A trial finally tested blood thinners in the atrial fibrillation grey zone

For patients with atrial fibrillation and a single stroke risk factor, the guidance has been a shrug in formal clothing. US and European guidelines give oral anticoagulation a class IIa indication in this group — meaning it should be considered — while the trial evidence to support the recommendation has been missing [s1].

SINGLE-AF, published in the New England Journal of Medicine on 28 August and presented at ESC Congress 2026, is that trial. Its companion editorial carries the title "Anticoagulation for Atrial Fibrillation with a Single Risk Factor for Stroke", which is precisely the clinical question at issue [s2].

What the trial did

SINGLE-AF was a multicentre, open-label, adjudicator-masked superiority trial in South Korea [s1]. It enrolled patients with atrial fibrillation at intermediate risk of stroke, defined as a score of 1 in men and 2 in women on the CHA2DS2-VASc scale, which runs from 0 to 9 with higher scores indicating greater stroke risk [s1].

The sex-asymmetric threshold is not an oddity: female sex itself contributes a point to that score, so a woman scoring 2 and a man scoring 1 carry the same non-sex risk factor. Enrolling them together is the correct way to isolate the grey zone.

Patients were randomly assigned 1:1 to direct oral anticoagulant therapy or to no anticoagulation [s1]. The primary end point was a composite of stroke, systemic embolism, major bleeding, or death from cardiovascular causes at 24 months [s1] — a composite that deliberately puts the benefit and the harm of anticoagulation on the same scale.

Of 1,803 patients randomised, 902 received a DOAC and 901 received no anticoagulant therapy [s1]. Mean age was 60.4 years and 23.7% were women [s1].

What happened

At 24 months, a primary end-point event had occurred in 4 patients in the DOAC group (cumulative incidence 0.5%) and 13 in the no-anticoagulant group (cumulative incidence 1.5%) [s1]. The difference was -1.0 percentage points (95% CI -2.0 to -0.1, P = 0.03), with a hazard ratio of 0.31 (95% CI 0.10 to 0.94) [s1].

Stroke occurred in 3 patients in the DOAC group (cumulative incidence 0.3%) and 10 in the no-anticoagulant group (cumulative incidence 1.1%) [s1]. Systemic embolism and major bleeding appeared similar between groups, and no deaths from cardiovascular causes occurred in either group [s1]. Serious adverse events occurred in 80 patients (8.9%) on a DOAC and 84 (9.3%) without [s1].

Read the denominators before the hazard ratio

A hazard ratio of 0.31 is a dramatic-looking number. It rests on 4 primary end-point events in one arm and 13 in the other, across 1,803 patients over two years [s1].

The confidence interval reflects that: 0.10 to 0.94, an upper bound that sits only just below one [s1]. The absolute difference — the number that actually describes what a patient experiences — is 1.0 percentage point over 24 months, with an interval reaching 0.1 [s1].

Both readings are true simultaneously. The trial met its primary end point and did so in the direction guidelines had guessed. And the event count is small enough that the precision of the estimate is low, which is what an interval running from 0.10 to 0.94 is telling you.

The absence of cardiovascular deaths in either arm is a further sign of how low-risk this population was [s1]. That is the nature of the question — the grey zone exists precisely because these patients have few events — but it means a trial in this group is always going to be estimating a small difference from a small number of outcomes.

What it does not resolve

The trial was conducted in South Korea [s1]. Stroke and bleeding risk in atrial fibrillation vary by ancestry and by health system, and intracranial haemorrhage rates in East Asian populations have historically differed from European ones, so the balance struck here may not transfer unchanged.

It was also open-label, with outcome adjudication masked rather than treatment [s1]. For hard endpoints like stroke and death that matters less than for softer ones, but it is a real limitation on a composite that includes major bleeding.

And 24 months is short for a chronic decision. Anticoagulation in atrial fibrillation is typically lifelong; a two-year window catches early events and cannot describe the accumulating bleeding risk of a decade on treatment [s1].

The trial was funded by the Ministry of Health and Welfare, South Korea, and others, and registered as NCT04437654 [s1].

What to watch

Whether guideline committees move the intermediate-risk recommendation from class IIa upward, and whether they do so on a total of 4 and 13 primary end-point events [s1]. The editorial published alongside the trial signals that the field is weighing exactly that [s2]. A second trial in a different population, or a longer follow-up of this one, would do more to settle the question than the effect size alone.

This article describes trial results. It is not advice about any medicine, and nothing here should be used to start, stop or change treatment.

Sources

• Anticoagulation for Atrial Fibrillation with Intermediate Stroke Risk, New England Journal of Medicine, 28 August 2026
• Anticoagulation for Atrial Fibrillation with a Single Risk Factor for Stroke, New England Journal of Medicine, 28 August 2026