STROKE-AF
Stroke of Known Cause and Underlying Atrial Fibrillation
Clinical Question
Is long-term cardiac monitoring with an insertable cardiac monitor more effective than usual care for detecting atrial fibrillation in patients with ischemic stroke attributed to large- or small-vessel disease?
Bottom Line
In patients with non-cryptogenic ischemic stroke attributed to large- or small-vessel disease, ICM monitoring detected significantly more atrial fibrillation than usual care at 12 months (12.1% vs 1.8%, HR 7.4). Nearly all detected AF episodes (96%) were asymptomatic. However, the clinical importance of detecting AF in this population remains uncertain.
Major Points
- At 12 months, AF was detected in 12.1% of ICM patients vs 1.8% of controls (HR 7.4, 95% CI 2.6-21.3, P<0.001)
- At 6 months, AF was detected in 7.9% of ICM patients vs 0.8% of controls (HR 9.9, 95% CI 2.3-43.5, P=0.002)
- 96.3% of first AF episodes detected in the ICM group were asymptomatic (26 of 27 patients)
- Only 6 patients (2.6%) had AF detected in the first 30 days, meaning 78% of AF detected at 12 months would have been missed by 30 days of continuous monitoring
- AF detection rates were similar in large-vessel (11.7%) vs small-vessel (12.6%) stroke subtypes (P=0.74)
- Median duration of longest AF episode was 88 minutes (IQR 10-526 minutes); 55.5% had episodes >1 hour
- OAC prescription at 12 months was higher in ICM group: 15.7% vs 5.6% (OR 3.1, P<0.001)
- Recurrent stroke rates were numerically lower in ICM group: 7.2% vs 9.5% (HR 0.7, P=0.30), but study was not powered for this endpoint
Design
Study Type: Randomized controlled trial, multicenter, parallel-group, open-label
Randomization: 1
Blinding: Open-label; patients, physicians, and clinical events committee were not blinded to randomization assignment due to nature of intervention. However, aggregate efficacy results remained blinded until completion of primary endpoint analysis.
Enrollment Period: April 2016 to July 2019
Follow-up Duration: Primary endpoint at 12 months; planned follow-up up to 36 months
Centers: 33
Countries: United States
Sample Size: 492
Analysis: Intention-to-treat analysis. Kaplan-Meier survival estimates with Cox proportional hazards regression for hazard ratios. Log-rank test for between-group comparisons. Mixed-effects model for site effects. SAS version 9.4 and R version 3.6.0.
Inclusion Criteria
- Age ≥60 years, or age 50-59 years with at least 1 additional stroke risk factor (congestive heart failure, hypertension, diabetes, prior ischemic stroke >90 days before index stroke, or other ischemic vascular disease)
- Index ischemic stroke attributed to large-artery atherosclerosis (large-vessel disease) or small-vessel occlusion (small-vessel disease) per TOAST criteria
- Index stroke within 10 days prior to ICM insertion
- TOAST classification assigned by enrolling investigator as used in clinical practice
Exclusion Criteria
- Cryptogenic stroke or embolic stroke of undetermined source
- Cardioembolic stroke
- History of documented atrial fibrillation or atrial flutter
- Known indication or contraindication for long-term oral anticoagulation
- Untreated hyperthyroidism
- Myocardial infarction or cardiac surgery less than 1 month prior to index stroke
- Mechanical heart valve or valvular disease requiring immediate surgery
Arms
| Field | Control | ICM |
|---|---|---|
| Intervention | Site-specific usual care consisting of external cardiac monitoring such as 12-lead ECGs, Holter monitoring, telemetry, or event recorders at treating physician discretion. Follow-up visits at 1, 6, and 12 months. | Insertable cardiac monitor (REVEAL LINQ, Medtronic) inserted within 10 days of index stroke and after randomization. Device insertion per site standard procedures. ICMs programmed according to standardized requirements. Follow-up visits at device insertion, 1, 3, 6, 9, and 12 months. |
| Duration | 12 months (primary); up to 36 months planned | 12 months (primary); up to 36 months or end of ICM battery life |
Outcomes
| Outcome | Type | Control | Intervention | HR / OR / RR | P-value |
|---|---|---|---|---|---|
| Time to first detection of atrial fibrillation (episode of irregular heart rhythm without detectable P waves lasting >30 seconds, adjudicated by clinical events committee) through 12 months | Primary | 1.8% (4 patients) | 12.1% (27 patients) | 7.4 | <0.001 |
| AF detection at 6 months | Secondary | 0.8% (2 patients) | 7.9% (18 patients) | 9.9 | 0.002 |
| AF detection in large-vessel disease subgroup at 12 months | Secondary | 2.3% (3 patients) | 11.7% (15 patients) | 5.3 | <0.001 |
| AF detection in small-vessel disease subgroup at 12 months | Secondary | 1.0% (1 patient) | 12.6% (12 patients) | 13.8 | <0.001 |
| OAC prescription at 12 months | Secondary | 5.6% (14 patients) | 15.7% (38 patients) | 3.1 | <0.001 |
| OAC prescription among patients with detected AF | Secondary | 1.2% (3 patients) | 7.4% (18 patients) | 6.6 | <0.001 |
| Recurrent ischemic or hemorrhagic stroke at 12 months | Secondary | 9.5% (23 patients) | 7.2% (16 patients) | 0.7 | 0.30 |
| Recurrent TIA at 12 months | Secondary | 0.4% (1 patient) | 1.8% (4 patients) | 4.1 | 0.21 |
| Asymptomatic first AF episode (ICM group) | Secondary | N/A | 96.3% (26/27 patients) | ||
| ICM procedure-related adverse events (total) | Adverse | N/A | 1.8% (4/221 patients) | ||
| Site infection | Adverse | N/A | 0.5% (1 patient) | ||
| Incision site hemorrhage | Adverse | N/A | 0.9% (2 patients) | ||
| Implant site pain | Adverse | N/A | 0.5% (1 patient) | ||
| Death through 12 months | Adverse | 5 patients | 12 patients |
Subgroup Analysis
A Cox model including interaction term between stroke subtype and randomization group showed no significant interaction (P=0.42). The group effect remained significant (P=0.009) while stroke subtype was not significant (P=0.46). AF detection was similar in large-vessel (11.7%) vs small-vessel (12.6%) stroke in the ICM group (HR 0.9, 95% CI 0.4-1.9, P=0.74). Mixed-effects model treating study site as random effect showed site effect was not significant (P=0.10).
Criticisms
- Open-label design with no blinding of patients, physicians, or clinical events committee to randomization assignment
- Attribution of stroke mechanism is subjective and may have enrolled patients at higher risk of underlying embolism
- Cannot determine if AF was related to the stroke, unmasked by it, or coexistent background AF that would be detected without stroke
- Study not powered to detect differences in clinical outcomes such as recurrent stroke
- Higher frequency of congestive heart failure among patients with incomplete follow-up may have underestimated true AF incidence
- ICM detection algorithm requires episodes ≥2 minutes, while primary endpoint defined AF as >30 seconds
- ECG monitoring in control group was variable and limited, reflecting real-world practice but reducing AF detection sensitivity
- TOAST classifications applied per clinical practice may have greater interrater variation than validated algorithms
- No comparator group without stroke to establish background AF incidence
- Study cannot make conclusions about clinical value of testing without clinical outcome data
Funding
Medtronic Inc
Based on: STROKE-AF (JAMA, 2021)
Authors: Richard A. Bernstein, Hooman Kamel, Christopher B. Granger, ..., for the STROKE-AF Investigators
Citation: JAMA. 2021;325(21):2169-2177
Content summarized and formatted by NeuroTrials.ai.