John Mandrola, MD


November 10, 2015

I recently saw four patients in one morning who had atrial fibrillation and either one or two risk factors for stroke. None had symptoms warranting rhythm control or tachycardia requiring rate-control drugs. The only question was anticoagulation. "Doctor, should I take the drug? For the rest of my life? Really?"

The decision to use anticoagulants in patients with AF is a gamble. The drugs lower stroke risk but increase bleeding risk. If the risk of stroke is high, the benefit of stroke reduction exceeds the risk of bleeding. In these cases, we say the net clinical benefit is favorable.

The tricky part is that not all AF patients have the same stroke risk. Everyone who takes an anticoagulant incurs an increased risk of bleeding, so those patients at low risk for stroke may get no net clinical benefit from the drug. The number we must know in this gamble is the risk of stroke without anticoagulation.

Most experts use a threshold of 1% to 2% yearly risk of stroke as the threshold for net benefit of anticoagulation. Both the European and North American guidelines have adopted the CHADS-VASc score as a way to estimate stroke risk.

In the decision to recommend anticoagulation, the guideline writers assume CHADS-VASc scores correlate to fixed stroke rates. And these stroke rates come from Danish national registries.

But what if we looked at different cohorts? Would the rates of stroke be the same?

That is exactly what Drs Gene Quinn and Daniel Singer (Harvard Medical School, Boston, MA) set out to study. In an oral abstract presentation[1] here at the American Heart Association (AHA) 2015 Scientific Sessions, they described and contrasted reported rates of stroke in worldwide cohorts of patients with AF, specifically looking at patients who were not using anticoagulants.

They did a systematic review of all cohort studies and randomized clinical trials that included patients with AF who were not taking anticoagulants. They found 37 eligible cohorts with 563,872 patients from all parts of the world. The cohorts ranged in size from 186 to 186,570 patients.

In a series of slides, Dr Quinn listed the 37 cohort studies. One was immediately struck by the variation in rates of ischemic events across cohorts. In studies that included more than 5000 patients, Dr Quinn noted that stroke rates ranged from 0.45% to 1.97% in North American cohorts to 6.22% in Danish registries. When weighted by numbers of subjects, the average stroke rates for cohorts were North America, 1.86%; Europe, 4.39%; Asia, 3.88%; Middle East, 3.00%.

Dr Quinn then considered 17 of the 37 original studies because they reported stroke rates based on the CHADS-VASc score. The variation in stroke rates among patients with CHADS-VASc scores of 0 to 2 persisted. For instance, in CHADS-VASc 2 patients, baseline stroke rates ranged from 0.48 in the Women's Health Initiative study to 3.71 in the Danish National Patient Registry.

In the discussion section of the oral abstract, they considered the reasons for these large variations. One reason may be that the stroke rates really do vary between regions. But that's hard to explain, given that stroke rates from Denmark are two to three times higher than (phenotypically) similar patients in Sweden.

Dr Quinn felt the more likely reason for variation was due to different methods of data extraction. He cited the Friberg et al paper (published in the Journal of the American College of Cardiology[2]), in which the use of different quarantine periods and differing definitions of ischemic events produced varied stroke rates.

In the question-and-answer period, Dr Quinn showed that there was even variation within the same cohort. He cited two papers from the same Taiwanese database. The stroke rate for CHADS-VASc 2 patients was 0.91 in a 2011 paper[3] and 3.39 in a paper from 2015[4].


Dr Singer reminded me that these observations apply to the smaller fraction of patients with AF who have few risk factors for stroke. He estimated that for more than 80% of AF patients (those with CHADS-VASc scores of >3), the net clinical benefit favors anticoagulants.

That said, AF is a common problem, one that is on the rise. And, as we quantify more and more of the human condition, we will identify increasing numbers of patients like the four I saw on one clinic day.

For these millions of lower-risk patients with AF, there is already significant uncertainty surrounding the decision to use anticoagulants. This presentation adds even more uncertainty.

Like most important scientific observations do, these findings suggest we overestimate what we know. They raise more questions than they answer. Dr Singer likened this gap in knowledge to a large blind spot in the guidelines.

I'd say it is. Think slowly about this for a moment: If we don't know the odds of having a stroke without anticoagulants, we can't possibly know the net clinical benefit. And if we don't know that, we know very little.

That's not good, because when we put an asymptomatic person on a potent drug with the promise of future benefits, we approach the do-no-harm rule.

Perhaps more sobering, as the young Dr Quinn reminded me, is the burden of giving another human being a serious medical condition.

Drs Quinn and Singer made a plea for more uniform ways to assess stroke risk across cohorts. At least then, they argue, we can start the decision from a position of consistency.

That seems a reasonable request, one that we should grant sooner rather than later.



Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.