From Wiki Journal Club
Jump to: navigation, search
Gladstone DJ, et al. "Atrial Fibrillation in Patients with Cryptogenic Stroke". The New England Journal of Medicine. 2014. 370(26):2467-2477.
PubMedFull textPDF

Clinical Question

Does 30-day EKG monitoring compared with 24-h Holter monitoring improve detection of atrial fibrillation in patients following cryptogenic stroke?

Bottom Line

Ambulatory 30-day EKG monitoring improved the detection of atrial fibrillation compared to 24-h Holter monitoring.

Major Points

Approximately one-quarter of ischemic strokes and half of TIAs have an etiology that is not identified readily after the routine evaluation, and are labeled as cryptogenic. It has long been suspected that a significant portion of these patients have paroxysmal atrial fibrillation that is not identified during the conventional Holter monitoring period which is part of this standard evaluation. A prior randomized study demonstrated improved AF detection with extended EKG monitoring of up to 1 week following cryptogenic stroke.[1]

Published in 2014, the 30-Day Cardiac Event Monitor Belt for Recording Atrial Fibrillation after a Cerebral Ischemic Event (EMBRACE) trial sought to quantify the benefits of longer 30-day monitoring for AF detection following cryptogenic stroke. EMBRACE enrolled patients ≥55 years with a cryptogenic stroke within prior 6 months, with cryptogenic stroke defined based on TOAST criteria and requiring a stroke neurology evaluation that included EKG, 24-hour Holter monitor, vascular imaging, and echocardiography. Patients were randomized in a 1:1 ratio to either the intervention group which received extended EKG monitoring up to 30 days, or to the control group which underwent an additional 24 hours of Holter monitoring. The primary outcome was the detection of atrial fibrillation or flutter episodes lasting ≥30 seconds within 90 days after randomization. Secondary outcomes included duration of detected AF episodes and proportion of patients who initiated anticoagulation. At 90 days, the primary outcome of AF detection occurred in 16.1% of patients in the extended monitoring group compared to 3.2% in the control arm, resulting in a number needed to screen (NNS) of 8 to detect subclinical AF. Nearly all AF events were detected by monitoring rather than by clinical evaluation. AF episodes lasting ≥2.5 mins were detected in 9.9% of the extended monitoring group and 2.5% of the control group (NNS=14). At 90 days, more patients in the extended monitoring group were receiving anticoagulation compared to the control group (18.6% vs. 11.1%).

The authors conclude that subclinical AF is common following cryptogenic stroke and that 30-day EKG monitoring is superior to standard monitoring for detecting subclinical AF in this patient population. The concurrently published CRYSTAL-AF trial studied very long-term continuous EKG monitoring following cryptogenic stroke and similarly demonstrated a high rate of AF detection following cryptogenic stroke.[2]

Practice varies with respect to the optimal form of EKG monitoring following cryptogenic stroke. The fact that half of AF episodes occurred in the first week of monitoring suggests that this period is high yield for monitoring. A cost analysis by the EMBRACE steering committee suggested that 14-day monitoring is likely cost effective.[3]

Several questions remain unanswered on the basis of these studies' designs. Is the subclinical AF detected with extended EKG monitoring the likely cause of the stroke, or could this AF be a transient response to vascular changes following stroke? Do patients with brief episodes of occult AF (≥30 seconds) in the 6 months following cryptogenic stroke have a higher rate of cardioembolism warranting anticoagulation? Finally, it is important to recognize that while extended EKG monitoring (1) clearly identifies more patients with subclinical AF following cryptogenic stroke and (2) results in more patients placed on anticoagulation, no well designed randomized trial answers the question of whether extended EKG monitoring improves outcomes such as recurrent stroke or mortality in this patient group.


AHA/ASA Guidelines for the Prevention of Stroke in Patients With Stroke and Transient Ischemic Attack (2014, adapted)

  • For patients who have experienced an acute ischemic stroke or TIA with no other apparent cause, prolonged rhythm monitoring (≈30 days) for AF is reasonable within 6 months of the index event.

Canadian Stroke Best Practice Recommendations-secondary prevention of stroke guidelines (2014, adapted)

  • In cases where the ECG or initial cardiac rhythm (eg, 24- or 48-h EKG monitoring) does not show atrial fibrillation but a cardioembolic mechanism is suspected, prolonged EKG monitoring is recommended in selected patients for the detection of paroxysmal atrial fibrillation (eg, older patients with recent embolic stroke of undetermined source who are potential candidates for anticoagulant therapy).


  • Investigator-initiated, open-label, multicenter, randomized controlled trial
  • N=572 patients with cryptogenic stroke in prior 6 months
    • Extended EKG monitoring (n=287)
    • 24-hour Holter monitor (n=285)
  • Setting: 16 Canadian centers
  • Enrollment: 2009-2012
  • Follow-up: 90 days
  • Analysis: Intention-to-treat
  • Primary outcome: Detection of atrial fibrillation or flutter episode lasting ≥30 seconds within 90 days after randomization


Inclusion Criteria

  • Age ≥55 years
  • Cryptogenic stroke within prior 6 months based on TOAST criteria
  • Diagnosis of stroke by stroke neurologist after standard workup including:
    • 12-lead EKG
    • Ambulatory EKG monitoring for minimum 24 hours
    • Brain and neurovascular imaging
    • Echocardiography

Exclusion Criteria

  • Prior AF
  • Exclusive retinal TIA or stroke
  • Likely stroke etiology already identified
  • Planned carotid endarterectomy within 90 days
  • Indication for long-term anticoagulation
  • Endocarditis
  • Pacemakers or ICD
  • Known skin reactions to synthetic polymers or to silver

Baseline Characteristics

From the extended monitoring group.

  • Mean age: 72.8 years
  • Female: 46.2%
  • Race: White 89.9%, Asian 5.2%, Black 2.1%, Other 2.8%
  • Modified Rankin score <2: 95.8%
  • Medical history:
    • Hypertension: 71.3%
    • Diabetes: 19.2%
    • Hyperlipidemia: 66.8%
    • Smoking Status: Current 6.6%, Prior 49.3%
    • Stroke 15.7%, >1 stroke 4.2%, TIA 14.7%
    • CHF 1.7%
    • MI 16.8%
    • Coronary angioplasty or stenting: 8.4%
    • Coronary bypass surgery: 10.1%
    • Cardiac valve surgery 2.1%
  • Index event: Stroke 65.7%, TIA 34.3%
  • Oxfordshire classification of index event:
    • Total anterior circulation syndrome 2.4%
    • Partial anterior circulationg syndrome 70.3%
    • Posterior circulation syndrome 22.0%
    • Lacunar syndrome 5.2%
  • Number of days from index event to randomization: 76.6±37.5 days


  • Patients randomly assigned in 1:1 ratio to either:
    • Extended AF monitoring with a 30-day event-triggered loop recorded
    • Additional 24-hour Holter monitor
  • The specific monitoring device in the intervention group was the ER910AF Cardiac Event Monitor, Braemar, attached to a dry-electrode chest belt from Cardiac Bio-Systems. This dry electrode system was chosen to improve patient compliance over conventional adhesive electrodes.
  • The intervention group was asked to wear the monitor as much as possible for 30 days, and could discontinue use of the monitor if AF was detected prior to the end of the 30-day monitoring period.
  • All abnormal rhythms were adjudicated by a cardiologist and an internist. Disagreements were resolved by consultation with another cardiologist.


Comparisons are extended monitoring group vs. 24-h Holter control group. NNS, number needed to screen.

Primary Outcomes

Detection of atrial fibrillation or flutter episode ≥30 sec within 90 day
16.1% vs. 3.2% (absolute difference 12.9%, 95% CI 5.7 to 12.5%, P<0.001; NNS=8)

Secondary Outcomes

Detection of atrial fibrillation with duration ≥2.5 min
9.9% vs. 2.5% (absolute difference 7.4%, 95% CI 8.8 to 29.4%, P<0.001; NNS=14)
Detection of atrial fibrillation of any duration
19.7% vs. 4.7% (absolute difference 15.0%, 95% CI 4.9 to 10.2%; P<0.001; NNS=7)
Anticoagulant therapy at 90 days
18.6% vs. 11.1% (absolute difference 7.5%, 95% CI 1.6 to 13.3%, P=0.01)
Antiplatelet therapy only at 90 days
79.6% vs. 88.2% (absolute difference −8.6%, 95% CI −14.6 to −2.5, P=0.006)
Therapy at randomization changed by 90 days from antiplatelet therapy to anticoagulant therapy
13.6% vs. 4.7% (absolute difference 8.9%, 95% CI 4.2 to 13.6%, P<0.001)
Therapy at randomization changed by 90 days from anticoagulant therapy to antiplatelet therapy
1.1% vs. 0.7%, (absolute difference 0.4%, 95% CI −1.2 to 1.9%, P=0.66)

Subgroup Analysis

AF detection rate was higher among patients who underwent randomization within 3 months after the index stroke or TIA compared with those who underwent randomization 3-6 months after index event (18.5% vs. 9.0%, P=0.049 for linear association).

Adverse Events

One patient in the 30-day EKG monitoring group had an adverse skin reaction.


  • Authors conceded that cutoff of 30 seconds for duration of atrial fibrillation used in this study was arbitrary. Duration of 2.5 minutes was chosen as a secondary endpoint because it was the maximum time of recording available on the device chosen for the study. Previous trials such as ASSERT[4] and a MOST substudy[5] showed atrial tachyarrhythmia lasting ≥5 mins were associated with a clinically significant risk of stroke and death.
  • Patients with high stroke burden which is more likely due to a cardioembolism embolism were underrepresented in this study compared to the population.
  • Direction of causality not determined by this study, thus it's not possible from this study alone to identify whether AF caused the stroke or the stroke caused the AF. However, the authors note that it is unlikely that the observed AF episodes were transient post-stroke events given that the median time from index stroke and study enrollment was about 2.5 months.
  • Compared with CRYSTAL AF, which had an average delay from index event to recording of 38 days, the delay from index event to recording was 75 days in EMBRACE. This may limit the applicability of the EMBRACE trial to patients for whom extended EKG monitoring is pursued immediately following the ischemic event.


Canadian Stroke Network

Further Reading

  1. Higgins P et al. Noninvasive cardiac event monitoring to detect atrial fibrillation after ischemic stroke: a randomized, controlled trial. Stroke 2013. 44:2525-31.
  2. Sanna T et al. Cryptogenic stroke and underlying atrial fibrillation. N. Engl. J. Med. 2014. 370:2478-86.
  3. Yong JH et al. Potential Cost-Effectiveness of Ambulatory Cardiac Rhythm Monitoring After Cryptogenic Stroke. Stroke 2016. 47:2380-5.
  4. Healey JS et al. Subclinical atrial fibrillation and the risk of stroke. N. Engl. J. Med. 2012. 366:120-9.
  5. Glotzer TV et al. Atrial high rate episodes detected by pacemaker diagnostics predict death and stroke: report of the Atrial Diagnostics Ancillary Study of the MOde Selection Trial (MOST). Circulation 2003. 107:1614-9.