- Design
- Multicentre, open-label, adjudicator-masked randomised superiority trial in South Korea (SINGLE-AF, NCT04437654)
- Population
- 1,803 patients with atrial fibrillation at intermediate stroke risk (CHA2DS2-VASc 1 in men, 2 in women); mean age 60.4 years, 23.7% women
- Primary outcome
- Composite of stroke, systemic embolism, major bleeding or cardiovascular death at 24 months
- Effect
- 0.5% (4 patients) vs 1.5% (13); difference -1.0 percentage points (95% CI -2.0 to -0.1), HR 0.31 (0.10-0.94), p=0.03
Anticoagulating atrial fibrillation at a CHA2DS2-VASc score of 1 in men or 2 in women is a class IIa recommendation resting on observational data. SINGLE-AF randomised 1,803 patients in South Korea to a direct oral anticoagulant or to nothing, open-label with masked adjudication, and followed them for 24 months.
The composite of stroke, systemic embolism, major bleeding or cardiovascular death occurred in 4 patients on a DOAC (0.5%) and 13 without (1.5%): difference -1.0 percentage points (95% CI -2.0 to -0.1), hazard ratio 0.31 (0.10-0.94), p=0.03. Stroke accounted for most of it, 3 versus 10. Major bleeding and systemic embolism looked similar between groups, no cardiovascular deaths occurred at all, and serious adverse events were 8.9% against 9.3%.
The result is positive and the evidence is thin, and both statements are true at once. Seventeen primary events in 1,803 patients is a very small number to carry a hazard ratio, the confidence interval reaches 0.94, and the trial was open-label in a single country with a mean age of 60.4 - younger than the patients most of us are deciding about. The absence of a bleeding difference is reassuring but was never going to be detectable at this event rate.
What it changes is the tone of the conversation rather than the threshold. A patient with one non-sex risk factor who asks whether anticoagulation is worth it can now be told there is randomised evidence pointing to yes, with an absolute benefit of about one event per hundred patients over two years, and with bleeding risk that looked no worse in the trial. For older Indian patients with hypertension as their single risk factor, that is a real conversation, but the trial population was not theirs, and their bleeding risk is not the trial's either.
- Apply this to patients resembling the trial: intermediate risk, younger, without high bleeding risk
- Quote the absolute figures - about one fewer event per hundred patients over two years
- Do not read 'no bleeding difference' as 'no bleeding risk'; the trial was too small to show one
- Re-score CHA2DS2-VASc at every review - most patients cross the threshold with time anyway
- In older or frail patients, weigh a bleeding assessment the trial's cohort did not test
The statistics, in plain English
A hazard ratio of 0.31 sounds dramatic but rests on 4 events against 13; with numbers that small, a handful of events either way would move it substantially, and the confidence interval reaching 0.94 says the data are only just incompatible with no effect. The absolute difference, 1.0 percentage point over two years, is the honest figure for a patient. An open-label design cannot bias adjudicated stroke much, but it can bias how hard clinicians look for events, and bleeding in particular tends to be reported differently when everyone knows who is on the drug.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for neurology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free