Antithrombotic Therapy To Prevent Stroke in Patients with Atrial Fibrillation

Nonvalvular atrial fibrillation is an important independent risk factor for stroke. Since 1989, 16 published clinical trials have conducted 36 separate randomized comparisons of antithrombotic agents in approximately 10 000 participants with atrial fibrillation (1-17). Previously published meta-analyses and pooled analyses of individual patient data (18-20) have considered, in various combinations, the first 6 clinical trials to be published. We present a meta-analysis of all currently available trials to further characterize the comparative efficacy and safety of antithrombotic therapy for the prevention of stroke in patients with atrial fibrillation. Methods Randomized trials testing long-term (>3 months) use of antithrombotic agents in patients with atrial fibrillation were sought by a computerized search of the OVID/MEDLINE databases (from 1966 to 1999, not restricted by language) and by inquiries to the Cochrane Collaboration Stroke Review Group and Antithrombotic Trialists Collaboration. Studies of atrial fibrillation associated with prosthetic cardiac valves or mitral stenosis were not considered; trials reporting results for subgroups of participants with atrial fibrillation among other participants without atrial fibrillation were included (14, 16). Double-blind and nonblinded trials were included, and sensitivity analysis was used to compare pooled results, as appropriate. We excluded one randomized trial in which results for participants with atrial fibrillation (approximately half of all participants) were not reported separately (21). We have anecdotal knowledge of two additional trials that are ongoing or have not been published (Table 1). Table 1. Randomized Clinical Trials for Nonvalvular Atrial Fibrillation Two reviewers independently extracted data from published sources on the number of patients treated, total follow-up exposure, and the occurrence of five outcomes by intention-to-treat analysis: all stroke (hemorrhagic and ischemic), ischemic stroke, intracranial hemorrhage, all-cause mortality, and major extracranial bleeding. The criteria for each of these outcomes were those used in each individual trial; major deviations are noted in footnotes to the tables. Intracranial hemorrhage included spontaneous and traumatic subdural hematomas. Transient ischemic attacks (TIAs) were not considered. The percentage of participants who underwent neuroimaging or autopsy to reliably distinguish between ischemic or hemorrhagic stroke varied, and this percentage was not always reported. Therefore, all stroke (hemorrhagic and ischemic) was chosen as the primary outcome. Effects of antithrombotic therapy on combinations of events (for example, stroke, myocardial infarction, or vascular death) and effects of age and sex were not consistently reported and are not considered here. Intention-to-treat results were used for the main analyses, but only results noted during treatment were available for one small trial that tested antiplatelet agents (14). Primary prevention refers to patients without previous stroke or TIA; secondary prevention refers to patients with previous stroke or TIA. Detailed consideration of the design and execution of individual trials can be found in our systematic reviews prepared for the Cochrane Collaboration Stroke Review Group (16, 22). The authors participated in the Stroke Prevention in Atrial Fibrillation (SPAF) I, II, and III clinical trials (1987 to 1999), which were funded by the National Institute of Neurologic Disorders and Stroke. Meta-analyses of the results of the trials are presented as relative risk reductions and absolute risk reductions for treatment groups compared with control groups. To estimate the relative risk reduction, the combined odds ratio was computed by using the modified Mantel-Haenszel (Peto) method (23), and the estimate was then subtracted from 1. For a study in which no events were observed for a specific outcome and treatment, 0.5 was added to the empty cell. The absolute risk reduction is a weighted estimate of the difference in annualized event rates (24). Before we estimated risk reduction, we tested the assumption of the statistical homogeneity of the treatment effect (across trials and within a specific scenario) by using the QL statistic for the relative odds scale (24) or the QW statistic with unequal weights for the absolute risk scale (24). Lack of homogeneity across trials precluded estimation of the treatment effect, as noted. Homogeneity was assessed for all meta-analyses of relative risk reduction and absolute risk reduction; the exact P value was reported for all main analyses and for analyses in which the P value was less than 0.2. Estimates of relative risk reduction in individual trials were computed by subtracting the estimated odds ratio from 1. We estimated absolute risk reductions in individual trials by calculating the absolute difference in annualized event rates (events per person-years of observation). A P value less than 0.05 was considered statistically significant; all tests and CIs are two-sided. Calculations were done by using SPSS software (SPSS, Inc., Chicago, Illinois) and EGRET software (Cytel Software Corp., Cambridge, Massachusetts). Results Sixteen randomized trials published between 1989 and 1999 included 9874 participants with nonvalvular atrial fibrillation, with 2239 participants assigned to placebo (Table 1) (1-16). Total reported exposure was about 16 400 person-years (mean follow-up, 1.7 years). Fourteen trials included only participants with atrial fibrillation (1-13, 15); 2 trials included participants with atrial fibrillation in larger trials of secondary stroke prevention (14, 16). Warfarin was used exclusively in 10 trials testing oral anticoagulant agents; other derivatives of coumarin were also used in 2 additional trials (9, 15). Adjusted-Dose Warfarin Compared with Placebo In six trials involving 2900 patients with a total of 186 strokes, anticoagulation with oral vitamin K antagonists was compared with placebo (in five trials [1, 3, 7-9]) or control (in one trial [6]) (Table 2). The mean age of participants at study entry was 69 years (approximately 20% of participants were>75 years of age). Twenty-nine percent of participants were women, 45% had hypertension, and 20% had previous stroke or TIA. The target range for the international normalized ratio (INR) varied (Table 2); the mean achieved INR ranged from 2.0 to 2.6 in the five primary prevention trials and was 2.9 in the single secondary prevention trial. Mean duration of follow-up ranged from 1.2 to 2.3 years (overall average, 1.6 years per participant). The rate of stroke among participants who were not assigned to receive anticoagulation averaged 4.6% per year for primary prevention and 12.3% per year for secondary prevention. Four of these six trials were stopped at an interim analysis because of treatment efficacy (1, 3, 6, 8). Table 2. Adjusted-Dose Warfarin Compared with Placebo Meta-analysis showed that therapy with adjusted-dose warfarin reduced the relative risk for stroke by 62% (95% CI, 48% to 72%). This effect was statistically significant in four individual trials (3, 6, 8, 9) (Figure). The pooled result of primary prevention trials (59% reduction) was similar to that from the secondary prevention trial (68% reduction); results from the two double-blind trials (58% reduction) (7, 8) and the four open-label trials (63% reduction) (1, 3, 6, 9) were also similar. Warfarin was associated with similar relative risk reductions in disabling stroke (59%) and nondisabling stroke (61%). The absolute risk reduction for all stroke was 2.7% per year (number needed to treat [NNT] for 1 year to prevent one stroke, 37) for primary prevention and 8.4% per year (NNT, 12) for secondary prevention. Figure. Effects on all stroke (ischemic and hemorrhagic) of therapies for patients with atrial fibrillation. Top. Middle. Bottom. When only ischemic strokes were considered, treatment with adjusted-dose warfarin was associated with a 65% (CI, 52% to 74%) relative risk reduction. Twice as many intracranial hemorrhages were identified in participants who were assigned to receive warfarin (six compared with three), but the difference was not statistically significant. The rate of intracranial hemorrhage averaged 0.3% per year during anticoagulation and was 0.1% per year with placebo. The rate of major extracranial hemorrhage averaged 0.6% per year in patients who received placebo. The relative risk for major extracranial hemorrhage was 2.4 (CI, 1.2 to 4.6; absolute risk increase, 0.3% per year) for participants who received warfarin (excluding one trial in which almost half of the control group took aspirin [6]). All-cause mortality decreased in participants who received warfarin (relative risk reduction, 26% [CI, 4% to 43%]; P>0.2 for homogeneity; absolute risk reduction, 1.6% per year). Antiplatelet Therapy Compared with Placebo Six trials compared antiplatelet therapy with placebo (1, 3, 9, 13, 14, 16). In these trials, 3337 participants experienced a total of 376 strokes while being randomly assigned to receive antiplatelet therapy or placebo (Table 3). In five trials (1, 3, 9, 14, 16), assignment was double-blind. Approximately 90% of total follow-up exposure during antiplatelet therapy was with aspirin alone. Aspirin dosage ranged from 25 mg twice daily to 1300 mg/d. On the basis of available data from the three largest trials (1, 3, 9), the mean age of participants was 70 years (about 33% of participants were>age 75 years). Thirty-eight percent of participants were women, 46% had hypertension, and 40% had previous stroke or TIA. Mean duration of follow-up in these trials ranged from 1.2 to 4 years (overall average, 1.5 years per participant). The average rate of stroke among participants assigned to placebo was 5.2% per year for primary prevention and 12.9% per year for secondary prevention. Table 3. Antiplatelet Agents Compared with Placebo Meta-