Left atrial strain predicts incident atrial fibrillation in the general population: the Copenhagen City Heart Study

Abstract

Background

Left atrial (LA) strain parameters have been demonstrated to be valuable predictors of atrial fibrillation (AF) in several patient cohorts. The purpose of this study was to investigate whether LA strain, assessed by two-dimensional speckle-tracking echocardiography, can be used to predict the development of AF in the general population.

Methods and results

This prospective longitudinal study included 4466 participants from the fifth Copenhagen City Heart Study. All participants underwent a health examination, including echocardiographic measurements of LA strain. Participants with prevalent AF at baseline were excluded. The primary endpoint was incident AF. During a median follow-up period of 5.3 years, 154 (4.3%) participants developed AF. In univariable analysis, peak atrial longitudinal strain (PALS), peak atrial contraction strain (PACS), and LA strain during the conduit phase were significantly associated with the development of AF. PALS [hazard ratio (HR) 1.05, 95% confidence interval (CI) (1.03–1.07), P < 0.001, per 1% decrease] and PACS (HR 1.08, 95% CI (1.05–1.12), P < 0.001, per 1% decrease] remained independent predictors of AF in multivariable analysis. In addition, PALS and PACS remained significantly associated with AF development even in participants with normal-sized atria and normal left ventricular (LV) systolic function.

Conclusion

In the general population, PALS and PACS independently predict incident AF. These findings remained consistent even in participants with normal-sized LA and normal LV systolic function.

Open in new tabDownload slide

Introduction

Atrial fibrillation (AF) is the most common cardiac arrhythmia and is characterized by a quick, disorganized excitation of the atrium followed by an irregular activation of the ventricles.¹ AF is a common condition in the general population and is known to be associated with increased morbidity and mortality.² The prevalence of AF is estimated to be ∼3% in the general population.³ Cardiovascular disease is the most common cause of death in AF patients, and cerebral infarction is the most common specific disease.⁴ The consequences of stroke and cardiovascular disease can significantly reduce the quality of life of the affected individuals. In addition, undiagnosed AF is likely to be responsible for up to 25% of cryptogenic ischaemic strokes.⁵ Hence, to plan preventive strategies, it is of the utmost importance to identify early diagnostic signs indicating an increased risk of AF.

Since AF is a condition arising from the atrium, the association between AF and several echocardiographic measures of both atrial structure and function has been investigated. The maximum volume of the left atrium has already been validated as a marker for the development of AF in several studies.⁶ Additionally, the association between AF and left atrial (LA) emptying fraction, a measure of LA function rather than structure, has been studied extensively.^7–9 However, recent studies suggest that LA strain—a measure of LA wall deformation—is a stronger and more sensitive marker for the development of AF.^10–12 LA strain is measured using two-dimensional speckle-tracking echocardiography (2DSTE) analysis in which both the magnitude and rate of the myocardial deformation can be determined, and three strain parameters are obtained. Peak atrial longitudinal strain (PALS) is a measure of LA reservoir function, peak atrial contraction strain (PACS) represents the late diastolic contraction phase of the LA and LA strain during the conduit phase (LACS) is the difference between PALS and PACS and represents the passive filling of the left ventricular (LV) in early ventricular diastole.

Most of the studies investigating LA strain have been conducted on relatively small study samples and only in patients with prevalent heart disease. We hypothesize that PALS, PACS, and LACS are superior to conventional echocardiographic measurements in predicting AF. Hence, this study aimed to investigate whether LA strain can be used to predict the development of AF in a large general population cohort.

Methods

Population

This study is part of the fifth Copenhagen City Heart Study (CCHS5), which is a prospective longitudinal cohort study concerning cardiovascular disease and risk factors in the general population.¹³ The inclusion phase of CCHS5 ran from 2011 to 2015. The participants were randomly invited people at the age of 20–99 years. The inclusion process has been described in detail previously.¹⁴ Informed written consent was obtained from all participants. CCHS5 was conducted in accordance with the Second Helsinki Declaration and was approved by the regional scientific ethics committee.

Initially, 4466 patients were included. Exclusion criteria were prevalent AF during echocardiography (N = 51), prevalent AF at baseline (N = 192), and inadequate echocardiographic image quality (N = 719). As 86 participants had more than one exclusion criterium in common, the final study sample consisted of 3590 participants.

Health examination and baseline information

In the CCHS5, all participants underwent a physical examination and completed an extensive questionnaire. Hypertension, heart rate, the presence of ischaemic heart disease (IHD), and diabetes mellitus were defined as previously described.¹⁵ Systolic and diastolic blood pressure (BP) was measured with the London School of Hygiene sphygmomanometer. The concentrations of plasma blood glucose and cholesterol were measured on non-fasting venous blood samples. Obesity was defined as body mass index (BMI) ≥30 kg/m². Participants with moderate and severe valvular disease were excluded.

Transthoracic echocardiography

Echocardiography was performed by experienced sonographers using the Vivid 9 ultrasound system (GE Healthcare, Horten, Norway). All echocardiographies were analysed offline using commercially available post-processing software, EchoPAC version 202 (GE Healthcare). One investigator performed all analyses and was blinded to all clinical data including follow-up information.

Conventional echocardiography

LV ejection fraction (LVEF) was measured by the Simpson’s biplane method using a semi-automatic function in the apical four- and two-chamber views by delineating the endocardial border and tracking volume changes throughout the cardiac cycle as previously described.¹⁶ The investigator could adjust the endocardial border if necessary. LA volume was measured using the area-length method and indexed to body area to obtain the LA volume index (LAVi). Tricuspid annular plane systolic excursion was measured in the apical four-chamber view using M-mode. In the apical four-chamber view, peak transmitral early (E-wave) and late (A-wave) diastolic inflow velocity and deceleration time were measured using pulsed wave Doppler at the tips of the mitral valve leaflets. Peak early diastolic tissue velocity (e′) was obtained with pulsed wave tissue Doppler imaging from the septal and lateral wall at mitral annulus level. Mean value was used to calculate e′. E′ was indexed to E-wave velocity to obtain E/e′.

Two-dimensional speckle-tracking echocardiography

2DSTE analysis of the LA was performed in the apical four-chamber and two-chamber view. In each projection, the LA was divided into six regional segments (two annular, two mid-wall, and two basal segments) resulting in a global 12-segment model of the LA. For all patients, the QRS complex was chosen as the baseline reference point. The following LA strain parameters were measured: PALS, PACS, and LACS (Figure 1). PACS is measured at the electrocardiogram p-wave. The myocardium of the LA was tracked using a semi-automated function, which outlined the region of interest. Modifications to the region of interest could be made manually by the investigator in instances of inaccurate tracing. Global longitudinal strain was assessed as previously described.¹⁴

Outcome

The outcome was new incident of AF. The Danish National Patient Register was used to retrieve information on the development of AF using ICD-10 codes.¹⁷ Follow-up was 100%.

Statistical analysis

STATA statistics/data analysis, SE 16.1 (StataCorp, College Station, TX, USA) was used for all statistical analyses. All variables showing Gaussian distribution are displayed as mean values ± standard deviation (SD) in Table 1. Data were compared using analysis of variance (ANOVA) for Gaussian-distributed variables. Categorical variables were compared using Pearson’s χ² test and are reported as total numbers or percentages. All continuous non-Gaussian-distributed variables are shown as median value with interquartile range and were compared using Kruskal–Wallis test. Univariable and multivariable linear regression was utilized to assess the relationship between conventional cardiovascular risk factors and LA strain parameters. Age, male sex, BMI, physical activity level in leisure time, pack-years, systolic and diastolic BP, serum cholesterol, serum creatinine, diabetes, and history of previous IHD were included in the multivariable model. Univariable Cox proportional hazard regression models were used to assess the association between AF development and LA strain parameters. Multivariable models were subsequently used to control for possible confounding covariates (age, sex, heart rate, number of pack-years, cardiac valve disease, BMI, hypercholesterolaemia, diabetes, IHD, hypertension, global longitudinal strain (GLS), and E/e′). Additionally, a sensitivity analysis was conducted in which all participants with LAVi >34 mL/m², reduced LVEF (<50%), a history of IHD, or development of acute myocardial infarction prior to the development of AF during follow-up were excluded from the analyses. The association between the development of AF and LA strain parameters were visualized using restricted cubic spline models constructed using a Poisson model with the rates of events calculated as the number of events divided by person-time at risk. The number of knots for the restricted cubic splines was chosen according to the lowest Akaike information criterion.

Subgroup analyses with the population sub-grouped according to guideline-recommended cut-offs (LVEF < 50%, LV diastolic dysfunction grades, LV hypertrophy grades, LAVi > 35%) were carried out. Incremental value of LA strain parameters in relation to clinically used echocardiographic cut-offs (diastolic dysfunction, LAVi, LV hypertrophy, and abnormal LVEF) were carried out. To assess the incremental value of LA strain parameters when added to conventional echocardiographic cut-off values Harrel’s C-statistics were assessed and compared.

We assessed both intra- and interobserver reproducibility of LA strain measurements in 20 randomly selected patients by blinded investigators. Both intraclass correlation coefficient (ICC) and absolute biases (mean difference ± 1.96 SD) were assessed.

Results

Patient characteristics

A total of 3590 participants were included in this study. Mean age was 55.3 ± 17.4 years and 1551 (43.2%) were males. Mean PALS was 37.6 ± 12.1%, mean PACS was 17.2 ± 6.6%, and mean LACS was 20.6 ± 9.8%. Demographic, clinical, and echocardiographic characteristics for all participants and stratified by tertiles of PALS are displayed in Table 1. Lower PALS was associated with older age, male sex, higher BMI, a higher amount of pack-years, higher systolic and diastolic BP, higher prevalence of diabetes, higher cholesterol levels, higher prevalence of cardiac valve disease, higher creatinine, and higher glucose levels. Furthermore, the following echocardiographic parameters were associated with decreasing PALS: larger LV mass index, lower LVEF, higher E/e′, lower E/A, higher LAVi, and lower GLS.

The relationship between cardiovascular risk factors and PALS, PACS, and LACS

Table 2 shows univariable and multivariable linear regression models assessing the relationship between cardiovascular risk factors and LA strain parameters. Increasing age, BMI, pack-years, diastolic BP, diabetes, and previous IHD were all associated with reduced PALS after adjusting for potential confounders. Decreasing PACS was associated with increasing age, male sex, BMI, diastolic BP, cholesterol, and previous IHD. Increasing age, male sex, BMI, systolic BP, diastolic BP, diabetes, and previous IHD were associated with decreasing LACS.

LA strain parameters and development of AF

During a median follow-up period of 5.3 years, 154 (4.3%) participants developed AF. The prevalence of AF decreased significantly from the lowest to highest tertile of PALS (Table 1). The incidence rate of AF was 2.9 [95% confidence interval (CI) (1.9–4.6)], per 1000 person-years, 4.7 [95% CI (3.3–6.7)] per 1000 person-years, and 18.5 [95% CI (15.2–22.4)] per 1000 person-years in participants with PALS >41.1, PALS 32.0–42.1%, and PALS <32.0%, respectively.

Normal vs. abnormal LA strain

For patients with LA strain <23%, there was a 6.8 increased risk to develop AF related to those with LA strain ≥23%. Compared to the reference group (patients with LA strain ≥23%), the hazard ratio (HR) for the subgroup of patients with LA strain between 23% and 19% is 4.16 [95% CI (2.60–6.65)], the HR for the subgroup of patients with LA strain between 19% and 15% is 6.58 [95% CI (3.53–12.25)], and finally the HR for the subgroup of patients with LA strain <15% is 22.14 [95% CI (13.69–35.81)].

All uni- and multivariable Cox regression models, assessing the relationship between LA strain measurements and AF development, are depicted in Table 3. In univariable analysis, PALS, PACS, and LACS were found to be significantly associated with the development of AF. Following multivariable adjustments, only PALS [HR 1.05, 95% CI (1.03–1.07), P < 0.001, per 1% decrease] and PACS (HR 1.08, 95% CI (1.05–1.12), P < 0.001, per 1% decrease] remained independent predictors of AF. The incidence rate of AF according to PALS, PACS, and LACS are illustrated in restricted cubic spline curves in Figure 2.

Predictive value of LA strain in patients with a normal LA size

The same multivariable Cox regression models were used to assess the predictive value of LA strain when restricting the analysis to participants with normal-sized LA (LAVi < 34 mL/m²), preserved LVEF (>50%), and no prior history of IHD and those who did not suffer acute myocardial infarction prior to the development of AF during follow-up. A total of 2701 participants were included (69 developed AF). In this analysis, PALS (HR 1.06, 95% CI (1.03–1.09), P < 0.001, per 1% decrease] and PACS [HR 1.08, 95% CI (1.04–1.12), P < 0.001, per 1% decrease] remained independent predictors for AF development.

PALS remained significantly associated with the development of AF in subgroup analyses (normal LVEF, LVEF > 40% and LVEF < 50%, LVEF < 40%, diastolic dysfunction grade 0, diastolic dysfunction grade 1, diastolic dysfunction grade 2, LAVi < 34 mL/m², LAVi ≥ 34 mL/m² and LAVi < 45 mL/m², LAVi ≥ 45 mL/m², arterial hypertension, non-arterial hypertension, <70 years old, and ≥70 years old) (see Supplementary data online, Table S2). PACS remained significantly associated with AF development in all subgroups, except for LVEF <40%, diastolic dysfunction grade 0, diastolic dysfunction grade 2, LAVi ≥45 mL/m², and <70 years old. LACS remained significantly associated with AF development in all subgroups, except for LVEF <40%, diastolic dysfunction grade 2, and LAVi ≥45 mL/m².

The potential incremental value of LA strain parameters to clinical cut-offs for LVEF <52 and 54 for men and women, respectively, LAVi >35 mL/m², along with grades of LV hypertrophy and LV diastolic function were analysed and LA strain parameters were found to provide significantly incremental prognostic information in all analyses, except for LV hypertrophy vs. LV hypertrophy and PACS (see Supplementary data online, Table S3). Incremental value of LA strain parameters in relation to clinically used echocardiographic cut-offs (diastolic dysfunction, LAVi, LV hypertrophy, and abnormal LVEF) were found to provide significantly incremental prognostic information in all analyses (see Supplementary data online, Table S4).

Intra- and interobserver reproducibility

For PALS, intraobserver variability was 3.3 ± 6.2% (ICC = 0.95) and interobserver variability was 2.9 ± 9.1% (ICC = 0.88). For PACS, intraobserver variability was 1.2 ± 4.7 (ICC = 0.83) and interobserver variability was 0.7 ± 4.4% (ICC = 0.82). For LACS, intraobserver variability was 2.0 ± 4.6% (ICC = 0.97) and interobserver variability was 2.1 ± 6.5% (ICC = 0.94).

Discussion

In this prospective cohort study of a large sample from the general population, we found that PALS, PACS, and LACS become incrementally impaired with the presence of cardiovascular risk factors. In addition, we found that PALS and PACS are independently associated with the development of AF. These findings did not change when restricting the analysis to participants with normal-sized LA, normal LV systolic function, and no history of IHD.

This is the first study that investigates the prognostic value of LA strain and the development of AF in the general population. As previously mentioned, LA strain parameters have been demonstrated to be valuable predictors of AF in several different patient cohorts. In previous studies from the fourth Copenhagen City Heart Study, Alhakak et al. found that PALS provides prognostic information on the long-term risk of AF (and ischaemic stroke) in participants aged <65 years and Olsen et al. found that LA functional measures predict AF in the general population.^17,18 In another study, containing 2812 healthy participants, Liao et al. found similar, however, slightly higher PALS values. This difference might be due to a significantly younger age amongst the participants in Liao et al.’s study (47.4 ± 9.9 vs. 55.3 ± 17.4 years in our study). LA functional measures have been shown to predict AF following ischaemic stroke,⁵ following ST-elevation myocardial infarction (STEMI),¹⁸ and recurrence of AF following catheter ablation.¹⁹ Since all these studies are based on cardiac patients, the prognostic value of LA strain in the general population regarding the development of AF remains unknown.

AF is associated with structural and functional remodelling of the left atrium. This remodelling is often accompanied by an increase in interstitial fibrosis.²⁰ Previously, LA structure has been used to stratify the risk of AF.⁶ The maximal LA volume obtained at LV end-systole indexed to body surface area (LAVi) is currently the only LA measure included in the clinical guidelines.²¹ However, recent studies have shown that LA function is a more sensitive measure to stratify patients according to the risk of AF development.^9,22 Changes in LA structure are generally caused by chronic pressure overload mainly caused by diastolic dysfunction. LA functional properties are affected by LA contractility and instantaneous LA loading conditions, which are dependent on pulmonary veins compliance and LV properties. Therefore, LAVi tells more about LV diastolic dysfunction, whereas LA function is more sensitive to trace subtle pathologic conditions in the LA.

In the present study, we found that both PALS and PACS were significantly associated with the development of AF. Both are measures of LA function. Meanwhile, we did not observe LACS to be significantly associated with AF. This suggests that the prognostic value of PALS regarding the risk of AF is driven by the prognostic value of PACS. Consequently, PACS seems to be the most sensitive diagnostic marker for risk of AF development in the general population.

Whereas PALS reflects LA compliance as well as atrial fibrosis, PACS represents LA intrinsic contractility. A reduction in contractile function may be a manifestation of LA remodelling leading to AF development. These findings suggest that intrinsic contractile strain may be a better early marker for LA function alterations than LACS. The fact that PACS remained significantly associated with AF development when restricting our analysis to participants with normal-sized LA, only emphasizes the prognostic advantage of PACS.

In a prospective study including 580 patients who were taking medication for conditions such as hypertension, type 2 diabetes mellitus, dyslipidaemia, coronary artery disease, mild mitral regurgitation or other valvular heart disease, heart failure, or arrhythmias other than atrial arrhythmias, Hirose et al.²² revealed how LA booster pump function assessed by LA active emptying fraction was superior to LA size in predicting AF development. Since LA active emptying fraction expresses the same as PACS, this is in concordance with the results of the present study.

Identification of paroxysmal AF with prolonged rhythm monitoring is expensive and time-consuming and has shown somewhat varied results.²³ Our results indicate that 2DSTE of the LA may be useful in risk stratification regarding AF in the general population. This could also facilitate the identification of the best candidates for prolonged rhythm monitoring. Therefore, integrating 2DSTE into a standard echocardiographic examination could help detect people at risk for developing AF at an earlier stage.

Study limitations

This study has several limitations. The study sample was primarily composed of Caucasians, which limits the generalizability of the findings to other races and ethnicities. Another limitation is the number of participants who were excluded due to inadequate image quality for measuring strain of the LA, thus leading to selection bias (Supplementary data online, Table S1). Furthermore, no tool for LA deformation analysis is currently available in EchoPAC. Therefore, the software for LV analysis was applied to obtain LA strain as has previously been done.^24,25 It is a clear limitation of this study, that we do not have data on the type of AF (paroxysmal AF, persistent AF, permanent AF, respectively) but only a composite. Furthermore, one must keep in mind that normal ranges for LA contractile strain are not present in current guidelines yet. However, systematic reviews have estimated PACS between 16% and 19%.²⁶ Our study shows that LA contractile strain has prognostic significance, whether one knows its true normal range yet or not. The rate of AF development could be underestimated since asymptomatic episodes may not have been detected during follow-up. Furthermore, as patients diagnosed with AF by a general practitioner are not registered in the National Patient Registry, we might underestimate patients diagnosed with AF. As most patients with newly discovered AF from general practice are referred to the hospital (emergency department or cardiology outpatient clinic), they still get registered in the National Patient Registry. Another limitation in this study is that we only have data for patients with significant valvular disease, which prevents a more precise inclusion of patients with the grading of valvular disease.

Conclusion

In the general population, PALS and PACS independently predict the development of AF. These findings remained true in participants with normal-sized LA, normal LV systolic function, and no history of IHD.

Supplementary data

Supplementary data are available at European Heart Journal - Cardiovascular Imaging online.

Funding

The Copenhagen City Heart Study is funded by The Danish Heart Foundation and The Metropolitan Region of Denmark.

Conflict of interest: T.B.-S. is a steering committee member of the Amgen financed GALACTIC-HF trial. Additionally, he is a member of an advisory board in Sanofi Pasteur and Amgen and has received a speaker honorarium from Sanofi Pasteur and Novartis. All other authors declared no conflict of interest.

Data availability

Data not available due to ethical/legal restrictions.

References