Echocardiographic diagnosis of atrial cardiomyopathy allows outcome prediction following pulmonary vein isolation

Background Relevant atrial cardiomyopathy (ACM), defined as a left atrial (LA) low-voltage area ≥ 2 cm2 at 0.5 mV threshold on endocardial contact mapping, is associated with new-onset atrial fibrillation (AF), higher arrhythmia recurrence rates after pulmonary vein isolation (PVI), and an increased risk of stroke. The current study aimed to assess two non-invasive echocardiographic parameters, LA emptying fraction (EF) and LA longitudinal strain (LAS, during reservoir (LASr), conduit (LAScd) and contraction phase (LASct)) for the diagnosis of ACM and prediction of arrhythmia outcome after PVI. Methods We prospectively enrolled 60 consecutive, ablation-naive patients (age 66 ± 9 years, 80% males) with persistent AF. In 30 patients (derivation cohort), LA-EF and LAS cut-off values for the presence of relevant ACM (high-density endocardial contact mapping in sinus rhythm prior to PVI at 3000 ± 1249 sites) were established in sinus rhythm and tested in a validation cohort (n = 30). Arrhythmia recurrence within 12 months was documented using 72-h Holter electrocardiograms. Results An LA-EF of < 34% predicted ACM with an area under the curve (AUC) of 0.846 (sensitivity 69.2%, specificity 76.5%) similar to a LASr < 23.5% (AUC 0.878, sensitivity 92.3%, specificity 82.4%). In the validation cohort, these cut-offs established the correct diagnosis of ACM in 76% of patients (positive predictive values 87%/93% and negative predictive values 73%/75%, respectively). Arrhythmia recurrence in the entire cohort was significantly more frequent in patients with LA-EF < 34% and LASr < 23.5% (56% vs. 29% and 55% vs. 26%, both p < 0.05). Conclusion The echocardiographic parameters LA-EF and LAS allow accurate, non-invasive diagnosis of ACM and prediction of arrhythmia recurrence after PVI. Graphic abstract Supplementary Information The online version contains supplementary material available at 10.1007/s00392-021-01850-x.


Introduction
Atrial fibrillation (AF) is the most common supraventricular arrhythmia worldwide with increasing incidence [1,2]. Persistent forms of AF are associated with atrial cardiomyopathy (ACM) [3][4][5]. ACM, in turn, was demonstrated to be related to higher arrhythmia recurrence rates after pulmonary vein isolation (PVI) and an increased risk of stroke [6][7][8][9][10]. The current ESC guidelines on AF emphasize the importance of ACM diagnosis for therapeutic management and guidance of PVI [2]. Invasive endocardial contact mapping of the left atrium has been established for detection and quantification of ACM [7,8,11]. As a non-invasive diagnostic tool, late gadolinium-enhanced magnetic resonance imaging (MRI) was introduced over the last years [6,9,10]. However, the widespread use of both is limited due to clinical and practical restrictions.
The aim of the current study was to validate transthoracic echocardiography (TTE) in relation to high-density endocardial contact mapping in a homogenous population of ablation-naive patients with persistent AF and determine cut-off values for ACM diagnosis to identify patients at risk for arrhythmia recurrence following PVI.

Methods
A total of 60 consecutive patients undergoing their first PVI for persistent AF were included in this prospective study. After completion of the first recruitment phase involving 30 patients, the derivation cohort, an interim analysis was performed. Subsequently, the validation cohort (n = 30) was enrolled. All patients underwent electrical cardioversion 4-6 weeks prior to PVI. In case of AF recurrence, patients were again electrically cardioverted in the early morning at admission. Subsequently, a TTE (GE ultrasound system E95, M5Sc probe, GE Healthcare, Solingen, Germany) was performed in sinus rhythm in the late afternoon one day prior to PVI in all patients. The following day, high-density voltage maps were acquired in sinus rhythm using an endocardial contact mapping system (CARTO-3, Biosense Webster, Irvine, CA, USA) prior to PVI. All patients underwent follow-up visits 6 and 12 months after PVI to assess arrhythmia recurrence. The study was approved by the institutional ethics committee and all patients provided written informed consent prior to enrollment.

Transthoracic echocardiography
Standardized TTE was performed in all patients during sinus rhythm in accordance with current guidelines [23]. LA diameter was measured in parasternal long-axis at endsystole. LA volume index was calculated from four-and 1 3 two-chamber views using Simpson's biplane method indexed for body surface area. Left ventricular cavity dilatation was assessed using left ventricular end-diastolic dimension (LVEDD) measured in M-mode echocardiograms derived from 2D images in the parasternal long axis. Simpson´s method was used to determine left ventricular ejection fraction (LV-EF). Left ventricular diastolic function was assessed as described previously [24]. Biplane left atrial emptying fraction (LA-EF) was measured as LA maximum volume minus LA minimum volume divided by LA maximum volume [25]. Functional mitral regurgitation was defined as ratio of regurgitation jet area to LA area ≥ 0.1 in apical four-chamber view [26]. Two-dimensional speckle tracking was analyzed in accordance with the current guidelines [27]. For calculation of the LV global longitudinal strain apical four, two-chamber, and long-axis views were used (average of the curves from the 3 apical views with a frame rate between 57 and 90 frames per second). Furthermore, LA global longitudinal strain (LAS) was automatically analyzed offline in four-and two-chamber views using TomTec software (AutoStrain, TomTec Imaging Systems, Unterschleissheim, Germany): A complete R-R cycle (end-diastole to end-diastole) was automatically selected and endocardial borders were automatically placed by the software. LAS was measured automatically in the reservoir phase (LASr, between mitral valve closure and mitral valve opening), conduit phase (LAScd, between mitral valve opening and onset of LA contraction) and contraction phase (LASct, between onset of LA contraction and mitral valve closure). LAS values were calculated as average from twoand four-chamber views.

Endocardial contact mapping and ablation procedure
High-density endocardial voltage maps (mean 3000 ± 1249 mapped sites per atrial chamber) were acquired in sinus rhythm using either a 20-polar Lasso-Nav (variable diameter: 15-25 mm) or a PentaRay-Nav catheter (electrode size: 1 mm, spacing: 2-6-2 mm, Biosense Webster, Irvine, CA, USA) in combination with the electro-anatomical contact mapping system CARTO-3 as described previously [11,28]. ACM extent was quantified by measurement of areas with a bipolar left atrial low-voltage substrate (LA-LVS) < 0.5 mV. Relevant ACM was defined as a LA-LVS ≥ 2 cm 2 in accordance with the prior studies [28,29].Representative endocardial voltage maps and the respective echocardiographic measurements in patients without and with relevant ACM are illustrated in Fig. 1.

Follow-up
All patients underwent ambulatory clinical visits including 12-lead electrocardiogram (ECG) and 72-h Holter ECG 6 and 12 months after PVI. Arrhythmia recurrence was defined as any documented episode of AF, atypical atrial flutter or atrial tachycardia lasting > 30 s after a three-month blanking period. If an arrhythmia recurrence could not be detected by 12-lead ECG and Holter ECG even though the patient was symptomatic, an event recorder was given to the patient to record a single-lead ECG during symptomatic episodes.

Endpoints
The primary endpoints were (a) to investigate the relation of ACM and echocardiographic parameters in the derivation cohort for definition of cut-off values for ACM diagnosis and (b) to apply these cut-off values to the validation cohort.
The secondary endpoint was to assess the predictive value of the echocardiographic ACM markers regarding arrhythmia recurrence at 12 months following PVI.

Statistical analysis
SPSS Statistics 23 (IBM, New York, USA) and Graphpad Prism 8 (Graphpad Software, San Diego, CA, USA) were used for statistical analysis. Test for normality was assessed by the Shapiro-Wilk test. Normally distributed data are given as mean ± standard deviation and skewed distributed data as median with interquartile range (IQR, 1st and 3rd quartiles). Depending on the number of groups and distribution, group comparison was performed using Student´s t-test and Mann-Whitney U test. Fisher's exact test was used to compare categorical variables. Receiver-operating curves (ROC) were acquired to identify the optimal cut-off values for ACM diagnosis using echocardiographic parameters after inclusion of the first 30 patients and validated in the following 30 patients. Univariate linear regression analysis was conducted for all independent variables. In the multivariate model, all variables with p < 0.05 in the univariate models were selected for analysis. Kaplan-Meier curves were used to illustrate arrhythmia recurrence and compared using the log-rank test. Impact of clinical covariates on arrhythmia recurrence was analyzed using Cox proportional hazard regression models. A two-tailed p < 0.05 was considered significant.

Results
Clinical and procedural characteristics are listed in Table 1. Patients were on average 66 ± 9 years old and 80% were males. Presence of ACM was evenly distributed with 50% of patients meeting the prespecified criteria for relevant ACM. Patients with and without ACM did not differ significantly with regard to clinical symptoms (NYHA and CCS classifications) and conventional cardiovascular risk factors such as atrial hypertension, diabetes mellitus or coronary artery disease. In both groups, only a minority of patients suffered from left ventricular dysfunction or relevant valvulopathies.

Derivation cohort
The derivation cohort showed a moderate correlation between ACM extent quantified by endocardial voltage mapping and LA-EF using standard 2D echocardiography (r = 0.49, p = 0.006). Similar results were obtained for speckle tracking parameters LASr (r = 0.51, p = 0.004) and LAScd (r = 0.54, p = 0.002). Correlation of ACM with LASct was low (r = 0.30, p = 0.1104). ROC analyses for prediction of ACM yielded a c-statistic of 0.846 for LA-EF with the highest sensitivity and specificity obtained at a LA-EF of < 34%. The corresponding cut-off values were < 23.5% for LASr (c-statistic of 0.878), < 13.4% for LAScd (c-statistic of 0.871) and < 5.4% for LASct (c-statistic of 0.742) (Fig. 2a).

Predictors of ACM
Univariate analysis for predictors for relevant ACM was based on the whole study cohort. Correlations between ACM extent and echocardiographic parameters are shown in Fig. 3a-

Arrhythmia recurrence
Arrhythmia recurrence occurred in 24 (40%) patients after a median of 203 (168-316) days during a median follow-up of 355 (200-386) days. Impaired LA-EF, LASr, and LASct were associated with higher arrhythmia recurrence rates within 12 months (log-rank: p = 0.047 for LA-EF, p = 0.030 for LASr and 0.006 for LASct, respectively, Fig. 4). There Application of these cut-offs to the validation cohort showed a significantly increased left atrial low-voltage substrate (LA-LVS) extent in patients with pathological echocardiography criteria (b) and allowed accurate differentiation between patients with and without relevant ACM (c). Whiskers depict median with 25% and 75% interquartile range. Dashed line marks border between absence (< 2 cm 2 LA-LVS at < 0.5 mV) and presence of relevant ACM was a non-significant trend for higher recurrences with reduced LAScd (log-rank: p = 0.061, Fig. 4c). In line with these findings, patients with relevant ACM as diagnosed using endocardial voltage mapping also demonstrated higher arrhythmia recurrence (log-rank: p = 0.001; Supplementary  Online Fig. 1).

Discussion
We report three main findings: First, LA function as quantified using standard and speckle tracking echocardiography in sinus rhythm is related to the extent of ACM in patients with AF. Second, echocardiography can stratify patients into those with and those without relevant ACM. Third, echocardiographic ACM diagnosis is of clinical relevance as it predicts arrhythmia recurrence following PVI.

Atrial cardiomyopathy
Histology of the left atrium can be considered the gold standard for diagnosis of ACM, but is, for obvious reasons, of little clinical relevance [3]. In contrast, endocardial contact mapping is widely available and well established in the context of AF ablation. However, its invasiveness makes it unsuitable for screening and pre-procedural risk stratification [7,8,11]. These limitations gave rise to approaches to non-invasively diagnose and quantify ACM: Late gadolinium-enhanced MRI has been evaluated for ACM diagnosis over the last years [6,9,10], but standardized acquisition and analysis methods are lacking, which affects reproducibility and the need for extensive expertise. Furthermore, costs are high, the examination is time-consuming and several patient groups have contraindications to MRI (e.g., renal insufficiency, claustrophobia, implanted cardiac devices) [30,31]. Another recently introduced non-invasive method is measurement of the amplified p-wave duration in sinus rhythm [28,29]. Its accurate measurement necessitates highquality (low noise) digital acquisition of 12-lead ECG and high expertise in recognition of the low-voltage components of the amplified p-wave, and it is also limited to patients with sufficient signal-to-noise ratio in ECG.

Echocardiography and LA function in ACM
Echocardiography has significant potential advantages for ACM diagnosis: It is widely available, relatively easily applicable, and delivers instantaneous results at low cost. Previous studies showed some value for standard and speckle tracking imaging to diagnose ACM: Cameli et al. reported lower LASr in patients with evidence of ACM diagnosed by atrial histology who were undergoing mitral valve surgery for severe regurgitation [18]. The generalizability of these findings, however, may be limited due to the very specific studied patient population. Kuppahally et al., on the other hand, used MRI to detect ACM, and also reported lower LASr in ACM patients [19]. Contrary to the current recommendations of the European Association of Cardiovascular Imaging (EACVI)/American Society of Echocardiography (ASE)/Industry Task Force [27], they limited LASr measurements to specific parts of the LA (the midlateral and midseptal wall) and performed LASr quantifications partly in AF, and partly in sinus rhythm. In the current study, we investigated the usefulness of echocardiography for ACM diagnosis in AF patients. Therefore, we compared echocardiographic LA function to endocardial voltage mapping of the left atrium in sinus rhythm. Our study differs from previous studies in significant ways: First, we used multi-polar, small-electrode mapping catheters, resulting in high-density LA voltage maps with substantially improved delineation of diseased and healthy atrial tissue as compared to standard low-density mapping with large, single-tip electrode. In contrast to previous studies, this allows a very detailed quantification of the true extent of ACM [20][21][22]. Second, we performed echocardiography in sinus rhythm only, while previous studies reported strain values which were acquired partly during AF and partly during sinus rhythm [21,22]. This is of particular importance as the presence of ACM not only affects LA function and, hence, echocardiographic measurements, but also the likelihood that such patients will be in AF during echocardiography. The correlation of functional echocardiographic parameters with ACM may, therefore, be significant due to collinearity of ACM and presence of AF. Third, all patients included in the current study are ablation-naive and present ACM that is associated with the natural course of AF. This is again in contrast to previous reports which included a substantial part of patients who had undergone prior LA ablations causing iatrogenic ACM, which make their findings unsuitable in the context of naturally occurring ACM [32]. The current study demonstrates that ACM is associated with presence of functional mitral regurgitation. This is in line with previous studies, which assessed that functional mitral regurgitation is related with new-onset AF, the presence of ACM and arrhythmia recurrence following PVI [26,34]. It is, however, controversial whether functional mitral regurgitation is a potential cause or the result of ACM. Furthermore, patients with ACM had more frequent evidence for a left ventricular diastolic dysfunction. Several studies also demonstrated an association between diastolic dysfunction and new-onset AF, LA remodeling and AF recurrence following PVI [35][36][37].

Clinical relevance of ACM diagnosis
The most recent guidelines emphasize the importance of ACM in the context of new-onset AF and arrhythmia recurrence following AF ablation [2]. The current study underlines these recommendations and validates the use of standard two-dimensional and speckle tracking echocardiography for ACM diagnosis and prediction of arrhythmia recurrence following PVI for persistent AF.
ACM as a disease is not only related to the development and persistency of AF but also seems to play an important role in many cardiovascular pathologies such as heart failure with preserved ejection fraction (HFpEF) [29] and ischemic stroke [6]. Previous studies investigating LAS demonstrated not only an association between reduced LAS values and new-onset AF [12][13][14] and higher arrhythmia recurrence rates after PVI [15][16][17], but also an increased risk for HFpEF [38] and stroke [39]. These findings may also be explained by presence of ACM in the investigated patients, resulting in reduced LAS values, and may, therefore, extend the pathophysiological understanding of the above-mentioned cardiovascular diseases.

Limitations
First, the study's sample was relatively small. Future largescale studies are needed to confirm the currently identified cut-off values for LA-EF and LAS for ACM prediction. Second, high quality of the two-dimensional echocardiographic images required for proper quantification of the above-mentioned parameters may be insufficient in some patients. However, no patient was excluded due to insufficient image quality. Third, it is unclear which results manual LA function measurements would have produced compared to those obtained by using automatic strain analysis software as in the current study. Fourth, the cut-off values presented here are valid in sinus rhythm only, requiring cardioversion 1 3 in certain patients. Sinus rhythm is recommended to ensure automatic analysis due to regular and comparable RR cycles.

Conclusion
Standard echocardiography and speckle tracking imaging enable accurate diagnosis of ACM and prediction of arrhythmia recurrence after PVI. Therefore, non-invasive echocardiographic measurement of ACM has the potential to become a widespread screening tool in patients with suspected ACM.