Intracranial Hemorrhage Mortality in Atrial Fibrillation Patients Treated With Dabigatran or Warfarin
Background and Purpose—In randomized trials, patients with atrial fibrillation (AF) receiving dabigatran, a direct oral anticoagulant, had lower risk of intracranial bleeding (ICB) than those on warfarin. However, concerns exist about potential worse outcomes in dabigatran users if bleeding occurs, given the lack of approved reversal agents. Thus, we examined in-hospital mortality in AF patients with ICB being treated with dabigatran versus warfarin in a real-world population in the United States.
Methods—We analyzed healthcare utilization claims in the Truven Health Marketscan Research Databases. The study sample included patients with AF admitted to a hospital with a primary diagnosis of ICB. Information on medications, inpatient, and outpatient diagnoses was obtained from available claims. Propensity score–adjusted risk ratios and 95% confidence intervals of in-hospital mortality comparing current users of dabigatran versus warfarin were estimated using relative risk regression.
Results—Among 2391 AF patients admitted with ICB (2290 on warfarin, 101 on dabigatran), 531 died during their admission. In-hospital mortality was similar in those treated with warfarin (22%) or dabigatran (20%). Compared with warfarin users, the propensity score–adjusted risk ratio (95% confidence interval) of mortality in dabigatran users was 0.93 (0.62–1.37). Associations were similar across different ICB subtypes (intracerebral hemorrhage, subarachnoid hemorrhage, and subdural hematoma).
Conclusions—In this sample of AF patients with ICB on oral anticoagulants, dabigatran was not associated with higher in-hospital mortality compared with warfarin. Hence, reluctance to use dabigatran because of a lack of approved reversal agents is not supported by our results.
Atrial fibrillation (AF) is a common cardiac arrhythmia associated with an increased risk of ischemic stroke and other cardiovascular diseases.1 Current guidelines for its treatment recommend chronic oral anticoagulation in patients with at least a moderate risk of ischemic stroke.2 Until recently, the only available drug for oral anticoagulation in AF was vitamin K antagonists (mostly warfarin in the United States). This landscape has changed in the past few years, with the approval by the Food and Drug Administration of 3 new oral anticoagulants for the prophylaxis of ischemic stroke and other cardioembolic complications. These new drugs—the direct thrombin inhibitor dabigatran and the direct factor X inhibitors rivaroxaban and apixaban—have been shown to be noninferior or superior to warfarin with respect to the prevention of stroke while being associated, in general, with lower rates of hemorrhage, particularly intracranial bleeding (ICB).3–5
In contrast to warfarin, the new oral anticoagulants lack approved, commercially available antidotes that could reverse their anticoagulant effect in case of acute hemorrhage (although there are several in different stages of development). This limitation has been highlighted as a major disadvantage of the new drugs.6 The main concern is that absence of specific reversal agents would increase the severity of any bleeding, resulting in higher mortality and overall worse outcomes. The lack of antidote is particularly troubling in the setting of ICB, possibly the most feared complication of anticoagulant treatment. The evidence addressing the severity of intracranial hemorrhages in patients using new oral anticoagulants, however, is limited. In a secondary analysis of the 150 intracranial hemorrhages occurring in the Randomized Evaluation of Long-Term Anticoagulation Therapy (RE-LY) trial, mortality was similar in patients randomized to receive dabigatran or warfarin.7 Similarly, receiving rivaroxaban or warfarin was not associated with mortality in the 172 intracranial hemorrhages identified in the Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation (ROCKET-AF) trial.8 These results, however, derive from a randomized trial and might not be directly applicable to less controlled environments in real-world populations. To date, data from other settings have been limited to case reports and several small case series of patients receiving new oral anticoagulants, without direct comparison with warfarin-treated patients.9
To provide additional evidence that could inform prescribing decisions for clinicians and patients, we studied the in-hospital mortality of AF patients with intracranial hemorrhages who were receiving dabigatran or warfarin in a real-world population, using a large healthcare utilization database.
We used claims data from the Truven Health MarketScan Commercial Claims and Encounters Database and the Medicare Supplemental and Coordination of Benefits Database (Truven Health Analytics Inc, Ann Arbor, MI) for the period January 1, 2009, to December 31, 2012 (the Food and Drug Administration–approved dabigatran for stroke prophylaxis in AF in October 2010). The MarketScan Commercial Database includes health insurance claims spanning all levels of care, as well as enrollment data from large employers and health plans across the United States, providing private healthcare coverage for employees, their spouses, and dependents. The MarketScan Medicare Supplemental Database includes claims from individuals and their dependents with Medicare supplemental coverage. Both databases link medical and outpatient prescription drug claims and encounter data with patient enrollment data to provide individual-specific clinical utilization, expenditure, and outcomes information across inpatient and outpatient services and outpatient pharmacy services.
Our analysis was restricted to individuals enrolled in plans with available outpatient pharmaceutical data, admitted to a hospital with a primary diagnosis of ICB, with ≥6 months of continuous enrollment before the ICB hospitalization and with a prior history of AF. We defined ICB as the first hospitalization with any of the following International Classification of Diseases, Ninth Revision, Clinical Modification codes as the primary diagnosis: 430.x, 431.x, 432.x, 852.x, and 853.x. Validation studies have demonstrated these codes to have high positive predictive value for intracranial hemorrhage.10 The admission date for this hospitalization was considered the index date. A history of AF was defined as ≥1 inpatient claim or 2 outpatient claims ≥1 week apart with the International Classification of Diseases, Ninth Revision, Clinical Modification codes 427.31 or 427.32 in any position before the index date.11 Patients with AF enrolled in the MarketScan Medicare Supplemental Database have similar demographic characteristics to patients with AF in the general fee-for-service Medicare population.12,13
All patient information was Health Insurance Portability and Accountability Act–compliant, deidentified, commercially available secondary data, and therefore, the Institutional Review Board at the University of Minnesota deemed this analysis exempt from review.
In both MarketScan databases, each individual outpatient pharmaceutical claim includes information on National Drug Code, date of service, and days supplied, among other variables. We identified all prescriptions for dabigatran and warfarin occurring before the index date (ICB hospitalization). Patients were considered current users of a specific anticoagulant at the index date if the ICB hospitalization occurred during an active prescription for oral anticoagulation. In a sensitivity analysis, we added to the current users those with an oral anticoagulant prescription ending in the 7-day period before the index date (ICB hospitalization). Only 12 ICB events (3 deaths) occurred in rivaroxaban users, and therefore they were excluded from the analysis.
Outcome and Other Covariates
The outcome of interest, in-hospital mortality, was directly obtained from the inpatient claims. Other covariates were ascertained from inpatient and outpatient claims before the index date using International Classification of Diseases, Ninth Revision, Clinical Modification codes and previously published algorithms.14,15 These covariates comprised those included in scores for the prediction of stroke (CHADS2 [congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolisms], CHA2DS2-VASc [congestive heart failure, hypertension, age 65-74, age ≥75, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolism, vascular disease, sex]) and hemorrhagic complications in patients with AF (ATRIA [anticoagulation and risk factors in atrial fibrillation], HEMORR2HAGES [hepatic or renal disease, ethanol abuse, malignancy, older age (>75), reduced platelet count or function, rebleeding risk, hypertension, anemia, genetic factors, excessive fall risk, stroke], HAS-BLED [hypertension, abnormal renal function, abnormal liver function, stroke, bleeding, labile international normalized ratios, elderly (age ≥65), drug therapy, alcohol intake]; see Tables I and II in the online-only Data Supplement for score definitions, diagnostic codes, and bibliographic references). Similarly, we assessed the presence of prescription for the following medication groups before the index date: antiplatelets, digoxin, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, type I and III antiarrhythmics, β-blockers, calcium channel blockers, lipid-lowering medications, and diuretics.
The primary analysis included current users of oral anticoagulants at the index date. We estimated the association between type of oral anticoagulant and in-hospital mortality using relative risk (log-binomial) regression (PROC GENMOD in the SAS software), comparing current users of dabigatran with current warfarin users.
In all analyses, an initial model was adjusted for age, sex, and hemorrhage subtype (intracerebral hemorrhage, subarachnoid hemorrhage, unspecified intracranial hemorrhage, nontraumatic subdural hemorrhage, traumatic subdural hemorrhage, unspecified traumatic intracranial hemorrhage). A second model additionally adjusted for CHA2DS2-VASc and ATRIA bleeding scores. We adjusted for these scores because, among the considered stroke and bleeding predictive models in AF, they were the strongest predictors of in-hospital mortality in our study sample. Third, we ran a model adjusted for age, sex, hemorrhage type, and deciles of propensity score (as a categorical variable). Propensity scores were calculated from a logistic regression model with current use of warfarin (versus dabigatran) as the dependent variable and all variables in Table 1 as the independent variables. A final analysis was performed in propensity score–matched warfarin and dabigatran users. Each dabigatran user was matched with ≤3 warfarin current users by index date (60 days caliper), age at index date (3 years caliper), sex, and propensity score (0.03 caliper).16 Analyses stratified by sex, age (below or above median), history of kidney disease, and ATRIA bleeding score (below or above median) were also conducted. Interactions were tested, including multiplicative terms in the models. We conducted 2 additional sensitivity analyses. First, we included patients who had been using their current anticoagulant for ≤6 months and, in a different model, for ≤1 year. Second, we restricted the analysis to nontraumatic intracranial bleeds.
We identified 2391 hospitalizations with a primary diagnosis of ICB among patients with AF who were current users of one anticoagulant at the time of the event (2290 were on warfarin, 101 on dabigatran) (Figure I in the online-only Data Supplement). Four additional events were excluded because of being current users of >1 anticoagulant. Patient characteristics by type of anticoagulant are presented in Table 1. Overall, no statistically significant differences in mean age, sex, or mean risk of ischemic stroke by the CHADS2 or CHA2DS2-VASc scores were found across groups. According to all bleeding scores, risk of hemorrhage was higher in dabigatran users compared with warfarin users. Dabigatran users had a higher prevalence of prior history of gastrointestinal bleeding.
Among the 2391 ICB events, 531 in-hospital deaths occurred (22% mortality). In-hospital mortality was similar in warfarin (22%) and dabigatran users (20%). Adjusting for age and sex, in-hospital mortality was similar in both groups (risk ratio, 0.89; 95% confidence interval, 0.60–1.33). Similarly, in multivariable and propensity score–adjusted or matched models, current use of dabigatran compared with current use of warfarin was not associated with mortality (Table 2). Results were comparable in sensitivity analyses adding to the current users those with their most recent oral anticoagulant prescription ending in the 7 days before the ICB hospitalization date (which added 187 cases; Table III in the online-only Data Supplement), restricting the analysis to patients with <6 months or 1 year on their current anticoagulant regime (Table IV in the online-only Data Supplement), and including only nontraumatic hemorrhages (Table V in the online-only Data Supplement). Hemorrhage subtype analyses were limited because of small numbers and imprecise estimates but suggested lower mortality associated with dabigatran use in subdural hemorrhage and higher mortality in intracerebral hemorrhage, compared with warfarin use (Table 3).
In stratified analysis (Table 4), associations between anticoagulant type and mortality were similar in younger and older patients, with or without kidney disease history, and with low or high bleeding risk assessed by the ATRIA score. A significant interaction with sex was observed (P=0.03): among men, mortality was lower in dabigatran users compared with warfarin users (risk ratio, 0.54; 95% confidence interval, 0.25–1.14), whereas the opposite pattern was observed for women (risk ratio, 1.49; 95% confidence interval, 0.96–2.29).
In this retrospective analysis of healthcare utilization data, in-hospital mortality in AF patients using oral anticoagulants hospitalized with ICB was similar in warfarin and dabigatran users. The lack of association is unlikely to be because of confounding by indication, given the extensive adjustment for clinical covariates. Our results do not support the notion that, in patients with ICB, use of dabigatran is associated with worse prognosis, assessed as in-hospital mortality, relative to use of warfarin.
The present analysis is based on a relatively small number of ICB events in current dabigatran users (101 cases with 20 deaths), but is consistent with reports from the RE-LY trial, where mortality in patients experiencing an ICB did not differ by treatment allocation. In the RE-LY trial, mortality was 38% and 36% in patients with ICB assigned to dabigatran and warfarin, respectively.7 Of note, mortality in the RE-LY trial was higher than that reported in the present analysis because of differences in mortality follow-up (in-hospital mortality in our analysis versus extended follow-up in RE-LY) and because of a higher proportion of intracerebral hemorrhages, which have worse prognosis, among all ICB events in the RE-LY trial population. No other studies have examined directly the outcomes and prognosis of intracranial hemorrhages in dabigatran versus warfarin users in real-world populations, outside the more controlled setting of clinical trials. Case reports and several small cases series of intracranial hemorrhages in dabigatran users have been published, mostly focusing on their clinical management, but they only provide limited evidence of the comparative effectiveness of dabigatran and warfarin.9
Our findings run counter to the belief that patients receiving dabigatran who experience severe bleeding may have worse outcomes than warfarin users, given the lack of approved reversal agents for the former.6 Several reasons could explain our results. First, in the setting of an ICB event, even the theoretical availability of reversal agents for vitamin K antagonists might not be enough to prevent the poor outcome of these episodes if their administration is delayed, leading to similar mortality rates. In addition, the half-life of dabigatran is shorter than that of warfarin, potentially reducing its impact in the setting of bleeding.17 Finally, evidence from animal models indicates that intracerebral hemorrhage volume may be lower and that active bleeding may terminate earlier in dabigatran-treated than warfarin-treated mice.18–20
The lack of association between type of oral anticoagulant and in-hospital mortality was observed across subgroups defined by age, history of kidney disease, or bleeding risk. However, we found evidence that sex modified the mortality risk associated with dabigatran versus warfarin. Specifically, compared with warfarin use, dabigatran use was associated with higher mortality in women but lower mortality in men. No obvious explanation can be provided for this difference. It is possible that metabolism of dabigatran may differ by sex; alternately, this observation could also be because of chance. Replication in independent samples is required.
Two additional considerations should be taken into account when interpreting our findings. First, if a particular anticoagulant is associated with severity of bleeding leading to higher out-of-hospital mortality, we would underestimate mortality in that group. Unfortunately, our data do not include information to test this hypothesis. Second, although all patients included in our analysis were current users of oral anticoagulation, it is possible that warfarin users have been using their medication for a longer time period than dabigatran users, survived other complications, and therefore might be a selected, more resilient group than the dabigatran users. As we have shown, however, characteristics of warfarin and dabigatran users were similar. Also, results did not change in analyses adjusting for time since first anticoagulant prescription.
Strengths of our study include the relatively large number of ICB events and the availability of an extensive array of clinical information for confounding adjustment. Some limitations need to be highlighted, however. Although previous studies have shown high specificity of claims for the diagnosis of intracranial hemorrhage,10 we did not have access to medical charts to validate the diagnosis. Similarly, we did not have information on bleeding-related prognostic variables (eg, bleeding location or size) or in the clinical management of these patients, including hemostatic approaches, which could have been different in warfarin and dabigatran users. Despite our efforts to capture and adjust for relevant potential confounding factors associated with mortality, residual confounding may have masked true differences in mortality between anticoagulant groups. In addition, our sample size may have been insufficient to identify relatively small differences in mortality between groups and to estimate precisely the associations within specific ICB subtypes (for instance, we only identified 25 intracerebral hemorrhages among dabigatran users). Last of all, our outcome was in-hospital mortality, and we did not have the possibility of assessing longer term mortality and functional outcomes.
In conclusion, despite the lack of approved reversal agents, in-hospital mortality among AF patients admitted with an ICB in a real-world setting in the United States was comparable in warfarin and dabigatran users. Our results may help clinicians and patients with AF make informed decisions, adequately balancing risk and benefits, when choosing an oral anticoagulant for the prevention of thromboembolic complications.
Sources of Funding
Research reported in this publication was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health Award Number UL1TR000114. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This work was additionally funded by a small grant from the University of Minnesota Academic Health Center.
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.006016/-/DC1.
- Received May 2, 2014.
- Accepted June 6, 2014.
- © 2014 American Heart Association, Inc.
- Magnani JW,
- Rienstra M,
- Lin H,
- Sinner MF,
- Lubitz SA,
- McManus DD,
- et al
- January CT,
- Wann LS,
- Alpert JS,
- Calkins H,
- Cleveland JC,
- Cigarroa JE,
- et al
- Ansell J
- Hart RG,
- Diener HC,
- Yang S,
- Connolly SJ,
- Wallentin L,
- Reilly PA,
- et al
- Hankey GJ,
- Stevens SR,
- Piccini JP,
- Lokhnygina Y,
- Mahaffey KW,
- Halperin JL,
- et al
- Piccini JP,
- Hammill BG,
- Sinner MF,
- Jensen PN,
- Hernandez AF,
- Heckbert SR,
- et al
- Won SY,
- Schlunk F,
- Dinkel J,
- Karatas H,
- Leung W,
- Hayakawa K,
- et al
- Lauer A,
- Cianchetti FA,
- Van Cott EM,
- Schlunk F,
- Schulz E,
- Pfeilschifter W,
- et al