Finding More AFib Has Not Clearly Meant Fewer Strokes
Patch screening lifted AFib diagnoses from 5.4% to 6.8%. Strokes ran 2.7% against 2.5%, a difference too small to call.
Evidence rating for the proposition “Screening asymptomatic individuals for atrial fibrillation using ambulatory ECG monitoring reduces stroke or systemic embolism”: Uncertain
Not clearly. In AMALFI, 5,040 older adults at elevated stroke risk were randomized to a 14-day ECG patch or usual care. Across those 5,040 adults, after 2.5 years AFib was on record for 6.8% of the screened group and 5.4% of controls, and stroke occurred in 2.7% against 2.5%, not a significant difference. LOOP and GUARD-AF found the same pattern: more detection, no significant stroke drop. The trial that used a smartwatch measured how often alerts were right, not strokes.
Does finding more AFib mean fewer strokes?
Not in the trials so far. Detection and stroke are different results. A screening test can find more AFib and still leave the stroke rate close to unchanged.
The chain has three links. The device must find AFib. The finding must lead to anticoagulation. Anticoagulation must then prevent strokes in people who would not otherwise have been treated. Trials have confirmed the first link and, in part, the second. The third is where results fall short.
How much more AFib did screening find?
Modestly with patches, and about three times as much with implants. AMALFI recruited 5,040 people aged 65 or older with elevated stroke risk from 27 UK primary care practices. Half were mailed a 14-day ECG patch. In the screened group, 84.4% wore and returned it.
After 2.5 years, AFib was on record for 172 people (6.8%) in the screened group and 136 (5.4%) in the control group. The ratio was 1.26 (95% CI 1.02 to 1.57). The authors call it a modest increase. Of the 89 people with AFib on the patch, 55% had an AFib burden under 10%.
LOOP implanted loop recorders in adults aged 70 to 90 with stroke risk factors. AFib was diagnosed in 31.8% of the recorder group and 12.2% of controls. GUARD-AF used a 14-day patch in US primary care. Diagnosis was 5% against 3.3%.
Anticoagulation followed diagnosis. In AMALFI, mean exposure by 2.5 years was 1.63 months against 1.14 months. In LOOP, 29.7% started anticoagulants against 13.1%. In GUARD-AF, 4.2% against 2.8%.
3 studies
- In 5,040 adults aged 65 and older with elevated stroke risk, mail-based 14-day patch screening was followed by an AFib record in 6.8% against 5.4% at 2.5 years (ratio 1.26, 95% CI 1.02 to 1.57). Mean anticoagulant exposure was 1.63 against 1.14 months.
- In 6,004 adults aged 70 to 90 with stroke risk factors, loop recorder screening raised AFib diagnosis from 12.2% to 31.8% and anticoagulant initiation from 13.1% to 29.7%.
- In 11,905 participants aged 70 and older, 14-day patch screening raised AFib diagnosis from 3.3% to 5% and anticoagulant initiation from 2.8% to 4.2%.
Did the extra diagnoses lead to fewer strokes?
Not significantly in the individual trials. In AMALFI, stroke occurred in 69 people (2.7%) screened and 64 (2.5%) controls. The rate ratio was 1.08 (0.76 to 1.53). Stroke was an exploratory outcome there.
LOOP found stroke or systemic embolism in 4.5% against 5.6% (hazard ratio 0.80, 0.61 to 1.05). GUARD-AF found stroke hospitalization in 0.7% against 0.6% (1.10, 0.69 to 1.75) and stopped enrolling early.
Combining seven trials gave a 7% relative reduction in stroke or systemic embolism (risk ratio 0.932, 0.873 to 0.996). That result is borderline and is covered in whether AFib screening reduces stroke.
A 2022 review by the AF-SCREEN collaboration reached a matching view. It said clinician-led trials with repeated or continuous ECG monitoring had failed to show a significant stroke reduction. It said companies should avoid claims about clinical outcomes until randomized trials show improvement.
3 studies
- In AMALFI, stroke occurred in 69 of 2,520 screened participants (2.7%) and 64 of 2,520 controls (2.5%), a rate ratio of 1.08 (95% CI 0.76 to 1.53), as an exploratory outcome.
- In LOOP, stroke or systemic embolism occurred in 4.5% of the recorder group and 5.6% of controls (hazard ratio 0.80, 95% CI 0.61 to 1.05).
- The AF-SCREEN Frontier review stated that clinician-led randomized studies using multiple ECG recordings or continuous monitoring had failed to demonstrate a significant reduction in stroke, and that companies should avoid claims about clinical outcomes unless randomized trials show improvement.
What did the big smartwatch study measure?
Alert accuracy, not strokes. The Apple Heart Study recruited 419,297 people without self-reported AFib. Over a median of 117 days, 2,161 (0.52%) received an irregular pulse notification.
Among 450 notified people who returned analyzable ECG patches, 34% had AFib on the patch. The positive predictive value was 0.84 (95% CI 0.76 to 0.92) for a notification confirmed by a simultaneous ECG. Apple funded the study.
That says whether an alert means something. It does not say whether acting on it changes stroke risk. Detection accuracy is covered in how accurately a wearable ECG detects AFib. The overdiagnosis question is in smartwatch AFib detection and overdiagnosis.
What does this not settle?
None of the stroke trials used a consumer smartwatch as the screening tool. They used patches and implants, mostly in adults aged 65 or 70 and older with stroke risk factors. Whether a watch alert in a younger or lower-risk person changes stroke risk has not been tested in a randomized trial in this evidence.
GUARD-AF reported low event rates and did not reach its planned size. A small real benefit could need larger or longer trials to show.
AMALFI, LOOP and GUARD-AF enrolled adults aged 65 or 70 and older, many with stroke risk factors, and screened with ECG patches or implanted recorders. They do not establish what a smartwatch alert does to stroke risk in a younger or lower-risk person.
Common questions
Does finding AFib with a wearable prevent strokes?
Not shown so far. In AMALFI, patch screening raised AFib records from 5.4% to 6.8% at 2.5 years, while stroke was 2.7% against 2.5%, a difference that was not significant. LOOP and GUARD-AF also found more detection without a significant stroke drop.
Does AFib screening lead to more anticoagulant use?
Yes in the trials. In LOOP, 29.7% of the recorder group started anticoagulants against 13.1% of controls. In GUARD-AF it was 4.2% against 2.8%. AMALFI reported mean exposure of 1.63 against 1.14 months.
What did the Apple Heart Study measure?
How often irregular pulse notifications were confirmed. Of 450 notified people with analyzable ECG patches, 34% had AFib on the patch, and the positive predictive value against a simultaneous ECG was 0.84. It did not measure strokes.
Did combining trials show fewer strokes?
Slightly. A meta-analysis of seven trials found a 7% relative reduction in stroke or systemic embolism (risk ratio 0.932, 95% CI 0.873 to 0.996), which was not significant when the implantable-recorder trial was excluded.
Do these trials apply to smartwatch alerts?
Not directly. The stroke trials used ECG patches and implanted loop recorders in older adults with stroke risk factors, not consumer smartwatch alerts.
Sources
- Remote Screening for Asymptomatic Atrial Fibrillation: The AMALFI Randomized Clinical Trial.
- Implantable loop recorder detection of atrial fibrillation to prevent stroke (The LOOP Study): a randomised controlled trial.
- Effect of Screening for Undiagnosed Atrial Fibrillation on Stroke Prevention.
- Clinical outcomes of atrial fibrillation screening: a meta-analysis of randomized controlled trials.
- Consumer-Led Screening for Atrial Fibrillation: Frontier Review of the AF-SCREEN International Collaboration.
- Large-Scale Assessment of a Smartwatch to Identify Atrial Fibrillation. [author affiliations include Apple; funding listed from Apple]