Cerebral Embolic Protection Devices During Transcatheter Aortic Valve Replacement: A Meta-analysis of Randomized Controlled Trials

Background Stroke is a feared complication of transcatheter aortic valve replacement (TAVR), which embolic protection devices (EPDs) may mitigate. This systematic review and meta-analysis synthesized randomized controlled trials (RCTs) to evaluate the effect of EPDs in TAVR. Methods All RCTs comparing EPDs with control during TAVR were systematically identified. Prespecified primary end points were all stroke, disabling stroke, nondisabling stroke, and all-cause mortality. Safety and neuroimaging parameters were assessed. Sensitivity analyses were stratified by EPD type. Study registration was a priori (CRD42022377939). Results Eight trials randomizing 4043 patients were included. There was no significant difference between EPDs and control for all stroke (relative risk [RR], 0.88; 95% CI, 0.65-1.18; P = .39; I2 = 0%), disabling stroke (RR, 0.67; 95% CI, 0.31-1.46; P = .32; I2 = 8.6%), nondisabling stroke (RR, 0.99; 95% CI, 0.71-1.40; P = .97; I2 = 0%), or all-cause mortality (RR, 0.87; 95% CI, 0.43-1.78; P = .71; I2 = 2.3%). There were no differences in safety end points of bleeding, vascular complications, or acute kidney injury. EPDs did not result in differences in total lesion volume or the number of new lesions. The Sentinel EPD significantly reduced the risk of disabling stroke (RR, 0.42; 95% CI, 0.20-0.88; P = .022; I2 = 0%) but did not affect all stroke, nondisabling stroke, or all-cause mortality. Conclusions The totality of randomized data for EPDs during TAVR demonstrated no safety concerns or significant differences in clinical or neuroimaging end points. Analyses restricted to the Sentinel EPD demonstrated large, clinically meaningful reductions in disabling stroke. Ongoing RCTs may help validate these results.


Introduction
][4][5][6][7] Therefore, reducing procedural complications remains paramount given the extension of TAVR to low-risk patient populations.Encouragingly, contemporary registry studies of TAVR show reductions in mortality and length of hospital stay over time. 8troke is a feared complication of TAVR, conferring disabling morbidity and mortality.Patients consider stroke a worse outcome than death. 9Despite improvements in device technology and refinements in procedural technique, stroke rates remain largely inert since the advent of TAVR at %2%. 8 Embolization likely represents the dominant mechanism whereby particulate valvular or vascular wall tissue and atherosclerotic plaque may dislodge after instrumentation of the aorta and subsequent valve deployment.Cerebral embolic protection devices (EPDs) were developed on the predicate that minimizing procedural embolization by capturing or deflecting debris from carotid and vertebral arteries would reduce stroke during TAVR.
We previously synthesized RCTs investigating cerebral EPDs as a method of mechanical stroke prophylaxis illustrating no benefit of EPDs on clinical outcomes or neuroimaging parameters. 10Three new RCTs have since been published, [11][12][13] with PROTECTED-TAVR 13 recently reporting on 3000 randomized patients.Given these new data, we performed an updated systematic review and meta-analysis to investigate the effect of EPDs on clinical efficacy, safety, and neuroimaging parameters.

Methods
This systematic review and meta-analysis was conducted according to PRISMA guidance. 14The study was registered a priori on the PROSPERO database (CRD42022377939).

Search strategy
We performed a systematic search of the MEDLINE and Embase databases from inception through October 24, 2022, for RCTs comparing EPDs with control during TAVR.No filters or limits were applied.Search strings included ("transcatheter aortic valve implantation" OR "transcatheter aortic valve replacement") AND ("embolic protection" OR "cerebral protection").Hand search of included studies was used.All authors independently participated in searching and screening with disputes resolved by consensus.

Inclusion and exclusion criteria
All randomized trials comparing EPD with controls were considered eligible for inclusion.

End points
The primary end point was risk of all stroke.Secondary clinical end points included risk of death, disabling stroke, nondisabling stroke, all bleeding, life-threatening or disabling bleeding, vascular complications, and acute kidney injury.
Neuroimaging end points included total lesion volume (TLV) and number of ischemic magnetic resonance imaging (MRI) lesions.Outcomes with the longest follow-up were used.Definitions for individual end points provided by the included studies are summarized in Supplemental Table S1.Further details on diffusion-weighted MRI protocols used in the included studies and usage of imaging core laboratories are summarized in Supplemental Table S2.

Data extraction and quality assessment
Three authors (R.K.R., Y.A., J.P.H.) independently participated in data extraction with data abstracted onto study-specific collection forms.Relevant study characteristics included first author, study acronyms, year of publication, study regions, number of patients, mean age, follow-up durations, inclusion and exclusion criteria, device type, TAVR type, and primary outcomes.Disputes were resolved by consensus discussion.The Cochrane risk-of-bias tool 15 was applied to each included trial.

Statistical analysis
Tests for publication bias were planned in the event of 10 or more studies being eligible for inclusion.For binary end points, we extracted event rates and pooled relative risks (RR).For primary analyses on neuroimaging end points, TLVs were extracted when studies reported values based on whole-brain MRI rather than those on protected-brain MRI because this was deemed more clinically relevant.For inclusion in sensitivity analysis, TLVs were extracted when studies additionally reported values based on protected-brain MRI.TLVs are presented as mean difference (MD) AE SD.If individual studies reported medians with corresponding interquartile range or 95% CI, these values were converted to mean AE SD using the methods described by Luo et al 16 and Wan et al 17 to facilitate quantitative synthesis.
Random-effects meta-analyses were performed using the restricted maximum likelihood estimator.All analyses were performed based on the intention-to-treat principle.The I 2 statistic was used to assess heterogeneity.
Given that different EPDs are available, sensitivity analyses were performed to synthesize trials studying the same device to consider clinical heterogeneity caused by the device type.All analyses were performed within the statistical programming environment R using the metafor package.P values of <.05 were considered statistically significant.

Results
Eight RCTs [11][12][13][18][19][20][21][22] enrolling 4043 patients met the inclusion criteria, with 2175 randomly assigned to EPDs and 1868 to control. Th following EPDs were studied: the Sentinel (Boston Scientific), the Montage (Claret Medical), the TriGUARD (Keystone Heart) and the EMBOL-X (Edwards Lifesciences).A flow chart representing the study selection process is displayed in Figure 1.Full study characteristics are displayed in Table 1.Because only 8 studies were included, tests for publication bias were not performed, as per Cochrane recommendations.23

Risk-of-bias assessment
Overall study quality after risk-of-bias assessment was high for 6 studies [11][12][13]19,20,22 and moderate for 2. 18,21 The complete results of risk-of-bias assessment are displayed in Table 2.

Discussion
This study synthesized all available randomized data for EPDs in patients who underwent TAVR, including 4043 patients (Central Illustration).When all trials were pooled, there were no differences in the rates of stroke, mortality, safety, or neuroimaging end points with EPDs compared with those of controls.In sensitivity analyses focusing solely on trials evaluating the Sentinel EPD (which is the device used in clinical practice today), a significant reduction in disabling stroke of ~60% was seen.There were no signals of harm detected with EPD, confirming a favorable safety profile.
Stroke rates in the decade since the inception of TAVR have remained >2% despite advancements in understanding, operative technique, and technology. 8Approximately half of these occur immediately or a few hours after TAVR 23 and are mostly attributed to intraprocedural embolism.In addition to clinically evident strokes, TAVR causes "silent" brain lesions, 27,28 relevant for their potential association with subtle functional and neurocognitive deficits that may go unnoticed.Furthermore, silent infarcts more than double the risk of subsequent stroke and dementia. 29Therefore, there has been interest in the potential role of EPDs.
This study demonstrated that when all RCTs investigating EPDs were pooled, there were no differences in stroke, whether disabling or nondisabling, and all-cause mortality.Similarly, there were no safety concerns identified with EPD.These findings have been reported in previous, smaller meta-analyses of solely RCTs. 10,30,31This may be due to low event rates across the trials for the relatively rare end points of death and stroke, which the meta-analysis technique helps to overcome through quantitative synthesis.However, this meta-analysis still likely remains underpowered for these low-frequency clinical events.Moreover, differences in intraprocedural heparinization use may have affected trial results, salient because EPD implantation results in longer procedural duration, greater device manipulation within the aortic arch, and increased thrombogenic surfaces.Thus, particular care is required surrounding intraprocedural anticoagulation.If not accounted for, the additional upfront embolic risk presented by EPDs may paradoxically increase stroke rates, precluding any benefit and biasing toward the null.Similarly, differing postprocedural anticoagulation and antiplatelet strategies or adherence may have resulted in a bias across trials.There were also no differences in whole-brain neuroimaging end points, with high heterogeneity demonstrated in these analyses.Reasons for this may include differences in measuring or reporting ischemic lesions and TLV, the timing of acquisition and imaging systems used.However, there was a reduction in protected-brain TLV, which is plausible given that EPDs can only prevent embolic   warrants attention.A recent meta-analysis demonstrated that early stroke risk is lower in low-risk TAVR vs surgical aortic valve replacement, although this difference was not significant 1-year postprocedure and is still nonzero. 32The finding that the Sentinel device did not reduce all-cause mortality may be considered surprising given that stroke incidences after TAVR are independently associated with markedly worse 30-day cardiovascular and all-cause mortality. 33owever, this analysis was underpowered for mortality.Furthermore, nonfatal but disabling stroke may lead to a longer-term increase in mortality, which is not captured in the randomized trials with short-term follow-up.Similarly, the risk of nondisabling stroke was not reduced, which is biologically plausible given the Sentinel's mesh pore size of 140 μm may not catch all embolized particulate matter smaller than this.In addition, the Sentinel device protects the brachiocephalic trunk and left common carotid artery but not the left subclavian artery, representing incomplete cerebral coverage.
There are a few reasons why the analysis focusing on the Sentinel device may have demonstrated significantly reduced disabling strokes compared with the overall pooled result.Practically, the Sentinel requires a smaller 6F delivery sheath compared with the 9F TriGuard and 17F EMBOL-X filter sheaths, presumably resulting in less trauma to both vessel and valve during aortic arch navigation.Mechanistically, the Sentinel is distinct in that it captures particulate debris and facilitates retrieval rather than relying on deflection as with other EPDs.Most importantly, the Sentinel trial is the beststudied device owing to the PROTECTED-TAVR 13 trial with 3000 patients.The trials of other devices were likely too small to suggest any effect on clinical events.
Unlike previous smaller meta-analyses, we investigated individual clinical end points rather than pooling data and generating composites, negating the risk of counting events twice when trials provide time-to-event data.Statistical heterogeneity was mostly low.We considered clinical heterogeneity posed by differing device types by performing sensitivity analyses including only trials evaluating the Sentinel device, which contributed ~90% of patients to the overall pooled analysis.
This study-level meta-analysis is naturally subject to the inherent limitations of the constituent RCTs, which include but are not limited to low clinical event rates, treatment crossover, and loss-to-follow-up, particularly for neurologic assessment and imaging end points.Owing to the use of aggregate data, hypothesis-generating subgroup analyses based on prerandomization demographic characteristics were inappropriate.An individual participant data meta-analysis is planned to combine PROTECTED-TAVR and the ongoing British Heart Foundation-funded PROTECT-TAVI (ISRCTN16665769).PROTECT-TAVI aims to enroll nearly 8000 randomized patients by 2026 and will define further the role of the Sentinel EPD.Subgroup analyses may identify higher-risk patients deriving benefit from EPDs, such as those with heavily calcified aortic anatomy or bicuspid aortic valves.Furthermore, in observational analyses, independent baseline predictors of periprocedural stroke include age older than 85 years, body mass index of <25 kg/m 2 , and a history of coronary artery disease, whereas previous cerebrovascular events and an estimated glomerular filtration rate of <30 mL/min/1.73m 2 independently associate with 30-day stroke risk. 34hese relevant subgroups should be further interrogated.Finally, future EPD iterations might improve efficacy by permitting complete cerebral coverage.Until then, this study represents the best available evidence within the field.
In conclusion, in this analysis of all 8 published RCTs evaluating all EPDs in TAVR, there was no evidence for differences in rates of stroke, mortality, or safety outcomes compared with controls.Importantly, in analyses limited to the Sentinel trials, ~60% reduction in disabling stroke was demonstrated.However, equipoise remains, and efforts should focus on enrolment into ongoing RCTs alongside development of newer devices enabling complete cerebral coverage.

Declaration of competing interest
Yousif Ahmad is a consultant for Shockwave Medical and Cardiovascular Systems Inc and serves on the medical advisory board for Boston Scientific.All other authors have no relevant disclosures.

Figure 1 .
Figure 1.Flowchart detailing the study selection process.

Figure 3 .
Figure 3. Clinical safety of cerebral embolic protection devices during transcatheter aortic valve replacement.(A) Life-threatening or disabling bleeding; (B) all bleeding; (C) vascular complications; (D) acute kidney injury.

Figure 4 .
Figure 4. Neuroimaging parameters of cerebral embolic protection devices during transcatheter aortic valve replacement.(A) Total cerebral lesion volume in the whole brain; (B) difference in the number of ischemic lesions.

Figure 5 .
Figure 5. Sensitivity analysis of clinical efficacy of the Sentinel cerebral embolic protection device during transcatheter aortic valve replacement.(A) All stroke; (B) disabling stroke; (C) nondisabling stroke; (D) all-cause mortality.

Table 1 .
Characteristics of the included studies.Mean age AE SD given for overall population if provided; otherwise given for each group.If mean age not available, then median with IQR given.
a R.K. Reddy et al. / Journal of the Society for Cardiovascular Angiography & Interventions 2 (2023) 101031