“Left Atrial Appendage Closure Versus Medical Therapy in Atrial Fibrillation: A Systematic Review and Meta‐Analysis of Randomized Controlled Trials”

ABSTRACT

We aim to compare the impact of left atrial appendage closure vs. medical therapy on clinical outcomes in patients with atrial fibrillation, focusing on key clinical outcomes relevant to thromboembolic prevention. We conducted a PRISMA‐guided systematic review and meta‐analysis of studies comparing LAAC with medical therapy in AF adults. PubMed, Embase, Scopus, ClinicalTrial. gov, and Cochrane Library were searched from inception to March 2026. Outcomes included all‐cause mortality, any stroke, ischemic stroke, hemorrhagic stroke, major bleeding, systemic embolism and cardiac death. Risk ratios (RRs) with 95% confidence intervals (CIs) were pooled using random‐effects models. Reconstructed individual patient‐level survival data from published Kaplan–Meier curves were used for time‐to‐event analyses. Trial sequential analysis (TSA) assessed the conclusiveness of cumulative evidence. Six randomized controlled trials involving 7004 patients were included. LAAC demonstrated comparable outcomes to medical therapy for all‐cause mortality, any stroke, major bleeding, systemic embolism, and cardiac death. There was a non‐significant trend toward increased ischemic stroke and reduced hemorrhagic stroke with LAAC. Reconstructed time‐to‐event analyses showed similar cumulative incidence of stroke and bleeding outcomes over follow‐up. TSA demonstrated that the required information size was not reached for any major endpoint, indicating that current evidence remains underpowered and additional randomized trials are needed. LAAC provides similar overall efficacy and safety compared with medical therapy for stroke prevention in AF but should currently remain reserved for selected patients rather than replacing oral anticoagulation broadly. Further adequately powered trials with longer follow‐up are required.

1. Introduction

Atrial fibrillation (AF) is the most common cardiac arrhythmia worldwide, with prevalence increasing from 33.5 million in 2010 to nearly 60 million in 2019. There can be a fivefold increase in risk of ischemic stroke, higher mortality, and progressive functional and cognitive decline in such patients [1]. In non‐valvular atrial fibrillation, there is mostly left atrial thrombus formation. It predominantly originates from the left atrial appendage [2] which is an embryological remnant of the left atrium and thus, accounts for over 90% of cases [3, 4]. Blood stasis due to reduced atrial contractility, endothelial dysfunction and hypercoagulability, lead to thrombus formation while anatomical variations such as multilobed morphology and altered emptying velocities greatly increase stroke risk [4, 5].

Oral anticoagulant (OAC) therapy is the standard treatment of stroke prevention in AF and is recommended within integrated management strategies such as the ABC pathway, which also addresses symptom control and cardiovascular risk factors. However, residual thromboembolic risk persists despite anticoagulation, partly due to comorbid conditions and underlying atherosclerotic disease [6]. Evidence such as the AFIRE trial supports OAC monotherapy as a safer strategy compared with combination antithrombotic regimens in patients with concomitant cardiovascular disease [7], although optimization of long‐term outcomes continues to be evaluated.

Left atrial appendage closure (LAAC) is emerging as a mechanical intervention to prevent stroke in NVAF patients specially those at high risk of thromboembolism or bleeding [2, 6]. Recent ESC and ACC/AHA guidelines recommend this approach in selected patients with contraindications to long‐term anticoagulation [5, 7]. While LAAC eliminates the primary source of thrombus formation, issues like peridevice leak and incomplete closure may affect outcomes [3, 4, 8]. Existing evidence from randomized trials and observational studies comparing LAAC with OAC keeps varying.

This study aims to compare left atrial appendage closure with oral anticoagulation therapy in patients with non‐valvular atrial fibrillation, focusing on key clinical outcomes relevant to thromboembolic prevention.

2. Methods

2.1. Study Design and Reporting Standards

We conducted this systematic review and meta‐analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA 2020) guidelines [9]. The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO registration number: CRD420261362617).

2.2. Eligibility Criteria

We included randomized controlled trials comparing percutaneous left atrial appendage closure (LAAC) (e.g., WATCHMAN, WATCHMAN FLX, Amulet) with medical therapy, defined as oral anticoagulation (OAC), including vitamin K antagonists (e.g., warfarin) and direct oral anticoagulants (DOACs), in adult patients (≥ 18 years) with atrial fibrillation.

The outcomes were: all‐cause mortality, any stroke, ischemic stroke, hemorrhagic stroke, major bleeding, systemic embolism and cardiac death.

We excluded observational studies, case reports, reviews, meta‐analyses, editorials, conference abstracts, animal studies, single‐arm studies, and trials that did not report relevant clinical outcomes. In cases of multiple publications from the same trial, the report with the longest follow‐up or most comprehensive data was included.

2.3. Search Strategy and Study Selection

A comprehensive literature search was performed in PubMed, Embase, Scopus, ClinicalTrial.gov and the Cochrane Library, from database inception to 31 March 2026, without language restrictions.

The search strategy incorporated both controlled vocabulary (e.g., MeSH terms) and free‐text keywords related to: (1) atrial fibrillation, (2) left atrial appendage closure, (3) medical therapy (including oral anticoagulation with vitamin K antagonists and direct oral anticoagulants), and (4) randomized controlled trials. The full search strategy is provided in Supporting Information S1: Table 1. Additionally, the reference lists of relevant reviews and included studies were manually screened to identify any potentially eligible studies not captured in the database search. Study selection was conducted independently by two reviewers, who screened titles and abstracts for eligibility. Full‐text articles were subsequently assessed to confirm inclusion. Discrepancies were resolved through discussion or consultation with a third reviewer.

2.4. Data Extraction and Outcomes

Two investigators independently extracted data from each eligible study using a standardized electronic Google sheet. Extracted information included study characteristics (first author, year, country, design, and data collection period), population details (inclusion criteria, baseline demographics, comorbidities, and sample size), intervention and comparator details (type of LAAC device, type of anticoagulation). The clinical outcomes of interest were all‐cause mortality, any stroke, ischemic stroke, hemorrhagic stroke, major bleeding, systemic embolism and cardiac death. Two outcomes reported as time‐to‐event data, time‐to‐event data were extracted from the published Kaplan–Meier curves using WebPlotDigitizer [10]. We then reconstructed the curves to obtain individual patient data (IPD) following two possible pathways. When studies reported the numbers at risk, IPD reconstruction was performed using the IPDfromKM R package developed by Liu et al. [11]. In the absence of a risk table, we applied the reconstruction procedure and accompanying R code outlined by Guyot et al. [12]. These reconstructed data were used to derive pooled hazard estimates and to visualize cumulative incidence over time, allowing a more detailed comparison between LAAC and medical therapy.

2.5. Risk of Bias and Certainty of Evidence

Methodologic quality was assessed using the Cochrane Risk of Bias 2.0 tool for randomized trials [13] assessing the following domains: (1) Bias arising from the randomization process, (2) Bias due to deviations from intended interventions, (3) Bias due to missing outcome data, (4) Bias in measurement of outcomes, and (5) Bias in selection of reported results. Each domain was classified as low risk, some concerns, or high risk of bias. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework was applied to assess the certainty of evidence for each clinical outcome, considering study limitations, inconsistency, indirectness, imprecision, and publication bias [14].

2.6. Statistical Analysis

Pooled effect estimates were calculated as risk ratios (RRs) with 95% confidence intervals (CIs) for dichotomous outcomes. Heterogeneity among studies was assessed using the I ² statistic, with values of 25%, 50%, and 75% representing low, moderate, and high heterogeneity, respectively. A random‐effects model was applied to account for potential clinical and methodological variability across studies. Statistical significance was defined as a two‐tailed p‐value < 0.05. Sensitivity analyses were performed using a leave‐one‐out approach to evaluate the stability of the pooled results. To explore potential sources of clinical heterogeneity, subgroup analyses were performed according to the type of oral anticoagulant used in the control group. Trials were categorized as those comparing left atrial appendage closure (LAAC) with vitamin K antagonists (warfarin) or with direct oral anticoagulants (DOACs). Because the medical therapy arm of the CLOSURE‐AF trial included a mixed anticoagulation strategy consisting of both vitamin K antagonists and DOACs, this trial was not included in the comparator‐specific subgroup analyses but remained included in the primary meta‐analysis. Subgroup analyses were performed using the same statistical approach as the primary analyses. Differences between subgroups were evaluated using the χ ² test for subgroup differences implemented in Review Manager (RevMan), with a two‐sided p value < 0.05 considered statistically significant. All analyses were performed using Review Manager (RevMan) version 5.4, and figures were refined using appropriate statistical software. Trial sequential analysis (TSA) was performed using TSA Viewer (http://www.ctu.dk/tsa/) with a random‐effects DerSimonian–Laird [15] model, two‐sided α = 5%, 80% power [16], and O'Brien–Fleming monitoring boundaries. TSA was conducted to account for repeated significance testing and assess whether the accumulated evidence was sufficient for firm conclusions. Outcome‐specific required information sizes were calculated using pooled control‐event rates, anticipated intervention effects, and adjustments for heterogeneity [17]. The assumed effects were 10.62% for all‐cause mortality, 5.36% for cardiac death, 3.78% for any stroke, 3.42% for ischemic stroke, 0.65% for hemorrhagic stroke, 0.24% for systemic embolization, and 6.38% for major bleeding; corresponding heterogeneity corrections were 36%, 68%, 29%, 0%, 16%, 0%, and 0%, respectively. These assumptions were selected according to the control‐event frequency, observed effect magnitude, and clinical plausibility of each outcome. Evidence was considered conclusive when the cumulative Z‐curve crossed a monitoring boundary or reached the required information size. Time‐to‐event endpoints were analyzed using a one‐stage framework built on the reconstructed IPD for freedom from all‐cause mortality. We fitted a Cox proportional hazards model, stratifying by study, to derive hazard ratios. All statistical procedures were executed in R (version 4.3.3) with the IPDfromKM package and supporting libraries.

2.7. Reconstructed Time‐To‐Event Analysis

To evaluate long‐term time‐to‐event outcomes, reconstructed individual patient‐level data (IPD) analyses were performed for outcomes with published Kaplan–Meier curves of sufficient quality for reconstruction. Because Kaplan–Meier curves were not consistently reported across all randomized controlled trials or for all outcomes, reconstructed analyses were restricted to studies providing analyzable survival curves together with adequate numbers‐at‐risk information.

For the stroke outcome, reconstructed analyses included the PROTECT AF, CLOSURE‐AF, and CHAMPION‐AF trials. For major bleeding, reconstructed analyses included the PRAGUE‐17 and CHAMPION‐AF trials.

Published Kaplan–Meier curves were digitized using WebPlotDigitizer [10]. Individual patient‐level time‐to‐event data were subsequently reconstructed using the IPDfromKM package in R according to the algorithm described by Guyot et al., which reconstructs individual survival data from digitized Kaplan–Meier coordinates, reported numbers at risk, and total events.

The accuracy of the reconstructed data sets was assessed by visually comparing the reconstructed Kaplan–Meier curves with the original published curves and by verifying agreement between reconstructed and reported numbers at risk and cumulative event estimates whenever these data were available. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated from the reconstructed IPD using Cox proportional hazards regression models.

The proportional hazards assumption was assessed by visual inspection of the Kaplan–Meier curves and log‐minus‐log survival plots for evidence of substantial crossing of survival functions. No major violations of the proportional hazards assumption were identified.

Because reconstructed analyses were limited to trials reporting analyzable Kaplan–Meier curves for the outcome of interest, the reconstructed population represented only a subset of the total randomized population included in the aggregate‐data meta‐analysis. Accordingly, the reconstructed time‐to‐event analyses were considered complementary to the primary meta‐analysis rather than a replacement for it.

3. Results

3.1. Study Selection and Database Results

Our systematic review identified 7438 records from PubMed, Embase, Scopus, ClinicalTrial.gov, and Cochrane. After removing 2667 duplicate records, 4771 records were screened. A total of 4659 records were excluded through title/abstract screening, followed by the exclusion of 106 articles during the full‐text review for wrong study design, non‐randomized design, lack of appropriate comparator, absence of relevant outcomes or being abstracts. Ultimately, six studies [18, 19, 20, 21, 22, 23] were included in the systematic review. The PRISMA flow diagram illustrates the study selection process (Figure 1), Finally, six trials, comprising six RCTs (PROTECT‐AF, PREVAIL, PRAGUE‐17, OPTION, CLOSURE‐AF and CHAMPION‐AF), met the predefined inclusion criteria and were included in our meta‐analysis. The baseline characteristics were generally well‐balanced between the LACC and medical therapy groups across the six studies. Patients were predominantly elderly (mean age 69–79 years) with a high prevalence of cardiovascular comorbidities, including hypertension, heart failure, and coronary artery disease. Representing study characteristics, population demographics and characteristics presented in (Table 1).

Figure 1.

The preferred reporting items for systematic reviews and meta‐analyses (PRISMA) flow diagram.

Table 1.

Summary of the included studies.

study number	Study ID	Year	Country	Study Design	Arms	Arms (n)	Follow‐Up Duration	key Outcomes	NCT
1	PROTECT‐AF	2014	United States and Europe (MS)	RCT	LAAC (WATCHMAN)	463	4 years	Stroke, systemic embolism, or cardiovascular death	NCT00129545
					Warfarin	244
2	PREVAIL	2018	United States	RCT	LAAC (WATCHMAN)	269	1.5 years	Composite of stroke, cardiovascular death, and systematic embolism	NCT01182441
					Warfarin	138
3	PRAGUE‐17	2022	Czech Republic (MS)	RCT	LAAC (WATCHMAN)	201	4 Years	primary endpoint was a composite of cardioembolic events (stroke, transient ischemic attack, or systemic embolism), cardiovascular death, clinically relevant bleeding, or procedure‐/device‐related complications	NCT02426944
					DOACs	201
4	OPTION	2025	Multicenter (USA, Europe, Australia, global sites) (MS)	RCT	LAAC (WATCHMAN)	803	3 years	all‐cause death, stroke, or systemic embolism	NCT03795298
					OACs	797
5	CLOSURE‐AF	2026	Germany (MS)	RCT	LAAC (WATCHMAN)	446	medianof 3 years (IQR 1.7 to 4.7)	composite of stroke (ischemic or hemorrhagic), systemic embolism, major bleeding, or cardiovascular or unexplained death	NCT03463317
					Medical therapy	442
6	CHAMPION‐AF	2026	International (MS)	RCT	LAAC	1499	3 years	Death from cardiovascular causes, stroke, or systemic embolism	NCT04394546
					OACs	1501

3.2. Risk of Bias Assessment and Certainty of Evidence

For all randomized trials (PROTECT‐AF, PREVAIL, PRAGUE‐17, OPTION, CLOSURE‐AF and CHAMPION‐AF), the overall risk of bias was judged as low across most domains (Supporting Information S1: Table S2). According to the GRADE approach, the certainty of evidence was moderate for all‐cause mortality, major bleeding, ischemic stroke, hemorrhagic stroke, any stroke and systemic embolization while it is low for cardiac death. A detailed summary of findings is presented in (Supporting Information S1: Table S3) (Table 2).

Table 2.

The baseline characteristics.

Study number	Study ID	Arms	Age, mean (SD)	Atrial fibrillation, n (%)	Classification of AF, No. (%)				Comorbidities
					Paroxysmal	Persistent	Permanent	Unknown	DM n (%)	Prior stroke/TIA n (%)	HF n (%)	MI n (%)	PAD n (%)	Renal Disease n (%)	CHA2DS2‐VASc score, mean (SD)
1	PROTECT‐AF	LAAC	71.7 ± 8.8	463 (100)	200 (43.2)	97 (21.0)	160 (34.6)	6 (1.3)	113 (24.4)	82 (17.7)	124 (26.8)	—	—	—	2.2 ± 1.2
		Warfarin	72.7 ± 9.2	244 (100)	99 (40.6)	50 (20.5)	93 (38.1)	2 (0.8)	72 (29.5)	49 (20.1)	66 (27.0)	—	—	—	2.3 ± 1.2
2	PREVAIL	LAAC	74.0 ± 7.4	—	—	—	—	—	—	—	—	—	—	—	2.6 ± 1.0
		Warfarin	74.9 ± 7.2	—	—	—	—	—	—	—	—	—	—	—	2.6 ± 1.0
3	PRAGUE‐17	LAAC	73.4 ± 6.7	201 (100)	53 (26.4)	65 (32.4)	83 (41.3)	—	73 (36.3)	66 (32.8)	88 (43.8)	30 (14.9)	—	—	4.7 ± 1.5
		DOACs	73.2 ± 7.2	201 (100)	67 (33.3)	62 (30.9)	72 (35.8)	—	90 (44.8)	63 (91.3)	90 (44.8)	39 (19.4)	—	—	4.7 ± 1.5
4	OPTION	LAAC	69.6 ± 7.3		477 (59.4)	326 (40.6)	—	—	222 (27.6)	80 (10.0)	161 (20.0)	—	—	—	3.5 ± 1.3
		OACs	69.4 ± 7.9		501 (62.8)	296 (37.1)	—	—	221 (27.7)	92 (11.5)	166 (20.8)	—	—	—	3.5 ± 1.3
5	CLOSURE‐AF	LAAC	78.5 ± 6.8	—	—	—	—	—	175 (39.2)	142 (31.9)	—	—	59 (13.3)	114 (25.6)	5.2 ± 1.5
		Medical therapy	77.3 ± 7.3	—	—	—	—	—	186 (42.1)	151 (34.2)	—	—	49 (11.2)	100 (22.6)	5.1 ± 1.6
6	CHAMPION‐AF	LAAC	71.6 ± 7.5	1498 (100)	1038 (69.3)	358 (23.9)	102 (6.8)	—	—	123 (8.2)	—	—	—	—	3.5 ± 1.2
		NOAC	71.8 ± 7.5	1501 (100)	1028 (68.5)	382 (25.4)	91 (6.1)	—	—	124 (8.3)	—	—	—	—	3.5 ± 1.3

3.3. Outcomes

3.3.1. Survival Free From Stroke and Bleeding, Assessed Using Time‐To‐Event Methods

Reconstructed time‐to‐event analyses based on Kaplan–Meier curves demonstrated similar cumulative incidence of stroke and bleeding between left atrial appendage closure and medical therapy over follow‐up (Figure 2). Reconstructed time‐to‐event analyses were performed only for studies reporting analyzable Kaplan–Meier curves. Stroke analyses included the PROTECT AF, CLOSURE‐AF, and CHAMPION‐AF trials, whereas major bleeding analyses included the PRAGUE‐17 and CHAMPION‐AF trials. Consequently, fewer participants contributed to the reconstructed analyses than to the conventional aggregate‐data meta‐analysis. The reconstructed Kaplan–Meier curves demonstrated good agreement with the original published survival curves, supporting the validity of the reconstructed individual patient‐level datasets.

Figure 2.

Kaplan–Meier curve: stroke and bleeding, follow‐up in months.

3.3.2. All‐Cause Mortality

A total of 6 studies were included in this analysis, with 3681 participants in the LAAC group and 3323 in the control group. The overall RR was 0.90 [95% CI: 0.76 to 1.07, p = 0.23], indicating no significant difference between LAAC and the medical therapy group. It was associated with low heterogenicity (I ² = 36%, p = 0.17) (Figure 3A).

Figure 3.

forest plots to (A) all‐cause mortality, (B) major bleeding, (C) cardiac death.

3.3.3. Major Bleeding

A meta‐analysis of 6 studies involving 3681 participants in LAAC group and 3323 in the medical therapy group revealed no statistically significant difference between both groups with RR of 0.96 [95% CI: 0.81 to 1.15, p = 0.67] and was associated with no heterogenicity (I ² = 0%, p = 0.52) (Figure 3B).

3.3.4. Cardiac Death

A total of 5 studies were included in this analysis, with 3412 participants in the LAAC group and 3185 in the medical therapy group. The overall pooled effect was comparable between both groups with RR 0.97 [95% CI: 0.80 to 1.17, p = 0.73], and the overall effect was non‐significant associated with a high heterogeneity (I ² = 68%, p = 0.01) (Figure 3C).

In leave‐one‐out analysis, excluding the PROTECT‐AF trial from the analysis reduced the heterogenicity to 26% but the effect did not reach significance.

3.3.5. Ischemic Stroke

A meta‐analysis of 6 studies involving 3681 and 3323 participants in the LAAC and medical therapy group respectively revealed that that the ischemic stroke was greater in the LAAC group as compared to the medical therapy group, but the effect did not reach significance 1.30 [95% CI: 0.99 to 1.72, p = 0.06]. This pooled effect was associated with no heterogenicity (I ² = 0%, p = 0.67) (Figure 4A).

Figure 4.

forest plots to (A) ischemic stroke, (B) hemorrhagic stroke, (C) any stroke, (D) systemic embolization.

3.3.6. Hemorrhagic Stroke

A meta‐analysis of 6 studies involving 3681 and 3323 participants in the LAAC and medical therapy group respectively revealed that the probability of hemorrhagic stroke was reduced in the LAAC group as compared to the medical therapy group, but the effect did not reach significance 0.56 [95% CI: 0.31 to 1.02, p = 0.06]. This pooled effect was associated with low heterogeneity (I ² = 16%, p = 0.31) (Figure 4B).

3.3.7. Any Stroke

A total of 5 studies were included in this analysis, with 3412 participants in the LAAC group and 3185 in the medical therapy group. The overall pooled effect was comparable between both groups with RR 1.03 [95% CI: 0.76 to 1.40, p = 0.85], and the overall effect was non‐significant associated with a low heterogeneity (I ² = 29%, p = 0.23) (Figure 4C).

3.3.8. Systemic Embolism

A meta‐analysis of 6 studies involving 3681 and 3323 participants in the LAAC and medical therapy group respectively revealed that that the ischemic stroke was greater in the LAAC group as compared to the medical therapy group, but the effect did not reach significance 1.36 [95% CI: 0.44 to 4.19, p = 0.59]. This pooled effect was associated with no heterogenicity (I ² = 0%, p = 0.66) (Figure 4D).

3.4. Trial Sequential Analysis (TSA)

Trial sequential analysis (TSA) was conducted to assess the conclusiveness of evidence. The cumulative Z‐curve did not reach the required information size (RIS) for any of the mentioned outcomes (see Figure 5). Therefore, both type I and type II errors cannot be excluded. This suggests that additional randomized trials are required to achieve conclusive evidence. For all‐cause mortality and cardiac death, the cumulative evidence represents only 1.62% and 0.88% of the RIS, respectively.

Figure 5.

showing trial sequential analysis for (A) any stroke, (B) hemorrhagic stroke, (C) ischemic stroke, (D) major bleeding, and (E) Systemic embolization.

3.5. Subgroup Analysis According to Comparator Therapy

To explore potential sources of clinical heterogeneity, subgroup analyses were performed according to the type of oral anticoagulant comparator (warfarin vs. direct oral anticoagulants [DOACs]). The CLOSURE‐AF trial was excluded from these analyses because the medical therapy arm included a mixed anticoagulation strategy.

No statistically significant subgroup differences were observed for all‐cause mortality, major bleeding, ischemic stroke, any stroke, or systemic embolization (all P for subgroup difference > 0.05). However, significant subgroup interactions were identified for cardiac death (P for subgroup difference = 0.004) and hemorrhagic stroke (P for subgroup difference = 0.02). Specifically, trials comparing LAAC with warfarin demonstrated a greater reduction in hemorrhagic stroke and cardiac death than trials comparing LAAC with DOACs. Given the limited number of studies within each subgroup and the exploratory nature of these analyses, these findings should be interpreted with caution. The detailed subgroup analyses are presented in Supporting Information S1: Figures S1 and S2.

4. Discussion

This systematic review and meta‐analysis included six RCTs, involving 7004 patients with AF. Our aim was to determine if LAAC is more effective and/or safer than medical treatment in patients with AF. The studies included were PROTECT‐AF, PREVAIL, PRAGUE‐17, OPTION, CLOSURE‐AF, CHAMPION‐AF, some comparing LAAC with warfarin and some comparing LAAC with a DOAC.

Our primary results found LAAC was associated with similar outcomes compared with medical therapy. We found no difference in all‐cause mortality, major bleeding, cardiac death, any stroke or systemic embolism. There was a trend towards more ischemic stroke with LAAC and less hemorrhagic stroke with LAAC, but this was not statistically significant. Furthermore, we reconstructed the Kaplan‐Meier curves for stroke and bleeding, which also showed good agreement with the original published curves and showed comparable cumulative incidences of stroke and major bleeding over time, supporting the validity of the reconstructed individual patient‐level data sets despite the smaller number of included participants. Subgroup analyses according to comparator therapy (warfarin vs. DOAC) demonstrated no significant subgroup differences for all‐cause mortality, major bleeding, ischemic stroke, any stroke, or systemic embolization, indicating that the overall treatment effect of LAAC was generally consistent regardless of the comparator anticoagulant. However, significant subgroup interactions were observed for cardiac death and hemorrhagic stroke, with greater reductions favoring LAAC when compared with warfarin than with DOACs. These findings suggest that the relative clinical benefit of LAAC may be more pronounced against warfarin, particularly for bleeding‐related outcomes, although the limited number of studies within each subgroup warrants cautious interpretation. Crucially, TSA demonstrated that the required information size was not achieved for all outcomes, implying the current evidence is still underpowered and more RCTs are needed.

These results are supported by current guidelines, which recommend LAAC primarily for a select group of patients rather than as first‐line treatment. The 2023 ACC/AHA/ACCP/HRS AF guideline [24] and the 2024 ESC guideline for AF management [25] recommend LAAC in patients with non‐valvular AF who are at high risk of stroke and contraindicated to long‐term use of anticoagulation, especially patients at high risk of bleeding. OAC, particularly DOACs, is the mainstay approach for stroke prevention in most patients with AF [26, 27, 28]. Thus, current evidence supports LAAC as an alternative for selected patients, but not a replacement for all patients.

Our results are in line with previous landmark trials and meta‐analyses [29, 30, 31, 32, 33, 34, 35]. The PROTECT‐AF and PREVAIL trials showed non‐inferiority of the WATCHMAN device to warfarin, with specific reduction in hemorrhagic stroke and cardiovascular death in long‐term follow‐up [32, 36]. However, these trials were conducted in the era of warfarin, which may be beneficial for LAAC given warfarin is associated with more bleeding complications than contemporary DOACs [37, 38, 39]. In contrast, PRAGUE‐17 compared LAAC versus DOACs and found non‐inferiority for the composite endpoint, but unclear superiority for individual outcomes [40]. Our meta‐analysis supports this finding, including more recent trials, such as OPTION [21], CLOSURE‐AF [22], and CHAMPION‐AF [23], which are more representative of current clinical practice. Despite these findings, important clinical heterogeneity should be considered when interpreting the pooled estimates. The included trials differed with respect to comparator therapy (warfarin vs. DOACs), patient selection, bleeding risk, device generations, procedural experience, follow‐up duration, and clinical indications. For example, OPTION enrolled patients after AF catheter ablation, whereas CLOSURE‐AF focused on patients at high bleeding risk or with contraindications to long‐term anticoagulation, while the earlier WATCHMAN trials primarily compared LAAC with warfarin. Although statistical heterogeneity was generally low for many outcomes, these clinical differences may have influenced the magnitude of the pooled treatment effects.

The increased risk of ischemic stroke with LAAC needs to be put in context. It is important to note that warfarin offers systemic thromboembolic protection [41, 42], but LAAC is specific for preventing thrombus formation from the LAA [43], This suggests that other non‐LAA thromboembolic sources, such as atrial cardiomyopathy, vascular disease, or other embolic mechanisms, may remain after LAAC. Also, thrombus on a device, incomplete closure of the appendage and leaks around the device may be a source of stroke. This is likely why LAAC did not reduce ischemic stroke despite closure of the LAA.

In addition, we found LAAC reduced the risk of hemorrhagic stroke, which is both biologically plausible and clinically relevant. Chronic anticoagulation carries a cumulative bleeding risk [44, 45]. In fact, this was shown to be higher for vitamin K antagonists (notably warfarin) but not for direct OAC (DOAC). The risk of major bleeding with DOAC was found to be lower than warfarin, while the safest DOAC depends on other factors such as the type of bleeding and the race of the population [46, 47]. However, the major bleeding event of interest is intracranial hemorrhage [48], a consequence of OAC therapy [49], LAAC may reduce the long‐term risk of this bleeding by avoiding the need for lifelong anticoagulation in many patients. Although our combined estimate was not significant, the direction of effect is consistent with current evidence and justifies LAAC in patients at risk of major bleeding.

The lack of significant difference in major bleeding in our study is due to the combination of early bleeding events with LAAC compared with the long‐term bleeding events due to medication. Early complications, including access‐site bleeding, pericardial effusion and device complications may have negated the subsequent benefit of reduced bleedings. In contemporary experience, risk of these complications has declined with operator and device experience, which may enhance the benefit/risk profile of LAAC in future trials.

5. Closure‐AF Trial: Comparative Efficacy of LAAC and OAC

The CLOSURE‐AF trial [22], a multicenter randomized noninferiority study including 912 high‐risk AF patients, showed that LAAC did not show noninferiority to conventional oral anticoagulation and was linked to an increased risk of the composite outcome of stroke, systemic embolism, cardiovascular or unknown cause of death and major bleeding. The increased risk in the LAAC group was due to immediate periprocedural complications and increased cardiovascular death; some of the bleeding events also occurred in the early post‐implantation period. This raises doubts about the superiority of LAAC in high‐risk patients and suggests the continued effectiveness of current state‐of‐the‐art anticoagulation with DOAC. Overall, the findings support the need for personalized treatment and are in line with the current guidelines that classify LAAC as a Class IIb device for which larger randomized studies are needed.

6. Clinical Implications and Future Directions

Our findings suggest that our research has shown that patient selection is critical when considering LAAC compared with medical care. LAAC is not a better alternative to OAC in all patients with AF, but it is of great value in those with contraindications to OAC, recurrent bleeding, lack of adherence to medications or a strong preference for avoiding OAC in the long term. It's crucial to work with patients to consider the risk of stroke, bleeding, procedural complications, existing medical conditions, and treatment preferences. Thus, the use of LAAC in the management of AF should remain personalized

Future investigations should include well‐powered, randomized trials of LAAC vs contemporary DOAC approaches, particularly in patients at high risk of bleeding. Further follow‐up is required to more accurately estimate late stroke protection, longevity of the device and mortality. Furthermore, reporting of device thrombus, peri‐device leak and post‐procedural antiplatelet therapy is needed, as these factors impact outcomes. Besides, trials of newer devices, such as WATCHMAN FLX and Amulet may better reflect current practice than older WATCHMAN trials. Although the observed direction of effect for ischemic and hemorrhagic stroke is biologically plausible, these findings did not reach statistical significance and should therefore be interpreted as hypothesis‐generating rather than conclusive evidence of differential efficacy. LAAC may reduce hemorrhagic stroke by avoiding long‐term exposure to anticoagulation, whereas residual ischemic stroke risk may persist because LAAC does not provide systemic thromboembolic protection and does not address non‐LAA sources of embolism. However, given the nonsignificant pooled estimates and the underpowered cumulative evidence demonstrated by TSA, these trends should be interpreted cautiously.

Procedural risk is also essential when interpreting bleeding outcomes after LAAC. Any potential long‐term reduction in anticoagulation‐related bleeding may be partially offset by early peri‐procedural complications, including access‐site bleeding, pericardial effusion, device‐related thrombus, peri‐device leak, and procedure‐related stroke or bleeding. Therefore, the overall clinical value of LAAC should be judged according to net clinical benefit rather than late bleeding outcomes alone. Future randomized trials should standardize the reporting of peri‐procedural complications and incorporate them into composite safety and net clinical benefit analyses to better define which patients derive the greatest advantage from LAAC.

7. Strengths and Limitations

Our study has several key strengths. We incorporated only RCTs, which reduces selection bias and improves internal validity. We included the newest large‐scale RCTs, including CLOSURE‐AF [22] and CHAMPION‐AF [23] which provide an updated analysis for the current evidence. The use of reconstructed individual patient data from Kaplan–Meier curves allowed more comprehensive time‐to‐event assessment. Besides, we conducted TSA, which strengthened our findings by addressing evidence sufficiency and the risk of false‐positive in our results.

However, some limitations should be addressed. First, only six RCTs were available, and several outcomes remained underpowered, as confirmed by TSA. Second, heterogeneity in comparator therapy, as some RCTs used warfarin and others used DOACs. This may affect pooled estimates of our analysis. Third, device generations and operator experience evolved across trials may affect procedural safety and efficacy. Fourth, follow‐up period differed between RCTs, and some procedural outcomes, such as device‐related thrombosis were not consistently reported. Fifth, a limitation of the reconstructed time‐to‐event analyses is that they depended on the availability and quality of published Kaplan–Meier curves. Not all randomized trials reported analyzable survival curves for every outcome; therefore, reconstructed analyses included only a subset of the total randomized population. Furthermore, reconstructed individual patient‐level data are estimates derived from published aggregate information rather than original trial datasets. Although validated reconstruction methods were used and reconstructed curves closely matched the published Kaplan–Meier curves, these analyses should be interpreted as complementary to the primary aggregate‐data meta‐analysis. Finally, individual patient‐level subgroup analyses were not applicable.

8. Conclusion

The efficacy and safety of LAAC for stroke prevention in AF is similar to medical treatment, with similar rates of death, major bleeding and any stroke. The observed nonsignificant trends toward increased ischemic stroke and reduced hemorrhagic stroke with LAAC may suggest a potential trade‐off, but these findings should be considered hypothesis‐generating rather than definitive. Subgroup analyses showed consistent effects across most outcomes, with greater reductions in cardiac death and hemorrhagic stroke versus warfarin than DOACs. Existing data indicate that LAAC is a valuable option for a well‐selected subset of patients, especially those who cannot receive long‐term OAC, rather than a direct substitute for OAC. Further long‐term and large‐scale RCTs are required to establish the best use of LAAC in AF.

Author Contributions

Ali Alsajad Hussein Al‐Janabi: validation, visualization, Screening, extraction, quality assessment, certainty of evidence, reviewing, writing. Ali Saad Al‐Shammari: screening, extraction, writing, reviewing, editing, certainty of evidence, and quality assessment. Abdullah Zeyad Hameed Al‐Tuaama: screening, extraction, reviewing. Ahmed W. Hageen: Analysis, writing, reviewing, editing. Abdullah Muataz Taha Al‐Ibraheem: screening, extraction, reviewing. Muhammad Shahzaib: extraction, analysis, writing, reviewing. Khadeeja Ali Hamzah: extraction, certainty of evidence, quality assessment, writing, editing. Mohammedsadeq A. Shweliya: analysis, writing, reviewing, editing. Yousif Hameed Kurmasha: validation, visualization, screening, extraction, reviewing, writing, editing.Diya Rathi: analysis, writing, reviewing, editing. Mohamed Wagdy: analysis, writing, reviewing, editing. Marwan M. Refaat: supervision, validation, and writing revision.

Funding

The authors have nothing to report.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.