CRYSTAL AF
(2014)Objective
To assess whether long-term ECG monitoring with an insertable cardiac monitor (ICM) is more effective than conventional follow-up for detecting atrial fibrillation in patients with cryptogenic stroke
Study Summary
• Most detected AF episodes (74-79%) were asymptomatic and would have been missed by symptom-triggered monitoring
• Detection led to more than doubling of oral anticoagulant use in the ICM group
Intervention
Insertable cardiac monitor (REVEAL XT) vs conventional follow-up with ECG monitoring at physician discretion
Inclusion Criteria
Patients ≥40 years with cryptogenic stroke or TIA within 90 days after extensive workup including ≥24 hours ECG monitoring, TEE, and vascular imaging
Study Design
Arms: ICM group (insertable cardiac monitor) vs Control group (conventional follow-up)
Patients per Arm: ICM: 221, Control: 220
Outcome
• AF detected at 12 months: 12.4% ICM vs 2.0% control (HR 7.3, P<0.001)
• Oral anticoagulant use at 12 months: 14.7% ICM vs 6.0% control (P=0.007)
Bottom Line
Among patients with recent cryptogenic stroke, continuous ICM monitoring detected AF in 8.9% vs 1.4% with conventional follow-up at 6 months (HR 6.4; P<0.001), 12.4% vs 2.0% at 12 months (HR 7.3; P<0.001), and 30.0% vs 3.0% at 36 months (HR 8.8; P<0.001). Most AF episodes were asymptomatic (74-79%). Oral anticoagulant use was doubled in the ICM group (14.7% vs 6.0% at 12 months; P=0.007). NNT to detect one AF: 14 at 6 months, 10 at 12 months, 4 at 36 months.
Major Points
- ICM detected 6x more AF at 6 months: 8.9% (19/221) vs 1.4% (3/220); HR 6.4 (95% CI 1.9-21.7; P<0.001).
- Superiority increased over time: 12 months 12.4% vs 2.0% (HR 7.3); 36 months 30.0% vs 3.0% (HR 8.8) — all P<0.001.
- Most AF was asymptomatic: 74% at 6 months, 79% at 12 months — explaining why conventional monitoring fails.
- Conventional monitoring was inadequate: control group had only 88 ECGs in 65 patients and 20 Holters in 17 patients over 6 months.
- ICM led to more anticoagulation: 10.1% vs 4.6% at 6 months (P=0.04); 14.7% vs 6.0% at 12 months (P=0.007). 97% with detected AF received anticoagulants.
- NNT to detect one AF: 14 at 6 months, 10 at 12 months, 4 at 36 months.
- Trend toward fewer recurrent strokes: 5.2% vs 8.6% at 6 months, 7.1% vs 9.1% at 12 months — not powered for this endpoint.
- AF burden substantial when detected: median max single-day AF 11.2 hours; 46% had episodes >12 hours.
- ICM safe: only 2.4% removed for infection/erosion. Retained in 98.1% at 6 months.
- Consistent across all subgroups: no interaction by age, sex, PFO, CHADS2, or index event type.
Study Design
- Study Type
- Randomized, controlled, parallel-group, open-label trial
- Randomization
- Yes
- Blinding
- Open-label (ICM insertion precludes blinding). Permuted blocks, stratified by index event type (stroke/TIA) and PFO status.
- Sample Size
- 441
- Follow-up
- Primary at 6 months; secondary at 12 months; long-term up to 36 months. Total 815.5 patient-years.
- Centers
- 55
- Countries
- United States, Canada, Germany, Italy, Belgium
Primary Outcome
Definition: Time to first detection of AF (>30 seconds) at 6 months
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| 3/220 (1.4%) | 19/221 (8.9%) | 6.4 (1.9-21.7) | <0.001 |
Limitations & Criticisms
- Causality uncertain — newly discovered AF may not be causally related to the index stroke.
- Clinical significance of brief AF episodes unknown.
- Device memory limitation — once full, oldest episodes overwritten.
- AF detection algorithm accuracy 98.5% — not perfect.
- Open-label design — could influence monitoring intensity in control group.
- Underpowered for stroke recurrence endpoint.
- Low detection in control (1.4%) reflects inadequate conventional monitoring (only 88 ECGs, 20 Holters).
- Predominantly white (87%) — limited diversity.
- No mandate for anticoagulation upon AF detection — physician discretion.
- Small number of control AF events (n=3 at 6mo) yields wide confidence intervals.
Citation
N Engl J Med 2014;370:2478-86.