Intracranial Hemorrhage in Atrial Fibrillation Patients During Anticoagulation With Warfarin or Dabigatran
The RE-LY Trial
Background and Purpose—Intracranial hemorrhage is the most devastating complication of anticoagulation. Outcomes associated with different sites of intracranial bleeding occurring with warfarin versus dabigatran have not been defined.
Methods—Analysis of 18 113 participants with atrial fibrillation in the Randomized Evaluation of Long-term anticoagulant therapY (RE-LY) trial assigned to adjusted-dose warfarin (target international normalized ratio, 2–3) or dabigatran (150 mg or 110 mg, both twice daily).
Results—During a mean of 2.0 years of follow-up, 154 intracranial hemorrhages occurred in 153 participants: 46% intracerebral (49% mortality), 45% subdural (24% mortality), and 8% subarachnoid (31% mortality). The rates of intracranial hemorrhage were 0.76%, 0.31%, and 0.23% per year among those assigned to warfarin, dabigatran 150 mg, and dabigatran 110 mg, respectively (P<0.001 for either dabigatran dose versus warfarin). Fewer fatal intracranial hemorrhages occurred among those assigned dabigatran 150 mg and 110 mg (n=13 and n=11, respectively) versus warfarin (n=32; P<0.01 for both). Fewer traumatic intracranial hemorrhages occurred among those assigned to dabigatran (11 patients with each dose) compared with warfarin (24 patients; P<0.05 for both dabigatran doses versus warfarin). Independent predictors of intracranial hemorrhage were assignment to warfarin (relative risk, 2.9; P<0.001), aspirin use (relative risk, 1.6; P=0.01), age (relative risk, 1.1 per year; P<0.001), and previous stroke/transient ischemic attack (relative risk, 1.8; P=0.001).
Conclusions—The clinical spectrum of intracranial hemorrhage was similar for patients given warfarin and dabigatran. Absolute rates at all sites and both fatal and traumatic intracranial hemorrhages were lower with dabigatran than with warfarin. Concomitant aspirin use was the most important modifiable independent risk factor for intracranial hemorrhage.
- atrial fibrillation
- intracranial hemorrhage
- intracerebral hemorrhage
- subdural hematoma
Intracranial hemorrhage is the most feared complication of warfarin anticoagulation in older patients with atrial fibrillation and is responsible for the bulk of disability and death from anticoagulation-associated bleeding.1 About two thirds of intracranial hemorrhages during warfarin anticoagulation are intracerebral hemorrhages, and most of the remainder are subdural hematomas. In recent randomized trials testing antithrombotic therapies in atrial fibrillation patients, the primary efficacy outcome included intracerebral hemorrhages combined with ischemic strokes, whereas subdural hematomas were separately categorized with major hemorrhage.2–6 The morbidity and mortality of different sites and precipitants of intracranial hemorrhage have been incompletely characterized in these studies.
In the Randomized Evaluation of Long-term anticoagulant therapY (RE-LY) randomized trial, the incidence of intracerebral hemorrhage in atrial fibrillation patients was significantly lower with dabigatran, a novel oral direct thrombin inhibitor, compared with warfarin.4 Little is known about the full clinical spectrum of intracranial bleeding in atrial fibrillation patients given dabigatran. The absence of an antidote to reverse emergently its antihemostatic effect has prompted concern that intracranial hemorrhages with dabigatran could carry a worse prognosis than could those associated with warfarin. We analyze intracranial hemorrhages occurring during anticoagulation, including sites, rates, risk factors, associated trauma, and outcomes among participants in the RE-LY trial.
The design and main results of the RE-LY trial have been published.4,7,8 The trial was funded by Boehringer Ingelheim and was coordinated by the Population Health Research Institute (Hamilton, Canada). Between 2005 and 2007, 18 113 patients from 951 sites in 44 countries who had documented atrial fibrillation and at least 1 additional stroke risk factor were randomized to receive, in a blinded fashion, fixed doses of dabigatran—110 mg or 150 mg twice daily; or, they were given open-label, adjusted-dose warfarin with a target international normalized ratio of 2.0 to 3.0.4
Intracranial hemorrhages were identified by local investigators and were submitted for central adjudication by neurologists.4 Source documents were translated into English, and information about antithrombotic therapy was expunged. Neuroimaging confirmation was present in 97% of intracranial hemorrhages: computed tomography (CT; 87%), magnetic resonance imaging (MRI; 3%), or both (7%); the diagnosis was based on spinal fluid examination in 1 patient. For this project, 2 stroke neurologists (R.G.H., H-C.D.) independently re-evaluated each case to identify the primary site of intracranial bleeding by review of imaging reports, presence of associated head trauma, and neurological outcomes, with differences resolved by consensus. Details about reversal of anticoagulation were inconsistently available in the source documents.
Intracranial hemorrhage was classified by the site of bleeding into intracerebral hemorrhages (classified as hemorrhagic strokes in the RE-LY main results), subdural hematomas, and subarachnoid hemorrhages based on imaging features from review of local interpretation of brain imaging (Figure). Patients with traumatic intracranial hemorrhage frequently had multiple sites of hemorrhage and were categorized according to the most clinically important site. Traumatic hemorrhagic contusions were classified as traumatic intracerebral bleeds. For patients with massive head trauma and multiple sites of intracranial bleeding for whom the most clinically relevant site could not be determined, the hemorrhages were analyzed as intracerebral hemorrhages if present (n=3) and otherwise as subdural hematomas (n=2). Large intracerebral hemorrhages with secondary rupture into the ventricular system were categorized as intracerebral, as well as the single patient with a primary intraventricular hemorrhage. Patients with secondary hemorrhagic transformation of ischemic strokes were excluded based on consideration of neuroimaging reports and clinical scenario—evidence of mottled hemorrhage within a larger area of apparent infarction and/or appearance of hemorrhage on follow-up imaging. These policies resulted in 1 intracranial hemorrhage per patient categorized at a single site of bleeding; the single exception was a patient with a traumatic intracerebral hemorrhage who several months later had a spontaneous subdural hematoma.
For reproducibility of classification as traumatic with differing amounts of clinical information, we applied specific criteria: the level of trauma must have been that which would have warranted immediate medical attention. Examples:
a. Spontaneous subdural hematoma: the patient slipped and fell at home, but did not seek medical attention. One week later, headache and mild hemiparesis led to CT and diagnosis of chronic subdural hematoma.
b. Traumatic subdural hematoma: the patient fell at home and was taken to a local emergency clinic where he was evaluated, treated for contusions, and released. One week later, headache led to CT and diagnosis of subacute subdural hematoma.
c. If a subdural hematoma had CT features of chronicity (eg, hypodense relative to brain) at the time of evaluation for acute head trauma, it was assumed not to be the result of the acute episode of head trauma and was classified as spontaneous.
Using these criteria, diagnosis was influenced by access to urgent medical care and likely resulted in misclassification of some trauma-related subdural hematomas as spontaneous. There were 7 patients with neuroimaging evidence of intracranial hemorrhage (3 intracerebral hemorrhage, 4 subdural), but too little clinical information to ascertain the role of trauma, and these patients were arbitrarily classified as spontaneous. Outcomes were categorized as full recovery, survival with neurological deficit, or fatal based on all available follow-up information.
All analyses are based on the intention-to-treat paradigm unless otherwise designated. The t test was used for the continuous variables, and χ2 test was used for categorical variables. The analyses of independent predictors of intracranial hemorrhage were based on Cox proportional hazards regression models, which considered variables that had probability values <0.1 by univariate analysis; only those that were significant are presented. Hazard ratios, 95% confidence intervals, and nominal probability values were calculated for outcome events comparing treatment arms. Mortality from intracranial hemorrhage between treatments was compared by χ2 test. All analyses were performed with SAS version 9.1 (SAS Inc). Two-sided probability values <0.05 were considered statistically significant. No adjustments were made for multiple comparisons.
Among all participants, the mean age was 71 years, 20% had previous stroke or transient ischemic attack (TIA), 50% received warfarin before study entry, and blood pressure at entry averaged 131/77 mm Hg. Concomitant aspirin was used at the first study follow-up visit by 28% of patients, and the time-in-therapeutic-range for those assigned to warfarin averaged 64%, with a mean achieved international normalized ratio of 2.41.
During follow-up, 154 intracranial hemorrhages occurred in 153 patients, with an overall 30-day mortality of 36% (Table 1). Intracranial hemorrhages included intracerebral hemorrhages (46%, with 49% mortality), subdural hematomas (45%, with 24% mortality), and subarachnoid hemorrhages (8%, with 31% mortality; Figure 1). Associated trauma was present in 30% of intracranial hemorrhages, but differed by site: 11% of intracerebral hemorrhages were classified as traumatic versus 44% of subdural hematomas (Table 1). Of 108 spontaneous intracranial hemorrhages, 58% were intracerebral hemorrhages (52% mortality), 36% were subdural hematomas (21% mortality), and 6% subarachnoid hemorrhages (67% mortality).
Compared with those without intracranial hemorrhage, patients with intracranial hemorrhage were, on average, older (P<0.001) with a history of stroke or TIA (P=0.001), more often took aspirin during follow-up (P=0.001), less often had heart failure (P=0.02), and had, on average, lower estimated creatinine clearances (P<0.001; Table 2); these differences were consistent between treatment arms (online-only, Supplemental Data, Appendix I). By multivariate analysis, assignment to warfarin (relative risk [RR], 2.9; P≤0.001), aspirin use during follow-up (RR, 1.6; P=0.01), age (RR, 1.1 per year; P<0.001), previous stroke/TIA (RR, 1.8; P=0.001) and white race (RR, 0.68; P=0.02) were independent predictors of intracranial bleeding (Table 3). By on-treatment analysis, the independent predictors were similar, except the relative risk associated with assignment to warfarin was larger (RR, 3.8; P=0.001) and white race was no longer significant (online-only, Supplemental Data, Appendix III).
The rate of intracranial hemorrhage was 0.76% per year among those assigned to warfarin and was significantly lower for those assigned to dabigatran 150 mg (0.31% per year; RR, 0.40; 95% CI, 0.27–0.59) and dabigatran 110 mg (0.23% per year; RR, 0.30; 95% CI, 0.19–0.45; Table 4). Mortality associated with intracranial hemorrhage was similar between treatment arms (36% warfarin, 35% dabigatran 150 mg, 41% dabigatran 110 mg; Table 5). Fatal intracranial bleeding occurred in 32 patients assigned to warfarin versus 13 patients and 11 patients with dabigatran 150 mg and 110 mg, respectively (P<0.01 for both comparisons with warfarin). Independent predictors of intracranial hemorrhage with warfarin were the same as those listed above with the addition of time-in-therapeutic-range (P=0.05; Table 3). Only age (RR, 1.06 per year; P=0.002) was independently predictive of intracranial hemorrhage among dabigatran-assigned patients (Table 3). The relative risk of intracranial hemorrhage was lower with dabigatran (either dose) compared with warfarin among subgroups defined by independent predictors of intracranial hemorrhage (online-only, Supplemental Data, Appendix V).
Most intracerebral hemorrhages (89%) were spontaneous. Of traumatic intracerebral hemorrhages (n=8), almost all were associated with major head trauma. Traumatic intracerebral hemorrhages usually consisted of dense hematoma within more widespread areas of cerebral contusion, and the contribution of the hematoma to neurological status and outcome was difficult to define.
Independent predictors of developing spontaneous intracerebral bleeding were: assignment to warfarin (RR, 4.1; P<0.001), previous stroke/TIA (RR, 2.7; P<0.001), aspirin use (RR, 1.8; P=0.02), and age (1.04 per year; P=0.02; Table 3), with similar independent predictors from on-treatment analysis (online-only, Supplemental Data, Appendix III). Aspirin use and previous stroke/TIA predicted intracerebral hemorrhage in 42 warfarin-assigned patients, but there were no significant predictors for the 21 events in those assigned dabigatran (Table 3).
The rate of spontaneous intracerebral hemorrhage was 0.36% per year (n=42) among those assigned to warfarin and was substantially lower for those assigned to dabigatran 150 mg (0.09% per year, n=11; RR, 0.26; 95% CI, 0.13–0.50) and dabigatran 110 mg (0.08% per year, n=10; RR, 0.23; 95% CI, 0.12–0.47; Table 4). The mortality associated with spontaneous intracerebral hemorrhage averaged 52%, with no significant differences between treatment arms (Table 5). Fatal spontaneous intracerebral bleeding occurred in 19 patients assigned to warfarin versus 7 patients each with dabigatran 150 mg and 110 mg (P<0.01 for both comparisons with warfarin).
Considering locations of spontaneous intracerebral bleeding, patients with hemorrhage in the basal ganglia/thalamus were, on average, younger (P=0.04) and more often had diabetes (P=0.02) compared with those with lobar bleeding (online-only, Supplemental Data, Appendix VI).
Subdural hematomas accounted for 45% of intracranial hemorrhages and were associated with trauma in 44%—an identical percentage for warfarin-assigned (44%, 16/36) and dabigatran-assigned (44%, 15/34) participants (Table 4). Risk factors for subdural hematomas categorized as spontaneous (n=39) and traumatic (n=31) were not different (online-only, Supplemental Data, Appendix II), and all subdural hematomas were considered together in subsequent analyses. The identification of subdural hematomas was distributed relatively evenly throughout follow-up in all treatment arms, and paralleled the occurrence of intracerebral hemorrhage (online-only, Supplemental Data, Appendix VII).
Participants with subdural hematomas were older (P<0.001), were assigned to warfarin (P=0.001), more often used aspirin during follow-up (P=0.001), and had reduced creatinine clearance (P<0.001) compared with patients without intracranial hemorrhage (Table 2). A history of falls before study entry was not significantly predictive (Table 2); only age was an independent predictor of developing subdural hematomas for patients assigned to warfarin (RR, 1.05; P=0.04), whereas assignment to the higher dabigatran dosage (RR, 2.4; P=0.02) and male sex (RR, 2.5; P=0.03) independently predicted subdural hematomas among dabigatran-assigned patients (Table 3).
The rate of subdural hematoma was 0.31% per year (n=36) among those assigned to warfarin versus 0.20% per year for those assigned to dabigatran 150 mg (n=24; RR, 0.65; P=0.10) and 0.08% per year for dabigatran 110 mg (n=10; RR, 0.27; P<0.001; Table 4). The rate of subdural hematomas was significantly higher with dabigatran 150 mg compared with the 110 mg dosage (RR, 2.4; P=0.02; Table 4). Fatal subdural bleeding occurred in 10 patients assigned to warfarin versus 5 patients and 2 patients to dabigatran 150 mg and 110 mg, respectively (P<0.05 for dabigatran 110 mg compared with warfarin).
Traumatic Intracranial Hemorrhages
Of 46 traumatic intracranial hemorrhages, 67% were subdural hematomas (Table 1). Significantly fewer traumatic intracranial hemorrhages occurred among those assigned to either dosage of dabigatran (11 patients for both dosages) compared with warfarin (24 patients; (P<0.05 for both dabigatran dosages versus warfarin; Table 4). Fatal traumatic intracranial hemorrhages occurred in 5 patients, 3 patients, and 3 patients assigned to warfarin, dabigatran 150 mg, and dabigatran 110 mg, respectively (Table 5).
The major findings of these analyses are the similar frequencies of spontaneous intracerebral hemorrhage and subdural hematoma in warfarin-assigned patients, the importance of concomitant aspirin use as a risk factor for spontaneous intracerebral hemorrhage, and the substantially lower rates at all sites of intracranial hemorrhage and of absolute mortality rates because of intracranial bleeding with dabigatran compared with high-quality warfarin anticoagulation. In addition, this first analysis of the role of trauma in intracranial hemorrhage in dabigatran-treated patients reveals significantly lower absolute rates compared with patients assigned to warfarin.
Intracerebral hemorrhage is the most devastating complication of anticoagulation, with mortality rates exceeding 50% in most studies (52% in RE-LY).1,9–11 This highly lethal stroke subtype cannot be considered equivalent to ischemic stroke when assessing clinical trial outcomes.9,12 Absolute rates of intracerebral hemorrhage during warfarin anticoagulation of atrial fibrillation patients in recent studies ranged from 0.3% per year to 0.6% per year (0.39% per year in RE-LY).2,3,9,12,13 Consequently, the lower intracranial hemorrhage rates during anticoagulation with dabigatran compared with warfarin are not explained by unduly high rates during warfarin anticoagulation, but rather by low rates with dabigatran that approximate the rate in nonanticoagulated atrial fibrillation patients.14 Concomitant aspirin use has been associated with intracranial hemorrhage during warfarin anticoagulation in most,13–16 but not all,17 previous studies and was confirmed as an independent risk factor here.
Subdural hematomas are collections of blood between the dura and leptomeninges, most often seen in older people and resulting from a tear in the veins bridging the meninges. Consequently, all subdural hematomas are theoretically traumatic, although the trauma causing the dural tear is often trivial and subclinical. In previous studies of older patients taking warfarin, subdural hematomas made up about 30% of intracranial hemorrhages with absolute rates between 0.1% to 0.3% per year and associated mortality rates averaging 20% to 30%.1–3 The rate of subdural hematoma in RE-LY was 0.31% per year with warfarin and was significantly lower among those assigned dabigatran 110 mg (RR, 0.27; P<0.001), but not for those given dabigatran 150 mg (RR, 0.65; P=0.10; Table 4). Predisposition to falling was found to be an independent risk factor for intracranial hemorrhage in a large study of Medicare beneficiaries with atrial fibrillation,11 but a history of falling was not independently predictive among RE-LY participants.
All sites of intracranial hemorrhage were less frequent in patients assigned to dabigatran compared with warfarin. It has been hypothesized that warfarin interferes with tissue factor VIIa-mediated thrombosis that may be especially important for hemostasis within the brain, whereas novel oral anticoagulants do not because of their more selective mechanisms of action. Any explanation for the lower rates of intracranial hemorrhage seen with dabigatran must account for reduced rates associated with all sites of intracranial hemorrhage and with traumatic versus atraumatic hemorrhages.
There is concern that patients with intracranial hemorrhages during treatment with dabigatran could have a worse prognosis than could those with warfarin because of the absence of a proven treatment to reverse emergently the antithrombotic effect. Mortality from intracranial hemorrhage was not increased in dabigatran-treated patients compared with those given warfarin (Table 5). This observation, coupled with the substantially lower absolute rates of intracranial hemorrhage with dabigatran, explains why the likelihood of dying from intracranial bleeding is significantly lower (P<0.01) during anticoagulation with dabigatran versus warfarin. Although no intervention has convincingly been shown to reduce the mortality of warfarin-associated intracranial bleeding, treatments differ worldwide,18 and it is unclear whether these results of the international RE-LY study apply to centers in which reversal of warfarin anticoagulation is aggressively undertaken in this setting.
In summary, in this cohort of older atrial fibrillation patients with well-controlled blood pressure, the clinical spectrum of intracranial hemorrhages was similar for patients given warfarin versus dabigatran, but with substantially lower absolute rates of all sites of intracranial hemorrhage and of traumatic intracranial hemorrhage with dabigatran. Almost half of intracranial hemorrhages were subdural hematomas, which were frequently associated with trauma. Fatal intracranial hemorrhages were substantially less frequent with dabigatran than with warfarin. Aspirin use independently predicted spontaneous intracerebral hemorrhage with warfarin. The underlying mechanism(s) accounting for the low risk of all sites of intracranial bleeding with dabigatran and other novel oral anticoagulants are critical to understand, but remain to be fully elucidated.
Sources of Funding
This study was funded by Boehringer Ingelheim.
P.A.R. is an employee of Boehringer Ingelheim, and all other coauthors have served as consultants, except S.Ya.
Louis Caplan, MD, was the Guest Editor for this paper.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.650614/-/DC1.
- Received January 11, 2012.
- Accepted February 22, 2012.
- © 2012 American Heart Association, Inc.
- Connolly S,
- Pogue J,
- Hart R,
- Pfeffer M,
- Hohnioser S,
- Chrolavicius S,
- et al
- Huhtakangas J,
- Tetri S,
- Juvela S,
- Saloheimo P,
- Bode MK,
- Hillbom M
- Hart RG,
- Tonarelli SB,
- Pearce LA
- Shireman TI,
- Howard PA,
- Kresowik TF,
- Ellerbeck EF