Utility of Speckle-Tracking Echocardiography For Predicting Atrial Fibrillation Following Ischemic Stroke: A Systematic Review And Meta-Analysis

Purpose: Undiagnosed atrial fibrillation (AF) is one of the main sources of cryptogenic stroke. And strain indices measured by speckle-tracking echocardiography are associated with atrial remodeling supposed to be the substrate of AF. Therefore, there is a strong need for evaluating the utility of speckle-tracking echocardiography to predict the likelihood of AF in patients with cryptogenic stroke. Methods: PubMed, Embase and Cochrane Database were searched for studies. The random-effects model was used to calculate the pooled results, and summary receiver operating characteristic curve (SROC) analysis was performed to show the overall predictive value. Results: There were 1483 patients with cryptogenic stroke from 8 studies. Meta-analysis showed that strain indices including global longitudinal strain (GLS) (mean difference [SMD]: -0.22, 95% confidence interval [95% CI]: -0.40 to -0.04) , left atrial reservoir strain (εR), (SMD: -0.87, 95% CI: -1.26 to -0.48, conduit strain (εCD) (SMD: -0.56, 95% CI: -0.81 to -0.30), contractile strain (εCT) (SMD: -1.00, 95% CI: -1.39 to -0.61), and left atrial reservoir strain rate (SRe) (SMD: -0.54, 95% CI: -0.80 to -0.28) measured at the period of cryptogenic stroke was significantly decreased in patients with AF occurrence compared to without. SROC analysis suggested an acceptable predictive efficiency of εR for AF occurrence (AUC = 0.799). Conclusion: For patients after cryptogenic stroke, GLS, εR, εCD, εCT and SRe were significantly decreased in AF occurrence compared with non-occurrence. But there was no value in left atrial reservoir strain rate (SRs) and contractile strain rate (SRa) for predicting AF.


Introduction
Ischemic stroke is a fatal condition which lead it to be one of the five leading causes of death worldwide [1]. So, it is critical to find the cause and the embolic source of ischemic stroke. Approximately 20% of thromboembolic events, including stroke, transient ischemic attack and systemic thromboembolism, are attributed to a cardioembolic source [2]. And atrial fibrillation (AF) is an independent risk predictor of ischemic stroke [2]. However, studies have shown that 20-30% of ischemic strokes could not find the source of the incident and were classified as cryptogenic stroke [3]. It has been reported that undiagnosed AF accounts for 20-30% of cryptogenic stroke [4]. Thus, many researchers explored how to predict AF in patients with ischemic stroke.
The use of cardiac implanted electronic devices (CIEDs) has significantly improved the detection of AF but the invasive characteristic restricts its widespread clinical adoption [5][6][7]. By contrast, echocardiography which could assess left atrium anatomy and function is noninvasive, and speckletracking echocardiography as a novel technology accurately evaluates regional and global left atrial strain which is deemed associated with atrial remodeling [8][9][10]. Atrial remodeling including atrial enlargement, heterogeneity of the conduction tissue and alterations of atrial electrical and contractile properties provides substrate of AF occurrence and promotes its persistence [11,12].
Parameters of atrial speckle-tracking echocardiography mainly include left ventricular global longitudinal strain (GLS) and left atrial reservoir strain (εR), conduit strain (εCD), contractile strain (εCT) [13]. Besides, left atrial reservoir strain rate (SRs), conduit strain rate (SRe) and contractile strain rate (SRa) are also used for evaluation of atrial and ventricular function. It has been reported that these parameters could predict and evaluate AF and its complications [14][15][16][17]. Multiple researches explored their predictive value for AF following cryptogenic stroke during follow-up. However, the results of these studies were not consistent. Therefore, the objectives of current review and meta-analysis of the published literature were to determine the exact value of parameters of speckle-tracking echocardiography to predict the likelihood of AF occurrence in patients with cryptogenic stroke.

Search strategy
We searched the online databases of PubMed, Embase, and Cochrane Database to identify relevant researches from inception to September 18th 2021. The search terms used were as follows: ("speckle tracking" OR "velocity vector imaging" OR "edge tracking" OR "strain" OR "function" OR "deformation" OR "stiffness") AND ("left atrial" OR "atrial" OR "atrium") AND ("atrial fibrillation" OR "AF") AND ("cryptogenic stroke" OR "ischemic stroke" OR "thrombotic stroke" OR "brain infarction" OR "cerebral infarction" OR "stroke"). We also manually searched the reference lists of relevant studies. Two investigators (Qinggele Gao and Peng Liu) independently screened the potentially studies and extracted the data from these studies using a standardized extraction form. Discrepancies between investigators were judged by the third experienced investigator (Tingting Lv).

Selection criteria
The inclusion criteria for this study were as follow: (a) the study design was cohort analysis; (b) studies that reported left ventricular GLS, and left atrial εR, εCD, εCT, SRs, SRe or SRa measured by speckle-tracking echocardiography. (d) studies concerned with the association with strain parameters of speckle-tracking echocardiography and occurrence of AF following cryptogenic stroke. The exclusion criteria including: (a) the identified studies were case reports, letters, comments, reviews or metaanalyses; (b) studies were animal studies. The Newcastle-Ottawa Quality Assessment Scale (NOS) was used for quality assessment of the included studies.

Data extraction
Two investigators (Qinggele Gao and Peng Liu) independently extracted data from identified studies using a standardized extraction form and disagreements were judged by the third investigator. The data

Statistical analysis
To perform the meta-analysis, we used STATA MP 14.0 software. For the demographic information, continuous variables were expressed as the mean ± standard deviation (SD), and categorical variables were expressed as numbers and percentages. Strain parameters of speckle-tracking echocardiography were described as the mean ± SD. The effect measure of the differences of left atrial GLS, εR, εCD, εCT, SRs, SRe or SRa between patients with and without AF occurrence was given as standard mean difference (SMD) with 95% confidence intervals (95% CI). Statistical heterogeneity across studies was assessed by I 2 statistic, which was determined from standard chi-square test. A random-effects model was used in the analysis. Sensitivity analysis using leave-one-out method was applied to identify the source of the heterogeneity. Statistical significance was defined as a 2-tailed p value of 0.05.

Study characteristics
The flow diagram of the search strategy of studies was shown in Figure 1. A total of 718 potentially relevant citations were retrieved from PubMed, Embase and the Cochrane Library, of which eight studies containing 1483 patients met the inclusion criteria and were therefore included [1,[18][19][20][21][22][23][24]. Among these studies, five, seven, four, six studies investigated the differences in GLS, εR, εCD, εCT, respectively, between patients with and without occurrence of AF following cryptogenic stroke; And three studies investigated the differences in SRs, SRe or SRa, between patients with and without occurrence of AF following cryptogenic stroke. The baseline characteristics of included studies and of the study populations are shown in Table 1. The proportion of female participants ranged from 36% to 50.8%, and mean age from 50 to 80 years.

Difference in GLS between patients with and without occurrence of AF
Five studies [19,20,[22][23][24] assessed the difference in GLS between patients with and without occurrence of AF following cryptogenic stroke. One study [24] revealed statistically decreased GLS in patients with AF occurrence compared to patients without. But the other four studies [19,20,22,23] indicated there was no difference in GLS. As meta-analysis, the pooled analysis showed a significant decrease in GLS measured at the period of cryptogenic stroke in patients with AF occurrence compared to patients without AF occurrence. And the pooled SMD was -0.22 (95% CI: -0.40 to -0.04; p = 0.016) (I 2 = 0.0%, p = 0.684) (Figure 2).

Difference in εR between patients with and without occurrence of AF
Seven studies [1,[18][19][20][21][22][23] assessed the difference in εR between patients with and without occurrence of AF following cryptogenic stroke. Five studies [1,19,[21][22][23] revealed that εR was significantly decreased in patients with AF occurrence compared to patients without. In contrast, the other two studies [18,20] reported there was no difference in εR. As meta-analysis, the pooled analysis showed a significant decrease of εR in patients with AF occurrence compared to patients without AF occurrence.

Difference in εCD between patients with and without occurrence of AF
Four studies [19][20][21][22] assessed the difference in εCD between patients with and without occurrence of AF following cryptogenic stroke. Two studies [21,22] showed a statistically significant decrease of εCD in patients with AF occurrence. But the other two studies [19,20] reported there was no difference.

Difference in εCT between patients with and without occurrence of AF
Six studies [1,[19][20][21][22][23] assessed the difference in εCT between patients with and without occurrence of AF following cryptogenic stroke. Four studies [1,19,21,23] revealed εCT was decreased in patients with AF occurrence compared to patients without. But the other two studies [20,22] reported there was no difference in εCT. The pooled analysis showed a significantly decrease of εCT in patients with AF occurrence compared to patients without. SMD was -1.00 (95% CI: -1.39 to -0.61, p < 0.001) with high heterogeneity (I 2 = 80.6%, p < 0.001). Upon sensitivity analysis by removing one study at a time, Rasmussen et al. [22] study was found to be the cause of heterogeneity and removal of this study reduced heterogeneity to I 2 66.4%, p = 0.018, but this did not change the SMD significantly. (Figure 3C)

Difference in SRs between patients with and without occurrence of AF
Three studies [18,20,24] assessed the difference in SRs. Among these studies, one study [18] reported a statistically significant decrease of SRs in patients with AF occurrence compared to patients without AF occurrence. But the other two studies [20,24] reported there was no difference in SRs. As meta-analysis, the pooled analysis showed no significantly different SRs in patients with AF occurrence compared to patients without AF occurrence with SMD of -0.46 (95% CI: -1.04 to 0.11, p = 0.114) and high heterogeneity (I 2 = 80.4%, p = 0.006). Removal of Deferm et al. [18] study reduced heterogeneity to I 2 21.2%, p = 0.260, but this did not change the SMD significantly ( Figure 4A).

Difference in SRe between patients with and without occurrence of AF
Three studies [18,20,24] assessed the difference in SRe. The three studies indicated statistically decreased SRe in patients with AF occurrence compared to patients without AF occurrence. As metaanalysis, the pooled analysis also showed significantly decreased SRe in patients with AF occurrence compared to patients without AF occurrence. And the SMD was -0.54 (95% CI: -0.80 to -0.28, p < 0.001) and low heterogeneity (I 2 = 12.2%, p = 0.320) ( Figure 4B).

Difference in SRa between patients with and without occurrence of AF
Three studies [18,20,24] assessed the difference in SRa between patients with and without occurrence of AF following cryptogenic stroke. One study [18] reported statistically decreased SRa in patients with AF occurrence. But the other two studies [20,24] revealed there was no difference in SRa.
As meta-analysis, the pooled analysis showed no significantly different SRa in patients with and without AF occurrence with SMD of -0.35 (95% CI: -0.92 to 0.21, p = 0.220) and high heterogeneity (I 2 = 79.9%, p = 0.007). Removal of Deferm et al. [18] study reduced heterogeneity to I 2 0.0%, p = 0.547, but this did not change the SMD significantly. (Figure 4C)

Discussion
This is the first meta-analysis exploring the value of parameters measured by speckle-tracking echocardiography to predict the likelihood of AF occurrence in patients with cryptogenic stroke.
Echocardiography is a noninvasive and relatively low-cost method to estimate atrial and ventricular function as well as anatomy. And speckle-tracking echocardiography has been used as a quantitative assessment tool, by which regional and global left atrial function, atrial strain could be evaluated accurately [25].
AF could induce structural and functional remodeling in atrium, and atrial remodeling also could provide substrate of AF occurrence and persistence, the phenomenon which is called "AF beget AF" [26,27]. Thus, it is meaningful to find an indicator of atrial remodeling to predict AF or diagnose subclinical AF. Atrial strain alteration is expression of atrial remodeling and could be measured quantitatively by speckle-tracking echocardiography [10,28]. respectively. In the current study, the pooled analysis demonstrated that εR εCD and εCT were significantly decreased in patients with AF occurrence compared to patients without. So, for the patients with cryptogenic stroke, decreased εR, εCD or εCT could represent severe atrial remodeling and reduced atrial compliance, which were at high risk of occurrence of AF or complicated with subclinical AF.
Meanwhile, in our pooled analysis, left ventricular GLS which was changed in HFpEF, stage B heart failure and mitral regurgitation significantly decreased in patients with AF occurrence [29]. We supposed that patients with decreased GLS were more likely to occur AF due to the patients complicated with heart diseases that could induce AF [16,30].
SRs, SRe and SRa as the rate of εR, εCD and εCT, respectively, are also the critical indices of speckle-tracking echocardiography to estimate atrial compliance [10]. In our analysis, the results showed that pooled analyses of SRs and SRa were at high heterogeneity caused by Deferm et al. study. We concluded that the heterogeneity of this study was due to shorter interval between occurrence of cryptogenic stroke and echocardiography compared with other studies, besides, the rhythm monitoring method was 30-day mobile cardiac outpatient telemetry for AF detection. However, removal of this study did not change the result that there were no significantly differences of SRs and SRa between patients with AF occurrence and without. So, it is meaningless for SRs and SRa to predict AF in patients with cryptogenic stroke. SRe, as a parameter of left ventricular relaxation and atrial compliance, were proved significantly decreased in patients with AF occurrence. Thus, SRe is more effective to predict AF than εR and εCT. Further studies are required to seek the cut-off value of strain and strain rate indices after cryptogenic stroke to evaluate risk of AF occurrence.

Limitations
This study has several limitations. Firstly, the current meta-analysis included relatively small number of studies, which was also due to that there were limited studies involved in association between indices of speckle-tracking echocardiography and occurrence of AF after cryptogenic stroke. Secondly, parameters of strain measured by speckle-tracking echocardiography have some methodologic variations including option of chamber views, timing of initial onset on ECG and measuring roof of the left atrium or not, which was a critical factor of high degree of heterogeneity in our effect estimate. Thirdly, in the identified studies and this meta-analysis, it was hard to figure out if the etiology of stroke was AF or not.
We concluded that some patients with cryptogenic stroke were due to subclinical AF, but there was limited way to prove it. Finally, only εR was analyzed with predictive effect quantitatively due to limitation of number of relevant studies.

Conclusion
Speckle-tracking echocardiography is useful to predict AF in patients with cryptogenic stroke. strain parameters of GLS, εR, εCD, εCT were exhibited decreased compared with patients without AF occurrence. And strain parameters were more predictive than indices of strain rate in which only SRe were proved decreased in patients with AF occurrence.   Figure 2. Forest plot pooled analysis of standard mean difference of GLS between patients with AF occurrence and without after cryptogenic stroke. Abbreviation: GLS, global longitudinal strain. AF, atrial fibrillation. Figure 3. Forest plot pooled analysis of standard mean difference of εR between patients with AF occurrence and without after cryptogenic stroke (a). Forest plot pooled analysis of standard mean difference of εCD between patients with AF occurrence and without after cryptogenic stroke (b); Forest plot pooled analysis of standard mean difference of εCT between patients with AF occurrence and without after cryptogenic stroke (c). Abbreviation: εR, left atrial reservoir strain. εCD, conduit strain. εCT, contractile strain. AF, atrial fibrillation. Figure 4. Forest plot pooled analysis of standard mean difference of SRs between patients with AF occurrence and without after cryptogenic stroke (a). Forest plot pooled analysis of standard mean difference of SRe between patients with AF occurrence and without after cryptogenic stroke (b); Forest plot pooled analysis of standard mean difference of SRa between patients with AF occurrence and without after cryptogenic stroke (c). Abbreviation: SRs, left atrial reservoir strain rate. SRe, conduit strain rate. SRa, contractile strain rate. Figure 5. The SROC of decreased εR for AF occurrence. SROC, summary receiver operating characteristic curve. AF, atrial fibrillation.