Circulation Journal
Online ISSN : 1347-4820
Print ISSN : 1346-9843
ISSN-L : 1346-9843

This article has now been updated. Please use the final version.

Evaluation of the Feasibility and Efficacy of a Novel Device for Screening Silent Atrial Fibrillation (MYBEAT Trial)
Yousaku OkuboTakehito TokuyamaSho OkamuraYoshihiro IkeuchiShunsuke MiyauchiYukiko Nakanofor the MYBEAT Trial Investigators
Author information
JOURNAL OPEN ACCESS FULL-TEXT HTML Advance online publication
Supplementary material

Article ID: CJ-20-1061

Details
Abstract

Background: myBeat is a novel cutaneous patch device that continuously records electrocardiography and automatically detects atrial fibrillation (AF) by using a new algorithm based on RR intervals. We aimed to test the diagnostic ability of this novel device for screening silent AF in asymptomatic patients.

Methods and Results: A multicenter randomized prospective clinical study was performed. To be eligible for inclusion in the study, patients had to be ≥65 years of age and have ≥1 of the following risk factors: hypertension, diabetes, heart failure, ischemic heart disease, stroke, and transient ischemic attack. Patients with prior AF, an implantable pacemaker, and previous palpitation or syncope were excluded. The 300 participants were divided into 2 groups, those using myBeat (n=150) or those undergoing 24-h Holter monitoring (control group; n=150), for AF screening. The rate of AF detection was significantly higher in the myBeat than control group (16 [10.7%] vs. 7 [4.7%], respectively; P=0.04). Multivariable logistic regression analysis revealed that prior heart failure was an independent predictor of silent AF (odds ratio 12.07; 95% confidence interval 1.67–86.27; P=0.01). A 7.7-fold difference in silent AF was found between subjects with CHA2DS2-VASc scores of 1 point and those with scores ≥4 points.

Conclusions: The novel patch device using an original algorithm was beneficial for screening of silent AF.

Atrial fibrillation (AF) is the most common supraventricular arrhythmia; it is associated with high mortality and can lead to adverse cardiovascular events. The annual incidence of stroke in individuals with non-valvular AF is approximately 5% per year, which is approximately 2- to 7-fold higher than the incidence of stroke in those without AF.1,2 Appropriate interventions, including anticoagulation therapy, can reduce the risk of stroke by approximately two-thirds.3 Therefore, stroke risk stratification and criteria for anticoagulation therapy for AF are based on risk scores (e.g., CHADS2 and CHA2DS2-VASc scores) that have been generally accepted in recent years.4,5

Editorial p ????

However, 30–50% of patients with AF are asymptomatic, which makes early detection of AF challenging. It is not unusual for the first manifestation of AF to be cerebral infarction.6,7 Moreover, the proportion of patients experiencing adverse cardiovascular events and dying was reported to be 3-fold higher in patients with asymptomatic than symptomatic AF.8

In a previous study, over a 3-month follow-up period, Healey et al found supraventricular tachycardia in 10% of 2,580 patients who were implanted with a pacemaker or defibrillator without any prior history of AF; the median time to detection of the first asymptomatic arrhythmia was 36 days after pacemaker implantation.6 That study suggested that early detection of AF may be difficult even when Holter monitoring is used once.6

myBeat (Union Tool Co., Niigata, Japan) is a novel patch-type, non-invasive, multiuse, continuously recording cardiac rhythm monitoring device (Figure 1). myBeat records electrocardiography (ECG) results for up to 5 days and can automatically analyze heartbeat fluctuations using an original algorithm based on RR intervals (RRI). Although myBeat can record single-lead ECG results, only the RRI is acquired to reduce the amount of data. In a previous pilot study, Matsui et al validated the novel patch-type device (myBeat) in 129 participants with normal sinus rhythm (NSR) and 108 patients with persistent AF.9 Matsui et al reported that differences in RRI normalized by RRI before and after the indexing beats (normalized DRs) were distributed within a narrower range in subjects with NSR. However, in subjects with AF, the normalized DRs were distributed over a wide range.9 Using the difference in normalized DRs, Matsui et al discriminated between NSR and AF with a high sensitivity and specificity. Thus, that study confirmed the ability of myBeat to identify AF automatically using an original algorithm based on the RRI.9

Figure 1.

The myBeat is placed over the patient’s left pectoral region. (Images courtesy of Union Tool Co., Niigata, Japan.)

The aim of the present study was to test the diagnostic ability of this novel device in screening for silent AF in asymptomatic patients without prior AF.

Methods

This study was a multicenter prospective randomized clinical trial designed to evaluate the efficacy of an original algorithm to automatically detect AF using the myBeat compared to a conventional screening method (single 24-h Holter monitoring). The study enrolled 358 outpatients who underwent treatment for various diseases related to the development of AF (e.g., hypertension, diabetes, sleep apnea syndrome, and chronic kidney disease) at 16 medical facilities in Hiroshima between October 2017 and December 2018 and who agreed to take part in the MYBEAT trial. The inclusion criteria were age ≥65 years and ≥1 of the following risk factors: hypertension, diabetes, heart failure (HF), ischemic heart disease, stroke, or transient ischemic attack (TIA). Patients with prior AF, an implantable pacemaker or defibrillator, and previous palpitation or syncope were excluded (Figure 2). Ultimately, 300 participants were enrolled and randomized into 2 groups to screen for AF. One group was given the myBeat wearable device (n=150) and the other group underwent 24-h Holter monitoring (control group; n=150). In the myBeat group, most participants continued to be monitored for 5 days.

Figure 2.

Flow diagram of the study participants. In all, 358 outpatients who underwent treatment for various diseases related to the development of AF (e.g., hypertension, diabetes, sleep apnea syndrome, and chronic kidney disease) at 16 medical facilities in Hiroshima between October 2017 and December 2018 and agreed to participate in the myBeat trial were enrolled in the study. The inclusion criteria were age ≥65 years and ≥1 of the following risk factors: hypertension, diabetes, heart failure, ischemic heart disease, stroke, or transient ischemic attack. Patients with prior atrial fibrillation (AF), an implantable pacemaker or defibrillator, and previous palpitation or syncope were excluded. Ultimately, 300 participants were enrolled and randomized into 2 groups for AF screening using the myBeat wearable device (n=150) or 24-h Holter monitoring (control group; n=150).

The myBeat algorithm based on the RRI is useful for detecting AF. This algorithm was constructed in the previous pilot study.9 In that study, Matsui et al noticed that the differences in RRI normalized by RRI before (RRIn-1) and after (RRIn) the indexing beats (normalized DRs) were distributed within a narrow range in subjects with NSR compared with AF patients.9 Matsui et al formulated the differences using the following equation:

Normalized DR = (RRIn-1 − RRIn) / (RRIn-1 + RRIn) / 2

Matsui et al also found that as the number of normalized DRs falling outside the range of NSR increased, the probability of AF increased.9 However, because the number of normalized DRs falling outside the range of NSR depends on the sampling time as well as fluctuations in RRI, Matsui et al determined the appropriate the sampling number (s) and the number of normalized DRs falling outside the range of NSR (i) by calculating sensitivity and specificity as follows:

Sensitivity = sCi × PAFi × (1.000 − PAF)S−i

Specificity = sCi × PNSRi × (1.000 − PNSR)S−i

where sCi is the number of ways to choose a sample of (i) from a set of (s), PAF is the percentage of normalized DRs divided by 100 (i.e., the probability of the number of normalized DRs falling outside the NSR range in the AF group), and PNSR is the percentage of normalized DRs divided by 100 (i.e., the probability of the number of normalized DRs falling within the NSR range in the AF group).9 When Matsui et al set the number of normalized DRs falling outside the range of NSR to 7 and the sampling number to 20, the algorithm showed good discrimination for the diagnosis of AF or NSR, with a sensitivity of 98% and specificity of 98.2%.9 However, that study was specifically designed to detect AF under certain conditions. Therefore, in the present study we performed 2 weeks of Holter monitoring in patients diagnosed as having AF by using myBeat to investigate various patterns of RRI for other arrhythmias and to verify the accuracy of myBeat.

This study was approved by the Institutional Ethics Committee of the Graduate School of Biomedical Science at Hiroshima University (Approval no. E-844, registered September 30, 2017) and was conducted in accordance with the tenets of the Declaration of Helsinki. All participants provided written informed consent.

Definitions

AF was defined as an irregular rhythm without P waves recorded for at least 30 s by the novel patch-based device (myBeat) and within 2 weeks by the Holter ECG. All diagnoses of new AF were confirmed and reviewed by an experienced cardiologist. In this study, we compared the detection rate of AF between the myBeat and control groups and investigated significant predictors of the occurrence of silent AF.

Hypertension was defined as a systolic blood pressure (BP) ≥130 mmHg or diastolic BP ≥80 mmHg, or the use of antihypertensive drugs. Diabetes mellitus (DM) was defined as HbA1c ≥6.5%, fasting plasma glucose concentrations ≥126 mg/dL, or a medical history of DM. Dyslipidemia was defined as low-density lipoprotein cholesterol concentrations ≥140 mg/dL, high-density lipoprotein cholesterol concentrations <40 mg/dL, triglycerides concentrations ≥150 mg/dL, or the use of lipid-lowering drugs. Chronic kidney disease was defined as an estimated glomerular filtration rate <60 mL/min/1.73 m2. Subjects with HF were defined as those with a history of at least 1 hospitalization for decompensated HF or using diuretics for HF. Ischemic heart disease was defined as a history of myocardial infarction, a medical history of angina, or the use of nitroglycerin.

Statistical Analysis

Continuous variables are summarized as the mean±SD or the median with interquartile range (IQR). Categorical variables are presented as proportions. The significance of between-group differences was analyzed using Fisher’s exact test for categorical variables or the Mann-Whitney U-test for continuous variables. Multivariate logistic regression analysis was used to assess independent predictors of the occurrence of AF, with results presented as odds ratios (ORs) and 95% confidence intervals (CIs). Variables that were statistically significant in the univariate analysis were included in the multivariate models.

All statistical analyses were performed using JMP version 14.0 (SAS Institute, Cary, NC, USA). Two-tailed P<0.05 was considered significant.

Results

In all, 300 patients were enrolled in this study between October 2017 and December 2018. The baseline clinical characteristics of the patients in both groups are presented in Table 1. We found no significant differences in baseline patient characteristics between the 2 groups. The most common comorbidity in the myBeat and control groups was hypertension (90 [60.0%] vs. 94 [62.6%], respectively; P=0.63), followed by dyslipidemia (69 [46.0%] vs. 64 [42.7%], respectively; P=0.56) and DM (34 [22.7%] vs. 27 [18.0%], respectively; P=0.31). Five percent of the participants in both groups had a history of cerebral infarction, but none had been diagnosed as having AF.

Table 1. Baseline Characteristics of the Study Subjects
  Overall
(n=300)
myBeat group
(n=150)
Control group
(n=150)
P value
Age (years) 73.2±6.5 73.9±6.8 72.9±5.9 0.15
Male sex 139 (46.0) 65 (43.3) 74 (49.3) 0.29
BMI (kg/m2) 22.9±2.8 22.6±2.9 23.1±2.7 0.11
Medical history
 Hypertension 184 (61.3) 90 (60.0) 94 (62.7) 0.63
 Dyslipidemia 133 (44.3) 69 (46.0) 64 (42.7) 0.56
 Diabetes 61 (20.3) 34 (22.7) 27 (18.0) 0.31
 Heart failure 5 (1.6) 3 (2.0) 2 (1.3) 0.25
 Ischemic heart disease 20 (6.7) 13 (8.7) 7 (4.7) 0.16
 Stroke/TIA 16 (5.3) 8 (5.3) 8 (5.3) 1.00
 Chronic kidney disease 24 (9.0) 11 (7.3) 13 (8.7) 0.57
 Thyroid disease 20 (6.7) 10 (6.7) 10 (6.7) 1.00
 Sleep apnea syndrome 14 (4.6) 5 (3.3) 9 (6.0) 0.27
Medications
 DOAC 7 (2.3) 5 (3.3) 2 (1.3) 0.24
 ACEI/ARB 110 (36.6) 52 (34.7) 58 (38.6) 0.56
 Calcium channel blocker 173 (57.7) 86 (57.3) 87 (58.0) 0.83
 β-blocker 31 (10.3) 18 (12.0) 13 (8.6) 0.34
CHA2DS2-VASc scores
 Mean±SD 2.9±1.2 3.0±1.2 2.9±1.2 0.35
 CHA2DS2-VASc score 1 38 (12.7) 16 (10.7) 22 (14.6)  
 CHA2DS2-VASc score 2 62 (20.7) 38 (25.3) 24 (16.0)  
 CHA2DS2-VASc score 3 113 (37.7) 57 (38.0) 56 (37.3) 0.27
 CHA2DS2-VASc score 4 57 (19.0) 24 (16.0) 33 (22.0)  
 CHA2DS2-VASc score >5 30 (10.0) 19 (12.6) 11 (7.3)  

Unless indicated otherwise, data are presented as the mean±SD or as n (%). ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker; BMI, body mass index; DOAC, direct oral anticoagulants; TIA, transient ischemic attack.

In the myBeat group, the median device wear time was 4.95±0.03 days and 98.6% of participants (n=148) were monitored for 5 days. Two patients dropped out on the second day because of itchiness requiring no medical intervention. In the present study, newly detected AF occurred in 23 patients (7.7%). The rate of AF detection was significantly higher in the myBeat than control group (16 [10.7%] vs. 7 [4.7%], respectively; P=0.04). A representative case is shown in Figure 3. In this case, AF was identified using the original algorithm based on the RRI. The cumulative detection rates of AF using the myBeat are shown in Figure 4A. The mean duration before the first episode of AF was 2.87 days, and most detections occurred within 3 days. In the myBeat group, we investigated the distribution of the total duration time of AF during the 5 days of continuous monitoring and found a non-normal distribution and a median value of 1.27 h (IQR 0.06–10.33 h; Figure 4B).

Figure 3.

A representative case. A 72-year-old woman with hypertension was enrolled in the myBeat group. She had no symptoms and no history of cardiac diseases, including arrhythmia. Atrial fibrillation (AF) was identified using the novel patch-type device, which enabled initiation of stroke prevention measures, including anticoagulation therapy, according to the CHA2DS2-VASc score. RRI, RR interval.

Figure 4.

(A) Cumulative detection rates of atrial fibrillation (AF) using the myBeat. The myBeat was used in 150 subjects without previously known AF. AF was detected in 10.6% of patients over a period of 5 days. The mean duration before the first episode of AF was 2.87 days, and most detections occurred within 3 days. (B) Distribution of the total duration times of AF detected with myBeat over a period of 5 days. The total duration time of AF during 5 days of continuous monitoring was non-normally distributed, and the median value was 1.27 h (interquartile range 0.06–10.33 h).

Predictors of the prevalence of silent AF were analyzed. A multivariate regression analysis that included the significant factors from the univariate analysis was performed.

The incidence rates of hypertension, prior of HF, and stroke were significantly higher in patients with silent AF than in those without AF (19 [82.6%] vs. 165 [59.6%; P=2.15×10–2], 3 [13.0%] vs. 2 [0.7%; P=1.30×10–3],and 4 [17.4%] vs. 12 [4.3%; P=1.38×10–2], respectively; Supplementary Table, Table 2). Significant independent associations were found between silent AF and prior HF (OR 12.1; 95% CI 1.67–86.27; P=0.01). Prior stroke (OR 3.5; 95% CI 0.90–13.37; P=0.06) was associated with silent AF. The incidence of silent AF increased with the increase in CHA2DS2-VASc score, and a 7.7-fold difference in ORs was found between the groups with the highest (CHA2DS2-VASc ≥4) and lowest (CHA2DS2-VASc=1) scores (Figure 5).

Table 2. ORs for Screening-Detected Diagnosis of Silent AF
Variable Univariate analysis Multivariate analysis
OR (95% CI) P value OR (95% CI) P value
Age ≥75 years 3.47 (1.37–8.71) 8.82×10−3 2.18 (0.81–5.88) 0.12
Male sex 0.89 (0.37–2.11) 0.80    
Hypertension 3.22 (1.23–14.62) 2.15×10−2 2.43 (0.76–7.78) 0.13
Heart failure 20.63 (1.51–82.6) 1.30×10−3 12.07 (1.67–86.24) 0.01
Stroke 4.64 (1.36–15.80) 1.38×10−2 3.49 (0.90–13.37) 0.06

AF, atrial fibrillation; CI, confidence interval; OR, Odds Ratio.

Figure 5.

Relationship between the incidence of atrial fibrillation (AF) and the CHA2DS2-VASc score. The incidence of silent AF increased with increases in the CHA2DS2-VASc score, and there was a 7.7-fold difference in odds ratios between the groups with the highest (CHA2DS2-VASc ≥4) and lowest (CHA2DS2-VASc=1) CHA2DS2-VASc scores.

No adverse events or side effects, such as skin ulcers or dermatitis, were seen in the present study as a result of wearing the myBeat or Holter ECG.

Discussion

The incidence of AF increases with age, and generally the prevalence of AF increases as the population ages.10 AF has many adverse effects that affect patient quality of life and increase mortality, and its early detection is essential. According to several guidelines, opportunistic pulse checks are recommended for patients aged >65 years for the primary prevention of stroke associated with AF. However, the detection rate of previously unknown AF was approximately 1.4% with single-time ECG or pulse palpation, indicating low sensitivity.11 Patients without symptoms were less likely to check their pulse because they did not know when the AF occurred.

Various attempts have been made to improve the detection rate of unknown AF. Grond et al reported that the detection rate of unknown AF using 72-h ECG monitoring was approximately 2-fold higher than that using conventional 24-h ECG monitoring.12 Another recent prospective study that used continuous stroke unit ECG monitoring to screen for AF showed similar results.13 These results indicate that a longer monitoring period increases the AF detection rate. Increasing the frequency of ECG testing is also effective for AF detection. Halcox et al reported that AF was diagnosed in 3.8% of participants (19/500) who underwent mobile ECG twice a week for >12 months.14 Gladstone et al showed that ECG monitoring using a 30-day event-trigger recorder increased the detection rate of AF by a factor of >5-fold compared with conventional 24-h Holter ECG monitoring.15 In that study, the detection rate of AF in the first week in patients with the event recorder was 7.4%.15

Portable ECG devices are simple, convenient, and inexpensive. These cost-effective devices also have the advantage of frequent monitoring whenever desired. Portable ECG devices have disadvantages such as a lack of Holter function and difficulty monitoring the timing for asymptomatic patients. An insertable cardiac monitor (ICM) is more effective than conventional ECG monitoring devices for detecting AF.16 However, it is not practical to use ICMs in all patients because they are invasive and expensive for screening purposes. In the Japanese health system, ICMs are only used in patients with repetitive syncope or cryptogenic stroke. Patch-based wearable devices that provide continuous long-term ECG monitoring are non-invasive, inexpensive, and useful for asymptomatic patients. Barrett et al demonstrated that using a 14-day adhesive patch monitor detected more arrhythmia events, including supraventricular tachycardia, than using a conventional Holter monitor.17 Turakhia et al reported that silent AF was detected in 5.3% of subjects by using the wearable continuous ambulatory ECG monitoring patch-type device.18 The device was worn for up to 2 weeks in a high-risk population who had at least 2 of the following risk factors: coronary disease, HF, hypertension, diabetes, or sleep apnea.18

The novel patch device (myBeat) and its original algorithm based on RRI have been shown to accurately differentiate sinus rhythm from AF with a sensitivity of 98% and a specificity of 98.2%.9 However, we could not validate these results because subjects with other arrhythmias (e.g., frequent ventricular premature contractions and sino-atrial and atrioventricular blocks of varying degrees) were excluded in the previous pilot study. Thus, the present study evaluated the feasibility and efficacy of this novel device for screening AF in the general population.

myBeat is a small and convenient device with many advantages over conventional devices. Using the myBeat device means that storing long-term heart function information is less stressful and less expensive than for a long-term solutions or Holter ECGs. A disadvantage of the myBeat was the difficulty discerning other arrhythmias from AF, but we have improved the accuracy of the algorithm for AF detection one step at a time. Finally, the detection rate of AF using myBeat was significantly higher than that using the conventional Holter monitoring device.

In this study we assessed the risk factors and predictive parameters of silent AF and found that prior HF and stroke were associated with silent AF. The prevalence of silent AF increased significantly in participants with CHA2DS2-VASc scores >4. In a previous study, the prevalence of asymptomatic AF was higher in patients with multiple risk factors of stroke and AF; AF was found in 3% of patients with hypertension and no other risk factors of AF, but in only 7% of patients with a previous stroke and at least 1 of 2 risk factors (diabetes and hypertension).19 Siontis et al demonstrated that patients with atypical symptoms of AF and asymptomatic AF were older and had higher CHA2DS2-VASc scores than those with typical AF.8 These results have important clinical implications, namely that patients with clinical features such as being elderly, having a history of stroke, and having a history of HF, need further screening with more conventional methods, even if they have no symptoms.

This study has several limitations. First, the AF events in the myBeat group were carefully determined using our improved algorithm by an experienced electrophysiologist, but it was difficult to discriminate AF from the other arrhythmias correctly in patients with excessive atrial ectopy, and short atrial runs occurred frequently. However, because excessive supraventricular ectopic activity was reported to be associated with an increased risk of ischemic stroke, these findings have significant clinical implications for early AF screening and the prevention of stroke.20 Second, the sample size of the present study was underpowered to evaluate individual risk factors or create risk models for the detection of silent AF. In this study, the participants were randomized 1 : 1 to receive myBeat or 24-h Holter ECG monitoring. The clinical backgrounds of the patients after randomization were well balanced, which minimized the selection bias. However, a crossover design is the best way to evaluate the efficacy of devices because it removes variations between participants that exist in parallel trials. Third, we could not obtain data on medication doses and echocardiographic parameters, which may have affected the overall accuracy of our results. The accuracy of diagnosing AF using the myBeat has not been established because of these study limitations. Therefore, an external validation study is necessary before using the myBeat in clinical practice. Fourth, the clinical benefit and cost-effectiveness of screening for silent AF were not evaluated owing to the limitations of the study design. Larger studies are needed to further evaluate the clinical effectiveness of this novel device in screening for silent AF to construct a clinical risk model for AF.

Despite these limitations, the present study is the first study to used and validate this novel device based on a landmark algorithm of RRI analysis for screening silent AF.

In conclusion, the novel patch-type device using an original algorithm based on RRI analysis was beneficial in screening for silent AF. The elderly population with multiple risk factors will benefit from screening for unknown AF using this novel device.

Acknowledgments

We thank the members of the clerical and medical staff at Hiroshima University Hospital and ENAGO Group (English editing system) for their linguistic assistance.

Sources of Funding

This trial was supported by UNION TOOL Co. (Niigata, Japan), JMS Co., Ltd., SYSTEM FRIEND Inc. (Hiroshima, Japan), and the Center for the Promotion of Medicine-Technological Industries Collaboration (Hiroshima, Japan). They provided the device (myBeatTM) and patch electrodes, but were not involved in study design or execution, data analysis, or manuscript preparation. Y.N. was supported by JSPS KAKENHI Grant no. 17K09501.

Disclosures

The authors have no conflicts of interest to disclose.

IRB Information

This study was approved by the Institutional Ethics Committee of the Graduate School of Biomedical Science at Hiroshima University (Approval no. E-844, registered September 30, 2017).

Data Availability

The deidentified participant data will not be shared.

Supplementary Files

Please find supplementary file(s);

http://dx.doi.org/10.1253/circj.CJ-20-1061

References
 
© 2021, THE JAPANESE CIRCULATION SOCIETY

This article is licensed under a Creative Commons [Attribution-NonCommercial-NoDerivatives 4.0 International] license.
https://creativecommons.org/licenses/by-nc-nd/4.0/
feedback
Top