AF Duration? Another Big Story from 2023

In 2023, two large randomized trials addressed the use of oral anticoagulation in short duration atrial fibrillation. This was a really big story. We learned a lot.

I wrote this on my TheHeart.org | Medscape Cardiology recap piece:

Changing Ideas About Atrial Fibrillation

We learned this year that not all atrial fibrillation (AF) is the same.

For an older patient with stroke risk factors and 3 hours of AF on a cardiac device, pre-2023 thinking would have us leaning toward anticoagulation. The results of the ARTESIA and NOAH-AFNET 6 trials strongly question that idea.

NOAH-AFNET randomly assigned about 2500 older patients with short-duration device-detected AF (median duration, 2.8 hours) to edoxaban vs placebo. The trial was stopped early at 21 months for perceived futility and excess bleeding in the edoxaban arm. Edoxaban reduced the primary endpoint of stroke, systemic embolism, and cardiovascular death by 19% (HR, 0.81; 95% CI, 0.60-1.08; P = .15) but increased bleeding by 31% (HR, 1.31; 95% CI, 1.02-1.67; P = .03).

ARTESIA randomly assigned about 4000 older patients with short-duration AF (median duration, 1.5 hours) to apixaban vs aspirin. More patients and longer follow-up of 3.5 years translated to more events. Apixaban reduced the primary endpoint of stroke or systemic embolism by 37% (HR, 0.63; 95% CI, 0.45- 0.88; P = .007). This was countered by an 80% increase in major bleeding (HR, 1.80; 95% CI, 1.26- 2.57; P = .001).

The results of both trials were consistent: Direct-acting oral anticoagulants reduce thrombotic events but increase bleeding rates. The biggest discovery — and reason for debate about net benefit — was the low stroke rates (< 1% per patient-year) in both trials. What clinicians need, and is probably forthcoming, is an analysis correlating duration of AF and net benefit. Right now, anticoagulation decisions with device-detected short-duration AF require high doses of judgement and patient preference.

Translation and Further Comments

In days of old, before smart watches and high-tech pacemakers, a patient diagnosed with AF had to have the arrhythmia long enough to have symptoms and seek medical attention. The doctor would do an ECG and that would diagnose AF.

We now call that “clinical AF.” Patients like these were enrolled in RCTs of anticoagulation, which then established the net benefit of oral anticoagulation over nothing or aspirin.

The original trials tested warfarin vs placebo or aspirin. Then the direct acting oral anticoagulants (DOACs) came along and were found in 4 big RCTs to be either similar to or better than warfarin. DOACs are surely more convenient than warfarin.

Modern technology has upset this treatment approach. Devices can now detect even minutes to a few hours of AF. These shorter arrhythmias usually create no symptoms. But it is real AF. You can see the atrial signal fibrillating.

NOAH and ARTESIA enrolled these patients and tried oral anticoagulation vs either placebo (NOAH) or aspirin (ARTESIA). A key factor with these patients was the median duration of AF was very short—1.5-2.8 hours.

A meta-analysis that combined the two trials found consistent results. Yes, oral anticoagulation reduced the risk of future clots, by about 32%. But the drugs also increased major bleeding by 62%.

Some experts have argued that strokes are worse than bleeds because strokes can lead to permanent disability while bleeds rarely disable anyone.

I don’t know. That can be a hard sell. Because the absolute stroke risk reduction in ARTESIA (the trial with the most strokes) was only 1.5%. The number needed to harm by increased bleeding was similar at 1.3%

Two Take-home Lessons

SCAF confers a low stroke risk: So-called short-duration or device-detected AF or subclinical AF confers a very low yearly stroke risk. About 1% in both trials. This, despite the older age and multiple stroke risk factors in these patients.

Oral anticoagulation will lower stroke risk, but if it starts out really low, the amount of risk reduction is countered by elevated bleeding risk.

If baseline stroke rate is confusing, consider a patient with a yearly stroke risk 10x higher at 10%. Now apply that same 30% risk reduction from oral anticoagulants to that patient. It now goes from 10% to 7%. That’s a big drop in absolute risk.

NOAH and ARTESIA have redefined AF. The trials have shown us that old-school AF, wherein it is detected due to symptoms, in a doctor’s office, and by a standard ECG, is different from short duration AF detected on a device. Different meaning lower stroke risk.

This has big implications for screening for AF. Two thoughts come to mind.

One is the common practice of putting in implantable loop recorders in patients with stroke of unknown cause. ILRs, as they are called, are costly. They pick up any duration of AF. The idea is that if a stroke patient has detected AF, then that is the likely cause, and we should use oral anticoagulation. But these two trials have shown that very short duration AF confers a very low stroke risk. And we may not be helping post-stroke patients by detecting one hour of AF.

These trials also dampen excitement of using smart watches to detect AF. Consider that these were patients sick enough to have pacemakers or defibrillators. The mean age was mid 70s. Their mean CHADSVASC score was 4. So, even in this high-risk group, SCAF only led to a yearly stroke rate of 1%.

Contrast that with the average smart-watch wearing 55-year-old who detects some AF. That person will have an even lower stroke risk, and, I think, would be at-risk of an over-zealous doctor doing things that could lead to more cost and potential harm.

What we need now?

The authors of these trials are likely going to combine raw data. This will allow them to construct a plot of duration of AF and stroke risk and net benefit. Consider that in ARTESIA, more than 40% of patients had less than an hour of AF.

We may soon learn that AF that lasts over a certain number of hours will associate with a high enough stroke rate to gain a net benefit from anticoagulation. For instance, 10 hours may be enough, but 2 hours is not. That’s just a guess.

For now, today, when I go to clinic and get a message that a 75-year-old patient with a pacemaker had 4 hours of AF over the holidays, I will have to rely on judgement and patient preferences. The two trials taught us a lot but did not provide an algorithm that a robot could use. JMM

---

P.S. This newsletter is often critical of industry-related research. Both these trials were at least partially supported by the drug makers. In this case, the industry incentives (to hopefully gain a larger market share) aligned well with patients’ incentives. These were well-conducted trials that taught us a lot. They helped re-define one of the most common conditions in all of medicine—AF.

My message should never be that industry-funded research is nefarious. Rather, the onus is on us to be good appraisers of the evidence.