Current issues of ACP Journal Club are published in Annals of Internal Medicine


Stroke prevention by aspirin or warfarin for atrial fibrillation

ACP J Club. 1991 Nov-Dec;115:66. doi:10.7326/ACPJC-1991-115-3-066

Source Citation

Stroke Prevention in Atrial Fibrillation Investigators. Stroke prevention in atrial fibrillation study. Final results. Circulation. 1991 Aug;84:527-39.



To determine the effectiveness of warfarin and aspirin for preventing stroke and systemic embolism in patients with nonrheumatic atrial fibrillation.


Randomized controlled trial.


15 centers in the United States.


1330 patients with electrocardiographically documented atrial fibrillation but neither prosthetic heart valves nor mitral stenosis. Consenting patients with no requirements or contraindications for warfarin or aspirin were placed in group 1 (n = 627), and patients with contraindications for warfarin but not aspirin were placed in group 2 (n = 703). The mean duration of atrial fibrillation at time of entry was 4.9 ± 7.1 (SD) years.


Group 1 patients were randomized to one of three groups: open-label warfarin (at a dose sufficient to prolong prothrombin times to international normalized ratios [INR] of 2.0 to 4.5, n = 210), enteric aspirin (325 mg/d, n = 206), or placebo (n = 211)—the latter two in a double-blind fashion. Group 2 patients were randomized (double-blind) to either enteric aspirin (n = 346) or placebo (n = 357). Follow-up occurred every 3 months, for a mean duration of 1.3 years.

Main outcome measures

Primary outcome events were ischemic stroke or systemic embolism. Secondary events were death, myocardial infarction, transient ischemic attack, or unstable angina pectoris requiring hospitalization. All reports of events were verified by a central committee unaware of treatment allocation.

Main results

With respect to the efficacy of warfarin, in group 1, patients on warfarin had fewer primary events than those on placebo (2.3% vs 7.4% per year, P = 0.01; relative risk reduction, 67%; 95% CI 27% to 85%); for primary events or death from any cause, the relative risk reduction from warfarin was 58% (CI 20% to 78%, P = 0.01). In the combined groups 1 and 2, fewer patients on aspirin had primary events than those on placebo (3.6% vs 6.3% per year, P = 0.02; relative risk reduction, 42%; CI 9% to 63%); for primary event or death from any cause, the relative risk reduction from aspirin was 32% (CI 7% to 50%, P = 0.02). Relative risk reductions for disabling ischemic stroke or vascular death were 54% for warfarin (P= 0.11) and 22% for aspirin (P > 0.2). Rates of major complications from treatment were less than 1 % in all treatment arms.


Either warfarin or aspirin reduces the risk of ischemic stroke and systemic embolism in patients with nonrheumatic atrial fibrillation.

Source of funding: National Institute of Neurological Disorders and Stroke.

Address for article reprint: Ms. R McBride, Statistics and Epidemiology Research Corporation, 1107 NE 54th Street, Suite 520, Seattle, WA 98105.


This is the third nonblinded, randomized, controlled trial of warfarin versus placebo in persons with nonvalvular atrial fibrillation to be published since 1989 (1, 2). Each of these methodologically sound studies has shown the superiority of warfarin over placebo in preventing thromboembolic events. Primary outcome events were reduced by 67% per year in the SPAF trial, 64% per year in the AFASAK trial (1), and 86% per year in the BAATAF trial (2). There was a small, but clinically acceptable, increase in important hemorrhagic events among persons taking warfarin in each study.

Two of these trials also compared aspirin with warfarin or placebo, but these results are conflicting. In the AFASAK trial no beneficial effect was seen in persons taking aspirin, 75 mg a day. The SPAF trial reported a 42% risk reduction in patients taking aspirin, 325 mg a day. The direct comparison of aspirin and warfarin in this study is not yet available.

Warfarin can be expected to reduce the risk for stroke in persons with chronic, nonvalvular atrial fibrillation from about 6% to 2% per year. Based on current information, most persons with nonvalvular atrial fibrillation who can safely be anticoagulated should be treated with warfarin adjusted to an INR between 2 and 3. At this time aspirin should be reserved for those persons who will not or cannot safely take warfarin.

James M. Kitchens, MD
Toronto General Hospital Toronto, Ontario