Patients with nonvalvular atrial fibrillation at low risk of stroke during treatment with aspirin 

Stroke Prevention in Atrial Fibrillation III study

Authors The SPAF III writing committee for the Stroke Prevention in Atrial Fibrillation investigators. 
Source JAMA. 279:1273-7. April 22/29, 1998. 
Institutions Multi-institutional in the U.S. and Canada. 
Support National Institute of Neurologic Disorders and Stroke, Bethesda, Maryland. 


Atrial fibrillation is a common condition, which carries with it a significant risk for stroke. Treatment with adjusted-dose warfarin significantly reduces this risk, but at the price of an increased risk for hemorrhage. Treatment with aspirin also reduces the risk for stroke, but less effectively than warfarin. 

Atrial fibrillation patients are a heterogeneous group with varying degrees of risk for CVA. Treatment with aspirin could be an attractive modality for those at low risk for CVA. A number of risk factors for stroke in atrial fibrillation have been identified. This study, a subgroup of the SPAF study, was designed to evaluate the stroke risk in a group of patients without several risk factors, who were prospectively treated with aspirin only. 



Patients were eligible if they satisfied all of the following criteria: 

  • Documented atrial fibrillation within the preceeding 6 months (paroxysmal or chronic). 

  • They would not be classified as "lone atrial fibrillation" (less than 60 years old, without a history of hypertension or cardiovascular disease and with a normal echocardiogram). 

  • They did not have a prosthetic heart valve, mitral stenosis, another indication for anticoagulation or a contraindication to aspirin therapy. 

  • They did not have any of the following 4 risk factors, which would have placed them at higher risk for thromboembolism: 

    • Impaired LV function, judged by a reduced fractional shortening (<25%) on echocardiogram or recent clinical CHF. 

    • Active hypertension (BP greater than 160/90 on two separate days). A history of hypertension was not an excluding risk factor. 

    • Prior stroke, transient ischemic attack or arterial embolism. 

    • Women older than 75 years of age. 

All patients who were eligible and participated in the study received enteric-coated aspirin, 325 mg daily. 

Patients were recruited at outpatient clinics in the United States and Canada. They were followed up by clinic visits every six months and a telephone contact between clinic visits. At clinic visits blood pressure was determined. Patients were withdrawn from the study if any of the four risk factors developed (congestive heart failure, hypertension, emboli/CVA/TIA or women reaching age 76). Patients withdrawn from the study continued to be followed up for events; their further treatment (in particular warfarin therapy) was up to their personal physicians. 


The primary endpoint was the rate of ischemic strokes and systemic embolism. Other pre-specified endpoints were disabling ischemic strokes (modified Rankin score of II or greater at 1 to 3 months) and a stratification of outcomes according to the presence or absence of a history of hypertension. 

A number of other outcomes were also reported, including TIA's, cerebral and other hemorrhage, vascular events. 

The hypothesis was that these patients would have a primary event rate under 3% per year, and the sample size was determined accordingly. 




892 patients were enrolled between 1993 and 1996. Mean follow-up time was 2.0 years. Withdrawals from active treatment occurred at a rate of about 10% per year; 6.5% because of the development of risk factors for embolism and 3.9% for "other" reasons. 

Some patient characteristics included: 

  • Male sex: 78%; mean age: 67 years. 
  • Mean BP: 130/70; history of hypertension: 46%. 
  • Onset of AF less than one year: 77%; paroxysmal AF: 28%. 
  • History of CHF: 9%; history of MI: 7%; diabetes: 13%. 
  • Mean left atrial diameter: 4.6 cms; patients with LA > 5.0 cms: 25%. 
Event rates 

The primary event rate (ischemic stroke and systemic emboli) was 2.2% per year (95% CI 1.6-3.0%). 

The rate of ischemic stroke was 2.0%, disabling stroke 0.8% and TIA (not a primary event) 1.3%. The annualized rate of intracranial hemorrhage was 0.1% and of major, non-CNS hemorrhage 0.6%. 

Among the 54% of patients without any history of hypertension, the annual event rates were substantially lower than among those with such a history: the primary event rate was 1.1% per year vs. 3.6%, and the rate of disabling strokes was 0.5% vs. 1.4%. 

In multivariate analysis, only a history of hypertension and age were significant predictors of a primary event. 


Author's discussion

The authors suggest that, based on the results of this study, the risk of stroke in patients without valvular heart disease can be classified as low, moderate or high. 

The risk is low in patients similar to those enrolled in this trial (without any of the four risk factors) who, in addition, have no history of hypertension. In this group, the stroke rate is around 1% per year if these patients are treated with aspirin alone, which is similar to the stroke rate in the general age-matched population. Treatment with warfarin may not add any significant benefit. 

In patients without any of the four risk factors, but with a history of hypertension are at moderate risk (around 3.5% per year), and therapy should take into account the risk of anticoagulation and patient preferences. 

Finally, in patients with any of the four risk factors, data from other branches of the SPAF trials indicate that the event rate is much higher, around 8% per year even when treated with aspirin and low-dose warfarin. These patients should be considered for adjusted-dose warfarin therapy unless there are contra-indications. 

The authors note that this trial did not directly compare aspirin with either placebo or warfarin. Thus, the degree of benefit conferred by aspirin treatment cannot be inferred from the data presented here. The added protection that might be gained by using warfarin in these patients can also not be directly deduced, although the authors use comparisons to other studies to suggest that it would be relatively low, and at a price of an increased risk of hemorrhage. 



Patients with "lone atrial fibrillation", who are under 60 years of age, not hypertensive, without evidence of cardiovascular disease and who have normal echocardiograms, are at sufficiently low risk for stroke that warfarin therapy is unlikely to be of significant benefit. Patients who do not fall into this category (the vast majority) have been shown to benefit from anticoagulation. However, this large group is not homogeneous. Within it, some patients are at much higher risk than others. It would make sense that some of these patients are at sufficiently low risk that they could be treated with aspirin, or not at all. 

Previous trials have identified a number of risk factors for stroke in the setting of atrial fibrillation. This study made use of this fact to define a subgroup of patients who did not have lone atrial fibrillation but were nevertheless felt to be at relatively low risk. These patients were then treated with aspirin alone and followed up. 

Patients who did not have valvular disease, were not actively hypertensive, did not have left ventricular dysfunction, did not have a history of embolic phenomena or TIA's and were not women over the age of 75 were enrolled. Their stroke risk on aspirin alone was 2.2%. More importantly, if patients with even a past history of hypertension were excluded, the stroke rate was as low as 1%. It would seem reasonable to withhold coumadin and treat with aspirin in this subgroup. In the subgroup with a history of hypertension but without any of the other risk factors, the stroke rate was about 3.5%. These patients would almost certainly benefit from warfarin therapy, although not as much as those with one or more of the above mentioned risk factors. 

This study should not be interpreted as showing that warfarin therapy in atrial fibrillation isn't as important as it would seem to be. The recent trend in therapy for atrial fibrillation has been to be much more liberal with anticoagulation, and the current study does not invalidate this trend. What it does do, however, is define a very specific subgroup of patients who can probably be managed with aspirin alone. This subgroup represents a minority of patients, but a significant one. 

To be eligible for this strategy, patients must not have mitral stenosis, any history of hypertension, congestive heart failure (clinically or by echocardiography), any history of CVA, TIA or embolism and they should not be women over 75 years of age. 

Since this trial did not compare aspirin therapy to warfarin, it is very possible that warfarin therapy would reduce the stroke rate in this population even further. Given the low incidence of stroke on aspirin however, and the not insignificant risk of hemorrhage with warfarin, the overall benefit is not likely to be large. 

The important information gleaned from this trial points out one of the problems with randomized controlled clinical trials: the populations which are shown to benefit from a particular intervention are often heterogeneous, and the benefit may not extend to all subgroups. Prospective analysis of therapeutic interventions in subgroups is one way to deal with this problem, and can enable us to target our interventions more appropriately. The RCT is not the final answer to every possible question. 

May 10, 1998 


References related to this article from the NLM's PubMed database. 

Reader Comments

Date: Sun, 10 May 1998 
From: steven shumak <> 
Dear Michael: 

Re: Patients with nonvalvular atrial fibrillation at low risk of stroke during treatment with aspirin - Stroke Prevention in Atrial Fibrillation III study 

In this study, the rate of withdrawal that occurred because of the development of risk factors for embolism was significant at 6.5% and actually exceeded the rate of any of the primary endpoints. It is reasonable to speculate that had those patients who developed such risk 
factors remained in the study, the overall stroke rate in the study patients would have increased. This speculation does not threaten the study's results but does suggest that we should be cautious in applying its conclusions to similar patients in our own practices. In particular, in "real world", non-study, patients, there may be less intensive follow up and monitoring. It would then become more likely that the development of risk factors for stroke would be overlooked and treatment with aspirin continued inappropriately. Phrased differently, the prescription of aspirin for some patients with atrial fibrillation must not lull us 
into therapeutic complacency. 

Steven L. Shumak MD, FRCPC 
Sunnybrook Health Science Center 
University of Toronto 

    This is a very important point  --mj 

Date: Mon, 11 May 1998 
From: "Dr G N Lakshminarayan" <> 
I am a teaching neurologist in practice also. 

Don't you think aspirin is only antithrombotic whereas we need anticoagulants to prevent "clotting"? Though it is cheap and "good", I feel ACs, if monitored properly with PT/INR, should be better in "lone" (really lone) AFs. 


    Anticoagulants are probably always better than aspirin for the prevention of thromboembolism in atrial fibrillation. The problem with AC's is the risk of bleeding, so that if the absolute risk of thromboembolism is sufficiently low, the risk of bleeding begins to outweigh the risk of embolism. Obviously, the risk of bleeding depends on how stable the INR is, which will depend on multiple factors, including patient compliance. 
    Again, this study does not demonstrate that treatment with warfarin should be withheld from any of the patients studied here, only that it is probably safe to do so in the lowest risk subgroup. For patients who are very compliant, it remains reasonable to treat with adjusted dose warfarin. Patient preferences should also be taken into account in this situation. 
    For patients with truly lone atrial fibrillation (not studied here) the risk of embolism is even lower, and the benefit of anticoagulation is thus likely to be even smaller.  -- mj 
Submit a comment 
Journal Club homepage

Site Meter