An effective defense against intravascular thrombosis involves a dynamic interplay between the vasculature, platelets, the formation of fibrin, and fibrinolysis (4). Vascular endothelial cells are an essential barrier to thrombosis. For example, the endothelial surface contains a key glycoprotein, thrombomodulin, that supports the activation of protein C, a potent natural anticoagulant. Activated protein C rapidly destroys activated factors V and VIII, major participants in the formation of fibrin. Moreover, the glycosaminoglycan heparan is widely distributed on the endothelial surface and avidly binds antithrombin, another natural anticoagulant. When bound to the endothelial surface, antithrombin rapidly neutralizes the clotting enzyme thrombin, as well as activated factor X and other prothrombotic serine proteases.

Vascular endothelial cells also inhibit platelet adhesion and platelet aggregation. When the endothelium is activated by local injury, inflammation, or other thrombogenic stimuli, prostacyclin (PGI₂), a potent inhibitor of platelet plug formation, is released. Finally, under appropriate circumstances, blood vessels may markedly enhance local fibrinolysis via the synthesis and release of tissue plasminogen activator (t-PA).

Injury to normal vascular endothelium serves as a potent stimulus to thrombosis. For example, in a patient with a transmural MI, endothelial cell damage may develop overlying the ischemic area of endocardium. The loss of the protective endothelium with the exposure of a thrombogenic
surface may subsequently culminate in a ventricular thrombus. The ultimate size of the intracavitary clot will be limited by the antithrombotic potential of the surrounding intact endothelium. A similar sequence of events may occur following the rupture of an atherosclerotic plaque within a coronary artery. Platelets and fibrin rapidly accumulate at the area of injury, which may lead to acute coronary artery occlusion and, ultimately, infarction of the myocardium.

Sluggish or disturbed blood flow resulting from ventricular failure or atrial fibrillation greatly enhances the likelihood of thrombosis. The supply of coagulation inhibitors such as antithrombin or protein C to an area of local tissue injury is reduced, and activated clotting factors accumulate instead of being flushed into the circulation, where they can be cleared by the liver.

Coagulation proteins such as prothrombin or factor X normally circulate in their nonactivated (zymogen) configuration. However, a variety of stimuli may convert them to an activated form, predisposing to thrombosis. Examples include surgical procedures, tissue necrosis following infarction, infections, or inflammatory reactions with the release of thrombogenic cytokines such as interleukin-1 or tumor necrosis factor.

The fibrinolytic balance may be altered in some patients with heart disease, contributing to thrombosis. Plasminogen activator inhibitor (PAI-1), a fibrinolytic inhibitor released from endothelial cells, is the major inhibitor of endogenous t-PA, and high levels may impair the fibrinolytic response, predisposing to coronary thrombosis (5). Elevated PAI-1 levels have been associated with an increased risk for first and recurrent MI in young men in several studies (6,7). A specific allele (4G) of a common polymorphism in the promoter of the PAI-1 gene (4G/5G) is associated with higher PAI-1 activity levels in vitro and in vivo. Molecular studies indicate that the increased PAI-1 activity may reflect differential binding of transcription activators and repressors to the promoter polymorphic site, resulting in a higher basal level of transcription of PAI-1 with the 4G allele (8). The 4G/4G genotype was significantly more prevalent in a group of Swedish men with a first MI before age 45 in comparison with controls, suggesting a twofold increase in coronary risk (8). However the 4G/5G polymorphism was not predictive of arterial or venous thrombosis in a cohort of healthy men in the Physicians’ Health Study (9). Elevated levels of D-dimer and t-PA antigen were associated with an increased risk for MI in Physicians’ Health Study participants, reflecting activation of the fibrinolytic system (10,11). However, after adjustment for total and high-density lipoprotein (HDL) cholesterol and other atherosclerotic risk factors, these associations were no longer statistically significant, suggesting that elevated t-PA and D-dimer levels may be a consequence rather than the cause of progressive atherosclerotic vascular disease.

Recent studies demonstrate the importance of vascular tone and its relationship to thrombosis. Intense vasoconstriction appears to promote thrombogenesis, whereas vasodilation may increase blood flow, inhibiting thrombosis. The regulation of vascular tone is complex, and several important mediators are now well characterized. Nitric oxide (NO), previously called endothelium-derived relaxing factor (EDRF), is a free radical product released by vascular endothelial cells in response to a variety of agonists, including thrombin, adenosine diphosphate (ADP), catechols, and various cytokines (12). NO is a potent vasodilator that also inhibits platelet adhesion, activation, secretion, and aggregation, and it promotes platelet disaggregation (13,14,15,16).

Endothelins are a family of small peptides produced by several cell types. Endothelin-1, produced by endothelial cells, is the most potent known vasoconstrictor; it increases vascular tone by its local effects on vascular smooth-muscle cells. It is also a mitogen for smooth-muscle cells. Abnormalities in the NO-endothelin system may contribute to endothelial dysfunction in several different types of heart disease (17).

For example, patients with severe heart failure have impaired NO-mediated vasodilation in response to stimuli (18). Plasma endothelin levels are elevated in severe heart failure (19,20) and are thought to contribute to myocardial damage after MI. There are reports of elevated endothelin levels in asymptomatic patients with atherosclerosis and after acute MI (21,22). Pretreatment of rats with antiendothelin antibodies reduced the degree of myocardial damage induced by MI in an experimental model (23).

The hypothesis that increased platelet reactivity may contribute to arterial thrombosis has been difficult to test, in part because of the lack of sensitive and specific laboratory assays for platelet activation. Patients with extensive peripheral atherosclerosis have evidence of increased platelet-vascular interactions, based on the excretion of
platelet and vascular prostaglandin metabolites in the urine (24). Transient but markedly enhanced platelet reactivity is observed after fibrinolytic therapy for acute coronary thrombosis (25,26). More recently, measurement of platelet surface P-selectin expression by flow cytometry has been used as a marker for platelet activation. One study showed that P-selectin expression increased after the addition of several thrombolytic agents, and suppression of activation required higher doses of aspirin (660 mg per day) (27). Other studies have shown that plasma from patients with acute MI or heart failure enhances platelet aggregation mediated by von Willebrand’s factor (28,29). Patients with severe heart failure, rheumatic valvular disease, and atrial fibrillation have elevated levels of platelet-specific proteins (platelet factor 4, β-thromboglobulin) that reflect platelet activation (30,31,32,33). There is also evidence of a diurnal variation in platelet reactivity that may contribute to acute coronary occlusion (34,35,36,37).

Antiphospholipid antibodies (APL Ab) react with epitopes on phospholipid binding proteins. This heterogenous group of antibodies is divided into two major categories based on the laboratory tests used to identify them. Lupus inhibitors, by definition, are antibodies that interfere with phospholipid-dependent coagulation tests in vitro. Anticardiolipin antibodies (ACL Ab) are detected by an immunologic assay in which cardiolipin-coated, enzyme-linked immunosorbent assay (ELISA) plates are used. Patients often have multiple APL Ab, each of which reacts with a specific antigen. Although a number of target proteins have been identified, β₂-glycoprotein-1 and prothrombin are the most common.

APL Ab are strongly associated with both arterial and venous thrombosis, and they have also been linked to a variety of cardiac complications. Cardiac valve abnormalities are found in up to 36% of patients with the primary antiphospholipid antibody syndrome (PAPS) by echocardiography; these usually involve the mitral and aortic valves (38). Leaflet thickening and irregularity and valvular insufficiency are the most common abnormalities. Although these lesions are often clinically silent, they may predispose to thrombosis and increase the risk for systemic embolism. Patients with APL Ab and thrombosis of histologically normal valves have also been reported (39). There are multiple reports of intracardiac mural thrombi involving both atria and ventricles in association with these antibodies (40,41,42,43).

MI has also been linked to APL Ab. One study found that a significant proportion of young men with an acute MI had elevated levels of APL Ab that persisted for 2 years (44). Additional thromboembolic complications developed in approximately one-third of these men. More recently, two prospective studies showed that ACL Ab are an independent risk factor for future MI and early death in healthy, asymptomatic middle-aged men (45,46). Another recent prospective study reported that the presence of ACL Ab at the time of acute MI predicted subsequent recurrent MI and thromboembolic events (47). There are also multiple reports of myocardial dysfunction or infarction resulting from thrombotic occlusions of the myocardial micro-circulation (48,49).

The mechanisms of thrombosis in patients with APL Ab are still not well understood. Evidence is accumulating that these antibodies may produce an acquired form of resistance to activated protein C (50). Reduced levels of free protein S, which functions as a cofactor for activated protein C (APC), are common in patients with lupus inhibitors, although the mechanism is unknown. The observation that some lupus inhibitors promote increased prothrombin binding to cell and phospholipid surfaces suggests the possibility that accelerated thrombin generation may contribute to thrombosis (51). The addition of APL Ab to cell cultures results in increased tissue factor synthesis by monocytes and endothelial cell expression of adhesion molecules (52,53). APL Ab also bind to oxidized lipoproteins, such as oxidized low-density lipoprotein (LDL), suggesting a potential role in the progression of atherosclerosis and arterial thrombosis (54).

Homocysteine (HC) is an amino acid generated as a by-product during the metabolism of methionine. Normal plasma contains a small amount of HC (average, 10 μM). Hyperhomocysteinemia is strongly associated with premature atherosclerotic vascular disease and both arterial and venous thrombosis (55,56). Even mildly elevated HC levels significantly increase the risk for MI, stroke, and peripheral vascular disease. In the Physicians’ Health Study, the relative risk for MI was increased to 3.4 with HC levels above 15.8 μM (57). Accumulating evidence regarding arterial disease suggests a graded effect on vascular risk rather than a threshold effect, such that the risk would not increase until HC levels exceeded the threshold. A recent meta-analysis concluded that each 5-μM increase in the HC level results in a twofold to fourfold increase in the odds ratio for vascular disease (58). Hyperhomocysteinemia is also an independent risk factor for venous thromboembolism (59).

Hyperhomocysteinemia has several effects that promote atherothrombosis, although the primary mechanism(s) responsible are still unclear. In experimental models, high HC levels have detrimental effects on endothelial cells, platelets, and components of the coagulation and fibrinolytic system that predispose to vascular disease and thrombosis (55,56). There is growing evidence that endothelial injury and dysfunction may be the final common pathogenetic pathway responsible for many of the atherothrombotic effects of HC. HC interferes with the natural antithrombotic properties of endothelium and may impair vascular reactivity by interfering with endothelial cell NO production. High levels also impair fibrinolysis and increase the binding of lipoprotein(a) [Lp(a)] to vascular endothelial cells (60).

Elevated levels of HC result from deficiencies or defects in either the enzymes (cystathionine-β-synthase, N⁵,N¹⁰-methylenetetrahydrofolate reductase) or the vitamin cofactors (folate, vitamins B₁₂ and B₆) involved in the two major metabolic pathways of HC metabolism. There is accumulating evidence that inadequate folate is a major factor contributing to most of the cases of mild to moderate hyperhomocysteinemia in the general population. Vitamin supplementation (folate with or without vitamins B₁₂ and B₆) will reduce
HC levels in most cases, although it remains to be shown that lowering HC levels will prevent vascular complications.

Lp(a) has been identified as a powerful predictor of premature atherosclerotic vascular disease in several large prospective trials (61). Lp(a) is composed of an LDL-like particle and apolipoprotein(a), a large protein with striking homology to plasminogen. Accumulating experimental and clinical evidence suggests that Lp(a) has both atherogenic and prothrombotic effects. For example, Lp(a) accumulates in atherosclerotic plaque, stimulates smooth-muscle cell proliferation, and promotes cholesterol accumulation (62). The structural homology between apo(a) and plasminogen may enable Lp(a) to interfere with fibrinolysis and act as a procoagulant. Lp(a) has been shown to stimulate the release of plasminogen activator inhibitor (i.e., PAI-1) from endothelial cells and to compete with plasminogen for binding on fibrin on the surface of vascular endothelial cells, inhibiting fibrinolysis (62,63,64). It also competes with t-PA in converting plasminogen to plasmin. In an experimental model, Lp(a) transgenic mice were resistant to t-PA-mediated thrombolysis of fibrin thrombi (65). HC increases Lp(a) deposition on a fibrin surface, which suggests an adverse synergistic interaction promoting atherothrombosis (60). Elevated Lp(a) levels (defined as >30 mg/dL) are associated with an increased risk for coronary artery, cerebrovascular, and peripheral vascular disease. Thus, it is well established that high levels of Lp(a) are associated with accelerated atherosclerosis and cardiovascular disease, although it is not yet clear whether they also predispose to thrombosis.

Many patients with heart failure have associated conditions for which antithrombotic therapy is specifically indicated, such as atrial fibrillation, prosthetic heart valves, recent myocardial infarction, and atherosclerotic vascular disease. Warfarin should be standard therapy unless otherwise contraindicated.

The presence of a prothrombotic milieu and the known risk of thromboembolic events in chronic heart failure raise the question of whether broader use of anticoagulation may be indicated. Indeed, several early trials reported improved survival in heart failure patients treated with chronic anticoagulation (70,71,72,73).

However, these studies included many patients with primary valvular heart disease and/or atrial fibrillation, as well as others with dilated cardiomyopathies who were managed with prolonged activity limitation or even bedrest. These patients are at higher risk for thromboembolic events than the usual heart failure patient, and thus, generalization of these findings to patients with heart failure due to either ischemic or nonischemic cardiomypathy in sinus rhythm is not appropriate. Nonetheless, a post hoc analysis of the Studies of Left Ventricular Dysfunction (SOLVD) resurrected this question. It reported that in a combined analysis, patients receiving anticoagulation (in a nonrandomized manner) had a 24% lower incidence of death from any cause and an 18% lower risk for hospitalization for worsening heart failure (74).

The mechanism is not immediately obvious, since as will be discussed later, clinically apparent thromboembolic
events are relatively uncommon in heart failure patients in sinus rhythm. One potential mechanism is the prevention of so-called silent MIs and their contribution to the progression of heart failure. Indeed, one of the few recent postmortem studies in heart failure patients found that a substantial proportion of patients experiencing both sudden death and death attributed to worsening heart failure had recent, unrecognized MIs (75).

Importantly, several trials have demonstrated that anticoagulation is at least as effective as antiplatelet therapy with aspirin in preventing vascular events during long-term treatment post-MI; this may be the reason for the apparent benefit in heart failure patients (76,77).

Although the routine use of anticoagulation in chronic heart failure patients is not recommended, controversy persists as to whether anticoagulation is indicated for patients considered to be at high risk for embolic events. Intracardiac thrombi are common in patients with dilated cardiomyopathies, and they form in dilated atria as well as ventricles. Mural thrombi have been found in 53% to 75% of patients with nonischemic cardiomyopathies in autopsy studies (78), and multiple thrombi involving more than one cardiac chamber are found in nearly 30% of cases, although these data are several decades old and there is a general perception that thrombi now occur much less frequently in this population. The pathogenesis of thrombosis is in large part related to the marked stasis and aberrant blood flow in the dilated and hypokinetic atrium or ventricle. Once formed, the fibrin clot may remain highly thrombogenic because of the residual thrombin bound to its surface, which is relatively inaccessible to inactivation (79).

Patients with heart failure also have evidence of endothelial injury or dysfunction, which may contribute to activation of platelets and coagulation. Several reports of elevated levels of biochemical markers of platelet activation and thrombin generation suggest that patients with moderate to severe heart failure may have a hypercoagulable or prothrombotic state. In one study, patients with the most severe heart failure (reflected by high plasma catechol levels and a low ejection fraction) were the most likely to have biochemical evidence of platelet and coagulation activation (30). In a follow-up study, high baseline levels of activated coagulation markers were reduced by low-dose warfarin therapy, suggesting that anticoagulation may suppress the prothrombotic state associated with advanced heart failure (80). Patients with advanced heart failure have elevated plasma levels of tumor necrosis factor and other cytokines, which may also activate coagulation (81,82). However, it has not yet been shown that elevated levels of these biochemical markers can be used to identify high-risk patients or predict clinical thromboembolic events.

An echocardiographic study suggested that intraventricular thrombi usually do not change in size or motion profile the absence of anticoagulation (83). However, thrombi tend to diminish in size gradually or resolve completely within weeks to months when anticoagulants are given. Other studies suggest that a new thrombus will develop in 10% to 20% of patients with a cardiomyopathy during the subsequent 2 years, whereas 10% to 20% of thrombi resolve within the same time period (84).

The rate of clinically apparent systemic cardiogenic emboli is approximately 2% to 3% yearly, from several older studies (85,86,87). However, these trials are generally retrospective and involve patients referred for echocardiography, in some cases because of suspected intracardiac thrombi. Atrial fibrillation patients were sometimes included. A review of data from the older literature in heart failure patients not receiving anticoagulation reported an average annual incidence of systemic thromboembolism of 1.9%, with a range of 0.9% to 5.5% yearly (88). Prospective data from large clinical trials in the last two decades suggest a rate of systemic embolic events around 1.5% (89,90), although approximately 75% of these are strokes.

The embolic rates in patients with idiopathic dilated cardiomyopathy are similar to those in studies of more heterogeneous populations of heart failure patients, suggesting that thromboembolism results from heart failure itself rather than the specific underlying cause. Although the overall risk for thromboembolism is low, coexisting atrial fibrillation or a history of one or more embolic events in the previous 2 years confers a much higher risk, which approaches 16% to 20% a year (91).

Intuitively, one might think that embolic events occur more frequently in patients with the poorest LV function or the largest left ventricles. However, this has not been confirmed in most analyses. In a retrospective analysis of 6,378 patients enrolled in the SOLVD Prevention and Treatment Trial, no relationship between ejection fraction and embolic risk was observed in men, although in women a 10% reduction in ejection fraction was associated with a 53% increase in thromboembolism (92). One trial in which a significant relationship was observed was the Survival and Ventricular Enlargment (SAVE) trial, which enrolled 2,231 patients with LV dysfunction early after MI (93). There was an 18% increase in stroke risk for each 5% reduction in ejection fraction, and patients with an ejection fraction of 28% or less had a twofold higher risk for stroke than did patients with an ejection fraction of 35% or more. However, this relationship was probably more related to the size of the recent MI than to the severity of LV dysfunction and cannot be extrapolated to chronic heart failure.

Patients with dilated cardiomyopathy are also at risk for pulmonary emboli, with a reported frequency of 5% to 11% (94), and these events are often underdiagnosed. In the analysis of data from the SOLVD trial, pulmonary emboli accounted for 24% and 14% of the total thromboembolic events in women and men, respectively (92). Although right ventricular thrombi are relatively rare, there have been several case reports of biventricular thrombi complicated by pulmonary emboli (95,96).

Several recent and ongoing trials have sought to evaluate the benefit of warfarin anticoagulation in patients with sinus rhythm. The Warfarin and Aspirin Study in Heart
Failure (WASH) randomized 279 patients to warfarin, aspirin, or placebo (97). There were no significant differences in death or other outcomes between the warfarin and placebo group, although the trial was underpowered and embolic events were rare. The Warfarin and Antiplatelet Therapy in Chronic Heart Failure trial (WATCH) was a much larger, three-arm trial comparing warfarin, aspirin, and clopidogrel in 1,587 patients with heart failure and LV ejection fractions ≤35% who were in sinus rhythm (3). Although the final results of this trial await publication, there were no significant differences in the primary end-points of death and nonfatal MI or nonfatal stroke, and there appeared to be fewer strokes in the warfarin-treated patients (98). One further trial, the National Institutes of Health (NIH)-sponsored trial is evaluating warfarin and aspirin in patients with heart failure, ejection fractions ≤35% who are in sinus rhythm, with endpoints of death and stroke (99).

Thus, there are no conclusive outcome data from the modern era to support the routine use of anticoagulation in chronic heart failure patients, either to improve overall outcomes or to prevent embolic events. Therefore, current guidelines do not recommend routine anticoagulation in the general heart failure population (90,100,101,102).

However, there are groups of heart failure patients in whom anticoagulation is indicated because of associated conditions. These are numerated below.