**1. Introduction**

Stroke is the leading cause of death and disability worldwide [1]. Ischemic stroke, which accounts for more than 70% of the overall incidence of stroke in developed countries, has various causes, such as large artery atherosclerosis in cerebral circulation, occlusion of cerebral small vessels, and cardiac embolism [2]. Of these causes, cardiac embolism contributes most to the increasing incidence of ischemic stroke [3]. Atrial fibrillation (AF) also independently contributes to the increased occurrence of ischemic stroke and is the most common sustained arrhythmia in older adults. In nonvalvular AF, the left atrial appendage (LAA) is the location most susceptible to thrombus formation, accounting for more than 90% of cases [4].

Even though LAA is the prime location of thrombus formation in AF patients, accumulative evidence shows that LAA thrombus may also occur in patients with sinus rhythm or even subclinical AF [5]. Through the advances and widespread use of medical devices, more cryptogenic strokes have been found to be related to subclinical AF [6].

**Citation:** Cheng, Y.-Y.; Tan, S.; Hong, C.-T.; Yang, C.-C.; Chan, L. Left Atrial Appendage Thrombosis and Oral Anticoagulants: A Meta-Analysis of Risk and Treatment Response. *J. Cardiovasc. Dev. Dis.* **2022**, *9*, 351. https://doi.org/ 10.3390/jcdd9100351

Academic Editor: Giovanni Cimmino

Received: 22 August 2022 Accepted: 5 October 2022 Published: 13 October 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Identifying the cause of stroke is vital in achieving optimal therapeutic strategies for the treatment and prevention of recurrent stroke [7]. Initiation of anticoagulation therapy with vitamin K antagonist (VKA) is the most common and conventional strategy employed for LAA thrombus (LAAT) [8]. However, these practices are slowly changing after the launch of the nonvitamin K direct oral anticoagulant (DOAC) in 2002. The introduction of the IIa inhibitor, dabigatran, and Xa inhibitors rivaroxaban, apixaban, and edoxaban in the millennium year has proved that these anticoagulants were at least as effective as VKA in AF for stroke prevention [9].

The safety profiles of DOAC have been highly recognized in many meta-analytic studies and healthcare databases [10,11]. Given that VKA requires regular coagulation monitoring and the potential effects from its interactions with drugs and food [12], DOAC's high efficacy and reliable safety profile are preferred over VKA in current clinical settings. Therefore, DOAC is now generally accepted as the treatment of choice over VKA in patients with nonvalvular AF [11,13].

This trend of switching from VKA to DOAC is not limited to the prevention of strokes from nonvalvular AF; it also extends to other forms of thromboembolism, such as deep venous thrombosis. Moreover, many emerging studies assessed the comparability of DOAC against VKA for LAAT prevention and resolution. However, the optimal treatment for LAAT is yet to be established.

To the best of our knowledge, not many large-scale randomized controlled trials have attempted to verify the differences between the roles of VKA and DOAC in the risk of LAAT formation and rate of thrombus resolution. Furthermore, the lack of large-scale cohort studies has impeded guidelines from being developed that would provide high-level recommendations for LAAT medication management.

The present study is a systematic review of the outcomes of VKA and DOAC use and was performed through an examination of real-world evidence. Further, a meta-analysis of available data was also performed to compare the effectiveness of VKA and DOAC for primary prevention and resolution of LAAT. In this meta-analysis, we included studies providing specific data on the incidence of LAAT and the LAAT resolution rate under VKA or DOAC use.

#### **2. Methods**

#### *2.1. Research Question and Objectives*

In this meta-analysis, we aimed to synthesize evidence to systematically review realworld evidence for a comparison of VKAs and DOAC with respect to their influence on the (i) risk of LAAT and (ii) resolution of LAAT.

#### *2.2. Selection of Articles*

Relevant studies, including case series and clinical trials published before July 2022 were identified from the PubMed, Embase, Google Scholar, and Cochrane databases. Only publications in English were included. We used the following sets of terms in our search: (warfarin (Title/Abstract)) OR (novel oral anticoagulant (Title/Abstract)) OR (oral anticoagulant (Title/Abstract)) OR (anticoagulant (Title/Abstract)) OR (direct oral anticoagulant (Title/Abstract)) OR (vitamin K anticoagulant (Title/Abstract)) OR (non-vitamin K oral anticoagulant (Title/Abstract)) OR (dabigatran (Title/Abstract)) OR (rivaroxaban (Title/Abstract)) OR (apixaban (Title/Abstract)) OR (edoxaban (Title/Abstract)) AND (left atrial appendage thrombus (Title/Abstract)) OR (left atrial thrombus (Title/Abstract)).

The syntax used in the database searches is detailed in the Supplementary Information (Table S1). Duplicate articles from different databases were excluded. The selection process is illustrated in Figure 1. All search records from all databases were downloaded and merged into Endnote.

**Figure 1.** Flow diagram of study selection and search results.

#### *2.3. Study Design*

We included studies (1) with LAA thrombus diagnosed using transesophageal echocardiography (TEE); (2) with clear records of VKA or DOAC anticoagulant use and in which patients were appropriately anticoagulated; (3) that were cohort studies published as original articles, or case series; and (4) that were publications in English.

#### *2.4. Data Extraction*

The selected studies were independently retrieved by two reviewers (Y-Y.C. and C-C.Y.) and were further reviewed by another author (S.T.). The selected studies were reviewed to identify the type of study, year of publication, total patient population, mean patient age, percentage of patients taking VKA, percentage of patients taking DOAC, percentage

of male participants, and mean duration of anticoagulant use. Any disagreements were resolved by a fourth reviewer (C-T.H.).

#### **3. Outcomes**

The efficacy of the primary prevention method was evaluated based on the risk of LAAT. A second comparison was made of the potency of VKA and DOAC in resolving LAAT.

#### *3.1. Synthesis of Results and Measures of Inconsistency*

A random-effects model was implemented to assess LAAT incidence and resolution under VKA and DOAC use. *Q* and *I* <sup>2</sup> were used to assess the level of heterogeneity between the included studies [14]. *Q* is a measure of the weighted sum of the squared deviations of the effect size of each study from the overall mean effect size and thereby serves as a test of heterogeneity significance (*p* ≤ 0.05) [15]. *I* <sup>2</sup> is a measure of relative heterogeneity, estimating the percentage of the variability of effect estimates that occurs due to heterogeneity rather than due to chance. *I* <sup>2</sup> ranges from 0% to 100%, with a value of 0% indicating no observed heterogeneity and a value greater than 50% representing moderate heterogeneity [16]. Furthermore, tau-squared (*T*2) measures the variance of the true effect as an estimate of absolute heterogeneity in effect sizes. When the observed variance increases or when the variance within studies decreases, *T*<sup>2</sup> increases accordingly [15].

#### *3.2. Publication Bias*

We used funnel plots [17], Egger's test [18], and the Begg and Mazumdar rank correlation test [19] to assess publication bias.

#### *3.3. Statistical Analysis*

All statistical analyses were performed using Stata 17. The meta-analysis is registered with PROSPERO (CRD42022319759) and was performed in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines [20]. Standard deviation was calculated using the provided confidence interval (CI) limits, standard errors, or interquartile ranges. The overall risk/odds ratios were pooled using a random-effects model. Publication bias was assessed using the funnel plot of each study's effect size against precision (1/SE). Publication bias was investigated using Egger's test at *p* < 0.10.

#### **4. Results**

#### *4.1. Study Selection*

After removing duplicate studies, we identified 811 articles for screening. After the exclusion of ineligible studies, 48 studies were included in the full-article assessment, and additional 23 studies were excluded because they were case reports, animal studies, or did not contain original data. Finally, 25 studies were included for qualitative synthesis. We further segregated the 25 studies into two categories: (1) cross-sectional risk analysis of developing LAAT under VKA or DOAC use and (2) analysis of LAAT resolution under VKA or DOAC use.

#### *4.2. Study Characteristics*

The characteristics of the included studies on the incidence of LAAT under VKA or DOAC use are listed in Table 1, and the characteristics of those on the LAAT resolution rate under VKA or DOAC use are listed in Table 2. 13 studies were cross-sectional analyses of the incidence of LAAT under VKA and DOAC use (involving 8609 individuals), and 13 were longitudinal analyses of the LAAT resolution rate under VKA and DOAC use (involving 922 individuals).

**Table 1.** Baseline characteristics of included studies for systematic review and meta-analysis of the incidence of left arterial appendage thrombus (LAAT) in patients with atrial fibrillation under vitamin K antagonist (VKA) or direct oral anticoagulant (DOAC) treatment.


nLAAT, patients without LAAT; Dab, dabigatran; Riv, rivaroxaban; Api, apixaban; Edo, edoxaban; INR, ≥ international normalized ratio.


**Table 2.** Baseline characteristics of included studies for systematic review and meta-analysis of LAAT resolution in patients under VKA or DOAC treatment.

Dab, dabigatran; Riv, rivaroxaban; Api, apixaban; Edo, edoxaban; INR, ≥ international normalized ratio.

#### *4.3. LAAT Incidence under VKA and DOAC*

A total of 13 cross-sectional analyses involving 8609 individuals were included in the first meta-analysis of the associations between VKA and DOAC use with LAAT. These studies were published between 2016 and 2022. Among the studies, four were conducted in Poland [27,29,30,33], two were conducted in Japan [25,32], two were conducted in the United States [24,28], and one was conducted each in the EU [26], Germany [31], Canada [21], Turkey [23], and Italy [22]. The selected studies were mostly conducted in Western countries, and only two were conducted in an Asian population (Japan). Six studies were prospective [21–23,26,29,33], and seven were retrospective [24,25,27,28,30–32].

Only the ENSURE-AF trial was a randomized, multicenter, global investigation [26]. The others were mainly single-center studies. Generally, patients were considered sufficiently anticoagulated after at least 3 or 4 weeks of administration of VKA or DOAC. TEE was performed for all study participants to detect the presence of LAAT. The allocation methods for VKA and DOAC were based on the clinician's decision.

The studies included different parameters for predicting the risk of LAAT. A higher CHA2DS2-VASc score [22,23,25,33], reduced left ventricular ejection fraction [22,23,27,33], reduced left atrial flow velocity [23,25,30,32], reduced B-type natriuretic peptide, and larger left atrium [25] were associated with the risk of LAAT. The demographic predictors of the risk of LAAT included aging, a lower body weight, lower creatinine clearance, heart failure, and diuretic treatment were also listed. [26].

#### *4.4. LAAT Resolution Rate under VKA and DOAC Use*

A total of 13 studies involving 922 patients were included in the second meta-analysis. These studies included three in Japan [25,38,40], five in European countries [30,37,39,43,44], three in China [34,36,42], and two in the United States [35,41]. One of the studies, the X-TRA study, was a multinational large-scale, prospective, single-arm, open-label, multicenter study. The X-TRA study evaluated a 6-week rivaroxaban treatment for left atrial and LAA thrombus resolution. Another study, the CLOT-AF study, retrospectively examined standard anticoagulation care provided to patients with left atrial and LAA thrombus for 3 to 12 weeks. These studies were published between 2015 and 2022. When the included study period was extended to 7 months, nine patients were identified as having LAAT [42]. When the included duration and dosage of anticoagulation were increased and considering the transition to DOAC, 12 patients (5%) were identified as experiencing LAAT resolution [41]. Of the 13 included studies, Kawabata, Karwowski, Durmaz, Feickert, and Shiraki studies contained data on the use of dabigatran, rivaroxaban, apixaban, and edoxaban, respectively. Edoxaban was used less frequently than the other three DOACs.

#### *4.5. LAAT Incidence under VKA and DOAC Use*

In the meta-analysis, 2963 patients were in the VKA arm, and 5646 were in the DOAC arm. The overall risk of LAAT under either VKA or DOAC treatment was 5.56% (479/8609). The risk ratio was derived from individual studies. The respective relative risks (RRs) with 95% CIs are listed on the right side of the forest plot (Figure 2A). The meta-analysis of 13 studies on LAAT incidence revealed significant superiority for DOACs (pooled RR = 0.65, 95% CI = 0.47–0.90, *p* = 0.009). Nearly moderate and significant heterogeneity (*Q*<sup>12</sup> = 22.97, *p* = 0.028; *I* <sup>2</sup> = 47.8%; *T*<sup>2</sup> = 0.13) was identified. The funnel plot (Figure 2B) revealed symmetric distribution. Egger's test (intercept = 0.684, t = 0.97, 2-tailed *p* = 0.352) and Begg's test (z = 0.18, *p* = 0.855) did not reveal any publication bias. The meta-regression analysis showed that none of the between-study variables significantly predicted the LAAT incidence under VKA and DOAC use (mean age of participants: β = 0.071, *p* = 0.333; male ratio: β = 0.354, *p* = 0.808).

#### *4.6. LAAT Resolution Rate under VKA and DOAC Use*

The VKA and DOAC arms included 484 and 438 patients, respectively. The summarized mean percentages of LAAT resolution for VKA and DOAC were 55.4% (268/484) and 67.6% (296/438), respectively. The odds ratio was derived from the individual studies (Figure 3A). This meta-analysis revealed that DOAC significantly increased the probability of LAAT resolution compared with VKA (pooled odds ratio = 1.52, 95% CI = 1.02–2.26, *p* = 0.040). In addition, no significant heterogeneity (*Q*<sup>12</sup> = 17.62, *p* = 0.128; *I* <sup>2</sup> = 31.9%; *T*<sup>2</sup> = 0.16) was identified. The funnel plot (Figure 3B), although slightly asymmetric, did not indicate a high risk of publication bias. Egger's test (intercept = 0.383, t = 0.44, 2-tailed *p* = 0.671) and Begg's test (z = 0.55, *p* = 0.583) did not reveal any publication bias. The meta-regression analysis showed that none of the between-study variables significantly
