Diagnostic Accuracy of Left Atrial/Left Atrial Appendage Thrombus in Patients with Atrial Fibrillation: A Systematic Review and Network Meta-Analysis

Background: This paper aimed to appraise the diagnostic precision of assorted methodologies to identify left atrial/left atrial appendage (LA/LAA) thrombus through a network meta-assessment. Methods: Methodologically, we conducted a comprehensive literature search across multiple databases. Utilizing the risk of bias tool from the Cochrane Collaboration, methodological quality of included studies was critically assessed and potential publication bias was examined via funnel plots. The subsequent data analysis was executed using Stata software, with the most efficacious diagnostic modalities being determined based on cumulative ranking curve (SUCRA) values. Results: We scrutinized a sum of 18 papers, comprising 4102 subjects and utilizing 10 different diagnostic techniques. The hierarchical results derived from the network meta-analysis indicated that in regards to sensitivity, the dual-source cardiac computed tomography (DSCT) was superior (with a SUCRA value of 71.7%), it was succeeded by 3-minute delayed cardiac computed tomography (CCT) (scoring 66.8%), which surpassed the transesophageal echocardiography (TEE) (holding a SUCRA value of 57.5%). In terms of specificity, DSCT was the best (SUCRA value of 84.3%), followed by three dimensional (3D) cardiac magnetic resonance imaging (3D-CMRI) (SUCRA value of 78.0%), which was better than TEE (SUCRA value of 66.6%). In terms of positive likelihood ratio (PLR), 6-minute delayed CCT (SUCRA value of 85.6%) was superior to 3-minute delayed CCT (SUCRA value of 80.1%), both of which were superior to TEE (SUCRA value of 69.1%). DSCT (SUCRA value of 89.3%) had the best negative likelihood ratio (NLR), while DSCT (SUCRA value of 79.9%) had the highest accuracy. Conclusions: This study demonstrated that DSCT outperformed TEE in sensitivity, specificity, NLR, and accuracy in identifying thrombus of LA/LAA among patients suffering from atrial fibrillation. Our conclusion is that DSCT is the best in diagnosing LA/LAA. In addition, 3D-CMRI and 3-minute delayed CCT are expected to replace TEE.


Introduction
In the spectrum of clinical arrhythmias, atrial fibrillation (AF) predominates.Around 59.7 million people worldwide have AF (including atrial flutter) as of 2019 [1].AF has an increased all-cause mortality by 1.5-fold in men and 2-fold in women [2].Currently, the treatment of AF consists mainly of drugs and catheter ablation.In the last decade, some antiarrhythmic drugs have been found to have a risk of causing arrhythmias [3,4], and therefore have limitations in their clinical application.Catheter ablation offers significant advantages in maintaining sinus rhythm nevertheless, it doesn't render an exception for the left atrial/left atrial appendage (LA/LAA) thrombus.More than 90% of the LA thrombus is present in the LAA, a special structure of the LA in which blood flow is slow and stagnant, leading to thrombus formation.The thrombus and emboli circulate through the blood stream to the cerebral arteries, blocking the blood supply to the brain and leading to ischemic cerebral infarction.The risk rates of ischemic stroke and sys-temic circulation artery embolism caused by AF are 1.92% and 0.24%, respectively.This results in a 20% mortality rate and a 60% disability rate [5], increases the number of cardiovascular diseases (49%), non-cardiovascular diseases (43%), and bleeding hospitalizations (8%) [6].
Currently, the main diagnostic methods for thrombus of LA/LAA include transesophageal echocardiography (TEE), cardiac magnetic resonance imaging (CMRI), multidetector computed tomography cardiac computed tomography, dual-source cardiac computed tomography (DSCT), computed tomography of the heart delayed, as well as cardiac computed tomography angiography (CCTA).However, there is minimal relevant research to determine the overall diagnostic efficacy of these various methods.TEE is decisive for determining LA/LAA thrombosis [7,8].Because TEE has certain complications and some patients cannot tolerate it because of Esophageal stricture, esophageal ulcer, anesthetic allergy/hypertension. Therefore, it is necessary to find alternative detection methods for TEE for TEE [9].We use network meta-analysis (NMA) to make a comparison of the diagnostic result between various thrombus of LA/LAA detection techniques in order to offer solid suggestions for patients and clinicians.

Registration
This study using network meta-analysis (NMA) has been registered on the INPLASY-International Platform, Invoice Number: 2022120041.

Search Strategy
As of September 2022, the literature was retrieved using PubMed, EMBASE, an electronic database of Cochrane Controlled Trials, WOS (Web of Science), as well as other databases.The PICOS ( Patient, Intervention, Control ,Outcome and Study design) tool served as the foundation for the investigators' search strategy: (P) Patients with atrial fibrillation who are receiving cardiac radiofrequency ablation, electrical cardioversion/cardiac evaluation for other rea-sons; (I) Interventional procedures include CMRI, cardiac computed tomography (CCT), TEE, transthoracic echocardiography (TTE), and other diagnostic modalities.(C) Control: Within one month, all patients underwent TEE examinations.(O) LAA/LA thrombus; (S) Study design: observational test.The search approach is shown in Supplementary Table 1.

Data Extraction
Ruirui Song and Jun Chen are responsible for searching the literature and importing the search results into file manager EndNote software (version 20.2.1, Bld 1574, Thomson ResearchSoft, USTC, USA).After Ruirui Song eliminated the duplicate literature, Ruirui Song, Jun Chen, Jian Huang and Xiaojing Shi screened the literature by reading the title, abstract and full text, and finally got the included literature after discussion and communication.Xuefeng Guo and Hongmei Gao extracted data from the in-cluded literature, respectively.If there was any doubt, they agreed on their opinions after consultation.Main extracted information: Author, Year, Country, Reference standard, Diagnostic method and main indicators Se, Sp, PLR, NLR, TP, FP, FN and TN.

Literature Quality Assessment
Hongmei Gao and Xiaojing Shi respectively used QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies) diagnostic accuracy research quality evaluation software [10] to determine the quality of the selected literature, discussed and unified the evaluation results.The scale evaluation included bias risk assessment and clinical applicability assessment.Bias risk is defined as "low", "high"/"uncertain".Discuss and resolve any disagreement with Fang Liu during quality evaluation.

Analysis of Data
In studies with different diagnostic methods used as interventions, all variables were continuous and represented by the mean and standard deviation.The study of continuous variable will take the mean differences (MD equals Mean difference between gold standard TEE and other diagnostic methods, calculated using the same scale), 95% confidence interval (CI), and analysis.Because there are differences among various research, the random effects model was chosen to analyze [11].
Stata software (version 15.1, StataCorp LLC, College Station, TX, USA) was recruited to perform mesh meta-analysis using a Bayesian framework based on the PRISMA NMA User Manual [12,13].Data were inputted into Stata15.1 software and p-values were obtained.Node method was used to quantify the consistency of the included study.If the p-value was above 0.05, it passed the consistency test [14].
Stata15.1 software was used to draw the network diagram, forest diagram and funnel diagram of LA/LAA with various diagnostic tools.Every node marks a type of diagnostic way, and the lines banding the nodes represent a straightforward, positive comparison between other diagnostic ways.The larger the node, the thicker the line, the more the number of studies, and vice versa [15].
Based on Bayesian method, the cumulative ranking curve (SUCRA) values can accurately represent the percentage of different diagnostic methods and intuitively display the performance of each diagnostic method.Through ranking, the most efficient diagnostic way can be obtained.A high SUCRA value indicates a high diagnostic performance rating for this diagnostic method [16].The network funnel plot can show whether there is publication bias in the included literature, and the symmetry criterion is used for visual examination.If the funnel plot is asymmetrical, publication bias is likely to occur [17].

Sensitivity
This study showed a big difference in the sensitivity of DSCT [MD equals 0.65, 95% confidence interval: (0.01, 1.29)] compared to CCTA.

Publication Bias Examination
Funnel plots were drawn for Se, Sp, PLR, NLR and Accuracy to determine possible publication bias.Visually, the overall publication bias of the literature was small.The details are shown in Fig. 9.

Discussion
Among various diagnostic methods for LA/LAA thrombosis in clinical practice, TEE is still the gold standard for recommendations.In this study, network metaanalysis was performed on different diagnostic methods to compare the diagnostic value of different diagnostic techniques for LA/LAA thrombosis detection, so as to provide evidence for clinical application.Our study participants included 18 articles (Ref.[17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32]34,35]), including 4102 patients and 10 kinds of diagnostic methods.The ranking results of the NMA demonstrated that in terms of sensitivity, DSCT is the best (SUCRA value of 71.7%), followed by 3minute delayed CCT (66.8%), which was better than TEE (SUCRA value of 57.5%).In terms of specificity, DSCT is the best (SUCRA value of 84.3%), followed by 3D-CMRI (SUCRA value of 78.0%), which was better than TEE (SU-CRA value of 66.6%).In terms of PLR, 6-minute delayed CCT (SUCRA value of 85.6%) was superior to 3-minute delayed CCT (SUCRA value of 80.1%), both of which were superior to TEE (SUCRA value of 69.1%).DSCT (SU-CRA value of 89.3%) had the highest NLR, while DSCT (SUCRA value of 79.9%) had the highest accuracy.This study demonstrated that besides the gold standard, DSCT was the best in diagnosing LA/LAA thrombosis.In addition, 3D-CMRI and 3-minute delayed CCT are expected to replace TEE.Through the literature review, several metaanalyses were conducted on two/ more methods for diagnosing thrombus of LA/LAA.However, this study conducted a comprehensive network meta-analysis of these diagnostic methods.
Because of LAA's complex structure and physiological characteristics, finding a substitution to TEE for the diagnosis of LAA thrombosis is challenging.CMRI is a safe and non-invasive examination method that does not require sedation, iodized contrast agents/ionizing radiation.A meta-analysis showed that [36] CMRI has a sensitivity of 44-100%, a specificity of 67-100%, an active forecast result of 50-100%, a passive forecast result of 29-100%, and the SUCRA value is 0.93 in the diagnosis of thrombus of LA/LAA.This confirms that CMRI is a reliable diagnostic method to detect thrombus of LA/LAA.It was shown that CMRI can evaluate thrombus of LA/LAA among patients suffering from non-rheumatic atrial fibrillation as well as patients who once had a stroke [37].They concluded that the TEE and CMRI 100% of consistency in terms of detection LAA blood clots.Rathi et al. [38] showed that CMRI detection of LAA/LA thrombus is specific to TEE and can detect intracardiac thrombus other than the LAA/LA.In addition, 3D-CMRI can clearly display   pulmonary vein anatomy, identify structural variations such as pulmonary vein stenosis, and provide additional cardiac electrophysiological information such as LA fibrosis [39].CMRI is safe and non-invasive for the diagnosis of thrombus of LA/LAA in patients with atrial fibrillation, but compared with TEE, CMRI has high cost and high technical requirements, which hinders its wide clinical development.
A computer tomography (CT) meta-analysis consisting of 9 researches [40] revealed that the average sensitivity as well as specificity of CCT within diagnosing thrombus of LA/LAA in patients with AF were 81% (95% confidence interval: 70-90%) and 90% (95% confidence interval: 88-91%), respectively.Five types of CT were included in this study, including MDCT, DSCT, CCTA, delayed CT and non-delayed CT.The outcomes demonstrated that DSCT was better than TEE in terms of sensitivity, specificity, NLR and accuracy.DSCT is the latest technology emerging in the development process of CT.Due to the fact that respiration, heartbeat, and AF can cause PV and LA movements, which affects the detection ability of conventional CT, AF is considered a contraindication for single source CT cardiac angiography [41].DSCT improves the time resolution of synchronous cardiac scanning and proves that DSCT can successfully image the heart and coronary arteries at high heart rates and complex rhythms.In addition, the combination of DSCT and electrocardiogram gating technology quickly covers the heart, resulting in a reduction in radiation exposure compared to standard MDCT [42,43].Due to the shortened examination time and reduced use of contrast agents, the safety for renal insufficiency patients has improved [44].A meta-analysis suggests that low-dose angiography also maintains the same accuracy [27].With the continuous development of technology, DSCT needs to further improve image reconstruction and cross scattering radiation technology, providing greater utilization for clinical and academic applications.
The limitations are as follows.First, most of the selected researches performed TEE and another test method within one week, but one study completed both tests within one month.LAA/LAA thrombus may have formed/ dissolved between the two examinations.Second, the sample sizes of some studies were small.Some patients have been excluded due to examination contraindications, which may have affected the prevalence of LAA/LAA thrombus.Hopefully, there will be more studies to increase the sample size of different diagnostic methods to further verify the accuracy of the test.

Conclusions
This study included 10 diagnostic methods: 2D-CMRI, 3D-CMRI, CCTA, MDCT, DSCT, non-delayed CCT, 1-minute delayed CCT, 3-minute delayed CCT, 6minute delayed CCT, and 3D-TEE; and the results showed that DSCT is superior to TEE in the sensitivity, specificity, NLR, and accuracy to determine thrombus of LA/LAA in AF patients.The outcomes will need to be substantiated by further research.