoma.20,21,23 The risk of subdural hematoma was significantly higher (meta-analysis OR, 2.9; 95% CI, 0.2.1, 4.1; I2 index, 0%; P=0.6 for heterogeneity) for those assigned VKA versus factor Xa inhibitors.
VKA Versus Oral Direct Thrombin Inhibitors
Five trials (3 atrial fibrillation and 2 venous thromboembolism) involving 30 866 patients with a mean age of 69 years compared warfarin with either ximelagatran or dabigatran and included 95 subdural hematomas (Tables 1 and 3). The risk of subdural hematoma was significantly higher (meta-analysis OR, 1.8; 95% CI, 1.2, 2.7; I2 index, 0%; P=0.5 for heterogeneity) for those assigned VKA versus direct thrombin inhibitor. The risk of subdural hematoma was significantly lower (OR, 0.42; 95% CI, 0.20, 0.88; P=0.02) among participants assigned to dabigatran 110 mg BID versus 150 mg BID in the Randomized Evaluation of Long-Term Anticoagulation Therapy (RE-LY) trial (Table I in the online-only Data Supplement).3,27
Rate of Subdural Hematoma During VKA Therapy
The rate of subdural hematoma for 24 485 atrial fibrillation patients (mean age, 72 years) treated with VKAs from 6 trials testing direct-acting oral anticoagulants (Table 3) was 2.9 per 1000 patient-years (95% CI, 2.5, 3.5).
Subdural hematomas are an important hemorrhagic complication of anticoagulation therapy. They account for ≈10% of all major hemorrhages and for about one third of intracranial hemorrhages during warfarin anticoagulation of elderly patients with atrial fibrillation (Table II in the online-only Data Supplement). The relative risks of subdural hematoma associated with anticoagulation therapies have not been defined previously. These analyses demonstrate that VKAs increase the risk of subdural hematoma by ≈3-fold compared with antiplatelet therapy (meta-analysis OR, 3.0; 95% CI, 1.5, 6.1) and with factor Xa inhibitors in the dosages tested (meta-analysis OR, 2.9; 95% CI, 2.1, 4.1) (Table 4; Figure 2). Compared with direct thrombin inhibitors, the increased risk with VKAs is smaller (meta-analysis OR=1.8; 95% CI, 1.2, 2.7), albeit with a significant dosage-dependent effect seen with dabigatran (Table I in the online-only Data Supplement). In short, the use of VKAs double or triple the risk for subdural hematoma compared with other antithrombotic therapies.
The achieved mean INRs in the included VKA trials fell into a relatively narrow range (2.1–3.2) that approximates the target therapeutic range for most long-term clinical indications. The antithrombotic effects of aspirin seem to be independent of dosage within the ranges tested (50–1300 mg/d) in trials included in these analyses. Consequently, the increased relative risk for subdural hematoma with VKA therapy relative to aspirin that emerged from this analysis (OR, 2.3; 95% CI, 1.1, 5.0) is likely to be generalizable to most clinical settings. This relative risk translates into an absolute risk of ≈2 additional subdural hematomas per 1000 elderly atrial fibrillation patients given VKAs per year.
In contrast to aspirin therapy, the intensity of anticoagulation is clearly different for different dosages of individual direct-acting oral anticoagulants (as demonstrated by the different effects of the 2 dosages of dabigatran and edoxaban on ischemic stroke and bleeding in the RE-LY27 and Effective Anticoagulation With Factor Xa Next Generation in Atrial Fibrillation [ENGAGE AF] trials)24 and perhaps between agents with similar anticoagulant mechanism. A major caveat pertaining to this meta-analysis concerns pooling data from trials testing different direct-acting oral anticoagulants to estimate their effect on the occurrence of subdural hematoma relative to VKA. A single estimate for the entire class of factor Xa inhibitors or of direct thrombin inhibitors may well be misleading, and indirect comparison between these classes of selective anticoagulants are likely unreliable. However, a general consistency emerged from the available data that the direct-acting oral anticoagulants in the dosages tested are associated with a reduced risk of subdural hematoma compared with conventional intensities of warfarin anticoagulation (Table 4).
The mechanism(s) for the substantially reduced risks of subdural hematoma associated with the direct-acting oral anticoagulants relative to high-quality warfarin anticoagulation is unclear. Comparing the direct-acting oral anticoagulants with warfarin, intracerebral hemorrhage shows similar relative reductions, but major extracranial bleeding is not reduced (and probably increased).30 Vascular bed–specific hemostasis seems to differ between oral VKAs and direct-acting oral anticoagulants.31,32
These best available estimates of the relative risks of subdural hematoma are similar when VKAs are compared with antiplatelet therapy (OR, 3.0; 95% CI, 1.5, 6.1) and with factor Xa inhibitors (OR, 2.9; 95% CI, 2.1, 4.1), supporting similar risks of these 2 types of antithrombotic therapies on precipitating subdural hematoma (Table 4). When the 9 trials comparing VKA versus antiplatelet therapy are included in a meta-analysis with the 5 trials comparing VKA versus direct factor Xa inhibitors, there is no suggestion of heterogeneity of treatment effect (meta-analysis OR, 2.9; 95% CI, 2.1, 4.0; I2 index, 0%; P=0.9 for heterogeneity). However, CIs are relatively wide, the effect of factor Xa inhibitors on subdural hematomas may be dose dependent (Table I in the online-only Data Supplement), and indirect comparisons are problematic, particularly with different patient populations. Only limited data (6 total subdural hematomas from 1 randomized trial) are available from direct randomized comparisons of factor Xa inhibitors with aspirin.33
Additional limitations include that subdural hematomas were not a primary or secondary outcome in any of the published trials, necessitating that the numbers of subdural hematomas in the published reports be accepted without details about imaging (eg, acute versus chronic), without distinguishing those associated with trauma from those presumed to be atraumatic, and without description of symptoms prompting their discovery. These caveats apply particularly to the unpublished data. VKA use was reported to be more frequently associated with atraumatic subdural hematomas versus traumatic in a recent surgical series.34 It is uncertain whether there is an early high-risk period for subdural hematoma after initiation of VKA in VKA-naive patients, similar to that seen with intracerebral hemorrhages.35 A unique strength of the study is inclusion of unpublished data from 11 trials, comprising 52% of the subdural hematomas analyzed, that were previously inaccessible. Data about the risk of subdural hematoma were available from all large recent randomized trials testing VKAs except for a single trial comparing warfarin with apixaban from which number of subdural hematomas was not available.36
Subdural hematomas are an important complication of warfarin anticoagulation of the elderly. Subdural hematomas are well-known to recur (≈20% during short-term follow-up in neurosurgical series).37,38 The risks of subdural hematoma associated with direct-acting oral anticoagulants are significantly lower that for VKAs (odds ratio 0.34 for pooled results of trials comparing factor Xa inhibitors and 0.55 for pooled results of trials comparing direct thrombin inhibitors), offering the opportunity of reducing the morbidity and mortality associated with other subdural hematoma in elderly patients requiring anticoagulation.
We gratefully acknowledge the following colleagues for providing unpublished data regarding subdural hematomas from previously reported randomized clinical trials: Dr Francisco Perez-Gomez (Hospital Clinico San Carlos, Madrid, Spain) for National Study for Prevention of Embolism in Atrial Fibrillation (NASPEAF) data, Dr J.L.P. Thompson, R. Buchsbaum, Dr Shunichi Homma, and Dr J.P. Mohr (Columbia University, New York, NY) for Warfarin-Aspirin Recurrent Stroke Study (WARSS) and Warfarin vs Aspirin in Reduced Cardiac Ejection Fraction (WARCEF) data, Dr Joseph F. Collins (Veterans Administration Medical Center, Perry Point, MD) for the Warfarin and Antiplatelet Therapy in Chronic Heart Failure (WATCH) data, Dr Ale Algra (University Medical Center Utrecht, Utrecht, The Netherlands) for European/Australasian Stroke Prevention in Reversible Ischemia Trial (ESPRIT) data, Professor Masatsugu Hori (Osaka Medical Center for Cancer and Cardiovascular Research, Osaka, Japan) for Japanese-Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation (J-ROCKET) data, Professor Harry Buller (Academic Medical Center, Amsterdam, The Netherlands) for EINSTEIN PE data and Hokusai-Venous Thromboembolism (Hokusai-VTE) study, Dr Sam Schulman (McMaster University, Hamilton, Canada) for RECOVER II, and Dr Robert P. Giugliano (Brigham and Women’s Hospital, Boston, MA) for Effective Anticoagulation With Factor Xa Next Generation in Atrial Fibrillation (ENGAGE AF).
Dr Hart received research support from Bayer Healthcare (rivaroxaban). The other authors report no conflicts.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.005430/-/DC1.
- Received March 10, 2014.
- Revision received April 13, 2014.
- Accepted April 16, 2014.
- © 2014 American Heart Association, Inc.
- Majeed A,
- Kim YK,
- Roberts RS,
- Holmström M,
- Schulman S
- Hart RG,
- Diener HC,
- Yang S,
- Connolly SJ,
- Wallentin L,
- Reilly PA,
- et al
- Singer DE,
- Chang Y,
- Fang MC,
- Borowsky LH,
- Pomernacki NK,
- Udaltsova N,
- et al
- Connolly SJ,
- Eikelboom JW,
- Ng J,
- Hirsh J,
- Yusuf S,
- Pogue J,
- et al
- Mant J,
- Hobbs FD,
- Fletcher K,
- Roalfe A,
- Fitzmaurice D,
- Lip GY,
- et al
- Pérez-Gómez F,
- Alegría E,
- Berjón J,
- Iriarte JA,
- Zumalde J,
- Salvador A,
- et al
- Massie BM,
- Collins JF,
- Ammon SE,
- Armstrong PW,
- Cleland JG,
- Ezekowitz M,
- et al
- Olsson SB
- Schulman S,
- Kakkar AK,
- Goldhaber SZ,
- Schellong S,
- Eriksson H,
- Mismetti P,
- et al
- Fisher MJ
- Hylek EM,
- Evans-Molina C,
- Shea C,
- Henault LE,
- Regan S