Mechanisms of thrombogenesis in atrial fibrillation

INTRODUCTION — Atrial fibrillation (AF) is associated with substantial mortality and morbidity, largely due to thromboembolism, particularly stroke. This complication can occur with either paroxysmal (intermittent) or chronic (permanent) AF. A number of randomized trials have demonstrated the efficacy of warfarin in reducing this risk both during the course of chronic AF and in the period prior to and after the restoration of sinus rhythm. (See "Atrial fibrillation in adults: Use of oral anticoagulants" and "Prevention of embolization prior to and after restoration of sinus rhythm in atrial fibrillation".)

The factors that promote thromboembolism in AF will be reviewed here.

FACTORS PROMOTING THROMBOEMBOLISM — Almost 150 years ago, Virchow proposed that three conditions should be present for development of thrombosis [1]:

●Abnormalities in blood flow

●Abnormalities in the blood vessel wall

●Interaction with blood constituents

Each of these abnormalities may contribute to thromboembolism in AF [2,3].

As an example, dilated atria, poorly contracting dilated left ventricles, valvular heart disease (particularly mitral stenosis), and congestive heart failure are clinical features commonly associated with stroke and thromboembolism in patients with AF [2]. These abnormalities in blood flow and vessels (the first two components of Virchow's triad) can be related to the presence of structural heart disease or extrinsic interventions such as cardioversion. As will be described below, AF also may confer a hypercoagulable or prothrombotic state [3].

Pooled data from a meta-analysis have demonstrated that independent clinical risk factors for stroke in nonvalvular AF include a history of hypertension and diabetes [4]. Patients with heart failure are also at high risk, particularly those with left ventricular systolic dysfunction or aneurysm formation [5-7]. (See "Atrial fibrillation in adults: Selection of candidates for anticoagulation".)

These clinical criteria are complemented by echocardiographic findings, which have demonstrated that a dilated left atrium, impaired left atrial function, and impaired left ventricular systolic function are independent risk factors for stroke in patients with AF (table 1 and table 2) [7]. (See "Role of echocardiography in atrial fibrillation".)

Left atrial abnormalities — A dilated left atrium and reduced left atrial and left atrial appendage (LAA) blood flow on echocardiography are independent risk factors for thromboembolism. Patients with these abnormalities are more likely to have stasis of blood as demonstrated by the presence of spontaneous echo contrast or "smoke" on transesophageal echocardiography; this increase in echogenicity is thought to represent aggregation of red cells at low shear stress (image 1 and movie 1 and movie 2) [8].

Spontaneous echo contrast has been related to hemodynamic and hemostatic abnormalities and an increased risk of stroke and thromboembolism [8-13]. The presence of both left atrial spontaneous echo contrast and chamber enlargement among patients with nonrheumatic AF is strongly associated with an increased risk for cerebral ischemic events (odds ratio 33.7 in one report) [11]. Left atrial spontaneous echo contrast does not appear to be affected by anticoagulant therapy [14]. (See "Role of echocardiography in atrial fibrillation".)

The fibrosis and inflammation seen in the left atrium of patients with AF are particularly intense in the LAA and may predispose to adjacent thrombosis. In addition, the fibrillating LAA is the only area within the left atrium that is comprised of pectinate muscle and can create an appropriate milieu for blood stasis and thrombus formation.

Atrial stunning after cardioversion — Cardioversion of AF leads to an increased risk of thromboembolism, particularly if patients are not anticoagulated before, during, and after cardioversion. In addition to dislodgement of pre-existing thrombi, embolization may result from de novo thrombus formation induced by impaired left atrial systolic function. (See "Prevention of embolization prior to and after restoration of sinus rhythm in atrial fibrillation".)

The transient atrial contractile dysfunction is also known as atrial "stunning," and can occur whether sinus rhythm is restored spontaneously, by external or internal DC (electric) cardioversion, or by drugs. The duration of the left atrial dysfunction appears to be related in part to the duration of AF. In one report, full recovery of atrial mechanical function was attained within 24 hours in patients with AF for ≤2 weeks, within one week in patients with AF for two to six weeks, and within one month with more prolonged AF [15]. The time course of recovery of left atrial function could explain why the great majority of embolic events in patients who remain in sinus rhythm occur within the first 10 days after cardioversion [16,17]. (See "Hemodynamic consequences of atrial fibrillation and cardioversion to sinus rhythm", section on 'Atrial stunning'.)

Paroxysmal atrial fibrillation — The risk of thromboembolic events in patients with paroxysmal AF and relationship to AF burden (percentage of time in AF) is discussed separately. (See "Paroxysmal atrial fibrillation", section on 'Risk of embolization'.)  

Paroxysmal AF is associated with abnormalities in atrial function, as evidenced by the appearance of spontaneous echo contrast on transesophageal echocardiography [18], and in hemostasis which presumably contribute to the increase in thrombotic risk. One report supported the importance of AF duration as levels of beta-thromboglobulin and platelet factor 4 (markers of platelet activation) were significantly increased during episodes more than 12 hours in duration; there was also a trend toward an elevation in fibrinogen levels in these patients [19]. In another series, patients with paroxysmal AF had intermediate values of fibrinogen and fibrin D-dimer between normals and elevations seen in patients with chronic AF [20]. In contrast, patients with paroxysmal supraventricular tachycardia, who retain active atrial contraction and have a low risk of stroke, had levels of hemostatic markers that were similar to controls in sinus rhythm.

Left ventricular dysfunction — Heart failure by itself confers a risk of stroke and thromboembolism; the risk is additive to that of AF [5,7,21]. A report from the SAVE trial, for example, found a progressive increase in stroke risk during a 42-month follow-up in patients with left ventricular dysfunction; furthermore, every 5 percentage point decrease in left ventricular ejection fraction (LVEF) was associated with an 18 percent increase in the risk of stroke [21]. In addition, there was a significant reduction in longitudinal stroke risk associated with the use of warfarin (relative risk 0.19) or aspirin (relative risk 0.44). (See "Antithrombotic therapy in patients with heart failure".)

Similar findings were noted in a prospective study of 1066 patients entered into three clinical trials evaluating the role of anticoagulation in nonvalvular AF (BAATAF, SPINAF, and SPAF) [22]. The incidence of a stroke was 9.3 percent per year in patients with moderate to severe left ventricular dysfunction, even in the absence of congestive heart failure, compared to 4.4 percent per year in those with normal or mildly abnormal left ventricular function (figure 1).

The presence of a poorly contracting, dilated left ventricle is likely to promote stasis of blood and lead to an increased risk of intracardiac thrombus formation and subsequent embolism. A left ventricular aneurysm has both diastolic and systolic bulging or dyssynergy which result in severe stasis of blood [5]. The incidence of thrombi within left ventricular aneurysms ranges from 14 to 68 percent at postmortem, an observation that is consistent with a 50 to 95 percent incidence at the time of aneurysmectomy [23]. The reported incidence of systemic embolism in patients with a left ventricular aneurysm ranges widely, from 0 to 52 percent [23-26].

Patients with heart failure, particularly those with a left ventricular aneurysm, also demonstrate abnormalities of blood rheology, coagulation, and endothelial function, suggesting the presence of a prothrombotic or hypercoagulable state. As an example, both plasma fibrinogen and von Willebrand factor concentrations may be elevated in heart failure, and platelet abnormalities are evident [27,28]. Since AF also confers a hypercoagulable state, this may be additive to the hemodynamic and hemostatic abnormalities conferred by heart failure [29].

Hypertension — Hypertension is a risk factor for stroke (usually thrombotic) and it increases the risk of stroke associated with AF twofold [2,4,6]. How this occurs is unclear. Hypertension may be associated with a hypercoagulable state due in part to abnormalities in blood rheology and endothelial function [30].

Valvular disease — Valvular heart disease, especially mitral stenosis, increases the risk of stroke in AF 17-fold [31]. There is some evidence that the presence of mitral regurgitation is protective against the development of intracardiac thrombi in chronic AF, presumably due to enhanced turbulence and decreased stasis within the left atrium [5]. As an example, one transesophageal echocardiography study of 169 patients with rheumatic heart disease, 63 percent of whom were in AF, found a preoperative incidence of left atrial spontaneous contrast echo of 1, 30, and 54 percent and of thrombus of 1, 13 and 17 percent for those with mitral regurgitation, combined mitral regurgitation and stenosis, and isolated mitral stenosis, respectively [32]. Further support for this observation comes from a report of 32 patients with mitral regurgitation undergoing mitral valve repair; atrial indexes of hypercoagulability were significantly lower than peripheral venous levels, consistent with the clinical observations of reduced echo contrast and left atrial thrombosis [33].

These findings correlate with a reduced incidence of embolism in mitral regurgitation. The Stroke Prevention in Atrial Fibrillation (SPAF) study, for example, reported that the annual rate of clinical thromboembolism in patients with an enlarged left atrium and abnormal left ventricular wall motion was lower in those with than in those without mitral regurgitation (7.2 versus 15.4 percent) [34]. Chronic mitral regurgitation often leads to left atrial dilatation and dysfunction and associated complications. However, even in the presence of left atrial enlargement, severe mitral regurgitation is associated with a lower incidence of embolism (9 versus 25 percent at 7.4 years in those in whom regurgitation was absent or mild) [35].

Hypercoagulable state — A number of studies that measured indices of coagulability suggest that AF confers a hypercoagulable state, increasing the risk of thromboembolism and stroke [3,12,29,36]. One report of 109 patients with AF fibrillation evaluated the relationship between hemostatic and hemodynamic parameters obtained with transesophageal echocardiography (TEE) and the presence of left atrial thrombus [12]. Compared to patients without evidence of thrombus, those with thrombus had spontaneous echo contrast, reduced LAA velocity, increased plasma concentrations of markers of platelet activation (beta-thromboglobulin and platelet factor 4), increased plasma markers of thrombogenesis (thrombin-antithrombin complexes, D-dimers), and evidence of endothelial damage/dysfunction (elevated plasma and endocardial levels of von Willebrand factor, which is released from damaged endothelium) [37]. (See "Coronary endothelial dysfunction: Clinical aspects".) A multiple logistic model identified LAA velocity, beta-thromboglobulin, and von Willebrand factor, but not spontaneous echo contrast, as independent associates of left atrial thromboembolism [12].

Activation of the coagulation system has been identified in other reports of both paroxysmal [19,20] and chronic AF [38-44]. This effect is independent of the presence or absence of underlying structural heart disease. As noted above, reversion to sinus rhythm results in normalization of hemostatic markers within two to four weeks [45,46].

Anticoagulation in patients with AF also alters the hypercoagulable state as illustrated by the following observations:

●Fibrin D-dimer levels are increased in patients with AF. In one study, fibrin D-dimer levels were highest in patients who were not receiving any antithrombotic therapy, intermediate in those on aspirin, and lowest in those treated with warfarin [41]. Similar findings were noted in another report in which patients with AF treated with warfarin had lower levels of prothrombin fragment F1 + 2 and thrombin-antithrombin complex than those treated with aspirin or no antithrombotic therapy [42].

●High fibrin D-dimer levels have been associated with an increased rate of embolic events in patients with AF on oral anticoagulant therapy [47]. In one study, elevated D-dimer levers (≥0.5 mcg/mL) were associated with more thromboembolic events compared to those with low D-dimer levels (hazard ratio 15.8, 95% CI 3.33-75.5) [48]

●In a substudy from the AFASAK trial, 100 patients with chronic AF were randomized to treatment with fixed minidose warfarin 1.25 mg daily alone or in combination with aspirin 300 mg/day, conventional warfarin therapy with dose adjusted to maintain an International Normalized Ratio (INR) between 2.0 and 3.0 or aspirin 300 mg daily. Patients treated with warfarin at any dose demonstrated a significant rise in the INR with a corresponding reduction in prothrombin fragment F1 + 2 [49].

●The degree of anticoagulation with warfarin appears to be important. In one report, ultra low-dose warfarin (1 mg/day) did not significantly alter plasma fibrin D-dimer or beta-thromboglobulin levels [50]. A second study found that treatment with aspirin (300 mg daily) plus low-dose warfarin (1 or 2 mg daily) or low-dose warfarin alone (2 mg daily) did not significantly reduce any of the hemostatic markers; in contrast, there was a significant reduction of fibrinogen and fibrin D-dimer with dose-adjusted warfarin [51].

Evidence of endothelial damage/dysfunction in AF is provided by the following observations:

●Endocardial damage and disorganization of the LAA endocardium has been described in the setting of mitral valve disease, especially where AF is present [52].

●Abnormal plasma indices of endothelial damage/dysfunction, such as vWf, which have been related to thrombogenesis [41], stroke risk [53], and adverse prognosis [54]. (See 'Left ventricular dysfunction' above.)

●Increased levels of circulating endothelial cells (CECs), an index of endothelial damage in the setting of AF and target organ damage (heart failure, stroke, myocardial infarction) [55].

Mechanisms — Although these observations suggest that AF is associated with a hypercoagulable state and endothelial dysfunction, the precise mechanisms by which this might occur are uncertain [3,12,56]. Abnormalities in cardiac blood flow (sluggish, slow flow within the atria) may be partly responsible, adding to an endothelial disturbance in the pulmonary vasculature [41,43]. The latter effect may stimulate lung macrophages to produce hepatocyte stimulating factor (now known to be interleukin IL-6), increasing the hepatic synthesis of fibrinogen, perhaps in a similar manner to smoking [57].

Other data suggest that inflammation may contribute to the hypercoagulable state in AF [58]. As an example, high plasma levels of C-reactive protein (CRP) and interleukin-6 (IL-6) among patients with AF are independently related to indices of the prothrombotic state in AF (eg, CRP to fibrinogen, IL-6 to tissue factor) [59]. CRP has also been related to the presence of dense spontaneous echo contrast (SEC) in the left atrium or LAA on transesophageal echocardiography [60]. SEC is a well-recognized independent predictor for stroke and thromboembolism in AF. Indices of the prothrombotic state and inflammatory markers (IL-6, but not CRP) may predict stroke and vascular events in AF [54,61]. In a larger series of 880 AF patients, CRP was positively correlated to stroke risk and related to stroke risk factors and prognosis (mortality, vascular events) [62].

Other mechanisms stimulating the prothrombotic state in AF that have been explored include matrix metalloproteinases (MMPs) and their inhibitors (tissue inhibitors of matrix metalloproteinases [TIMPs]) as well as growth factors. In one report, plasma levels of MMP-1 were lower and plasma levels of TIMP-1 were higher in 48 consecutive patients with permanent nonvalvular AF compared with controls [63]. However, these values were not independently associated with the presence of AF on multivariate analysis. Instead, clinical factors (that is, age, ischemic heart disease, or hypertension) and echocardiographic variables (end-diastolic left ventricular diameter or left ventricular mass index) were found to be independently associated with MMP system.

Abnormal growth factors, such as vascular endothelial growth factor (VEGF, also a marker of angiogenesis) have been related to tissue factor (TF) upregulation — and therefore coagulation — at least in cancer pathophysiology. Among AF patients, TF levels have been shown to be significantly correlated with plasma VEGF levels [64]. In separate studies, plasma levels of TF and VEGF have also been shown to be increased in atherosclerosis, a condition whose pathophysiology involves a tendency to thrombosis and angiogenesis [65,66]. Similar observations have been noted for angiopoietin, another index of angiogenesis [67].

Other possible mechanisms that have been proposed for the hemostatic abnormalities include neuroendocrine activation [68], slow flow itself [69], increased expression of markers of platelet activation including P-selectin and CD63 [70,71], and elevated serum levels of lipoprotein(a) which is structurally similar to plasminogen and may have antifibrinolytic action [72].

In a series that included 121 patients with AF, 78 with a history of AF but in sinus rhythm at the time of the study, and 65 control subjects, markers of platelet activation were increased in patients with AF and in those with a history of AF [71]. However, multivariable analysis suggested that the degree of platelet activation was probably due to the underlying cardiovascular diseases associated with AF rather than the AF itself. Indeed, another marker of platelet activation, soluble CD40 ligand levels, was not related to the risk of stroke nor to prognosis [62].

Prognostic value of elevated biomarkers — Measures of the prothrombotic state in patients with AF, using biomarkers such as von Willebrand factor (vWF) or D-dimer (both of which are discussed in the section above), have been evaluated for their prognostic value. (See "Clinical use of coagulation tests", section on 'Fibrin D-dimer'.)

With regard to vWF, the following observations have been made:

●Among 1321 patients with AF, there was a significant stepwise increase in plasma vWf levels as the risk of stroke increased [53].

●Two studies have suggested that the addition of plasma vWf levels to traditional risk predictors may improve the ability to predict stroke and vascular events [54,73].

●In a study of 229 patients with permanent AF who were stabilized on warfarin therapy for at least six months, high plasma vWF levels (>221 international units/dl) were an independent predictor for adverse events, including death, stroke, and bleeding during two years of follow-up [74].

Two studies have found that D-dimer levels added to the ability of traditional risk factors to predict adverse outcomes [47,48], while one did not [74].

Other biomarkers have been tested in highly selected clinical trial cohorts to predict the risk of stroke, thromboembolism, or bleeding. These have included troponin, natriuretic peptides, and growth differential factor 15 (GDF15) [75-77]. These have led to proposals of biomarker-based stroke and bleeding risk scores, ABC-stroke, and ABC-bleeding, respectively [78]. In the real world clinical setting, the ABC-stroke and ABC-bleeding scores did not offer any advantage over clinical factor-based scores, such as CHA₂DS₂-VASc and HAS-BLED [79,80]. Also, the small incremental value of improved prediction of high risk (at least statistically) should be balanced against the simplicity and practicality for everyday clinical use [78].  

In another study, addition of multiple biomarkers enhanced the predictive value of CHA₂DS₂-VASc and HAS-BLED, although the overall improvement was modest, the added predictive advantage over original scores was marginal, and decision curve analyses found lower net benefit compared with the original clinical scores [81].

Many of these biomarkers are predictive of thromboembolism, bleeding, death, heart failure, and hospitalizations, as well as noncardiovascular conditions, such as glaucoma progression. This may lead to confusion amongst clinicians over which end point to focus on.

Use of indices of a prothrombotic or hypercoagulable state to test anticoagulation therapies — Indices such as fibrin D-dimer are indicative of a prothrombotic state in AF. They have been used as biomarker surrogates of thrombogenesis in AF when testing anticoagulation regimes in Phase 2 clinical trials and to help with dose selection for large Phase 3 clinical trials [82-85].

SUMMARY — The mechanisms leading to an increased risk of stroke, thrombus, and embolism in atrial fibrillation (AF) are multiple, complex, and closely interact with each other. Many of these factors can be explained by Virchow's triad for thrombogenesis. Consideration must be given to the patient's age, the presence of structural heart disease, and clinical risk factors such as previous hypertension, diabetes, various biomarkers, or heart failure (particularly depressed left ventricular systolic function). In addition, AF appears to be associated with a reversible hypercoagulable state, which complements the clinical and echocardiographic features in explaining the mechanisms of thrombosis in this common disorder.