Accepted Manuscript

Cerebral Embolic Protection During Transcatheter Aortic Valve Replacement

Samir R. Kapadia, MD, Susheel Kodali, MD, Raj Makkar, MD, Roxana Mehran, MD,Ronald M. Lazar, PhD, Robert Zivadinov, MD, PhD, Michael G. Dwyer, MD, HasanJilaihawi, MD, Renu Virmani, MD, Saif Anwaruddin, MD, Vinod H. Thourani, MD,Tamim Nazif, MD, Norman Mangner, MD, Felix Woitek, MD, Amar Krishnaswamy,MD, Stephanie Mick, MD, Tarun Chakravarty, MD, Mamoo Nakamura, MD, JamesM. McCabe, MD, Lowell Satler, MD, Alan Zajarias, MD, Wilson Y. Szeto, MD, LarsSvensson, MD, PhD, Maria C. Alu, MS, Roseann M. White, MS, Carlye Kraemer, MS,Azin Parhizgar, PhD, Martin B. Leon, MD, Axel Linke, MD

PII: S0735-1097(16)36768-7

DOI: 10.1016/j.jacc.2016.10.023

Reference: JAC 23128

To appear in: Journal of the American College of Cardiology

Received Date: 1 October 2016

Revised Date: 21 October 2016

Accepted Date: 22 October 2016

Please cite this article as: Kapadia SR, Kodali S, Makkar R, Mehran R, Lazar RM, Zivadinov R, DwyerMG, Jilaihawi H, Virmani R, Anwaruddin S, Thourani VH, Nazif T, Mangner N, Woitek F, KrishnaswamyA, Mick S, Chakravarty T, Nakamura M, McCabe JM, Satler L, Zajarias A, Szeto WY, Svensson L,Alu MC, White RM, Kraemer C, Parhizgar A, Leon MB, Linke A, Cerebral Embolic Protection DuringTranscatheter Aortic Valve Replacement, Journal of the American College of Cardiology (2016), doi:10.1016/j.jacc.2016.10.023.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service toour customers we are providing this early version of the manuscript. The manuscript will undergocopyediting, typesetting, and review of the resulting proof before it is published in its final form. Pleasenote that during the production process errors may be discovered which could affect the content, and alllegal disclaimers that apply to the journal pertain.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Cerebral Embolic Protection During Transcatheter Aortic Valve Replacement Samir R. Kapadia, MD*a, Susheel Kodali, MD*b, Raj Makkar, MDc, Roxana Mehran, MDd, Ronald M. Lazar, PhDb, Robert Zivadinov, MD, PhDe, Michael G. Dwyer, MDe, Hasan Jilaihawi, MDf, Renu Virmani, MDg, Saif Anwaruddin, MDh, Vinod H. Thourani, MDi, Tamim Nazif, MDb, Norman Mangner, MDj, Felix Woitek, MDj, Amar Krishnaswamy, MDa, Stephanie Mick, MDa, Tarun Chakravarty, MDc, Mamoo Nakamura, MDc, James M. McCabe, MDk, Lowell Satler, MDl, Alan Zajarias, MDm, Wilson Y. Szeto, MDh, Lars Svensson, MD, PhDa, Maria C. Alu, MSb, Roseann M. White, MSn, Carlye Kraemer, MSo, Azin Parhizgar, PhDp, Martin B. Leon, MDb, and Axel Linke, MDj,q *co-first authors From: aCleveland Clinic, Cleveland, OH ; bColumbia University Medical Center, New York, NY; cCedars-Sinai Medical Center, Los Angeles, CA; dMount Sinai School of Medicine, New York, NY; eBuffalo Neuroimaging Analysis Center, Buffalo, NY; fNYU Langone Medical Center, New York, NY; gCV Path Institute, Gaithersburg, MD; hUniversity of Pennsylvania, Philadelphia, PA; iEmory University, Atlanta, GA; jHerzzentrum Leipzig GmbH –Universitätsklinik, Leipzig, Germany; kUniversity of Washington, Seattle, WA; lMedstar Washington Hospital Center, Washington, DC; mWashington University School of Medicine, St. Louis, MO; nDuke Clinical Research Institute, Durham, NC; oNAMSA, Minneapolis, MN; pClaret Medical Inc, Santa Rosa, CA; qLeipzig Heart Institute, Leipzig, Germany Short Running Title: Embolic protection in TAVR Sources of Funding: The SENTINEL Trial was funded by Claret Medical, Inc. Relationships with Industry: Authors disclose the following relationships with industry:Dr. Kodali is a consultant for Edwards Lifesciences. Dr. Makkar has received research support from Edwards Lifesciences and St. Jude Medical and is a consultant for Abbott Vascular, Cordis, and Medtronic. He also holds equity in Entourage Medical. Dr. Mehran receives research grant support from Eli Lilly/DSI, BMS, AstraZeneca, The Medicines Company, OrbusNeich, Bayer, and CSL Behring; is a consultant for Janssen Pharmaceuticals, Osprey Medical, Watermark Research Partners, and Medscape, a member of the Scientific Advisory Board of Abbott Laboratories, and holds Equity or Stock Options in Claret Medical Inc. and Elixir Medical Corporation. Dr. Zivadinov has received personal compensation from Teva Pharmaceuticals, Biogen Idec, EMD Serono, Genzyme-Sanofi, Claret Medical, IMS Health and Novartis for speaking and consultant fees. He received financial support for research activities from Teva Pharmaceuticals, Genzyme-Sanofi, Novartis, Claret Medical, Intekrin and IMS Health. Dr. Zivadinov serves on editorial board of J Alzh Dis, BMC Med, BMC Neurol, Vein and Lymphatics and Clinical CNS Drugs. He is Executive Director and Treasurer of International Society for Neurovascular Disease. Dr. Jilaihawi is a consultant for Edwards Lifesciences, St. Jude Medical, and Venus Medtech. Dr. Anwaruddin is a speaker, consultant, and proctor for Edwards Lifesciences and Medtronic and a consultant for the American College of Radiology. Dr. Thourani is a consultant for Edwards Lifesciences, Sorin Medical, St. Jude Medical, and DirectFlow. Dr. Nazif is a consultant for Edwards Lifesciences. Dr. Nakamura is a consultant for Edwards Lifesciences. Dr. McCabe is a consultant for Edwards Lifesciences. Dr. Zajarias is a

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

consultant for Edwards Lifesciences. Dr. Szeto is a consultant for MicroInterventional Devices. Dr. Svensson holds equite in Cardiosolutions and ValvXchange and Intellectual Property Rights for Posthorax. Ms. Alu is a consultant for Claret Medical. Ms. White is a consultant for Claret Medical through Duke Clinical Research Institute. Ms. Kraemer is a consultant for Claret Medical. Dr. Parhizgar is an employee of Claret Medical. The other authors report no relevant relationships with industry. Acknowledgements: The authors would like to thank Matthew T. Finn, MD (Columbia University Medical Center) for assistance with manuscript preparation. Address for Correspondence: Samir Kapadia, MD Cleveland Clinic 9500 Euclid Avenue, Desk J2-3 Cleveland, OH 44195 Tel: 216-444-6697 Email: kapadis@ccf.org

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Abstract Background Neurological events and brain infarction after transcatheter aortic valve replacement (TAVR) are concerns which may be reduced with transcatheter embolic protection (TEP). Objective Evaluate the safety and efficacy of TEP during TAVR. Methods Nineteen centers randomized 363 patients undergoing TAVR to safety (n=123), device imaging (n=121), and control imaging (n=119). The primary safety endpoint was major adverse cardiac and cerebrovascular events (MACCE) at 30 days and the primary efficacy endpoint was reduction in new lesion volume in protected brain territories on MRI scans at 2-7 days. Patients underwent neurocognitive assessments and the debris captured was analyzed. Results MACCE (7.3%) was non-inferior to the performance goal (18.3%, pnoninferior<0.001) and not statistically different from control (9.9%, p=0.41). New lesion volume was 178.0 mm3 in controls and 102.8 mm3 in the device arm (p=0.33). A post hoc multivariable analysis identified preexisting lesion volume and valve type to be predictors of new lesion volume. Strokes at 30d were 9.1% in controls and 5.6% in device patients (p=0.25). Although neurocognitive function was similar in control and device patients, there was correlation between lesion volume and neurocognitive decline (p=0.0022). Histopathologic debris, found within filters in 99% of patients, included thrombus, calcification, valve tissue, artery wall and foreign material. Conclusions TEP was safe, captured embolic debris 99% patients, and did not change neurocognitive function. Reduction in new lesion volume on MR scans was not statistically significant. Keywords: Cerebral embolic protection, transcatheter aortic valve replacement (TAVR), stroke, neuroimaging. Trial Registration: ClinicalTrials.gov #NCT02214277 Abbreviations ANOVA Analysis of Variance AS Aortic Stenosis DW-MRI Diffusion-weighted Magnetic Resonance Imaging FLAIR Fluid-attenuated Inversion Recovery ITT Intention to Treat MACCE Major Adverse Cardiac and Cerebral Events MMSE Mini Mental State Exam MRI Magnetic Resonance Imaging mRS Modified Rankin Score MSCT Multi Slice Computed Tomography NIHSS National Institutes of Health Stroke Scale STS Society of Thoracic Surgeons TAVR Transcatheter Aortic Valve Replacement TCEP Transcatheter Cerebral Embolic Protection

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Introduction

Transcatheter aortic valve replacement (TAVR) is an important new therapy for high and

intermediate-risk patients with severe aortic stenosis (AS).(1-8) However, strokes remain a

concerning complication after TAVR and are associated with increased mortality and

morbidity.(1,2,4,9-13) In addition, clinically “silent” brain infarctions seen on magnetic

resonance (MR) scans have been associated with neurocognitive function changes(14-17) and

are present in as many of 80% of patients after TAVR.(18-21) Although the etiology of strokes

and MR perfusion abnormalities is multifactorial in this elderly co-morbid population, the

majority are due to embolization of particulate debris during the procedure.(22,23) Previous

exploratory studies have attempted to minimize procedural embolization using either

transcatheter filters or deflection devices.(24-30) The present randomized trial was designed to

assess the safety of transcatheter cerebral embolic protection (TCEP) during TAVR and the

efficacy in reducing the effects of cerebral embolization.

Methods

Patients

The study population included 363 patients with severe symptomatic AS and planned TAVR

who were at high risk for surgery from 17 centers in the U.S. and 2 centers in Germany. All

patients had multi-slice computed tomography (MSCT) scans which were analyzed by a core

laboratory and reviewed by a committee to determine treatment eligibility using the Sentinel®

TCEP device (Claret Medical, Santa Rosa, CA, USA). Other exclusions included patients with

known contraindications for right radial or brachial artery access and patients unable to undergo

MRI evaluation of their brain for any reason.

Study Device and Procedure

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

The SentinelT TCEP device consists of 2 filters within a single 6 French delivery catheter

percutaneously placed from the right radial (preferred) or brachial artery over a 0.014” guide

wire (Figure 1). The filters are positioned in the brachiocephalic and the left common carotid

arteries before TAVR and are withdrawn into the catheter and removed after TAVR. Details of

the device and the procedure have been previously described.(26,31)

Study Design

In this prospective multicenter randomized trial, patients undergoing TAVR were randomized

1:1:1 into a safety arm (TCEP only) and two imaging cohorts, in which patients were randomly

treated with TCEP (device arm) or without TCEP (control arm) (Figure 2). TCEP safety was

assessed in the safety and device arms (device safety cohort). Reason to include safety arm was

show clinical safety without increasing cost of the trial by eliminating MRI cost. Patients were

blinded to treatment assignment. Blinded diffusion weighted MRI (DW-MRI) and

neurocognitive function assessments were performed in the device and control arms. Particulate

debris from the extracted filters was studied in the device arm. All patients underwent rigorous

neurologic evaluations post-TAVR, at 30 days and at 90 days.

Brain MRI Studies

Brain MRI using a 3 Tesla scanner was performed in both imaging arms (device and control) at

baseline, and post-TAVR at 2-7 days and at 30 days. All MRI studies were analyzed by a core

laboratory in a blinded manner. Appendix 1 in the Supplementary Appendix provides a

description of the methodology for MRI acquisition and analysis. Post-TAVR studies were

matched with baseline scans and subtraction analyses were performed to identify new lesions.

Protected territories were defined as brain territories entirely perfused by vessels protected by

TEP and “all” territories refers to the entire brain.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Neurocognitive Function

Device and control arms underwent comprehensive neurocognitive assessment tailored for

TAVR patients and designed to evaluate seven domains of neurocognitive function, including bi-

hemispheral and hemisphere-specific attention, executive function, processing speed, working

memory, visual memory, mental status, and depression (Appendix 2, Supplementary Appendix).

Trained and certified test administrators and neurocognitive core lab personnel were blinded to

randomization assignment.

Histopathologic Assessment of Debris

All filters from the device arm were stored in formalin and sent to a histopathology core lab for a

blinded analysis. Extracted debris was stained, examined with light microscopy and the debris

was catalogued as thrombus, calcium, valve tissue, or catheter fragments. A quantitative

analysis was also performed to measure the size of embolic debris.

Study Endpoints

The primary safety end point was the occurrence of major adverse cardiac and cerebrovascular

events (MACCE) at 30 days compared to a historical performance goal. MACCE was defined as

all death, all strokes (Disabling and Non-Disabling, VARC-2), and acute kidney injury (stage 3,

VARC-2).(32) The occurrence of stroke was assessed by neurologist-administered National

Institutes of Health Stroke Score (NIHSS) and modified Rankin score (mRS) at Baseline (< 14

days pre-procedure), Discharge, and 30 days. For those experiencing a stroke within 30 days, 90-

day NIHSS and mRS were also administered by a neurologist for the purposes of determining

stroke severity.

The primary efficacy endpoint was the reduction in median total new lesion volume in protected

territories between the device and control arms assessed by DW-MRI at 2-7 days after TAVR.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Minimum treatment effect of 30% reduction in median total new lesion volume in protected

territories was also pre specified as an observational success criteria. Total new lesion volume

was defined as the sum of all diffusion-positive new cerebral lesion volumes in post-procedural

scans relative to the pre-TAVR scans.

Other pre-specified secondary endpoints included device success, vascular complications,

new lesion number in protected and all territories, and the correlation of lesion volume with

neurocognitive function changes, and histopathology evaluations.

Statistical Analysis

Fisher’s exact test was used to compare categorical variables. Continuous variables, which are

presented as means with standard deviations or medians with interquartile ranges, as appropriate,

were compared with the use of analysis of variance (ANOVA), non-parametric ANOVA, or the

Wilcoxon rank-sum test. The point estimate for the historical performance goal for the primary

safety endpoint at 30 days post-TAVR was derived from a review of the published literature of

30-day TAVR outcomes.(1,2,4) The boundary was selected by first weighting the published

MACCE rates by the expected proportion of transfemoral and transapical cases using the

following formula: weighted MACCE Rate = [20.2% x 20% (TA) + 12.0% x 80%(TF)]*2/3 +

[12.62]*1/3 = 13.3%. The performance goal of 18.3% was derived by adding a conservative non-

inferiority margin of 5% to the weighted literature rate of 13.3%. Sample size estimates for

comparing the total new lesion volume from the protected territories between the two

randomized imaging arms were obtained based on a Wilcoxon-Mann-Whitney test, assuming

data with a lognormal distribution and the following means: Raw Mean (SD): Test 474.2 (813.6)

vs. Control 1029.7 (2424.12); Lognormal Mean (SD): Test 5.4 (1.2) vs. Control 6.0 (1.3). Under

these assumptions, seventy-two subjects per arm were required, with an 80% power and an alpha

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

of 0.05 (two-sided). With an estimated loss allowance of 35%, 120 subjects were planned to be

randomized to each imaging arm to achieve 75 evaluable subjects. The primary efficacy

endpoint, new median lesion volume differences in the test and control arms were compared

using the Wilcoxon rank-sum test.

A z-score for each neurocognitive function domain was calculated based on normative

means and standard deviations for each neurocognitive test. Change scores were calculated (by

domain) by subtracting baseline scores from the 30-day or 90-day post-TAVR scores.

Comparison of the change in composite neurocognitive z-scores was performed controlling for

mini mental state examination (MMSE), education, and the depression scores.

Multivariable analysis was undertaken to determine covariates of new lesion volumes,

starting with identifying all baseline univariate predictors with a p-value < 0.1. Stepwise linear

regression was performed to identify independent predictors. Adjustment models to account for

the effect of multivariable predictors on new lesion volume are described in Appendix 3,

Supplementary Appendix.

Statistical analyses were performed on the intention-to-treat (ITT) population using SAS

version 9.3 (SAS Institute, Cary, NC).

Results

Study Characteristics

240 patients were randomized to the imaging cohort (119 control and 121 device) and 123

patients to the safety arm (Figure 2). Within the imaging cohort, MRI studies at baseline and 2-

7 days post-TAVR were performed in 189 (78.8%) patients and neurocognitive assessments

were completed at baseline and 30 days in 185 (77.1%) patients. Baseline characteristics of the

primary safety cohort and of those with and without paired MRI studies (primary efficacy

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

cohort) are presented in Tables 1 and 2 of the supplementary appendix. Of note, there were no

significant differences between groups in the primary safety cohort and the only baseline

characteristics which differed between those with and without paired MRI were history of prior

coronary artery bypass graft and mean gradient.

Overall, the study population was elderly (median age 83.4 years), the majority (52.1%)

were females, median Society of Thoracic Surgeons (STS) score was 6.0%, and there were

frequent co-morbidities including atrial fibrillation (31.7%) and prior strokes (5.8%). Baseline

characteristics were generally well balanced for the entire population (Table 1) and for the paired

MRI and paired neurocognitive function cohorts (Tables 3 and 4, Supplementary Appendix).

Due to timing of FDA approval and operator choice, four different TAVR devices were used in

this trial; Sapien XT (17.8%) and Sapien 3 (52.4%) (both Edwards Lifesciences, Irvine, CA,

USA) as well as CoreValve (3.9%) and Evolut R (25.9%) (both Medtronic, Minneapolis, MN,

USA). TAVR systems were used with similar distribution across all three randomized treatment

groups.

Procedural Details and Clinical Outcomes

TAVR was performed via the femoral artery in 94.7% of cases and TCEP was delivered from the

radial and brachial arteries in 93.2% and 5.6% of cases. Delivery and retrieval of both filters was

successful in 94.4% of patients. In the device arm vs. control, there was an increase in total

procedure time (p=0.01) and fluoroscopy time (p=0.007) (Table 1).

MACCE in the device and safety arms was 7.3% and the upper bound of the 95%

confidence interval (11.4%) was less than the 18.3% performance goal (p<0.001 for non-

inferiority) (Figure 3, Table 2). MACCE in the control arm (9.9%) was not statistically different

compared with the device and safety arms (p=0.405). Strokes were not significantly different in

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

the device and safety arms vs. the control arm (5.6% vs. 9.1%, p=0.25). There were no

differences in other important endpoints including acute kidney injury or vascular complications

(Table 2).

MRI Efficacy Primary Outcomes

The median total new lesion volume in protected territories was 42% lower, meeting the 30% pre

specified treatment effect success criteria, but was not significantly different in device vs. control

arms (102.8 mm3 versus 178.0 mm3, p=0.33) (Figure 2). Total new lesion volume in all

territories was also not statistically different in device vs. control arms (294 mm3 vs. 309.8 mm3,

p = 0.81). New lesion number in device vs. control arms in both protected and all territories was

unchanged (Table 3). When analyzed by valve type, there were significant differences in new

lesion volume and number in both protected and all territories (Table 5, Supplementary

Appendix). The median total new lesion volume at 30 days was 0 for in both protected and all

territories for both the Device and Control arms (Table 6, Supplementary Appendix).

Post Hoc Multivariable Analysis

Univariate and multivariable analyses indicated that baseline T2/FLAIR lesion volume on MRI

(a marker of prior injury and gliosis, often called white matter disease) was the strongest baseline

predictor of new lesion volume after TAVR (Table 7, Supplementary Appendix). After

adjusting for valve type, baseline T2/FLAIR lesion volume, and an interaction between valve

type and treatment arm, there was a significant reduction in new lesion volume in both protected

and all territories in the device vs. control arms (p=0.025 for protected and p=0.050 for all

territories (Table 8, Supplementary Appendix). After similar adjustments for baseline

T2/FLAIR lesion volume, there were variable responses associated with specific valve types

(Table 9, Supplementary Appendix).

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Neurocognitive Function and Histopathology

Neurocognitive test batteries were similar in all 7 domains with no difference in overall

composite scores at baseline, 30 days or 90 days between device and control arms (Table 4).

However, the change in neurocognitive scores from baseline to 30-day follow-up correlated with

median new lesion volume in protected territories (r = -0.20275, R2 = 4.1, p = 0.0109) and all

territories (r = -0.23562, R2 = 5.6, P = 0.003) (Figure 1, Supplementary Appendix).

Histopathologic debris was found in retrieved filters in 99% of patients (Figure 4). Most

frequently captured debris components included acute thrombus with tissue elements, artery

wall, calcification, valve tissue, and foreign materials. More than 80% of the debris was 150–

500 microns (maximum diameter) and < 5% was greater than 1000 microns (Figure 4).

Discussion

The main results from this important multicenter randomized trial evaluating the role of embolic

protection using the Sentinel™ device during TAVR are as follows: (1) transcatheter placement

of a dual filter device was successful in most patients without safety concerns; (2) the primary

efficacy endpoint (reduction in median new lesion volume on MRI at 2-7 days in protected

territories) was not met; however, after adjusting for valve type and baseline T2/FLAIR lesion

volume in a post hoc analysis, there were significant differences in new lesion volumes favoring

embolic protection; (3) neurocognitive function was not significantly improved, but there was

correlation between new lesion volume and number and neurocognition at 30 days; (4)

particulate debris during TAVR was found in almost all patients including diverse biologic and

foreign materials.

The first hurdle of accessory therapies to improve TAVR outcomes is the demonstration

of technical feasibility and clinical safety. As suggested in prior smaller studies,(26,31) the

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Sentinel® dual filter device was easily delivered, usually from the right radial artery, and was

compatible with standard TAVR workflow. The total procedure was longer by approximately 13

minutes with an additional 3 minutes of fluoroscopy time. Importantly, clinical safety outcomes

were much lower than the pre-specified performance goal, and MACCE point-estimates were

lower in the device arms compared to control. Although the largest difference was in minor

strokes, it has been shown that all strokes, and even TIAs, confer increased risk of mortality risk

among TAVR. (9)

Historically, demonstration of clinical efficacy with embolic protection to reduce

deleterious target organ effects has been problematic. The proposed or accepted use of embolic

protection devices for the brain, heart, kidneys, and legs has been based largely on observational

studies indicating device safety combined with surrogate clinical efficacy endpoints.(18,19,33-

36) Early TAVR trials showing increased peri-procedural stroke frequency(1-3) and MRI

examinations revealing concordant ischemic deficits(18,19,21) heightened the need to develop

brain sparing therapies and encouraged the use of quantitative MRI analyses as a surrogate

endpoint. In the present trial, with embolic protection there was a 38% reduction in all strokes at

30 days, which did not reach statistical significance. There was a 42% reduction in MRI median

new lesion volume at 2-7 days in the device arm compared with controls, which also did not

reach statistical significance.

Several limitations of the study likely contributed to the lack of statistical significance

despite this observed 42% reduction in new lesion volume in protected territories. First, despite

the use of 3T MRI scanners to improve the accuracy of characterizing new lesions and

subtraction imaging methodology to provide unbiased quantitative analyses, there was

considerable variance in MRI post-procedure results, in part due to rapidly changing new lesion

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

volumes and numbers during the broad 2-7 day follow-up window. In addition, 3T MRI is

potentially more prone to scanner signal (B0 and B1) inhomogeneity across the brain,

particularly on FLAIR imaging, although the increased inhomogeneity is offset by the sensitivity

increase and by software correction using the N3 algorithm. Second, there was little benchmark

MRI data on which to base control arm assumptions and the observed new lesion volume and

number were less than predicted from the recently published CLEAN-TAVI trial.(25) This may

be the result of using CoreValve exclusively in Clean TAVI, whereas the use of self-expanding

valves was ~20% in the Sentinel trial. Third, the impact of baseline T2/FLAIR lesion volume on

subsequent new lesion volume was not accounted for in the trial design. Prior neurology studies

have demonstrated that baseline disease burden is an important predictor of subsequent clinical

events after interventions (37-39). Fourth, different TAVR devices were included in this trial and

the randomization scheme was not stratified according to valve category. Both the control arm

MRI results and the response to embolic protection appeared to differ with varying TAVR

systems.

At the time the Sentinel Trial was designed, only one device was commercially available

in the US, and therefore no stratification by device type was anticipated in the randomization

structure. However, several new devices became available over the course of trial. Additionally,

risk factors for stroke, especially findings on baseline MRI, were incompletely understood at the

time at the time the study was designed. When evaluation of univariate predictors of new lesion

volume revealed that device type and baseline T2/FLAIR lesion volume were potentially

important confounders or effect modifiers of the relationship between TCEP and new lesion

volume, we determined that is was necessary to perform multivariable analysis to adjust for the

unanticipated baseline differences in brain infarction volume and valve type. After adjusting for

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

these variables, there was a significant difference in new lesion volume favoring neuroprotection

in both protected (p = 0.025) and all territories (p = 0.050). The study provides a very interesting

observation regarding differences in MRI findings due to implanted valve type. In the control

group of patients, the volume of new MRI lesions was lower in S3 compared to Evolut R or

Sapien XT. The overall treatment effect, after adjustment for TAVR Device and the interaction

between TAVR and Treatment, is a 49% reduction in post procedure new lesion volume in

protected areas. However, SAPIEN 3 generated the lowest post procedure new lesion volume

(30% - 50% lower than the other TAVR devices (Supplemental Table 6). Therefore, SAPIEN 3

derived the least benefit from the use of the Sentinel device, resulting in little to no difference

between the treatment arms. This is responsible for the significant interaction effect between the

device and treatment. The treatment effect of the Sentinel device was significant in non-S3

valves. The reasons for these differences are not clear, but will be an important future question

for clinical trials. There may be other factors that may explain these differences like use of pre-

dilatation or post-dilatation, operator experience, or patient selection for different valves.

Although these variables did not reach statistical significance in multivariate modeling, limited

power of the study and possible interaction with valve types do not allow us to rule out their

contribution to observed differences between valve types. It is important to note that first

generation balloon expandable and self-expanding valves have reported similar stroke rates. The

clinical stroke rate reported from S3 has been very low, but has not been directly compared to

new generation self-expanding valves.

The neurocognitive function test domains utilized in this trial were rigorously obtained by

trained examiners and were customized to optimize the sensitivity of identifying changes

associated with diffuse cerebral embolization. Although there was no difference in

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

neurocognition at 30 days between the embolic protection and control arms, there was an

important relationship linking cumulative neurocognition scores with new lesion volumes and

numbers.

As has been noted in other trials using similar filters for neuroprotection during TAVR

(22,23), there was a striking consistency of retrieved materials in almost all patients in this study.

The observation of frequent thrombus, artery wall, valve tissue, and calcification suggests that

overly aggressive device manipulation within the aortic valvar complex should be avoided,

whenever possible.

There are several limitations associated with this clinical trial. The Sentinel® dual filter

device appears safe and feasible but the embolic protection afforded excludes the territory of the

left vertebral artery. The observation that residual new lesions are still present in protected

territories after neuroprotection indicates that either the current transcatheter devices are

suboptimal in debris capture or that post-procedure particulate embolization is ongoing and

occurs after removing the filters. Further, it is possible that that some of the retrieved material in

the filters was not directly related to TAVR, but rather was due to placement of TEP. Follow-up

MRI studies were not obtained in 25% of patients from the imaging cohort due to patient

compliance and the need for new pacemakers post-TAVR. Despite being the largest randomized

trial examining neuroprotection during TAVR, the sample size was clearly too low to assess

clinical outcomes, and in retrospect, was also too low to evaluate follow-up MRI findings or

neurocognitive outcomes. Finally, the analyses of valve type differences and multivariable

analysis to account for confounders should be viewed as hypothesis generating and non-

definitive.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

There are important lessons from this trial, which should impact future research on

neuroprotection during TAVR. In particular, the use of quantitative MRI analyses as a surrogate

endpoint must be further clarified, including stricter time windows for follow-up studies and

more accurate assumptions of expected control arm results, such that sample sizes can be

adjusted appropriately. The requirement of baseline MRI studies to account for prior lesion

volume and the need to adjust for differences in valve type (e.g. stratification of valve types

during randomization) cannot be overemphasized.

In conclusion, this randomized trial investigating the SentinelTM dual filter

neuroprotection therapy provides reassuring evidence of device safety and confirms the high-

frequency of embolic debris capture.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

REFERENCES

1. Leon MB, Smith CR, Mack M et al. Transcatheter aortic-valve implantation for aortic

stenosis in patients who cannot undergo surgery. N Engl J Med 2010;363:1597-607.

2. Smith CR, Leon MB, Mack MJ et al. Transcatheter versus surgical aortic-valve

replacement in high-risk patients. N Engl J Med 2011;364:2187-98.

3. Popma JJ, Adams DH, Reardon MJ et al. Transcatheter aortic valve replacement

using a self-expanding bioprosthesis in patients with severe aortic stenosis at extreme

risk for surgery. J Am Coll Cardiol 2014;63:1972-81.

4. Adams DH, Popma JJ, Reardon MJ et al. Transcatheter aortic-valve replacement with

a self-expanding prosthesis. N Engl J Med 2014;370:1790-8.

5. Leon MB, Smith CR, Mack MJ et al. Transcatheter or Surgical Aortic-Valve

Replacement in Intermediate-Risk Patients. N Engl J Med 2016;374:1609-20.

6. Thourani VH, Kodali S, Makkar RR et al. Transcatheter aortic valve replacement

versus surgical valve replacement in intermediate-risk patients: a propensity score

analysis. Lancet 2016;387:2218-25.

7. Reardon MJ, Kleiman NS, Adams DH et al. Outcomes in the Randomized CoreValve

US Pivotal High-risk Trial in Patients With a Society of Thoracic Surgeons Risk

Score of 7% or Less. JAMA Cardiology 2016 doi: 10.1001/jamacardio.2016.2257.

[Epub ahead of print].

8. Holmes DR, Jr., Nishimura RA, Grover FL et al. Annual Outcomes With

Transcatheter Valve Therapy: From the STS/ACC TVT Registry. J Am Coll Cardiol

2015;66:2813-23.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

9. Kapadia S, Agarwal S, Miller DC et al. Insights Into Timing, Risk Factors, and

Outcomes of Stroke and TIA After Transcatheter Aortic Valve Replacement in the

PARTNER Trial. Circulation Cardiovasc Interv 2016;9.

10. Athappan G, Gajulapalli RD, Sengodan P et al. Influence of transcatheter aortic valve

replacement strategy and valve design on stroke after transcatheter aortic valve

replacement: a meta-analysis and systematic review of literature. J Am Coll Cardiol

2014;63:2101-10.

11. Athappan G, Gajulapalli RD, Tuzcu ME, Svensson LG, Kapadia SR. A systematic

review on the safety of second-generation transcatheter aortic valves.

EuroIntervention 2016;11:1034-43.

12. Auffret V, Regueiro A, Del Trigo M et al. Predictors of Early Cerebrovascular Events

in Patients With Aortic Stenosis Undergoing Transcatheter Aortic Valve

Replacement. J Am Coll Cardiol 2016;68:673-84.

13. Gleason TG, Schindler JT, Adams DH et al. The risk and extent of neurologic events

are equivalent for high-risk patients treated with transcatheter or surgical aortic valve

replacement. J Thorac Cardiovasc Surg 2016;152:85-96.

14. Floyd TF, Giovannetti T. Neurocognitive outcomes in older adults after transcatheter

aortic valve replacement. J Thorac Cardiovasc Surg 2012;144:1539.

15. Vermeer SE, Prins ND, den Heijer T, Hofman A, Koudstaal PJ, Breteler MM. Silent

brain infarcts and the risk of dementia and cognitive decline. N Engl J Med

2003;348:1215-22.

16. Vermeer SE, Longstreth WT, Jr., Koudstaal PJ. Silent brain infarcts: a systematic

review. Lancet Neurology 2007;6:611-9.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

17. Svensson LG, Blackstone EH, Apperson-Hansen C et al. Implications from

neurologic assessment of brain protection for total arch replacement from a

randomized trial. J Thorac Cardiovasc Surg 2015;150:1140-7.e11.

18. Ghanem A, Muller A, Nahle CP et al. Risk and fate of cerebral embolism after

transfemoral aortic valve implantation: a prospective pilot study with diffusion-

weighted magnetic resonance imaging. J Am Coll Cardiol 2010;55:1427-32.

19. Kahlert P, Knipp SC, Schlamann M et al. Silent and apparent cerebral ischemia after

percutaneous transfemoral aortic valve implantation: a diffusion-weighted magnetic

resonance imaging study. Circulation 2010;121:870-8.

20. Fairbairn TA, Mather AN, Bijsterveld P et al. Diffusion-weighted MRI determined

cerebral embolic infarction following transcatheter aortic valve implantation:

assessment of predictive risk factors and the relationship to subsequent health status.

Heart 2012;98:18-23.

21. Spaziano M, Francese DP, Leon MB, Genereux P. Imaging and functional testing to

assess clinical and subclinical neurological events after transcatheter or surgical aortic

valve replacement: a comprehensive review. J Am Coll Cardiol 2014;64:1950-63.

22. Van Mieghem NM, Schipper ME, Ladich E et al. Histopathology of embolic debris

captured during transcatheter aortic valve replacement. Circulation 2013;127:2194-

201.

23. Schmidt T, Schluter M, Alessandrini H et al. Histology of debris captured by a

cerebral protection system during transcatheter valve-in-valve implantation. Heart

2016; 102:1573-80.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

24. Lansky AJ, Schofer J, Tchetche D et al. A prospective randomized evaluation of the

TriGuard HDH embolic DEFLECTion device during transcatheter aortic valve

implantation: results from the DEFLECT III trial. Eur Heart J 2015;36:2070-2078.

25. Haussig S, Mangner N, Dwyer MG et al. Effect of a Cerebral Protection Device on

Brain Lesions Following Transcatheter Aortic Valve Implantation in Patients With

Severe Aortic Stenosis: The CLEAN-TAVI Randomized Clinical Trial. JAMA

2016;316:592-601.

26. Van Mieghem NM, van Gils L, Ahmad H et al. Filter-based cerebral embolic

protection with transcatheter aortic valve implantation: the randomised MISTRAL-C

trial. EuroIntervention 2016;12:499-507.

27. Campelo-Parada F, Regueiro A, Dumont E et al. Embolic protection in patients

undergoing transaortic transcatheter aortic valve replacement: initial experience with

the TriGuard HDH embolic deflection device. J Card Surg 2016 doi:

10.1111/jocs.12822. (epub ahead of print).

28. Pagnesi M, Martino EA, Chiarito M et al. Silent cerebral injury after transcatheter

aortic valve implantation and the preventive role of embolic protection devices: A

systematic review and meta-analysis. Int J Cardiol 2016;221:97-106.

29. Samim M, van der Worp B, Agostoni P et al. TriGuard HDH embolic deflection

device for cerebral protection during transcatheter aortic valve replacement. Catheter

Cardiovasc Interv 2016 doi: 10.1002/ccd.26566. [Epub ahead of print].

30. Wendt D, Kleinbongard P, Knipp S et al. Intraaortic Protection From Embolization in

Patients Undergoing Transaortic Transcatheter Aortic Valve Implantation. Ann

Thorac Surg 2015;100:686-91.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

31. Naber CK, Ghanem A, Abizaid AA et al. First-in-man use of a novel embolic

protection device for patients undergoing transcatheter aortic valve implantation.

EuroIntervention 2012;8:43-50.

32. Kappetein AP, Head SJ, Genereux P et al. Updated standardized endpoint definitions

for transcatheter aortic valve implantation: the Valve Academic Research

Consortium-2 consensus document. EuroIntervention 2012;8:782-95.

33. Iqbal MB, Nadra IJ, Ding L et al. Embolic protection device use and its association

with procedural safety and long-term outcomes following saphenous vein graft

intervention: An analysis from the British Columbia Cardiac registry. Catheter

Cardiovasc Interv 2016;88:73-83.

34. Mendes BC, Oderich GS, Fleming MD et al. Clinical significance of embolic events

in patients undergoing endovascular femoropopliteal interventions with or without

embolic protection devices. J Vasc Surg 2014;59:359-367.e1.

35. Paul TK, Lee JH, White CJ. Renal embolic protection devices improve blood flow

after stenting for atherosclerotic renal artery stenosis. Catheter Cardiovasc Interv

2012;80:1019-22.

36. Roberts D, Niazi K, Miller W et al. Effective endovascular treatment of calcified

femoropopliteal disease with directional atherectomy and distal embolic protection:

final results of the DEFINITIVE Ca(+)(+) trial. Catheter Cardiovasc Interv

2014;84:236-44.

37. Ederle J, Davagnanam I, van der Worp HB et al. Effect of white-matter lesions on the

risk of periprocedural stroke after carotid artery stenting versus endarterectomy in the

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

International Carotid Stenting Study (ICSS): a prespecified analysis of data from a

randomised trial. Lancet Neurology 2013;12:866-72.

38. Mashour GA, Moore LE, Lele AV, Robicsek SA, Gelb AW. Perioperative care of

patients at high risk for stroke during or after non-cardiac, non-neurologic surgery:

consensus statement from the Society for Neuroscience in Anesthesiology and

Critical Care*. J Neurosurg Anesthesiol 2014;26:273-85.

39. Ng JL, Chan MT, Gelb AW. Perioperative stroke in noncardiac, nonneurosurgical

surgery. Anesthesiology 2011;115:879-90.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

FIGURE LEGENDS

Figure 1: Sentinel Device

Panel A shows the Sentinel device. Panel B shows the proximal filter is placed in the innominate

artery and the distal filter is placed in the left carotid artery. The pore size of the filters is 140

micron. Panel C shows fluoroscopic image of the device.

Figure 2: Study Flow

Sentinel study randomized patients in 3 arms including imaging control arm, imaging device arm

or safety arm with 1:1:1 randomization. Patients in the safety arm were followed for clinical

events without MRI (magnetic resonance imaging) or neurocognitive (NC) testing. The figure

shows the patients and reasons for not having MRI or NC testing. TAVR: Transcatheter aortic

valve replacement, F/U: follow-up, W/D: study withdrawal, PPM: Permanent pacemaker, TEP:

Thromboembolic protection.

Figure 3: Primary safety (a) and efficacy (b) endpoints

Panel A shows MACCE in the device and safety arms was 7.3% and the upper bound of the 95%

confidence interval (11.4%) was less than the 18.3% performance goal (p<0.001 for non-

inferiority). MACCE in the control arm (9.9%) was not statistically different compared with the

device and safety arms (p=0.405).

Panel B shows median total new lesion volume in protected territories in control and device arm.

The median total new lesion volume in protected territories was 42% lower, meeting the 30% pre

specified treatment effect success criteria, but was not significantly different in device vs. control

arms (102.8 mm3 versus 178.0 mm3, p=0.33). This is a box plot with central line representing

median and box representing the interquartile range (25th-75th percentile).

Figure 4: Histopathology Particulate Debris Analysis

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Study analyzed the particulate debris from filters. Panel A shows rate of debris capture by type

and panel B shows average number of particles captured by size in maximum diameter.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table 1. Baseline Characteristics and Procedural Details

Control Arm (N=119)

Device Arm (N=121)

Safety Arm (N=123)

Total Randomized (N=363)

P-value1

Age 85.0 [78.4, 89.4] 83.1 [77.2, 87.2] 82.5 [76.4, 87.5] 83.4 [78.0, 88.2] 0.1371 Male 51.3% (61/119) 47.9% (58/121) 44.7% (55/123) 47.9% (174/363) 0.6062 BMI 27.0 [23.8, 30.5] 27.0 [23.7, 32.1] 26.2 [22.7, 30.4] 26.7 [23.4, 30.8] 0.3103 STS PROM Score 6.6 [4.5, 8.6] 5.6 [3.9, 8.0] 5.8 [3.9, 8.0] 6.0 [4.2, 8.1] 0.0860 History of Atrial Fibrillation

30.3% (36/119) 34.7% (42/121) 30.1% (37/123) 31.7% (115/363) 0.6932

History of PVD 15.1% (18/119) 14.0% (17/121) 16.3% (20/123) 15.2% (55/363) 0.9003 History of CAD 55.5% (66/119) 50.4% (61/121) 53.7% (66/123) 53.2% (193/363) 0.7341 Previous CABG 21.0% (25/119) 18.2% (22/121) 14.6% (18/123) 17.9% (65/363) 0.4289 Previous PCI 16.8% (20/119) 17.4% (21/121) 15.4% (19/123) 16.5% (60/363) 0.9367 History of Diabetes

37.8% (45/119) 40.5% (49/121) 26.8% (33/123) 35.0% (127/363) 0.0577

Previous Stroke2 5.0% (6/119) 4.1% (5/121) 8.1% (10/123) 5.8% (21/363) 0.4376 Previous TIA 6.7% (8/119) 7.4% (9/121) 8.1% (10/123) 7.4% (27/363) 0.9678 Heavily Calcified Aorta

2.5% (3/119) 1.7% (2/121) 3.3% (4/123) 2.5% (9/363) 0.7791

NYHA Class III/IV

82.8% (96/116) 84.9% (101/119) 81.7% (98/120) 83.1% (295/355) 0.8077

Lesion Volume as calculated on FLAIR (mm3)

7916.7 [3865.4, 17315.3]

7377.5 [2562.9, 19181.5]

N/A 7847.9

[3243.2, 17854.5] 0.43063

Echocardiographic Findings Valve Area (cm2) 0.7 ± 0.20 (118) 0.7 ± 0.17 (119) 0.7 ± 0.18 (122) 0.7 ± 0.18 (359) 0.6603 Mean aortic valve gradient (mmHg)

41.0 [33.0, 47.0] 42.7 [33.6, 52.0] 41.0 [31.9, 49.0] 41.0 [33.0, 49.0] 0.3334

Procedural Details Sentinel Device Access

0.4918

Radial N/A 91.2%(104/112) 95.0% (114/119) 93.2% (218/231) Brachial N/A 7.0% (8/112) 4.2% (5/119) 5.6% (13/231) Both Filters Deployed

N/A 92.0% (103/112) 96.6% (115/119) 94.4% (218/231) 0.1570

At Least One Filter Deployed

N/A 99.1% (111/112) 100.0% (119/119) 99.6% (230/231) 0.4848

Procedure Time4 68.0 [41.0,

96.0] 83.5 [54.0,

118.0] 78.0 [52.0,

101.0] 74.0 [52.0, 106.0] 0.0050

Fluoroscopy Time 15.0 [9.0, 20.0] 18.0 [12.0, 29.0] 14.0 [10.0, 27.0] 15.0 [10.0, 26.0] 0.0402 TAVR Device Used 0.7176 Sapien XT 16.9% (20/118) 17.5% (21/120) 19.0% (23/121) 17.8% (64/359) Sapien 3 53.4% (63/118) 55.8% (67/120) 47.9% (58/121) 52.4% (188/359) CoreValve 5.9% (7/118) 2.5% (3/120) 3.3% (4/121) 3.9% (14/359) CoreValve Evolut R

23.7% (28/118) 24.2% (29/120) 29.8% (36/121) 25.9% (93/359)

BMI, Body Mass Index; STS PROM, Society of Thoracic Surgeons Predicted Risk of Operative Mortality; PVD, peripheral vascular disease; CAD, coronary artery disease; CABG, coronary artery bypass graft; PCI, percutaneous coronary intervention; TIA, transient ischemic attack; NYHA, New York Heart Association Note: Continuous data presented as Mean ± SD (n) or Median [IQR]. Categorical data presented using % (n/N). 1 P-values are testing for statistical differences across randomized arms. Continuous data are compared using ANOVA for mean, non-parametric ANOVA for median; categorical data are compared using Fisher's exact test.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

2 Defined as neurological deficit lasting >24 hours confirmed by imaging. 3Based on two-sided Wilcoxon test 4Defined as time from first vascular access puncture to achievement of hemostasis at the TAVR access site.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table 2. Clinical Outcomes

Control Arm Safety + Device Arm p-value 30-day Clinical Outcomes

Any MACCE† 9.9% (11/111) 7.3% (17/234) 0.40

Death (all-cause) 1.8% (2/111) 1.3% (3/234) 0.65

Stroke 9.1% (10/110) 5.6% (13/231) 0.25

Disabling 0.9% (1/109) 0.9% (2/231) 1.00

Non-disabling 8.2% (9/110) 4.8% (11/231) 0.22

AKI (Stage 3) 0% 0.4% (1/231) 1.00

TIA 0% 0.4% (1/231) 1.00

Major Vascular Complication 5.9% (7/119) 8.6% (21/244) 0.53

Radial/Brachial N/A 0.4% (1/244)

Femoral 5.9% (119) 8.2% (20/244) †MACCE defined as Death (any cause), Stroke (any), Acute Kidney Injury (Stage 3)

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table 3. Median total new lesion volume and number of new lesions (unadjusted analysis, Day 2-7)

Control Arm

n=98 Device Arm

n=91 Hodges-Lehmann Estimate of Location Shift (95% CI) P-value

Median Total New Lesion Volume in Protected Territories, mm3

178.0 (34.3, 482.5)

102.8 (36.9, 423.2)

-21.1 (-94.9, 21.8) 0.33451

Median Total New Lesion volume in All Territories, mm3

309.8 (105.5, 859.6) 294 (69.2, 786.4)

-8.6 (-110.7, 68.6) 0.80761

Median Number of New Lesions in Protected Territories

3 (1, 6) 2 (1, 6) 0 (-1, 0) 0.89792

Median Number of New Lesions in All Territories

5 (2, 10) 3 (2, 10) -1 (-2, 1) 0.76672

Data are presented as median (IQR) 1 Based on Wilcoxon Test 2Based on negative binomial regression model

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table 4. Neurocognitive Assessments

Control Arm Device Arm

Mean ± SD (n) Change from

Baseline Mean ± SD (min,

max), n Change from

Baseline P-value1

Attention Baseline -0.17 ± 0.88 (117) NA -0.14 ± 0.96 (117) NA NA

2 to 7 Days Post-TAVR

-0.28 ± 1.1 (65) -0.01 ± 0.59 (65) -0.5 ± 1.02 (65) -0.32 ± 0.7 (65) 0.0334

30 Day Follow-Up

0.03 ± 0.88 (92) 0.14 ± 0.51 (92) -0.14 ± 0.93 (93) 0.03 ± 0.55 (93) 0.1778

90 Day Follow-Up

0.11 ± 0.87 (76) 0.23 ± 0.55 (76) 0.06 ± 0.87 (77) 0.2 ± 0.49 (77) 0.6103

Executive Function Baseline -1.36 ± 1.36 (117) NA -1.28 ± 1.3 (117) NA NA

2 to 7 Days Post-TAVR

-1.36 ± 1.36 (63) 0 ± 1 (63) -1.7 ± 1.55 (65) -0.36 ± 1.19 (65) 0.0865

30 Day Follow-Up

-0.99 ± 1.34 (91) 0.25 ± 0.86 (91) -1.2 ± 1.4 (93) 0.14 ± 0.86 (93) 0.4692

90 Day Follow-Up

-0.79 ± 1.21 (76) 0.39 ± 0.86 (76) -0.94 ± 1.16 (77) 0.32 ± 0.79 (77) 0.4585

Processing Speed Baseline -0.23 ± 0.95 (117) NA -0.24 ± 0.91 (117) NA NA

2 to 7 Days Post-TAVR

-0.47 ± 0.87 (63) -0.05 ± 0.38 (63) -0.51 ± 0.94 (65) -0.06 ± 0.5 (65) 0.9698

30 Day Follow-Up

-0.01 ± 0.86 (90) 0.12 ± 0.39 (90) -0.11 ± 1 (92) 0.14 ± 0.43 (92) 0.5470

90 Day Follow-Up

0.14 ± 0.81 (76) 0.27 ± 0.43 (76) -0.05 ± 0.86 (77) 0.21 ± 0.46 (77) 0.7272

Verbal Memory Baseline -0.64 ± 1.07 (117) NA -0.85 ± 0.94 (117) NA NA

2 to 7 Days Post-TAVR

-1.06 ± 1.06 (66) -0.63 ± 0.79 (66) -1.34 ± 1.3 (66) -0.7 ± 1.03 (66) 0.8534

30 Day Follow-Up

-0.88 ± 1.18 (91) -0.32 ± 0.8 (91) -1.09 ± 1.13 (93) -0.28 ± 0.85 (93) 0.4644

90 Day Follow-Up

-0.61 ± 1.11 (76) -0.13 ± 0.78 (76) -0.86 ± 1.05 (77) -0.02 ± 0.78 (77) 0.2933

Visual Memory Baseline -0.72 ± 0.96 (117) NA -0.83 ± 0.85 (115) NA NA

2 to 7 Days Post-TAVR

-0.7 ± 1.01 (65) 0.06 ± 0.86 (65) -0.87 ± 0.94 (66) -0.19 ± 0.96 (65) 0.1340

30 Day Follow-Up

-1.02 ± 1.03 (92) -0.36 ± 0.79 (92) -1.28 ± 0.94 (93) -0.46 ± 0.91 (92) 0.4282

90 Day Follow-Up

-0.53 ± 0.98 (76) 0.12 ± 0.81 (76) -0.58 ± 0.98 (77) 0.17 ± 0.86 (77) 0.6942

Mental Status2 Baseline 26.07 ± 3.32 (116) NA 26.12 ± 2.95 (114) NA NA

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Control Arm Device Arm

Mean ± SD (n) Change from

Baseline Mean ± SD (min,

max), n Change from

Baseline P-value1

2 to 7 Days Post-TAVR

26.02 ± 2.81 (62) -0.25 ± 2.47 (61) 25.41 ± 3.58 (64) -0.73 ± 2.95 (63) NA

30 Day Follow-Up

26.82 ± 2.74 (89) 0.52 ± 2.55 (89) 26.24 ± 2.84 (92) 0.41 ± 2.67 (91) NA

90 Day Follow-Up

27.24 ± 2.47 (76) 0.96 ± 2.42 (76) 26.56 ± 2.6 (77) 0.3 ± 2.76 (76) NA

Depression2 Baseline 2.7 ± 2.28 (116) NA 3.33 ± 2.62 (114) NA NA

2 to 7 Days Post-TAVR

2.26 ± 2.53 (62) -0.48 ± 1.41 (61) 2.77 ± 2.83 (64) -0.57 ± 2.12 (63) NA

30 Day Follow-Up

2.07 ± 2.14 (89) -0.73 ± 1.57 (89) 2.38 ± 2.43 (91) -0.68 ± 2.02 (90) NA

90 Day Follow-Up

2.37 ± 2.75 (76) -0.49 ± 2.16 (76) 2.53 ± 2.66 (77) -0.75 ± 2.22 (76) NA

Overall Composite Score

Baseline -0.63 ± 0.79 (117) NA -0.66 ± 0.75 (117) NA NA

2 to 7 Days Post-TAVR

-0.81 ± 0.93 (66) -0.16 ± 0.58 (66) -1 ± 0.95 (66) -0.33 ± 0.65 (66) 0.1894

30 Day Follow-Up

-0.59 ± 0.79 (92) -0.03 ± 0.37 (92) -0.77 ± 0.82 (93) -0.09 ± 0.44 (93) 0.4207

90 Day Follow-Up

-0.34 ± 0.72 (76) 0.18 ± 0.35 (76) -0.47 ± 0.76 (77) 0.18 ± 0.38 (77) 0.9409

1P-values based on model adjusted for education, baseline Geriatric Depression Score, and baseline Mini Mental State Score. 2Raw score provided for Mental State and Depression.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Online Supplemental Appendix Cerebral Embolic Protection During Transcatheter Aortic Valve Replacement

Kapadia, Kodali, et al.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPTTable of Contents

List of Study Sites and Investigators

Appendix 1: Description of the MRI acquisition and analyses techniques

Appendix 2: Description of the neurocognitive function domains and methods

Appendix 3: Multivariable modeling methodology

Table 1: Baseline characteristics of the Safety Cohort – ITT, evaluable for MACCE

Table 2: Baseline characteristics of patients with and without paired baseline and 2-7 day MRI

Table 3: Baseline patient characteristics for the paired MR imaging population

Table 4: Baseline patient characteristics for the paired neurocognitive function population

Table 5: MR new lesion volume and number data by valve type (2-7 days)

Table 6: MR new lesion volume (30 days)

Table 7: Univariate and Multivariable analysis of predictors of new lesion volume

Table 8: Multivariable adjustment model new lesion volume

Table 9: Multivariable adjustment model new lesion volume by valve type

Figure 1: Correlation of neurocognitive function and MRI new lesion volume

References

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPTThe SENTINEL Trial Site and Investigator Listing

Cedars-Sinai Medical Center PI: Raj Makkar

Sub-I: Hasanian Al-Jilaihawi Los Angeles, CA United States

Cleveland Clinic

PI: Samir Kapadia Sub-I: Amar Krishnaswamy

Sub-I: E. Murat Tuzcu Sub-I: Stephanie Mick

Cleveland, OH United States

Columbia University Medical Center

PI: Susheel Kodali Sub-I: Tamin Nazif

New York, NY United States

Emory University Hospital Midtown

PI: Vinod Thourani Sub-I: Vasilis Babaliaros

Sub-I: Chandan Devireddy Sub-I: Kreton Mavromatis

Atlanta, GA United States

MedStar Washington Hospital Center

PI: Ron Waksman Sub-I: Lowell Satler

Sub-I: Augusto Pichard Washington, DC

United States

University of Pennsylvania PI: Wilson Szeto

Sub-I: Saif Anwaruddin Sub-I: Prashanth Vallabhajosyula

Sub-I: Jay Giri Sub-I: Howard Herrmann

Philadelphia, PA United States

Barnes - Jewish Hospital

PI: Alan Zajarias Sub-I: John Lasala

St. Louis, MO United States

Henry Ford Health System PI: Adam Greenbaum Sub-I: William O’Neill

Sub-I: Marvin Eng Detroit, MI

United States

Morton Plant Hospital PI: Joshua Rovin Sub-I: Lang Lin

Sub-I: Douglas Spriggs Clearwater, FL United States

Weill Cornell Medical Center

PI: Shing-Chiu Wong Sub-I: Geoffrey Bergman

Sub-I: Arash Salemi New York, NY United States

The University of Texas - Health Science Center

PI: Richard Smalling Sub-I: Biswajit Kar

Sub-I: Pranav Loyalka Houston, TX

United States

University of Virginia Health System PI: D. Scott Lim

Sub-I: Michael Ragosta Charlottesville, VA

United States

University of Washington PI: Mark Reisman

Sub-I: James McCabe Sub-I: Creighton Don

Seattle, WA United States

Icahn School of Medicine at Mount Sinai

PI: Samin Sharma Sub-I: Annapoorna Kini Sub-I: George Dangas

New York, NY United States

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Sentara Norfolk General Hospital PI: Paul Mahoney

Norfolk, VA United States

St. Thomas Hospital

PI: Andrew Morse Sub-I: Mark Stankewicz Sub-I: Evelio Rodriguez

Nashville, TN United States

Herzzentrum Leipzig GmbH -Universitätsklinik

PI: Axel Linke Sub-I: Norman Mangner

Sub-I: Felix Woitek Leipzig

Germany

Asklepios Klinik St. Georg PI: Christian Frerker

Hamburg Germany

St. Luke's Hospital

PI: David Cohen Kansas City, MO

United States

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Appendix 1: MRI acquisition and analysis methodology

STUDY PROCEDURES

Magnetic Resonance Imaging

Brain MRI assessments were performed at baseline, 2-7 days and 30 days post procedure. MRI scans were obtained

according to a protocol provided by the MRI reading center (Buffalo Neuroimaging Analysis Center, Buffalo, NY,

USA) that also performed all sequential co-registration of MRI scans in a blinded manner. MR images at each site

were acquired only on a 3 Tesla certified and validated system at 0, 2-7 and 30 days. Diffusion weighted images

(DWI) were acquired with a 2D echo planar sequence with one b=0 image and 3 orthogonal diffusion directions

with b=1000 s/mm1. Additional parameters were: repetition time (TR) = 13000ms, echo time (TE) = 100ms, slice

thickness = 3mm (no gap), acquisition matrix 204 x 156, final voxel size = 1.25mm x 1.25mm x 3.0mm. The DWI

images were required at baseline and 2-7 days post procedure on all evaluable imaging cohort patients. Fluid

attenuated inversion recovery (FLAIR) images were acquired with a 2D spin echo inversion recovery sequence with

an inversion time (TI) of 2580ms. Additional parameters were: TR = 9730ms, TE=92ms, slice thickness = 2mm (no

gap), acquisition matrix 256 x 186, final voxel size = 0.94mm x 1.17mm x 2.0mm. The FLAIR images were

required at baseline and 30 days post procedure on all evaluable imaging cohort patients. High resolution T1-

weighted images (hires-T1) were acquired with an MP-RAGE sequence. Additional parameters were: TR = 1690ms,

TE=2.57ms, flip angle (FA) = 12, TI=1100ms, slice thickness = 1.5mm (no gap), acquisition matrix 256 x 224, final

voxel size = 1.00mm x 1.00mm x 1.5mm. Finally, either a manufacturer-based dual-echo GRE sequence was used to

acquire B0 field maps (voxel size = 4.00mm x 4.00mm x 5.00mm.), or the DWI images were acquired with two

different phase encoding directions. Minor site-specific deviations were allowed to accommodate individual scanner

capabilities, provided they were approved by MR physicists at the reading center and were acquired consistently

within the site.

DWI acquisitions are subject to substantial artifacts, including eddy current distortions, susceptibility-induced

warping, and signal dropout. Although these do not have a substantial impact on clinical assessment of large lesions

associated with stroke or transient ischemic attack (TIA), they are quite large relative to the small embolic lesions

resulting from the TAVI procedure – distortions may easily be on the order of 1cm, while lesions may be as small as

a few mm. Therefore, a number of pre-processing steps were taken to improve image quality and subsequent

analysis. First, the raw DWI images were corrected for distortions using FMRIB’s FSL FDT library.2 This was

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

accomplished using either directly acquired fieldmaps or by inferring the field map from paired, phase-reversed

DWI acquisitions.3 Next, the diffusion b=0 (b0) and three corrected b=1000 diffusion-encoded raw images were

combined to create trace and apparent diffusivity coefficient (ADC) images.

Because the lesions are often so small, subtraction imaging was also employed to increase lesion salience.4 Baseline

DWI and FLAIR images were voxel-wise subtracted from follow-up images to produce direct change maps. To

facilitate this subtraction approach, additional pre-processing steps were performed. First, low-frequency spatial

intensity inhomogeneities on FLAIR images were corrected using N3.5 Corrected FLAIR and DWI trace images

were further standardized by applying a piecewise-linear histogram adjustment method to compensate for scan-to-

scan variability in absolute intensity.6 Finally, to facilitate direct longitudinal analysis, all within-subject scans were

co-registered to each subject’s baseline FLAIR image using FLIRT with 6 degrees of freedom.7

Lesions were delineated on corrected and aligned 2-7, and 30 day DWI trace images and 30 day FLAIR images

using a semi-automated contouring technique provided by the JIM software package, with simultaneous reference to

the ADC and subtraction images.8 Using this approach, a trained operator identified lesions individually, and for

each lesion an assistive algorithm delineated a highly reproducible iso-contour at the maximum local gradient. The

operator viewed all images and change maps simultaneously to increase confidence, and also coded lesions as new

or persistent.

In addition to lesion counts and volumetry, vascular territory was also assessed using an atlas-based technique. For

this purpose, a vascular territory atlas was manually created in the standard MNI 152 template space9 based on

existing literature,10 and including 28 separate regions. Individual hires-T1 images were used to non-linearly align

this atlas to individual lesion maps. First, individuals’ hires-T1 images were corrected for intensity inhomogeneity

using N3, then aligned to the MNI 152 template using a two-stage process consisting of an initial rigid-body co-

registration followed by composition with a warp field obtained from a non-linear warping technique.11 These

transforms were then inverted, and applied to the original atlas. Lesion number and volume within each vascular

territory were than assessed separately.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Appendix 2: Neurocognitive test battery The Stroke and Cerebrovascular Disease Division, Department of Neurology, Columbia University Medical Center designed a neurocognitive test battery tailored to the presumed mechanisms of cognitive impairment in the setting of transcatheter aortic-valve implantation (TAVI) and the deployment of this investigational cerebral protection device: perfusion failure in the cerebral circulation and microembolism to the brain. In the former case, cognitive dysfunction is global (bi-hemispheral) in nature, producing deficits in processing speed, executive function, and attention/concentration12. In the instance of microembolism, these particles are too small to cause the branch occlusions needed for focal neurocognitive syndromes, such as aphasia (left hemisphere) or visual neglect (right hemisphere).13 Rather, there is a diffuse impact that can be either bi-hemispheral or within a single hemisphere across a wide swath of brain territory in the MCA and ACA distribution territories.12 A weakness in prior studies examining the neurocognitive impact of TAVI is the lack of specificity and sensitivity in the tests used. Either measures are too brief (e.g. the Mini Mental State Exam, or MMSE)14, or have not been validated in the setting of diffuse vascular disease (e.g., RBANS)15, raising concern about Type II error. The battery employed in this study, and its variants, have been used in both NIH-funded (NHLBI, NINDS) and industry-sponsored studies, assessing neurocognitive sequelae in end-stage heart failure and LVAD support16, 17, carotid artery disease18-20, and recently piloted in the PARTNER 2 trial to establish feasibility in the target demographic. All tests are standardized, well-validated instruments in the literature21, and provide extensive normative data permitting baseline analysis and change over time, at 30 and 90 days post implantation. Both efficacy (improvement) and safety (worsening) can be derived from these measures. The MMSE was given only to establish comparability to prior work and to serve as a covariate for baseline mental status. Depression is also a covariate.

Neurocognitive Test Laterality Domain Trails A Bi-hemispheral Attention Trails B Bi-hemispheral Executive Function Digit Span Bi-hemispheral Attention Digit Symbol Bi-hemispheral Processing Speed Letter-Number Sequencing Bi-hemispheral Working Memory Controlled Oral Word Association Left Hemisphere Processing Speed Hopkins Verbal Learning Test Left Hemisphere Verbal Memory Rey Complex Figure (Copy) Right Hemisphere Executive Function Brief Visual Memory Test Right Hemisphere Visual Memory Mini Mental State Exam --- Mental Status Geriatric Depression Scale --- Depression Developed by Ronald M. Lazar, PhD, FAHA, FAAN, Professor and Marykay Pavol, PhD, ABCN, Assistant Professor, Doris & Stanley Tananbaum Stroke Center Neurological Institute of New York

Primary Neurocognitive Analysis: Comparison of the change in composite neurocognitive z-scores from Baseline to 30-days post-TAVR between the group in whom the Sentinel device was used and the group that did not receive distal protection, controlling for MMSE and the depression scores.

Composite Score Calculation: A z-score for each domain was calculated based on the normative means and standard deviations for each neurocognitive test supplied by the Neurocognitive Core Lab at Columbia. These norms were stratified by age and education (when possible). When there is more than one test for a given domain (e.g., Trails A and Digit Span for “Attention”), an average was computed from the z-scores comprising the tests for that domain. When there is more than one outcome for a given test (e.g., Total Recall, Delayed Recall and Recognition for “Verbal Memory”), a mean z-score was derived from these outcomes. The composite

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

neurocognitive z-score for each treatment group is the average z-score from all domains (Attention, Executive Function, Processing Speed, Verbal Memory, Visual Memory). Change scores were calculated (by domain) by subtracting post-surgical exam scores from the baseline scores.

Secondary Neurocognitive Analyses:

1. The interaction of the 30-day composite neurocognitive change score for each treatment group with lesion outcomes from post-TAVR DWI, 30-Day FLAIR, and Baseline FLAIR.

2. Change in neurocognitive composite scores from baseline to 7-days and baseline to 90-days post-TAVR for each treatment group.

3. Baseline composite neurocognitive scores for all study participants to characterize pre-TAVR cognitive function and correlations between these baseline composite scores and baseline FLAIR imaging variables.

4. Change in individual domain scores from Baseline to 7-days, 30-days and 90-days post-TAVR between the group in whom the Sentinel device was used and the group that did not receive distal protection, and the relationships among these domain scores to Baseline FLAIR, post-TAVR DWI and 30-Day FLAIR imaging variables.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Appendix 3. Multivariable model methodology

Introduction During the time of design and initiation of the Sentinel trial, only one TAVR device was commercially available in the U.S. However, several additional TAVR devices became available during the course of the trial and were not stratified in the randomization. In addition, given that this is a first-in-kind study and the data evidenced significant skew and large variance, it was possible that a key baseline that was not pre-identified. This supplement will describe the steps taken to address these concerns in a post-hoc analysis.

Analysis Population The ITT population used in this analysis are all randomized patients that have had a MRI follow-up post TAVR irrespective of whether the follow-up falls within the 2-7 day windows as specified by the protocol.

Normalizing the Outcome: New Volumes in Protected or All Territories Post TAVR Procedure A log transformation was required to normalize the skewed end point lesion volume data, baseline T2-FLAIR volumes, and baseline HU-850 aortic valve calcification measures. The high frequency of zero values required a correction to the logarithmic transformation. Zero values were set to 5 mm3, the midpoint of zero the MRI detection limit of 10 mm3. In the rest of this supplement, the prefix LNZ will indicate that the data has been transformed as described in this section. Figure S3-1 shows the distribution of the LNZ transformed lesion volumes in protected and all territories.

Figure S3-1: Histogram of New Volumes on the Log Base E scale where zero volumes are set to 5 mm3 on the arithmetic scale and Log Base E[5] on the log scale.

Identifying Key Baseline Covariates A comprehensive univariate analysis of all baseline characteristics (characteristics that existed before the TAVR procedure) was performed with the baseline characteristic as predictor and the transformed new volumes in protected territories as the outcome. Table S3-1 provides the list of baseline characteristics that were tested.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table S3-1: List of Baseline Characteristics used for Univariate analyses

Characteristic Characteristic Characteristic

Acute Success [Y/N] Distal Vessel Diameter Pre-Dilation [Y/N]

Age Dual Antiplatelet Therapy [Y/N] Procedure Time

Antithrombotic Medications [Y/N] Education Proximal Vessel Diameter

Any Adverse Event(s) during the procedure

Ethnicity Race

Any Diabetes [Y/N] General anesthesia used [Y/N] Region[USA/Germany]

Any other clinically significant infections, allergies, diseases, or surgeries [Y/N]

Heart Rate Respiration

AVA Indexed Height (inches) Rosenhek2

Baseline Geriatric Depression Score History of Atrial Fibrillation [Y/N] Sex [Female/Male]

Baseline Modified Rankin Scale History of Peripheral vascular disease [Y/N]

STS

Baseline Neurocognitive z Overall lnz[baseline volumes] Systole Blood Pressure

BMI lnz[HU_850Volume] TAVR Access

BSA Mean aortic valve area Temperature

Center Modified Allen Result [Normal/Abnormal]]

TIA [Y/N]

Coronary Artery Disease [Y/N] Neurological Deficit [Y/N] Weight (lbs)

CT Result[Normal/Abnormal] NIH Stroke Scale

Currently Taking Medications [Y/N] Peak Aortic Jet Velocity

Diastole Blood Pressure Porcelin Aorta [Y/N]

Dilatation[Yes/No] Post Dilatation

Baseline T2-FLAIR and HU850 Volume measurements were transformed using the LNZ transformation previously described.

Any baseline characteristic with a p-value less than or equal to 0.1 (see Table S3-1 for a list of characteristics tested) were included in a step-wise multi-variable regression model (

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table S3-2). The step wise regression model minimized the AIC and variables entered the model if they had a p-value <=0.1 and exited the model if the p-value was > 0.05. The results of the forward and backward stepwise regression model is shown in Table S3-3. Only LNZ[Baseline T2-FLAIR Volumes] was retained out of all the variables listed in

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table S3-2. Only the LNZ[Baseline T2-FLAIR Volumes] was predictive of LNZ[New Volumes in Protected Territories].

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table S3-2: Results of Univariate Analysis of Baseline Characteristics predicting transformed New Volumes in Protected Territories

Source Pr > F Source Pr > F Baseline T2-FLAIR Volume Strata (High/Low)

<.0001 Weight (lbs) 0.0568

LNZ[Baseline Lesion Volume] <.0001 Neurological Deficit [Y/N] 0.0594LNZ[Baseline T2-FLAIR Lesion Number]

0.0024 Currently taking medications (Y/N) 0.0632

LNZ[HU_850 Volume] 0.009 Peak Aortic Jet Velocity 0.0636Age 0.0155 STS Risk score 0.0643Dilitation Post TAVR Procedure 0.034 Rosenhek2 0.071Temperatue 0.0495 Neurocognitve Verbal Memory Score 0.078BSA 0.0517 History of Percutaneous Coronary Intervention (Y/N) 0.0884

Table S3-3: Results of step wise regression model

Parameter DF Estimate StandardError

t Value P-Value

Intercept 1 0.88 1.041516 0.85 0.39.

LNZ[Baseline T2-FLAIR Volumes] 1 0.44 0.116631 3.83 0.0002

Error! Reference source not found.3- 2 shows a scatter plot of Post Procedure New Volumes in Protected Territories vs. Baseline T2-FLAIR Volumes on a log base 10 scale (zero values are set to 1). There is a clear relationship between the two. However, there is a great deal of variability. Therefore, to reduce the additional variability that might be added to the model if one uses the continuous data, the baseline data was transformed into a binary variable, volumes above the median of the pooled baseline data (high) or below the median of the baseline data (low).

The covariate assessment was repeated for the new volumes in all territories with the same results.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Figure S3-2: Post Procedure New Volumes in Protected Territories vs. Baseline T2-FLAIR Volumes

Building the model to assess treatment effects In the previous sections, baseline T2-FLAIR volume was identified as the key baseline covariate. Further, the randomization was not stratified by TAVR device. Potential differences due to physician selection preferences or design differences may affect the outcome and are of scientific interest in the burgeoning TAVR market, therefore the TAVR Device was added to the model as described below.

~ + + +

Where Yijk is either the LNZ[New Volumes in Protected Territories] or LNZ[New Volumes in All Territories]

In order to get stable TAVR device estimates from the model, Medtronic CoreValve which only contributed 7 patients with imaging follow-up, were excluded from the analysis.

Results The p-values in reported in this section represent the adjusted p-values within each ANOVA. The results of the analysis of variance for LNZ[New Volumes in Protected Territories] are in Table S4 and Figure S3 1. Once the model is adjusted for baseline T2-FLAIR volumes, TAVR device and the interaction between TAVR device and treatment arm, treatment arm as a main effect is shown to be a significant predictor of the outcome (p=0.025).

It should be noted that there is a significant interaction effect between treatment arm and TAVR device. The overall treatment effect, after adjustment for TAVR Device, and the interaction between TAVR and Treatment, is a 42% reduction in post procedure new lesion volume in protected areas [see Supplemental Table 6 following this appendix). However, SAPIEN 3 generated the lowest post procedure new lesion volume (30% - 50% lower than the other TAVR devices; see Supplemental Table 6). Therefore, SAPIEN 3 derived the least benefit from the use of the Sentinel device resulting in little to no difference between the treatment arms. This is responsible for the significant interaction effect between the device and treatment. Unfortunately, since the randomization was not stratified by TAVR device, the low new lesion volumes in the S3 TAVR device could be due to a number of reasons including the influence of the use of clinician judgement when matching device to patient.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

This interaction effect may be due to unknown factors associated with clinician judgement as to which device to use at the time of the procedure. Therefore, while the impact of the treatment and its interaction with TAVR device on the outcome can be surmised from this model, the size of the treatment effect requires a more rigorous design that controls for baseline volume and randomizes the TAVR device to the patient.

Table S4-3: Results of Analysis of Variance Using LNZ[New Volumes in Protected Territories]

Effect Numerator DF Denominator DF F Value P-value

BL T2-FLAIR Lesion Volume (High/Low) 1 175 19.08 <.0001

Valve Type 2 175 1.81 0.1666

Arm 1 175 5.13 0.0248

Valve Type*Arm 2 175 4.24 0.0159

Figure S3 1: Diagnostic Plot for the Analysis of Variance for LNZ[New Volumes in Protected territories]

Residual by Predicted Plot

-

Sensitivity analyses The analysis of variance for both the LNZ[New Volumes in Protected Territories] and LNZ[New Volumes in All Territories] was repeated with using the LNZ[Baseline Volumes] as a continuous variable. The overall results were the same except the p-value for treatment arm increased to 0.041 and 0.078 for LNZ[New Volumes in Protected Territories] and LNZ[New Volumes in All Territories] respectively.

The analysis of variance was also repeated using the ranks of the New Volumes in Protected or All territories as outcome, essentially making the outcome a uniform distribution. The p-values for the treatment arm was the similar, 0.031 and .066 for LNZ[New Volumes in Protected Territories] and LNZ[New Volumes in All Territories] respectively.

-4

-2

0

2

4

2 4 6 8 10

Resid

ual

Predicted

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Table 1. Baseline characteristics of the Safety Cohort – ITT, evaluable for MACCE

Safety+Test

(N=234) Control (N=111)

Total (N=345) P-value1

Age - years 83.1 [77.2, 87.5] n=234

85.0 [78.5, 89.0] n=111

83.8 [78.0, 87.9] n=345

0.0896

Male 47.4% (111/234) 49.5% (55/111) 48.1% (166/345) 0.7305

BMI 26.4 [23.2, 30.9] n=231

27.1 [24.0, 30.6] n=109

26.9 [23.4, 30.8] n=340

0.6580

STS PROM Score 5.8 [4.0, 8.0] n=234

6.4 [4.5, 8.6] n=111

6.0 [4.2, 8.1] n=345

0.0565

History of Atrial Fibrillation

33.3% (78/234) 31.5% (35/111) 32.8% (113/345) 0.8063

History of PVD 15.4% (36/234) 14.4% (16/111) 15.1% (52/345) 0.8732

History of CAD 52.1% (122/234) 55.0% (61/111) 53.0% (183/345) 0.6456

Previous CABG 17.1% (40/234) 21.6% (24/111) 18.6% (64/345) 0.3737

Previous PCI 16.7% (39/234) 16.2% (18/111) 16.5% (57/345) 1.0000

History of Type II Diabetes

32.5% (76/234) 38.7% (43/111) 34.5% (119/345) 0.2761

Previous Stroke2 5.6% (13/234) 1.8% (2/111) 4.3% (15/345) 0.1577

Previous TIA 8.1% (19/234) 7.2% (8/111) 7.8% (27/345) 0.8336

Heavily Calcified Aorta 2.1% (5/234) 1.8% (2/111) 2.0% (7/345) 1.0000

NYHA Class III/IV 83.5% (192/230) 82.4% (89/108) 83.1% (281/338) 0.9077

Echocardiographic Findings

Valve Area (cm2) 0.7 ± 0.17 (231) 0.7 ± 0.20 (110) 0.7 ± 0.18 (341) 0.9542

Mean aortic valve gradient (mmHg)

41.5 [33.0, 50.0] n=229

41.0 [32.0, 47.0] n=110

41.0 [32.4, 49.0] n=339

0.3564

Note: Continuous data presented as Mean ± SD (n) or Median [IQR] (n). Categorical data presented using % (n/N). 1 P-values are testing for statistical differences across randomized arms. Continuous data are compared using ANOVA; categorical data are compared using Fisher's exact test. 2 Defined as neurological deficit lasting more than 24 hours confirmed by imaging.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Table 2. Baseline characteristics of patients with and without paired baseline and 2-7 day MRI

Observed MRI

(N=189) Missing MRI

(N=51) Total

(N=240) P-value1

Age - years 84.1 [78.5, 87.6] (189)

83.1 [75.4, 89.4] (51)

83.8 [78.1, 88.3] (240)

0.5216

Male 47.6% (90/189) 56.9% (29/51) 49.6% (119/240) 0.2712

BMI 27.0 [23.7, 30.8] (186)

27.7 [24.5, 32.1] (50)

27.0 [23.8, 31.0] (236)

0.2921

STS PROM Score 6.0 [4.4, 8.0] (189) 6.9 [3.8, 9.3] (51) 6.0 [4.2, 8.2] (240) 0.6404

History of Atrial Fibrillation

33.3% (63/189) 29.4% (15/51) 32.5% (78/240) 0.7364

History of PVD 14.8% (28/189) 13.7% (7/51) 14.6% (35/240) 1.0000

History of CAD 53.4% (101/189) 51.0% (26/51) 52.9% (127/240) 0.7550

Previous CABG 23.3% (44/189) 5.9% (3/51) 19.6% (47/240) 0.0048

Previous PCI 15.9% (30/189) 21.6% (11/51) 17.1% (41/240) 0.4012

History of Type II Diabetes

37.6% (71/189) 41.2% (21/51) 38.3% (92/240) 0.6309

Previous Stroke2 3.2% (6/189) 9.8% (5/51) 4.6% (11/240) 0.0590

Previous TIA 7.4% (14/189) 5.9% (3/51) 7.1% (17/240) 1.0000

Heavily Calcified Aorta 2.1% (4/189) 2.0% (1/51) 2.1% (5/240) 1.0000

NYHA Class III/IV 83.7% (154/184) 84.3% (43/51) 83.8% (197/235) 0.2825

Lesion Volume as calculated on FLAIR (mm3)

7783.6 [2982.1, 18224.4] (189)

10285.9 [4638.0, 17038.5] (39)

7847.9 [3243.2, 17854.5] (228)

0.3318

Echocardiographic Findings

Valve Area (cm2) 0.7 ± 0.18 (187) 0.7 ± 0.18 (50) 0.7 ± 0.18 (237) 0.2974

Mean aortic valve gradient (mmHg)

40.6 [32.0, 47.7] (186)

44.5 [40.0, 56.0] (50)

41.1 [33.5, 49.2] (236)

0.0029

Note: Continuous data presented as Mean ± SD (n) or Median [IQR] (n). Categorical data presented using % (n/N). 1 P-values are testing for statistical differences across randomized arms. Continuous data are compared using ANOVA for mean or two-sided Wilcoxon text for median; categorical data are compared using Fisher's exact test. 2 Defined as neurological deficit lasting more than 24 hours confirmed by imaging.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Table 3. Baseline and procedural characteristics of patients with paired baseline and 2-7 day MRI – ITT, complete case population

Control Arm

(N=98) Device Arm

(N=91) Total

(N=189) P-value1

Age 85.0 [79.7, 89.0] (98)

83.2 [77.8, 87.2] (91)

84.1 [78.5, 87.6] (189)

0.1397

Male 50.0% (49/98) 45.1% (41/91) 47.6% (90/189) 0.5605

BMI 27.0 [23.7, 30.6] (96)

26.8 [23.7, 31.8] (90)

27.0 [23.7, 30.8] (186)

0.5113

STS PROM Score 6.4 [4.6, 8.0] (98) 5.6 [4.2, 8.0] (91) 6.0 [4.4, 8.0] (189) 0.2999History of Atrial Fibrillation

32.7% (32/98) 34.1% (31/91) 33.3% (63/189) 0.8780

History of PVD 16.3% (16/98) 13.2% (12/91) 14.8% (28/189) 0.6826

History of CAD 55.1% (54/98) 51.6% (47/91) 53.4% (101/189) 0.6635

Previous CABG 22.0% (20/91) 24.5% (24/98) 23.3% (44/189) 0.7324

Previous PCI 16.5% (15/91) 15.3% (15/98) 15.9% (30/189) 0.8447

History of Diabetes 36.7% (36/98) 39.6% (36/91) 38.1% (72/189) 0.7647

Previous Stroke2 3.1% (3/98) 3.3% (3/91) 3.2% (6/189) 1.0000

Previous TIA 6.1% (6/98) 8.8% (8/91) 7.4% (14/189) 0.5827

Heavily Calcified Aorta 2.0% (2/98) 2.2% (2/91) 2.1% (4/189) 1.0000

NYHA Class III/IV 80.0% (76/95) 87.6% (78/89) 83.7% (154/184) 0.1696

Lesion Volume as calculated on FLAIR (mm3)

7800.1 [3248.4, 17315.3]

7551.6 [2127.8, 19875.6]

7783.6 [2982.1, 18224.4]

0.5881

Echocardiographic Findings

Valve Area (cm2) 0.7 ± 0.17 (89) 0.7 ± 0.20 (98) 0.7 ± 0.18 (187) 0.5969

Mean aortic valve gradient (mmHg)

40.0 [31.8, 46.5] (98)

41.0 [32.3, 50.2] (88)

40.6 [32.0, 47.7] (186)

0.2929

Note: Continuous data presented as Mean ± SD (n) or Median [IQR] (n). Categorical data presented using % (n/N). 1 P-values are testing for statistical differences across randomized arms. Continuous data are compared using ANOVA or two-sided Wilcoxon test; categorical data are compared using Fisher's exact test. 2 Defined as neurological deficit lasting more than 24 hours confirmed by imaging.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Table 4. Baseline characteristics of patients with paired baseline and 30 day neurocognitive assessment

Control Arm

(N=92) Device Arm

(N=93) Total

(N=185) P-value1

Age 84.8 [78.4, 89.0] (92)

83.1 [77.8, 87.2] (93)

83.8 [78.1, 88.3] (185)

0.2585

Male 50.0% (46/92) 49.5% (46/93) 49.7% (92/185) 1.0000

BMI 27.3 [24.5, 31.1] (90)

27.0 [23.7, 32.6] (92)

27.1 [24.1, 31.8] (182)

0.7377

STS PROM Score 6.4 [4.5, 8.5] (92) 5.3 [3.9, 7.9] (93) 6.0 [4.2, 8.0] (185) 0.1214 History of Atrial Fibrillation

31.5% (29/92) 31.2% (29/93) 31.4% (58/185) 1.0000

History of PVD 12.0% (11/92) 15.1% (14/93) 13.5% (25/185) 0.6679

History of CAD 55.4% (51/92) 50.5% (47/93) 53.0% (98/185) 0.5568

Previous CABG 22.8% (21/92) 19.4% (18/93) 21.1% (39/185) 0.5927

Previous PCI 12.0% (11/92) 21.5% (20/93) 16.8% (31/185) 0.1145

History of Diabetes 39.1% (36/92) 41.2% (39/93) 40.5% (75/185) 0.7651

Previous Stroke 1.1% (1/92) 3.2% (3/93) 2.2% (4/185) 0.6209

Previous TIA 5.4% (5/92) 8.6% (8/93) 7.0% (13/185) 0.5669

Heavily Calcified Aorta 1.1% (1/92) 1.1% (1/93) 1.1% (2/185) 1.0000

NYHA Class III/IV 80.9% (72/89) 84.6% (77/91) 82.8% (149/180) 0.5574

Echocardiographic Findings

Valve Area (cm2) 0.7 ± 0.21 (91) 0.7 ± 0.17 (91) 0.7 ± 0.19 (182) 0.7162

Mean aortic valve gradient (mmHg)

40.8 [33.0, 47.0] (91)

41.5 [32.3, 53.0] (90)

41.0 [32.4, 49.0] (181)

0.4658

Note: Continuous data presented as Mean ± SD (n) or median [IQR] (n). Categorical data presented using % (n/N). 1 P-values are testing for statistical differences across randomized arms. Continuous data are compared using ANOVA or two-sided Wilcoxon test; categorical data are compared using Fisher's exact test. 2 Defined as neurological deficit lasting more than 24 hours confirmed by imaging.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Table 5. DW-MRI Median new lesion volume and number of new lesions by TAVR device type (unadjusted analysis, Day 2-7)

Control Arm Device Arm P-value1

Median New Lesion Volume (mm3), Protected Territories

SAPIEN XT 322 (31, 785.7), n=14 55.4 (0, 181.9), n=15 0.0455

SAPIEN 3 143.4 (33, 233.3), n=56 104.7 (50.1, 384.3), n=52 0.4454

CoreValve Evolut R 324.3 (141.1, 1077.8), n=24 147.7 (18.5, 479.9), n=21 0.1185

Median New Lesion Volume (mm3), All Territories

SAPIEN XT 518.1 (134.5, 1170.7), n=14 114 (29, 321.7), n=15 0.0341

SAPIEN 3 193.5 (59.3, 464.1), n=56 265 (131.8, 962.1), n=52 0.0516

CoreValve Evolut R 807.2 (326.6, 1716.5), n=24 369.1 (63.3, 786.4), n=21 0.0465

Median New Lesion Number, Protected Territories

SAPIEN XT 3.5 (1, 9), n=14 2 (0, 3), n=15 0.0347

SAPIEN 3 2 (1, 4), n=56 2 (1, 6.5), n=52 0.0425

CoreValve Evolut R 5.5 (2.5, 11.5), n=24 2 (1, 6), n=21 0.0189

Median New Lesion Number, All Territories

SAPIEN XT 7 (1, 12), n=14 2 (1, 5), n=15 0.0300

SAPIEN 3 4 (1.5, 6), n=56 3 (2, 10), n=52 0.0124

CoreValve Evolut R 8.5 (5.5, 15.5), n=24 3 (1, 12), n=21 0.0346

Note: Medtronic CoreValve (which only contributed 7 patients with imaging follow-up) was excluded from this analysis. 1 Lesion volume analyses based on Wilcoxon test; lesion number analyses based on negative binomial regression model

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Table 6. 30-day median total new lesion volume (unadjusted analysis)

Test Arm

(n=78) Control Arm

(n=80)

Observed Treatment Difference

(test - control)

Hodges-Lehmann Estimate of

Location Shift (95% CI) P-value1

Protected Territories ITT, mm3

0 (0, 52.7) 0 (0, 83.1) 0.0 0 (0, 0) 0.8255

All Territories ITT, mm3

0 (0, 73.8) 0 (0, 108.1) 0.0 0 (0, 0) 0.8507

Data are presented as median (IQR) 1 Based on Wilcoxon Test

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Table 7. Significant Results of (A) Univariate Analysis and (B) Multivariable Analysis of Baseline Characteristics Predicting Transformed New Volumes in Protected Territories

Source Pr > F Source Pr > F Baseline Volume Strata (High/Low) <.0001 Weight (lbs) 0.0568LNZ[Baseline Lesion Volume] <.0001 Neurological Deficit [Y/N] 0.0594LNZ[Baseline Lesion Number] 0.0024 Currently taking medications (Y/N) 0.0632LNZ[HU_850 Volume] 0.009 Peak Aortic Jet Velocity 0.0636Age 0.0155 STS Risk score 0.0643Dilitation Post TAVR Procedure 0.034 Rosenhek2 0.071Temperatue 0.0495 Neurocognitve Verbal Memory Score 0.078BSA 0.0517 History of Percutaneous Coronary Intervention (Y/N) 0.0884A. Univariate Analysis

Parameter DF Estimate StandardError

t Value P-Value

Intercept 1 0.88 1.041516 0.85 0.39.

Baseline T2-FLAIR Volume 1 0.44 0.116631 3.83 0.0002B. Multivariable Analysis

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Table 8. Least Squares Means and Test of Main Effects (adjusted by Interaction of Valve Type and Treatment Arm, and High/Low BL Lesion Volume) – ITT Population

Effect Mean Estimate(95% CI) p-value

Least Squares Mean new lesion volume (mm3), Protected territories

Control Arm 162.8(107.9, 245.5)

0.0248

Device Arm 83.3(55.0, 126.1)

Least Squares Mean new lesion volume (mm3), all territories

Control Arm 311.1(212.2, 456.3)

0.0500

Device Arm 180.6(122.7, 265.8)

Least Squares Mean new lesion volume (mm3), Protected territories

Evolut R (n=45) 172.6

(103.7, 287.2)

0.166 SAPIEN 3 (n=108) 113.6

(81.8, 157.8)

SAPIEN XT (n=29) 80.5

(42.6, 152.0)

Least Squares Mean new lesion volume (mm3), all territories

Evolut R (n=45) 408.0

(253.9, 655.6)

0.021 SAPIEN 3 (n=108) 228.2

(168.0, 309.9)

SAPIEN XT (n=29) 143.1

(79.1, 258.7)

Note: In order to get stable TAVR device estimates from the model, Medtronic CoreValve (which only contributed 7 patients with imaging follow-up) was excluded from the analysis.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Table 9. Least Squares Means and Test of Interaction Effects (Arm within Valve Type adjusted by Treatment Arm, Valve Type and and High/Low BL Lesion Volume) – ITT, Lesion Volume

Effect Control Arm

Mean Estimate (95% CI)

Device Arm Mean Estimate

(95% CI) p-value*

Least Squares Mean new lesion volume (mm3), Protected territories

Evolut R (n=45) 279.1(139.2, 559.4)

106.7 (50.7, 224.7) 0.0642

SAPIEN XT (n=29) 162.7(65.3, 405.4)

39.8 (16.5, 96.1) 0.0296

SAPIEN 3 (n=108) 95.0(60.2, 149.9)

135.8 (84.7, 217.9) 0.2834

Least Squares Mean new lesion volume (mm3), all territories

Evolut R (n=45) 691.0(361.6, 1320.4)

240.9 (120.4, 481.7)

0.0297

SAPIEN XT (n=29) 268.1(114.5, 627.7)

76.3(33.6, 173.4)

0.0369

SAPIEN 3 (n=108) 162.5(106.2, 248.7)

320.3 (206.3, 497.4)

0.0299

Note: In order to get stable TAVR device estimates from the model, Medtronic CoreValve (which only contributed 7 patients with imaging follow-up) was excluded from the analysis. *No adjustment for multiple comparisons was made.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Supplemental Figure 1. Correlation between neurocognitive score and MRI lesions

a. Protected territories

r = -0.2028

p = 0.0109

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

b. All territories

r = -0.2356

p = 0.003

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

References

1. Smith, S., Jenkinson, M., & Woolrich, M. Advances in functional and structural MR image analysis and

implementation as FSL. Neuroimage. 2004

2. Jezzard, P., & Balaban, R. Correction for geometric distortion in echo planar images from B0 field variations.

Magnetic Resonance in Medicine. 1995;34(1), 65–73.

3. Andersson, J.L.R., Skare, S., Ashburner, J. How to correct susceptibility distortions in spin-echo echo-planar

images: application to diffusion tensor imaging. NeuroImage, 20(2):870-888, 2003.

4. Liguori, M., Meier, D. S., Hildenbrand, P., Healy, B. C., Chitnis, T., Baruch, N. F., Guttmann, C. R. G. (2011).

One year activity on subtraction MRI predicts subsequent 4 year activity and progression in multiple sclerosis.

Journal of Neurology, Neurosurgery, and Psychiatry. 2011;82(10), 1125–31. doi:10.1136/jnnp.2011.242115

5. Sled, J. G., Zijdenbos, A. P., & Evans, A. C. A nonparametric method for automatic correction of intensity

nonuniformity in MRI data. IEEE Transactions on Medical Imaging. 1998;17(1), 87–97. doi:10.1109/42.668698

6. Nyul, L., Udupa, J., & Zhang, X. New variants of a method of MRI scale standardization. Medical Imaging,

IEEE. 2000;19(2), 143–150.

7. Jenkinson, M., & Smith, S. A global optimisation method for robust affine registration of brain images. Medical

Image Analysis, 2001;5(2), 143–156.

8. Tavazzi, E., Dwyer, M. G., Weinstock-Guttman, B., Lema, J., Bastianello, S., Bergamaschi, R., Munschauer, F.

E. Quantitative diffusion weighted imaging measures in patients with multiple sclerosis. Neuroimage,

2007;36(3):746–54. doi:10.1016/j.neuroimage.2007.03.056

9. Grabner, G., Janke, A., & Budge, M. (2006). Symmetric atlasing and model based segmentation: an application to

the hippocampus in older adults. In Medical Image Computing and Computer-Assisted Intervention (pp. 58–66).

Berlin.

10. Möller, T. B., & Reif, E. (2007). Pocket Atlas of Sectional Anatomy: Head and neck (p. 264). Thieme.

11. Hartkamp, N. S., De Cocker, L. J., Helle, M., van Osch, M. J. P., Kappelle, L. J., Bokkers, R. P. H., &

Hendrikse, J. In vivo visualization of the PICA perfusion territory with super-selective pseudo-continuous arterial

spin labeling MRI. NeuroImage, 2013;83:58–65. doi:10.1016/j.neuroimage.2013.06.070

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

12. Marshall RS, Lazar RM. Pumps, aqueducts, and drought management: Vascular physiology in vascular cognitive impairment. Stroke; a journal of cerebral circulation. 2011;42:221-226

13. Mohr J, Lazar R, Marshall R. Middle cerebral artery disease. In: Mohr J, Wolf P, Grotta J, Moskowitz M, Maybert M, von Kummer R, eds. Stroke: Pathophysiology, diagnosis and management. New York: Elsevier; 2011:385-424.

14. Kahlert P, Knipp SC, Schlamann M, Thielmann M, Al-Rashid F, Weber M, et al. Silent and apparent cerebral ischemia after percutaneous transfemoral aortic valve implantation: A diffusion-weighted magnetic resonance imaging study. Circulation. 2010;121:870-878

15. Ghanem A, Kocurek J, Sinning JM, Wagner M, Becker BV, Vogel M, et al. Cognitive trajectory after transcatheter aortic valve implantation. Circulation. Cardiovascular interventions. 2013

16. Lietz K, Brown K, Ali SS, Colvin-Adams M, Boyle AJ, Anderson D…Lazar, RM. The role of cerebral hyperperfusion in postoperative neurologic dysfunction after left ventricular assist device implantation for end-stage heart failure. The Journal of thoracic and cardiovascular surgery. 2009;137:1012-1019

17. Festa JR, Jia X, Cheung K, Marchidann A, Schmidt M, Shapiro PA…Lazar, RM. Association of low ejection fraction with impaired verbal memory in older patients with heart failure. Archives of neurology. 2011;68:1021-1026

18. Lazar RM, Marshall RS, Pile-Spellman J, Young WL, Hacein-Bey L, Sloan RP, et al. Continuous time estimation as a behavioural index of human cerebral ischaemia during temporary occlusion of the internal carotid artery. Journal of neurology, neurosurgery, and psychiatry. 1996;60:559-563

19. Marshall RS, Festa JR, Cheung YK, Chen R, Pavol MA, Derdeyn CP…Lazar, RM. Cerebral hemodynamics and cognitive impairment: Baseline data from the recon trial. Neurology. 2012;78:250-255

20. Festa JR, Schwarz LR, Pliskin N, Cullum CM, Lacritz L, Charbel FT…Lazar, RM. Neurocognitive dysfunction in adult moyamoya disease. Journal of neurology. 2010;257:806-815

21. Lezak D, Howieson D, Loring D. Neuropsychological assessment. Oxford; 2004.