Guidelines for antithrombotic therapy in atrial fibrillation: understanding the reasons for non-adherence and moving forwards with simplifying risk stratification for stroke and bleeding

This editorial refers to ‘The likelihood of decreasing strokes in atrial fibrillation patients by strict application of guidelines' by R. Pisters et al ., on page 779.

Randomized trials in patients with atrial fibrillation (AF) have consistently demonstrated that compared with placebo, adjusted-dose vitamin K antagonists (VKAs) reduce the incidence of stroke by ∼65%, whereas antiplatelet agents reduce stroke by ∼20%. ¹ Despite existing evidence and guidelines, several observational studies have demonstrated suboptimal use of oral anticoagulation in clinical practice. For example, the EuroHeart Survey in AF reported that although the application rate of VKAs had generally improved, antithrombotic drug therapy was only modestly tailored to guidelines. ² Appreciation that adherence to guidelines is suboptimal is one thing; understanding the reasons for this situation and its consequences and finally attempting at better adherence to guidelines are a different issue.

In this issue of the Journal, Pisters et al . ³ report that many patients with known AF admitted for ischaemic stroke to a university medical centre did not receive adequate antithrombotic treatment on admission and 51% patients with known AF who were eligible for VKA did not actually receive it. Indeed, an improved antithrombotic guideline adherence could have potentially prevented 22% of the ischaemic strokes. Moreover, at least 10% of patients at discharge were still insufficiently protected against recurrent stroke.

These observations are even more alarming than those from the EuroHeart Survey on AF in whom participating centres chose to include or not their patients on an individual basis, ² thus resulting in a possible bias. Nevertheless, these findings represent a systematic analysis of a ‘real world’ cohort of patients. Additional analyses of the factors leading to undertreatment would have been of interest, because the parameters associated with undertreatment may be broadly classified in two categories: those related to the patient and those (unfortunately) related to the physician.

Older age and renal failure are recognized risk factors for bleeding and are commonly associated with undertreatment. ^{4 , 5} Patients whose only risk factor is their greater age are less often treated with anticoagulants and this reflects the continued reluctance to treat elderly patients with VKAs. The pattern of AF (permanent or non-permanent) and/or time spent in AF (or ‘AF burden’) should not influence the decision to prescribe an anticoagulant, in the presence of risk factors. However, it is not an uncommon finding that the prescription rates of VKAs are lower rate in paroxysmal compared with permanent AF, or in persistent AF once sinus rhythm has been achieved. Indeed, patients with permanent AF are more often treated with VKAs than those with paroxysmal AF, ^{4 , 5} perhaps as a result of the erroneous belief that paroxysmal AF would lead to fewer embolic events.

A take-home message from the paper by Pisters et al . ³ is that almost 25% of strokes might have been prevented by guideline adherence. This finding is not the result of a direct comparison of two medical managements as in usual clinical research, but an evaluation, based on what is known from medical literature, of what would have been the benefit with an appropriate management of the AF patients in whom the guidelines were not applied, while there was no reason not to apply them. Such a method remains somewhat speculative but should probably be used more often. The calculation of the benefit—a 65% reduction in risk of stroke if VKA was used instead of no antithrombotic agent and a 40% reduction in risk if VKA was used instead of an antiplatelet agent—might not be accurate for all subgroups of patients or on an individual basis. However, there is no reason to think it is an inappropriate viewpoint since the benefit of VKAs in terms of relative risk reduction is similar in the different groups of patients at low, medium, and high risk of thromboembolic events. ⁶ It is possible that the benefit obtained with a better guideline adherence was rather underestimated in the work by Pisters et al ., ³ since at least 20% of the reasons not to prescribe VKA could be considered invalid.

Interestingly, the AF patients who were classified as low risk of thromboembolism accounted for a substantial part of the whole population with stroke. This observation does not support a less vigorous antithrombotic regimen in these seemingly ‘low risk’ patients because their risk of ischaemic stroke is, in fact, high and aspirin does not provide a sufficient protection on a population basis. In patients with AF and a CHADS ₂ score of 1, a lower incidence of stroke and/or death from all causes has been found among patients treated with VKAs when compared with other patients. ^{5 , 6} In contrast, prescription of an antiplatelet agent was not associated with a lower risk of events compared with no antithrombotic therapy. ⁵ In most cases, an antiplatelet strategy runs a similar risk of major bleeding as oral anticoagulant therapy, and it is also associated with a much lower benefit in reducing death and stroke. Such ‘not so low risk’ patients should be treated with oral anticoagulant whenever possible unless there is a high risk of haemorrhagic event.

In patients with the CHADS ₂ score ≥2, anticoagulant should then systematically be prescribed. Even so, most published stroke risk stratification systems are of modest value in predicting patients at high risk of thromboembolism. A paradigm shift is needed in the primary prevention of stroke in AF, whereby identification of the ‘low risk’ category is improved so these patients should be truly low risk . This has been achieved with the CHA ₂ DS ₂ -VASc, [cardiac failure, hypertension, age ≥ 75 (doubled), diabetes, stroke (doubled)—vascular disease, age 65–74, and sex category (female)] schema, which is based on a point system in which two points are assigned for a history of stroke or TIA, or age ≥75 and one point each is assigned for age 65–74 years, a history of hypertension, diabetes, recent cardiac failure, vascular disease [myocardial infarction, complex aortic plaque, and peripheral arterial disease (PAD), including prior revascularization, amputation due to PAD or angiographic evidence of PAD, etc], and female gender ⁷ ( Table  1 ). This schema improves on the CHADS ₂ schema, classifies a low proportion of subjects into the ‘moderate risk’ category, and helps better determine the truly 'low risk' patients who have very low event rates and no need for anticoagulation.

Nonetheless, most physicians seem more concerned with a greater risk of bleeding, whereas the patients' major concern is thromboembolic stroke. ⁸ Patients who are at high risk of stroke due to AF (i.e. those with a previous thromboembolic event) accept anticoagulation, despite its increased risk of bleeding, to avoid stroke. What is more concerning, is that higher thrombotic stroke risk, defined by CHADS ₂ score, is also coupled with significant bleeding risk.

Until recently, bleeding risk has been less well stratified due to the lack of available validated and user-friendly bleeding risk stratification systems. Indeed, a simple bleeding risk predictor schema would assist physicians in making decisions about commencing anticoagulation therapy. There are currently four published bleeding risk score systems (only two of which have been validated in AF populations) but their major drawback is that they have dissimilar characteristics and (sometimes) complex scoring systems, which hamper their usage in daily clinical practice. ⁹

Recently, a new simple bleeding risk score (acronym HAS-BLED: Hypertension, Abnormal renal/liver function, Stroke, Bleeding history or predisposition, Labile INR, Elderly (>65), Drugs/alcohol concomitantly) has been derived and validated in a European cohort ¹⁰ ( Table  1 ). Application of this new simple bleeding risk score (HAS-BLED) gave similar c-statistics to other published schema in the overall cohort for major bleeding (0.72) but in some patient subgroups, who received antiplatelet agents alone or no antithrombotic therapy, there was even superior predictive ability for major bleeding (with c-statistics of 0.91 and 0.85, respectively). A HAS-BLED score of ≥3 indicates ‘high risk’ of bleeding and some caution (rather than contraindication) is needed, following initiation of oral anticoagulation therapy or aspirin, given the same rates of major bleeding with the latter, especially in the elderly.

Clearly, the relationship between stroke and bleeding risks factors and thromboembolic events and major bleeding now have to be examined prospectively in AF cohort. This will help to establish the best bleeding risk index and finally a unifying consensus on how to treat AF patients, particularly those with a relatively low risk of thromboembolic event.

Given the availability of innovative oral anticoagulant drugs which may have a better stroke prevention effect than VKAs and a lower risk of bleeding, ¹¹ simplification of the approach to stroke prevention is overdue. With the CHA ₂ DS ₂ -VASc score, those ≥2 are clearly high risk and should receive oral anticoagulation (whether with a VKA or new agents, such as dabigatran), whereas those with a CHA ₂ DS ₂ -VASc score of 1 are at moderate risk and while ‘oral anticoagulation or aspirin’ is recommended, oral anticoagulation is probably preferred. ⁵ Finally, those with a CHA ₂ DS ₂ -VASc score of 0 are so low risk that no antithrombotic therapy is a possible approach, given the lack of evidence for the effectiveness (and possible harm, from bleeding) with aspirin in these patients. ¹² This paradigm shift leads to a ‘yes/no’ approach for anticoagulation based on the CHA ₂ DS ₂ -VASc score, whereby a CHA ₂ DS ₂ -VASc score ≥1 means ‘prescribe oral anticoagulation’, whereas those at low risk (score = 0) may not need antithrombotic therapy. Such a simplification would improve guideline adherence and our approach to thromboprophylaxis in AF patients.

Conflict of interest: none declared.

References

Author notes