EMBRACE
Atrial Fibrillation in Patients with Cryptogenic Stroke
Year of Publication: 2014
Authors: David J. Gladstone, Melanie Spring, Paul Dorian, ..., for the EMBRACE Investigators
Journal: The New England Journal of Medicine
Citation: N Engl J Med 2014;370:2467-77.
PDF: Clinical Question
In patients ≥55 years with recent cryptogenic stroke/TIA, does 30-day ambulatory ECG monitoring detect more atrial fibrillation than a single 24-hour Holter?
Bottom Line
30-day ambulatory ECG monitoring detected AF (≥30 sec) in 16.1% vs 3.2% with 24-hour Holter (absolute difference 12.9 percentage points; P<0.001; NNS=8). Prolonged monitoring nearly doubled anticoagulation rate at 90 days (18.6% vs 11.1%; P=0.01). AF detection was incremental: 2.2% at 24h rising to 14.8% by 4 weeks.
Major Points
- AF (≥30 sec) detected in 16.1% (45/280) with 30-day monitoring vs 3.2% (9/277) with 24-hour Holter — >5-fold increase (P<0.001; NNS=8).
- Anticoagulation at 90 days: 18.6% vs 11.1% (P=0.01). Switch from antiplatelet to anticoagulant: 13.6% vs 4.7% (P<0.001).
- AF detection incremental over time: 2.2% at 24h → 7.4% at 1 week → 11.6% at 2 weeks → 14.8% at 4 weeks.
- 59% of AF patients had ≥2 episodes; 64% had episode ≥2.5 minutes.
- Earlier monitoring (within 3 months of event) detected more AF: 18.5% vs 9.0% (P=0.049).
- 82% completed ≥3 weeks of monitoring (adherence).
- AF detected by clinical means alone in only 0.5% — monitors essential.
- 572 patients, 16 Canadian stroke centers, randomized open-label.
- Device: ER910AF event-triggered loop recorder with dry-electrode belt. ECG data transmitted transtelephonically.
- Challenged standard 24-48h monitoring practice; supported prolonged outpatient cardiac monitoring post-stroke.
Design
Study Type: Randomized controlled trial (investigator-initiated, open-label)
Randomization: 1
Blinding: Open-label; ECG adjudicators blinded to patient demographics. 1:1, Web-based, variable block size.
Enrollment Period: June 2009 to March 2012
Follow-up Duration: 90 days
Centers: 16
Countries: Canada, Ireland
Sample Size: 572
Analysis: ITT (intention-to-monitor). Pearson's chi-square. No interim analyses.
Inclusion Criteria
- Age ≥55 years.
- No known atrial fibrillation.
- Cryptogenic ischemic stroke or TIA (TOAST) within previous 6 months.
- Diagnosed by stroke neurologist after standard workup: 12-lead ECG, ≥24h Holter, brain/neurovascular imaging, echocardiography.
Exclusion Criteria
- Etiologic diagnosis already determined (large-vessel, small-vessel, or other known cause).
Baseline Characteristics
| Characteristic | Intervention — 30-day monitor (N=286) | Control — 24h Holter (N=285) |
|---|---|---|
| Mean age (±SD) | 72.5±8.5 | 73.2±8.8 |
| Age ≥75 | 104 (36.4%) | 118 (41.4%) |
| Female | 132 (46.2%) | 125 (43.9%) |
| White | 257 (89.9%) | 260 (91.2%) |
| mRS ≤2 | 274 (95.8%) | 263 (92.3%) |
| Hypertension | 204 (71.3%) | 191 (67.0%) |
| Diabetes | 55 (19.2%) | 55 (19.3%) |
| Hyperlipidemia | 191 (66.8%) | 177 (62.1%) |
| Previous ischemic stroke | 45 (15.7%) | 36 (12.6%) |
| Index event — Stroke | 188 (65.7%) | 172 (60.4%) |
| Index event — TIA | 98 (34.3%) | 113 (39.6%) |
| Days from event to randomization (mean±SD) | 76.6±37.5 | 73.7±39.7 |
Arms
| Field | 30-Day Event-Triggered Loop Recorder | Control |
|---|---|---|
| Intervention | ER910AF Cardiac Event Monitor (Braemar) attached to dry-electrode belt (Cardiac Bio-Systems) worn around chest for 30 days. Auto-records AF based on R-R interval irregularity. ECG transmitted transtelephonically for central interpretation. | One additional round of conventional 24-hour ambulatory ECG monitoring at local site. All reports reviewed centrally. |
| Duration | 30 days (could stop if AF detected) | 24 hours |
Outcomes
| Outcome | Type | Control | Intervention | HR / OR / RR | P-value |
|---|---|---|---|---|---|
| Detection of AF ≥30 seconds within 90 days | Primary | 9/277 (3.2%) | 45/280 (16.1%) | 8 | <0.001 |
| AF ≥30 sec (study-monitor detected) | Secondary | 7/277 (2.5%) | 44/284 (15.5%) | Diff 13.0 pp | <0.001 |
| AF ≥2.5 minutes | Secondary | 7/277 (2.5%) | 28/284 (9.9%) | Diff 7.4 pp; NNS=14 | <0.001 |
| AF of any duration | Secondary | 13/277 (4.7%) | 56/284 (19.7%) | Diff 15.0 pp; NNS=7 | <0.001 |
| Anticoagulant therapy at 90 days | Secondary | 31/279 (11.1%) | 52/280 (18.6%) | Diff 7.5 pp | 0.01 |
| Switch antiplatelet → anticoagulant | Secondary | 13/279 (4.7%) | 38/280 (13.6%) | Diff 8.9 pp | <0.001 |
| Adverse skin reaction (ECG electrodes) | Adverse | 1/287 (0.3%) - led to intervention drop | |||
| Study withdrawal (non-medical) | Adverse | 8 | 10 | ||
| Death from ischemic stroke during follow-up | Adverse | 1 | 1 | ||
| Device / monitoring-related AEs | Adverse | No systemic AEs reported; monitoring generally well tolerated |
Subgroup Analysis
Earlier monitoring (within 3 months of event): 18.5% vs 9.0% AF detection (P=0.049). Patients with AF were older with more atrial ectopy on baseline Holter (P<0.001 for both).
Criticisms
- Cannot determine total AF burden — limited 2.5-minute recording capacity per episode.
- Monitoring started late (mean 75 days post-event) — earlier initiation may detect more.
- 30 days may underestimate true yield (implanted devices detect more with longer monitoring).
- Cryptogenic stroke heterogeneous — not all patients had intracranial imaging or TEE.
- Mild nondisabling strokes overrepresented (mRS ≤2: 94%).
- Open-label design (though ECG adjudicators blinded).
- 16% AF prevalence likely conservative underestimate.
- Detection of AF after stroke does not prove causation.
Funding
Canadian Stroke Network; Heart and Stroke Foundation of Ontario. Device manufacturers purchased equipment but had no study role.
Based on: EMBRACE (The New England Journal of Medicine, 2014)
Authors: David J. Gladstone, Melanie Spring, Paul Dorian, ..., for the EMBRACE Investigators
Citation: N Engl J Med 2014;370:2467-77.
Content summarized and formatted by NeuroTrials.ai.