Home SVCC                                                   Area: English - Español - Português

Stroke, Atrial Fibrillation and
Antithrombotic Therapy

Robert G. Hart, MD

University of Texas Health Sciences Center,
San Antonio, Texas, USA

   The text of this lecture is linked to the sequence of Figures, as described below. The presentation is based on a recent review, which includes the key citations (Hart RG, Halperin JL. Atrial fibrillation and stroke: Concepts and controversies. Stroke 2001; 32: 803-808).

Figure 1: Atrial Fibrillation and Thromboembolism: A Decade of Progress in Stroke Prevention

   Atrial fibrillation (AF) is a common cardiac dysrhythmia in the elderly and the most frequent substrate for cardioembolic stroke. After decades of neglect, randomized trials testing antiplatelet and anticoagulants for stroke prevention were initiated in the mid-1980s. During the past ten years, a score of randomized clinical trials combined with high-quality epidemiologic and observational studies have changed the antithrombotic management of millions of people with AF. AF that is not associated with valvular heart disease (that is, nonvalvular or nonrheumatic AF) is now appreciated to be an important, independent risk factor for stroke, for which effective prevention has been established.

   This lecture focuses on three issues relevant to stroke prevention in AF patients;

1. The efficacy and safety of antithrombotic agents for stroke prevention.
2. The relationship between the intensity of anticoagulation and its efficacy/safety.
3. Risk factors that identify AF patients at high vs. low absolute risks of stroke.

   Other strategies for prevention of stroke in AF patients, which include cardioversion/maintenance of sinus rhythm and surgical obliteration of the left atrial appendage, will not be discussed here. Clinical trials testing these interventions are underway, but at present their value for prevention of stroke has not been convincingly established.

Figure 2: Photograph of the Left Atrial Appendage with Thrombi

   Consistent results of epidemiological studies show that nonvalvular AF is associated with an approximately five-fold increased risk of stroke, after adjustment for age, gender, hypertension and other stroke risk factors. The underlying substrate of most stroke in AF patients is cardiogenic embolism. Stasis-precipitated thrombi form in the left atrial appendage (shown here - yellow arrows). Thrombi tend to be sufficiently large to produce substantial stroke if they embolize to the brain. While autopsy studies confirm that only about 20% of cardiogenic embolism involve the cerebral circulation (in proportion to its fraction of cardiac output), more than 80% of symptomatic emboli are to the brain, because most embolism to the systemic circulation is not clinically recognized.

Figure 3: Prevalence of AF

   In the United States, about 1% of the total population has sustained or recurrent AF, but importantly this frequency is about 5% of people over age 65 years. As the number of people living beyond age 65 years is rapidly increasing in many countries, the number of people with AF is expected to sharply increase during the next 10-20 years. Hence, the problem of AF-associated stroke is predicted to double in coming years in many countries.

Figure 4: AF and Stroke

   About 16% (i.e. one in six) of ischemic strokes in most population-based case series are associated with AF. Hospital-based case series show that this proportion is age-sensitive: the older the mean age of the stroke cohort, the higher the frequency of AF. Ischemic stroke in AF patients tend to be particularly disabling, affecting older patients with higher short-term mortality. Efforts at primary prevention of stroke, rather than waiting for stroke to occur and then dealing with its consequences, are important in AF-associated stroke.

Figure 5: The Stroke Prevention in Atrial Fibrillation (SPAF) Trials

   The SPAF trials were a series of randomized clinical trials sponsored by the U.S. NIH/NINDS between 1987-1997, in which I was heavily involved and which form the background and basis for many of the comments that follow.

Figure 6: The SPAF Randomized Trials

   The SPAF clinical trials consisted of a series of randomized clinical trials, as outlined on this Figure. After the initial trial (SPAF I) showed both warfarin and aspirin to be superior to placebo for prevention of stroke in AF patients, subsequent trials compared warfarin with aspirin (SPAF II) and then conventional intensity warfarin to low-intensity warfarin combined with aspirin (SPAF III).

Figure 7: Randomized Trials of Antithrombotic Therapy in Atrial Fibrillation

   At present, 25 randomized trials have been undertaken testing various types and combinations of antithrombotic therapies for prevention of stroke in AF patients. Results from clinical trials involving over 10,000 participants have been published and tell a reasonably consistent story (see the next three Figures). For further description of these studies, see a recent meta-analysis (Hart RG et al., Ann Intern Med 1999; 131: 492-501).

Figure 8: Adjusted-dose Warfarin vs. Placebo

   Six randomized trials have compared adjusted-dose warfarin (approximate INR range 2-4) to placebo or no antithrombotic treatment in patients with AF. All showed substantial reduction in stroke which, by intention-to-treat analysis, averaged about 60%. This overall 60% reduction in stroke includes a small increase in hemorrhagic stroke associated with warfarin use. By on-therapy analysis (i.e. excluding participants who were not receiving assigned anticoagulation), the overall reduction in ischemic stroke was 85%. Adjusted-dose warfarin, if taken carefully, virtually eliminates the increased risk of ischemic stroke associated with AF by inhibiting formation of atrial appendage thrombi (as confirmed by transesophageal echocardiographic correlations, see Ann Intern Med 1998; 128: 639-647).

Figure 9: Aspirin vs. Placebo

   Six randomized trials comparing aspirin to placebo show a small effect of aspirin, averaging about 20% for reduction in all stroke. All six trials have shown trends favoring aspirin, but statistical significance was apparent in only a single individual trial (SPAF I). Despite initial controversy, it is now generally acknowledged that aspirin offers a modest reduction in stroke in patients with AF. There is no clear evidence that the dosage of aspirin is related to its efficacy; dosages between 75mg/day to 325 mg/day have been used.

Figure 10: Warfarin vs. Aspirin

   As would be predicted from the previous two Figures, randomized trials directly comparing adjusted-dose warfarin to aspirin have shown warfarin to be superior, with a relative risk reduction of about 40% by intention-to-treat analysis that considered all stroke (ischemic and hemorrhagic).

   A recently published meta-analysis (Taylor FC et al., BMJ 2001; 322: 321-6) claimed that there is "considerable uncertainty about the value of long term anticoagulation compared with antiplatelet treatment." I disagree. This meta-analysis did not include the most "positive" randomized trial comparing the two treatments (the European Atrial Fibrillation Trial), but it did include a randomized trial comparing warfarin to indobufen, an antiplatelet agent, combining indobufen with aspirin (analyzed together as "antiplatelet agents). All antiplatelet agents do not have the same effects, and meta-analysis that combined indobufen with aspirin is potentially misleading, in my view.

Figure 11: Ischemic Stroke Subtypes in Atrial Fibrillation

   Not all ischemic strokes that occur in AF patients are due to embolism of left atrial appendage thrombi. Other ischemic stroke etiologies account for perhaps one-third of stroke in these elderly, usually hypertensive patients with AF. By the best available clinical estimates (and these are imperfect), about 10% of ischemic strokes (and about 2/3rds of ischemic stroke in AF patients) are due to embolism of appendage thrombi. Considering the different ischemic stroke mechanisms in AF-associated stroke is important for understanding the mechanisms underlying the effects of aspirin and warfarin.

Figure 12: Ischemic Stroke Subtypes vs. Antithrombotic Therapy in the SPAF Studies

   These exploratory analyses of the SPAF studies (Cerebrovascular Disease 2000; 10: 39-43) shows an 83% reduction by adjusted-dose warfarin vs. aspirin in strokes classified as cardioembolic, but an essentially equal rates of presumed noncardioembolic stroke (the latter observation consistent with the recently reported results of the large NIH-sponsored Warfarin-Aspirin Recurrent Stroke Study). The effect of aspirin in AF patients is primarily to prevent noncardioembolic strokes. AF patients at highest risk have the highest frequency of cardioembolic strokes, and hence warfarin is particularly efficacious for high-risk AF patients.

Figure 13: Summary of Efficacies of Antithrombotic Therapies for Stroke Prevention

   In summary, pooled analysis of multiple randomized clinical trials show that adjusted-dose warfarin reduces stroke by about 60% in AF patients (preventing larger, cardioembolic strokes), aspirin reduces stroke by about 20% (mainly noncardioembolic strokes), and warfarin reduces stroke by about 40% compared to aspirin.

   Low-dose warfarin plus aspirin is much less efficacious than adjusted-dose warfarin (from the SPAF III randomized trial). Results of trials comparing other antithrombotic strategies are inconclusive, due to small numbers of patients or other methodologic vagaries (see Ann Intern Med 1999: 492-501).

Figure 14: Lowest Effective Intensity of Anticoagulation in Atrial Fibrillation

   What is the optimal intensity of anticoagulation for stroke prevention in AF patients? The widely recommended target INR range is 2.0 - 3.0. I believe that a target INR of 2.0 (target range of 1.6-2.5) is more appropriate for primary prevention in AF patients over age 75, and the next few Figures offer evidence in support of this viewpoint.

   Shown on this Figure are the results of a case-control study in AF patients correlating the odds of stroke with achieved INR (Hylek et al., NEJM 1996; 335: 540-546). The graph shows a steep rise in the odds of stroke beginning at INRs below about 1.7.

Figure 15: INR vs. Ischemic Stroke in the SPAF III Trial

   A similar analysis, carried-out independently on the SPAF III randomized trial, showed substantial protection against stroke with INRs between 1.5 and 2.0.

Figure 16: INRs vs. Protection from Stroke in Atrial Fibrillation

    Here, the results of the case-control study (Figure 14) and the SPAF III data (Figure 15) are plotted as a fraction of maximum protection (assumed to be INRs between 2.0-3.0). Achieving a target INR of range of 1.6-2.5 results in over 90% of the protection afforded by higher INRs. From these data, it appears that achieved INRs between 1.6-3.0 offer substantial protection against ischemic stroke in AF patients.

Figure 17: Bleeding Risk, Age and INR in the SPAF III Trial

   If INRs between 2.0-3.0 offer maximal protection, why consider use of the lower limits of effective intensities? Because of the risk of serious bleeding is continuously related to the intensity of anticoagulation, with no threshold level below which anticoagulation is completely safe. Shown here are data from the SPAF III randomized trial. The risk of serious bleeding was strongly and independently related to the achieved INR and to patient age. The rate of serious bleeding in patients over age 75 who achieved INRs over 3.0 was intolerably high. Use of the lowest efficacious intensity of anticoagulation is particularly important for AF patients over age 75. A target INR of 2.0 results in fewer major hemorrhages than an INR target of 2.5 (Yamaguchi et al., Stroke 2000; 31: 817-821).

   In summary, choosing the lowest efficacious intensity of anticoagulation is particularly important for AF patients over age 75, in whom major and minor hemorrhage are especially frequent (the latter often leading to discontinuation of anticoagulation). While most current guidelines recommend a target INR range of 2.0-3.0 for all patients with atrial fibrillation, it is my view that the lower end of this range may be optimal for primary prevention in AF patients over age 75 years.

Figure 18: Which Patients with Atrial Fibrillation Should be Anticoagulated?

   Now that the relative benefits of warfarin and aspirin have been reasonably well defined by recent randomized trials, the most important and controversial question remains "which AF patients are at sufficiently high risk for stroke to warrant lifelong anticoagulation?"

   While the relative risk of stroke is increased by AF, the absolute rate of stroke varies widely - nearly 25 fold - between identifiable subgroups of AF patients. Young patients with "lone" AF have a low rate of stroke (0.5% per year), compared to 12% per year for older AF patients with previous stroke. Adjusted-dose warfarin reduces stroke both for AF patients with low and high stroke rates, but the absolute rate reduction is small for those at low risk. Which AF patients benefit sufficiently to warrant the bleeding risks, need for regular medical monitoring and other disutility associated with anticoagulation? I contend that many, including most who are under age 75, AF patients do not benefit substantially from anticoagulation to warrant its use (despite its striking efficacy). This apparent paradox is developed and explained in the next few Figures.

Figure 19: Predictors of Stroke Risk in Atrial Fibrillation

   Clinical features associated independently with increased risk of stroke in AF patients have emerged from multivariate analysis of several large clinical trial datasets as well as observational studies and are listed on the Figure. The mechanism(s) by which these features contribute to stroke risk is not entirely clear. Hypertension, the most prevalent factor, appears to enhance stasis in the left atrial appendage, presumably through effects on left ventricular function. Left atrial diameter and volume assessed by precordial echocardiography does not appear to be a useful independent predictor when analyzed with hypertension and other clinical variables.

Figure 20: Hormone Replacement Therapy and Stroke Risk in SPAF III

   Stasis is fundamental for the formation of atrial appendage thrombi in AF patients. Prothrombotic (i.e. "hypercoaguable") states have been postulated to also contribute, but these have not been adequately defined. Some evidence of the contribution of altered coagulation emerged from exploratory analysis of SPAF III, in which use of postmenopausal hormone replacement therapy was found to be an independent predictor of stroke risk. Hormone replacement therapy is known to alter coagulation and to be a risk factor for venous thromboembolism. Confirmation of the association between use of hormone replacement therapy and stroke in AF requires independent confirmation before influencing clinical management because of the exploratory nature of these analyses.

Figure 21: Pathological Specimen of Left Atrial Appendage Thrombus

   Atrial thrombi in nonvalvular AF form in the appendage, rather in the body of the atrium proper, where they cannot be detected with precordial echocardiography but can be visualized using transesophageal echocardiography. This pathologic specimen shows enlargement of the smooth-walled left atrium (upper right area) with a thrombus evident in the trabeculated atrial appendage (upper left). While reduced appendage flow velocities and spontaneous echodensities ("smoke") detected by transesophageal echocardiography are associated with an increased risk of stroke in AF patients (Ann Intern Med 1998; 218: 639-647), it remains unclear whether these features add important predictive information when clinical features are also considered. The role of transesophageal echocardiography for stratifying stroke risk in AF is uncertain, and additional studies are needed.

Figure 22: Risk Stratification Schemes in Atrial Fibrillation

   Based on the clinical features independently associated with stroke risk in AF, four risk stratification schemes have been developed and published, with the two most widely cited shown here (the ACCP and SPAF III schemes).. While there are clear similarities, differences are important for individual patient management. For example, all AF patients over age 75 are considered high-risk on the basis of age alone by the ACCP scheme, while in the SPAF-generated scheme, about one-third of patients over age 75 are classified as low risk. Yet another risk stratification scheme called CHADS2 will be published during the summer of 2001 (Gage BF et al., JAMA, in press).

   In this rapidly evolving area, the "best" risk stratification scheme is unclear. The Atrial Fibrillation Investigator scheme (1994) focuses on identification of a small subset of AF patients with a very low risk of stroke. The SPAF III scheme (1995), on the other hand, seeks to identify a large group of AF patients at moderate risk, for whom the decision to use anticoagulants is influenced by bleeding risk and patient preferences/values. Comparison of three of these schemes has been undertaken (Pearce LA et al., Am J Med 2000; 109: 45-51), but the last word has yet to be written on this critical issue in patient management.

Figure 23: Stroke Risk with Intermittent (Paroxysmal) Atrial Fibrillation

   Surprisingly, intermittent (also called paroxysmal) AF was not an independent predictor of stroke risk in any of the studies using multivariate analysis, to date. The data shown are from the SPAF study (Hart RG et al., JACC 2000; 35: 183-187). Matched for other clinical risk factors, intermittency vs. sustained AF did not significantly alter observed stroke rates, indicated by the similar heights of the bars in each risk category. Of note, patients with intermittent AF included in the SPAF trials were relatively elderly and had multiple, recurrent attacks; whether these observations apply to younger AF patients with brief, rare recurrences is unknown.

Figure 24: Testing of the SPAF III Criteria for Stratification of Stroke Risk in Atrial Fibrillation

   At present, the SPAF III scheme (1995) has been the most extensively validated and has been found to be predictive in two cohorts of patients with AF, one of which was population-based. The soon-to-be-published CHADS2 scheme has also been tested in a hospital-based cohort of AF patients, and it was found to be predictive and relatively simple to use. Research is ongoing in this important area, and no firm recommendation about an optimal scheme can be made. At present, the SPAF III (1995), ACCP (1998) and CHADS2 scheme are reasonable choices for clinical use, in my view.

Figure 25: Risk Stratification in Atrial Fibrillation: Numbers Needed to Treat

   The importance of risk stratification for selecting AF patients who benefit substantially from anticoagulation is illustrated on this Figure and on the one that follows. The number of AF patients who would need to be treated for one year with warfarin instead of aspirin in order to prevent one stroke is shown in the table. For a typical cohort of unselected AF patients, 40 would need to be treated with warfarin instead of aspirin for one year to save one stroke, and 80 to save one disabling stroke. Using a risk stratification approach, the 200 low-risk AF patients would need to be treated with warfarin to save one stroke per year, but only 10-20 high-risk AF patients. The threshold for use of anticoagulation for those at moderate risk varies with availability of optimal anticoagulation and patient values/preferences.

Figure 26: Stroke Rates During Antithrombotic Therapy Stratified by Risk

   Another way of illustrating the effect of risk stratification is to consider the absolute stroke rates during aspirin and warfarin therapy, depending on the intrinsic risk of stroke. Among high-risk AF patients represented on the left, the difference in height of the bars for aspirin compared to warfarin is substantial. Among low-risk AF patients (on the right), the rate of stroke on warfarin is only 50% of that on aspirin, but the difference in the height of the bars (i.e. the absolute rate reduction) is small. This illustrates that even a large relative risk reduction applied to low-risk patients results in only small absolute rate reductions.

   In short, estimating the intrinsic risk of stroke in AF patients, employing one of the available risk stratification schemes, is a necessary precursor to the optimal use of anticoagulation. High-risk and many moderate-risk AF patients should undergo anticoagulation, while the benefits of anticoagulation are minimal for those at low-risk.

Figure 27: Patient Values and Preferences

   An additional important factor in the decision to anticoagulate AF patients is the individual patient's preferences after explanation of the benefits, risks and disutility of therapy. The underuse of warfarin in AF patients documented in many recent surveys is typically blamed on the physician, but the patient refusal is an underestimated factor, in my experience. It is certainly true that the patient's choice is strongly influenced by how the benefits and risks are presented, but studies (like the one shown here) using standardized educational materials show that the threshold of benefit for which patients elect to take warfarin varies (of no surprise to experienced clinicians). Hence, particularly for AF patients with moderate-risks of stroke, patient preferences are often a key factor in the decision to pursue anticoagulation.

Figure 28: Antithrombotic Therapy in Atrial Fibrillation: Summary

   Based on the concepts developed in this presentation, the following seem clear:

- Left atrial thrombi in AF patients account for about 10% of all ischemic stroke.
- Adjusted-dose warfarin (INRs between 1.6-3.0) is highly efficacious for stroke prevention.
- Aspirin is modestly efficacious, reducing mainly noncardioembolic strokes occurring in AF patients.

Figure 29: Intensity of Anticoagulation in Atrial Fibrillation: Recommendations

   While a target INR of 2.5 (target INR range of 2.0-3.0) is most often recommended for all AF patients, it is my policy to aim for a lower INR target of 2.0 (target range 1.6-2.5) for AF patients over age 75 who have not had prior stroke/TIA (i.e. for primary prevention).

Figure 30: Antithrombotic Therapy in Atrial Fibrillation: Recommendations

   The general conceptual approach to selection of antithrombotic therapy for prevention of stroke in AF patients, as developed during this presentation, is outlined here.

Figure 31: Crucial Unsettled Issues

   An enormous amount has been learned about atrial fibrillation-associated stroke and its prevention during the past decade. Efficacious antithrombotic therapy for prevention, tailored according to the inherent risk of stroke, has been established as standard of care. Many issues important for patient management remain. Particularly, whether successful and sustained treatment of hypertension reduces the risk of cardioembolic stroke in AF patients, reversing the left ventricular abnormalities that augment left atrial stasis.


Your questions, contributions and commentaries will be answered
by the lecturer or experts on the subject in the Stroke list.
Please fill in the form (in Spanish, Portuguese or English) and press the "Send" button.

or commentary
Name and Surname:
E-Mail address:


2nd Virtual Congress of Cardiology

Dr. Florencio Garófalo
Steering Committee
Dr. Raúl Bretal
Scientific Committee
Dr. Armando Pacher
Technical Committee - CETIFAC

Copyright© 1999-2001 Argentine Federation of Cardiology
All rights reserved


This company contributed to the Congress: