43 Antithrombotic agents

Heart failure (HF) and the causes of HF provide, theoretically, a powerful pathophysiological substrate for thrombosis but there is remarkably little evidence that the rate of thrombotic events can be reduced by antithrombotic therapy.^1,–4 HF is often associated with coronary artery disease and with atrial fibrillation,^5,6 two conditions for which long-term antithrombotic agents are traditionally given. Accordingly, many consider it unnecessary to investigate whether antithrombotic agents are effective in HF, since they consider the result a foregone conclusion. Some would even consider it unethical to withhold antithrombotic agents in patients with HF. Such an opinion reflects a general lack of rigour in the analysis and interpretation of long-term trials of antithrombotic agents for cardiovascular disease, undermining the scientific basis of medicine.

It is widely perceived that patients with HF have high rates of vascular occlusive events, including myocardial infarction (MI) and stroke,⁷ but this may not be true. Reported rates of MI and stroke are certainly low compared to mortality in patients with HF.¹ Moreover, treatment directed at reducing the risk of vascular events, including aspirin,^8,9 statins,¹⁰ and coronary revascularization,¹¹ has not yet been shown to reduce mortality in patients with HF, even though statins reduce the rate of nonfatal vascular events.

In summary, the contribution of vascular occlusive events to the morbidity and mortality of HF is uncertain. Too many opinions have been based on too few facts, resulting in recommendations about management that are only weakly supported by evidence. A lot of this stems from misinterpretation of the effects of aspirin in long-term trials of patients with atherosclerotic vascular disease.¹²

While the precise history and components of Virchow’s triad are disputed, in principle it describes the three categories of factors that contribute to thrombosis: blood stasis; reduced integrity of the walls of chambers and vessels containing blood; and hypercoagulability of the blood itself (Fig. 43.1).¹³ Changes in one factor will affect the others.

Cardiac output is usually normal at rest until HF is advanced. Low cardiac output is unlikely to play a role in vascular occlusive events in most patients. Myocardial disease may cause ventricular dilatation, regional wall motion abnormalities and, more rarely, aneurysm formation, which may cause local stasis and increase thrombotic risk. However, the atria are likely to be a more common source of thrombi. The atria dilate and hypertrophy in response to atrial hypertension secondary to valve disease or ventricular dysfunction, either systolic or diastolic. This is a powerful stimulus to the development of atrial fibrillation, which further contributes to sluggish flow around the atrial walls and left atrial appendage, greatly increasing the risk of stroke.¹⁴ Devices to occlude the left atrial appendage have been advocated as means to avoid blood stasis as an alternative to anticoagulation. Further research is required before this approach can be recommended as a mainstream therapy, although it might be considered for patients who cannot receive anticoagulants.

In the arterial system, atheroma and aneurysms may cause turbulence and pockets of stasis, provoking changes in the vessel wall that promote atherogenesis and predisposing to thrombotic risk. High venous pressure combined with reduced mobility will increase the propensity to venous stasis and thrombosis.

Atheroma will affect most patients with HF, even if coronary artery disease is not the primary cause of cardiac dysfunction, and will alter vascular compliance, create turbulent flow, and contribute to focal arterial endothelial dysfunction. However, there are other causes of systemic arterial and venous endothelial dysfunction in patients with HF, including oxidative stress, and activation of neuroendocrine, inflammatory, and haemostatic systems. Endothelial dysfunction is complex and reflects activation of systems likely both to promote and retard thrombosis. Keeping the activation of these countervailing systems in balance is important.^15,–17 Dysfunctional endothelium secretes more endothelin, von Willebrand factor, and adhesion molecules that may cause vasoconstriction and thrombosis.¹⁵ Nitric oxide metabolism is compromised, leading to impaired flow-mediated dilatation and increased platelet adhesion.¹⁸ On the other hand, there is marked activation of vasodilator prostaglandins in HF that mitigates the thombotic tendency, provided clumsy physicians do not intervene. Inhibition of vasodilator prostaglandin synthesis by nonsteroidal anti-inflammatory drugs (NSAIDs), including aspirin in doses as low as 75 mg/day, may cause arteriolar and venous constriction (Fig. 43.2),^16,17 enhance endothelin-induced vasoconstriction,¹⁹ exacerbate renal dysfunction,²⁰ accelerate atherosclerosis,^21,22 and might also increase the risk of vascular occlusion.²³

There is a widespread view that the primary event causing arterial vascular occlusion is thrombotic. This may not be true. Histology suggests that haemorrhage into plaque may often be the primary occlusive event, with thrombosis a secondary feature.^24,25 Clearly, antithrombotic drugs could increase the risk of plaque haemorrhage, accelerating plaque growth²⁵ or causing acute events. This may be why there is so little evidence that long-term antiplatelet or antithrombotic therapy reduces mortality in patients with (or at risk of) cardiovascular disease. In the acute setting the haemorrhage has already occurred and the main risk is thrombosis, whereas in the chronic setting the increased risk of haemorrhage into plaque may negate any benefit acquired from inhibition of thrombosis.

As with the endothelium, there is evidence of activation of both pro- and antithrombotic factors.^26,–28 The most consistently elevated haemostatic factors are D-dimer and von Willebrand factor, the former reflecting increased breakdown of crosslinked fibrin and the latter a manifestation of endothelial dysfunction. Both are independently related to an adverse prognosis in patients with HF and are similarly affected in patients with or without ischaemic heart disease. The severity of HF, particularly right HF leading to hepatic congestion, is an important determinant of haemostatic dysfunction. Disturbance of other haemostatic factors has been investigated less often, but is generally consistent with a prothrombotic state held in check by increases in thrombolytic activity.

While knowledge of the factors that predispose to thrombosis may be of value in deciding theoretically which interventions might be of value for patients with HF, the clinical setting is sufficiently complex that theory should be put to the clinical test before making firm recommendations.

As with all received wisdom, the perception that patients with HF are at increased risk of many different types of arterial and venous occlusive events should be questioned. Indeed, careful scrutiny of clinical trial data indicates that patients with HF have somewhat lower rates of nonfatal vascular events than those who do not. This may be deceptive. Sudden death could be the most common manifestation of vascular events in patients with HF.^29,–31 Alternatively, vascular events may be less likely to be diagnosed in patients with HF.

The low rate of MI in patients with HF could be explained by a change in its presentation and the shortened life expectancy of HF. A patient who dies of an arrhythmia or of HF is no longer at risk of having a vascular event. Alternatively, patients with HF may be at increased risk of both painless MI and of sudden death as manifestations of vascular events.³² Increases in troponin are common during exacerbation of HF and augur a poor outcome.³³ Although raised troponin may reflect myocardial stress due to HF, it also probably reflects coronary vascular occlusion. Prior MI, comorbid diabetes, and HF itself may all conspire to denervate the heart and reduce the pain associated with myocardial ischaemia and infarction. In studies of hypertension and diabetes that included few patients with HF, about one-third of all MIs diagnosed were silent.^3,34,35 Presumably, many more silent infarcts remained undiagnosed and the rate of undiagnosed MI in patients with HF may be substantially higher than in other patients. Vascular occlusion may also be more likely to trigger ventricular arrhythmias as a result of either worsening ventricular function or increased electrical instability, resulting in sudden death.²⁹ An effective treatment for reducing vascular occlusion might reduce the rate of worsening HF and sudden death but might have little impact on the rate of classical MI.

The rate of clinically overt strokes is similar in patients with vascular disease with and without HF. Thus the ratio of stroke to MI in patients with HF is higher than in those without. This is probably because strokes are difficult to conceal, whether or not HF is present. The origin of strokes in HF is uncertain. Some will be cardioembolic and some due to cerebrovascular disease, but atherosclerosis in the aortic arch may also be an important source. Age, atrial fibrillation, diabetes, and atheroma are common risk factors for stroke and HF. Well-treated patients with HF are now unlikely to have a high arterial pressure and this, as well as improved management of atrial fibrillation, may account for the decline in the rate of stroke in patients with HF over the last 30 years.¹ Patients with more severe HF or left ventricular dysfunction are at greater risk of stroke.^7,36 The risk of haemorrhagic stroke will be increased by either aspirin or warfarin. MRI suggests a high prevalence of subclinical cerebral infarction in patients with HF,³⁷ but similar rates might exist in age-matched patients with vascular disease but not HF.

Deep venous thrombosis and pulmonary embolism are rare diagnoses in ambulant patients with HF, but common in patients confined to bed because of worsening HF or concomitant illness.³⁸ The proportion of sudden deaths due to pulmonary embolism is hard to quantify but is generally considered low.

Ultimately, the two most common manifestations of vascular events in HF may be worsening cardiac function and sudden death. Autopsy data show a high rate of fresh coronary occlusion amongst patients who have died in either of these ways.^31,39

MI is an important cause of ventricular dysfunction leading to HF. Surprisingly, no substantial, long-term trial of aspirin after MI at doses less than 300 mg/day exists, and no long-term trial of aspirin at any dose has ever shown a reduction in mortality(Fig. 43.3).³ Among patients with HF after MI, the two largest long-term trials showed trends to a higher mortality in patients assigned to aspirin compared to placebo.^3,40,41 The validity of the meta-analysis of trials of aspirin is undermined by strong evidence of publication bias.¹² Many patients stop their aspirin therapy within a few months of a MI, perhaps because the patients have a better grasp of the evidence than their doctors! Indeed, a large trial of older patients likely to have stable coronary disease showed that aspirin increased the risk of fatal or nonfatal MI.²³

The data on warfarin after MI are only slightly more encouraging. Trials of warfarin conducted in the 1960s were widely considered to be neutral because they failed to reduce mortality by more than 75%—they reduced mortality only by 60%, with mortality dropping from 9.9% to 4.0% amongst the 4814 patients enrolled.⁴² How times and expectations have changed in clinical trials! The landmark WARIS trial was conducted in only two hospitals and using one central laboratory.⁴³ Although the study was positive, it is not clear that the results can be replicated. Indeed, the WARIS-II trial failed to show that aspirin, warfarin, or their combination had substantially different effects on outcome.⁴⁴

Atrial fibrillation (AF) is another common condition that commonly complicates, and is complicated by, HF.^45,–47 There have been numerous trials of antithrombotic agents in patients with AF. There are two reasons why the findings are relevant to HF. Firstly, most of these patients probably had HF,^45,–47 although many protocols paid insufficient attention to the diagnosis. Moreover, the event rates for patients without HF were low and probably not that much above the background population risk for stroke or other cardiovascular events (Fig. 43.4).⁴⁶ The risk of events and of death resided almost entirely in the patients who had some manifestation of HF. Although there is a lot of evidence that HF confers increased risk on patients with AF, the converse is far from certain.⁴⁸ Indeed, it might be considered that the trials were really trials of HF among patients who just happened to have AF. In other words, people have jumped to the conclusion that anticoagulants should be used for AF, but it might be that AF is just a surrogate marker for HF and the trials could be interpreted as evidence for an effect of anticoagulants in HF.

Aspirin generally failed to reduce thromboembolic events in patients with AF and was inferior to warfarin in patients with concomitant major ventricular dysfunction or HF.⁴⁹ Aspirin did not reduce mortality in these trials. Although recent trials have shown that the combination of aspirin and clopidogrel is inferior to warfarin, the combination is superior to aspirin alone for patients with AF who are ineligible for treatment with warfarin.^50,51 On the other hand, clopidogrel alone might be a better choice than in combination with aspirin for patients with vascular disease.⁵² More recently, direct thrombin antagonists proved as, or more, effective than warfarin in reducing stroke and death in studies of AF that included many patients with HF.⁵³

There are biologically plausible reasons why aspirin may have adverse effects in patients with HF.³ Neuroendocrine systems that cause vasoconstriction and other adverse effects are activated in patients with HF, but so too are systems that could be considered defence mechanisms, including natriuretic peptides and vasodilator prostaglandins.⁵⁴ Although aspirin might reduce platelet activation associated with HF, it may also reduce vascular prostaglandin production leading to an increase in vascular resistance, sodium retention, worsening renal function, and reduced vascular defences against platelet adhesion.^16,20,55,56 Also, ACE inhibitors may exert some of their benefit through increased prostaglandin synthesis which aspirin may counteract. Even small doses of aspirin may be sufficient to block both the arteriolar and venous response to a variety of prostaglandin-mediated vasodilator agents and such effects may last for 24 h or more,^16,55,56 suggesting that assertions, largely based on in vitro evidence, that 75 mg/day affects only platelets and not vessel wall prostaglandins, are wrong.

Aspirin remains the most widely used antithrombotic agent in patients with HF and in sinus rhythm.¹⁰ Small randomized controlled trials have generally, but not always, shown that aspirin impairs the haemodynamic response to ACE inhibitors.^3,57,58 Small randomized studies also suggest that plasma concentrations of natriuretic peptides (cardiac stress hormones) are lower when patients take clopidogrel rather than aspirin,^59,60 that arterial function is better on clopidogrel than aspirin,⁶¹ and that oxygen uptake during exercise is impaired by aspirin^62,63 and is better on clopidogrel.⁶⁴

Observational trials can be used to argue any preferred point of view about the effect of aspirin on outcomes in HF, but such an approach to the prediction of drug effects is now generally discredited. Retrospective analysis of the SOLVD study suggested that aspirin use was associated with a lower morbidity and mortality,⁶⁵ with more prominent effects observed in the prevention arm rather than the treatment arm of the study, despite the poorer prognosis of the latter group. However, these observations many not be evidence of a therapeutic benefit from aspirin but simply reflect the fact that aspirin was given to patients at lower risk. Patients treated with aspirin had higher baseline ejection fraction, were less likely to have New York Heart Association (NYHA) class III/IV HF and more likely to be on a β-blocker. Other observational datasets have reported conflicting results.³

Cardiovascular prophylactic aspirin is a powerful risk factor for gastrointestinal bleeding, increasing with age and accounting for 30% or more of all major gastrointestinal haemorrhage in patients aged over 60 years.⁶⁶ There is little evidence that the risk is altered by reducing the dose of aspirin⁶⁶ or switching to enteric-coated preparations.⁶⁷ Compared to patients who do not have HF, those with HF are at markedly increased risk (odds ratio 5.9; 95% CI 2.3–13.1) of a major gastrointestinal haemorrhage when taking aspirin.⁶⁸

In large randomized trials of ACE inhibitors in patients with HF or ventricular dysfunction, the effect of ACE inhibitors on mortality was reduced in the presence of aspirin (from 26% to 14%, test for interaction p = 0.04)⁶⁹. In the most relevant trial, SOLVD (n = 6512),⁶⁵ patients taking aspirin derived no mortality benefit from enalapril (hazard ratio for death with enalapril compared to placebo if taking aspirin 1.10 [0.93–1.30] compared to 0.77 [0.67–0.87] if not taking aspirin; a striking p = 0.0006 for the interaction. There was a similar trend for the composite of death or hospitalization for HF (HR 0.81 vs 0.71; p = 0.09). More recently, a meta-analysis of the HOPE and EUROPA studies (n = 25 515) suggested that the reduction of fatal and nonfatal vascular events by ACE inhibitors was halved (from about 40% to 17%) in the presence of aspirin (test for interaction p = 0.003) (Fig. 43.5).⁷⁰

The reason for the interaction between aspirin and ACE inhibitors is unclear. It could reflect a reduction in the benefit of ACE inhibitors by aspirin. ACE inhibitors block the production of angiotensin II and impede the breakdown of bradykinin. The latter action may enhance the production of vasodilator prostaglandins and this action may be lost in patients taking aspirin. Alternatively, aspirin and ACE inhibitors may exert benefits by a similar pathway and therefore the benefits of the treatments could be less than additive.³ Aspirin has also been reported to reduce the improvement in ventricular function with carvedilol,⁷¹ although this may just reflect the association between aspirin and aetiology of ventricular function.⁷²

The most important aspect of managing an adverse interaction between therapies is to be sure that both components are required. We do not know that long-term aspirin has ever been effective for the management of vascular disease. With the concurrent use of ACE inhibitors, aspirin may have been rendered worse than useless or, less likely, may finally be in an environment in which it works. This needs clarification. In the meantime, there is no evidence to support its long-term use in patients who require ACE inhibitors.

Antiplatelet agents that do not block cyclooxygenase, such as dipyridamole and clopidogrel, may be preferred in patients with HF (Fig. 43.6).

One small randomized trial of 28 patients followed for 1 year suggested that dipyridamole was associated with better cardiovascular function.⁷³ Another study suggested that aspirin plus dipyridamole was associated with a slightly lower incidence of HF compared to clopidogrel.⁷⁴ This could reflect a benefit from the vasodilator effects of dipyridamole. However, the incidence of HF was so low that it suggests many cases were missed.

Clopidogrel exerts its actions by inhibiting the binding of ADP to platelet receptors, thereby reducing activation of the glycoprotein GPIIb/IIIa which is the binding site for fibrinogen. As noted above, compared to aspirin, treatment with clopidogrel is associated with better exercise capacity, lower serum creatinine and lower plasma natriuretic peptide concentrations. A post-hoc analysis of the CAPRIE study conducted on patients with ischaemic heart disease who had had cardiac surgery and who had a much higher prevalence of ventricular dysfunction and HF, suggested a strikingly greater effect of clopidogrel, compared to aspirin, on vascular morbidity and mortality.⁷⁵

Coumadin anticoagulants act specifically on vitamin-K dependent clotting factors. They are not known to have direct effects on the vasculature, renal function, or electrolytes and do not appear to interact with neuroendocrine systems (Fig. 43.7). By reducing thrombin formation and fibrin deposition, anticoagulants may additionally reduce platelet activation. Anticoagulants may reduce the risk of dying from cancer, either by acting as an early warning system for bowel and bladder tumours, which are picked up earlier because of the increased risk of bleeding, or because of the more intense haematological surveillance associated with anticoagulant monitoring.⁷⁶

For decades there have been strong advocates, supported by guidelines, for the routine use of anticoagulants in patients with HF, and surveys and trials in patients with HF suggest that more than 20% of patients in sinus rhythm receive such therapy.^5,6 In SOLVD, warfarin was associated with a reduction in both morbidity and mortality without evidence of any interactions with ACE inhibitor therapy.⁷⁷ Anticoagulants are inconvenient because they require regular monitoring: they gained a reputation as dangerous drugs before modern monitoring techniques were introduced, and there were few conclusive trials to support their use other than in AF. Aspirin was ‘sold’ as a more convenient, safer alternative that was just about as effective. In the context of HF, warfarin was thought to be an agent only to reduce the risk of stroke, a relatively uncommon event, and for the management of ventricular thrombus, although there is no evidence to support such practice.¹

HF complicates the management of vitamin-K-dependent anticoagulants. Hepatic congestion reduces the synthesis of clotting factors and may result in over-anticoagulation. Amiodarone and other drugs interfere with warfarin metabolism. Intercurrent illness requiring antibiotics, and procedures that require treatment to be stopped, also make accurate control more difficult, adding to the hazard of therapy. Direct thrombin antagonists may give rise to fewer problems.

The Warfarin-Aspirin Study in Heart Failure (WASH)^8,78 was the first substantial randomized controlled trial of antithrombotic therapy for HF. It compared no antithrombotic therapy, aspirin (300 mg/day), or warfarin (target INR 2.5) in patients with HF who also had left ventricular systolic dysfunction (LVSD) and required diuretic therapy. The study was not blinded and was designed as a pilot study to address the feasibility of doing a trial that included a placebo or no-antithrombotic therapy group. The trial was slow to recruit, afflicted by concerns (on the part of patients as well as investigators) not only about withholding antithrombotic therapy but also about randomization to warfarin without a compelling reason, given the perceived risk and inconvenience. The concerns influenced the design of the WATCH and WARCEF studies (see below) which both avoided the problem of a no-antithrombotic treatment group but still struggled to recruit patients, who were reluctant to accept warfarin without a strong mandate.

About 60% of patients in WASH had definite ischaemic heart disease and only 15% of patients were proven to be free of coronary disease. Over a mean follow-up of 27 months, 70 (25%) of the 279 patients enrolled died but only 18 had a MI and only 4 had a stroke (Fig. 43.8). There were no significant differences in these outcomes amongst assigned groups but a trend to a worse outcome in those assigned to aspirin. Similar proportions of patients were hospitalized in those assigned to no antithrombotic treatment (48%) and warfarin (47%) but a higher proportion of those assigned to aspirin (64%; p = 0.044). The difference was driven by an increase in hospitalizations for HF (19%, 20%, and 34% of patients were hospitalized for HF respectively; p = 0.032). When the endpoint of both hospitalization for cardiovascular reasons and haemorrhage was considered, then the group randomized to no antithrombotic therapy fared better than either active intervention (22%, 25%, and 40% of patients were hospitalized, respectively; p = 0.019).

Clearly, WASH is not large enough to be definitive but does question the received wisdom that patients with HF should generally be on an antithrombotic agent, especially if they have coronary artery disease. The rate of overt atherothrombotic events was much lower than the rate of death, suggesting that the primary target of ant

ithrombotic therapy in HF is to reduce mortality. It is possible that the dose of aspirin in WASH was too high, but there are no long-term trials of aspirin at lower doses in patients with ventricular dysfunction.

HELAS had a complex design to try and circumvent the issue of withholding antithrombotic agents in patients with coronary artery disease.⁷⁹ It enrolled a cohort of 115 patients with HF and LVSD due to ischaemic heart disease, who were then assigned to aspirin (325 mg/day) or warfarin, and a cohort of 82 patients who were not considered to have coronary artery disease, who were then assigned to placebo or warfarin. Over 22 months, 28 patients (14%) died, but only 2 had a MI (one a patient who was thought not to have coronary artery disease who had been assigned to warfarin) and only 5 had a stroke (4 of these on antithrombotic therapy). There was no difference in outcome depending on assigned group, with patients who had ischaemic heart disease assigned to aspirin doing marginally worse than other groups. Overall, the limited tconclusions that can be drawn from HELAS support those from WASH.

The WATCH trial compared aspirin, at a lower dose than the above trials (162.5 mg/day), to clopidogrel 75 mg/day double-blind and, open-label, to warfarin titrated to an INR of between 2.5 and 3.0.⁹ The study was stopped prematurely because of slow recruitment and showed no difference between aspirin, clopidogrel, and warfarin for the primary endpoint (20.7% on aspirin, 21.6% on clopidogrel, and 19.6% on warfarin) which was a composite of death, MI, or stroke. The reasons for slow recruitment mainly reflected the reluctance of patients with HF to consent to warfarin. However, patients assigned to warfarin had a lower rate of stroke (2.3%, 2.3%, and 0.6% respectively; p = 0.01) and of hospitalization for HF (22.2%, 18.5%, and 16.5% respectively; p = 0.0186), very similar to the WASH trial (Fig. 43.9). Clopidogrel was not quite as good as warfarin and not quite as bad aspirin. Whether lower doses of aspirin would have produced a different result is unclear.

The WATCH study result is unsatisfactory because, although it suggests that warfarin is the best option, warfarin is not sufficiently superior to aspirin to warrant its being the clear preferred option. Indeed, given the neutral primary outcome of WATCH, the preferred option could be considered to be aspirin, since it is not inferior, most convenient and, arguably,⁸⁰ the least expensive option.

SPORTIF was a study in patients with AF, half of whom had HF, comparing warfarin and ximelagatran, a direct thrombin antagonist.⁴⁶ The risks of hospitalization for HF or of death were both strongly related to the presence of markers of HF at baseline. Ximelagatran was associated with a lower risk of the composite outcome of a vascular event, hospitalization for HF, or death (HR = 0.84, CI 0.72–0.98). This was driven by a 27% reduction in hospitalization for worsening HF (p = 0.02) (Fig. 43.10). Patients with HF were also less likely to develop abnormal liver function tests on ximelagatran than patients who did not have HF.

WARCEF is an ongoing study comparing aspirin 325 mg/day with warfarin at a target INR of 2.5–3 in patients with HF regardless of symptom severity, in sinus rhythm and with a left ventricular ejection fraction below 35%.⁸¹ The study is double-blind which, for warfarin, is complex to manage. The study aims to recruit 2860 patients, of whom 70% or more have already been recruited.

The CACHE study intends to compare open-label aspirin 75 mg/day with clopidogrel 75 mg/day in 3000 patients with HF and an NT-proBNP level greater than 400 pg/mL. Most of the patients should be treated with diuretics, ACE inhibitors, and β-blockers. The hypothesis is that aspirin and clopidogrel will both inhibit platelet aggregation but that clopidogrel will not inhibit cyclooxygenase. If cyclooxygenase inhibition leads to adverse effects on renal function, sodium retention, and lung, neuroendocrine, arteriolar, and venous function, then this should translate into a clinical difference between aspirin and clopidogrel. The primary outcome is all-cause mortality with hospitalization for HF, vascular events, and quality of life as key secondary outcomes.

For patients with HF and in sinus rhythm, the weight of evidence suggests that doctors should generally avoid using any antithrombotic agent. If they feel compelled to treat, then there is less evidence of harm with clopidogrel or warfarin than with aspirin, and yet aspirin is most widely used. It is likely to be several years before this therapeutic mess is sorted out. For those who have the opportunity, engaging with a randomized controlled trial is clinically, ethically, and scientifically appropriate.