• atrial fibrillation
• stroke
• pacemakers
• implanted cardiac defibrillators
• ECG/electrocardiogram

The prevalence of atrial fibrillation (AF) among adults over the age of 20 years is estimated to be 3% with 25% of middle-aged adults expected to develop AF during their lifetime.1 However, undiagnosed AF is common, especially in higher risk patient populations, including the elderly and patients with concomitant cardiovascular comorbidities. For example, patients who presented with a cryptogenic stroke and no history of AF were diagnosed with AF at rates of 8.9% at 6 months and 12.4% at 12 months poststroke, when an implantable cardiac monitor was used.2 A screening programme in Sweden checked intermittent electrocardiograms (ECGs) over a 2-week period in the general population of people aged 75 years and found that 3% had undiagnosed AF with 0.5% being diagnosed with the first ECG.3

Despite the available data on the rates of undiagnosed AF, the optimal screening strategy for AF remains debated and an active area of investigation. The European Guidelines for AF recommend ‘opportunistic screening’ for patients ≥65 years of age by checking a pulse for irregularity or ECG rhythm strip and at least 72 hours of monitoring in patients with ischaemic stroke or transient ischaemic attack.1 In the lack of any large outcome screening studies, the American College of Cardiology/American Heart Association/Heart Rhythm Society guidelines do not provide recommendations for screening, while an international collaboration of providers recommend: (1) screening all patients ≥65 years of age with a single time point, single lead ECG or pulse check for irregularity followed by an ECG; (2) patient activated ECG screening for a 2-week period in patients ≥75 years of age or in younger patients at a high risk for AF; and (3) long-term continuous monitoring in patients with an embolic stroke of undetermined source (ESUS).4

Beyond the question of whom to screen, physicians need to decide which devices to use for AF screening, for how long and how to manage the results. Traditionally, providers have access to holter monitors, external loop recorders (ELR), mobile cardiac outpatient telemetry and implantable cardiac monitors (figure 1). However, the types of outpatient rhythm monitoring continue to expand with the availability of wearable and Smartphone-based technology (figure 1). Despite cardiac implantable electronic devices and wearable devices becoming more prevalent, the duration of subclinical AF that requires treatment with oral anticoagulation remains poorly defined and the management of subclinical AF is still controversial.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Screening technologies for atrial fibrillation.

Given the rapid expansion of monitoring for AF, it is critical to understand the diagnostic yield of the available modalities. In theirHeart paper, Sejr et al 5 present data from a prospective, consecutive cohort study evaluating the diagnostic yield for AF of ELR compared with continuous ECG recording over a 48-hour period. The study included 1412 Danish patients from a single centre with an acute ischaemic stroke (n=849, 60%) or transient ischaemic attack (n=562, 40%) between 2013 and 2017, and patients had no history of AF. Patients were predominantly men (n=797, 56%) with hypertension (n=825, 58%), and the mean age was 73 (SD 7.7) years. Continuous ECG monitoring for a 48-hour period was considered the gold standard and identified 2.7% (n=38) of patients with previously undiagnosed AF. The ELRs were processed with an automatic detection algorithm for AF, based on the irregularity of R-R intervals, and the recordings were then over-read by a cardiologist. When this technique was compared with the continuous ECG monitoring, 6 patients (15.8%) with AF were not diagnosed with the ELR, and 25 patients (1.8%) without AF were incorrectly diagnosed with AF by the ELR. After the ELRs were over-read by a cardiologist, the sensitivity and specificity for AF detection were 84.2% and 98.2%, respectively, while the positive and negative predictive values were 56.1% and 99.6%, respectively.

How should clinicians interpret these results? The answer to this question likely varies based on the risk factors of the patient and the goals of the provider in ordering rhythm monitoring. The current study demonstrates that the sensitivity is only 84.2%, so nearly one in six patients with AF was not detected by the ELR, which is concerning in a poststroke or transient ischaemic attack population in which the goal is not missing a diagnosis of AF. The findings from the authors also raise the interesting and challenging question of how to respond to a finding of AF on an ELR, as more than two in five patients with AF reported on ELR did not have AF on continuous ECG. It is important to note that the study showed that there was a higher mean duration of AF in cases when AF on the ELR was confirmed by continuous ECG monitoring (approximately 12 hours) versus unconfirmed cases of AF noted on ELR (approximately 1 hour). Although it was not reported in the manuscript, the positive predictive value, presumably, increases with longer episodes of AF. This is intuitive clinically, as episodes of sinus arrhythmia, premature atrial contractions and atrial runs are sporadic and are not persistent, so they are less likely to affect the diagnostic accuracy of longer episodes of irregular R-R intervals.

The positive predictive value is highly dependent on the event rate of true AF episodes. If the goal of a monitor in a patient after a stroke or transient ischaemic attack is to identify any episodes of AF, the ideal monitor would have a high sensitivity. The implantable cardiac monitors have been shown to have meaningfully higher sensitivity than the ELR findings in this study. Data from the Medtronic Reveal XT found a sensitivity and specificity of 96.1% and 85.4%, respectively, with improvement in the sensitivity and specificity with the Medtronic Reveal LINQ algorithms to 97.4% and 97.0%, respectively.6 The specificity with Medtronic Reveal LINQ is similar to the specificity of the ELR (98.2%) in the Sejr study.

Many of the cases in which AF was diagnosed by ELR and was then not confirmed by continuous ECG monitoring were due to frequent premature atrial contractions. However, the risk of subsequent AF is higher in patients with frequent premature atrial contractions.7 Therefore, it is possible that clinical outcomes of patients with AF detected by ELR are more similar to outcomes of patients with AF detected by continuous ECG, despite the differences in diagnostic yield.

Ultimately and more importantly, the clinical question is how to manage patients with an asymptomatic AF detected on an implantable device. The randomised trial of atrial arrhythmia monitoring to guide anticoagulation in patients with implanted defibrillator and cardiac resynchronization devices (IMPACT study) evaluated the use of oral anticoagulation around the time of subclinical AF detection on a cardiac implantable electronic device. Patients were randomised to a control versus an intervention group.8 The intervention group received oral anticoagulation, depending on CHADS₂ score, for 30 days, 90 days or indefinitely after a device detected AF event, which was defined based on any AF, AF ≥24 hours duration or AF ≥48 hours duration, again dependent on CHADS₂ score. The study found no difference between the two strategies in the primary endpoint of the composite of stroke, systemic embolism and major bleeding.8 Similarly, the trials (RESPECT-ESUS and NAVIGATE-ESUS) of oral anticoagulation versus aspirin in patients with an ESUS have not found a reduction in recurrent stroke with oral anticoagulation when compared with aspirin.9

The data from the ESUS trials suggest that only a portion of the embolic strokes were due to AF and potentially modifiable by oral anticoagulation, while data from the IMPACT study, despite several limitations including a composite endpoint of stroke and major bleeds, suggested that shorter duration of AF and longer use of oral anticoagulation may have more impact on long-term clinical events. There are two ongoing clinical trials evaluating the use of an oral anticoagulant versus aspirin or placebo in patients with subclinical AF for the prevention of stroke or systemic embolism. The Apixaban for the Reduction of Thrombo-Embolism in patients with device-detected Sub-clinical Atrial fibrillation (ARTESiA) trial (NCT02840201) is studying apixaban in 4000 patients, while the Non-vitamin k antagonist Oral Anticoagulants in patients with atrial High rate episodes (NOAH) trial (NCT02618577) is studying edoxaban in 3400 patients.

The key questions in this field are (1) how to screen for AF?; (2) for how long should we screen?; (3) what patient populations most benefit from screening?; (4) does screening for AF ultimately improve clinical outcomes?; and (5) what duration of subclinical AF should be treated with an oral anticoagulant to improve long-term outcomes. The findings by Sejr et al are quite informative about the first question of how to screen patients, particularly in patients with a stroke or transient ischaemic attack, where the need for a high-sensitivity screening test should move physicians away from ELRs and towards using implantable cardiac monitors for longer-term monitoring. More studies around screening for AF are needed to answer these important questions, and unfortunately, the clinical community will have to wait for the results of ARTESiA and NOAH before understanding more about the optimal management of patients with subclinical AF lasting <24 hours.