Left Atrial Strain as Evaluated by Two-Dimensional Speckle Tracking Predicts Left Atrial Appendage Dysfunction in Chinese Patients with Atrial Fibrillation

Left atrial appendage (LAA) dysfunction identified by transesophageal echocardiography (TEE) is a powerful predictor of stroke in patients with atrial fibrillation (AF). The aim of our study is to assess if there is a correlation between the left atrial (LA) functional parameter and LAA dysfunction in the AF patients. This cross-sectional study included a total of 249 Chinese AF patients who did not have cardiac valvular diseases and were undergoing cardiac ablation. TEE was performed in all the patients who were categorized into two groups according to their left atrial appendage (LAA) function. A total of 120 of the 249 AF patients had LAA dysfunction. Univariate and multivariate logistic regression was conducted to assess the independent factors that correlated with the LAA dysfunction. Different predictive models for the LAA dysfunction were compared with the receiver operating characteristic (ROC) curve. The final ROC curve on the development and validation datasets was drawn based on the calculation of each area under the curves (AUC). Univariate and multivariate analysis showed that the peak left atrial strain (PLAS) was the most significant factor that correlated with the LAA dysfunction. PLAS did not show inferiority amongst all the models and revealed strong discrimination ability on both the development and validation datasets with AUC 0.818 and 0.817. Our study showed that a decrease in PLAS is independently associated with LAA dysfunction in the AF patients.


Introduction
AF is the most frequent type of arrhythmia, having a global prevalence of 1% to 2%. AF is a public health challenge associated with high comorbidity and an increased mortality risk [1][2][3][4]. For example, AF increases the risk of stroke by five times and is observed in ∼33.3% of all ischemic stroke patients [5,6]. In patients presenting with nonvalvular AF, over 90% of the thrombi originate in the LAA. Anatomical remodeling, atrial fibrosis, and a decline in the contractility of the LA myocardium are factors associated with left atrial thrombus/spontaneous echocardiographic contrast (LAT/ SEC), as well as reductions in LAA flow velocities by TEE, which has been found to predispose patients to stroke [7]. Conventionally, TTE has been used to observe anatomical changes of LA following remodeling in the AF patients [8]; however, it has been increasingly replaced with two-dimensional (2D) speckle tracking echocardiography in order to evaluate LA function [9][10][11]. Most studies that report the association between LA and LAA function [12][13][14][15] were carried out on occidental populations. Compared to Western countries, stroke incidence in China is universally higher and the number of cases undergoing AF ablation procedures has been rapidly increasing. Moreover, stroke, which is associated with AF, is the second leading cause of death in China [16]. Here, we aim to provide a noninvasive imaging modality to evaluate the LAA function in Chinese patients.

Methods
A retrospective and cross-sectional study was conducted at the Guangdong Provincial People's Hospital. is study has been approved by the institutional review board of the hospital. A total of 700 patients diagnosed with AF at the Guangdong Provincial People's Hospital from July 2014 to December 2017 ( Figure 1) were included in this study. Patients exhibiting the following criteria were excluded from this study: pregnancy (n � 105), moderate to severe valvular dysfunction (n � 212), dilated/hypertrophic cardiomyopathy (n � 50), failed TEE (n � 20), and cases containing unqualified images (n � 64), either through failure to capture the total LA geometry (mostly incomplete contours of the LA roof ) or image quality precluding speckle tracking. e patients were informed of research objectives and protocols, and they provided consent to participate in the study. Clinical information was collected at the time of their first hospital visit, including demographic data, current medications, and past medical history (PMH). e CHA2DS2-VASc score was calculated for each patient, and they underwent evaluation with TTE or/and TEE before proceeding to AF management.

Transthoracic Echocardiography.
A two-dimension gray scale echocardiography was used to obtain standard view images, parasternal long-axis view, apical four-chamber view, and apical two-chamber views, where the frame rate was set between 65 and 75 frames per second. In addition, foreshortening of the left atrium was avoided [17]. Briefly, echocardiographic parameters were averaged from a total of five cardiac cycles [12], all of which contained clear image quality. TTE was measured using a Vivid E9 ultrasound instrument (GE Healthcare, Wauwatosa, WI, USA), containing a sector transducer (2.5-3.75 MHz). e left atrial anteroposterior diameter was acquired at its greatest dimension for parasternal long-axis view. e LV end-diastolic diameter (LVEDd) and LV end-systolic diameter (LVESd) were obtained at the level of the mitral valve tips using the M-mode Doppler ultrasonography. Left ventricle ejection fraction (LVEF) was automatically calculated by the system. e biplane modified Simpson's method of discs was used to separately calculate the left ventricular end-diastolic volume (LVEDV), left ventricular end-systolic volume (LVESV), maximum LA volume (LAVmax), and LA minimum volume (LAVmin) from apical 4-and 2-chamber views. e LAEF was calculated using the following formula: (LAVmax − LAVmin)/LAVmax × 100%. e left atrial transverse and longitudinal diameters were measured in a 4-chamber view of the endventricular systole. A spectral Doppler tissue imaging was used to measure tissue Doppler velocities at the septal and lateral annuli while pulsed Doppler echocardiography was used to record flow velocities of the valves. e ratio of the peak early mitral annular velocity to E wave (E/e′) was calculated and recorded.

Transesophageal Echocardiography.
During TEE examination ( Figure 2), images were acquired using a transesophageal probe (2.9-7.0 MHz). Briefly, LAAeV was measured using a pulsed wave Doppler, where the sample volume was placed 1 cm distal from the mouth of the appendage. e maximum emptying velocity was calculated from an average of five well-defined emptying waves [18]. e LAT was defined as a fixed or mobile echogenic mass that clearly differed from the LA or LAA tissue. SEC was defined when dynamic, "smoke-like" echoes were seen within the atria and could not be eliminated by altering the gain settings [19,20]. According to TEE findings, (1) the LAAeV < 40 cm/s and (2) LAT/SEC was found regardless of LAAeV; if either one is positive, the condition of LAA dysfunction was met [7].

Speckle Tracking
Imaging. Cine loops were used to determine speckle tracking analysis ( Figure 2) At least 5 consecutive beats from the three standard apical views were collected for analyses. Average PLAS was measured at the end of the reservoir phase with the tracking location marker placed at the end of the QRS wave calculated from 5 beats in patients with AF rhythm [21]. e LA endocardial border was manually traced in both the four-and two-chamber views with the regions of interest (ROI) encompassing the endocardial border of the mitral annulus, carefully excluding pulmonary veins and the LA appendage orifices. e thickness of the ROI was adjusted to the thinnest part to adapt to the atria. e data were analyzed using EchoPAC software (version 201) by two independent echocardiologists, who were blinded to clinical as well as other identifiers revealing patient characteristics.

Statistical
Analysis. Data were analyzed using R (http:// www.R-project.org) and Empower Stats software (http:// www.empower.stats.com, X&Y solutions, Inc., Boston, MA, USA). Continuous variables were analyzed using the ttest (normal distribution) or Kruskal-Wallis rank sum test (nonnormal distribution), and categorical variables were analyzed using the χ 2 test. Variables associated with LAAeV were evaluated by univariate logistics regression analysis, and those with a p value <0.1 were further analyzed using the multivariable logistics regression model to test for independence. e receiver operating characteristic (ROC) curve was used to evaluate the prognostic predictive value of PLAS in LAA dysfunction. Two-sided p values with p < 0.05 are considered statistically significant.
Reproducibility of the PLAS measurement was assessed in 20 randomly selected subjects. e inter-and intraobserver correlation coefficients were 0.8 and 0.9, respectively, which are consistent with those of previous work [22].

Patient Characteristics.
e basic characteristics of patients used in this study are depicted by a flow diagram ( Figure 1). On the basis of inclusion and exclusion criteria, a total of 249 patients undergoing both TEE and TTE were included in the study. e mean age of the patients was 59.7 years (ranging from 49.7 to 69.7 years), and the majority of the patients were male (72.7%). Following TEE measurement, it was observed that LAA dysfunction was present in 120 patients, with an incidence rate of 48.2%, where 118 patients exhibited LAAeV < 40 cm/s. Out of the 120 identified patients with the LAA dysfunction, 43 had LAT/SEC and 41 showed a LAAeV < 40 cm/s measurement. e basic characteristics of patients with and without LAA dysfunction are presented in Table 1. Patients with LAA dysfunction were older (61.4 ± 9.7 vs. 58.1 ± 12.0, p � 0.021) and had a higher heart rate (83.6 ± 24.2 vs. 76.7 ± 22.9, p � 0.006) compared to controls. e number of patients presenting both LAA dysfunction and AF rhythm (OR 3.99, 95% CI: 2.36, 6.76, p < 0.0001) were statistically significant.  (Table 2). ese identified variables were then further analyzed using multiple logistic regression, and PLAS (OR 0.90, 95% CI: 0.85, 0.95, p < 0.0005) was the only parameter significantly associated with LAA dysfunction after adjusting for sex, PMH of ablation, International Normalized Ratio (INR), AF, and isovolumic relaxation time (IVRT) ( Table 2). e multivariate regression analysis showed that LAA dysfunction was consistently related to the tertials of the PLAS (Table 3).

Univariate and Multiple Logistic Regression
ere was a linear relationship between PLAS and LAA dysfunction ( Figure 3).

Sensitivity Analysis.
e ROC curve for the predictive models of LAA dysfunction are presented in Figure 3. Model 1 includes the CHA2DS2-VASc score only, while Model 2 includes both the CHA2DS2-VASc score and LAEF. Model 3 includes all the three factors, viz., CHA2DS2-VASc score, LAEF, and PLAS, while Model 4 includes PLAS only. Different models performances were compared, and it was observed that PLAS of Model 4 demonstrated similar accuracy and discriminability with Model 3, including CHA2DS2-VASc score, LAEF, and PLAS. For analyses, all subjects were categorized into two equal groups, viz., development and validation groups (Figure 4). e AUC for the development group was 0.818 (95% CI: 0.744, 0.818), yielding a sensitivity of 63.33% with a specificity of 91.18% at the optimal cutoff value (PLAS � 20.44%). In the validation set, the AUC was 0.817 (95% CI: 0.741, 0.894), along with a sensitivity of 78.33% and a specificity of 78.33% at the corresponding threshold.

Discussion
In this study, we performed a cross-sectional study of 249 participants to investigate the association between the functions of LA and LAA. Our results revealed a positive correlation between the reduced peak LA longitudinal strain, assessed by STE, and the presence of LAA reduced emptying velocity and/or thrombus on TEE examination in AF patients. Multivariate logistic regression analysis showed that a decrease in PLAS was an independent predictor of LAA dysfunction. By comparing the different models, it was observed that PLAS alone showed strong discriminability. ese results indicate that in real life settings, PLAS can alternatively be used to rule out the possibility of LAA thrombus and to determine subsequent needs for other imaging assessments.
CHA2DS2-VASc scores, including risk factors of age, hypertension, heart failure, and diabetes mellitus, are clinically measurable indicators promoting atrial remodeling and are widely used to predict the risk of stroke in patients with AF. In our study, we did not find the CHA2DS2-VASc score to be significantly associated with LAA dysfunction, and this score was not an ideal platform to predict risks associated with stroke using the LAT value as a comparator. is inconsistency was also observed in previous studies [23,24]. It is clear that the scoring system includes an assessment of the risks of early stroke, including those that cause vascular wall injury and hypercoagulable    [25,26]. However, the scoring system did not analyze the effect of AF itself on cardiac remodeling which is a continuous present risk factor for stroke. Other possible variations include AF duration, structural cardiac abnormalities, race, anticoagulation adequacy, and other cardiovascular risk factors that might be associated with the observed effects. Yet, these indicators are relatively difficult to define. Consequently, recent efforts are focused on studies to refine the assessment of the risk of stroke based on cardiac structure and function [27,28]. AF has been associated with electromechanical remodeling of the LA. LAA plays a role in maintaining LA pressure as an actively contracting structure [29]. As result of the intravascular volume status and remodeling in AF, LAA has become the primary location for the formation of thrombus [30,31]. In the current study, we observed LAA as the site of thrombi in all patients. TEE is considered as the gold standard for the assessment of the LAA function [32]. However, as it is a semi-invasive procedure not tolerated by all patients, it remains controversial as to whether it is rational to use it as a measurement tool for the serial resolution of LAT or SEC in patients with a history of LAT or SEC [33]. Direct assessments of LAA function by TTE have also been proposed. Conventional 2D TTE has a low sensitivity of only 3-19% for the detection of thrombi in LA and especially in LAA [34]. Studies, such as those performed by Tamura et al. [35] as well as Uretsky et al. [36], were carried out to resolve the problem. ese studies found that TTE-measured LAA tip Doppler tissue velocity was an independent predictor of LAA thrombus formation and may be a useful indicator for risk stratification of AF. Nevertheless, LAA is generally difficult to be clearly and totally visualized on TTE, especially in patients with small LAA.
Kuppahally et al. [13] showed that the LA strain inversely correlates with LA fibrosis in AF patients, suggesting that fibrosis decreases with LA compliance. Replacement of healthy atrial tissue with fibrotic tissue in AF leads to a reduction in the atrial contractile function and blood stasis, which results in the process of thrombus formation [37][38][39]. Our study observed that an impaired PLAS independently correlates with a decreased LAA function. It was also found to be the strongest independent predictor of LAA dysfunction multivariate analysis. Other conventional TTE parameters were found to be significant in univariate analysis but no longer maintained their significance in multivariate analysis. e associations between LA strain and TEE changes were also described in a previous study performed by Meng-Ruo Zhu et al. [40]. e difference here is that we defined LAA dysfunction as both decreased LAAeV and presence of LAT/ SEC considering patient PMH may reduce the incidence of LAT/SEC. erefore, we had a higher LA strain threshold for predicting LAA dysfunction compared to the one used in other studies [40]. Furthermore the diagnostic performance of PLAS to detect LAA dysfunction is strong. e addition of the CHA2DS2-VASc score and LAEF to PLAS offered no further discrimination in the detection of LAA dysfunction to  Figure 4: A receiver operating characteristic curve of PLAS that was used to predict the LAA dysfunction. e receiver operating characteristic curve of PLAS predicting the LAA dysfunction, having 50% of cases as modeling data and 50% of cases as validation data; the AUC were 0.818 and 0.817. e model showed a good discrimination ability on the validation datasets. that offered by PLAS alone, which suggests that in real-world practice, when an AF patient cannot undergo TEE, the PLAS can alternatively be used to rule out the possibility of thrombus and to determine the subsequent needs for other imaging assessments.
Our study has few limitations including the fact that it is composed of a small sample size with infrequent LAT events, which impaired our ability to analyze many covariates within a single regression model. us, we meticulously chose the covariates to be analyzed in the logistic regression models. It is still possible that some weakly associated parameters may have been missed. Secondly, the PLAS was calculated using EchoPAC software (version 201). Vendor differences should be considered when using different machines. Finally, we used the current software for LV analysis to study the LA pattern strain because a dedicated software for LA analysis has not yet been released.
us, it can be concluded that a decrease in PLAS is significantly associated with LAA dysfunction in patients with nonvalvular AF, providing a parameter to identify patients with nonvalvular AF at high risk of stroke.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request. e data are not publicly available due to the potential revealing of information compromising patient identity.

Conflicts of Interest
e authors declare no conflicts of interest.

Authors' Contributions
Yu Wang, Mingqi Li, and Lishan Zhong contributed equally to this work.