Reflective Learning on the Role of Cerebral Embolic Protection in TAVI Patients?

Rajesh K. Kharbanda


Eur Heart J. 2021;42(27):2680-2682. 

Graphical Abstract: Figure showing the relationship between the development of an intervention over time and the increasing strength of evidence to guide clinical practice.

The AllTrials campaign has emphasised the central importance of reporting findings from trials that are negative, neutral or terminated early, as well as those with positive findings.[1] The fact that findings from some trials are 'lost' due to positive publication bias is well recognised.[2] In this issue, Lansky et al report findings from the REFLECT I trial: a Randomized Evaluation oF the TriGuard HDH Cerebral Embolic Protection Device to Reduce the Impact of Cerebral Embolic LEsions after TransCatheter Aortic Valve ImplanTation. The publication of this trial highlights important lessons that will improve future studies.[3]

REFLECT I compared cerebral embolic protection (CEP) to prevent stroke during transcatheter aortic valve implantation (TAVI) with control. TAVI has developed as a safe and effective treatment for aortic stenosis, but stroke remains an unpredictable and currently unavoidable outcome in some patients. The CEP strategy is conceptually robust: reducing embolic material reaching the brain should reduce stroke and improve outcomes. REFLECT I recruited 258 participants—54 in the roll-in phase, 141 randomised to receive the TriGuard CEP device and 63 to control. The safety outcome was measured against a predefined performance goal, and the efficacy outcome was a hierarchical endpoint of (i) all-cause mortality or any stroke at 30 days, (ii) worsening of National Institutes of Health Stroke Scale (NIHSS) day 2–5 post procedure or Montreal Cognitive Assessment worsening at 30 days, and (iii) total volume of cerebral ischaemic lesions detected by diffusion-weighted magnetic resonance imaging (DW-MRI). The key findings were that the device group met the safety outcome but not the efficacy endpoint. However, the trial was terminated early on the recommendation of the Data and Safety Monitoring Board (DSMB), and the sponsor elected not to resume the trial with this device. The trial had a neutral result, is statistically underpowered, and now outdated due to the introduction of a next generation device (TriGuard 3). So, what can we learn from REFLECT I to improve our understanding of CEP, inform the testing of similar devices, and the design and conduct of future clinical trials?

The reasoning behind the DSMB recommendation is not provided in the publication although the rates of efficacy outcomes for all-cause mortality and stroke, and NIHSS score suggest that continuing with the first-generation device was futile, and the secondary safety outcome signal was also of concern. The trialists could have provided more information about the DSMB decision and in future authors should consider reporting the rationale for such recommendations, particularly when a trial is terminated early.[4]

The transition of a novel device therapy to routine clinical practice is a step-wise process whereby we learn about safety and efficacy of a device, refine the device and seek to acquire knowledge about its wider clinical and cost-effectiveness as a routine treatment. New medical interventions generate increasingly powerful evidence starting from small safety and feasibility studies, then proof-of-principle evidence, efficacy studies with surrogate endpoints through to the gold standard of randomised controlled trials with hard clinical endpoints. Finally, real-world registries provide information about the how these perform in routine clinical practice (Graphical abstract).

Currently there are two devices with 'Conformité Européenne' (CE) marking: Sentinel CPS (Boston Scientific, Natick, MA, USA) and TriGuard 3 (Keystone Heart, Caesarea, Israel). The two broad questions relating to these devices are: is CEP clinically effective and are there differences between devices?

The evidence for the safety and efficacy of the Sentinel device is available from three published randomised trials and a combined patient-level analysis.[5–8] The patient-level analysis reports that in patients undergoing TAVI with CEP, stroke and the combination of mortality and stroke are reduced at 72 hours (10/533 (1.88%) vs. 29/533 (5.44%), odds ratio (OR) 0.35, 95% confidence interval (CI) 0.17–0.72, P = 0.003 and 11/533 (2.06%) vs. 32/533 (6.00%), OR 0.34, 95% CI 0.17–0.68, P = 0.001, respectively).[8] However, caution is needed in interpreting these results because of the small numbers and heterogeneity in the included studies. Two large randomised trials adequately powered for clinical outcomes are currently underway to provide a definitive answer about the efficacy of CEP to improve hard outcomes: Stroke PROTECTion With SEntinel During Transcatheter Aortic Valve Replacement (NCT04149535) aims to recruit 3000 patients assuming a stroke rate of 4% in the control group and 2% in the device group, and British Heart Foundation Randomised Trial of Routine Cerebral Embolic Protection in Transcatheter Aortic Valve Implantation (ISRCTN16665769) aims to recruit 7730 participants assuming a stroke rate of 3% in the control group reducing to 2% with the device.

The DEFLECT studies investigated the TriGuard device. DEFLECT I and II were single-arm studies of device safety and performance.[9,10] There was no adverse safety signal, and device 'success', defined as access to the aortic arch, deployment and positioning across the cerebral vessels, and device retrieval, was 64–80%. A secondary endpoint was new embolic lesions detected by DW-MRI, and the findings suggested that CEP may reduce the volume of these. DEFLECT III was a randomised trial testing safety, performance and efficacy.[11] It was designed to inform the endpoints for an efficacy trial and collected information about neurological events using imaging, stroke assessment and neurocognitive testing. This demonstrated that device success could be achieved in nearly 90% of patients; there was a reduction in new lesions on brain MRI and better neurological outcomes with TriGuard CEP. This series of safety and proof-of-concept studies has provided the foundation for the REFLECT trial, which was designed to fulfil the regulatory requirements of the US Food and Drug Administration.

The stroke rates reported in REFLECT I (10.7% in the device arm and 6.8% in control) are considerably higher than have been reported in recently large national registries, and there is a consistent signal emerging that routine neurological assessment identifies more stroke than self-report.[12,13] The clinical severity of events defined by neurological assessment requires further evaluation, but it highlights the need for clinical vigilance to identify stroke after TAVI, and also raises questions about the selection of clinical trial endpoints. What are the implications for the design of future trials of using different case definitions for stroke?

Another insight from REFLECT I is the disconnect between lesions seen on MRI and neurological outcomes: 85% of patients had ischaemic lesions on MRI after TAVI whereas 10% had clinical stroke. Post-hoc analysis suggested that that the CEP device may reduce the size of lesions. The association between disability and lesion size identified in REFLECT I is an observation which should be evaluated further, as it is possible that we should focus on reducing larger lesions which appear to have greater clinical impact. On the other hand, it may not be wise to ignore the clinically silent lesions altogether as they could have late consequences which short-term post trial follow-up has not yet identified.[14]

The REFLECT I trialists chose to assess safety against an historical event rate of 35% which appears very high for contemporary TAVI practice, and some of the secondary endpoints are potentially unrelated to CEP, for example coronary obstruction or valve-related dysfunction. The higher event rate of 21.8% in the device arm, compared with 8.5% in the control group (P = 0.03), was largely driven by major vascular complication and bleeding rates. The authors suggest this was related to the TAVI access site, but we should not be complacent about the possibility that the larger contralateral femoral access required for CEP may also have additional complications. Future studies should carefully consider which safety endpoints are most relevant in the study of a specific CEP device, and ensure that they consistently report detailed information on all complications.

Other CEP devices are now in development.[15] In broad terms these devices reduce the risk of embolic debris reaching the brain by positioning a mesh across the cerebral blood vessels to either capture or deflect the material destined for the brain. Access site, sheath size, mesh pore size and extent of cerebral circulation protection vary across devices. The ideal device would protect the entire cerebral circulation, with ease of delivery and positioning, stability throughout the procedure, and be associated with clinical effectiveness and safety. REFLECT I found that complete CEP protection was achieved in only around half of the patients who received the device. This prompted further device development, and resulted in the next generation TriGuard 3 device. It is possible that this experience will be repeated in future and CEP devices will be remodelled during real-world clinical testing.

As new CEP devices enter clinical research, the lessons from Sentinel and TriGuard studies will help to refine the devices and the design of study protocols to evaluate them. A key lesson from the REFLECT I trial is that it is imperative that all studies involving novel devices are published so that safety and efficacy data are available to inform device evaluation in clinical trials and practice. It is commendable that the authors and Journal Editors have sought to publish the findings from REFLECT I, despite the early termination of the trial, to ensure these can meaningfully contribute to the ongoing discussion about the role of CEP in TAVI.