Predicting risk of AF in ischaemic stroke using sinus rhythm ECG abnormalities: A meta-analysis

Abstract

Objective:

To identify ECG changes in sinus rhythm that may be used to predict subsequent development of new AF.

Method:

We identified prospective and retrospective cohort or case control studies evaluating ECG patterns from a 12-lead ECG in sinus rhythm taken in hospital or community predicting subsequent development of new AF. For each identified ECG predictor, we then identify absolute event rates and pooled risk ratios (RR) using an aggregate level random effects meta-analysis.

Results:

We identified 609,496 patients from 22 studies. ECG patterns included P wave terminal force V1 (PTFV1), interatrial block (IAB) and advanced interatrial block (aIAB), abnormal P wave axis (aPWA), PR prolongation and atrial premature complexes (APCs). Pooled risk ratios reached significance for each of these; PTFV1 RR 1.48 (95% CI 1.04–2.10), IAB 2.54 (95% CI 1.64–3.93), aIAB 4.05 (95% CI 2.64–6.22), aPWA 1.89 (95% CI 1.25–2.85), PR prolongation 2.22 (95% CI 1.27–3.87) and APCs 3.71 (95% CI 2.23–6.16). Diabetes reduced the predictive value of PR prolongation.

Conclusion:

APC and aIAB were most predictive of AF, while IAB, PR prolongation, PTFV1 and aPWA were also significantly associated with development of AF. These support their use in a screening tool to identify at risk cohorts who may benefit from further investigation, or following stroke, with empirical anticoagulation.

Keywords

Stroke prevention atrial fibrillation atrial cardiopathy ESUS

Introduction

Embolic stroke of undetermined source (ESUS) is a term developed in 2014 referring to patients who present with non-lacunar embolic strokes where there is no other identifiable cause at presentation. These patients account for at least 25% of all stroke presentations.¹ While it was initially hypothesised that ESUS patients may benefit from anticoagulation, there is currently no evidence to support this treatment over antiplatelets. Two large studies failed to demonstrate benefit with dabigatran or rivaroxaban respectively over aspirin,^2,3 as has preliminary data from a study using apixaban in an enriched cohort.⁴ The lower stroke recurrence rate in the dabigatran arm of RESPECT-ESUS³ was not statistically significant, though in several post-hoc subgroup analyses of this and NAVIGATE-ESUS,² superiority of anticoagulation compared with aspirin was statistically significant,^5–7 including among patients with LV dysfunction,⁵ patients above the age of 75 or with renal impairment.⁷ It is therefore plausible that a subgroup of ESUS patients can be identified based on clinical markers, who may benefit from anticoagulation.

Atrial fibrillation (AF) is classically associated with 25% of all strokes,^8,9 though was found to be as high as 36% in one metropolitan study.¹⁰ The lack of clear temporal association between episodes of AF and stroke^11,12 has led to the hypothesis that AF-related stroke may be caused by atrial cardiopathy, and subsequent AF may be a consequence of this already-developing process. While identifying these patients is a priority, current methods are time-consuming, resource intensive and at times invasive.¹³ There is therefore an urgent unmet clinical need to find non-invasive and inexpensive biomarkers that can be used to identify patients who are at most benefit from rigorous investigations for AF, or empirical anticoagulation.

Several ECG patterns have been associated with risk of AF. These include atrial premature complexes (APCs) which are markers of atrial cardiopathy, and P wave terminal force V1 (PTFV1), interatrial block (IAB) (subcategorised to advanced IAB and partial IAB), PR prolongation and abnormal P wave axis (aPWA), which reflect an altered atrial conduction pathway. PTFV1 also occurs in the setting of pressure overload.^14,15

We conducted a systematic review that identified numerous ECG patterns that are predictive of AF.¹⁶ Following on from this, we have conducted a meta-analysis. There have been previous meta-analyses for limited individual ECG abnormalities and risk of AF, however many of these have focussed on populations with significant cardiac comorbidities. We aimed to present a comprehensive and comparative meta-analysis focussing on patients presenting either post-stroke or to primary care.

Methods

Search methodology

Our search was conducted on EMBASE and Medline using key search terms consisting of ‘electrocardiography’, ‘atrial fibrillation’, ‘risk factor’, ‘predictive value’ (see Supplemental Appendix 1 for full search terminology). Further studies were identified from the references of selected articles. We limited our search to articles published in English. Following an initial search, abstracts were screened to determine suitability for inclusion.

Inclusion criteria

We included studies published from January 2014 until September 2022, with the former date chosen due to the development of the term ESUS in that year.¹ We limited the search to prospective, retrospective and case-control cohorts where the exposure of interest included a single 12-lead ECG taken in hospital or community with the primary outcome being incidence of new AF, and where exposure (abnormality ‘present’ or ‘absent’) was measured as odds ratio (OR), hazard ratio (HR) or relative risk (RR). We included cohorts presenting following stroke as well as those recruited from primary care or employment health screening.

Exclusion criteria

We excluded studies of known AF or other heart disease. We also excluded studies which relied on exposures other than a 12 lead ECG (e.g. Holter monitor) or where the ECG was studied as a continuous variable. We also excluded study populations taken from a primarily cardiology cohort, as these patients were not felt to be representative of our population of interest and may have cardiac comorbidities that may independently influence our outcome measure.

Our search identified studies assessing multiple ECG patterns for risk of AF, including PTFV1, aIAB, IAB, aPWA, APCs, PR prolongation, P wave amplitude, P wave area, P wave dispersion, QRS prolongation, left ventricular hypertrophy, QRS-T angle, ST-T segment abnormalities and QTc prolongation. We focussed on ECG patterns associated with atrial and AV nodal pathology. An ECG pattern was included for meta-analysis if there were a minimum of two studies with key data. Key data that were screened included study type, ECG pattern and method of measuring this, number of participants, follow-up duration, quantified risk of AF and potential biases. If key data were not available in the paper, the authors were contacted for this information. Where key data remained unavailable, studies were excluded from meta-analysis. Information was collected in accordance with the Guidelines for Meta-analyses and systematic reviews of observational studies for statement and the preferred reporting items for systematic reviews and meta-analyses (PRISMA) for screening studies. The quality criteria included an assessment for bias and all studies had a quality score of good or fair.

Quality assessment and reducing risk of bias

Quality assessment was performed on all studies utilising the Newcastle Ottawa quality assessment scale according to (1) selection of patients, (2) comparability of groups or cohorts and (3) exposure evaluation for case control studies and outcome evaluation for cohort studies. This was performed independently by two investigators (AB and LB).

Data collection process

Data were entered in pre-designed spreadsheets using Microsoft Excel. Abstracts were retrieved as complete manuscripts and assessed against the inclusion criteria. Data extracted included publication details (author, year), study design, follow-up duration, endpoints, quality score, characteristics of the population and sample size. Two reviewers (AB and LB) independently reviewed the data, with disagreements resolved with input from a third reviewer (RS). Principally, information was extracted according to PRISMA guidelines and the PRISMA 2020 checklist was used. Data was then arranged by ECG pattern.

Statistical analysis

We performed separate random effect meta-analyses to derive summary estimates of the pooled risk ratios of AF for each ECG abnormality versus the reference category of ‘no ECG abnormality’. We calculated the I² statistic to investigate heterogeneity. Funnel plots (Begg and Mazumdar) were generated to investigate publication bias. Finally, where necessary we undertook meta-regression of confounding covariates of biological plausibility or with differences between the studies (average follow up, age, hypertension prevalence, diabetes, heart failure). All analyses were performed using STATA 17.0 (StataCorp LP, TX) and R studio 2022.12.0.

Results

The ECG abnormalities included in this metanalysis were P wave terminal force V1 (PTFV1), partial interatrial block (pIAB), interatrial block (IAB) and advanced interatrial block (aIAB), abnormal P wave axis (aPWA), PR prolongation and atrial premature complexes (APCs). These are defined below. Studies were included if parameters were provided allowing comparison between studies (Figure 2). Studies measuring APCs either used a 10 s or a 15 s strip.

We identified 46 publications reporting risk of developing AF based on analyses of ECG patterns. We screened the articles to only include a categorical variable (i.e. one that can be included in a risk prediction score) and where sufficient information was reported or available from the authors, yielding 22 separate articles (Figure 1). The relevant ECG patterns and the number of publications were as follows; P wave terminal force V1 (PTFV1) 5,^17–21 Interatrial block (IAB) 8,^{17,18,20,22–27} advanced interatrial block (aIAB) 4,^18,20,28,29 abnormal P wave axis (aPWA) 7,^{17,20,27,30–33} PR prolongation 5,^{17,20,22,34,35} Atrial premature complexes (APCs) 4,^35–38 (Figure 2). The pooled number of patients was 609,496 patients (22,947 patients for PTFV1, 30,720 for IAB, 26,834 for aIAB, 34,923 for aPWA, 94,662 for PR prolongation and 144,727 for APCs).

Figure 1.

Systematic review search strategy.

Figure 2.

ECG abnormalities included in the study: (a) aIAB was defined as P-wave duration >120 ms with biphasic P-waves in leads II, III and aVF, (b) IAB was defined as P-wave duration >120 ms with one or more biphasic P-waves, (c) PTFV1 is the area of the negative deflection of the P-wave in lead V1 × duration of this negative deflection (abnormal >4000 µV × ms), (d) PR prolongation defined as >200 ms, (e) APC and (f) APWA defined as P-wave axis outside 0°–75°.

The patient and data characteristics of each study are tabulated in Table 1.

Table 1.

Studies included in meta-analysis.

ECG features	Author	Population; (Heart failure/HTN/Diabetes %)	Study	Sample size	Follow up	AF (marker present)	AF (marker absent)	OR/HR	Newcastle-Ottawa Quality score (good, fair, poor)
P wave terminal force V1 (PTFV1)
PTFV1	Lehtonen et al.¹⁷	Community registry, Finland (representative) (3/16/6)	Prospective cohort	5813	11.9 years	82 (9.4%)	330 (6.7%)	HR 1.43; 95% CI 1.07–1.91, p = 0.02	Good
PTFV1	Eranti et al.¹⁸	Community registry, Finland (8/16/5)	Retrospective cohort	5489	10 years	36 (4%)	88 (2%)	HR 2.23; 95% CI 1.41–3.53, p = 0.001	Fair
PTFV1	Eranti et al.¹⁹	Community registry, Finland (3/51/1)	Retrospective cohort	10,647	35–41 years	148 (19%)	1458 (15%)	HR 1.91 (1.34–2.73 p < 0.001), p < 0.05	Fair
PTFV1	Kreimer et al.²⁰	ESUS cohort, Germany (3/70/17)	Retrospective cohort	366	2 years	51 (44%)	65 (26%)	HR 5.297 (95% CI 3.249–8.636), p < 0.001	Good
PTFV1	Rasmussen et al.²¹	Copenhagen Holter Registry (community registry, Denmark) (0/28/11)	Retrospective cohort	632	14.7 years	34 (11%)	34 (10%)	HR 1.073 (95% CI 0.79–1.34), p < 0.05	Good
Interatrial block (IAB)
IAB	Smith et al.²²	Atherosclerosis Risk in Communities (ARIC) community registry, USA (4/30/11)	Retrospective cohort	14,924	21.2 years	166 (24%)	1819 (13%)	HR 1.92 (95% CI 1.64–2.26)	Good
IAB	Wu et al.²³	Admitted to medical wards, China (8/38/29)	Retrospective cohort	1571	4.8 years	110 (20%)	22 (1%)	HR 8.66; 95% CI 5.27–14.23, p < 0.001	Good
IAB	Boccanelli et al.²⁴	Recruited from community, Italy (6/59/17)	Prospective cohort	1626	6.6 years	34 (8%)	66 (5%)	HR 1.50 (95% CI 0.99–2.29) – HR 3.05 (95% CI 1.51–6.18, p = 0.002)	Fair
IAB	Lehtonen et al.¹⁷	Community registry, Finland (3/16/6)	Prospective cohort	5813	11.9 years	77 (14%)	335 (6%)	HR 1.43 (95% CI 1.07–1.91), p = 0.02	Good
IAB	Kreimer et al.²⁰	ESUS cohort, Germany (3/70/17)	Retrospective cohort	366	2 years	35 (32%)	40 (15%)	HR 2.437; CI 1.546–3.839, p < 0.001	Good
IAB	Skov et al.²⁵	Community registry, Denmark (0/28/11)	Retrospective cohort	152,759	10 years	13 (20%)	55 (10%)	HR 1.25 (95% CI 1.19–1.30), p < 0.05	Good
IAB	Eranti et al.¹⁸	Community registry, Finland (8/16/5)	Retrospective cohort	5489	10 years	43 (6%)	81 (2%)	RR 4.43 (95% CI 2.89–6.8), p < 0.001	Fair
IAB	Ariyarajah et al.²⁶	Admitted to general medical ward, USA (21/69/24)	Prospective cohort	118	1 year	12 (29%)	7 (9%)	RR 4.2 (95% CI 1.2–14.4), p < 0.05	Fair
IAB	Li et al.²⁷	ESUS cohort, Singapore (3/75/36)	Retrospective cohort	181	2.1 years	5 (10%)	9 (7%)	RR 1.57 (95% CI 0.95–4.25)	Fair
Advanced Interatrial block (aIAB)
aIAB	Kreimer et al.²⁰	ESUS cohort, Germany (3/70/17)	Retrospective cohort	366	2 years	11 (73%)	64 (18%)	HR 5.014 (CI 2.638–9.528), p < 0.001	Good
aIAB	Istolahti et al.²⁸	Community registry, Finland (N/A/40/6)	Retrospective cohort	6354	15 years	27 (43%)	511 (8%)	HR 1.39 (95% CI 1.09–1.77), p < 0.05	Good
aIAB	O’Neal et al.²⁹	ARIC community registry, USA (4/30/11)	Retrospective cohort	14,625	18.6 years	100 (39%)	1829 (13%)	HR 3.09 (95% CI 2.51–3.79), p < 0.05	Good
aIAB	Eranti et al.¹⁸	Community registry, Finland (8/16/5)	Retrospective cohort	5489	10 years	11 (11%)	113 (2%)	RR 7.62 (95% CI 4.06–14.31), p < 0.001	Good
Abnormal P wave axis (aPWA) ⁵
aPWA	Maheshwari et al.³⁰	ARIC community registry, USA (5/31/12)	Retrospective cohort	15,102	20.2 years	494 (40%)	2124 (15%)	RR 2.34 (95% CI 2.12–2.58), p < 0.05	Good
aPWA	Rangel et al.³¹	Cardiovascular Health Study (CHS), Community registry, USA (3/63/14)	Retrospective cohort	4274	12.1 years	342 (30%)	932 (30%)	HR 1.17 (95% CI 1.03–1.33) p = 0.023	Fair
aPWA	Dhaliwal et al.³²	Recruited from community, diabetes (35/83/100)	Retrospective cohort	8965	4.9 years	31 (5%)	114 (1%)	HR 2.65 (95% CI 1.76–3.99), p < 0.0001	Fair
aPWA	Acampa et al.³³	Cryptogenic stroke cohort, Italy (4/58/19)	Prospective cohort	222	7 days	13 (39%)	31 (16%)	OR 3.31 (95% CI 1.49–7.35), p < 0.05	Good
aPWA	Lehtonen et al.¹⁷	Community registry, Finland (representative) (3/16/6)	Prospective cohort	5813	11.9 years	59 (10%)	353 (7%)	N/A	Good
aPWA	Li et al.²⁷	ESUS cohort, Singapore (3/75/36)	Retrospective cohort	181	2.1 years	1 (9%)	13 (8%)	N/A	Good
aPWA	Kreimer et al.²⁰	ESUS cohort, Germany (3/70/17)	Retrospective cohort	366	2 years	10 (33%)	65 (19%)	N/A	Good
PR prolongation ⁷
PR prolongation	Lehtonen et al.¹⁷	Community registry, Finland (3/16/6)	Prospective cohort	5813	11.9 years	43 (22%)	369 (7%)	HR 1.59 (95% CI 1.05–2.41), p < 0.05	Good
PR prolongation	Smith et al.²²	Atherosclerosis Risk in Communities (ARIC) community registry, USA (4/30/11)	Retrospective cohort	14,924 (ARIC)	21.2 years	174 (17%)	1811 (13%)	HR 1.19 (95% CI 1.02–1.4), p < 0.05	Good
PR prolongation	Cheng et al.³⁴	Framingham Heart Study (FHS) community registry, USA (9/33/2)	Prospective cohort	7575	20 years	25 (20%)	456 (6%)	HR 2.06; 95% CI 1.36–3.12, p < 0.0001	Good
PR prolongation	Hamada and Muto³⁵	Community cohort, Japan (1/13/3)	Retrospective cohort	96,841	7 years	14 (1%)	335 (1%)	HR 1.77 (95% CI 1.04–3.03) p = 0.036	Good
PR prolongation	Kreimer et al.²⁰	ESUS cohort, Germany (3/70/17)	Retrospective cohort	366	2 years	30 (28%)	45 (17%)	HR 1.986 (CI 1.249–3.156), p < 0.05	Good
Supraventricular extrasystoles ⁵
APC (1 or more)	Ntaios et al.³⁶	ESUS cohort, Greece (N/A/43/19)	Retrospective cohort	853	3.4 years	69 (31%)	56 (9%)	HR 1.8 (95% CI 1.06–3.05)	Good
(APC) (1 or more)	O’Neal et al.³⁷	Community registry (REGARDS study), USA (17/46/16)	Retrospective cohort	13,840	9.4 years	139 (15%)	876 (7%)	OR 1.92 (95% CI 1.57–2.35), p < 0.05	Good
APC (1 or more)	Murakoshi et al.³⁸	Community registry, Japan (n/a/19/5)	Prospective cohort study	63,197	3.4 years	112 (3%)	274 (0.5%)	RR 4.87 (95% CI 3.61–6.57), p < 0.05	Fair
APC (1 or more)	Hamada and Muto³⁵	Community cohort, Japan (1/13/3)	Retrospective cohort	96,841	7 years	21 (4%)	540 (1%)	HR 5.09 (95% CI 3.26–7.95) p < 0.001	Fair

Four studies were conducted in post-stroke populations, while 18 studies included community-based cohorts. One study included patients with diabetes. Eighteen studies were in Western (Caucasian) populations and four were from Eastern (Asian) populations. While rates of heart failure were between 3% and 8% in across most populations, there was heterogeneity in hypertension (rates between 13% and 75%) and diabetes (between 6% and 36%) across studies. Median follow up was 11.9 years (IQR 4.9–15). AF incidence was documented using an ECG in 21 studies and using a loop recorder in one study. Diagnosis of AF was made according to ICD criteria in 20 studies. In one study, it was defined as 5 min or longer on ECG monitoring, and in another study, 30 s or longer on implantable loop recorder.^20,33

P wave terminal force V1 (PTFV1)

Five studies were identified (n = 22,947) with PTFV1 defined as >4000 µV × ms. Median follow up was 11.9 years (IQR 10–14.7). Across the studies, the median age was 51.6 (IQR 50.3–62) and 52.19% were male. In total, 2326 patients developed AF over the study period (10.14%), 12% with PTFV1 and 9.86% in the control arm. The pooled risk ratio of developing AF in patients with PTFV1 was 1.48 (95% CI 1.04–2.10) (Figure 3). I² for heterogeneity was 67%. Funnel plots showed Eggers text p value 0.003 (see Supplemental Appendix).

Figure 3.

Pooled relative risks and forest plots: (a) aIAB, (b) IAB, (c) APCs, (d) aPWA, (e) PR prolongation and (f) PTFV1.

Interatrial block (IAB) and advanced interatrial block (aIAB)

Nine studies were identified assessing IAB (n = 30,720) and four studies for aIAB (n = 26,834). IAB was defined as pIAB with one or two biphasic P waves in leads II, III and aVF and aIAB was defined as pIAB with biphasic P waves in leads II, III and aVF. Median follow up was 6.6 years for IAB (IQR 2.1–11.9) and 12.5 years for aIAB (IQR 8–15.9). Across the studies, median age was 62 for IAB (54–65) and 57 for aIAB (52–62). 2929 patients with IAB and 2666 with aIAB developed AF during the study; this was a total of 9.5% and 9.9% across the two sets of studies. Among patients with IAB, 15% developed AF compared with 8.9% controls. Among those with aIAB this increased to 34% compared with 9.5% controls. Pooled risk ratios were 2.54 (95% CI 1.64–3.93) and 4.05 (95% CI 2.64–6.22) (Figure 3). I² test for heterogeneity was 86% for IAB and 75% for aIAB. Funnel plots showed Eggers test p value 0.288 for aIAB and 0.003 for IAB.

Abnormal P wave axis (outside 0°–75°)

Seven studies were identified (n = 34,923). Median follow up across studies was 4.9 years (IQR 2–12) and median age was 63 (58–68). Across the studies, 4882 patients developed AF during the study period (13%). 25.6% of patients with aPWA developed AF compared to 11.6% in the control group. Pooled risk ratio for development of AF was 1.89 (95% CI 1.25–2.85) (Figure 3). I² test for heterogeneity was 97%. Funnel pots showed Eggers test p value 0.01.

PR prolongation (>200 ms)

Five studies were identified (n = 94,662). Median follow up was 11.9 years (IQR 7–20) and median age was 52. Overall, 3302 patients developed AF during the study period (3.5%). Among patients with PR prolongation, this was 11.4% compared with 3.3% among controls. Pooled risk ratio for development of AF was 2.22 (95% CI 1.27–3.87) with I² for heterogeneity 92% (Figure 3). Funnel plots showed Eggers test p value 0.0004.

Supraventricular extrasystoles

Five studies were identified (n = 144,727) with median follow up 7 years (IQR 3.4–9.4) and median age 63. Overall, 2217 patients developed AF (1.5%). Among patients with at least one APC, 6.7% developed AF compared to 1.3% in the control group. Pooled risk ratio was 3.71 (95% CI 2.23–6.16) with I² test for heterogeneity 93% (Figure 3) and funnel plot Eggers test p value 0.047.

Effects of confounders on predicting AF

Meta regression was performed examining the effect of rates of heart failure, hypertension, sex, age, follow-up time and diabetes on the risk of AF for each of the ECG changes. Only the presence of diabetes attenuated the effect size of PR prolongation on the subsequent risk of AF (p = 0.0183).

Discussion

Our meta-analysis of 22 studies, including 609,496 patients from primary care or post-stroke cohorts, identified six ECG patterns independently associated with risk of subsequent AF. We found the most predictive ECG abnormality for the occurrence of AF was aIAB (RR 4.05, 95% CI 2.64–6.22), followed by presence of at least one atrial premature complex (RR 3.71, 95% CI 2.23–6.16). Presence of IAB, PR prolongation, aPWA or PTFV1 also conferred increased risk of subsequent AF. Our meta-analysis builds on previous studies by improving the statistical power and by providing a comparison of the statistical powers of different ECG abnormalities. Furthermore, we used meta-regression to adjust for confounders. This demonstrated that diabetes reduces the predictive value of PR prolongation, however none of the other demographics had a significant confounder effect.

Our findings support previous literature, demonstrating the predictive value of PTFV1, IAB, aIAB and PR prolongation in AF.^14,39,40 However, much of the previous literature included heterogeneous demographics with several studies including patients with significant cardiac comorbidities, which may have influenced the outcome measure. Our findings demonstrate that these ECG abnormalities remain significant predictors of AF when restricted to community-based or post-stroke populations. We also demonstrated that the utility of PR prolongation as a predictor of AF was reduced in the presence of diabetes. This is an interesting result as both PR prolongation and diabetes mellitus are predictors of AF, and diabetes has been recently associated with AV-nodal block,⁴¹ though it may suggest that pathophysiology of AF in diabetes is independent of AV-nodal dysfunction.⁴²

Further, we add to the previous literature by demonstrating that abnormal P wave axis and APCs on a 12-lead ECG are predictors of AF. While APCs have previously been found to predict AF on a Holter monitor, from previous studies it was not clear that their presence on a 12-lead ECG was a significant predictor.⁴³ In our study, the presence of at least one APC on an ECG conferred a four-times increased risk of AF.

In contrast to previous studies that have suggested age and hypertension may increase the predictive value of IAB^17,44 we did not demonstrate any other significant effect from age, diabetes, heart failure or hypertension on risk prediction in our study through metaregression.

PTFV1 has previously been recognised as a consistent marker of AF⁴⁵ and in a study comparing the predictive value of AF between values, PTFV1 was among the strongest predictors. In our study this association was comparatively weaker and it is possible that a cut-off for abnormal PTFV1 > 5000 µV × ms as has been used in some studies⁴⁶ would have a stronger predictive value.

Our study has several strengths, including being the first study to consider multiple ECG abnormalities together in their predictivity of subsequent AF. We did not include studies of poor quality based on the Newcastle-Ottawa scoring system. We included a large sample size from multiple cohorts and different continents, improving its generalisability. We excluded patients recruited from primarily cardiac cohorts, who would have unique characteristics influencing their risk of AF. We also included data on other risk factors that may have influenced the predictive value of the ECG abnormalities in question through meta regression.

Our study has limitations. We included patients recruited from employment health screening or primary care populations, in addition to post stroke populations. This was done as the overall number of studies using post stroke cohorts is small, and it was felt that the data from these cohorts is relevant to a post stroke cohort. This has however introduced heterogeneity to our study, which was reflected in high I² numbers across our study. Due to paucity of post-stroke studies, we were unable to compare strength of association between these two populations. Heterogeneity is increased by small variations in diagnosis of AF, with two studies employing prolonged ECG monitoring. Furthermore, diagnosis of AF was through convenience ECG sample in most studies, which likely underestimated the incidence of AF.

We have identified several ECG patterns that can be used to predict AF. APC and aIAB are more predictive than the other ECG patterns studied, followed by IAB, PR prolongation, aPWA and PTFV1. There is a need to identify enriched ESUS cohorts, and biomarkers have been suggested to guide this in clinical use and/or for research purposes.⁴⁷ In the setting of a patient presenting following a stroke, with an ECG in sinus rhythm, ECG patterns identified in this meta-analysis could be developed into a computed clinical risk score, combined with clinical factors, to identify one such enriched cohort of ESUS patients who will benefit from intensified cardiac monitoring, or to guide further studies into whether they may benefit from anticoagulation. Negative/neutral ESUS trials create an opportunity to reframe embolic stroke of undetermined source. Rather than assuming that ESUS is undetected AF, future trials could consider ESUS subgroups, with ECG changes high-risk for subsequent AF forming one such subgroup to guide future research. Collaborative prospective observational cohorts investigating the natural history of subsequent stroke could identify the ESUS subgroups where antiplatelet therapy is ineffective to help inform future ESUS trials.

Research Data

sj-docx-1-eso-10.1177_23969873231172559 – Supplemental material for Predicting risk of AF in ischaemic stroke using sinus rhythm ECG abnormalities: A meta-analysis

sj-docx-1-eso-10.1177_23969873231172559 for Predicting risk of AF in ischaemic stroke using sinus rhythm ECG abnormalities: A meta-analysis by Alexander Berry-Noronha, Luke Bonavia, Duncan Wilson, Antti Eranti, Maria Uggen Rasmussen, Ahmad Sajadieh, Fabienne Kreimer, Michael Gotzmann and Ramesh Sahathevan in European Stroke Journal

This article is distributed under the terms of the Creative Commons Attribution 4.0 License (https://creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).

Research Data

sj-docx-2-eso-10.1177_23969873231172559 – Supplemental material for Predicting risk of AF in ischaemic stroke using sinus rhythm ECG abnormalities: A meta-analysis

sj-docx-2-eso-10.1177_23969873231172559 for Predicting risk of AF in ischaemic stroke using sinus rhythm ECG abnormalities: A meta-analysis by Alexander Berry-Noronha, Luke Bonavia, Duncan Wilson, Antti Eranti, Maria Uggen Rasmussen, Ahmad Sajadieh, Fabienne Kreimer, Michael Gotzmann and Ramesh Sahathevan in European Stroke Journal

Supplemental Material

sj-docx-3-eso-10.1177_23969873231172559 – Supplemental material for Predicting risk of AF in ischaemic stroke using sinus rhythm ECG abnormalities: A meta-analysis

Supplemental material, sj-docx-3-eso-10.1177_23969873231172559 for Predicting risk of AF in ischaemic stroke using sinus rhythm ECG abnormalities: A meta-analysis by Alexander Berry-Noronha, Luke Bonavia, Duncan Wilson, Antti Eranti, Maria Uggen Rasmussen, Ahmad Sajadieh, Fabienne Kreimer, Michael Gotzmann and Ramesh Sahathevan in European Stroke Journal

Research Data

sj-pptx-4-eso-10.1177_23969873231172559 – Supplemental material for Predicting risk of AF in ischaemic stroke using sinus rhythm ECG abnormalities: A meta-analysis

sj-pptx-4-eso-10.1177_23969873231172559 for Predicting risk of AF in ischaemic stroke using sinus rhythm ECG abnormalities: A meta-analysis by Alexander Berry-Noronha, Luke Bonavia, Duncan Wilson, Antti Eranti, Maria Uggen Rasmussen, Ahmad Sajadieh, Fabienne Kreimer, Michael Gotzmann and Ramesh Sahathevan in European Stroke Journal

Footnotes

Acknowledgements

None

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Ethical approval

Not applicable

Informed consent

Not applicable

Guarantor

Alexander Berry-Noronha (AB)

Contributorship

Alexander Berry-Noronha: conception of work, data acquisition, analysis, drafting and revision. Luke Bonavia: data-acquisition, analysis and revision. Duncan Wilson: analysis and revision. Antti Eranti: data-acquisition and revision. Maria Uggen Rasmussen: data-acquisition and revision. Ahmad Sajadieh: data-acquisition and revision. Fabienne Kreimer: data-acquisition and revision. Michael Gotzmann: data-acquisition and revision. Ramesh Sahathevan: conception and revision.

ORCID iDs

Alexander Berry-Noronha

Fabienne Kreimer

Data availability statement

Further data is available from the authors on request.

Supplemental material

Supplemental material for this article is available online.

References

Hart

Diener

Coutts

, et al. Embolic strokes of undetermined source: the case for a new clinical construct. Lancet Neurol 2014; 13: 429–438.

Hart

Sharma

Mundl

, et al. Rivaroxaban for stroke prevention after embolic stroke of undetermined source. N Engl J Med 2018; 378: 2191–2201.

Diener

Sacco

Easton

, et al. Dabigatran for prevention of stroke after embolic stroke of undetermined source. N Engl J Med 2019; 380: 1906–1917.

Poli

Keller

Martus

, et al. Abstract 31: the ATTICUS Randomized Controlled Trial - subgroup analyses. Stroke 2023; 54: A31.

Merkler

Pearce

Kasner

, et al. Left ventricular dysfunction among patients with embolic stroke of undetermined source and the effect of rivaroxaban vs aspirin: a subgroup analysis of the NAVIGATE ESUS randomized clinical trial. JAMA Neurol 2021; 78: 1454–1460.

Toyoda

Uchiyama

Hagihara

, et al. Dabigatran vs. aspirin for secondary prevention after embolic stroke of undetermined source - Japanese subanalysis of the RE-SPECT ESUS randomized controlled trial. Circ J 2020; 84: 2286–2295.

Diener

Sacco

Easton

, et al. Antithrombotic treatment of embolic stroke of undetermined source: RE-SPECT ESUS elderly and renally impaired subgroups. Stroke 2020; 51: 1758–1765.

Cotter

Martin

Ring

, et al. Incidence of atrial fibrillation detected by implantable loop recorders in unexplained stroke. Neurology 2013; 80: 1546–1550.

Gladstone

Spring

Dorian

, et al. Atrial fibrillation in patients with cryptogenic stroke. N Engl J Med 2014; 370: 2467–2477.

10.

Leyden

Kleinig

Newbury

, et al. Adelaide stroke incidence study: declining stroke rates but many preventable cardioembolic strokes. Stroke 2013; 44: 1226–1231.

11.

Guichard

Nattel

. Atrial cardiomyopathy: a useful notion in cardiac disease management or a passing fad? J Am Coll Cardiol 2017; 70: 756–765.

12.

Brambatti

Connolly

Gold

, et al. Temporal relationship between subclinical atrial fibrillation and embolic events. Circulation 2014; 129: 2094–2099.

13.

Diamantopoulos

Sawyer

Lip

, et al. Cost-effectiveness of an insertable cardiac monitor to detect atrial fibrillation in patients with cryptogenic stroke. Int J Stroke 2016; 11: 302–312.

14.

Tse

Wong

Gong

, et al. Predictive value of inter-atrial block for new onset or recurrent atrial fibrillation: a systematic review and meta-analysis. Int J Cardiol 2018; 250: 152–156.

15.

Win

Venkatesh

Volpe

, et al. Associations of electrocardiographic P-wave characteristics with left atrial function, and diffuse left ventricular fibrosis defined by cardiac magnetic resonance: the PRIMERI study. Heart Rhythm 2015; 12: 155–162.

16.

Berry-Noronha

Bonavia

Song

, et al. “P wave predictors of AF: a systematic review”; STROKE 2022, 31st Annual Scientific Meeting of the Stroke Society of Australasia 2022, 31st August - 2nd September 2022, Christchurch, New Zealand. Int J Stroke 2022; 17: 3–39.

17.

Lehtonen

Langén

Porthan

, et al. Electrocardiographic predictors of atrial fibrillation in nonhypertensive and hypertensive individuals. J Hypertens 2018; 36: 1874–1881.

18.

Eranti

Carlson

Kenttä

, et al. Orthogonal P-wave morphology, conventional P-wave indices, and the risk of atrial fibrillation in the general population using data from the Finnish Hospital Discharge Register. Europace 2020; 22: 1173–1181.

19.

Eranti

Aro

Kerola

, et al. Prevalence and prognostic significance of abnormal P terminal force in lead V1 of the ECG in the general population. Circ Arrhythm Electrophysiol 2014; 7: 1116–1121.

20.

Kreimer

Aweimer

Pflaumbaum

, et al. Impact of P-wave indices in prediction of atrial fibrillation—insight from loop recorder analysis. Ann Noninvasive Electrocardiol 2021; 26: e12854.

21.

Rasmussen

Kumarathurai

Fabricius-Bjerre

, et al. P-wave indices as predictors of atrial fibrillation. Ann Noninvasive Electrocardiol 2020; 25: e12751.

22.

Smith

O’Neal

Shoemaker

, et al. PR-Interval components and atrial fibrillation risk (from the Atherosclerosis Risk in communities study). Am J Cardiol 2017; 119: 466–472.

23.

Wang

Chu

, et al. Usefulness of a combination of interatrial block and a high CHADS₂ score to predict new onset atrial fibrillation. Int Heart J 2016; 57: 580–585.

24.

Boccanelli

Mureddu

Cesaroni

, et al. Predictive value of interatrial block for atrial fibrillation in elderly subjects enrolled in the PREDICTOR study. J Electrocardiol 2019; 54: 22–27.

25.

Skov

Ghouse

Kühl

, et al. Risk prediction of atrial fibrillation based on electrocardiographic interatrial block. J Am Heart Assoc 2018; 7: e008247.

26.

Ariyarajah

Apiyasawat

Fernandes

, et al. Association of atrial fibrillation in patients with interatrial block over prospectively followed controls with comparable echocardiographic parameters. Am J Cardiol 2007; 99: 390–392.

27.

TYW

Yeo

LLL

JSY

, et al. Association of Electrocardiographic P-wave markers and atrial fibrillation in embolic stroke of undetermined source. Cerebrovasc Dis 2021; 50: 46–53.

28.

Istolahti

Eranti

Huhtala

, et al. The prevalence and prognostic significance of interatrial block in the general population. Ann Med 2020; 52: 63–73.

29.

O’Neal

Zhang

Loehr

, et al. Electrocardiographic advanced interatrial block and atrial fibrillation risk in the general population. Am J Cardiol 2016; 117: 1755–1759.

30.

Maheshwari

Norby

Soliman

, et al. Refining prediction of atrial fibrillation risk in the general population with analysis of P-wave axis (from the Atherosclerosis Risk in communities study). Am J Cardiol 2017; 120: 1980–1984.

31.

Rangel

O’Neal

Soliman

. Usefulness of the electrocardiographic P-wave axis as a predictor of atrial fibrillation. Am J Cardiol 2016; 117: 100–104.

32.

Dhaliwal

Upadhya

Soliman

, et al. Association of P-wave axis with incident atrial fibrillation in diabetes mellitus (from the ACCORD trial). Am J Cardiol 2020; 128: 191–195.

33.

Acampa

Lazzerini

Guideri

, et al. Electrocardiographic predictors of silent atrial fibrillation in cryptogenic stroke. Heart Lung Circ 2019; 28: 1664–1669.

34.

Cheng

Keyes

Larson

, et al. Long-term outcomes in individuals with prolonged PR interval or first-degree atrioventricular block. JAMA 2009; 301: 2571–2577.

35.

Hamada

Muto

. Simple risk model and score for predicting of incident atrial fibrillation in Japanese. J Cardiol 2019; 73: 65–72.

36.

Ntaios

Perlepe

Lambrou

, et al. Supraventricular extrasystoles on standard 12-lead electrocardiogram predict new incident atrial fibrillation after embolic stroke of undetermined source: the AF-ESUS Study. J Stroke Cerebrovasc Dis 2020; 29: 104626.

37.

O’Neal

Kamel

Judd

, et al. Usefulness of atrial premature complexes on routine electrocardiogram to determine the risk of atrial fibrillation (from the REGARDS study). Am J Cardiol 2017; 120: 782–785.

38.

Murakoshi

Sairenchi

, et al. Prognostic impact of supraventricular premature complexes in community-based health checkups: the Ibaraki Prefectural Health Study. Eur Heart J 2015; 36: 170–178.

39.

Huang

Zheng

, et al. Predictive value of P wave terminal force in lead V1 for atrial fibrillation: a meta-analysis. Ann Noninvasive Electrocardiol 2020; 25: e12739.

40.

Cheng

Huang

, et al. Electrocardiographic PR prolongation and atrial fibrillation risk: a meta-analysis of prospective cohort studies. J Cardiovasc Electrophysiol 2015; 26: 36–41.

41.

Rawshani

McGuire

Omerovic

, et al. Cardiac arrhythmias and conduction abnormalities in patients with type 2 diabetes. Sci Rep 2023; 13: 1192.

42.

Seyed Ahmadi

Svensson

Pivodic

, et al. Risk of atrial fibrillation in persons with type 2 diabetes and the excess risk in relation to glycaemic control and renal function: a Swedish cohort study. Cardiovasc Diabetol 2020; 19: 1–12.

43.

Himmelreich

JCL

Lucassen

WAM

Heugen

, et al. Frequent premature atrial contractions are associated with atrial fibrillation, brain ischaemia, and mortality: a systematic review and meta-analysis. Europace 2019; 21: 698–707.

44.

Cortez

Baturova

Lindgren

, et al. Atrial time and voltage dispersion are both needed to predict new-onset atrial fibrillation in ischemic stroke patients. BMC Cardiovasc Disord 2017; 17: 200.

45.

Yaghi

Kamel

Elkind

MSV

. Atrial cardiopathy: a mechanism of cryptogenic stroke. Expert Rev Cardiovasc Ther 2017; 15: 591–599.

46.

Kamel

Longstreth

Tirschwell

, et al. The AtRial Cardiopathy and Antithrombotic Drugs In prevention After cryptogenic stroke (ARCADIA) randomized trial: rationale and methods. Int J Stroke 2019; 14: 207–214.

47.

Rose

Kasner

. Forge AHEAD with stricter criteria in future trials of embolic stroke of undetermined source. Neural Regen Res 2022; 17: 1009.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB

0.02 MB

0.03 MB

0.31 MB