Newer technologies for detection of atrial fibrillationBMJ 2018; 363 doi: https://doi.org/10.1136/bmj.k3946 (Published 17 October 2018) Cite this as: BMJ 2018;363:k3946
- Nath Zungsontiporn, cardiology fellow,
- Mark S Link, professor of internal medicine
- UT Southwestern Medical Center, Department of Internal Medicine, Division of Cardiology, Dallas, TX, USA
- Correspondence to: M Link
Atrial fibrillation is a common arrhythmia that is associated with increased risk of stroke, which can be reduced with appropriate anticoagulation treatment. However, it remains underdiagnosed in contemporary clinical practice using conventional detection methods, resulting in missed opportunities to implement appropriate treatment. Newer technologies developed in recent years can potentially enhance the detection of atrial fibrillation and overcome certain limitations of the conventional methods. However, uncertainties remain about their use and the significance of atrial fibrillation detected by some of these newer technologies. This review examines the evidence supporting the use of some of these technologies and evaluates their applications in certain clinical scenarios.
Atrial fibrillation is the most common sustained cardiac arrhythmia globally.1 It is associated with an increased risk of cardiovascular morbidity and mortality, including a threefold to fivefold increased risk of stroke and a greater than threefold increased risk of heart failure.23
Atrial fibrillation is usually diagnosed when patients develop symptoms that prompt them to seek medical evaluation or incidentally when they attend a healthcare professional for another reason. More than a third of patients diagnosed as having atrial fibrillation have no attributable symptoms,45 and many patients do not seek medical evaluation routinely; thus, the diagnosis of atrial fibrillation is often delayed. Data from studies of screening for atrial fibrillation in populations at risk confirm that it is diagnosed in people without symptoms.678 As patients with undiagnosed atrial fibrillation remain at increased risk of cardiovascular events such as stroke,4 which could be reduced with appropriate treatment, a clear need exists for an improvement in its detection and diagnosis. The National Heart, Lung, and Blood Institute expert panel identified this as one of the high priority research areas.9
In recent years, newer technologies to detect atrial fibrillation have been exponentially developed. These technologies can enhance the ability to detect atrial fibrillation and overcome certain limitations of conventional methods. This review will examine evidence supporting the use of these technologies and evaluate clinical scenarios in which they may benefit patients.
Atrial fibrillation had an estimated global prevalence of 33.5 million people in 2010.1 In the US, more than 3 million adults are estimated to be living with atrial fibrillation.10 In China, the prevalence of atrial fibrillation in adults aged 20 years or over was assessed at 0.2/100 between 2001 and 2012.11 In the UK, the prevalence was estimated to be 14.5/1000 in 2010.12 The prevalence of atrial fibrillation is projected to increase in the future, with more than 5.6 million and 1.2 million adults predicted to have atrial fibrillation in the US by 2050 and in the UK by 2060, respectively.1012
The incidence of atrial fibrillation is estimated to double with each successive decade beyond the age of 50.13 Men have a 1.5-fold greater risk of developing atrial fibrillation than women.13 In the Framingham study, the biennial incidence of atrial fibrillation in men and women aged 55-64 years was 6.2 and 3.8 cases per 1000 person examinations, respectively, and was 75.9 and 62.8 cases per 1000 in men and women aged 85-94 years.13 The Multi-Ethnic Study of Atherosclerosis (MESA) reported that non-Hispanic white people had higher age specific incidence rates than other ethnic groups.14
The risk of atrial fibrillation is also related to heart failure, valvular heart disease, and coronary artery disease.1315 Echocardiographic abnormalities, such as left atrial enlargement,1516 left ventricular hypertrophy, and left ventricular systolic and diastolic dysfunction, have been associated with atrial fibrillation.1516171819 Other important risk factors include pulmonary diseases, such as obstructive sleep apnea and chronic obstructive pulmonary disease,2021 and systemic risk factors, such as hypertension, diabetes mellitus,1315 obesity,16 and cigarette smoking.22
Sources and selection criteria
We searched PubMed for relevant literature from 1 January 2010 to 15 March 2018. We used the search term “atrial fibrillation” in combination with (detect OR detection), (monitor OR monitoring), (smartphone OR “mobile phone”), (mobile AND telemetry), handheld, thumb, zenicor, alivecor, mydiagnostick, watch, photoplethysmography, and “implantable loop recorder”. We reviewed only studies published in English and included studies on the basis of quality and size. The priority of studies reviewed ranged from randomized controlled trials (RCTs), which were highest, followed by prospective cohort studies and cross sectional studies. However, we included retrospective and case-controlled studies when data were not available from higher quality studies. We also identified references from reference lists of relevant review articles and relevant studies. We reviewed approximately 4000 search results.
Conventional detection methods and their limitations
Twelve lead electrocardiography remains the gold standard for the diagnosis of an arrhythmia, but its ability to detect atrial fibrillation is limited by the short time frame of recording. Atrial fibrillation, especially in patients with new onset, is usually paroxysmal and short in duration and as a result has often terminated by the time a standard electrocardiogram is acquired. To overcome this limitation, several types of ambulatory electrocardiographic recording devices have been developed (table 1). These devices are traditionally categorized according to the continuity of recording into “continuous” and “intermittent” recorders.232425 Although these devices have increased the diagnostic yield for atrial fibrillation,26 certain limitations such as inadequate duration of monitoring, insufficient sensitivity of the atrial fibrillation detection method, and discomfort associated with use of the device remain important barriers for detection of atrial fibrillation.
Newer detection methods
Non-invasive monitoring devices
Continuous recording devices
Cutaneous patch monitors are exemplified by the Zio Patch (iRhythm Technologies, San Francisco, CA, USA) (fig 1). This device is a single use, water resistant, cutaneous patch that continuously records single lead electrocardiography for up to 14 days. It has an integrated button that, when pressed, marks the timing of symptoms on the recorded rhythm tracing. After completion of the recording, the device is returned to the manufacturer for data retrieval, analysis of arrhythmias by proprietary algorithm, and further examination of detected arrhythmias by technicians. A report of the cardiac electrical activity is then sent to the ordering physician.
Although the Zio Patch records only a single lead electrocardiograph, its ability to detect atrial fibrillation seems to be similar to that of a multi-lead conventional Holter monitor when both devices monitor patients for the same length of time.2728 As expected from the longer duration of monitoring, the Zio Patch detects 60-70% more arrhythmias, including atrial fibrillation, over the total wear time compared with conventional Holter monitors (table 2).2728
As the Zio Patch has no leads, it is generally more comfortable to wear than conventional Holter monitors. More than 90% of patients in a cohort study found the Zio Patch comfortable to wear as opposed to 52% for the Holter monitor. In addition, the Zio Patch affected activities of daily living in only 11% of patients as opposed to 76% of patients in the Holter group.28
The feasibility of using the Zio Patch to identify atrial fibrillation in patients at higher risk was evaluated in a small prospective screening study. The definition of patients at higher risk was those aged 55 years or over with at least two of the following risk factors: coronary disease, heart failure, hypertension, diabetes, and sleep apnea. Of 75 enrolled patients (100% male; mean age 69 (SD 8.0) years; CHA2DS2-VASc ≥2 in 97%), atrial fibrillation for at least 30 seconds was detected in four (5%) patients. However, these four participants had at least one episode in the first 48 hours and could have been detected by conventional Holter monitor (table 3).30
Mobile cardiac telemetry is a cardiac monitoring device system that consists of three to four electrocardiography electrodes, which generate two to three leads. Some of the newer systems consist of only a one lead patch sensor and a monitor. The device continuously monitors the patient and has a proprietary algorithm to detect arrhythmias automatically. Automatically detected arrhythmias as well as patient activated rhythms are transmitted in real time to a central monitoring station, which is staffed by trained electrocardiogram monitoring technicians.
CardioNet MCOT (Conshohocken, PA, USA) is one of the most well studied mobile cardiac telemetry systems. Its algorithm to detect arrhythmia is based on rate, rhythm irregularity, QRS morphology, and P wave analysis.44 In an RCT that included 266 patients with symptoms suggestive of a significant cardiac arrhythmia who were monitored for up to 30 days, this system has been shown to detect significantly more arrhythmias, including atrial fibrillation and atrial flutter (23%), than standard external loop recorders (8%) (table 2).29 Most of the additional arrhythmias detected by the CardioNet MCOT have no temporally related symptoms, which is not unexpected as the external loop recorder with an auto-trigger algorithm was used in only 16% of the patients. Although data from retrospective analyses suggest that this system may still detect more arrhythmias than a loop recorder with an auto-trigger function,2944 conclusive data are still lacking.
Retrospective studies reported detection rates of atrial fibrillation between 17% and 23% by CardioNet MCOT in patients with cryptogenic stroke or transient ischemic attack (TIA).4546 However, approximately 67-77% of these patients had only atrial fibrillation for less than 30 seconds, and the clinical significance of these short atrial fibrillations is not clear (as discussed below). To date, no prospective studies have evaluated the use of mobile cardiac telemetry to detect atrial fibrillation after stroke/TIA or in patients without symptoms at higher risk. Table 4 summarizes studies evaluating the use of mobile cardiac telemetry to detect atrial fibrillation after stroke/TIA.454647
Intermittent recording devices
Standalone handheld devices, such as MyDiagnostick (Applied Biomedical Systems BV, Maastricht, Netherlands) and Zenicor-ECG (Zenicor Medical System, Stockholm, Sweden), operate without additional hardware. MyDiagnostick is a stick with electrocardiography electrodes at both ends (fig 2). The device, when held, will generate a one minute single lead electrocardiographic recording. Then, an automated algorithm will determine whether the rhythm is an atrial fibrillation on the basis of analysis of R-R interval regularity, and the device will turn either red (indicating an atrial fibrillation rhythm) or green (indicating a non-atrial fibrillation rhythm). The device can also store up to 140 electrocardiogram strips, which can be uploaded to a computer for review.
Studies across diverse populations report the sensitivity of the MyDiagnostick automated algorithm for atrial fibrillation to be 80-100% and its specificity to be 93-99%, compared with interpretation of electrocardiograms by physicians (table 5).31575859 This device was used to screen for atrial fibrillation in a large cross sectional study in the Netherlands that included 3269 participants of an influenza vaccination program. In this study, new atrial fibrillation was identified in 1.1% of the screened population. Table 3 summarizes studies using MyDiagnostick to screen for atrial fibrillation in population at higher risk.3132
Zenicor-ECG has a display monitor at the center and two electrocardiography electrodes on each side (fig 3). Placing both thumbs on the electrodes generates a 30 second lead I electrocardiogram, which is transmitted to a web based central database for storage and processing with analysis and interpretation support.
Single lead electrocardiograms from Zenicor-ECG interpreted by a cardiologist had a sensitivity of 96% and a specificity of 92% in detecting atrial fibrillation compared with simultaneous 12 lead electrocardiograms interpreted by a cardiologist in a small cross sectional study.61 An automated interpretation of Zenicor-ECG had a sensitivity of 98% and a specificity of 88% in detecting atrial fibrillation compared with interpretation of the same electrocardiograms by specially trained nurses and physicians in a large retrospective analysis (table 5).60
The longitudinal performance of this device in detecting arrhythmia, including atrial fibrillation, compared with conventional methods was evaluated in a cohort of 95 patients (mean age 54.1 (SD 16.4) years; 44% male) referred for palpitations or dizziness who were monitored by both Holter monitor for 24 hours and intermittent Zenicor-ECG (twice daily and as needed for symptoms) over 28 days. The electrocardiograms in both groups were interpreted independently by a study investigator. The intermittent Zenicor-ECG strategy detected significantly more arrhythmias (14%, including nine patients with atrial fibrillation) than did the Holter monitor (3%, including two patients with atrial fibrillation) (P=0.009).69
The efficacy and cost effectiveness of an atrial fibrillation screening strategy with the Zenicor-ECG device in people aged 75-76 years is being evaluated in an ongoing RCT (the STROKESTOP study). An analysis of the baseline information of 7173 participants in the screening arm of this study showed that new atrial fibrillation was detected in 3.0% (95% confidence interval 2.7% to 3.5%) over two weeks of intermittent Zenicor-ECG monitor use (twice daily and as needed for symptoms), whereas only 0.5% were detected on their first electrocardiogram34). Studies using Zenicor-ECG to screen for atrial fibrillation in populations at higher risk and after stroke/TIA are summarized in table 3 and table 4, respectively.4849
Add-on accessory devices to mobile electronics, most commonly smartphones, allow them to record electrocardiograms. AliveCor KardiaMobile (AliveCor Inc, Mountain View, CA, USA) is an example of a smartphone based device. It is a thin card with two electrocardiography electrodes (fig 4). A brief (30 seconds to 5 minutes) lead I electrocardiographic tracing is generated by placing fingers of the right and left hands on the electrodes. This tracing is sent wirelessly to the smartphone for display in real time and analysis by an automated algorithm incorporated in the phone’s application. The algorithm can detect possible atrial fibrillation on the basis of criteria for the presence/absence of a P wave and the regularity of the R–R interval. The tracing is also stored for subsequent review.
The sensitivity and specificity of the current commercial automated algorithm (version 2.2.2) of this device to detect atrial fibrillation were reported to be close to 70% and above 99%, respectively, compared with the same single lead electrocardiogram interpreted by cardiologists in two large cross sectional studies of patients at higher risk for atrial fibrillation (table 5).3763Table 5 summarizes studies evaluating the sensitivity and specificity of other versions of the device’s automated algorithms, as well as the interpretation of the device’s single lead electrocardiograms by physicians.38576667
The performance of this device in screening for atrial fibrillation in populations at higher risk was evaluated in the SEARCH-AF study, a cross sectional study of opportunistic community based screening that enrolled 1000 pharmacy customers aged 65 or over (table 3).38 The participants in this study were screened with AliveCor KardiaMobile single lead electrocardiography, which was then interpreted by a cardiologist. Newly identified atrial fibrillation was found in 1.5% (95% confidence interval 0.8% to 2.5%) of the population, and all had CHA2DS2-VASc scores of 2 or above.
This device was also evaluated in the REHEARSE-AF Study, an RCT of screening for atrial fibrillation that enrolled 1001 participants aged 65 or over with a CHA2DS2-VASc score of at least 2 who were free from known atrial fibrillation at baseline (table 3).7 The participants were randomized to AliveCor KardiaMobile single lead electrocardiography screening twice weekly over 12 months (plus additional electrocardiography if symptomatic) or routine care. In this study, 19/500 patients in the screening group were diagnosed as having atrial fibrillation over the 12 month study period compared with 5/501 patients in the routine care group (hazard ratio 3.9; P=0.007). There was no statistically significant difference in the number of strokes or transient ischemic attacks, but the study was not powered to detect differences in these outcomes.
Kardia Band (AliveCor, Mountain View, CA, USA) is an accessory device for the Apple smartwatch (fig 5). This device records a 30 second, single lead electrocardiogram while the patient is wearing the Apple Watch on one wrist and putting the opposite thumb on the Kardia Band. This tracing will be stored and analyzed by the same automated algorithm as AliveCor KardiaMobile, installed in the Apple Watch to identify possible atrial fibrillation. Compared with 12 lead electrocardiograms interpreted by electrophysiologists, the automated algorithm was reported to have a sensitivity and specificity of 93% and 84%, respectively, in a small cross sectional study of patients with atrial fibrillation undergoing cardioversion (table 5).68 It should be noted that about one third of the electrocardiograms were uninterpretable and excluded from this analysis.
Automatic oscillometric blood pressure monitors with algorithms for detecting atrial fibrillation detect atrial fibrillation by checking the regularity of pulses.70 The Microlife BP monitor (Microlife USA, Dunedin, FL) is one of the best studied devices. This device identifies possible atrial fibrillation when at least two of three measurements show pulse irregularity. The reported sensitivity and specificity were approximately 80-100% and 89-99%, respectively (table 6).3662637172737475
In a large atrial fibrillation screening study that included 5969 patients with hypertension, diabetes mellitus, and/or age 65 or over in a primary care clinic, the Microlife WatchBP monitor identified atrial fibrillation in 58 of 72 patients with atrial fibrillation and produced false positive results in 79 patients (table 3).36Table 3 summarizes smaller studies evaluating the feasibility of this device as a screening tool for atrial fibrillation.3940
Photoplethysmography is an optical technology that detects atrial fibrillation by measuring and analyzing a peripheral pulse waveform. This technology measures the pulse waveform by detecting changes in the light intensity, which reflects the tissue blood volume, of a skin surface such as the fingertip, earlobe, or face.787980 The generated pulse waveform is subsequently analyzed by an automated algorithm to detect atrial fibrillation. This technology has been applied to use with smartphones; most commonly, the smartphone’s camera is used to measure fingertip pulse waveform.
In a large atrial fibrillation screening study that included 1013 patients with hypertension, diabetes mellitus, and/or age 65 or over in a primary care clinic, smartphone based photoplethysmography with the Cardiio Rhythm smartphone application (Cardiio Inc, Cambridge, MA, USA; fig 6) was found to have a sensitivity of 93% and a specificity of 98% to detect atrial fibrillation compared with a single lead electrocardiogram interpreted by two cardiologists (table 6).37 In this study, the Cardiio Rhythm application identified atrial fibrillation in 26/28 atrial fibrillation patients and produced 23 false positive results (table 3). Table 6 summarizes smaller studies from selected populations evaluating the sensitivity and specificity of smartphone based photoplethysmography.7677
Photoplethysmography technology has also been applied to smartwatches to measure heart rate. However, no published data evaluating smartwatch based photoplethysmography for detection of atrial fibrillation exist. The Apple Heart Study (NCT03335800) and the GARMIN AF study (NCT03566836) are evaluating smartwatch based photoplethysmography for detection of atrial fibrillation, and both are expected to be completed in January 2019.
Invasive monitoring devices
Implantable loop recorders (ILRs) are implanted in subcutaneous tissue. Similar to traditional external loop recorders, these devices continuously monitor cardiac electrical activity but record an electrocardiogram only shortly before and after activation by either the patient or an automated algorithm. The total period of monitoring lasts several years. In the past, ILRs have primarily been used for evaluation of symptoms suggestive of arrhythmia, such as palpitations and syncope, that occurs infrequently. Older generations of ILRs did not have dedicated automatic atrial fibrillation detection algorithms and thus reported unexpectedly low rates of detection of atrial fibrillation in a cohort study of patients with cryptogenic stroke.81
The newer generations of ILRs have dedicated atrial fibrillation detection algorithms, and interest in their use for evaluation of atrial fibrillation is increasing, especially after cryptogenic stroke. Several approved ILR devices are available (fig 7 and fig 8).82 In general, most automated algorithms use variability in the R-R interval to detect atrial fibrillation.838485 Some newer devices’ algorithms also incorporate the presence/absence of a P wave.86Table 7 summarizes studies evaluating the sensitivity and specificity of ILRs.83848587
The incidence of new atrial fibrillation detected by ILRs in cohort studies of patients at risk for stroke and patients with cryptogenic stroke or TIA varies widely,414243515253545556 which is likely because of differences in study population, duration of follow-up, definition of atrial fibrillation, and ILRs used (table 3 and table 4). The Cryptogenic Stroke and Underlying Atrial Fibrillation (CRYSTAL AF) trial, which enrolled 441 patients aged 40 or over with cryptogenic stroke or TIA, evaluated the Reveal XT (Medtronic, Minneapolis, MN, USA) in an RCT. In this study, atrial fibrillation for more than 30 seconds was detected in 9% of patients in the ILR group at six months compared with 1% of patients in the control group receiving routine care (P<0.001).50 However, this study was not powered to evaluate the effect on clinical outcomes such as recurrent stroke.
Validity of studies evaluating newer detection methods
Although studies have reported excellent sensitivity and specificity for many newer devices, these results need to be interpreted cautiously for several reasons. Firstly, studies evaluating the accuracy of newer devices are generally done in an ideal setting, which can be difficult to achieve in a real world situation. For example, electrocardiogram acquisition for handheld devices is usually done under supervision to ensure the quality of the tracings. Patients who cannot operate the devices and/or have uninterpretable recordings are often excluded from the study’s analysis. Secondly, automated algorithms for detection of atrial fibrillation in some devices have been modified from the ones that were evaluated in some published studies. For instance, the commercial automated algorithm of AliveCor KardiaMobile is biased toward higher specificity compared with the research algorithm used in some published studies.64 Lastly, certain devices were evaluated in selected patient populations that may not reflect the general patient population in which the devices might be used in the real world. These factors need to be considered when interpreting the studies evaluating newer devices.
Significance and uncertainty of duration of atrial fibrillation
Historical studies that associated atrial fibrillation with increased risk of stroke and showed the benefit of anticoagulation in patients with atrial fibrillation primarily documented atrial fibrillation by intermittent detection methods such as resting 12 lead electrocardiography.8889909192 Thus, only people with atrial fibrillation of substantial duration were likely to be included in these studies. In recent years, the development of devices that continuously monitor cardiac rhythm allows the detection of shorter episodes of atrial fibrillation. However, whether short episodes of atrial fibrillation detected by continuous monitoring devices are associated with the same level of thromboembolic risk as atrial fibrillation reported in historical studies remains uncertain.
Observational cohort studies of patients with a pacemaker or implantable cardioverter-defibrillator have associated short atrial fibrillation of different minimal durations (from >5 minutes to >5.5 hours) with increased thromboembolic risk.939495 However, the reported risk in these studies tends to be lower than expected from the mean CHADS2 scores of these cohorts. In addition, in a subgroup analysis of the Asymptomatic Atrial Fibrillation and Stroke Evaluation in Pacemaker Patients and the Atrial Fibrillation Reduction Atrial Pacing Trial (ASSERT), only atrial fibrillation of more than 24 hours was reported to be associated with an increased risk of ischemic stroke or systemic embolism, whereas atrial fibrillation with a duration between six minutes and 24 hours was not.96
In addition, the role of these short episodes of atrial fibrillation in the pathogenesis of stroke remains uncertain. An analysis of the ASSERT study showed that 26/51 patients who had stroke or systemic embolism during the follow-up had atrial fibrillation of more than six minutes. Among these 26 patients, only four had atrial fibrillation within 30 days before the embolic events. Moreover, eight of these 26 patients had atrial fibrillation only after the embolic events.97 Also, whether treating short episodes of atrial fibrillation with anticoagulation would decrease the thromboembolic risk remains unknown. This is being evaluated in three major clinical trials—Apixaban for the Reduction of Thrombo-Embolism in Patients with Device-Detected Sub-Clinical Atrial Fibrillation (ARTESiA: NCT01938248), Non-Vitamin K Antagonist Oral Anticoagulants in Patients With Atrial High Rate Episodes (NOAH: NCT02618577), and Atrial Fibrillation Detected by Continuous ECG Monitoring (LOOP: NCT02036450).
Applications of newer detection methods
Screening for atrial fibrillation in selected asymptomatic populations
Although screening for atrial fibrillation in asymptomatic populations has not been definitely shown to reduce the risk of stroke, data from retrospective cohort studies suggest that patients with asymptomatic atrial fibrillation incidentally detected by conventional detection methods are, at least, at similar risk of stroke as those with symptomatic atrial fibrillation.498 In addition, this risk of stroke could be reduced by anticoagulation to the same degree.9899 Thus, screening for atrial fibrillation in a population at higher risk seems to be a logical approach to reduce stroke and has been advocated by many experts.100101
The Screening for Atrial Fibrillation in the Elderly (SAFE) study was one of the first RCTs to evaluate atrial fibrillation screening. This study included 14 802 patients aged 65 or over in 50 primary care practices. It showed that both opportunistic screening for atrial fibrillation in patients attending a routine clinic visit by pulse palpation, followed by 12 lead electrocardiography when the pulse was irregular, and systematic screening by inviting the whole patient population to have 12 lead electrocardiography were more effective than routine care in detecting atrial fibrillation.6 Both screening strategies detected a similar number of patients with atrial fibrillation, but opportunistic screening was more cost effective.102 Thus, this method is generally preferred.
However, this approach has limitations. Although pulse palpation was reported to have a reasonable sensitivity (0.92, 95% confidence interval 0.85 to 0.96), its specificity was less than ideal (0.82, 0.76 to 0.88).103 When used as a screening tool, this method could generate substantial false positive results. This is further complicated by the logistic challenge of obtaining electrocardiograms and their interpretation in a primary care setting, which also limits the use of conventional electrocardiography as an initial screening tool.
Automatic blood pressure monitors with atrial fibrillation detection algorithms, smartphone based photoplethysmography, and standalone handheld and smartphone based electrocardiographic recording devices may improve the feasibility and accuracy of atrial fibrillation screening. These devices are generally easy to use and have excellent sensitivity and specificity for detecting atrial fibrillation. Electrocardiographic confirmation is still needed when the automatic blood pressure monitors with atrial fibrillation detection algorithms or smartphone based photoplethysmography detect possible atrial fibrillation. If a substantial delay in acquiring electrocardiographic confirmation occurs, paroxysmal atrial fibrillation may be missed. The standalone handheld and smartphone based recording devices generate electrocardiographic tracings and thus eliminate the need to perform additional confirmatory tests. In addition, the abnormal electrocardiographic tracings generated can be readily accessed and reviewed by experienced physicians. Several studies have shown the efficacy of these devices for atrial fibrillation screening in populations at risk (table 3).731323334353637383940
Modeled cost effectiveness analyses predict that atrial fibrillation screening with standalone handheld and smartphone based devices in populations at higher risk and subsequent anticoagulation treatment in people identified as having atrial fibrillation would be cost effective or even cost reducing in certain scenarios.38104105 However, these cost effectiveness analyses have certain limitations. The effect of atrial fibrillation screening and subsequent anticoagulation on clinical outcomes, such as stroke and bleeding, were generally extrapolated from outside studies as these clinical outcomes have not been adequately evaluated by the studies of these newer devices. The costs associated with atrial fibrillation screening, anticoagulation, and treatment of different clinical outcomes are generally based on local costs and might be different in other healthcare systems.
The Zio Patch and various ILRs have been shown to be feasible for use in screening for atrial fibrillation in higher risk populations (table 3).30414243 However, because of their single use, associated cost, and invasiveness (in the case of ILR), and the unclear clinical significance of short episodes of atrial fibrillation, they are not likely to be widely used for atrial fibrillation screening in asymptomatic populations. Figure 9 shows our proposed algorithm for screening in patients at risk for but with no symptoms of atrial fibrillation.
Evaluation for atrial fibrillation after stroke and TIA
Strokes are estimated to be the initial manifestation of atrial fibrillation at a rate of 2-5/10 000 person years.106 In addition, patients with atrial fibrillation and previous stroke or TIA are at high risk (averaging 10% per year) of developing recurrent stroke107 Thus, a certain period of cardiac monitoring after stroke or TIA to look for atrial fibrillation seems reasonable. Compared with a shorter duration of monitoring, prolonged monitoring for seven days or more was shown to detect more atrial fibrillation in RCTs evaluating patients with stroke or TIA,50108109110111 but no difference in recurrent stroke was shown. However, these studies were not adequately powered to detect differences in clinical outcomes. Thus, the optimal duration of cardiac rhythm monitoring after stroke and TIA remains unknown. Given the lack of definite evidence, the decision to recommend that a patient should undergo prolonged cardiac rhythm monitoring beyond the first few days and the duration of monitoring need to remain individualized. In our opinion, one should consider prolonged cardiac monitoring in patients with cryptogenic stroke and at increased risk of having atrial fibrillation on the basis of risk factors. Also, one should consider whether the detection of atrial fibrillation would alter the downstream management of the patient. For example, the utility of detecting atrial fibrillation is likely to be low in patients who are already taking anticoagulants for other reasons.
The optimal modality for cardiac rhythm monitoring in patients needing prolonged monitoring is influenced by factors such as the planned duration of monitoring as well as the patient’s clinical characteristics and preference. In general, the devices that monitor cardiac rhythm continuously are likely to detect more atrial fibrillation than the ones that monitor intermittently. However, the clinical significance of short episodes of atrial fibrillation that can be detected by the continuous monitoring remains uncertain.
For patients who need prolonged monitoring for less than a month, an external loop recorder with an automatic trigger function has traditionally been used, but a mobile cardiac telemetry device is a viable option. Although data specific for patients with cryptogenic stroke or TIA are still lacking, mobile cardiac telemetry was reported to detect more atrial fibrillation and atrial flutter than an external loop recorder in patients evaluated for arrhythmia.29 Also, the mobile cardiac telemetry allows minimal delay between acquisition and interpretation of the electrocardiogram. Thus, it may be more suitable in patients who may have coexisting malignant arrhythmias.
ILRs seem to be the most viable modality for patients who need monitoring for longer than a month, as external devices, especially those with leads, are generally not well tolerated. Selective use of ILRs may be reasonable in patients with cryptogenic stroke at moderate to high risk of atrial fibrillation. As discussed above, the Reveal XT was shown to detect more atrial fibrillation than routine care in the CRYSTAL AF trial.50 Also, it was shown to detect more atrial fibrillation than seven day Holter monitoring in a prospective cohort study of patients with cryptogenic stroke.56 Although a modeled cost effective analysis predicted that using ILR monitoring after cryptogenic stroke and subsequent anticoagulation treatment in patients identified as having atrial fibrillation were cost effective,112 this analysis has certain limitations similar to the cost effective analyses of standalone handheld and smartphone based electrocardiography recording devices. Also, which groups of patients, if any, should be monitored by ILR early after stroke and TIA or after negative prolonged monitoring by non-invasive methods remains largely unknown.
No study has directly compared the cutaneous patch monitor and standalone handheld or smartphone based electrocardiography recording devices against an external loop recorder or other methods of prolonged (≥7 days) electrocardiographic monitoring. The Zio Patch monitor used for up to 14 days was reported to detect a higher number of arrhythmias, including atrial fibrillation, over the total wear time compared with a conventional Holter monitor in cohort studies of patients evaluated for arrhythmia,2728 and it is generally better tolerated by patients.28 Intermittent Zenicor-ECG for 30 days seems to be able to detect atrial fibrillation in patients with recent ischemic stroke or TIA with results at least comparable to, if not better than, conventional Holter monitoring in prospective cohort studies.4849 The longer duration of monitoring of these newer devices likely underlies their higher yield compared with conventional Holter monitors. Figure 10 shows our proposed algorithm for evaluation of atrial fibrillation after stroke or TIA.
In 2010 the European Society of Cardiology (ESC) recommended opportunistic screening for atrial fibrillation by pulse check in patients aged 65 or over, followed by electrocardiography in case of irregularity (class I recommendation).113 This recommendation is largely based on the result of the SAFE study, which showed the efficacy of both opportunistic screening during routine clinic visits (by pulse palpation, followed by 12 lead electrocardiography when the pulse was irregular) and systematic screening of the whole population with 12 lead electrocardiography in patients aged 65 or over.6 However, as the opportunistic screening was more cost effective,102 this method was recommended by the guidelines. In 2016 the ESC guideline expanded the screening method to include electrocardiography rhythm strip (class I).114 It also recommended that systematic electrocardiographic screening may be considered to detect atrial fibrillation in patients aged over 75 or those at high risk of stroke (class IIb). This is largely based on the reported efficacy and modeled cost effectiveness of the Zenicor-ECG device in screening for atrial fibrillation in at risk populations.34104 The American College of Cardiology/American Heart Association (AHA)/Heart Rhythm Society guideline made no specific recommendation about screening for atrial fibrillation.115
Comprehensive physiologic monitoring, including cardiac monitoring, in specialized stroke units has been shown to improve clinical outcome in patients admitted for acute ischemic stroke.116117 At least a short period of cardiac monitoring after ischemic stroke or TIA is consistently recommended by professional societies’ guidelines. The AHA/American Stroke Association (ASA) guideline recommends cardiac monitoring for at least 24 hours in patients with TIA or ischemic stroke (class I).118 The minimum duration of 24 hours is largely based on general consensus and limited data. The ESC recommends screening for atrial fibrillation by short term electrocardiographic recording followed by continuous electrocardiographic monitoring for at least 72 hours in patients with TIA or ischemic stroke (class I).114 The minimum duration of 72 hours is based on the higher yield for detection of atrial fibrillation with a longer period of monitoring in prospective cohort studies of patients with stroke and TIA.119120
Although prolonged cardiac monitoring has been shown to detect more atrial fibrillation in patients with ischemic stroke,50108109110111 the recommendations for its use are less certain in the guidelines as its clinical benefit and cost effectiveness remain uncertain. The AHA/ASA guideline stated that the clinical benefit of prolonged cardiac monitoring to detect atrial fibrillation after acute ischemic stroke is uncertain (class IIb).118 However, prolonged cardiac monitoring may be reasonable in some patients with ischemic stroke, although the effect on outcomes is uncertain (class IIb).118 The ESC guideline recommended that additional electrocardiographic monitoring by long term non-invasive monitors or ILRs should be considered to document silent atrial fibrillation in stroke patients (class IIa).114
Mechanocardiography uses accelerometers and gyroscopes to sense mechanical activity of the heart. In a proof of concept study, the accuracy of this technology to detect atrial fibrillation by using a smartphone’s built-in accelerometer and gyroscope sensors was evaluated in 16 patients with atrial fibrillation and 23 healthy controls. The patient is advised to lie down in a supine position, and the smartphone is placed on the patient’s chest to detect atrial fibrillation by measuring the regularity of cardiac mechanical activity. The device was reported to have a sensitivity of 94% and a specificity of 100% for detection of atrial fibrillation compared with simultaneous electrocardiography in this study.121 A larger case-control study, which included 150 consecutive patients in atrial fibrillation and 150 age and sex matched patients in sinus rhythm, reported the sensitivity and specificity of this device for detection of atrial fibrillation to be 95% (95% confidence interval 91% to 98%) and 96% (92% to 99%), respectively, compared with simultaneous five lead telemetry electrocardiographic recording interpreted by two cardiologists.122
Upcoming clinical trials
Newer technologies have enhanced our ability to detect atrial fibrillation. However, whether their application will improve clinical outcomes remains uncertain. The STROKESTOP study (NCT01593553), LOOP study (NCT02036450), STROKESTOP II study (NCT02743416), and SAFE-PE Study (NCT03274401) are RCTs that will evaluate whether the applications of newer devices in selected clinical scenarios will improve clinical outcomes. Table 8 summarizes upcoming clinical trials involving newer technologies to detect atrial fibrillation.
Atrial fibrillation remains underdiagnosed in contemporary clinical practice using conventional detection methods, resulting in missed opportunities to implement appropriate treatment. Several recently developed technologies have been shown to enhance our ability to diagnose atrial fibrillation. These devices have different advantages and limitations primarily relating to the atrial fibrillation detection method of the devices as well as their accuracy, duration of monitoring, ease of use, and cost. The uncertainties about the significance of short episodes of atrial fibrillation detected by some of the newer technologies and the optimal management for these episodes complicate decisions about treatment.
Experts and professional societies’ guidelines advocate screening for atrial fibrillation in populations at increased risk, as people with undiagnosed asymptomatic atrial fibrillation remain at increased risk of stroke. Opportunistic screening by pulse palpation during a clinic visit, followed by 12 lead electrocardiography in case of irregular pulses, is effective in detecting atrial fibrillation but has several major limitations. Screening for atrial fibrillation with automatic blood pressure monitors with atrial fibrillation detection algorithms, smartphone based photoplethysmography, and standalone handheld or smartphone based electrocardiography recording devices can overcome some of these limitations and may improve the feasibility and accuracy of screening for atrial fibrillation.
Patients with atrial fibrillation and previous stroke are at high risk of developing recurrent stroke. Thus, evaluation for atrial fibrillation with at least 24-72 hours of cardiac rhythm monitoring in these patients is consistently recommended by professional society guidelines. Although longer duration of monitoring has been shown to detect more atrial fibrillation in these patients, the optimal duration of monitoring after stroke or TIA remains unknown. Selecting the modality and duration of monitoring for atrial fibrillation in patients with stroke or TIA must be individualized at this point. For patients who need prolonged monitoring, external loop recorders with automatic trigger functions are conventionally used but mobile cardiac telemetry and ILRs are viable options.
What is the minimum duration of episodes of atrial fibrillation associated with significant risk of stroke and thromboembolic risk, and does anticoagulation improve clinical outcomes in such cases?
What is the optimal duration of monitoring for people at risk of atrial fibrillation?
What is the most cost efficient means of detecting atrial fibrillation and preventing strokes?
Does screening for atrial fibrillation in people without symptoms but at increased risk of stroke improve clinical outcomes, and what is the most optimal strategy for screening?
Which patients should be evaluated by prolonged monitoring for atrial fibrillation after stroke or transient ischemic attack, and what is the optimal duration and modality?
Series explanation: State of the Art Reviews are commissioned on the basis of their relevance to academics and specialists in the US and internationally. For this reason they are written predominantly by US authors
Contributors: NZ conducted the search, and both authors selected the studies for inclusion. NZ drafted the manuscript, and MSL edited and approved the final version.
Competing interests: We have read and understood BMJ policy on declaration of interests and declare the following interests: none.
Provenance and peer review: Commissioned; externally peer reviewed.
Patient involvement: No patients were asked for input in the creation of this article.