What is the evidence for LAA closure?

There have been two prospective randomised non-inferiority trials (RCTs) on LAA exclusion.

Watchman Left Atrial Appendage System for Embolic Protection in Patients with AF (PROTECT AF) randomised patients in a 2:1 ratio to receive either Watchman device or warfarin (table 1). At 2 years follow-up, trial showed non-inferiority of LAA closure to warfarin. After 4 years of observation, LAA exclusion met prespecified criteria for both non-inferiority and superiority compared with warfarin.2 Despite these optimistic efficacy findings, the rate of periprocedural complications was high at 8.7% (table 1). Importantly, though LAA occlusion was superior to warfarin in the prespecified primary efficacy endpoint (composite of stroke, systemic embolism and cardiovascular/unexplained death), it did not meet the non-inferiority criteria for ischaemic stroke risk reduction (table 1). Indeed, there were less ischaemic strokes in warfarin group and superiority of LAA closure over warfarin was predominantly driven by lower incidence of haemorrhagic strokes in the device arm. A closer inspection of mortality reasons also shows that only death rates resulting from haemorrhagic stroke were significantly different between study groups (0.4% with Watchman vs 3.3% on warfarin, p=0.004). Of 10 subjects who experienced haemorrhagic stroke in warfarin group, 8 patients died, further contributing to a significant difference in cardiovascular and all-cause mortality.

Prospective Randomised Evaluation of the WATCHMAN LAA Closure Device in Patients with Atrial Fibrillation vs Long Term Warfarin Therapy (PREVAIL) enrolled an older population at significantly higher risk of stroke compared with PROTECT AF (table 1). Compared with PROTECT-AF, the incidence of procedure-related complications was lower, at 4.2%. This trial did not meet the prespecified non-inferiority primary efficacy criteria because of the lower number of ischaemic strokes in warfarin arm (table 1).3 However, there was also a second efficacy endpoint comprising of stroke or systemic embolism >7 days after randomisation, and non-inferiority of this endpoint was achieved.3

Of note, though the differential of periprocedural events to allow comparison of late-ischaemic endpoints may be appreciated, a patient has to consider all risks of the procedure overall, notwithstanding complication rates also decrease with operator experience.

One meta-analysis of two RCTs and their respective registry continuations comprising 2406 patients with mean follow-up of 2.69 years showed a similar incidence of all-cause stroke and systemic embolism, all-cause death and major bleeding events among both strategies.4 However, the type of stroke was different in both groups. Indeed, there were more ischaemic strokes in the device group (1.6 vs 0.9; HR 1.95; p=0.05), but significantly less haemorrhagic strokes (0.15 vs 0.96 events/100 patient-years; HR 0.22; p=0.004). The lower rate of haemorrhagic strokes in Watchman arm largely contributed to significantly fewer cardiovascular/unexplained deaths (1.1 vs 2.3 events/100 patient-years; HR 0.48; p=0.006) compared with warfarin.

LAA occlusion or oral anticoagulation?

The concept of stroke prevention by LAA exclusion was developed primarily to reduce the incidence of thromboembolism coming from LAA and thus mitigate the risk of ischaemic stroke. The data discussed above suggest that LAA closure may not be as good as systemic anticoagulation especially in AF patients with higher stroke risk profile (PREVAIL).

Perhaps this is unsurprising, given that AF is a systemic disease and thrombogenesis in AF is also multifactorial, including all aspects of Virchow's triad, that is, endothelial or endocardial damage/dysfunction, abnormal blood stasis (not limited to LAA only) as well as abnormal haemostasis, platelet function and fibrinolysis.5 LAA occlusion will also not diminish the risk of stroke arising from other vascular sources such as carotid disease or complex aortic plague. Indeed, one of the strongest independent predictors of ischaemic stroke is the presence of complex aortic plaque on the descending aorta, as imaged using transoesophageal echocardiography. By contrast, systemic anticoagulation mitigates the hypercoagulable state that is increased in many conditions, and thus, the overall mortality reduction is greater would be assumed by decrease of stroke-related deaths only.

However, trial data do indicate that though exclusion of LAA occlusion may not be as effective as warfarin in ischaemic stroke prevention, it causes significantly fewer intracranial haemorrhages. The risks of bleeding on warfarin are closely related to the quality of anticoagulation control, as reflected by the time in therapeutic range (TTR; where we should be aiming for a TTR>60%). In trials using NOACs, there was clear superiority of NOACs over warfarin in the reduction of stroke and all-cause mortality especially from a reduction in haemorrhagic strokes (rate ratio 0.48; 95% CI 0.39 to 0.59).6 Also, if patients need to be on antiplatelet therapy postprocedure, the data would suggest that NOACs and well-managed warfarin (TTR>70%) would have similar risks of major bleeding to aspirin, especially in the elderly.

Future directions

Despite clear advantages of LAA exclusion and growing enthusiasm in its clinical application, several issues have to be clarified. Meta-analyses show low-strength evidence that LAA closure with Watchman device is associated with similar risk of stroke and mortality compared with warfarin.7 Of 4998 screened patients in the PROTECT AF trial, only 707 (14.1%) were finally randomised, which raises question about method utility and findings generalisability. The RCTs used Watchman device only, and thus findings from these trials should not be easily extrapolated on other devices. There are no data to compare implantation success rate, successful LAA exclusion achieved or outcome benefit among different devices. The RCTs enrolled only patients who were eligible for long-term warfarin therapy and thus there is no evidence to recommend LAA closure in patients at high risk for stroke and contraindications to anticoagulation; indeed, current European AF guidelines on this point are based on reports from observational studies and expert consensus only.

In both RCTs, warfarin was used in the device arm for at least 45 days post implantation, followed by dual antiplatelet therapy (aspirin+clopidogrel) till month 6 and then by aspirin in a recommended dose of 325 mg indefinitely. Such chronic use of antiplatelet therapy following LAA closure creates some bias of the actual impact of the implanted device on ischaemic stroke risk reduction compared with warfarin. Also, such treatment protocol substantially limits method applicability in patients at high risk of bleeding.

Although there are numerous reports coming from non-randomised observational studies, these data do not provide decisive evidence on efficacy and safety of LAA closure, particularly when there is no control group and outcomes are compared with event rates projected from completely different (and often historical) patient populations. In addition, serious periprocedural adverse events were reported in 1.6–13.6% of patients.7 Finally, given the advantages of NOACs over warfarin, particularly the substantially lower risk of intracranial bleeds, RCTs comparing efficacy and safety of LAA occlusion versus NOACs are highly warranted.

In the context of limited evidence, the paper presented in this issue by Sahay et al8 is important. The authors performed a network meta-analysis of 19 RCTs comprising in total 87 831 patients with AF, to compare LAA closure versus placebo, antiplatelet therapy and NOACs. By indirect efficacy comparisons, Sahay et al found LAA closure to be superior to antiplatelet or no therapy, and (importantly) similar to NOACs, while rates of major bleeding were similar among all groups. This paper further suggests that LAA closure may be a reasonable treatment option for patients with contraindications to long-term anticoagulation with warfarin. While a comprehensive and well-performed network meta-analysis, this methodology is no substitute for properly conducted RCTs when testing the efficacy and safety of an intervention. Furthermore, clinical trials would need to be supported by large prospective ‘real-world’ registries to show effectiveness and safety when the intervention is used in everyday clinical practice outwith the confines of a selected population in clinical trials.

For now, oral anticoagulation (whether with an NOAC or warfarin with TTR>70%) remains the standard of care therapy for stroke prevention in non-valvular AF. LAA occlusion could possibly be an option, where patients really cannot take any antithrombotic drugs, notwithstanding the fact that some thromboprophylaxis periprocedure and postprocedure for LAA occlusion is often needed. The question remains that if the patient can tolerate aspirin or OAC during this time, some compelling reasons may be needed to justify why they cannot be used longer.