When to anticoagulate up for debate with stroke risk scores but no defined thresholds

The evidence available from trials is not directly relevant to the discussion of whether or not to anticoagulate patients with AF because those involved generally have more stroke risk factors than current guidelines recommend.

The CHA₂DS₂ stroke risk score — used in the American College of Chest Physicians and Canadian guidelines — that recommends a patient with at least 1 point be anticoagulated is a very good rule in terms of those who have a decent risk. However, many patients who by that recommendation are not anticoagulated, have a greater risk of thrombosis than the patients involved in trials where there was an indication of a benefit of antithrombotic treatment.

The CHA₂DS₂-VASc scale — used in European and National Institute of Clinical Excellence guidelines — comes in with additional risk factors, thereby more patients would be recommended for anticoagulation. The advantage of the CHA₂DS₂-VASc scale is that patients who truly have a score of 0 are recognized as being very low risk, and it is very plausible to recommend no anticoagulation whatsoever — no proper treatment and no aspirin.

When considering which of the scores is better for selecting patients for anticoagulation, it is important to emphasize there has been only indirect evidence from epidemiological studies with a focus not on whether anticoagulation is good, but the risk when a patient is not anticoagulated. We added substantially to that debate by drawing curves based on all the epidemiological studies to determine the risk at various scores.   

In some studies that suggest CHA₂DS₂-VASc is too sensitive, patients had an extremely low risk; in the ATRIA Study, patients seen in an outpatient clinic could have had AF for quite some time, and therefore the risk may not reflect a true clinical situation.

Conversely, in some Scandinavian studies involving populations admitted to a hospital for AF and then followed, patients had a higher risk — not very high but higher — even with 1 point on the CHA₂DS₂-VASc scale.

Based on this variability, it is not enough to look at CHA₂DS₂ or CHA₂DS₂-VASc. Prescribing clinicians must consider other circumstances. Is this a chronic patient who has had AF for a long time, come into the clinic frequently and more or less proven they do not get strokes from it? Was the patient newly diagnosed with AF during a hospital admission, and a higher risk is already known?

Stroke prevention with oral anticoagulant (OAC) treatment is an essential part to the management of atrial fibrillation. As the risk of stroke in AF patients is not homogeneous, different stroke risk stratification schemes (or risk scores) have emerged.

The CHA₂DS₂-VASc score is endorsed by the European Society of Cardiology and the American Heart Association/American College of Cardiology/Health Services Research guidelines, but the two guidelines do not suggest the same threshold for recommending OAC treatment.

The reasoning behind setting a threshold (e.g. stroke risk of 1% per year) emerges from different observational studies involving different study populations, from cohorts obtained from either randomized controlled trials or cohorts based on administrative databases.

Different study populations and different methodological approaches will inevitably entail different outcomes in terms of stroke rates. For example, the ATRIA cohort could be affected by selection bias by only including those patients with health care plans. Similarly, the Danish nationwide cohort could be affected by selection bias by only investigating AF patients who are referred to a hospital (not investigating general practitioner-diagnosed AF). More recently, a study published in the Journal of American College of Cardiology based on the Swedish nationwide cohort argued against recommendation of OAC treatment in patients with a CHA₂DS₂-VASc score of 1. However, the study design may well have imposed a bias that deflated the event rates by excluding patients if they initiated OAC treatment during follow-up, as noted by the Journal of American College of Cardiology editors.

An important caveat with these studies is the outcome of thromboembolism is affected by a competing risk of death. Hence, event rates are accordingly reported, but these cannot be regarded as absolute risks, which is really what is wanted for a treatment decision — to recommend anticoagulant treatment or not.

A loose interpretation on rates and risk translates into two different questions: for risks, “What is the probability that my patient will have a stroke during the next X years?” and for rates, “On average, if my patient is still alive at some unspecified time point during the next X years, what would the risk be of having a stroke the next day?” To comply with this catch, an arbitrary measure of net clinical benefit is introduced.

The net clinical benefit is a weighted sum of rate differences for thromboembolisms and intracranial haemorrhages. The net clinical benefit, balancing ischemic stroke reduction against serious bleeding, is positive for patients with 1 or more additional stroke risk factors.

Unquestionably, some treatment decisions need to be individualized, but prescribing clinicians really need to ask whether is it really worth taking the risk by not anticoagulating those AF patients with one or more additional stroke risk factors, that is, a CHA₂DS₂-VASc score of 1 (males) and 2 (females). 

References:

Singer DE, et al. J Am Heart Assoc. 2013; doi:10.1161/JAHA.113.000250.

Lip GYH, et al. Chest. 2014;doi:10.1378/chest.14-0533.

Lip GYH, et al. J Am Coll Cardiol. 2015;doi:10.1016/j.jacc.2015.01.044.

Neilsen PB and Chao TF. Thromb Haemost. 2015;doi:10.1160/TH15-03-0210.

Olesen JB and Torp-Pedersen C. Thromb Haemost. 2015;doi:10.1160/TH15-02-0154.