Should patients with atrial fibrillation and intermediate stroke risk receive anticoagulation?

Introduction

Atrial fibrillation(AF) increases the risk of stroke. We normally assess patients based on their CHA2DS2-VASc Score(0-9), with the higher the score indicating a greater the risk of stroke. For patients with a higher risk ie males >2 and females >3, oral anticogulation is strongly recommended. However for those patients with intermediate risk ie., a score of 1 in males and 2 in females, although guidelines recommend anticoagulation, there is conflicting evidence for this limited randomised trials.

THE STUDY
Kim D, Lee YS, Shim J, et al. Anticoagulation for Atrial Fibrillation with Intermediate Stroke Risk. N Engl J Med. Published August 28, 2026. DOI: 10.1056/NEJMoa2607978

Bottom Line

The SINGLE-AF trial provides the first direct randomised evidence that DOAC therapy may reduce thromboembolic events in patients with clinical AF and a single non-sex CHA₂DS₂-VASc risk factor.

Over two years, the primary composite outcome occurred in:

• 0.5% with a DOAC

• 1.5% without anticoagulation

• Absolute risk reduction: 1.0%

• NNT: approximately 100 over two years

• HR 0.31; 95% CI 0.10–0.94

The apparent benefit was driven predominantly by fewer ischaemic strokes. However, there were only 17 primary events, making the result uncertain and the magnitude of benefit unclear.

This supports offering anticoagulation to appropriately selected, low-bleeding-risk patients—but does not eliminate the need for shared decision-making.

What They Did

SINGLE-AF was an investigator-initiated, Multicentre, Open-label, Randomised, Superiority trial with Blinded endpoint adjudication, conducted at 18 centres in South Korea.

The study population was >19yo with AF and an intermediate risk score.

The trial excluded patients with:

• • Moderate-to-severe mitral stenosis

  • Mechanical prosthetic valves

  • Creatinine clearance below 30 mL/min

  • Significant liver disease

  • Clinically significant cardiomyopathy or amyloidosis

  • Another indication for anticoagulation

  These results therefore apply to relatively healthy patients with conventional clinical AF and low bleeding risk—not to all patients with AF.

A total of 1,803 patients were randomised 1:1 to:

• DOAC therapy the coice of which was left to the physician. (Apixaban 5mg po bd, or Rivaroxaban 20 mg daily left to physicians' choice), or

• No routine anticoagulation. However patients could have a temporary administration of a DOAC before and after rhythm control with cardioversion or ablation.

The intervention was:

• Apixaban 5 mg twice daily: 67.4%

• Rivaroxaban 20 mg daily: 14.1%

• Rivaroxaban 15 mg daily: 13.9%

Overall, 81.6% received a full DOAC dose.

Follow-up was 24 months, with 94.6% completing follow-up in each group.

Patient population

The cohort was relatively young and at low bleeding risk:

Primary outcome

The primary endpoint was a composite of:

• Stroke

• Systemic embolism

• Major bleeding

• Cardiovascular death

What They Found

At 24 months post randomisation a primary end-point event iccurred in 4 of 902 patients in the DOAC group and 13 of 901 patients in the no-anticoagulation group.

Treating approximately 100 patients for two years would prevent one primary composite event.

That estimate is imprecise because it is based on only:

• 4 events in the DOAC group

• 13 events in the control group

Individual outcomes

The reduction in the primary endpoint was principally explained by fewer ischaemic events:

• Ischaemic stroke or systemic embolism: 0.1% versus 1.1%

• HR 0.10; 95% CI 0.01–0.77

• Absolute difference: 1.0%

• Approximate NNT: 100 over two years

However, this was a post hoc exploratory endpoint and should not be interpreted as definitive.

Bleeding

Major bleeding was uncommon and appeared similar:

• DOAC: 3 patients

• No anticoagulation: 4 patients

There was nevertheless a numerical increase in less severe clinically relevant bleeding:

• DOAC: 2.5%

• No anticoagulation: 1.6%

There were also:

• Two haemorrhagic strokes in the DOAC group

• No haemorrhagic strokes in the control group

The numbers are far too small to draw reliable conclusions, but they remind us that anticoagulation is not risk-free, even in a carefully selected low-bleeding-risk population.

Strengths and Limitations

Strengths

• Randomised comparison of DOAC therapy against no anticoagulation

• Addresses a common and unresolved clinical decision

• Multicentre recruitment

• Excellent adherence in the DOAC group

• Blinded independent adjudication of outcomes

• Good follow-up completion

• Contemporary DOAC therapy, predominantly apixaban

• Sensitivity analyses supported the primary result

Limitations

1. Few events

Only 17 primary events occurred. The confidence interval around the hazard ratio was wide—0.10 to 0.94—and approached no effect.

The true benefit is probably more modest than the headline relative risk reduction of 69%.

2. Lower event rate than anticipated

The trial anticipated primary event rates of:

• 2.0% with DOAC therapy

• 4.5% without anticoagulation

The actual rates were only 0.5% and 1.5%. This substantially reduced the precision of the study.

3. The primary endpoint was changed during the trial

The original composite included all-cause death. In January 2025, it was changed to cardiovascular death after the data-monitoring board observed that deaths were non-cardiovascular.

The board remained blinded to treatment allocation, however this remains methodologically important.

4. Open-label treatment

Patients and clinicians knew the treatment allocation. Although events were independently adjudicated, knowledge of treatment could have affected investigations, reporting, co-interventions or admission decisions.

5. Antiplatelet imbalance

Antiplatelet therapy was used in:

• 0.4% of the DOAC group

• 36.5% of the no-anticoagulation group

This could increase bleeding in the control group and make the net-safety comparison more favourable to DOAC therapy.

Importantly, antiplatelets are not an effective substitute for anticoagulation for AF-related stroke prevention.

6. Limited generalisability

All participants were recruited in South Korea. The cohort also contained relatively few:

• Women

• Patients with non-paroxysmal AF. Only 30% of patients had nonparoxysmal AF. This suggests that patients with this phenotype have a higher burden of disease than patients without recurrent AF after catheter ablation

• Patients with higher bleeding risk

• Patients with significant renal, hepatic or structural heart disease

7. Crossovers and changing risk

During follow-up:

• 5.5% of the no-anticoagulation group received a DOAC

• Approximately 15% developed a higher CHA₂DS₂-VASc score

These changes would generally dilute the apparent treatment effect, although sensitivity analyses remained consistent.

8. Composite endpoint limitations

The primary outcome combines benefit outcomes—stroke and systemic embolism—with harms such as major bleeding.

This is a reasonable “net clinical benefit” strategy, but the components are not equivalent in severity. A disabling ischaemic stroke, gastrointestinal bleed and non-fatal minor intracranial event cannot be assumed to carry identical importance.

Conclusion

This trial shifts the evidence in favour of anticoagulation for clinical AF with one additional stroke risk factor.

But it is not a mandate to anticoagulate every patient with a CHA₂DS₂-VASc score of 1 in men or 2 in women.

The absolute untreated risk was low:

• Any stroke: approximately 0.55% per year

• Ischaemic stroke: approximately 0.55% per year

Consequently, the absolute benefit was also modest. Whether that benefit is worthwhile depends on:

• Which risk factor generated the score

• AF burden and persistence

• Bleeding risk

• Renal function

• Frailty and falls

• Concurrent antiplatelet therapy

• Patient values and preferences

The trial provides stronger evidence for a conversation about anticoagulation, rather than an automatic prescription.

"Given the lower-than-expected number of end-point events, the precision of the estimated treatment effect is limited, and these findings should be interpreted with caution."

Relevance to us in emergency medicine

This study does not directly answer whether a patient with recent-onset AF can be safely cardioverted without anticoagulation. It may push us to commence long term anticoagulation in patients with intermediate risk. Anecdotaly many do.

It addresses long-term stroke prevention, not pericardioversion anticoagulation or the immediate safety of cardioversion. Pericardioversion anticoagulation was permitted in this trial.

For the emergency clinician:

• Identify patients with intermediate stroke risk rather than labelling them “low risk.”

• Do not use aspirin as an alternative to anticoagulation for AF-related stroke prevention.

• Discuss or arrange consideration of a DOAC, particularly when bleeding risk is low.

• Reassess stroke risk over time—patients frequently move into a higher-risk category.

• Continue to follow guideline-specific anticoagulation requirements surrounding cardioversion.

Take-home points

1. DOAC therapy reduced the primary net clinical composite from 1.5% to 0.5% over two years.

2. The absolute benefit was modest: approximately one event prevented for every 100 patients treated for two years.

3. The apparent benefit was driven by fewer ischaemic strokes and systemic emboli.

4. Major bleeding was not increased, but clinically relevant non-major bleeding was numerically more common and two haemorrhagic strokes occurred with DOAC therapy.

5. Only 17 primary events occurred, so the impressive relative effect is imprecise and potentially exaggerated.

6. This trial supports offering anticoagulation—not automatically prescribing it—to patients with clinical AF and one non-sex stroke risk factor.

7. The results do not apply to moderate-to-severe mitral stenosis, mechanical valves, severe renal disease or immediate cardioversion decisions.

Verdict

SINGLE-AF strengthens the case for DOAC therapy in intermediate-risk AF, but the low absolute event rate and very small number of outcomes mean that anticoagulation should remain an individualised, shared decision rather than an automatic response to the score alone.

You may also be interested in the Garfield score