Dual-Phase Increase in Atrial Arrhythmia Prevalence Following Percutaneous Atrial Septal Defect Closure

Abstract

Background

Conflicting data surround the relationship between age at percutaneous atrial septal defect (ASD) closure and subsequent burden of atrial arrhythmias (AA), particularly in adults. This study aimed to determine the effect of age at ASD closure and other predisposing patient-specific factors on the burden of AA postpercutaneous ASD closure.

Methods

All patients who underwent percutaneous ASD closure at Cleveland Clinic from January 2010 to July 2022 were included. A nonlinear logistic temporal decomposition mixed-effects model was used to analyze the longitudinal AA.

Results

Among 197 patients, 63% (125) were female, with a mean age of 40 years (SD, 24 years) at ASD closure. A total of 177 patients (89% of the cohort) had 687 rhythm records. Postclosure AA exhibited a dual-phase pattern: early peaking phase followed by a late rise up to 6 years postclosure. Age 60 years or older was associated with higher likelihood of early (≤6 months) and late (>6 months) AA prevalence. Older age, lower E/A ratio, and lower left ventricular ejection fraction were associated with a higher likelihood of AA post-ASD closure. Greater than moderate tricuspid regurgitation was associated with a higher likelihood of early AA. Mild right ventricular dysfunction and more than moderate right ventricular dilation were associated with a higher likelihood of late AA.

Conclusions

Older patients have an ongoing dual-phase risk for AA postpercutaneous ASD closure. Our findings underscore the need for routine rhythm monitoring in patients aged 60 years or older.

Introduction

Atrial septal defects (ASDs) are one of the most common congenital heart defects, observed in ∼1 to 2 per 1000 live births and ∼30% of adult patients with congenital heart disease.1, 2, 3 ASDs often present in adulthood owing to subtle symptomatology and physical examination findings during childhood.⁴ The extent and direction of flow depends not only on defect size but also, importantly, on the differences in pressure and compliance of the atria and ventricles. The magnitude of shunting can change with time as these variables change.⁵ Without timely closure, life expectancy may be decreased due to pulmonary hypertension, atrial arrhythmias (AA), and heart failure.^2,6,7 A chronic left-to-right shunt and right-sided volume loading may lead to electrical remodeling, which can predispose to AA.⁸ Patients with ASDs may have underlying diastolic dysfunction that can be unmasked once the ASD is closed, potentially due to a sudden increase in left ventricular (LV) volume and pressure.9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23 This sudden increase in left atrial (LA) pressure, in addition to mechanical inflammation from device placement, has been hypothesized to contribute to the periprocedural AA burden.

There has been conflicting data regarding whether ASD closure changes AA burden, especially in adults aged older than 40 years, with some reporting reduction and others reporting no significant difference in overall burden.24, 25, 26, 27 Few studies have investigated preprocedural patient factors that confer the risk of post-ASD closure AA.

Objectives

Our primary aim was to describe the association between age at ASD closure and the prevalence of AA after percutaneous ASD closure. Our secondary aim was to identify other patient-specific factors that influence the risk of AA after percutaneous ASD closure.

Materials and methods

This was a retrospective cohort study of all patients who underwent percutaneous ASD closure at our institution between January 2010 and July 2022. Patients were excluded if they had a hemodynamically insignificant ASD closed for stroke prophylaxis or military clearance, Qp:Qs of <1.5 and normal right atrial and right ventricular (RV) size, iatrogenic ASD (eg, after transcatheter mitral valve replacement), and Eisenmenger syndrome (Qp:Qs < 1.0).

Data were initially collected via natural language processing to extract data from free-text reports, followed by a manual chart review for quality assurance. Missing data points were collected by a manual chart review where available. Demographic and anthropomorphic data collected included sex, age at ASD closure, weight, and body surface area (BSA). Comorbid conditions were collected at the time of ASD closure and incident during follow-up and defined as documented diagnosis in the patient electronic medical record. Conditions include hypertension, hyperlipidemia, diabetes, coronary artery disease, obesity, pulmonary disease, obstructive sleep apnea, acquired heart disease, and valvular heart disease.

The presence of pre-ASD closure AA was determined by documentation in the patient chart and available echocardiogram (ECG) and ambulatory rhythm monitors. Postclosure AA were determined solely by available ECG and ambulatory rhythm monitors. Rhythm diagnosis was initially documented by a cardiologist or emergency physician and subsequently confirmed by a cardiology fellow with the assistance of an electrophysiologist when unclear.

ASD size was characterized by echocardiogram, magnetic resonance imaging, or balloon sizing, where available. Echocardiographic parameters were collected from available reports and, where not documented, were independently assessed by a cardiologist. Variables collected included qualitative RV and LV size and function, LV ejection fraction, and degree of tricuspid and mitral valve regurgitation. Mitral E/A and E/e' were collected to objectively assess LV diastolic function. Hemodynamic catheterization data were collected from available catheterization reports, including shunt fraction, mean right atrial pressure, RV systolic pressure, and mean pulmonary artery pressure.

Statistical analysis

Simple descriptive statistics were used to summarize the data. Continuous variables are presented as mean ± SD and as 15th, 50th (median), and 85th percentiles if skewed. Categorical data are described using frequencies and percentages. Data were analyzed using SAS software version 9.4 (SAS Institute) and R software version 3.6.0. ECG and ambulatory rhythm data were categorized as sinus rhythm, atrial fibrillation (AF), atrial flutter (AFL), re-entrant supraventricular tachycardias, and others. AA was categorized as early (≤6 months) or late (>6 months) post-ASD closure.

We considered only sinus rhythm, AF, and AFL for longitudinal analysis. To assess the temporal trend of the likelihood of AA (AF or AFL) over time after the ASD closure, rhythm records were analyzed longitudinally for change in the patient-specific profiles of probabilities of AA. A nonlinear logistic mixed model was used to resolve the number of time phases in the odds domain, form a temporal decomposition model, and estimate the shaping parameters at each phase.²⁸ Each phase is modulated by a nonlinear (or linear) time function. We used a common patient-specific random effect to account for possible associations among the longitudinal rhythms collected within each patient. The nonlinear temporal decomposition model for repeated binary measurements was implemented using PROC NLMIXED (SAS). The prevalence of AA over time was estimated by averaging the patient-specific prevalence profiles. We performed a focused analysis to assess the association between age at ASD closure and the prevalence of AA. We evaluated the association after adjusting for sex and the presence of preoperative AA.

A boosting approach was used to identify patient and procedure factors associated with the likelihood of longitudinal binary data (AA vs sinus) using Boostmtree. Variables considered in the multivariable analysis are listed in Appendix A. Boostmtree is a boosting approach based on a marginal model for modeling longitudinal data. Boostmtree provides a flexible alternative to model complex relationships of multiple covariates and their possible interactions with time. The approach uses multivariable trees as a base learner for modeling relationships of multiple covariates and a response. A P-spline approach is used with estimated smoothing parameters to model covariate–time interactions.²⁹ Missing data in the covariates were imputed “on the fly” as a part of growing the forest object.³⁰ Variable selection is performed using the variable important approach,^30,31 which can separate an overall effect of covariates into a covariate main effect and a covariate–time interaction effect. That is, variable effects that do not change with time and effects that change with time. Partial dependency plots are then used to describe the relationship between the covariate of interest and the response by integrating the effect of all the other covariates (or averaging out or risk adjusted).³² The study was approved by the Cleveland Clinic institutional review board.

Results

A total of 197 patients underwent percutaneous ASD closure at our institution between January 2010 and July 2022. The mean age at the time of ASD closure was 40 years (SD, 24 years; 15th-85th percentile, 7.9-68 years), and 63% (n = 125) were female. Baseline patient characteristics are described in Table 1. All patients were noted to have secundum ASDs. Further, 127 patients (89% of the study cohort) had post-ASD closure ECGs or ambulatory rhythm monitors and were included in the longitudinal analysis. A total of 687 longitudinal rhythm records, including ECGs and ambulatory rhythm monitors, were collected. The frequency of rhythm types observed during follow-up is detailed in Table 2. Rhythm records were not reviewed after any post-ASD closure ablations. The median follow-up duration was 6.5 months (IQR, 2 years). Patients with earlier dates of ASD closure had longer follow-ups with more longitudinal rhythm data. The distribution of rhythm data collection over different time frames after ASD closure is detailed in Figure 1, with 96 patients having 161 rhythm records collected at least 1-year post-ASD closure.

Table 1.

Preoperative patient characteristics at the time of percutaneous ASD closure (n = 197)

Characteristics	N = 197
Demographics
Female sex	125/197 (63)
Age at ASD closure, y	40 ± 9.2 (n = 197)
Body surface area, m2	1.6 ± 0.58 (n = 128)
ASD characteristics
ASD size, mm	16 ± 6.0 (n = 154)
Electrophysiologic characteristics
History of atrial arrhythmias	28/182 (15)
Renal function
Creatinine, mg/dL	8.1 ± 4.1 (n = 197)
Comorbidities
Hypertension	50/195 (26)
Hyperlipidemia	36/195 (18)
Diabetes	10/195 (5.1)
Coronary artery disease	22/195 (11)
Obesity	20/195 (10)
Pulmonary disease	25/195 (13)
Obstructive sleep apnea	17/195 (8.7)
Acquired heart disease	20/195 (10)
Valvular heart disease	35/195 (18)
Echocardiography
LV ejection fraction, %	61 ± 8.3 (n = 133)
RA size: dilated (any)	114/165 (69)
RV dilation
Normal size	45/146 (31)
Mild	48/146 (33)
Moderate or severe	53/146 (36)
TV regurgitation
None	119/187 (64)
Mild	42/187 (22)
Moderate or severe	26/187 (14)
RV function
None	130/184 (71)
Mildly decreased	41/184 (22)
Moderately or severely decreased	13/184 (7.1)
E/A ratio	1.4 ± 0.64 (n = 112)
E/e' ratio	8.3 ± 2.7 (n = 106)
Invasive hemodynamics
RV systolic pressure, mm Hg	36 ± 17 (n = 137)
Shunt fraction (Qp/Qs)	1.8 ± 0.52 (n = 138)

Values are n/N (%) or mean ± SD. ASD, atrial septal defect; LV, left ventricular; RA, right atrial; RV, right ventricular; TV, tricuspid valve.

Table 2.

Rhythm monitoring data during the longitudinal follow-up

Rhythm type	ECGs	Ambulatory monitor
Sinus	519	64
Atrial fibrillation	24	6
Atrial flutter	38	3
Ectopic atrial tachycardia	0	1
SVT (AVRT, AVNRT)	1	2
Other	95	0

AVNRT, atrioventricular nodal reentrant tachycardia; AVRT, atrioventricular reentrant tachycarida;ECG, echocardiogram; SVT, supraventricular tachycardia.

Figure 1.

Summary of individual patient rhythm data over time with blue circles indicating sinus rhythm and red dots indicating AA. AA, atrial arrhythmia; AFIB, atrial fibrillation; ASD, atrial septal defect; AM ambulatory rhythm monitor; ECG, echocardiogram.

Estimated odds of post-ASD closure AF or AFL yielded an early peaking phase followed by a late rising phase (Figure 2A). The early phase is consistent with a peak in AA prevalence at 1 week post-ASD closure, increasing from 0.9% up to 14%, followed by a gradual decrease to 3.4% by 6 months post-ASD closure. This was followed by a late rising phase in AA prevalence starting 6 months post-ASD closure, increasing to 4.9% by 6 years post-ASD closure. The composite temporal trend of the prevalence of AA over time after ASD closure is seen in Figure 2B.

Figure 2.

(A) Temporal decomposition of occurrence of AA. Two phases are noted: an early peaking phase and a later phase of essentially random occurrence of AA after percutaneous ASD closure. (B) Temporal trend of prevalence of AA after percutaneous ASD closure. Solid lines represent parametric estimates of the percentage of patients’ average of the patient-specific profiles in AA after the ASD closure, enclosed within a 68% CI. Symbols represent data grouped (without regard to repeated measurements) within a time frame to provide a crude verification of model fit. (C) Temporal trend of prevalence of AA stratified by age (≤40 vs 40-60 vs ≥60 years). Solid lines represent parametric estimates of the percentage of patients (average of the patient-specific profiles) in AA after the ASD closure. Symbols represent data grouped (without regard to repeated measurements) within a time frame to provide a crude verification of model fit. AA, atrial arrhythmia; AF, atrial fibrillation; ASD, atrial septal defect.

The temporal trend of prevalence of AA was then stratified by age into 3 groups: 40 years or younger, 40 to 59 years, and 60 years or older. A parametric estimate of the prevalence of AA after ASD closure, stratified by age group, is summarized in Table 3 and visualized in Figure 2C. Age 60 years or older was associated with a higher likelihood of early (≤6 months) and late (>6 months) postprocedural AA prevalence.

Table 3.

Parametric estimate of the prevalence of AA after ASD closure.

Time	1 d	1 wk	2 wk	1 mo	6 mo	1 y	3 y	5 y	6 y
Overall	0.90	14	12	7.6	3.4	3.6	4.4	4.7	4.9
Stratified by age
<40 y	0.00	1.7	1.1	0.44	0.00	0.00	0.00	0.00	0.00
40-60 y	0.00	23	16	7.0	0.59	0.16	0.02	0.00	0.00
>60 y	2.1	66	54	29	9.8	11	13	14	15

AA, atrial arrhythmia; ASD, atrial septal defect.

The top 10 variables associated with increased likelihood of AA over time after ASD closure are depicted using variable important analysis (Figure 3A, Appendix A). Lower E/A ratio (Figure 3B), lower LV ejection fraction (Figure 3C), and older age at ASD closure (Figure 2C) were associated with a higher likelihood of AA after ASD closure. Of note, LV ejection fractions in this population were predominately within the normal range. Moderate or greater tricuspid valve regurgitation was associated with early AA post-ASD closure (Figure 3D). Other factors associated with a higher likelihood of early AA included lower shunt fraction, lower RV systolic pressure, higher baseline creatinine, and larger BSA. Mild RV dysfunction (Figure 3E) and moderate or more RV dilation (Figure 3F) were associated with a higher likelihood of late AA post-ASD closure.

Figure 3.

(A) Preprocedural and procedural variables associated with the likelihood of AA after ASD closure. Variable importance (VIMP) of the top 10 variables. The top upward-pointing bars depict VIMP for the main effect of variables; the downward-pointing bars depict VIMP for these same variables as their effect changes with time. (B) Association between echocardiographic E/A ratio and the likelihood of AA over time after ASD closure. The association is depicted using a risk-adjusted partial dependence plot. Symbols are risk-adjusted 1-month (green), 1-year (orange), and 5-year (brown) predicted probability of AA over time after percutaneous ASD closure for different values of preoperative echo E/A ratio. Solid lines are smoothed lowess lines of the symbols. (C) Association between preoperative LV ejection fraction and the likelihood of AA over time after percutaneous ASD closure. The association is depicted using a risk-adjusted partial dependence plot. Symbols are risk-adjusted 1-month (green), 1-year (orange), and 5-year (brown) predicted probability of AA over time after ASD closure for different values of preoperative LV ejection fraction. Solid lines are smoothed lowess lines of the symbols. (D) Association between preoperative TV regurgitation and the likelihood of AA over time after percutaneous ASD closure. The association is depicted using a risk-adjusted partial dependence plot. Solid lines depict the risk-adjusted predicted probability of AA over time after ASD closure stratified by different categories of preoperative TV regurgitation. (E) Association between preoperative RV dysfunction and the likelihood of AA over time after percutaneous ASD closure. The association is depicted using a risk-adjusted partial dependence plot. Solid lines depict the risk-adjusted predicted probability of AA over time after ASD closure stratified by different categories of preoperative RV dysfunction. (F) Association between preoperative RV dilation and the likelihood of AA over time after percutaneous ASD closure. The association is depicted using a risk-adjusted partial dependence plot. Solid lines depict the risk-adjusted predicted probability of AA over time after ASD closure stratified by different categories of preoperative RV dilation. AA, atrial arrhythmia; AF, atrial fibrillation; ASD, atrial septal defect; BSA, body surface area; LV, left ventricular; RV, right ventricular; RVSP, right ventricular systolic pressure; TV, tricuspid valve.

Interventions for AA after percutaneous ASD closure included cardioversion in 7 patients, with 12 cardioversions over the follow-up period. Indications for cardioversion were AF for 6 cases and AFL for 6 cases. A total of 7 ablations were performed in 5 patients. Indications for ablation were AF for 1 case, AFL for 5 cases, and atrial tachycardia in 1 patient. There were 10 arrhythmia-related admissions after ASD closure.

Discussion

Percutaneous closure has become the first-line therapy for the management of ASD. However, long-term data on the interaction of age at closure and atrial arrhythmic burden in adults is lacking. Some studies have suggested that transcatheter ASD closure before the onset of AA may be protective against the subsequent development of arrhythmia in patients younger than 40 to 55 years at the time of intervention.^25,33 Vitarelli et al²⁷ demonstrated reduced atrial flutter but not AF post-ASD closure. A meta-analysis conducted in 2010 found that surgical or transcatheter ASD closure was associated with a reduction in the postclosure prevalence of pre-existing AA in the short term to medium term (<30 days and <5 years, respectively).²⁴ A recent meta-analysis showed only a small decrease in AA after transcatheter closure for patients with age at closure of 40 years or older and no decrease at age 60 years or older.³⁴ Previous studies have been limited overall by the narrow age range of the patients included.34, 35, 36

In our study, there was a dual-phase increase in AA post-ASD closure (Central Illustration). A first-week early peak in AA postdevice closure up to 14%, followed by a gradual decrease to 3.4% by 6 months, was seen. This was followed by a slow re-increase in AA prevalence to 4.9% at 6 years, particularly evident in patients older than 60 years. The initial rise in AA postdevice closure may be multifactorial. Theoretically, there is an acute inflammatory response postdevice placement and a sudden rise in LA pressure. The slow increase in AA prevalence over the years that is more pronounced in older patients is possibly related to the ongoing risk of developing AA from other comorbidities, such as hypertension, diabetes, chronic kidney disease, and obesity.

Central Illustration.

Dual-phase increase in atrial arrhythmia prevalence following percutaneous atrial septal defect closure. AA, atrial arrhythmia; AF, atrial fibrillation; ASD, atrial septal defect; BSA, body surface area; LV, left ventricular; RV, right ventricular.

In addition to older age, we also demonstrate that a lower E/A ratio, which may be a marker of impaired relaxation versus expected age-related alterations in LV inflow indices, was also associated with increased AA prevalence. Baseline diastolic dysfunction likely plays an important role in AA outcomes after percutaneous ASD closure; however, it merits more careful preclosure and postclosure diastology assessment, not only by echocardiography which should include E/e', pulmonary venous waveforms, and LA volumes, but also by invasive hemodynamics. Of note, lower LV ejection fraction, even if within the lower range of normal, was associated with higher likelihood of AF or AFL. The acute rise in volume return to the left ventricle after ASD closure likely exacerbates underlying systolic and diastolic dysfunction in these patients, resulting in higher LA pressures, LA stretch, and AA. Moderate or greater tricuspid valve regurgitation was associated with the early peak in AA postclosure. This may be due to precapillary or postcapillary pulmonary hypertension. In turn, the tricuspid regurgitation leads to elevated right atrial pressures and right atrial dilation, exacerbating AA. Other factors associated with early AA included lower shunt fraction, which may be an indirect marker of elevated pulmonary vascular resistance, and lower RV systolic pressure, an indirect marker of RV dysfunction. In addition, higher baseline creatinine and larger BSA reflect comorbidities of chronic kidney disease and obesity, which can often lead to increased LA pressures, contributing to the risk of early AA. Finally, effects of longstanding shunting with RV dilation and/or dysfunction were associated with late AA prevalence post-ASD closure. This finding suggests that RV remodeling from chronic volume loading has lasting changes on the atrial myocardium even after ASD closure.

To our knowledge, this is the first study to use multiphase, longitudinal modeling to describe the odds of AA after percutaneous device closure. Our findings reinforce existing evidence describing the early rise in AA prevalence after device ASD closure and add to the existing body of data investigating the relationship between age and arrhythmia outcomes. We found the early peak in AA is more prominent with advancing age and most prominent in those aged 60 years or older. Patients aged 60 years or older were found to have ongoing increased odds for late AA post-ASD closure, which was not seen in those aged 40 and 40 to 59 years. This finding suggests that ASD closure before the age of 60 years is associated with improved AA outcomes, but the follow-up duration of this cohort limits the strength of this conclusion. Long-term follow-up in patients with ASD closure at younger than 59 years of age is needed to determine whether their late AA prevalence remains lower than those with ASD closure at 60 years and beyond. It is reasonable to suggest routine rhythm monitoring in patients 60 years or older at the time of percutaneous ASD closure, given the high prevalence of AA in both the early and late postclosure periods. We could not compare the prevalence of AF vs AFL as 2 distinct arrhythmias due to the limited patient population.

Limitations of this study include its retrospective nature and the loss of data with longer follow-up. There is likely bias toward symptomatic AA and underdetection of asymptomatic AA in our cohort given a lack of standardized rhythm monitoring. AFL was more common than AF in our cohort, potentially due to more pronounced symptomatology. Since 2018, our institution routinely obtains ambulatory monitors immediately postprocedure and at 6-month follow-up. There was insufficient echocardiographic data on diastology and insufficient arrhythmic records, which precluded analysis comparing patients with postclosure AF with those with AFL. There may be limitations in the analyses due to shorter follow-up. Without a control population, we cannot rule out the potential that our findings are the result of patient-specific risk factors rather than ASD closure. However, the use of the mixed-effects model strengthens our conclusion.

Conclusion

Following percutaneous ASD closure, there is a dual-phase increase in AA—an immediate peak in AA that declines by 6 months postclosure and a late-occurring gradual rise in AA by 6 years follow-up postclosure. Older patients have an ongoing risk for AA postpercutaneous ASD closure. Our findings underscore the need for routine rhythm monitoring in patients aged 60 years or older after ASD closure. Furthermore, focused monitoring of patients aged 40 to 59 years with risk factors for AA may improve the detection of occult AA after percutaneous ASD closure.