INTRODUCTION — Systemic embolization of atrial thrombi can occur with any form of atrial fibrillation (AF), either spontaneously or in association with cardioversion. As a result, chronic antithrombotic therapy with either anticoagulation (ie, a vitamin K antagonist) or antiplatelet therapy is considered for most of these patients. Both have been shown to be effective in preventing systemic embolization, although anticoagulant therapy is far more effective and the preferred therapy in all but the lowest-risk patients. The decision to use antithrombotic therapy is best made with an appreciation of the risks of both embolic events and bleeding during antithrombotic therapy [1].

Ischemic stroke is the most frequent clinical manifestation of embolization associated with AF. It can occur at any point during the clinical course of AF. (See "Stroke in patients with atrial fibrillation".)

Chronic antithrombotic therapy for the prevention of embolic events in patients with nonvalvular AF will be reviewed here. The role of anticoagulation in relation to restoration of sinus rhythm and the evaluation and treatment of patients with AF who have suffered a stroke are discussed separately. (See "Anticoagulation prior to and after restoration of sinus rhythm in atrial fibrillation" and "Stroke in patients with atrial fibrillation".)

Risk stratification and the incidence of embolization are discussed elsewhere. (See "Risk of embolization in atrial fibrillation".)

PATIENTS WITH VALVULAR HEART DISEASE — The major clinical trials of antithrombotic therapy and subsequent meta-analyses have largely been restricted to patients with nonvalvular or nonrheumatic AF [2-11].

Issues related to anticoagulation for AF in patients with valvular heart disease are discussed elsewhere in the appropriate topic reviews:

Mitral stenosis (see "Medical management and indications for intervention in mitral stenosis" section on Prevention of embolization.) .

• Mitral regurgitation or mitral valve prolapse (see "Overview of the management of chronic mitral regurgitation", section on 'Anticoagulation' and "Nonarrhythmic complications of mitral valve prolapse", section on 'Recommendations for antithrombotic therapy'.
• Aortic regurgitation and aortic stenosis, the indications for anticoagulation, are primarily determined by the AF (see "Course and management of chronic aortic regurgitation in adults" and "Medical therapy in asymptomatic aortic stenosis in adults", section on 'Atrial fibrillation'.
• Prosthetic heart valves (see "Antithrombotic therapy in patients with prosthetic heart valves".

RISK STRATIFICATION — Many studies have evaluated risk factors for embolization in AF, and multivariate risk models have been constructed [12]. The CHADS2 risk score is the most popular and has been best validated in different patient populations (calculator 1). The main advantage of the CHADS2 score compared to other risk models is its simplicity [13,14]. A detailed discussion of risk stratification is found elsewhere. (See "Risk of embolization in atrial fibrillation".)

Despite limitations of all current models, including CHADS2 [15], we recommend risk stratification in all patients as part of the process of choosing therapy. Consideration needs to be made for patient factors, such as stroke risk versus bleeding complication, as well as patient compliance and comorbidities.

MECHANISM OF THROMBOEMBOLISM — The mechanisms leading to an increased risk of thrombus, embolism, and stroke in AF are multiple and closely interact with each other. Blood stasis, especially in the left atrial appendage (LAA), and activation of the hemostatic system are thought to play primary roles in patients with nonvalvular AF. Coexistent aortic atherosclerosis may also play a role. These topics are discussed in detail separately. (See "Mechanisms of thrombogenesis in atrial fibrillation" and "Echocardiographic evaluation of the atria and appendages" and "Role of echocardiography in atrial fibrillation".)

ANTITHROMBOTIC PREVENTION — Multiple antithrombotic (anticoagulant and antiplatelet) strategies to decrease the risk of embolic events in AF have been evaluated in randomized trials. Warfarin and aspirin have been best studied [16]. More recently, oral thrombin inhibitors have received increased attention. Dabigatran is one such agent that shows promise as an alternative to warfarin. The decision to use these drugs is based upon assessment of the competing risks of thromboembolic events and bleeding related to antithrombotic therapy.

Clinical trials [2-8] and meta-analyses [9-11] have demonstrated that among patients with AF at high risk of thromboembolic events, warfarin, compared to placebo, significantly reduces the incidence of stroke. The evidence that aspirin is better than placebo is less robust. In addition, warfarin is approximately three times as effective.

The following discussion will concern the efficacy and safety of warfarin and aspirin in the medical management of patients with AF. There are two other issues related to chronic anticoagulation that are discussed separately:

• Role of anticoagulation in patients who have already had an embolic event (see "Stroke in patients with atrial fibrillation".
• The role of anticoagulation in patients who have undergone radiofrequency catheter ablation to prevent recurrent AF. (See "Radiofrequency catheter ablation to prevent recurrent atrial fibrillation".)

Warfarin — The decision to use antithrombotic therapy to prevent embolization in patients with nonvalvular atrial fibrillation (AF) is based upon an assessment of the comparative benefits and risks of such therapy for individual patients.

Efficacy — The efficacy of warfarin in the prevention of clinical stroke in patients with AF was studied in numerous trials, including SPAF- I, SPAF- II, and SPAF- III trials, and AFASAK, BAATAF, SPINAF, and CAFA [2-7]. Together, these trials randomly assigned more than 4000 patients with nonvalvular or nonrheumatic AF to aspirin, warfarin, or placebo, and demonstrated that anticoagulation with adjusted-dose warfarin significantly reduced stroke risk in patients with AF when compared with aspirin [3-5] or placebo (graph 1) [2,3,5-7]. Overall, adjusted-dose warfarin reduces the risk of stroke by two-thirds compared to no antithrombotic therapy, with the expected degree of absolute benefit dependent on baseline risk (table 1) [8,9,17]. Warfarin is effective in both men and women and in all age groups, including those over the age of 75 years (graph 2) [8,18-20]. In addition, mortality was significantly reduced by warfarin compared to no antithrombotic therapy [9].

In addition, warfarin provides benefit in patients who have already had a stroke. (See "Stroke in patients with atrial fibrillation".)

These trials were conducted more than 15 years ago, and the following concerns have been raised about whether the findings can be applied to current practice [21]:

• While more recent studies continue to support the efficacy of warfarin, lower absolute levels of stroke risk have been observed, resulting in a smaller absolute benefit of anticoagulant therapy [14,22].
• The real-world benefit may be lower than that demonstrated in clinical trials, since the patients are less likely to have a therapeutic INR. (See 'Underutilization and underanticoagulation' below.)

Safety — The major safety concern with the use of warfarin is the risk of major bleeding. Major bleeding includes bleeding that requires hospitalization, transfusion, or surgery, or involves particularly sensitive anatomic locations. Intracranial bleeding is the most serious bleeding complication with warfarin, since the likelihood of mortality or subsequent disability is substantially higher than bleeding at other sites [23].

Overanticoagulation (as defined as a supratherapeutic INR), concomitant use of aspirin, and patient age are three of the most important predictors of major bleeding [24,25]. In an attempt to estimate the probability of major bleeding in patients on chronic warfarin therapy, a number of risk models have been developed. The risk factors and risk models are discussed in depth separately. (See "Therapeutic use of warfarin", section on 'Bleeding' and "Risk of intracerebral hemorrhage in patients treated with warfarin".)

Both the risk of any bleeding and of major bleeding appears to be significantly higher with adjusted-dose warfarin compared with aspirin. Major risk factors for intracranial hemorrhage include older age and prior stroke [21,26].

In the individual patient meta-analysis of six prevention trials cited above, the absolute rate increase of major bleeding with warfarin compared with aspirin was 0.9 events per 100 patient-years (2.2 versus 1.3 events per 100 patient-years) [10]. The risk of intracranial bleeding is extremely high at an INR above ≥5.0 (graph 3) [27].

The increased risk of major bleeding is particularly relevant to older adults in whom varying combinations of frailty; poor mobility; poor compliance; the use of concomitant medications that can result in drug interactions; and frequent falls are common. The absolute risk of bleeding was determined in a review of over 10,000 Medicare recipients (mean age 77 years) who were treated with warfarin for AF [28]. At 180 days, the rate of major bleeding was 2.0 percent in patients treated with warfarin. Issues related to anticoagulation in older adults and bleeding rates according to CHADS2 score are discussed separately. (See "Anticoagulation in older adults".)

The combination of warfarin and aspirin increases the risk of bleeding. In the study of Medicare recipients discussed above, the rate of major bleeding was 2.8 percent in those treated with warfarin plus aspirin. Combination therapy was associated with a threefold increase in the risk of intracranial hemorrhage (0.9 versus 0.3 percent, odds ratio 2.95). In another meta-analysis, combination therapy was associated with a significant increase in the risk of intracranial hemorrhage (relative risk 2.4, 95% CI 1.2-4.8) [29]. (See 'Aspirin' below.) The addition of aspirin to warfarin to prevent embolization in AF should be avoided, given the lack of additive benefit in preventing stroke or cardiac events and the fact that bleeding is substantially increased [30].

Net clinical benefit — The decision to use antithrombotic therapy to prevent embolization in patients with nonvalvular AF is based upon an assessment of the absolute benefits and risks of such therapy. Combining ischemic events with major bleeding together to yield net clinical benefit attempts to balance benefits and risks.

The ATRIA observational study evaluated the net clinical benefit (NCB) of warfarin therapy in 13,559 patients with nonvalvular AF who were identified in an outpatient database in 1996 and 1997 [21]. NCB was defined as the difference between annualized rate of thromboembolic events prevented by warfarin, minus the annualized rate of intracranial hemorrhages (ICH) induced by warfarin, multiplied by a weighting factor. In the base model, ICH was weighted as 1.5 times the impact of ischemic stroke.

Outcomes were evaluated in the warfarin and no-warfarin groups (approximately 50 percent of the latter were on aspirin) over a median follow-up of six years. The NCB became statistically significant at a CHADS2 score of 2 (1 event prevented per 100 patient-years) and progressively increased at higher CHADS2 scores (2.2 events prevented per 100 patient-years at a CHADS2 score of 4 to 6). This relationship reflects the much greater reduction in embolic risk compared to increase in intracranial hemorrhage risk with higher CHADS2 scores.

The ATRIA study supports the benefits of warfarin therapy in patients with a CHADS2 score of 2 or higher. The NCB with warfarin in patients with a CHADS2 score of 1 was small, and the confidence intervals around the NCB included the possibility of harm.

Initiation of therapy — The choice of whether to start warfarin alone or in combination with unfractionated heparin or low-molecular-weight heparin (ie, bridging) is based upon an assessment of the risk of a thrombus developing within the next several days versus the risk of bleeding complications.

In most patients with nonvalvular AF, the risk of stroke during the several days typically required to reach a therapeutic INR is very low, and it is reasonable to administer warfarin as an outpatient without bridging. For patients with nonvalvular AF deemed to be at high risk of thrombus formation/thromboembolism (eg, prior cerebrovascular event/TIA or intracardiac thrombus) and low risk of intracranial bleeding, initiation of warfarin with a heparin bridging regimen is reasonable. This approach is in general agreement with the 2008 ACCP guidelines [31]. However, there are no data from randomized trials addressing such patients, and expert opinion is divided as to the benefit of heparin bridging in certain situations.

Patients with nonvalvular AF who present with acute stroke have a relatively high risk of recurrent embolism and/or progressive ischemia (approximately 5 percent during the first two weeks) [32,33]. Although early use of heparin reduced the rate of recurrent embolism and/or progressive ischemia in some trials, this was balanced by an increased incidence of intracranial bleeding, especially in patients with large strokes. The conclusion from these data is that there is no overall benefit to early heparin therapy [32,33], and we do not recommend heparin bridging in such patients.

Other issues surrounding the initiation of warfarin are found elsewhere. (See "Therapeutic use of warfarin", section on Loading doses of warfarin.)

Target INR — An INR between 2.0 and 3.0 has been generally recommended for most patients with AF who receive warfarin anticoagulation (table 2) [24,31]. This is based upon the increased risk of stroke observed with INR values significantly below 2.0 (four- to sixfold at an INR of 1.3 compared with an INR of 2.0 or above) and the increased risk of bleeding associated with higher intensities of anticoagulation (graph 3) [31,34-37]. (See 'Safety' above.)

The severity of stroke and the mortality rate in patients with AF are also increased when the INR is below 2.0. This was illustrated in a cohort of 596 patients with nonvalvular AF who experienced an ischemic stroke [35]. Those with an INR below 2.0 on admission had an increased likelihood of severe stroke (odds ratio 1.9) and of death at 30 days (16 versus 6 percent, hazard ratio 3.4) compared to those with an INR above 2.0.

Advanced age is an independent risk factor for bleeding during anticoagulation, and some experts argue that a lower INR target (1.8 to 2.5) is a reasonable compromise between toxicity and efficacy for some patients in this age group [24] since patients with an INR of 1.8 are afforded considerable protection from stroke [21,38,39]. However, the BAFTA trial, in which the mean age of participants was 81 years, found that warfarin (target INR 2.0 to 3.0) was superior to aspirin (75 mg/day) for stroke prevention, with no difference in major bleeding between the two treatment arms [20]. The benefits of warfarin were independent of age (75-79, 80-84, or 85+ years). Furthermore, aiming for an INR range of 1.8 to 2.5 will likely lead to a substantial number of patients whose INRs are below 1.8, which carries an attendant increased risk of stroke but does not significantly reduce the risk of intracranial hemorrhage [21]. Therefore, aiming for an INR of 2.0 to 3.0 provides the best target range for guarding against the sequelae of under- and overanticoagulation. (See "Anticoagulation in older adults".)

Some have suggested that target INR values higher than 2.0 to 3.0 be used for patients at particularly high risk, such as those with prior thromboembolism [24,36]. Evidence from the ATRIA study make it clear that the there is no merit in targeting INR levels higher than the standard range of INR 2.0 to 3.0 [40]. Patients with rheumatic mitral stenosis and AF have not been extensively studied, but INR 2.0 to 3.0 seems a reasonable target, although evidence has been provided that a combination of antiplatelet agent plus anticoagulation targeted at a lower INR level may be beneficial [41]. In patients with AF who have a prosthetic heart valve, the appropriate antithrombotic regimen will depend on the type of valve and its location. This issue is discussed in detail separately. (See "Antithrombotic therapy in patients with prosthetic heart valves".)

Monitoring — The frequency of INR monitoring soon after initiation of warfarin therapy is determined by how quickly the patient can get to the INR target consistently. The initial three months of adjusted-dose warfarin are a particularly high-risk period for bleeding; as a result, especially close anticoagulation monitoring is warranted during this period [42].

After stabilization at a therapeutic level, the INR should be monitored at least once a month, even after a steady state has been established, and more frequently if there has been a change in medication or diet that can alter the effect of warfarin. The method of follow-up will vary from patient to patient and practice to practice. (See "Therapeutic use of warfarin", section on 'Laboratory monitoring'.)

Temporary cessation of anticoagulation — Patients on long-term warfarin therapy may occasionally require cessation of warfarin anticoagulation to prevent bleeding in the periprocedural or perisurgical period or to avoid teratogenesis during pregnancy. Less often, reversal of anticoagulation may be required for emergent surgery or other urgent indications. These issues are discussed in detail elsewhere. (See "Management of anticoagulation before and after elective surgery" and "Gastroenterologic procedures in patients with disorders of hemostasis" and "Anticoagulation during pregnancy" and "Correcting excess anticoagulation after warfarin", section on 'Temporary reversal of warfarin'.)

For low-risk (no prior thromboembolism, mitral stenosis, mechanical prosthetic valve, intracardiac thrombus and preserved left ventricular systolic function) patients with AF undergoing surgery or an invasive procedure, we discontinue warfarin three to four days prior to the procedure and resume warfarin on the evening of the procedure. The 2006 ACC/AHA/ESC guidelines on the management of atrial fibrillation give a weak recommendation to the use of unfractionated or low-molecular-weight heparins as a "bridge" in patients who must stay off oral anticoagulants for more than one week [24]. The 2008 ACCP guidelines suggest considering a heparin "bridge" in higher-risk patients in whom the interruption will exceed two weeks [31].

Although there are no data demonstrating the efficacy of this approach, we often use LMW heparin in the periprocedural period in patients at especially high risk of an embolic event (prior thromboembolism, mitral stenosis, mechanical prosthetic valve, intracardiac thrombus, or severe left ventricular systolic dysfunction). (See 'Patients with valvular heart disease' above.)

The discussion of the management of anticoagulant and antiplatelet therapy in the patient undergoing PCI is found elsewhere. (See "Antithrombotic therapy for intracoronary stent implantation: General use", section on 'Patients who require warfarin'.)

Underutilization and underanticoagulation — Despite the compelling evidence that anticoagulation with warfarin reduces the risk of stroke in most patients with AF, warfarin therapy continues to be underutilized [11,43-49]. In a 1999 review of 13,428 ambulatory patients with nonvalvular AF without a known contraindication to warfarin, only 53 percent were receiving warfarin [44].

Even when anticoagulation is appropriately initiated, it is often discontinued, or the target INR is not achieved. Among 1005 patients in the FRACTAL registry with new AF, warfarin was initially prescribed in 65 percent. However, at 30 months, warfarin use had declined to 44 percent [46]. A similar trend (70 down to 50 percent) was noted in patients who were considered "ideal" candidates for anticoagulation. Recurrent episodes of AF were correlated with higher rates of continued warfarin therapy.

The risks of termination of warfarin therapy, even in patients in whom sinus rhythm has been restored, were highlighted in the AFFIRM and RACE trials. Seventy to eighty percent of ischemic strokes occurred either after discontinuation of warfarin or when the INR was subtherapeutic [48,49]. (See 'Patients with paroxysmal AF' below.)

Another problem is that the maintenance of the target INR is often not achieved, and the failure to maintain a therapeutic INR is associated with worse outcomes [50-53]. The following observations illustrate the range of findings:

• In the SPAF-III trial, only 61 percent of INR values were between 2.0 and 3.0, and 25 percent were <2.0, despite the use of a warfarin nomogram and nursing assistance with INR monitoring and warfarin dose adjusting [50].
• Similar findings were noted in a systematic review of four studies in which the INR was <2.0 on 26 percent of measurements [51]. An INR at this level was associated with a fivefold increase in ischemic events compared to values in the therapeutic range.
• In a retrospective analysis of the SPORTIF III and V trials, patients assigned to warfarin therapy were categorized as having poor, moderate, or good control of anticoagulation (therapeutic INRs <60 percent, 60 to 75 percent, or >75 percent of the time, respectively) [52]. Patients with poor control had, when compared to those with moderate or good control, significantly higher annual rates of mortality (4.2 versus 1.8 and 1.7 percent) and major bleeding (3.9 versus 2.0 and 1.6 percent). In addition, patients with poor control had a significantly higher annual rate of stroke or systemic embolization than those with good control (2.1 versus 1.1 percent).

The importance of time in therapeutic range (TTR) was further emphasized in a secondary analysis from the ACTIVE-W trial of clopidogrel plus aspirin versus vitamin K antagonists (VKA) [53]. A wide variation in TTR achieved across participating centers and countries was observed, and the relative benefit of VKA for patients with AF was strongly associated with average TTR . For example, in the analysis by participating country, an average TTR of 50 percent was needed to demonstrate any benefit of VKA therapy versus clopidogrel plus aspirin. This benefit increased with higher levels of average TTR, while at lower TTR levels VKA was the inferior therapy.

Stroke in anticoagulated patients — Embolic events occurring during adequate anticoagulation are more likely in patients with transesophageal echocardiographic evidence of dense spontaneous echo contrast and low left atrial appendage velocity [54], as well as complex aortic plaque. In addition, the INR is often subtherapeutic, even with careful monitoring, and subtherapeutic values are associated with increased embolic risk [51]. (See "Echocardiographic evaluation of the atria and appendages" and 'Underutilization and underanticoagulation' above.)

While most ischemic stroke in atrial fibrillation patients is due to embolism of left atrial thrombi, other stroke mechanisms should be considered.

Alternatives to chronic warfarin — Although conventional-dose warfarin reduces the risk of embolization in atrial fibrillation compared to no warfarin, its use is difficult because of the requirement for frequent international normalized ratio (INR) monitoring, the need to avoid dietary and medication interactions, the risk of major bleeding, including intracranial hemorrhage, and a suboptimal efficacy in real-world populations [55,56]. (See "Therapeutic use of warfarin" and "Risk of intracerebral hemorrhage in patients treated with warfarin".)

For these reasons, a number of alternatives have been studied, including aspirin, low-dose warfarin plus aspirin, and clopidogrel plus aspirin. Warfarin has been superior to these alternatives in patients at high risk of stroke, while aspirin therapy is considered a reasonable option in patients at low thromboembolic risk. (See "Risk of embolization in atrial fibrillation".)

In addition, novel factor Xa inhibitors and oral direct thrombin inhibitors are being evaluated for their safety and efficacy compared to warfarin. (See 'Dabigatran' below.)

Other oral vitamin K antagonists, such as tecarfarin, which may have advantages to warfarin due to differing metabolic pathways, are under investigation [57].

Aspirin — A meta-analysis of six randomized trials comparing aspirin with placebo found that aspirin reduced the incidence of stroke or TIA by 22 percent (95% CI 2-38 percent); the absolute risk reduction for prevention was 1.5 percent per year [9]. Similar findings, an almost significant 32 percent reduction in stroke risk (odds ratio 0.68, 95% CI 0.46-1.02), were noted in a meta-analysis [11,17]. However, the effect of aspirin appears to be more on minor, noncardioembolic strokes from which elderly atrial fibrillation patients are not spared.

A 2000 meta-analysis did not find conclusive evidence of a reduction in the risk of major stroke with aspirin compared to placebo (odds ratio 0.56, 95% CI 0.19, 1.65) [58].

Randomized trials have shown that aspirin is consistently and substantially less effective than warfarin (except in low-risk patients) (table 1) [4,9,10,17]. The magnitude of the difference was illustrated in an individual patient meta-analysis of six prevention trials (AFASAK 1, EAFT, PATAF, SPAF-II, AFASAK 2, SPAF-III) in patients with nonvalvular AF (76 percent primary prevention) [10]. Patients treated with warfarin were significantly less likely to experience an ischemic stroke (2.0 versus 4.3 per 100 patient-years, hazard ratio 0.55, 95% CI 0.45-0.71); the benefit was similar in patients with chronic and paroxysmal AF. It was concluded that treating 100 patients with warfarin rather than aspirin for one year would prevent 2.3 ischemic strokes. As mentioned above, however, the benefit varies by patient subgroup (graph 2) [18].

There is no solid evidence favoring one dosage of aspirin over another, although several trials have used 325 mg daily. We suggest a dose between 75 and 325 mg daily.

The use of aspirin for secondary prevention of cardiovascular disease in patients treated with warfarin for AF is discussed below. (See 'Patients with cardiovascular disease' below.)

Low-dose warfarin plus aspirin — In contrast to adjusted-dose warfarin, low-dose warfarin (1.25 mg/day or target INR between 1.2 and 1.5) in combination with aspirin (300 to 325 mg/day) should not be used to reduce stroke risk in patients with AF [17,50,59]. In the SPAF-III trial of 1044 patients with AF who were at high risk for embolism (women age >75 years, systolic blood pressure >160 mmHg, poor left ventricular function, or prior thromboembolic event), low-dose warfarin plus aspirin was associated with a much higher morbidity and mortality than adjusted-dose warfarin (graph 4 and graph 5) [50].

One trial (NASPEAF) using a low-dose vitamin K antagonist, acenocoumarol, in combination with the antiplatelet agent triflusal showed improved efficacy and safety of this combination compared to acenocoumarol alone [41]. Limitations of this study prevent a firm conclusion of benefit with this combination.

Aspirin plus clopidogrel — Two large randomized trials have investigated the safety and efficacy of dual antiplatelet therapy in patients with AF. ACTIVE W compared clopidogrel plus aspirin to warfarin and ACTIVE A compared clopidogrel plus aspirin to aspirin alone in patients who were not candidates for anticoagulation with a vitamin K antagonist. All of the patients in the two trials had AF and one or more risk factors for stroke (modal CHADS2 score 2.0 in both; (table 3). The primary end point in both trials was a composite outcome (the first occurrence of stroke, systemic (non-central nervous system) embolization, MI, or vascular death).

The ACTIVE W trial included 6706 patients who were randomly assigned to combined therapy with clopidogrel (75 mg/day) and aspirin (75 to 100 mg/day) or to oral anticoagulation with a vitamin K antagonist (target INR 2.0 to 3.0) [60]. The trial was stopped at an interim analysis after a median follow-up of 1.3 years because warfarin anticoagulation was associated with a significantly lower annual rate of the primary end point (3.9 versus 5.6 percent, relative risk 0.69, 95% CI 0.57-0.85). Although the overall rate of bleeding was significantly increased in the dual antiplatelet group (15.4 versus 13.2 percent per year), there was no significant difference in major bleeding (2.4 versus 2.2 percent per year).

Thus, in AF patients who are candidates for oral anticoagulation due to a high risk of bleeding, dual antiplatelet therapy should NOT be considered an alternative.

The ACTIVE A trial included 7554 patients with AF who were not candidates for anticoagulation and were randomly assigned to combined therapy with clopidogrel (75 mg/day) and aspirin (75 to 100 mg/day) or to aspirin alone at the same dose [61]. The reasons that patients were not considered candidates for anticoagulation included the physician's judgment that such treatment was inappropriate (50 percent), a specific risk for bleeding (23 percent), and strong patient preference (26 percent). Patients were excluded from participation in ACTIVE A if they had documented peptic ulcer disease in the previous six months, significant thrombocytopenia, prior intracranial hemorrhage, or ongoing alcohol abuse. The primary end point, as in ACTIVE W, was the first occurrence of stroke, systemic (non-central nervous system) embolization, MI, or vascular death.

After a median follow-up period of 3.6 years, patients treated with clopidogrel plus aspirin had a significantly lower annual rate of the primary combined end point (6.8 versus 7.8 percent, relative risk [RR] 0.89, 95% CI 0.81-0.98), which was primarily driven by a reduction in stroke (2.4 versus 3.3 percent, RR 0.72, 95% CI 0.62-0.83). On the other hand, dual antiplatelet therapy was associated with a significant increase in the incidence of major bleeding (2.0 versus 1.3 percent per year, RR 1.57, 95% CI 1.29-1.92).

Thus, dual antiplatelet therapy may be a reasonable alternative to therapy with aspirin alone in high-risk patients with AF who CANNOT be treated with warfarin anticoagulation because of lack of access to adequate INR monitoring, widely fluctuating and difficult-to-control INR, or because of strong patient preference following careful consideration of the advantages of oral anticoagulation. As shown in ACTIVE W, dual antiplatelet therapy and oral anticoagulation have similar bleeding risks. Therefore, a patient who would not be a candidate for oral anticoagulation because of bleeding risk is also not a candidate for dual platelet therapy.

Dabigatran — Direct factor Xa inhibitors and oral direct thrombin inhibitors, such as dabigatran etexilate, are being compared with adjusted-dose warfarin in large international trials involving patients with AF. (See "Anticoagulants other than heparin and warfarin".)

The Randomized Evaluation of Long-Term Anticoagulation Therapy (RE-LY) trial, which evaluated the efficacy and safety of dabigatran (at two doses) relative to warfarin, is the first trial to demonstrate that an alternative oral anticoagulant may be superior to adjusted-dose warfarin [62].

In RE-LY, 18,113 patients with AF and at least one risk factor for stroke (mean CHADS2 score 2.1; (calculator 1) were randomly assigned to receive dabigatran at one of two doses (110 or 150 mg) twice daily or to receive warfarin adjusted to an INR of 2.0 to 3.0. Aspirin use was allowed, and approximately 20 percent of patients in each of the three groups received it. Exclusion criteria included important valvular heart disease or a creatinine clearance of less than 30 mL/min/1.73 m2.

After a median follow-up of two years, the following findings were noted:

• The rate of the primary efficacy outcome of stroke (including hemorrhagic stroke) or systemic embolism was 1.53, 1.11, and 1.69 percent per year in the dabigatran 110 mg, dabigatran 150 mg, and warfarin groups, respectively. The differences in rates were driven principally by reductions in ischemic stroke.
• Dabigatran 110 mg met the criteria for noninferiority compared to warfarin (relative risk 0.91, 95% CI 0.74-1.11), while dabigatran 150 mg was significantly more effective than warfarin (relative risk [RR] 0.66, 95% CI 0.53-0.82) or dabigatran 110 mg (RR 0.73, 95% CI 0.58-0.91).
• The rates of hemorrhagic stroke were significantly lower in the dabigatran 110 and 150 mg groups compared to warfarin (0.12, 0.10, and 0.38 percent per year, respectively; RR 0.31, 95% CI 0.17-0.56 and 0.26, 95% CI.14-0.49, respectively).

With regard to the primary safety outcome of the rate of major bleeding, which was defined as a reduction in the hemoglobin concentration at least 20 g per liter, transfusion of at least two units of blood, or symptomatic bleeding in a critical area or organ (including hemorrhagic stroke):

• The rate of major bleeding, which was the primary safety outcome, was 2.7, 3.1, and 3.4 percent per year in dabigatran 110 mg, dabigatran 150 mg, and warfarin groups, respectively. The risk of major bleeding was significantly less with dabigatran 110 mg than warfarin, and dabigatran 150 mg was of equal safety to warfarin (RR 0.80, 95% CI 0.69-0.93 and 0.93, 95% CI 0.81-1.07).

Other findings included a trend toward a higher rate of myocardial infarction and a higher rate of dyspepsia with both doses of dabigatran compared to warfarin. All-cause mortality was significantly reduced by dabigatran 150 mg compared with warfarin.

In addition to an advantage at either dose compared to warfarin, dabigatran does not require monitoring of the INR, may be less susceptible to dietary and drug interactions, and does not have warfarin's narrow therapeutic window [55]. Disadvantages include twice-daily dosing, lack of long-term safety data, and high cost.

For patients with AF taking warfarin to prevent thromboembolism who have difficulty with warfarin therapy, including monitoring with INR testing, twice-daily dabigatran appears to be an attractive alternative. Until a full evaluation of current and pending evidence is available, at this time we strongly recommend anticoagulant therapy with warfarin, as opposed to dabigatran. Dabigatran is not approved for preventing the arterial embolism in patients with atrial fibrillation.

Non-pharmacologic alternatives — Interventional techniques to prevent thrombus formation in the LAA have been attempted. The data evaluating the efficacy of these procedures are presented separately. (See "Left atrial appendage amputation, ligation, or occlusion in patients with atrial fibrillation".)

Prevention summary — Prevention of embolization using long-term antithrombotic therapy with either aspirin (low risk) or warfarin (moderate/high risk; INR 2.0 to 3.0) should be considered for MOST patients with chronic nonvalvular AF who have risk factors for stroke as outlined above. (See 'Risk stratification' above and "Risk of embolization in atrial fibrillation".)

The decision to recommend anticoagulant, aspirin, or no therapy is based upon assessments of the risk of embolism without treatment and the risk of bleeding with antithrombotic therapy. While all current risk stratification schemes provide only modest capacity to predict stroke risk, the CHADS2 score is simple and well validated. As a result, we base our risk-based recommendations for antithrombotic therapy on the CHADS2 scheme [13,14]. (See "Risk of embolization in atrial fibrillation".)

We suggest the following approach:

• Patients with a CHADS2 score of 0 are at low risk of stroke (rate of 0.5 to 1.7 percent per year in the absence of warfarin) and can be managed with aspirin. A dose of 75 to 325mg daily is suggested.
• Patients with a CHADS2 score of 1 are at intermediate risk of stroke (2 percent per year, or perhaps less) and should be treated with either warfarin (INR 2.0-3.0) or aspirin (75 to 325 mg daily). Based on the studies discussed above, which show that warfarin is more effective than aspirin, and which show that aspirin may be better than placebo, we prefer warfarin to aspirin in most CHADS2 patients with a score of 1. The choice between the two therapies will depend upon many factors, including the clinician's assessment of risk, the ability to provide high-quality monitoring of the intensity of oral anticoagulation, the patient's risk of bleeding with oral anticoagulation, and patient preference.

We do not recommend routine screening for aortic plaque using transesophageal echocardiography (TEE) in patients with AF. However, for those patients who demonstrate evidence for complex plaque on TEE performed for another reason, the SPAF-III TEE substudy suggests warfarin is preferred to aspirin.

• - Patient preference is an important issue, since the absolute reduction in stroke risk is likely to be small, but stroke remains a feared outcome. Better patient education can improve understanding of the benefits and risks of warfarin therapy [63].
• - The role of dual antiplatelet therapy with aspirin plus clopidogrel remains unclear in patients with a CHADS2 score of 1 who do not wish to take warfarin. Such patients have a relatively low risk of stroke on aspirin alone (2 percent per year); as a result, the absolute benefit of dual antiplatelet therapy is small. Furthermore, dual antiplatelet therapy and oral anticoagulation have similar bleeding risks, as shown in ACTIVE W. (See 'Aspirin plus clopidogrel' above.)

• Patients with a CHADS2 score ≥2 are at high risk of stroke (4 percent per year and higher). Those with prior thromboembolism have an especially high risk that is nearly 10 percent per year. Anticoagulant therapy with a vitamin K antagonist such as warfarin is strongly recommended for most patients with a CHADS2 score of ≥2 [24].
• For CHADS2 ≥2 patients who cannot take warfarin anticoagulation, dabigatran therapy should be considered. (See 'Dabigatran' above.)

OTHER CONSIDERATIONS

Patients with paroxysmal AF — Patients with paroxysmal atrial fibrillation (AF) should be anticoagulated according to the same criteria if AF is present for a substantial portion of the time [1,50,64]. Embolic events can occur in patients with acute AF for as little as 72 hours [65]. Although the incidence of embolization had been thought to be lower in paroxysmal AF than in chronic AF, two large trials found a similar risk when adjusted for patient age and other risk factors [2,66]. In addition, the guideline statement described above found that the reduction in ischemic stroke risk with warfarin was similar in patients with paroxysmal and chronic AF [1]. (See "Paroxysmal atrial fibrillation".)

Patients treated with rhythm control — Patients treated with rhythm control (cardioversion with or without chronic antiarrhythmic therapy) remain at risk for embolization even when in sinus rhythm, for two main reasons: recurrent episodes of AF are common and asymptomatic in up to 90 percent [67,68], and some patients have other reasons for thromboembolic risk, such as aortic plaque or left ventricular systolic dysfunction. (See "Indications for anticoagulation in heart failure" and "Embolism from aortic plaque: Thromboembolism" and "Rhythm control versus rate control in atrial fibrillation".)

Nonpharmacologic strategies to prevent recurrent AF might reduce the risk of embolic stroke. Approaches that may be considered include the surgical atrial isolation/maze procedure in patients undergoing cardiac surgery for some other reason and pulmonary vein isolation by radiofrequency catheter ablation. However, there are only limited data confirming a reduction in embolic risk with these procedures, and their use should be considered only in selected patients and only by experienced operators. (See "Radiofrequency catheter ablation to prevent recurrent atrial fibrillation" and "The role of pacemakers in the prevention of atrial fibrillation" and "Surgical approaches to prevent recurrent atrial fibrillation".)

Based upon these findings, most patients with AF, regardless of whether a rate control or rhythm control strategy is chosen, should receive antithrombotic therapy.

Patients with hyperthyroidism — The role of warfarin is less well defined in patients in whom the underlying disease can be corrected, as in hyperthyroidism. (See "Causes of atrial fibrillation" and "Cardiovascular effects of hyperthyroidism", section on Cardiac arrhythmias.)

The 2006 ACC/AHA/ESC guidelines recommended oral anticoagulation with warfarin or other vitamin K antagonist to a target INR of 2.0 to 3.0 [24]. This recommendation was made even though it was noted that it remains controversial whether AF in hyperthyroidism is associated with increased thromboembolic risk compared to AF in other settings. Once the patient is euthyroid, the recommendations for antithrombotic prophylaxis if AF persists are similar to those in patients without hyperthyroidism. (See 'Prevention summary' above.)

The 2008 American College of Chest Physicians guidelines on antithrombotic therapy in atrial fibrillation did not make a separate recommendation for patients with hyperthyroidism [1].

Lower-than-normal warfarin doses are usually required, since hyperthyroidism is associated with increased clearance of vitamin K-dependent clotting factors [69].

Patients with cardiovascular disease — A separate issue from its role in the prevention of embolization is the use of aspirin for secondary prevention of cardiovascular disease in patients treated with warfarin for AF. There have been no high-quality randomized trials comparing aspirin plus warfarin to warfarin alone in patients with both AF and atherosclerotic cardiovascular disease in which the primary end point was either arterial embolization or an atherosclerotic event.

One randomized trial of fluindione (an oral anticoagulant) versus fluindione plus aspirin in AF patients was stopped early due to poor recruitment. However, it showed a substantially higher bleeding rate in the combination arm [70].

Indirect evidence argues against the efficacy of a combined strategy. In a post-hoc analysis of the SPORTIF trials, which included a high percentage of patients with cardiovascular disease or at high risk, combination therapy with warfarin plus aspirin, in comparison to anticoagulant alone, did not reduce the rate of MI [71]. The rate of myocardial infarction with aspirin and warfarin (0.6 percent per year) was not significantly different from that with warfarin alone (1.0 percent per year). In addition, warfarin plus aspirin was associated with a significant increase in the risk of major bleeding compared to warfarin alone (3.9 versus 2.3 percent).

Other indirect evidence of the detrimental effect of adding aspirin to anticoagulation in AF comes from recent trials of new anticoagulants in AF, where addition of aspirin increased bleeding rates. In one trial, the indirect factor Xa inhibitor idraparinux was noninferior to warfarin for the prevention of stroke and thromboembolism. However, there was a substantial increase in bleeding, which was nearly twofold higher among those taking concomitant aspirin or thienopyridines [72].

The 2008 ACCP Consensus Conference suggested that a low dose of aspirin (<100 mg per day) or clopidogrel (75 mg per day) may be given with warfarin [1], although the risk of bleeding may be increased, particularly in elderly patients [28]. The 2006 ACC/AHA/ESC guidelines for the management of atrial fibrillation do not address this issue [24].

In summary, there is no evidence that combined warfarin and antiplatelet therapy compared to warfarin alone should be used solely to increase protection against embolization in patients with AF or to improve survival in stable patients with cardiovascular disease [30]. However, combined therapy may be reasonable in selected patients with coronary artery disease, such as those with acute coronary syndromes or those who receive stents, in whom the potential benefits may outweigh the increased risk of hemorrhage [30,73,74]. (See 'Antithrombotic prevention' above and "Chronic anticoagulation after myocardial infarction" and "Antithrombotic therapy for intracoronary stent implantation: General use", section on 'Patients who require warfarin'.)

RECOMMENDATIONS OF OTHERS — Recommendations for the choice of therapy to prevent embolization have been published by the ACC/AHA/ESC and the American College of Chest Physicians (table 2) [1,24]. These guidelines recommend anticoagulant therapy with a vitamin K antagonist, such as warfarin, for most patients with a CHADS2 score of ≥2 and either aspirin or warfarin for patients with a CHADS 2 score of 1 [24].

INFORMATION FOR PATIENTS — Educational materials on this topic are available for patients. (See "Patient information: Atrial fibrillation" and "Patient information: Warfarin (Coumadin®)".) We encourage you to print or e-mail this topic review, or to refer patients to our public Web site, www.uptodate.com/patients, which includes this and other topics.

SUMMARY AND RECOMMENDATIONS — Systemic embolization from atrial thrombi can occur with any form of nonvalvular atrial fibrillation (AF), and many embolic events lead to ischemic strokes. Anticoagulation with warfarin, which has been shown to reduce this risk by almost 70 percent, should be considered for most of these patients. Aspirin reduces the risk by a much smaller amount and is used in some patients at lower risk, as it is associated with less bleeding and is a less burdensome therapy. (See 'Safety' above.)

Recent clinical trials, summarized above, have tested combination antiplatelet therapy with clopidogrel plus aspirin (shown to be superior to aspirin, inferior to warfarin, but with bleeding risk comparable to warfarin) and a novel oral direct thrombin inhibitor, dabigatran. (See 'Alternatives to chronic warfarin' above.)

The following points should be kept in mind when considering and implementing antithrombotic prophylaxis. Recommendations for the prevention of secondary embolism in AF patients with an acute stroke, for the antithrombotic management of patients with an acute embolic stroke, and for the use of antithrombotic therapy in AF patients with valvular heart disease are presented separately. (See "Stroke in patients with atrial fibrillation" and 'Patients with valvular heart disease' above.)

• The type of AF (paroxysmal, permanent, and persistent) does not influence decision-making regarding antithrombotic therapy.
• Risk factors for embolic events have been identified. Validated risk-assessment tools are available and should be used prior to deciding whether to initiate antithrombotic therapy and which agent to choose in all patients with AF. The CHADS2 score is simple and well validated and is commonly used. (See 'Risk stratification' above and "Risk of embolization in atrial fibrillation", section on 'CHADS2 score'.)

• The use of warfarin is associated with a significant risk of major bleeding; intracranial bleeding is the most serious type and is often lethal or disabling. This risk is particularly high in older individuals. However, there is clear evidence for net benefit of warfarin among the elderly, as long as prescribing physicians view such elderly patients as reasonably safe candidates for warfarin. Issues related to anticoagulation in elderly patients are discussed separately. (See "Anticoagulation in older adults".)

• Despite the evidence of benefit for the use of anticoagulation in many patients with AF, underutilization and underanticoagulation are common. For patients with nonvalvular AF, regardless of age and CHADS2 score, the anticoagulation intensity range of INR 2-3 optimizes reduction of ischemic stroke and risk of major hemorrhages. Patients should be in this INR range at least 50 percent of the time to benefit from anticoagulation therapy. (See 'Underutilization and underanticoagulation' above.)
• Newer anticoagulant therapies, in particular dabigatran, offer the promise of stroke prevention equivalent to or better than warfarin, with perhaps an even lower rate of major bleeding.
• For most patients with nonvalvular AF undergoing surgery or an invasive procedure, we discontinue warfarin three to four days prior to the procedure and resume warfarin on the evening of the procedure. Bridging therapy with heparin should be used for patients with mitral stenosis, those with prosthetic heart valves, prior thromboembolism, or severe left ventricular systolic dysfunction. (See 'Temporary cessation of anticoagulation' above.)

Our recommendations for use of antithrombotic therapy to prevent embolization in nonvalvular AF are as follows:

• In nearly all patients with nonvalvular AF and a CHADS2 score greater than 0, we recommend chronic antiplatelet or anticoagulant therapy (Grade 1B). (See 'Antithrombotic prevention' above.)

• For patients with CHADS2 score of 2 or greater :

• - In locations where dabigitran is not available, we recommend warfarin rather than antiplatelet therapy (aspirin, clopidogrel, or both) (Grade 1A).
• - In locations where dabigitran is available, we suggest warfarin rather than dabigitran (Grade 2B). (See 'Alternatives to chronic warfarin' above.)
• - If warfarin is not tolerated and dabigitran is not available, we suggest ASA 325 mg daily plus clopidogrel 75 mg daily, rather than aspirin alone (Grade 2A). However, if the reason that warfarin is not tolerated is bleeding, aspirin and clopidogrel are NOT an option as the combination produces a similar risk of bleeding to that seen with warfarin. (See 'Alternatives to chronic warfarin' above.)
• - For patients whose INR levels on warfarin are erratic or for those who will not comply with frequent INR monitoring, and dabigatran is available, we suggest dabigatran, rather than aspirin plus clopidogrel (Grade 2B). For those patients in whom dabigatran is chosen, we suggest the 110 mg dose; patients who are willing to accept an increase in bleeding risk for an increase in stroke prevention may reasonably choose the 150 mg dose. This recommendation is made while awaiting additional further evaluation, including long-term data on the efficacy and safety of dabigitran. (See 'Alternatives to chronic warfarin' above.)
• - In patients with a CHADS2 score of 1, we suggest warfarin (INR 2.0-3.0) in preference to aspirin (Grade 2B). In deciding between the two, it is particularly important to be sure patients are well informed about their choices and that patient preferences are clearly part of the decision. Aspirin is a reasonable choice for patients in whom the use of warfarin is problematic. Doses that have shown benefit range between 75 and 325 mg daily. (See 'Prevention summary' above.)

• In patients with a CHADS2 score of 0, we suggest antithrombotic therapy with aspirin, rather than either no therapy or anticoagulant therapy (Grade 2B). Doses that have shown benefit range between 75 and 325 mg daily. Patients who are more concerned with the small risk of bleeding while taking aspirin than with the small risk of embolization may prefer no antithrombotic therapy. (See "Lone and low-risk atrial fibrillation", section on 'Anticoagulation'.)
• We recommend a target INR between 2.0 and 3.0 in patients with AF who are taking warfarin (Grade 1B). (See 'Target INR' above.)

ACKNOWLEDGMENT — The editorial staff at UpToDate, Inc. would like to acknowledge Dr. J Philip Kistler, who contributed to an earlier version of this topic review.