How AF episodes cluster may matter as much as how much AF there is

Design

Observational, two device-monitored US cohorts, g-formula modelling and meta-analysis

Population

12,868 patients with implantable devices; mean age 72, median CHA2DS2-VASc 4.0

Primary outcome

One-year ischaemic stroke risk by AF density

Effect

High vs low density risk ratio 1.75 (95% CI 1.25-2.44), independent of burden

Using two US cohorts of patients with cardiac implantable devices (12,868 of 41,780 screened), investigators measured atrial fibrillation density, a 0-to-1 index of how tightly episodes cluster in time, separately from total burden, and related it to ischaemic stroke.

Over a median four years, 336 strokes occurred (6.3 per 1,000 person-years). High density showed a dose-response relationship with one-year stroke (risk ratio 1.75, 95% CI 1.25 to 2.44), consistent across device types, comorbidity, age and anticoagulation status. At any given burden, higher density carried greater risk.

This is prognostic modelling, not a trial, and density is not yet a routine metric. But it suggests that two patients with the same AF burden can carry different stroke risk depending on whether the fibrillation comes in consolidated runs, which is worth knowing as device data become richer. It does not yet justify changing anticoagulation decisions on density alone.

• AF density measures how tightly episodes cluster, separate from total burden.
• High density was linked to a 75% higher one-year stroke risk in a dose-response pattern.
• The association held across device type, comorbidity, age and anticoagulation.
• Density is not yet a routine metric and should not override standard risk scores.
• Expect device reports to start surfacing temporal pattern, not just burden.

Why it matters

It challenges the assumption that total AF burden is the whole story for stroke risk.

Don't overread it

This is observational and does not show that acting on density changes outcomes.

The statistics, in plain English

A risk ratio of 1.75 with a confidence interval from 1.25 to 2.44 is a moderate, statistically solid association; because it is observational it describes risk, not a treatment effect.

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.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.