Residual leaks following percutaneous left atrial appendage occlusion and outcomes: a meta-analysis

Abstract

Background and Aims

Residual leaks are not infrequent after left atrial appendage occlusion. However, there is still uncertainty regarding their prognostic implications. The aim of this study is to evaluate the impact of residual leaks after left atrial appendage occlusion.

Methods

A literature search was conducted until 19 February 2023. Residual leaks comprised peri-device leaks (PDLs) on transoesophageal echocardiography (TEE) or computed tomography (CT), as well as left atrial appendage patency on CT. Random-effects meta-analyses were performed to assess the clinical impact of residual leaks.

Results

Overall 48 eligible studies (44 non-randomized/observational and 4 randomized studies) including 61 666 patients with atrial fibrillation who underwent left atrial appendage occlusion were analysed. Peri-device leak by TEE was present in 26.1% of patients. Computed tomography-based left atrial appendage patency and PDL were present in 54.9% and 57.3% of patients, respectively. Transoesophageal echocardiography-based PDL (i.e. any reported PDL regardless of its size) was significantly associated with a higher risk of thromboembolism [pooled odds ratio (pOR) 2.04, 95% confidence interval (CI): 1.52–2.74], all-cause mortality (pOR 1.16, 95% CI: 1.08–1.24), and major bleeding (pOR 1.12, 95% CI: 1.03–1.22), compared with no reported PDL. A positive graded association between PDL size and risk of thromboembolism was noted across TEE cut-offs. For any PDL of >0, >1, >3, and >5 mm, the pORs for thromboembolism were 1.82 (95% CI: 1.35–2.47), 2.13 (95% CI: 1.04–4.35), 4.14 (95% CI: 2.07–8.27), and 4.44 (95% CI: 2.09–9.43), respectively, compared with either no PDL or PDL smaller than each cut-off. Neither left atrial appendage patency, nor PDL by CT was associated with thromboembolism (pOR 1.45 and 1.04, 95% CI: 0.84–2.50 and 0.52–2.07, respectively).

Conclusions

Peri-device leak detected by TEE was associated with adverse events, primarily thromboembolism. Residual leaks detected by CT were more frequent but lacked prognostic significance.

Structured Graphical Abstract

Incidence and clinical impact of residual leaks after left atrial appendage occlusion across different leak size thresholds and imaging modalities. Ao, aorta; CT, computed tomography; DAPT, dual antiplatelet therapy; LA, left atrium; LAA, left atrial appendage; LAAO, left atrial appendage occlusion; OAC, oral anticoagulation; OR, odds ratio; PA, pulmonary artery; PDL, peri-device leak; TEE, transoesophageal echocardiography.

Open in new tabDownload slide

Listen to the audio abstract of this contribution. See the editorial comment for this article ‘Residual leaks after transcatheter left atrial appendage closure: why and how to assess it?’, by O. De Backer and P. Garot, https://doi.org/10.1093/eurheartj/ehad843.

Introduction

Non-valvular atrial fibrillation (AF) is associated with a five-fold risk of cardioembolic events.¹ Transcatheter left atrial appendage occlusion (LAAO) has been established as an effective non-pharmacological alternative strategy for stroke prevention in selected patients with non-valvular AF.^2–6 Procedural success is associated with complete left atrial appendage (LAA) sealing, excluding the main source of cardiac thrombi from the circulation.⁷

Despite growing operators’ experience and device improvements, the incidence of peri-device leaks (PDLs) after LAAO remains considerable in real-world clinical practice.^8,9 This could, theoretically, undermine the therapeutic benefit,⁸ yet data on the association between PDL and outcomes are sparse and often contradictory.^10–12 This could be owing to the great uncertainty regarding the definition of clinically significant leaks, which has implications in medical management of patients after LAAO.⁹ Expert guidelines and recommendations use an arbitrary cut-off of 5 mm to classify significant vs. non-significant leaks, but cut-offs of 3 mm have also been suggested.^7,13 Different imaging modalities and device-surveillance protocols add to the puzzle.^14,15

This meta-analysis aims to address the gap in clinical knowledge regarding the clinical impact of residual leaks after LAAO across different thresholds and imaging modalities.

Methods

The current systematic review and meta-analysis was performed in accordance with a pre-specified research protocol registered a priori in the PROSPERO database (CRD42021280983: https://www.crd.york.ac.uk/prospero/display_record.php? RecordID=280983). The reporting follows the Preferred Reporting Items for Systematic reviews and Meta-Analyses 2020 reporting guidelines (see Supplementary data online, Table S1). Ethical approval was not applicable to this study.

Literature search

Literature search was conducted in MEDLINE (PubMed), EMBASE, Scopus by Elsevier, and Cochrane Central Register of Controlled Trials (CENTRAL) databases from database inception until 19 February 2023. Basic keywords used in search strings were ‘peri-device leak’, ‘residual leak’, ‘left-atrial appendage occlusion’, ‘left-atrial appendage closure’, and ‘stroke’ in both free text and Medical Subject Headings (MeSH) format (see Supplementary data online, Table S2). Finally, the reference lists of the eligible studies and relevant reviews were hand searched to identify further papers not previously detected. Corresponding study authors were also contacted via email to obtain any substantial missing information.

Outcomes of interest

The primary study outcome was the difference in thromboembolism incidence amongst LAAO patients with and without a residual leak. For the current analysis, thromboembolism was defined as the composite outcome of any ischaemic stroke, transient ischaemic attack (TIA) or systemic arterial thromboembolic episode. Any observed differences in all-cause mortality and major bleeding rates were considered as the secondary outcomes of this analysis.

Eligibility criteria

Observational (prospective or retrospective) cohort or randomized controlled studies were included in the meta-analysis if they reported the specific risk of clinical outcomes (i.e. thromboembolic episodes, major bleeding episodes, and all-cause death) by the presence or not of a residual leak amongst adult individuals with AF that underwent LAAO. The presence of a residual leak should have been confirmed with peri-procedural or follow-up imaging examination, either transoesophageal echocardiography (TEE) or computed tomography (CT). The presence of a comparison group without a leak was required.

Exclusion criteria of the meta-analysis were the following: (i) abstracts, case reports, reviews, editorials, and practice guidelines; (ii) studies not specifically reporting rates of events or absolute numbers for the outcome measure; (iii) studies lacking a control group of patients without a leak; and (iv) articles published in languages other than English.

Residual leak definition

Since the definition of a residual leak across the eligible studies was expected to be quite heterogeneous, it was assessed as follows: (i) PDL larger than 0 mm according to TEE examination; (ii) PDL larger than 1 mm according to TEE examination; (iii) PDL larger than 3 mm according to TEE examination; (iv) PDL larger than 5 mm according to TEE examination; (v) any reported PDL, as defined by the authors of each eligible study regardless of the PDL cut-off used to define it; (vi) LAA patency according to CT examination; or (vii) any PDL according to CT examination.

Regarding studies with more than one imaging examinations over the follow-up period, the examination closest to hospital discharge was selected for the main analyses. The clinical implications of leak assessment in subsequent imaging examinations were examined in meta-regression analyses.

Quality assessment

The methodological quality of the included studies was evaluated using the Quality In Prognosis Studies (QUIPS) tool.¹⁶ Three independent reviewers conducted the data extraction and quality assessment (A.B., A.O., and V.P.). Any disagreements were adjudicated by a fourth reviewer (A.S.). Accordingly, we evaluated the potential bias of each study in terms of study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding, statistical analysis, and reporting. Thereby, each study was classified as having a low, medium, or high risk of bias (see Supplementary data online, Figure S1).

Data synthesis and analysis

Rates of events have been recorded for both case (LAAO patients with a leak) and control (LAAO patients without a leak) groups. Random-effects meta-analyses were performed using Mantel–Haenszel weighting and the DerSimonian–Laird method as high heterogeneity was expected amongst the selected studies because of the different diagnostic and therapeutic procedures utilized in the included populations. For both primary and secondary outcomes of the study, pooled odds ratios (ORs) and 95% confidence intervals (CIs) are given for each analysis, with a two-sided significance level of P < .05. Evaluation of heterogeneity was conducted by calculating I². I² > 50% was considered to indicate significant heterogeneity. The possibility of publication bias was also examined by applying the Egger’s regression test for funnel plot asymmetry and using the random-effects meta-analysis model.¹⁷P < .1 was selected to indicate significant publication bias amongst included studies.¹⁸ R language version 4.0.5 via RStudio version 1.4.106 and Python IDLE Shell 3.9.5 were used for the statistical analysis and visualization of our findings.

Given the expected heterogeneity of the included studies, simple random-effects meta-regression analyses were executed to explain residual heterogeneity and reveal any difference that could be attributed to specific study/population characteristics [i.e. year of publication, mean age, mean CHA₂DS₂-VASc score, male percentage, percentage of permanent/persistent AF, hypertension, diabetes mellitus, prior thromboembolism, device-related thrombosis (DRT), LAA orifice size, oral anticoagulation (OAC) at discharge, and time to TEE exam, time to follow-up, time difference between TEE exam and follow-up, Amulet:Amplatzer Cardiac Plug (ACP) ratio, and Watchman:Amplatzer device ratio].

Results

Study selection and study characteristics

A total of 699 articles were initially retrieved. Ultimately, 48 studies satisfied the pre-defined eligibility criteria and were included in this analysis (37 studies using TEE examination, 5 studies using CT examination, and 6 others using both TEE and CT examinations). The detailed flow diagram of the study selection process is presented in Figure 1.

The patient sample size added up to a total of 61 666 AF patients undergoing LAAO. Peri-device leak by TEE was present in 26.1% of patients. Computed tomography-based LAA patency and PDL were present in 54.9% and 57.3% of patients, respectively. The cut-offs of TEE-based PDL size, as well as the period of imaging or outcome follow-up, widely varied across studies (Figure 2). Of the eligible studies, 31 of them provided evidence for a 0 mm PDL cut-off (26.5% patients had any PDL of >0 mm), 12 studies for a 1 mm PDL cut-off (15.0% patients had any PDL of >1 mm), 19 studies for a 3 mm PDL cut-off (9.6% patients had any PDL of >3 mm), and 21 studies for a 5 mm PDL cut-off (0.9% patients had any PDL of >5 mm). Mean age was 72.2 (±4.2) years, mean CHA₂DS₂-VASc score was 4.0 (±0.6), mean HAS-BLED score was 2.8 (±0.9), and mean percentage of females was 40%. The design and main characteristics of the selected studies are presented in Table 1.

Quality assessment and publication bias

Overall, no evidence of substantial publication bias was detected, by combining the asymmetry of funnel plots with the results of Egger’s weighted regression statistic, which used a P-value of <.1 to indicate significant publication bias amongst included studies (see Supplementary data online, Figures S2 and S3). In all analyses, Egger’s tests had a P-value of <.05. Moreover, the number of included studies was greater than 10 in all analyses, except bleeding (see Supplementary data online, Figures S2B).

Association of outcomes with peri-device leak presence

Transoesophageal echocardiography examination

The presence of any reported TEE-based PDL was significantly associated with almost two-fold increased odds of thromboembolism (pooled OR 2.04, 95% CI: 1.52–2.74; I² = 28%; see Supplementary data online, Figure S4), compared with the absence of PDL. Regarding the secondary study outcomes, PDL presence was also significantly associated with increased all-cause mortality (pooled OR 1.16, 95% CI: 1.08–1.24; I² = 0%; see Supplementary data online, Figure S5A) and major bleeding rates (pooled OR 1.12, 95% CI: 1.03–1.22; I² = 0%; see Supplementary data online, Figure S5B), compared with absence of TTE-based PDL.

The risk of thromboembolism had a positive graded association across cut-offs of PDL size. Specifically, as illustrated in Figure 3, the thromboembolic risk as indicated by pooled OR and 95% CI was as follows: (i) 1.82 (1.35–2.47) for patients with any PDL of >0 mm, compared with patients without PDL (Figure 3A); (ii) 2.13 (1.04–4.35) for patients with any PDL of >1 mm, compared with patients without PDL or with any PDL of <1 mm (Figure 3B); (iii) 4.14 (2.07–8.27) for patients with any PDL of >3 mm, compared with patients without PDL or with any PDL of <3 mm (Figure 3C); and (iv) 4.44 (2.09–9.43) for patients with any PDL of >5 mm, compared with patients without PDL or with any PDL of <5 mm (Figure 3D).

The heterogeneity observed for these analyses was small to moderate; however, we could not pool the odds for all-cause mortality or major bleeding episodes based on the PDL cut-offs due to the unavailability of systematically reported relevant evidence. The effect of PDL on the thromboembolic risk remained unchanged across all cut-offs of leak size, even after using only studies with at least 30 or 50 patients with PDL (see Supplementary data online, Table S3 and Figures S6–S10).

Computed tomography examination

The evaluation of the prognostic impact of PDL from CT-based studies yielded that LAA patency (nine studies) and PDL by CT (four studies) were not significantly associated with increased risk of thromboembolism. Specifically, as illustrated in Figure 4, the thromboembolic risk as indicated by pooled OR and 95% CI was as follows: (i) 1.45 (0.84–2.50) for patients with any LAA patency by CT, compared with patients without LAA patency by CT (Figure 4A); and (ii) 1.04 (0.52–2.07) for patients with any PDL by CT, compared with patients without PDL by CT (Figure 4B).

Similar results were produced when using studies with a minimum of 30 or 50 patients with LAA patency (see Supplementary data online, Table S3 and Figure S11).

Prognostic associations based on peri-device leak size categories

Patients with large leaks (PDL of >5 mm) had a significantly higher risk of thromboembolism when compared with (i) patients with no PDL (OR 4.37, 95% CI: 1.74–10.97, 10 studies) and (ii) patients with PDL of <3 mm or no PDL (OR 5.03, 95% CI: 1.09–24.78, 6 studies).

Smaller leaks were not benign and were also associated with high thromboembolic risk (Table 2). Specifically, (i) patients with PDL of 3–5 mm had a three-fold higher risk of thromboembolism compared with patients without PDL or with PDL of 0–3 mm; and (ii) patients with PDL of 0–5 mm had a significantly higher risk of thromboembolism (OR 1.38, 95% CI: 1.11–1.71, 15 studies) compared with patients without any PDL.

Meta-regression analyses

Most of the performed meta-regression analyses failed to explain a significant proportion of the heterogeneity, as described by R² (see Supplementary data online, Table S4). Studies with lower mean CHA₂DS₂-VASc score were linked with increased occurrence of thromboembolic episodes in patients with PDL of >5 mm. Studies with higher use of ACP compared with Amulet were moderately associated with a higher risk of thromboembolism in patients with any leak of >0 mm. Larger mean LAA orifice size was indirectly associated with a higher risk of thromboembolism in patients with any leak of >0 mm. Nevertheless, none of these associations explained consistently the heterogeneity across PDL categories, defined by cut-offs of PDL size (0, 3, or 5 mm).

Discussion

This is the first meta-analysis evaluating the clinical impact of residual leaks following LAAO. Amongst 61 666 patients who underwent LAAO, several interesting findings were noted: (i) any PDL of >0 mm after LAAO was present in one out of four patients by TEE and was marginally associated with worse outcomes, (ii) any PDLs of >3 and >5 mm by TEE were associated with a more than four-fold increase in thromboembolic events, (iii) smaller PDLs are not devoid of clinical risk but rather have comparable outcomes to larger PDLs, and (iv) more than half of LAAO patients had a patent LAA by CT examination, which was not associated with thromboembolic events (Structured Graphical Abstract).

Incidence and definition of leaks after left atrial appendage occlusion

The incidence of residual leaks across studies varied between 5% and 32% due to various definitions, implantation devices, and imaging protocols.⁹ Gradually increased operators’ experience reduced the incidence of leaks to <10% in more recent registries, while the presence of severe leaks (>5 mm) remains minimal.⁵³ The development of novel devices with enhanced conformability has shown promising results, even though one out of six patients still had PDL of 0–5 mm at 45-day TEE follow-up.¹¹

Computed tomography has been demonstrated to be more accurate than 2D or 3D TEE in detecting LAA leaks, especially with venous phase acquisitions.⁴³ Moreover, CT studies have used different metrics to evaluate incomplete LAAO, such as LAA patency,^14,38 PDL,³⁴ or mixed systems.¹⁵ In our meta-analysis, the incidence of leaks by CT was significantly higher than TEE.

Optimal cut-off for clinically relevant peri-device leak

The recommended 5 mm PDL cut-off to classify the procedure as technically successful and guide OAC therapy originated from the design of the PROTECT-AF trial when only data from surgical studies were available.¹⁰ This arbitrary cut-off was not supported by evidence and has drawn significant criticism over time. However, it has been widely used to steer downstream medical therapy and management following LAAO. Indeed, all three randomized trials comparing LAAO with medical therapy used a residual flow of <5 mm into the LAA as the criterion for procedural success.^2,62,63 In the recent PRAGUE-17 trial, a leak <5 mm was considered enough to de-escalate treatment from dual to single antiplatelet therapy.² Left atrial appendage occlusion consensuses have supported over time a threshold of 5 mm for clinically relevant leaks.^7,13

The 3 mm threshold for relevant PDL originated from the PLAATO trial and has not been adopted to the extent of its 5 mm counterpart.⁶⁴ Early registries with Amplatzer devices (mainly using ACP) have defined complete LAAO as the absence of any PDL of >3 mm,⁶ while others have classified a PDL of 3–5 mm at follow-up TEE as mild.¹² A report from experienced LAAO centres noted that half of their patients with any PDL of >3 mm remained on OAC therapy during follow-up.²⁶

Recent studies reported that PDLs, even those ≤3 mm, progressed to a more severe grade in 8%–14% of patients, while PDLs of >3 mm detected at 2- to 3-month follow-up were not likely to regress completely.^26,65 It has been recently reported that newly identified PDLs of ≥3 mm are not infrequent after LAAO, while PDLs of ≥3 mm persist even 12 months post-LAAO, especially in single occlusive mechanism devices.¹⁹ Thus, frequent device-surveillance imaging is of paramount importance until leak regression is confirmed.^26,65

Clinical impact of peri-device leak

Theoretically, a residual leak after LAAO may enhance platelet adhesion and thrombus formation, permitting the embolization of LAA thrombus to systemic circulation.¹² Indeed, incomplete LAAO after surgical procedures and epicardial approaches have both been associated with increased risk for thromboembolic events.^66–68 However, there is inconclusive evidence that incomplete LAAO after percutaneous closure with endovascular devices is associated with thromboembolic events.⁹

Previous studies on transcatheter LAAO have failed to find any significant association between large PDL and clinical events.^3,12,63 In the largest available LAAO database from the National Cardiovascular Data Registry (NCDR) LAAO Registry, large leaks (>5 mm) were not associated with adverse events.²² Of note, even though studies have not shown an increased incidence of TIA or stroke in patients with PDL, the need for continued anticoagulation was noted to be higher.^12,22,38

However, most studies deserve more scrutiny due to inherent problems. These include limited follow-up duration (<1 year), not consecutive imaging follow-up for leak assessment, small numbers of ischaemic events, and various antithrombotic regimens after LAAO (anticoagulants, antiplatelets, or a combination of both). Thus, these studies appeared to be underpowered to answer this issue and could not draw definitive conclusions to discard an association between PDL and risks of thromboembolic events.

Studies did not report consistently granulated data regarding the type or the severity of thromboembolic events. Out of the 48 studies, 10 of them reported a composite thromboembolic event (stroke/TIA/systemic embolism), while only four studies classified stroke as major or minor/non-disabling. The ratio of ischaemic stroke:TIA in studies with classified data was almost 2:1 (see Supplementary data online, Table S5). Further dedicated studies are warranted to examine the potential association between the size of the residual leak and the severity of thromboembolism.

Prognostic implications of small peri-device leak

Expert recommendations considered small leaks, especially those ≤3 mm, as not clinically relevant, suggesting switching from OAC therapy to antiplatelet agents upon their detection.^7,13 However, recent data from the NCDR LAAO registry reported that small leaks (0–5 mm) at short-term imaging follow-up were associated with a modest increase in thromboembolic events compared with no leak.²² Moreover, a recent study that combined data from three large trials reported that PDL of ≤5 mm after LAAO with the Watchman device was associated with an increased risk for ischaemic stroke or systemic embolism at long-term follow-up.²¹

This association is intriguing and could be explained by three hypotheses: (i) thrombus formation in small leaks is situated at the leak site (between the device and the LAA wall), where turbulent blood flow is prothrombotic,⁶⁹ or (ii) patients with small leaks are generally not maintained on prolonged courses of OAC therapy as per guidelines,⁷ or (iii) gradual progression of small leaks to a more severe grade.^26,65 Interestingly, it has been reported that more than half of clots retrieved from patients with cardioembolic ischaemic stroke are <5 mm in size.⁷⁰

In theory, smaller leaks (0–5 mm) should draw the attention over large leaks, given that the average diameter of the middle cerebral artery is 3 mm.⁷¹ Our study reported that PDL of 3–5 mm was associated with a more than three-fold risk of thromboembolism compared with no PDL or PDL of 0–3 mm. Moreover, a substantial thromboembolic risk persisted even at PDL cut-offs of 0 and 1 mm, indicating that smaller leaks might not be benign.

The prior perspective around small PDL was focused less on its direct and more on its indirect impact on thromboembolism through a potential association with DRT, which is undoubtedly associated with unfavourable outcomes.⁷² However, data regarding the incidence of DRT in patients with PDL are contradictory.^12,23,73 In our meta-analysis, there was no association between higher thromboembolic risk by increasing percentages of DRT across different PDL cut-offs. This implies that a PDL might have an independent association with higher thromboembolic risk, beyond a co-existing DRT. This was also consistent with a recent large study that reported no relationship between a PDL of <5 mm and DRT, suggesting that the mechanism of stroke when PDL is present is unrelated to DRT.²¹ In any case, not detecting a DRT following a thromboembolic event should not exclude the possibility that the source of an embolic stroke was a pre-existing DRT that escaped into the systemic circulation. Hence, early identification of predictors of DRT, such as lower left ventricular ejection fraction, deep implantation, hypercoagulable state, and large LAA diameter, as well as subsequent proper management is of paramount importance.^23,73

CHA₂DS₂-VASc score and thromboembolic risk in patients with residual leak

An interesting finding of our meta-analysis was that studies with lower mean CHA₂DS₂-VASc score were associated with increased occurrence of thromboembolic events, only in patients with a PDL of >5 mm. It is evident that the CHA₂DS₂-VASc score does not guide therapy in post-LAAO patients with large leaks, since this group of patients continue OAC until PDL regression, irrespective of their CHA₂DS₂-VASc score.⁷⁴ Hence, it could be speculated that the thromboembolic risk in patients with large PDLs (>5 mm) is not only beyond the CHA₂DS₂-VASc score but is also probably defined by the anatomical relationship between the patent LAA and the malpositioned device following LAAO.¹⁵ Targeting the predictors of PDL, such as low device compression rate or angled device lobe without co-axial alignment to the LAA wall, could also indirectly minimize the risk of thromboembolism following LAAO.¹⁵ On the other hand, there is a possibility that a prolonged OAC therapy in some patients with large PDLs cannot be feasible due to bleeding complications.²² Therefore, PDL closure might be the optimal solution in such cases.⁷⁴

Left atrial appendage occlusion devices and impact of peri-device leak

Peri-device leak sizing is challenging in disc-and-lobe devices, such as Amplatzer LAA occluders, due to the presence of two areas for potential residual leak. The very recent SWISS-APERO randomized clinical trial that used multimodality imaging for leak assessment reported that rates of any PDL were two-fold higher with Watchman than with Amulet device.⁷⁵ However, data on leaks of >5 mm were not available.⁷⁵

The clinical impact of PDL between different LAAO devices has not been analysed yet. The presence of two potential barriers confers a theoretical advantage, making it more difficult for a thrombus to escape the patent LAA. Our meta-analysis did not identify any difference in the prognostic significance of PDL between Watchman and Amplatzer devices.

Medical management of peri-device leak

Peri-device leaks of >5 mm have been generally considered as clinically relevant and managed with prolonged OAC and serial TEE follow-up.⁹ Alkhouli et al.²² reported that patients with large (>5 mm) leaks were less likely to discontinue OAC treatment at follow-up, while only one out of five patients with small (>0–5 mm) leaks remained on OAC at 6 months and 1 year. In studies with a patent LAA after LAAO, OAC therapy was not interrupted until complete LAA sealing, while in others until regression of PDL below the cut-off of 3 or 5 mm. However, the included studies in this meta-analysis reported only aggregate data; therefore, an analysis of the therapy in specific patients with gradations of PDL could not be conducted. Alternatively, the mean percentages of antithrombotic regimens based on the therapeutic protocols in each study at different follow-up periods were reported (see Supplementary data online, Table S6).

Expert recommendations suggest discontinuation of OAC treatment and switching to antiplatelet agents upon follow-up imaging that confirms regression of a previous PDL below the threshold of 5 mm.⁷ However, a recent study recommended readjusting the threshold at 3 mm of PDL size and still repeat imaging at 6–12 months.²⁶

The role and clinical significance of computed tomography imaging on left atrial appendage occlusion

Cardiac CT provides better understanding of the LAA anatomy and surrounding structures, which allows optimal device selection and sizing, reducing the risk of post-LAAO leaks.⁷⁶ Given its greater spatial resolution, the capabilities and definitions of CT have been developed primarily to maximize the sensitivity over TEE to detect a patent LAA.³⁴ Thus, it has been reported that the incidence of post-LAAO leak detected by CT could be as high as 60%.^15,42,43 However, the detection of contrast patency by CT is highly dependent on the phase and timing of the high-pitch single-heartbeat CT acquisition protocol, which creates discrepancies between early arterial and late venous phase acquisitions.⁴³ Despite its high sensitivity for detecting leaks, the specificity of CT is low for the degree of PDL and the detection of leaks of larger leaks, which are more likely to carry clinical implications.^9,14,38

Our meta-analysis reported that the incidence of PDL or LAA patency by CT was more than two-fold compared with TEE. Whether this disparity is because TEE is less sensitive in detecting leaks or because cardiac CT is ‘overdiagnosing’ residual LAA patency remains to be investigated.^15,77 Left atrial appendage patency by CT may correspond to intra-fabric leaks or delayed device endothelialization, which is not captured in TEE analysis.^14,34

Another interesting finding of our meta-analysis is that LAA patency and PDL detected by CT were not associated with thromboembolic events. The inadequate assessment and grading of a ‘true’ residual leak on post-LAAO CT scans appear to undermine the utility of cardiac CT as a valuable prognostic imaging tool for device surveillance following LAAO.⁷⁴ Left atrial appendage patency is most likely very simplistic as a grading method, while a novel PDL grading method merging LAA patency with peri-device gap could not stratify the risk of worse clinical outcome.¹⁵ Different mechanism-based classifications of leaks have been analysed by CT without reaching an association with adverse events.³⁰ Further research is warranted to develop a CT-based PDL grading system that provides accurate thromboembolic risk stratification following LAAO.

Implications of our results

To minimize the risk of future thromboembolic events, efforts should be steered towards achieving complete occlusion of the LAA, starting from an optimal pre-procedural planning. Even though TEE is the current gold standard, CT could provide improved imaging with high-quality multi-planar and 3D reconstruction that may enable better optimal device sizing and lower rates of PDL.⁷⁸ Computed tomography-based computational modelling could provide an added value when planning for transcatheter LAAO by improving procedural efficiency and minimizing LAA patency, as reported recently in the PREDICT-LAA randomized controlled trial.⁷⁹

During the procedure, a higher emphasis should be given to co-axial placement across multiple TEE views from 0° to 135°. Larger device sizing to avoid undercompression is recommended, given the observed lower rates of PDL after deliberate oversizing of the Watchman device by 20% or more.⁸⁰ Dedicated software packages could enable 3D printing, access route planning, and overlay/fusion imaging peri-procedurally.^74,81 These could optimize the transseptal location for optimal LAA access to obtain co-axial alignment between the delivery sheath and the LAA central axis.⁸¹ Moreover, identifying the optimal fluoroscopic angulation from pre-procedural CT planning to engage the LAA and then using the same C-arm angulation peri-procedurally could minimize the risk of PDL before final device deployment.⁷

Complex or prohibitive LAA anatomies often result in either failed implantation or pre-procedural screen. The NCDR LAAO Registry of 38 158 patients reported that 7% of the cases were cancelled or aborted, equating to successful closure at 91% of the intended at-risk population.⁸² Different device types may be more suitable for specific LAA anatomies, which may tailor the device selection for each patient.¹¹ This could guide clinical care towards the availability of more than one device types in the Cath Lab to allow the possibility of selecting the appropriate one for LAAO and, thus, increasing procedural success.

The optimal timing for post-LAAO TEE exam is also a matter of uncertainty. Dukkipati et al.²¹ reported that leaks of any size detected by 1-year TEE imaging were associated with higher rates of ischaemic events between 1 and 5 years of follow-up. Interestingly, outcomes were similar when including only events during the first year after LAAO, which implies that TEE follow-up should not be limited to 1 year.²¹ In our meta-analysis, the timing of TEE after LAAO was not significantly associated with the prognostic value of leak management (see Supplementary data online, Table S3), implying that meticulous evaluation of any leak after LAAO is warranted.

Post-procedural management should focus on prolonged use of OAC or dual antiplatelet therapy until complete occlusion is achieved or, at least, signs of PDL regression from consecutive follow-up imaging sessions are present. However, there is no evidence to support the continuation of an intense antithrombotic therapy indefinitely; therefore, such a decision should be carefully balanced to the major clinical implications that could occur in an LAAO population with an inherent high bleeding risk. Interventional PDL closure (e.g. coiling and plugging) should be considered to prevent thromboembolism in cases where PDL progresses to a more severe grade after consecutive over-time imaging follow-up. However, data on the interventional management of PDL are still limited.⁸ Nevertheless, maintaining patients with high bleeding risk on blood thinners or re-operating on them to treat a post-LAAO leak could be too cumbersome for a preventive procedure. Thus, targeting for a ‘true’ complete closure with the absence of any visible leak should be the main priority of interventionalists.

Limitations

The key methodological issue of our study is the exploitation of unadjusted risk estimates derived from observational studies, which could pose selection and performance bias.⁸³ We tried to partially safeguard against this problem by performing extensive meta-regression analyses, even though multivariable adjustment of pooled event rates could not be performed because the majority of studies did not classify potential confounders between subgroups of patients with or without leak presence. Hence, careful interpretation of our results is suggested when translating these findings into clinical decisions, given that the observational nature of the data provides associations without proven causation. Moreover, several factors, such as ‘chicken-wing’ LAA anatomy, indication for LAAO, or LA size, that could influence the thromboembolic risk could not be included in meta-regression analyses due to inconsistent reports of relevant information across studies. Another limitation is that the measurement of PDL is self-reported and operator dependent; therefore, it is subject to variability in assessment techniques across different study settings and image acquisition protocols. Different operators’ experience and preferences could also influence the incidence of leaks and indirectly the thromboembolic risk; this could pose a potential bias in the interpretation of the results. Finally, there is a possibility that some patients underwent subsequent interventional leak closure, which could interfere with the results. However, such data were not systematically reported in selected studies.

Conclusions

Residual leak after LAA occlusion remains an unresolved issue of LAAO devices. This large meta-analysis sheds light on the open question of whether incomplete LAA closure has any prognostic implications. Transoesophageal echocardiography-detected PDLs after LAAO were associated with adverse clinical events, primarily thromboembolism. Leaks were detected more frequently by CT (than TEE) but lacked prognostic significance. More research is warranted to determine the optimal detection method for PDL, as well as the appropriate cut-off for clinically relevant leak across different surveillance protocols and imaging modalities.

Supplementary data

Supplementary data are available at European Heart Journal online.

Declarations

Disclosure of Interest

J.S. is on the advisory board and a consultant for Abbott, Boston Scientific, and Baylis; M.A.A. is on the advisory board for Boston Scientific and Abbott; U.L. has received lecture and consultancy honorary from Abbott and Boston Scientific; A.T. is a clinical proctor for Abbott.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Funding

All authors declare no funding for this contribution.

Ethical Approval

Ethical Approval was not required.

Pre-registered Clinical Trial Number

None supplied.

References