Cumulative percentage of patients with major bleeding.
Douketis JD, Arneklev K, Goldhaber SZ, Spandorfer J, Halperin F, Horrow J. Comparison of Bleeding in Patients With Nonvalvular Atrial Fibrillation Treated With Ximelagatran or WarfarinAssessment of Incidence, Case-Fatality Rate, Time Course and Sites of Bleeding, and Risk Factors for Bleeding. Arch Intern Med. 2006;166(8):853-859. doi:10.1001/archinte.166.8.853
Ximelagatran is a novel direct thrombin inhibitor that can be administered as a fixed oral dose, without the need for anticoagulant monitoring.
We undertook a pooled analysis of 7329 patients with nonvalvular atrial fibrillation from the Stroke Prevention Using Oral Thrombin Inhibitor in Atrial Fibrillation III and V trials to compare bleeding outcomes in patients who received ximelagatran, 36 mg twice daily, or warfarin sodium (target international normalized ratio, 2.0-3.0). We determined annual risk of bleeding (any, major), case-fatality rate, time course and anatomic sites of major bleeding, and risk factors for major bleeding with ximelagatran and warfarin treatment.
Annual incidence of any bleeding was 31.75% with ximelagatran and 38.82% with warfarin (relative risk reduction, 18.2%; 95% confidence interval [CI], 13.0-23.1; P<.001). Annual incidence of major bleeding was 2.01% with ximelagatran and 2.68% with warfarin (relative risk reduction, 25.1%; 95% CI, 3.2-42.1; P = .03). Case-fatality rate of bleeding was comparable in ximelagatran- and warfarin-treated patients (8.16% vs 8.09%; P = .98). Cumulative incidence of major bleeding was higher with warfarin than ximelagatran after 24 months of treatment (4.7% vs 3.7%; P = .04). Anatomic sites of bleeding were comparable with both treatments. Risk factors for bleeding with ximelagatran were as follows (hazard ratios and 95% CIs in parentheses): diabetes mellitus (1.81; 1.19-2.77; P = .006), previous stroke or transient ischemic attack (1.78; 1.16-2.73; P = .008), age 75 years or greater (1.70; 1.33-2.18; P<.001), and aspirin use (1.68; 1.08-2.59; P = .02). Risk factors for bleeding in warfarin-treated patients were previous liver disease (4.88; 1.55-15.39; P = .007); aspirin use (2.41; 1.69-3.43; P<.001); and age 75 years or greater (1.26; 1.03-1.52; P = .02).
Treatment with ximelagatran, 36 mg twice daily, is associated with a lower risk of bleeding than warfarin in patients with nonvalvular atrial fibrillation. Aspirin use and increasing age were associated with an increased risk of bleeding in ximelagatran- and warfarin-treated patients.
In patients with nonvalvular atrial fibrillation, conventional therapy uses warfarin sodium to prevent intracardiac thrombosis and stroke.1 Ximelagatran is the first of a new class of oral anticoagulants, direct thrombin inhibitors, that bind to circulating and clot-bound thrombin.2 The Stroke Prevention Using Oral Thrombin Inhibitor in Atrial Fibrillation (SPORTIF) III and V trials found that fixed-dose ximelagatran, 36 mg twice daily, prevents stroke as well as dose-adjusted warfarin (target international normalized ratio [INR], 2.0-3.0) in patients with nonvalvular atrial fibrillation.3- 5
Managing warfarin therapy is problematic because of the need for anticoagulant monitoring, with INR testing required every 1 to 4 weeks for dose adjustments.6 Warfarin also interacts with other drugs, food, genetic polymorphisms, and acute illness, with the potential to cause excessive anticoagulation and bleeding.7- 11 Ximelagatran therapy, on the other hand, is administered as a fixed dose, without the need for laboratory monitoring; has a low potential for interactions with drugs or food; and is not known to be affected by genetic polymorphisms.12 These properties might reduce the likelihood of an excessive anticoagulant effect and the potential for bleeding compared with warfarin. Furthermore, differences in pharmacologic properties between ximelagatran and warfarin may also result in different predictors of bleeding with each drug.
Individually, the SPORTIF III and V trials were not primarily designed to compare the incidence of bleeding with ximelagatran and warfarin therapy, in part because intracerebral bleeding episodes were counted as strokes, irrespective of their being ischemic or nonischemic.5 Furthermore, these studies did not compare the case-fatality rate, time course and anatomic sites of bleeding, and risk factors for bleeding with ximelagatran and warfarin therapy. To address these issues, which were prespecified before the completion of the SPORTIF trials, we undertook a pooled analysis of patients with nonvalvular atrial fibrillation who participated in the SPORTIF III and V trials.
The study sample of the SPORTIF III and V trials consisted of 7329 patients with nonvalvular atrial fibrillation. The patient inclusion and exclusion criteria were the same for both trials and are described elsewhere.3,4 In brief, eligible patients were adults, aged 18 years or more, with permanent or paroxysmal nonvalvular atrial fibrillation confirmed by at least 2 electrocardiographic recordings, and with 1 or more of the following risk factors for stroke: hypertension (increased blood pressure requiring antihypertensive therapy but <180/100 mm Hg); age 75 years or greater; previous stroke, transient ischemic attack, or systemic embolism; left ventricular dysfunction (left ventricular ejection fraction <40% or symptomatic congestive heart failure); age 65 years or greater and coronary artery disease; and age 65 years or greater and diabetes mellitus. The major exclusion criteria were mitral stenosis; previous valvular heart surgery; transient atrial fibrillation caused by reversible disorders; severe renal insufficiency (calculated creatinine clearance, <30 mL/min [<0.50 mL/s]); active liver disease or serum liver enzyme levels 2 times the upper limit of normal or greater; and conditions associated with an increased risk of bleeding (eg, active peptic ulcer disease).
In SPORTIF III and V, patients were randomly allocated to receive fixed-dose oral ximelagatran, 36 mg twice daily, or dose-adjusted warfarin (target INR, 2.0-3.0). In SPORTIF III, treatment was open-label, with INR measurements obtained in the warfarin-treated patients at least every month and with dose adjustments done according to local clinical practice. In SPORTIF V, treatment was administered in a double-blind, double-dummy manner, where patients received active drug and matching placebo tablets. Patients in SPORTIF V underwent blood testing to provide a real INR (warfarin arm) or a sham INR (ximelagatran arm). For all patients, warfarin and placebo doses were adjusted according to local clinical practice. In SPORTIF III, the duration of treatment was up to 25 months (median, 17.7 months). In SPORTIF V, the duration of treatment was up to 31 months (median, 17.8 months). Concomitant aspirin therapy was permitted, in dosages of 100 mg/d or less, but other antithrombotic drugs were prohibited. Patients were allowed intermittent use of nonsteroidal anti-inflammatory drugs.
For this pooled analysis, bleeding was classified as major if it satisfied at least 1 of the following criteria: fatal bleeding; clinically overt bleeding associated with a reduction in hemoglobin level of 20 g/L or more; clinically overt bleeding requiring transfusion of 2 or more units of whole blood or erythrocytes; intracerebral bleeding; and bleeding involving a critical anatomic site (intracranial [not intracerebral], intraspinal, intraocular, retroperitoneal, pericardial, or atraumatic intra-articular). Intracerebral bleeding was defined as any bleeding that occurred within the cerebrum, cerebellum, brainstem, thalamus, or basal ganglia, irrespective of cause (ischemic or nonischemic). All other bleeding episodes that did not satisfy the criteria for major bleeding were classified as minor. An independent and blinded Central Event Adjudication Committee reviewed all major bleeding episodes.
We performed an exploratory analysis to identify conventional and novel risk factors for bleeding in ximelagatran-treated, warfarin-treated, and all patients, irrespective of treatment. Conventional candidate risk factors for bleeding assessed were greater age, lower creatinine clearance, previous gastrointestinal or genitourinary tract bleeding, previous peptic ulcer disease, previous liver disease, malignant disease, chronic cardiovascular disease, and concomitant aspirin, antiplatelet, or nonsteroidal anti-inflammatory drug use.13- 19 Novel candidate risk factors for bleeding assessed were race, body weight, body mass index, type of atrial fibrillation (permanent vs paroxysmal), and other concomitant drug use (diuretics, digoxin, β-blockers, calcium-channel blockers, angiotensin-converting enzyme inhibitors, or statins). Excessive warfarin anticoagulation was not considered a candidate risk factor for bleeding because this was relevant only for warfarin-treated patients. Furthermore, because INR testing frequency was not fixed, other than that it had to be done at least monthly, patients who tended to have excessive anticoagulation (INR >3.0) might have more frequent INR testing than patients without excessive anticoagulation, with the potential for preferential (biased) reporting of excessive anticoagulation in patients with bleeding than in those without bleeding.
We determined the annual incidence of any (major or minor), major, and intracerebral bleeding with ximelagatran and warfarin therapy during the study period, based on the time to first bleeding episode while patients were treated. Thus, no patient was counted as having more than 1 bleeding outcome, and no patient was counted as having a bleeding episode if it occurred off treatment, which was defined as not having received study medication for 30 consecutive days or 60 total days. The same approach was used for analyses of case-fatality rate, time course, anatomic site of major bleeding, and risk factors for bleeding.
The incidence of bleeding with ximelagatran and warfarin therapy was compared by Fisher exact test. This method uses the number of patient-years of treatment as the denominator and assumes a constant rate of bleeding over time. This comparison was also done with the log-rank test, which allows bleeding rate to vary over time. The incidence of bleeding was expressed as the number of bleeding episodes per year. The case-fatality rate of major bleeding was defined as the number of patients with fatal bleeding divided by the number of patients with either fatal or nonfatal major bleeding and was expressed as a percentage. The Kaplan-Meier product-limit technique estimated incidences of bleeding at 3, 6, 12, 18, and 24 months of treatment; the log-rank test compared incidences of bleeding with ximelagatran and warfarin therapy. The anatomic sites of major bleeding with ximelagatran and warfarin therapy were compared by Fisher exact test.
Univariate and multivariate stepwise forward Cox regression analyses were used to identify factors that were independently associated with an increased risk of major bleeding. The hazard ratio for major bleeding, and corresponding 95% confidence interval, was determined for each risk factor (variable) in the Cox regression model. Variables in the final multivariate regression models were selected on the basis of their achieved level of statistical significance, defined as P<.05.
The clinical characteristics of 3664 patients randomized to ximelagatran and 3665 patients randomized to warfarin are shown in Table 1. Patients who received ximelagatran or warfarin were comparable in terms of baseline clinical characteristics.
As shown in Table 2, in ximelagatran-treated patients, there were 1217 total (major or minor), 98 major, and 7 intracerebral bleeding episodes; in warfarin-treated patients, there were 1453 total, 136 major, and 11 intracerebral bleeding episodes. The annual incidence of any (major or minor) bleeding was 31.75% with ximelagatran therapy and 38.82% with warfarin therapy (relative risk reduction, 18.2%; 95% confidence interval, 13.0-23.1). The annual incidence of major bleeding was 2.01% with ximelagatran therapy and 2.68% with warfarin therapy (relative risk reduction, 25.1%; 95% confidence interval, 3.2-42.1). The annual incidence of intracerebral bleeding was 0.12% with ximelagatran therapy and 0.19% with warfarin therapy (relative risk reduction, 36.8%; 95% confidence interval, −75.0 to 61.0). These results assume a constant event rate. Repeat analyses without this assumption yielded similar results for all categories of bleeding: there was a significantly lower incidence of any bleeding (P<.001) and major bleeding (P = .03) with ximelagatran compared with warfarin therapy; there was no significant difference in intracerebral bleeding (P = .35) with ximelagatran and warfarin therapy.
Ximelagatran therapy was associated with a 0.67% absolute risk reduction in major bleeding per year compared with warfarin therapy. This equates to a number needed to treat of 149 patients treated with ximelagatran instead of warfarin for 1 year to prevent 1 major bleeding episode. Ximelagatran therapy was associated with a 7% absolute risk reduction in any bleeding compared with warfarin therapy. This corresponds to a number needed to treat of 14 patients treated with ximelagatran instead of warfarin for 1 year to prevent 1 bleeding episode.
As shown in Table 3, the case-fatality rate of major bleeding was 8.16% with ximelagatran and 8.09% with warfarin (P = .98). The case-fatality rate of intracerebral bleeding was 71.4% with ximelagatran and 45.4% with warfarin (P = .37).
The time course of major bleeding with ximelagatran and warfarin therapy is shown in the Figure, and the cumulative incidence of major bleeding after 3, 6, 12, 18, and 24 months of treatment is shown in Table 4. The cumulative incidences of major bleeding during treatment with ximelagatran and warfarin diverged with time, reaching a statistically significant difference at 24 months in favor of ximelagatran (P = .04).
The anatomic sites of major bleeding, as shown in Table 5, were comparable with ximelagatran and warfarin therapy.
Potential predictors of major bleeding were identified in the univariate analysis, as shown in Table 6. In ximelagatran-treated patients, risk factors for major bleeding, as shown in Table 7, were age 75 years or greater, previous stroke or transient ischemic attack, diabetes mellitus, and aspirin use. Use of 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins) was associated with a decreased risk of bleeding. In warfarin-treated patients, risk factors for major bleeding, as shown in Table 7, were age 75 years or greater, history of liver disease, and aspirin use. Statin use was associated with a decreased risk of bleeding. In the combined study sample, risk factors for bleeding, as shown in Table 7, were age 75 years or greater, Asian race, left ventricular dysfunction, diabetes mellitus, and aspirin use. Statin use was associated with a decreased risk of bleeding. Patients in SPORTIF III had a lower risk of bleeding than those in SPORTIF V, and treatment with ximelagatran was associated with a lower risk of bleeding.
The principal finding in this study is that, in patients with nonvalvular atrial fibrillation, treatment with oral ximelagatran, 36 mg twice daily, is associated with significantly less major bleeding than treatment with warfarin (target INR, 2.0-3.0). The validity of this finding is strengthened because all major bleeding episodes included in the analysis were based on standardized criteria and were reviewed by independent and blinded adjudication committees. Furthermore, the lower risk of major bleeding in ximelagatran-treated than in warfarin-treated patients was maintained in the multivariate regression analysis that assessed independent determinants of bleeding in the combined study sample.
Although an annual absolute risk reduction in major bleeding of 0.67% with ximelagatran appears modest, it is clinically important because atrial fibrillation is common, with more than 2 million prevalent cases and about 200 000 incident cases diagnosed annually in North America.20 Thus, for every 10 000 patients treated with ximelagatran instead of warfarin, about 70 major bleeding episodes might be prevented, including 5 to 6 fatal episodes (based on an 8% case-fatality rate of major bleeding). These potential benefits of ximelagatran therapy should be weighed against potential risks of treatment, which include elevated serum liver enzyme levels in approximately 6% of patients with atrial fibrillation and, rarely, severe and possibly fatal hepatitis.3,4
There are other noteworthy findings in comparing bleeding with ximelagatran and warfarin therapy. The case-fatality rate of major bleeding was approximately 8% with both ximelagatran and warfarin therapy, and is consistent with the 9% case-fatality rate of major bleeding observed in warfarin-treated patients with venous thromboembolism.21 The high case-fatality rate of intracerebral bleeding is consistent with studies involving warfarin-treated patients with venous thromboembolism or a mechanical heart valve, in whom the case-fatality rate was 40% to 50%.22,23 The gastrointestinal tract was the most common site of bleeding with both ximelagatran and warfarin therapy, and is consistent with other studies involving patients who were receiving warfarin.13- 15 This finding emphasizes the need for attention to risk factors for gastrointestinal tract bleeding before the start of anticoagulant therapy.
Aspirin use was the only modifiable risk factor associated with increased bleeding with both ximelagatran and warfarin therapy. This finding is not unexpected because concomitant use of multiple drugs that impair hemostasis increases the risk of bleeding.24- 26 Use of nonsteroidal anti-inflammatory drugs, although included as a candidate risk factor, was not associated with an increased risk of bleeding in the univariate analysis. Increasing age was the other risk factor for bleeding that occurred in both ximelagatran- and warfarin-treated patients, and has been observed in other studies.13- 15,27 Unexpectedly, concomitant statin therapy was associated with an approximately 40% decreased risk of major bleeding with both ximelagatran and warfarin therapy. This finding appears counterintuitive, as statins have antiplatelet and other antithrombotic properties that might promote bleeding and requires further study.28,29 Asian patients had a 2-fold greater risk of major bleeding than did non-Asian patients. The reason for this finding is not clear and may be related to differences in patient characteristics in study samples, but it is consistent with a high bleeding risk reported in patients of Chinese origin who were receiving anticoagulants.30 Finally, patients in SPORTIF III had a lower risk of bleeding than did those in SPORTIF V. This finding is likely due to differences in patient characteristics across studies: patients in SPORTIF V were more likely to be 75 years or older (42.3% vs 33.6%; P<.001), more likely to have a history of gastrointestinal or genitourinary tract bleeding (8.0% vs 4.3%; P<.001), and more likely to use aspirin (24.2% vs 18.4%; P<.001).
There are potential limitations of this study. First, the study sample was pooled from 2 studies that used different designs to compare the interventions (open-label design in SPORTIF III, double-blind design in SPORTIF V). Although the study designs were otherwise identical, there is potential for biased reporting of outcomes in the open-label study. This is unlikely to have influenced our findings, because in both studies, major bleeding outcomes were based on standardized criteria and were reviewed by independent and blinded adjudication committees that were unaware of patients' treatment allocation. Furthermore, when bleeding outcomes with ximelagatran and warfarin therapy across the pooled studies were compared, there was no significant heterogeneity of results (P = .59). Second, the inclusion of intracerebral bleeding episodes as major episodes might have exaggerated the number of major bleeding episodes if some intracerebral episodes initially were ischemic strokes with subsequent hemorrhagic transformation.31 However, when intracerebral bleeding episodes were excluded from the comparison of bleeding outcomes, the annual risk of bleeding remained lower with ximelagatran than with warfarin therapy (1.88% vs 2.46%), although the difference was not quite statistically significant (P = .054). Third, the risk factors for bleeding we identified express associations between putative risk factors and bleeding, but do not provide evidence of causality.32 We acknowledge that identifying bleeding risk factors is dependent on the baseline characteristics and cointerventions within the patient sample studied, and our findings do not preclude additional risk factors for anticoagulant-related bleeding, particularly in different patient samples. Thus, previous gastrointestinal tract bleeding and peptic ulcer disease were risk factors for bleeding in other studies13,16 but not in this study. Although across-study differences in patient characteristics might account for these discrepant findings, it is possible that increased use of proton pump inhibitors and anti–Helicobacter pylori therapy in recent years might weaken an association between previous gastrointestinal tract bleeding or peptic ulcer disease and the risk of anticoagulant-related bleeding.33,34
There are 2 main clinical implications of this study. First, in anticoagulated patients with nonvalvular atrial fibrillation, ximelagatran was associated with a lower risk of bleeding. It is likely that the superior safety margin for bleeding with ximelagatran vs warfarin was underestimated because of the high quality of anticoagulation control in warfarin-treated patients.3,4 In the SPORTIF III and V trials, excessive anticoagulation (INR >4.0) occurred during 1.5% of the period of warfarin therapy, whereas in a real-world setting, excessive anticoagulation occurs during 5% to 10% of the period of warfarin therapy.35- 37 Furthermore, minimizing excessive anticoagulation would also reduce anticoagulant-related mortality, which is more likely to occur in patients with excessive than nonexcessive anticoagulation with warfarin therapy.11,38 Second, in patients with nonvalvular atrial fibrillation who are receiving warfarin therapy, avoidance of aspirin might be considered to minimize the risk of bleeding, particularly in patients who have additional risk factors for bleeding or those who have developed bleeding during combined treatment with aspirin and an oral anticoagulant.
To summarize, in patients with nonvalvular atrial fibrillation who require long-term anticoagulation, treatment with oral ximelagatran, 36 mg twice daily, is associated with a lower risk of bleeding complications than is treatment with warfarin (target INR, 2.0-3.0). Aspirin use and increasing age were associated with an increased risk of bleeding with both ximelagatran and warfarin therapy.
Correspondence: James D. Douketis, MD, St Joseph's Hospital, Room F-541, 50 Charlton Ave E, Hamilton, Ontario, Canada L8N 4A6 (firstname.lastname@example.org).
Accepted for Publication: October 6, 2005.
Financial Disclosure: None.
Acknowledgment: We acknowledge the contributions of members of the SPORTIF Executive Steering Committee to all phases of study implementation and for their review of this manuscript.