Chris’s Corner: Why Is It So Hard to Change our Minds?

By Christopher P. Cannon, MD

One of the biggest problems I see in clinical practice today is the underuse of anticoagulants in patients with atrial fibrillation (AF). Multiple registries have shown that only 50% of patients with AF receive anticoagulation. What is so puzzling is that the benefit of anticoagulation in preventing strokes is not subtle—it provides higher than a 65% reduction in risk! We have adopted various antiplatelet therapies demonstrating 15-20% reductions in major adverse CV events in acute coronary syndromes (ACS), yet something three times more effective gets passed over half the time?

The reason appears to be a fear of bleeding. Of course, this is a reasonable concern; we always have to balance benefit and risk. However, in this case, our fears may be too strong. As I understand it, while the original trials of warfarin versus placebo found the huge relative benefit in stroke prevention, the risk of bleeding was a concern based on the absolute risk of bleeding. As such, the notion emerged that treating patients with CHADS2 risk scores ≥2, who had a higher absolute risk reduction of stroke, would have a clear benefit that outweighed the risk. For those with CHADS2 risk score of 1, the absolute risk of bleeding was about 1%, which is similar to the absolute benefit of stroke prevention, so perhaps that is not a good enough risk-benefit ratio. (Note: This equates an ISTH-defined major bleed with a stroke on equal terms, which may not be exactly equal in terms of clinical severity.)

As a result, treating lower-risk patients with aspirin became an accepted plan. However, aspirin has much more modest benefit, in the range of only 20% over placebo. More recently, anticoagulation has been shown to be vastly better than antiplatelet therapy. The ACTIVE W trial was stopped early due to the overwhelming benefit seen with anticoagulation; in the AVERROES trial, one of the new anticoagulants, apixaban, showed a greater than 50% reduction in stroke compared with aspirin in AF patients deemed unable or unwilling to take warfarin.

Given those data, and informed by the relative benefits of the newer anticoagulants over warfarin (even at lower CHADS2 risk scores) the European AF guidelines actually removed aspirin as an option for treatment!

Making Room for New Information in Our Decision-Making Process In addition to formal stroke risk assessment to determine treatment, a more general assessment of bleeding risk—in particular, the risk of falling—has also played a role. This is the most cited reason I see clinically for not wanting to start anticoagulants: the patient is a bit unsteady on their feet and falling carries a risk of subdural hematoma.

This risk-benefit calculation is the subject of one of my all-time favorite papers. In a paper in the Archives of Internal Medicine, Malcolm Man-Son-Hing, MD, and colleagues constructed a Markoff decision analytic model balancing the risks of stroke and bleeding for anticoagulation, aspirin, or no antithrombotic therapy.1 To quote the authors: "Persons taking warfarin must fall about 295 times in 1 year for warfarin to not be the optimal therapy." Kind of an amazing statistic!

Nonetheless, this remains an oft-cited reason for not treating patients on rounds. Indeed, last year I had a 4-day discussion back and forth with a primary care physician about a 92-year-old woman admitted to my cardiology service with newly-developed AF. Her CHADS2 risk score was 3, but the primary care physician resisted anticoagulation based on the patient's advanced age and the risk of falling—even though her daughter, a retired registered nurse, was able to be at her bedside 24/7. Ultimately, the patient and her daughter decided she wanted to receive the anticoagulant.

Maybe Thomas Bayes had it right all along in his prior probability theory: Our prior assessment of a problem is the dominant information used in making our decision. New information is used to move one's thoughts, but rarely totally replaces the older information.

According to the website of the International Society for Bayesian Analysis, the process involves assessing the current state of knowledge regarding the issue of interest, gathering new data to address remaining questions, and then updating and refining understanding to incorporate both new and old data.

Hopefully, the flood of new information from the larger trials involving the newer anticoagulants in AF patients will help bring new light onto this topic, and registries and quality improvement efforts can enhance the use of anticoagulation and improve outcomes for our patients.


1. Man-Son-Hing M, Nichol G, Lau A, Laupacis A. Arch Intern Med. 1999;159:677-85.

Christopher P. Cannon, MD, is a professor of medicine at Harvard Medical School in Boston, Massachusetts. He is also the Editor-in-Chief of CardioSource Science and Quality.

< Back to Listings