Revisit to the Prognostic Value of Premature Atrial Contraction Burden in 24-h Holter Electrocardiography for Predicting Undiagnosed Atrial Fibrillation　― A Propensity Score-Matched Study ―

Abstract

Background: The optimum cut-off value of premature atrial contraction (PAC) burden (CV-PACb) in 24-h Holter electrocardiography (24-h ECG) for predicting atrial fibrillation (AF) is debatable, with few validation data.

Methods and Results: We retrospectively analyzed 61 patients already diagnosed with AF (AD-AF) and 147 patients never diagnosed with AF (ND-AF), aged ≥50 years, free of heart disease, and who had undergone 24-h ECG and transthoracic echocardiography (TTE). Receiver operating characteristic analysis demonstrated that 0.4% was the optimal CV-PACb differentiating AD-AF from ND-AF, with 69% sensitivity and 72% specificity (area under the curve [AUC] 0.72; 95% confidence interval [CI] 0.65–0.79); however, the left atrial volume index was not significant (AUC 0.60; 95% CI 0.51–0.68). To verify the CV-PACb, new propensity-matched cohorts (i.e., subjects with a PAC burden ≥0.4% and <0.4%; n=69 in each group) were compared based on new detection of AF at a median follow-up of 50 months (interquartile range 12–60 months) Multivariable Cox regression analysis revealed that among 24-h ECG and TTE findings, only PAC burden ≥0.4% was independently associated with incident AF (hazard ratio 5.28; 95% CI 1.28–26.11; P=0.023).

Conclusions: A high PAC burden (≥0.4%) in 24-h ECG was a reliable indicator to identify undiagnosed AF, whereas TTE parameters did not show any predictive value.

Atrial fibrillation (AF) can present with symptoms such as palpitation, dizziness, chest pain, shortness of breath, or syncope and is considered a major cause of a cardioembolic stroke. Thus, early detection of AF is required. For this purpose, a 24-h Holter electrocardiography (24-h ECG) is routinely performed; however, it has a low diagnostic yield due to its short examination time.¹ Prolonged ECG monitoring techniques, such as a 2-week ECG recorder or intracardiac monitor for cryptogenic syncope attacks or embolic stroke of undetermined source (ESUS), lack feasibility due to their invasiveness and cost. Approximately 25% of patients with AF-related stroke were diagnosed with AF at the time of stroke.² Therefore, aggressive primary prevention and less-invasive techniques, such as 24-h ECG or transthoracic echocardiography (TTE), should be performed to predict covert or new-onset AF. In patients with paroxysmal AF, it has been observed that rapid runs of premature atrial contraction (PAC) precede the development of AF, and repeated episodes of AF can cause electrical or structural remodeling (or both) of the atrial muscle.^3,4 Previous studies have shown that a high PAC burden is an indicator for the development of AF; however, there is no universal agreement regarding the cut-off value for PAC burden, with few validation data.^5–10 Therefore, we sought to derive a specific PAC burden discriminating patients with and without AF using a single 24-h ECG, and to verify its predictive value, while exploring the diagnostic utility of TTE parameters representing enlargement of the left atrium (LA) or ventricular diastolic dysfunction, both of which are known predictors of AF.^11–13

Editorial p 1273

Methods

Study Design and Population

This was a single-center retrospective observational study. Patients aged ≥50 years who underwent 24-h ECG and TTE at St. Marianna University Hospital were enrolled. Patients who were <50 years old were excluded because AF is less prevalent in such patients and the conditions related to the development of AF in younger patients are different from those seen in older people.

The study protocol was approved by the Institutional Ethics Committee of St. Marianna University Hospital (Reference no. 4964).

To determine the cut-off value of PAC burden discriminating patients with and without AF, candidates recruited between January 2020 and April 2020 were classified into 2 groups: (1) patients who had already been diagnosed with AF (AD-AF); and (2) patients who had never been diagnosed with AF (ND-AF).

Patients with structural heart disease (e.g., prior myocardial infarction, cardiomyopathy, severe valvular heart disease, congenital heart disease, a history of congestive heart failure, and previous heart surgery), prior catheter ablation for AF, and prior pacemaker implantation were excluded from the study because these clinical backgrounds strongly affect the occurrence of AF and may prevent accurate estimation of the net effect of PAC on the incidence of AF. In addition, patients who had been treated with antiarrhythmic drugs other than β-blockers were excluded, as were patients who had suffered embolic stroke without a history of AF, because these patients may have subclinical AF and including them in the ND-AF group could overlook true differences between AD-AF and ND-AF. In this study, embolic stroke was defined as a non-lacunar brain infarct without proximal arterial stenosis or cardioembolic sources. Finally, patients with AF documentation during the 24-h ECG were also excluded from the study.

Receiver operating characteristic (ROC) analysis was used to determine an appropriate cut-off value for the final population; this phase was defined as the derivation study (Figure 1A). We also investigated the TTE parameters and their correlation with the outcomes.

Figure 1.

Flowchart of the study population. (A) Derivation study, (B) Validation study. AAD, antiarrhythmic drug; AD-AF, already diagnosed atrial fibrillation; AF, atrial fibrillation; ECG, electrocardiography; ND-AF, never diagnosed AF; PAC, premature atrial contraction; PMI, pacemaker implantation; PS, propensity score; pts, patients; PVI, pulmonary vein isolation; ROC, receiver operating characteristic; ROC, receiver operating characteristic; SHD, structural heart disease; TTE, transthoracic echocardiography.

For the validation study (Figure 1B), patients were recruited between January 2015 and May 2015. The purpose of this recruitment was to stratify a new cohort without a history of AF using the cut-off value determined in the derivation study and to corroborate the stratification by Cox regression analysis. Exclusion criteria for the validation study were the same as for the derivation study (see above), with the exception that patients with prior embolic stroke were now included because some of them may have covert AF. Using the cut-off value obtained from the derivation study, the subjects in the validation study were divided into 2 subsets and were followed up until June 2020. We then verified whether the cut-off values could predict AF before and after propensity score (PS) matching analysis.

Analysis of 24-h ECG

A 24-h ECG (FM-180, FM-180S, FM-190, and FM-980; Fukuda Denshi, Tokyo, Japan) was performed to: (1) examine the causes of symptoms, such as palpitations, chest pain, shortness of breath, dizziness, or syncope; (2) follow-up targeted arrhythmias (i.e., PAC, premature ventricular contraction, atrioventricular block, or sick sinus syndrome); and (3) identify potential causes of embolic events. The procedures were performed by experienced cardiac technicians and reviewed by the responsible investigating cardiologist. AF was defined as irregular supraventricular activity without monomorphic P-waves or irregular supraventricular activity in the presence of F-waves lasting for at least 30 s, and PACs as supraventricular complexes occurring >25% earlier than expected compared with the previous RR interval. The following parameters were examined: (1) PAC burden (%), calculated by dividing the number of PACs by total heartbeats (THB) in the index recording and multiplying by 100; (2) total number of PAC runs, defined as ≥3 consecutive PACs; and (3) longest PAC run. To estimate PAC frequency, we used the percentage of PACs instead of the total number of PACs because the number of PACs depended on the entire recording time or THB, which exhibited a certain interindividual variation.

Analysis of TTE

In all study populations, TTE examinations were performed simultaneously with 24-h ECG recordings. The following data were extracted: left ventricular (LV) ejection fraction (LVEF), LV diastolic function, LA diameter, and LA maximum volume index (LAVI). The LVEF was calculated using the biplane disc summation method. The diameter of the LA was measured at the end of LV systole from the parasternal long-axis image, and the LAVI was calculated using the biplane disk summation method. LAVI was calculated by dividing the LA volume by the body surface area. The LV diastolic function was evaluated by: (1) transmitral E and A peak velocities; (2) the E/A ratio; and (3) the mean ratio of E to e’, where e’ is early diastolic mitral annular velocity derived from tissue Doppler imaging.

Follow-up and Study Endpoint

Participants in the validation study were retrospectively followed up until June 2020. Clinical information of patients who did not undergo regular checkups at the St. Marianna University Hospital outpatient clinic was obtained by telephone interviews or contact with patients’ home doctors. The primary endpoint of this study was newly detected AF within the follow-up period, which was defined as any AF documented by standard 12-lead ECG or ambulatory ECG monitoring. We could not determine whether the AF revealed was new in onset or subclinical because continuous ECG monitoring was not performed in this study; thus, documentation was unattainable.

Statistical Analysis

Continuous variables are described as the mean±SD or the median with interquartile range (IQR), and were compared using Student’s t-test or the Wilcoxon rank-sum test depending on variable distribution. Categorical values are presented as counts and percentages and were compared using Fisher’s exact test. The analysis of ROC curve was used in the derivation study to determine the optimal cut-off value for PAC burden to distinguish AD-AF from ND-AF. The cut-off value was then used to stratify subjects in the validation study into 2 groups (i.e., frequent and infrequent PACs). The incidence of AF during the follow-up period was assessed using the Kaplan-Meier method and log-rank test. Multivariable Cox regression analyses were also performed to assess the relationship between 24-h ECG or TTE variables with P<0.10 in the univariate analysis and the time to detection of AF. These analyses were also performed after PS matching.

The PS for PAC burden was generated from a multivariable logistic regression model using 4 covariates that were strongly associated with the incidence of AF, namely age, sex, hypertension, and LAVI substituted for LA enlargement.^11,12,14,15 Patients in the 2 groups were matched on a 1 : 1 basis using a 4-digit nearest neighbor algorithm within a caliper of 0.2, resulting in 69 patient pairs. Baseline characteristics of the matched groups were re-evaluated, and the multivariable Cox proportional hazards method was used to compare the 2 groups based on the onset of AF.

Two-sided P<0.05 was considered statistically significant. Statistical analyses were performed using JMP^® 14 (SAS Institute Inc., Cary, NC, USA).

Results

Patient Characteristics in the Derivation Study

The derivation study cohort of 208 patients consisted of 61 with AD-AF and 147 with ND-AF. Comparisons of baseline characteristics between the 2 groups are summarized in Table 1. There were no significant differences between the 2 cohorts, except for older age (75±8 vs. 71±11 years; P=0.012), higher B-type natriuretic peptide (BNP) levels (median [IQR] 100.3 [55.1–167.0] vs. 25.8 [18.8–94.2] pg/mL; P=0.014) and more β-blocker use (27% vs. 12%; P=0.012) in the AD-AF group. The higher use of β-blockers was associated with the fact that patients in the AD-AF group were treated with these drugs. Hypertension, diabetes, dyslipidemia, high body mass index, high BNP, and CHADS₂ score ≥2, which are known predictive factors of new AF,^8,14–17 were not significantly different between the 2 groups. However, there was a tendency that hypertension was more prevalent in the AD-AF than ND-AF group (69% vs. 54%; P=0.064).

Table 1. Clinical Characteristics of Patients in the Derivation Study

	All (n=208)	AD-AF (n=61)	ND-AF (n=147)	P value
Age (years)	72±10	75±8	71±11	0.012
Female sex	100 (48)	25 (41)	75 (51)	0.223
BMI (kg/m2)	23±4	23±3	23±4	0.756
Hypertension	122 (59)	42 (69)	80 (54)	0.064
Diabetes mellitus	34 (16)	8 (13)	26 (18)	0.538
Dyslipidemia	53 (25)	15 (25)	38 (26)	1.000
CHADS2 score ≥2	89 (43)	29 (48)	60 (41)	0.442
eGFR (mL/min/1.73 m2)	61±20	58±22	62±19	0.209
BNP (pg/mL)	71.6 [30.9–148.9]	100.3 [55.1–167.0]	25.8 [18.8–94.2]	0.014
Medications
Calcium channel blocker	79 (38)	29 (47)	50 (34)	0.083
β-blocker	34 (16)	16 (27)	18 (12)	0.012
ACEI or ARB	61 (29)	20 (33)	41 (28)	0.501

Categorical variables are shown as absolute numbers and percentages of subgroups. Continuous variables are given as the mean±SD or median [interquartile range]. ACEI, angiotensin-converting enzyme inhibitor; AD-AF, already diagnosed atrial fibrillation; ARB, angiotensin-receptor blocker; BMI, body mass index; BNP, B-type natriuretic peptide; eGFR, estimated glomerular filtration rate; ND-AF, never diagnosed atrial fibrillation.

Findings of 24-h ECG and TTE in the Derivation Study

The findings of 24-h ECG and TTE for the derivation study are summarized in Table 2. Patients with AD-AF had lower THB (96,449±14,401 vs. 103,221±16,564 beats/record; P=0.006) and a lower mean heart rate (69±10 vs. 74±11 beats/min; P=0.002) than those with ND-AF, despite similar recording durations. These observations may be associated with increased β-blocker use in the AD-AF group. Patients with AD-AF had higher PAC burden (median [IQR] 0.82% [0.21–2.24%] vs. 0.14% [0.03–0.56%]), more PAC runs (median [IQR] 7 [3–36] vs. 2 [0–5]), and longer PAC runs (median [IQR] 8 [4–16] vs. 4 [2–9] beats) than those with ND-AF (all P<0.001). Regarding echocardiographic parameters, patients with AD-AF had a larger LA diameter (36±6 vs. 34±6 mm; P=0.014) and higher LAVI (36±16 vs. 30±11 mL/m²; P=0.015) than those with ND-AF. There were no significant differences in E/A and mean E/e’ ratio between the 2 groups (P=0.126 and 0.118, respectively).

Table 2. Findings of 24-h Holter Electrocardiography and Transthoracic Echocardiography for Patients in the Derivation Study

	All (n=208)	AD-AF (n=61)	ND-AF (n=147)	P value
24-h Holter ECG findings
Monitoring duration (h)	23.7±1.0	23.7±1.2	23.6±0.9	0.672
Total no. heart beats/record	101,235±16,222	96,449±14,401	103,221±16,564	0.006
Mean heart rate (beats/min)	73±11	69±10	74±11	0.002
No. PACs (beats/record)	204 [38–985]	733 [208–2,124]	138 [29–577]	<0.001
PAC burden (%)	0.21 [0.04–1.14]	0.82 [0.21–2.24]	0.14 [0.03–0.56]	<0.001
No. PAC runs (episodes/record)	3 [1–9]	7 [3–36]	2 [0–5]	<0.001
Longest PAC run (beats)	5 [3–11]	8 [4–16]	4 [2–9]	<0.001
TTE findings
LVEF (%)	65±7	66±8	65±6	0.252
LA diameter (mm)	35±6	36±6	34±6	0.014
LAVI (mL/m2)	32±13	36±16	30±11	0.015
E/A ratio	0.85±0.38	0.92±0.47	0.82±0.34	0.126
Mean E/e’ ratio	9.5±3.5	10.0±3.7	9.2±3.3	0.118

Categorical variables are shown as absolute numbers and percentages of subgroups. Continuous variables are given as the mean±SD or median [interquartile range]. ECG, electrocardiography; LA, left atrium; LAVI, left atrial volume index; LVEF, left ventricular ejection fraction; PAC, premature atrial contraction; TTE, transthoracic echocardiography. Other abbreviations as in Table 1.

Parameters Discriminating AD-AF From ND-AF

The ROC curve analysis revealed the following cut-off values for PAC parameters in the 24-h ECG: 0.4% PAC burden, 4 episodes of PAC run, and 4 successive beats of PAC. These parameters indicated the presence of AF, with an area under the curve (AUC) of 0.72 (95% confidence interval [CI] 0.65–0.79), 0.71 (95% CI 0.63–0.78), and 0.67 (95% CI 0.59–0.74), respectively (Figure 2A–C). The cut-off values derived from the TTE parameters were an LA diameter of 34 mm (AUC 0.61; 95% CI 0.52–0.67), LAVI of 30 mL/m² (AUC 0.60; 95% CI 0.51–0.68), an E/A of 0.72 (AUC 0.57; 95% CI 0.47–0.65), and a mean E/e’ ratio of 10.1 (AUC 0.57; 95% CI 0.48–0.66; Figure 2D–G). Therefore, it is reasonable to consider that 0.4% PAC burden and 4 episodes of PAC run were more optimal (AUC >0.7) cut-off points than value for other parameters in the 24-h ECG and TTE. Because there was a significant correlation between (log-transformed) PAC burden and the number of PAC runs (R²=0.72, P <0.001), we focused on evaluating PAC burden only, which was the primary objective of the present study.

Figure 2.

Receiver operating characteristic curves of (A) premature atrial contraction (PAC) burden, (B) number of PAC runs, (C) longest PAC run, (D) left atrial (LA) diameter, (E) LA volume index (LAVI), (F) E/A ratio, and (G) mean E/e’ for the incidence of atrial fibrillation. AUC, area under the curve; CI, confidence interval.

Prognosis of PS-Matched Groups Stratified by PAC Burden

Subjects in the validation study were stratified by the optimal PAC burden cut-off value into 2 subsets, namely those with a PAC burden ≥0.4% (frequent PACs) and those with a PAC burden <0.4% (infrequent PACs), and the prognoses (i.e., AF occurrence) investigated. Table 3 summarizes baseline characteristics of patients in the validation study. Significant differences were observed in clinical characteristics between the comparison groups before PS matching. Specifically, subjects with a PAC burden ≥0.4% were older, with a tendency for a greater proportion of males in the group, and had higher BNP levels, a higher prevalence of CHADS₂ score ≥2, and a lower estimated glomerular filtration rate. After matching, there were no significant differences between the 2 groups in any of the covariates. Regarding TTE parameters, LA enlargement was greater in the frequent than infrequent PACs group before PS matching, but not after. With regard to 24-h ECG findings, monitoring duration, THB, and mean heart rate did not differ significantly, unlike in the derivation study.

Table 3. Baseline Characteristics of Patients in the Validation Study

	Before propensity score matching			After propensity score matching
	PAC burden ≥0.4% (n=80)	PAC burden <0.4% (n=246)	P value	PAC burden ≥0.4% (n=69)	PAC burden <0.4% (n= 69)	P value
Age (years)	75±9	69±10	<0.001	75±9	75±8	0.728
Female sex	30 (38)	126 (51)	0.039	27 (39)	27 (39)	1.000
BMI (kg/m2)	23±3	23±4	0.664	23±3	23±3	0.779
Hypertension	50 (63)	125 (51)	0.072	42 (61)	44 (64)	0.861
Diabetes	14 (18)	44 (18)	1.000	14 (20)	19 (28)	0.425
Dyslipidemia	25 (31)	83 (34)	0.785	24 (35)	30 (43)	0.383
CHADS2 score ≥2	45 (56)	94 (38)	0.006	41 (59)	39 (57)	0.863
eGFR (mL/min/1.73 m2)	61±22	67±20	0.025	62±23	62±18	0.539
BNP (pg/mL)	50.5 (33.6–135.6)	28.9 (14.9–67.2)	<0.001	50.9 (33.8–131.8)	38.1 (16.0–100.1)	0.119
Medications
Calcium channel blocker	35 (44)	84 (35)	0.180	30 (44)	29 (43)	1.000
β-blocker	5 (6)	18 (8)	0.808	4 (6)	3 (4)	1.000
ACEI or ARB	29 (37)	71 (30)	0.265	24 (35)	25 (37)	0.859
24-h Holter ECG findings
Monitoring duration (h)	23.8±0.6	23.8±1.0	0.546	23.8±0.7	23.7±1.4	0.543
Total no. heart beats/record	101,597±16,063	103,229±16,063	0.440	103,819±16,716	100,309±17,719	0.233
Mean heart rate (beats/min)	73±11	74±11	0.345	74±10	72±11	0.263
No. PACs (beats/record)	1,435 (649–2,864)	66 (26–133)	<0.001	1,508 (651–2,863)	78 (43–170)	<0.001
TTE findings
LVEF (%)	66±8	67±6	0.861	67±8	67±7	0.974
LA diameter (mm)	37±6	35±5	0.001	36±5	36±5	0.425
LAVI (mL/m2)	34±14	27±9	<0.001	31±12	30±9	0.673
E/A ratio	0.77±0.30	0.83±0.62	0.227	0.73±0.25	0.73±0.24	0.964
Mean E/e’ ratio	9.3±3.2	8.6±2.8	0.081	9.0±3.1	9.2±3.3	0.811

Categorical variables are shown as absolute numbers and percentages of subgroups. Continuous variables are given as the mean±SD or median [interquartile range]. Abbreviations as in Tables 1 and 2.

After a median [IQR] follow-up of 50 (12–60) months, AF was newly detected in 17 patients (13 after PS matching), and the event rate was significantly higher in the frequent PAC group before (log-rank P<0.001) and after (log-rank P=0.010) PS matching. Figure 3 shows the Kaplan-Meier AF-free survival curves in PS-matched groups stratified by PAC burden. Frequent PACs were associated with a higher incidence of AF than infrequent PACs.

Figure 3.

Kaplan-Meier curves showing the incidence of atrial fibrillation (AF) in propensity score-matched patients stratified by premature atrial contraction (PAC) burden. AF, atrial fibrillation; PAC, premature atrial contraction.

Independent Risk Factors for the Incidence of AF

Table 4 shows the hazard ratios (HRs) for patient characteristics for the incidence of AF before and after PS matching. The multivariable Cox proportional hazards regression model demonstrated that a PAC burden ≥0.4% among baseline characteristics was significantly associated with a higher risk of AF occurrence (HR 6.21; 95% CI 1.90–20.26; P=0.003). After PS matching, aging (HR 1.99 [per 10-year increment]; 95% CI 1.01–4.21; P=0.045) and PAC burden ≥0.4% (HR 5.28; 95% CI 1.28–26.11; P=0.023) were identified as indicators of AF. No echocardiographic parameters were found to have a predictive value for the incidence of AF.

Table 4. Hazard Ratios of Patient Characteristics for the Incidence of Atrial Fibrillation Before and After Propensity Score Matching

	Before propensity score matching				After propensity score matching
	Univariate		Multivariable		Univariate		Multivariable
	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value
Age (per 10-year increment)	2.29 (1.33–4.11)	0.002	1.60 (0.93–2.86)	0.093	1.95 (0.96–4.36)	0.065	1.99 (1.01–4.21)	0.045
Female sex	0.69 (0.26–1.80)	0.444			0.92 (0.30–2.81)	0.884
Hypertension	1.11 (0.42–2.91)	0.833			0.88 (0.29–2.69)	0.825
eGFR (per 10-mL/min/ 1.73 m2 increment)	0.84 (0.68–1.07)	0.150			0.88 (0.70–1.14)	0.313
BNP (per 10-pg/mL increment)	1.02 (0.93–1.09)	0.565			1.00 (0.90–1.06)	0.968
PAC burden ≥0.4%	9.38 (3.06–28.76)	<0.001	6.21 (1.90–20.26)	0.003	5.40 (1.20–24.39)	0.028	5.28 (1.28–26.11)	0.023
LAVI (per 1.0-mL/m2 increment)	1.05 (1.01–1.08)	0.011	1.01 (0.98–1.05)	0.431	1.02 (0.97–1.07)	0.367
E/A ratio (per 0.1 increment)	1.00 (0.88–1.04)	0.950			1.09 (0.87–1.31)	0.409
Mean E/e’ ratio (per 1.0 increment)	1.18 (1.02–1.33)	0.023	1.05 (0.89–1.23)	0.359	1.10 (0.92–1.28)	0.287

CI, confidence interval; HR, hazard ratio. Other abbreviations as in Tables 1 and 2.

Discussion

Main Findings

The main findings of this study are that: (1) among the 24-h ECG and TTE parameters, PAC burden was the best in differentiating AD-AF from ND-AF; (2) PAC burden ≥0.4% successfully predicted the occurrence of AF in the PS-matched AF-free cohorts stratified according to PAC burden during a median of 50 months (IQR 12–60 months) of follow-up; and (3) PAC burden ≥0.4% was an independent risk factor for the incidence of AF, but echocardiographic parameters representing LA enlargement or LV diastolic dysfunction were not.

Parameters in the 24-h ECG Discriminating and Predicting AF

The present study demonstrated that a PAC burden of 0.4% in the 24-h ECG reasonably and accurately discriminated AD-AF from ND-AF with a sensitivity and specificity of 69% and 72%, respectively (AUC 0.72). To the best of our knowledge, only one other study has performed ROC analysis to differentiate patients with AF using 24-h ECG parameters, although the data after pulmonary vein isolation is reported occasionally.^9,18 In previous reports, there is no consensus definition of frequent PACs. Defined cut-off values vary between studies, with these studies defining each cut-off point by referring to the distribution in PAC quartiles. For example, Chong et al and Suzuki et al reported that PAC burden with increased risk of AF occurrence was >100 and ≥102 beats/day, respectively,^7,8 and Wallmann et al reported that ≥70 PACs/day was associated with a greater incidence of AF in patients with ischemic stroke.⁵ In contrast, Binici et al reported that ≥720 PACs/day predicted the development of AF in apparently healthy subjects.⁶ To describe the association between PACs and AF, we used a dichotomized PAC count, as in previous studies. It was expected that this association was continuous and independent of an arbitrary PAC cut-off point, and there was no lower PAC threshold below which AF could be excluded. In addition, it does not seem feasible to determine specific values perfectly stratifying the risk of AF because they greatly depend on the clinical background of the study population. However, establishing a certain criterion can be useful in determining which patients require follow-up for the detection of covert AF. Further, it may have value as a triage tool to identify high-risk subjects who may benefit from additional monitoring. The cut-off point determined in this study (i.e., a PAC burden of 0.4%) seems not so strict that it will lead to over-triaging and not so lenient as to overlook underlying AF.

As shown in the Kaplan-Meier survival curves in Figure 3, stratification with a PAC burden of 0.4% successfully predicted the incidence of AF in subjects without a prior diagnosis of AF. Further, we verified the stratification by Cox proportional hazards models before and after PS matching. It is evident that the prevalence of PAC increased with aging,¹⁹ which was a significant predictor for AF in addition to frequent PAC burden in the present study. Therefore, it was very likely that the longer the follow-up period, the more AF patients were detected. In the present study, the cut-off point of a PAC burden of 0.4% showed high predictive value; however, the PAC burden could potentially be a surrogate marker reflecting prevalent subclinical AF or a precursor indicating subsequent AF occurrence. Demonstrating that a patient now has AF is generally unfeasible, especially in asymptomatic cases, unless repeated or prolonged ECG monitoring is performed. Considering the relatively short follow-up period in the present study, the AF identified may indicate covert rather than new-onset AF.

Repetitive PACs were also reported to be an independent predictor of AF.^6,9 However, we excluded data regarding PAC runs for the following reasons. First, there was a general correlation between PAC burden and the extent of repetitive PACs; thus, the problem of multicollinearity could appear in the multivariable analysis. Second, there was a possibility that PAC runs may be underestimated because of the presence of blocked PAC. If a supraventricular beat was not conducted to the ventricles because of atrioventricular conduction impairment or a very short coupling interval, the QRS complex corresponding to the beat was left out and the contraction was not counted as a PAC.

Parameters of TTE Discriminating and Predicting AF

None of the echocardiographic parameters mentioned in the present study showed statistical significance for predicting new AF. As previously mentioned, LA enlargement or LV diastolic dysfunction is a known indicator of AF.^11–13 In fact, the parameters regarding them (i.e., LA diameter, LAVI, or E/e’) were significantly different between patients with and without AF in the derivation study. However, they were almost useless as an AF discriminator, although LA dilatation or LV diastolic dysfunction was expected to be associated with the occurrence of AF from electrophysiological and hemodynamic perspectives. This may be because LA enlargement or LV diastolic dysfunction does not easily appear until AF is converted from paroxysmal to the vicinity of persistent AF; thus, the presence of AF is hard to document with TTE findings, at least during an early stage of paroxysmal AF. In the present study, we enrolled patients who had less advanced atrial remodeling, because the mean LA diameter of AD-AF in the derivation study was 36±6 mm. Further, one of the major reasons why TTE parameters failed to show predictivity was thought to be attributable to interobserver variability of measurement or physical differences between individuals; this was not evaluated in the present study. In contrast, PAC burden is a more objective parameter, and the assessment of PAC is readily available and easily quantifiable via 24-h ECG.

Clinical Implications

Twenty-four-hour ECG is routinely used to diagnose AF; however, it has relatively low sensitivity, especially in patients with ESUS. If AF is expected in those patients, anticoagulants can be used to prevent AF-related ischemic strokes. For patients with a PAC burden ≥0.4% showing a greater risk for developing or having AF, intense diagnostic intervention with repeated or prolonged ECG monitoring for the early detection of AF should be performed. Repeated or prolonged ECG monitoring significantly improves AF detection; however, it is unclear how often the monitoring should be performed. Further research on the optimal management of patients with frequent PACs is urgently needed, and non-invasive or minimally invasive monitoring technologies for AF detection should be deployed.

Study Limitations

Our study has several limitations. First, this was a retrospective study based on medical records from a single center, and systematic follow-up, such as regular 24-h ECG checkups, was not performed. Thus, we cannot ignore the possibility of underestimating new AF, especially in asymptomatic patients. Second, there was a selection bias. Most of the study population had symptoms such as palpitation, dizziness, syncope, or shortness of breath, and thus represented a sicker cohort. In addition, this study was performed in middle-aged and elderly Japanese patients. Therefore, the application of the cut-off value in the present study to subjects with other clinical backgrounds should be done with caution. Third, 24-h ECG may not be as reproducible as a 1-week ECG or longer ECG monitoring, because the number of PACs is likely to be affected, to some extent, by the circadian rhythm or autonomic activities. Fourth, we did not assess several risk factors associated with the occurrence of AF, such as alcohol intake or smoking status because of incomplete data. Finally, data regarding the coupling interval of PACs or P-wave morphology, which should provide useful information for detecting AF, were not available.

Conclusions

A high PAC burden (≥0.4%) assessed by 24-h ECG independently predicted the occurrence of AF in patients without a history of AF. This association remained significant after PS matching.

Acknowledgments

None.

Disclosures

Y.J.A. is a member of Circulation Journal’s Editorial Board. None of the authors has any financial disclosures.

IRB Information

This study was approved by the Institutional Ethics Committee of St. Marianna University (Reference no. 4964).

References

• 1.   Schuchert A, Behrens G, Meinertz T. Impact of long-term ECG recording on the detection of paroxysmal atrial fibrillation in patients after an acute ischemic stroke. Pacing Clin Electrophysiol 1999; 22: 1082–1084.
• 2.   Wolf PA, Kannel WB, McGee DL, Meeks SL, Bharucha NE, McNamara PM. Duration of atrial fibrillation and imminence of stroke: The Framingham study. Stroke 1983; 14: 664–667.
• 3.   Kolb C, Nürnberger S, Ndrepepa G, Zrenner B, Schömig A, Schmitt C. Modes of initiation of paroxysmal atrial fibrillation from analysis of spontaneously occurring episodes using a 12-lead Holter monitoring system. Am J Cardiol 2001; 88: 853–857.
• 4.   John AG, Hirsch GA, Stoddard MF. Frequent premature atrial contractions impair left atrial contractile function and promote adverse left atrial remodeling. Echocardiography 2018; 35: 1310–1317.
• 5.   Wallmann D, Tüller D, Wustmann K, Meier P, Isenegger J, Arnold M, et al. Frequent atrial premature beats predict paroxysmal atrial fibrillation in stroke patients: An opportunity for a new diagnostic strategy. Stroke 2007; 38: 2292–2294.
• 6.   Binici Z, Intzilakis T, Nielsen OW, Køber L, Sajadieh A. Excessive supraventricular ectopic activity and increased risk of atrial fibrillation and stroke. Circulation 2010; 121: 1904–1911.
• 7.   Chong BH, Pong V, Lam KF, Liu S, Zuo ML, Lau YF, et al. Frequent premature atrial complexes predict new occurrence of atrial fibrillation and adverse cardiovascular events. Europace 2012; 14: 942–947.
• 8.   Suzuki S, Sagara K, Otsuka T, Kano H, Matsuno S, Takai H, et al. Usefulness of frequent supraventricular extrasystoles and a high CHADS₂ score to predict first-time appearance of atrial fibrillation. Am J Cardiol 2013; 111: 1602–1607.
• 9.   Acharya T, Tringali S, Bhullar M, Nalbandyan M, Ilineni VK, Carbajal E, et al. Frequent atrial premature complexes and their association with risk of atrial fibrillation. Am J Cardiol 2015; 116: 1852–1857.
• 10.   Lin CY, Lin YJ, Chen YY, Chang SL, Lo LW, Chao TF, et al. Prognostic significance of premature atrial complexes burden in prediction of long-term outcome. J Am Heart Assoc 2015; 4: e002192.
• 11.   Psaty BM, Manolio TA, Kuller LH, Kronmal RA, Cushman M, Fried LP, et al. Incidence of and risk factors for atrial fibrillation in older adults. Circulation 1997; 96: 2455–2461.
• 12.   Olshansky B, Heller EN, Mitchell LB, Chandler M, Slater W, Green M, et al. Are transthoracic echocardiographic parameters associated with atrial fibrillation recurrence or stroke?: Results from the Atrial Fibrillation Follow-Up Investigation of Rhythm Management (AFFIRM) study. J Am Coll Cardiol 2005; 45: 2026–2033.
• 13.   Tsang TS, Gersh BJ, Appleton CP, Tajik AJ, Barnes ME, Bailey KR, et al. Left ventricular diastolic dysfunction as a predictor of the first diagnosed nonvalvular atrial fibrillation in 840 elderly men and women. J Am Coll Cardiol 2002; 40: 1636–1644.
• 14.   Benjamin EJ, Levy D, Vaziri SM, D’Agostino RB, Belanger AJ, Wolf PA. Independent risk factors for atrial fibrillation in a population-based cohort: The Framingham Heart Study. JAMA 1994; 271: 840–844.
• 15.   Conen D, Tedrow UB, Koplan BA, Glynn RJ, Buring JE, Albert CM. Influence of systolic and diastolic blood pressure on the risk of incident atrial fibrillation in women. Circulation 2009; 119: 2146–2152.
• 16.   Schnabel RB, Larson MG, Yamamoto JF, Sullivan LM, Pencina MJ, Meigs JB, et al. Relations of biomarkers of distinct pathophysiological pathways and atrial fibrillation incidence in the community. Circulation 2010; 121: 200–207.
• 17.   Wanahita N, Messerli FH, Bangalore S, Gami AS, Somers VK, Steinberg JS. Atrial fibrillation and obesity: Results of a meta-analysis. Am Heart J 2008; 155: 310–315.
• 18.   Inoue H, Tanaka N, Tanaka K, Ninomiya Y, Hirao Y, Oka T, et al. Burden and long firing of premature atrial contraction early after catheter ablation predict late recurrence of atrial fibrillation. Circ J 2020; 84: 894–901.
• 19.   Fleg JL, Kennedy HL. Cardiac arrhythmias in a healthy elderly population: Detection by 24-hour ambulatory electrocardiography. Chest 1982; 81: 302–307.