Author + information
- Received July 13, 2017
- Revision received October 26, 2017
- Accepted October 30, 2017
- Published online February 19, 2018.
- Ronald M. Lazar, PhDa,∗ (, )
- Marykathryn A. Pavol, PhDb,
- Tobias Bormann, PhDc,
- Michael G. Dwyer, PhDd,
- Carlye Kraemer, MSe,
- Roseann White, MAf,
- Robert Zivadinov, MD, PhDd,
- Jeffrey C. Wertheimer, PhDg,
- Angelika Thöne-Otto, PhDh,
- Lisa D. Ravdin, PhDi,
- Richard Naugle, PhDj,
- Dawn Mechanic-Hamilton, PhDk,
- William S. Garmoe, PhDl,
- Anthony Y. Stringer, PhDm,
- Heidi A. Bender, PhDn,
- Samir R. Kapadia, MDo,
- Susheel Kodali, MDp,
- Alexander Ghanem, MDq,
- Axel Linke, MDr,
- Roxana Mehran, MDs,
- Renu Virmani, MDt,
- Tamim Nazif, MDp,
- Azin Parhizgar, PhDu and
- Martin B. Leon, MDp
- aDepartment of Neurology, University of Alabama at Birmingham, Birmingham, Alabama
- bDepartment of Neurology, Columbia University Medical Center, New York, New York
- cDepartment of Neurology, Medical Center–University of Freiburg, Freiburg, Germany
- dBuffalo Neuroimaging Analysis Center, SUNY/Buffalo, Buffalo, New York
- eNorth American Science Associates, Minneapolis, Minnesota
- fDuke Clinical Research Institute, Durham, North Carolina
- gDepartment of Physical Medicine and Rehabilitation, Cedars-Sinai Medical Center, Los Angeles, California
- hClinic for Cognitive Neurology, University of Leipzig, Leipzig, Germany
- iDepartment of Neurology, Weill Medical College of Cornell University, New York, New York
- jNeurological Institute, Cleveland Clinic, Cleveland, Ohio
- kDepartment of Neurology, University of Pennsylvania, Philadelphia, Pennsylvania
- lPsychology Department, MedStar Health, Washington, DC
- mDepartment of Rehabilitation Medicine, Emory University, Atlanta, Georgia
- nDepartment of Neurology, Mount Sinai School of Medicine, New York, New York
- oDepartment of Cardiovascular Medicine, Cleveland Clinic, Cleveland, Ohio
- pDepartment of Medicine, Columbia University Medical Center, New York, New York
- qDepartment of Cardiology, Asklepios St Georg, Hamburg, Germany
- rHerzzentrum Leipzig–Universitätsklinik, Leipzig, Germany
- sDepartment of Medicine, Mount Sinai School of Medicine, New York, New York
- tCV Path Institute, Gaithersburg, Maryland
- uClaret Medical, Santa Rosa, California
- ↵∗Address for correspondence:
Dr. Ronald M. Lazar, Department of Neurology, SC650K, University of Alabama at Birmingham, Birmingham, Alabama 35294.
Objectives The authors sought to determine baseline neurocognition before transcatheter aortic valve replacement (TAVR) and its correlations with pre-TAVR brain imaging.
Background TAVR studies have not shown a correlation between diffusion-weighted image changes and neurocognition. The authors wanted to determine the extent to which there was already impairment at baseline that correlated with cerebrovascular disease.
Methods SENTINEL (Cerebral Protection in Transcatheter Aortic Valve Replacement) trial patients had cognitive assessments of attention, processing speed, executive function, and verbal and visual memory. Z-scores were based on normative means and SDs, combined into a primary composite z-score. Brain magnetic resonance images were obtained pre-TAVR on 3-T scanners with a T2 fluid-attenuated inversion recovery (FLAIR) sequence. Scores ≤−1.5 SD below the normative mean (7th percentile) were considered impairment. Paired t tests compared within-subject scores, and chi-square goodness-of-fit compared the percentage of subjects below −1.5 SD. Correlation and regression analyses assessed the relationship between neurocognitive z-scores and T2 lesion volume.
Results Among 234 patients tested, the mean composite z-score was −0.65 SD below the normative mean. Domain scores ranged from −0.15 SD for attention to −1.32 SD for executive function. On the basis of the ≥1.5 SD normative reference, there were significantly greater percentages of impaired scores in the composite z-score (13.2%; p = 0.019), executive function (41.9%; p < 0.001), verbal memory (p < 0.001), and visual memory (p < 0.001). The regression model between FLAIR lesion volume and baseline cognition showed statistically significant negative correlations.
Conclusions There was a significant proportion of aortic stenosis patients with impaired cognition before TAVR, with a relationship between baseline cognitive function and lesion burden likely attributable to longstanding cerebrovascular disease. These findings underscore the importance of pre-interventional testing and magnetic resonance imaging in any research investigating post-surgical cognitive outcomes in patients with cardiovascular disease.
Aortic stenosis (AS) is a progressive disease, resulting in angina, syncope, and dyspnea, and advancing toward death from heart failure in those left untreated (1,2). Transcatheter aortic valve replacement (TAVR) has developed into a new treatment option for the older, high-risk cohort, and has since been shown to reduce mortality versus medical therapy in patients at extreme high risk for surgical aortic valve replacement (3).
Despite cardiovascular efficacy and improvement in quality of life (4), most previous studies of TAVR and the brain have focused on procedure-related ischemic injury (5). The stroke rate may have declined marginally as the TAVR technology has evolved (6), but questions have emerged about the significance of microemboli producing silent infarction resulting in clinically covert events, including changes in cognition. The overall rate of ischemic lesions, as detected by diffusion-weighted imaging (DWI) changes after TAVR, has ranged from 66% to 90% of cases (7–10). Most TAVR studies to date, however, have not shown a correlation between DWI lesion number or lesion volume (LV) with cognitive decline (8,11–13).
The SENTINEL (Cerebral Protection in Transcatheter Aortic Valve Replacement) trial compared distal protection vs no protection for brain embolization with TAVR, and found that major adverse cardiac and cerebrovascular events (MAACE) at 30 days were non-inferior to the performance goal (14). New DWI lesion burden, however, was not statistically different between the control and device arms. There was also no difference between global cognitive outcomes between the 2 treatment arms, although there was an overall correlation between new DWI LV and lesion number with cognitive decline at 30 days. Several previous studies appeared to show that a significant number of patients had cognitive impairment at baseline (12,15), but none of this prior work used a comprehensive test battery or examined their pre-TAVR cognitive status in conjunction with 3-T T2 fluid-attenuated inversion recovery (FLAIR) imaging as a marker of pre-procedure brain lesion burden. The purpose of this subanalysis of the SENTINEL data was to determine the full extent of cognitive function before TAVR, and the degree to which there is an association with pre-procedural brain pathology as measured by 3-T T2-FLAIR LV.
The SENTINEL trial included 363 patients, of whom 240 were randomized into the imaging cohort. The study population comprised subjects with severe symptomatic calcified native aortic valve stenosis who met the commercially approved indications for TAVR as described elsewhere, including their baseline characteristics (14). In general, there had to be high operative risk for mortality, no stroke or transient ischemic attack within the last 6 months, English or German speaking, and ability to undergo brain magnetic resonance imaging (MRI) 3-T examination. As part of their baseline evaluation, all patients underwent administration by certified assessors of the National Institutes of Health Stroke Scale (16) for assessment of neurological impairment and the modified Rankin Scale for evaluation of pre-TAVR neurological disability (17).
Patients underwent a 50-min cognitive battery consisting of 9 standardized neuropsychological tests, administered by a neuropsychologist or trained technician, all of whom were certified by the combined Neurocognitive Core Laboratory at Columbia University Medical Center in New York, and the Department of Neurology and Clinical Neurosciences, Medical Center–University of Freiburg, Freiburg, Germany. The battery was designed to assess neuropsychological sequelae associated with widely distributed pathology of vascular origin (18). All tests had published, age-adjusted norms (19–24), and the battery and its variants have been used extensively in studies of vascular cognitive impairment (25–28). Tests in German were validated translations of their English equivalents, and analyzed with North American norms, which is the standard of care. There were 17 U.S. and 2 German sites. All patients were administered the complete neurocognitive battery. In addition, each received the Mini-Mental State Examination (MMSE) (29) to control for general mental status and the Geriatric Depression Scale-15 (30) to control for depression. The order of test administration remained constant. The test battery, summarized in Table 1, was scored at the core lab. Neither the test examiners at local clinical sites nor the core laboratory personnel had knowledge of the imaging findings.
A z-score for each domain was calculated based on the normative means and SDs for each neurocognitive test. Normative groups were stratified by age and education (when possible). When there was more than 1 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 was more than 1 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 cognitive z-score for each treatment group was the average z-score from all domains (attention, executive function, processing speed, verbal memory, visual memory). Scores for the MMSE and the Geriatric Depression Scale were not included in the composite cognitive score. The threshold for abnormality was defined as −1.5 SD below the normative mean (31–33).
As described previously (14), brain MRIs were obtained before TAVR on 3-T scanners and performed at the clinical sites using standard parameters according to a protocol provided by the MRI reading center (Buffalo Neuroimaging Analysis Center, Buffalo, New York). The sequence of interest for this substudy was the T2-FLAIR because of its sensitivity to measure pre-existing lesions and white matter changes associated with underlying cerebrovascular disease (34). T2-FLAIR was acquired with a 2-dimensional spin echo inversion recovery sequence with an inversion time of 2,580 ms. Additional parameters were: relaxation time = 9,730 ms, echo time = 92 ms, slice thickness = 2 mm (no gap), acquisition matrix 256 × 186, and final voxel size = 0.94 mm × 1.17 mm × 2.0 mm. The imaging outcome of interest was T2-LV on the SENTINEL baseline FLAIR MRI scan, quantified via manual identification and semiautomated isocontour-based edge delineation. The imaging investigators, based in Buffalo, New York, were blinded to cognitive outcomes, and the test examiners and the staff at the Columbia Neurocognitive Core, none of whom were based in Buffalo, were blinded from the imaging findings.
Z-scores were calculated by dividing the difference in the observed individual mean and the normative mean by the normative SD for a given age range. In contrast to the post-procedural comparison in cognition between the test (device) and control groups in the SENTINEL trial, a normative population was used in the present analysis to characterize the cognitive status of the total patient population before intervention. A chi-square goodness-of-fit test was used to compare the percentage of subjects below −1.5 SD from the normative mean to the expected proportion. Correlation and regression analyses were employed to assess the relationship between neurocognitive z-scores and T2-LV. Due to the skewness in the distribution of T2-LV, a natural log transformation was utilized. Statistical analyses were generated using SAS software version 9.4 (SAS Institute, Cary, North Carolina). Because the SENTINEL trial was designed and powered to test for the difference in median new LV between the 2 treatment arms, it was not powered for any objectives in the present analysis.
A total of 240 patients were randomized to the imaging portion of the SENTINEL trial, of whom 234 (97.5%) underwent cognitive assessment and 219 had 3-T imaging (91.2%). Of the 234 patients tested, 186 were from U.S. sites. The baseline demographics of the total cohort, typical of the elderly, high-risk group who have planned TAVR, have been presented elsewhere (14). Patients were 83 years of age, one-half of whom were male, and mainly (97%) White/Caucasian. Just under 5% had a history of stroke. The mean National Institutes of Health Stroke Scale score at baseline was 0.9 (0 to 1 = normal), and the mean modified Rankin Scale was 0.4 (0 = no symptoms at all; 1 = no significant disability despite symptoms; able to carry out all usual duties and activities).
Table 2 depicts the means and SDs for the composite z- score, and for the 5 cognitive domains, respectively. These scores differ slightly from those reported in the SENTINEL outcomes analysis because baseline scores here were combined from both treatment arms into a single, pre-TAVR cohort (14). The mean composite z-score was −0.62 SD below the normative mean, equivalent to the 25th percentile. The 5 individual domain scores ranged from −0.15 SD below the normative mean for attention to −1.32 SDs for executive function. The mean MMSE score, used as a covariate, was 26.3 (SD ± 3.15). The mean score on the Geriatric Depression Score was 3.01 (SD ± 2.47), with 97.4% reporting no worse than mild depression. Using a paired t test, we found no statistical difference between verbal and visual memory (p = 0.62). There was no difference in T2-LV between the U.S. and German sites (p = 0.94).
The baseline composite z-score and the individual domain scores, however, obscure distributions demonstrating lower functioning than was evident by the mean values. Figure 1 shows the percentage of z-scores above and below the normative mean (z = 0), revealing a shift to the lower end of the scales. To quantify the number of patients who were impaired at baseline, we performed chi square calculations for goodness of fit by examining the number of study participants who scored ≤−1.5 SD below the normative means for the composite and domain scores, as seen in Table 3. The frame of reference was the 7% of the age-matched, normative population who would have been expected to perform at or below this level of function. We found that more than 13% of our cohort performed in the impaired range on the composite z-score, which was statistically significant (p ≤ 0.001), and nearly twice the rate of the normal population. The frequencies of patients impaired in executive function, verbal memory, and visual memory before TAVR were far greater than would have been expected on a normative basis (p ≤ 0.001). Indeed, 41.9% had impaired executive function. Not every individual domain, however, was affected. Mean scores for attention and processing speed were not different from those found in the normative group.
To determine one of the possible underlying mechanisms for these baseline performances, we explored whether there was an association between baseline neurocognition z-scores and T2-LV as an index of pre-TAVR severity of cerebrovascular disease, as shown in Figure 2. In every case, there was a statistically significant negative correlation between cognition and T2-LV. For the composite measure of cognition (Figure 2A), r = −0.27 (p < 0.001). The strongest correlations across the individual domains involved those with the highest proportion of impaired cognitive scores: executive function (Figure 2C), verbal memory (Figure 2E), and visual memory (Figure 2F). Lesser significant correlations were found in attention (r = −0.18) and processing speed (r = −0.14).
In the largest cohort of pre-TAVR patients studied to date, we found that a significant proportion of these older patients with high surgical risk had impaired cognitive domains before intervention. Moreover, there was a significant relationship between baseline cognitive function and T2-LV, which is likely attributable to underlying cerebrovascular disease. To our knowledge, this is the first study to demonstrate the extent to which this patient cohort carries the burden of vascular cognitive impairment, supported by imaging before their TAVR procedure.
Nearly every TAVR study to date has failed to show either a correlation between lesion number or LV with cognitive decline (8,11–13). These largely negative studies, however, have not adequately characterized their patient groups before intervention with respect to cognition and neurovascular status, failing to ask questions whether baseline patient characteristics may have accounted for their findings at follow-up.
One limitation in prior studies is the manner in which cognition was assessed. Most previous TAVR studies have used measures designed for cognitive screening, such as the MMSE (9,35–38) or the Montreal Cognitive Assessment Test (13,15,39). Kahlert et al. (8), for example, evaluated 32 patients before and after TAVR, with MMSE scores within the normal range at baseline and at 3 months, inferring that patients were functioning well before and after intervention. Vogels et al. (40), however, showed that whereas patients with New York Heart Association functional class II and III heart failure had MMSE scores no different than healthy controls, a comprehensive neuropsychological examination revealed significantly lower scores among the heart failure patients in a more detailed assessment. In the primary SENTINEL outcome paper (14), we showed that the mean baseline MMSE scores among the SENTINEL cohort for the device and control arms, respectively were >26, but the scores on the comprehensive assessment in the current analysis of the same cohort revealed domains of cognitive impairment. Thus, the assertion by Schoenenberger et al. (38) that, on the basis of the MMSE, only a few of their study patients had moderate-to-severe cognitive impairment is not supported when patients undergo more in-depth cognitive assessment. Our findings provide additional support for the NeuroARC consensus guidelines, recommending a comprehensive, baseline cognitive assessment for studies with neurological outcomes (41), and the need for baseline FLAIR imaging.
Recently, Ghanem et al. (12) reported the results of serial neuropsychological assessments of 111 patients before and after TAVR and correlated cognitive findings with serial 1.5-T DWI to determine the clinical impact of cerebral embolization. They reported that long-term cognitive performance was preserved in the great majority (91%) of patients throughout the first 2 years after TAVI. But, as noted by Browndyke and Mathew (42), a close examination of their baseline data showed a significant skew toward significant cognitive impairment. On the basis of the impairment criterion of ≤−1.5 SD below the normative mean, 27% had cognitive dysfunction at baseline, which is almost 4 times the expected proportion in the normative population (p < 0.001). That there was no further loss over time after TAVR is, therefore, unsurprising, given the level of pre-TAVR cognitive dysfunction producing a floor effect.
There are several reasons that could account for the low cognitive performances at baseline in our cohort. Given the age of this patient population, some of them could have been in the early-to-moderate stages of a neurodegenerative disease. The results of the MMSE in our protocol showed that the mean score was above 26, suggesting few patients had frank dementia. Depression, a known cause of cognitive decline among the elderly (43), did not account for our findings because nearly 98% of our cohort reported no worse than mild depressive symptoms. The second most common class of causes of cognitive impairment in the elderly is cerebrovascular in nature, including white matter disease typically caused by vascular risk factors (44,45). In our elderly patient group with severe aortic stenosis, more than a third had these risk factors. Our FLAIR 3-T imaging allowed us to address the question of the importance of vascular brain lesion burden pre-TAVR and its relationship to baseline cognition. We found for both the composite cognitive outcome and for the 5 domains statistically significant inverse correlations between baseline T2-LV and z-scores. These data are among the first demonstrations that baseline lesion burden has a measurable cognitive impact in this cohort of older patients, and thus suggest a mechanism for the relatively small effect of TAVR-related microembolization in the SENTINEL trial and no effect in other studies. These findings also validate the sensitivity of our cognitive battery in the assessment of brain-related function in the absence of any other neurological manifestations. Our results suggest strongly that future MRI studies designed to assess cognition in the setting of any cardiac-related procedures employ a T2-FLAIR sequence to ascertain the pre-procedure health status of the brain.
Although we found statistically significant correlations between T2-FLAIR LV and all of our cognitive domains, the multiple regression models showed that only a small proportion of the variance was accounted for by this association. Thus, the degree of cognitive impairment in selective domains, as shown in Table 3, is mediated by factors not measured in our study. Figure 2 showed that there was considerable variability. We did not measure the presence of cerebral microbleeds on imaging, which is known to affect cognition. We also did not look for the presence of the APOE E4 genotype, which is also a known risk factor for cognitive decline. We did not take lesion location or dispersion into account, or differentiate nonspecific T2-LV into underlying pathological categories. Finally, we were unable to address the intriguing question of how the subgroup of patients with preserved baseline cognitive function performed at the post-TAVR follow-up because the sample size was so low.
There is a greater degree of pre-procedural cognitive impairment among older, high-surgical-risk patients with planned TAVR than previously known. Moreover, this impairment is related to pre-TAVR cerebral lesion burden that is affected by cerebrovascular risk factors. Determining whether patients who have lower surgical risk have comparable degrees of cognitive loss requires further research.
WHAT IS KNOWN? Most prior research examining the effect of TAVR on neurocognition has shown little impact.
WHAT IS NEW? Our study from the SENTINEL trial showed that before TAVR, comprehensive neurocognitive assessment revealed significant cognitive impairment in this older, high-risk cohort that correlates with pre-TAVR MRI brain lesion burden.
WHAT IS NEXT? Future research will determine whether lower risk patients with presumably less intercurrent cerebrovascular disease, will be at higher risk for post-TAVR cognitive sequelae.
The SENTINEL trial was funded by Claret Medical. Dr. Lazar is a consultant to and holds stock options in Claret Medical. Dr. Bormann has been a paid advisor to Claret Medical. Dr. Dwyer has received consultant fees from Claret Medical and EMD Serono; and has received research grant support from Novartis. Ms. Kraemer has received financial support from Claret Medical. Ms. White is a consultant to and former employee of Abbott Vascular. Dr. Zivadinov has received speaking and consultant fees from EMD Serono, Novartis, Claret Medical, Celgene, and Genzyme-Sanofi; has received research support from Biogen Idec, Teva Pharmacuticals, EMD Serono, Novartis, Claret Medical, IMS Health, and Genzyme-Sanofi; and serves on the editorial boards of Journal of Alzheimer's Disease, BMC Medicine, BMC Neurology, Veins and Lymphatics, and CNS Drugs. Dr. Wertheimer is a paid consultant to Medtronic, and the Parkinson’s Alliance. Dr. Thöne-Otto has received financial support from Keystone Heart. Dr. Stringer has received research funding from the U.S. Department of Veterans Affairs. Dr. Kodali is a consultant for Edwards Lifesciences; has served on scientific advisory boards of Thubrikar Aortic Valve and Dura Biotech; and holds equity in Thubrikar Aortic Valve. Dr. Linke has received grant support from Medtronic, Edwards Lifesciences, and Boston Scientific; and has been a consultant to Abbott Vascular. Dr. Mehran receives institutional research support from Eli Lilly/Daiichi Sankyo, Inc., Bristol-Myers Squibb, AstraZeneca, The Medicines Company, OrbusNeich, Bayer, Beth Israel Deaconess, Novartis Pharmaceuticals, Medtronic, and CSL Behring; is a consultant for Janssen Pharmaceuticals, Osprey Medical, Watermark Research Partners, Boston Scientific, Shanghai BraccoSine Pharmaceutical, Cardiovascular Systems, Inc., and Medscape; is a consultant (paid to the institution) for Abbott Laboratories, CardioKinetix, and Spectranetics; serves on the scientific advisory board of Abbott Laboratories; serves on executive committees/data safety monitoring boards for Janssen Pharmaceuticals, Osprey Medical, and Watermark Research Partners; and holds equity or stock options in Claret Medical and Elixir Medical. Dr. Virmani is a consultant for 480 Biomedical, Abbott Vascular, Medtronic, and W. L. Gore. Dr. Nazif is a consultant to Edwards Lifesciences. Dr. Parhizgar is president and CEO of and has equity in Claret Medical. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- aortic stenosis
- diffusion-weighted image
- fluid-attenuated inversion recovery
- lesion volume
- Mini-Mental State Examination
- magnetic resonance imaging
- transaortic valve replacement
- Received July 13, 2017.
- Revision received October 26, 2017.
- Accepted October 30, 2017.
- 2018 American College of Cardiology Foundation
- Ghanem A.,
- Muller A.,
- Nahle C.P.,
- et al.
- Kahlert P.,
- Knipp S.C.,
- Schlamann M.,
- et al.
- Rodes-Cabau J.,
- Dumont E.,
- Boone R.H.,
- et al.
- Kahlert P.,
- Al-Rashid F.,
- Dottger P.,
- et al.
- Ghanem A.,
- Kocurek J.,
- Sinning J.M.,
- et al.
- Kapadia S.R.,
- Kodali S.,
- Makkar R.,
- et al.
- Auffret V.,
- Campelo-Parada F.,
- Regueiro A.,
- et al.
- Brott T.,
- Adams H.P. Jr..,
- Olinger C.P.,
- et al.
- van Swieten J.C.,
- Koudstaal P.J.,
- Visser M.C.,
- Schouten H.J.,
- van Gijn J.
- Gorelick P.B.,
- Scuteri A.,
- Black S.E.,
- et al.
- Lezak M.,
- Howieson D.B.,
- Loring D.W.
- Brandt J.,
- Benedict R.H.B.
- Meyers J.E.,
- Meyers K.R.
- Mitrushina M.,
- Boone K.B.,
- Razani J.,
- D’Elia L.F.
- Wechsler D.
- Warach S.,
- Baird A.E.,
- Dani K.A.,
- Wintermark M.,
- Kidwell C.S.
- Stortecky S.,
- Schoenenberger A.W.,
- Moser A.,
- et al.
- Schoenenberger A.W.,
- Zuber C.,
- Moser A.,
- et al.
- Haussig S.,
- Mangner N.,
- Dwyer M.G.,
- et al.
- Lansky A.J.,
- Messe S.R.,
- Brickman A.M.,
- et al.
- Browndyke J.N.,
- Mathew J.P.
- Morimoto S.S.,
- Alexopoulos G.S.
- Breteler M.M.,
- van Amerongen N.M.,
- van Swieten J.C.,
- et al.