Results from The VSI Study employing multimodal neuroimaging procedures will provide a deeper understanding of the structural and functional neural correlates of VSI in older adults with normal and preclinical AD. Here, preclinical AD will be defined as manifesting impaired cognitive performance [performance worse than 1.5 standard deviations from the mean on standardized neuropsychological tests] and presence of elevated Aß in plasma at baseline using established cut scores (Bateman et al., 2019). We hypothesize that the magnitude of VSI will be correlated with gray matter volume, cortical thickness, and blood-oxygen-level-dependent (BOLD) signal activation in subcortical and cortical regions of interest including dorsal lateral prefrontal cortex (DLPFC), rostral middle frontal, and superior frontal gyrus at study baseline (Year 1). We predict that older adults with preclinical AD will manifest reduced magnitude of VSI (worse), decreased cortical volumetrics, decreased functional connectivity, and lower BOLD responses when compared to older adults with normal cognition.

Individuals with Alzheimer’s disease are at greater risks for falls and mobility disability, but specific causes of AD and the temporal onset of functional changes across systems are currently not known. We have shown a mediating effect of dementia and mild cognitive impairment on the relationship between VSI and motor outcomes (Mahoney and Verghese, 2018; Mahoney et al., 2019). These results suggested that individuals with cognitive impairments manifested poor VSI and poor balance/slow gait. Using an fMRI task where participants are asked to respond as quickly as possible to unisensory visual, unisensory somatosensory and combined visual-somatosensory stimuli in a 3-Tesla (3T) magnet, our second specific aim will determine whether visual-somatosensory task-related BOLD activation in the prefrontal cortex at baseline predicts future mobility (gait) declines and risk of incident falls. We hypothesize that preclinical AD causes disruptions in subcortical and cortical (multisensory, motor, and cognitive) regions that modulate multisensory, motor, and cognitive functions necessary for efficient mobility.

The validity of VSI as a novel marker for AD will be established by correlating the magnitude of VSI with presence of Aβ using plasma-based measures at baseline. In Year 2, positron emission tomography (PET) measures affording localization of Aβ deposits (Piramal Imaging) to estimate Aβ neuritic plaque density will also be examined in relation to magnitude of VSI. Aβ protein deposition has been documented in both sensory and cognitive areas (Thal et al., 2002; Jack et al., 2018, 2019; Bateman et al., 2019). Therefore, we hypothesize that increased Aβ accumulation in sensory and cognitive areas, areas related to increased AD pathology, will be associated with decreased magnitude of VSI.

In keeping with the NIA-AA research framework (Jack et al., 2018), our innovative and timely project will distinguish AD symptomology (presence of mild cognitive impairment) from AD pathology (Aβ accumulation), while also applying the AT (N) classification system [Aβ (A), tau (T), and neurodegeneration (N)] to attain more direct assessment of neuropathologic changes. More specifically, and in keeping with the goals of establishing whether magnitude of VSI is a novel and early biomarker of AD, associations of VSI with plasma-based total and phosphorylated Tau, neurofilament (NfL), ApoE, and multimodal neuroimaging measures of Neurodegeneration will also be examined for study completeness.

Participants will complete a simple reaction time (RT) test employing three bilaterally presented conditions (visual, somatosensory, and multisensory visual-somatosensory) and a control (i.e., “catch”) condition where no stimulation is presented, and no response is expected. The four stimulus conditions will be randomly presented with equal frequency (15 trials per condition per block, 3 blocks, yielding a total of 180 trials). The addition of “catch” trials and variable inter-stimulus-interval (ranging from 1–3 s) impedes anticipatory effects (see Mahoney and Verghese, 2018; Mahoney and Verghese, 2019, 2020; Mahoney et al., 2019; for details). Participants will be instructed to respond to all stimuli as quickly as possible. Performance accuracy will be defined as the number of accurate stimulus detections divided by 45 trials per condition. Using our established methodology, robust probability (P) models that compare the cumulative distribution function (CDF) of combined unisensory visual (V) and unisensory somatosensory (S) reaction times with an upper limit of 1 min [P (RT_V ≤ t) + P (RT_S ≤ t), 1] to the CDF of multisensory visual-somatosensory (VS) reaction times [P (RT_VS ≤ t)] will be implemented. For any latency t, the inequality holds when the CDF of the actual multisensory VS condition [P (RT_VS ≤ t)] is less than or equal to the predicted CDF {min [P (RT_V ≤ t) + P (RT_S ≤ t), 1]}. When the actual CDF is greater than the predicted CDF (i.e., positive value), the model is violated, and the RT facilitation is the result of multisensory interactions that allow signals from redundant information to integrate or combine non-linearly. Predicted CDF will be subtracted from the actual CDF to form a difference curve. The area-under-the-curve of the group-level violated portion of the difference curve will serve as the continuous measure of magnitude of VSI.

All neuroimaging procedures will be conducted at the Gruss Magnetic Resonance Research (MRRC) Center at Albert Einstein College of Medicine under the direction of Dr. Lipton. The MRRC offers state-of-the-art multimodal neuroimaging on Philips whole-body Ingenia Elition 3.0 Tesla Magnetic Resonance Imaging (MRI) scanner equipped with 32-channel head coil. Multimodal MRIs will be captured at baseline (i.e., study Year 1) and processed by our neuroimaging team consisting of Drs. Blumen, Fleysher, and Hoptman. Our non-invasive multimodal MRI imaging techniques are reliable and have been used extensively in both healthy aging and dementia studies in our division/department. For the VSI Study, specific MRI outcome measures are listed by modality in Table 3. Structural MRI (sMRI; ∼5 min) will be acquired using high-resolution T1-weighted whole head structural imaging using axial 3D-MP-RAGE acquisition over a 240 mm field of view (FOV) with 1.0 mm isotropic resolution. TE = 4.6 ms, TR = 9.9 ms, α = 8°, and SENSE factor = 2.6 (left-right) x 2 (head-foot). Functional MRI (∼10 min total) will be acquired using whole brain T2* weighted images with echo planar weighted images with echo planar imaging over a 224 mm FOV on a 112 × 112 acquisition matrix, 3 mm slice thickness (no gap); TE = 30 ms, TR = 2,000 ms, flip angle = 90°, SENSE factor = 2 and 42 trans-axial slices per volume. The fMRI procedures will measure BOLD activation (outcome measure) during the VSI task. This event-related design emulates our established psychophysical protocol where 60 trials of visual alone, somatosensory alone, multisensory visual-somatosensory (VS; Mahoney and Verghese, 2019) will be presented in the scanner, but will also include the above-mentioned “catch” trials (60 trials). The visual (V) stimulus will be bilateral asterisks presented for 100 ms on a VisuaStim digital visor (Resonance Technology, Inc., Northridge, CA, USA). The somatosensory (S) stimulus will be bilateral pneumatic pulses presented for 100 ms through the Somatosensory Stimulus Generator system (4-D Neuroimaging) which is compatible in the MRI scanner. These stimuli will be presented alone and concurrently in the case of the concurrent VS stimulus. The critical contrast here will examine differences in BOLD activation between the multisensory VS condition vs. the sum of the two unisensory conditions (V + S). Resting-state (rs)-fMRI (10 min) will also be captured while participant lay still and relax (i.e., a passive no-task condition) with their eyes open. Fluid Attenuated Inversion Recovery (FLAIR; ∼5 min total) will account for white matter hyperintensities (WHI) indicative of small vessel disease. FLAIR will be acquired using whole head imaging sagittal 3D-TSE-IR acquisition over a 250 mm FOV with 1 mm isotropic resolution. TE = 338 ms, TR = 4,800 ms, TI = 1,650 ms TSE Factor = 182, compressed SENSE acceleration factor 3.5. FLAIR results will account for presence/absence of small vessel disease and will be considered as a covariate. Additional multimodal neuroimaging procedures to be included for study completeness, beyond the scope of the specific aims include: Susceptibility Weighting Imaging (SWI); Pseudo-Continuous Arterial spin labeling (pc-ASL); and Neurite orientation and dispersion density imaging (NODDI) – see Table 3 for details.

In terms of our statistical approach for Aim 1, the magnitude of VSI (independent variable) will be analyzed and quantified using established probabilistic modeling procedures (Mahoney and Verghese, 2019). The dependent measures of structural and functional neural integrity include: (1) cortical thickness: (2) volumetric measures for regions of interest extracted from structural MRI; (3) Beta-weights for the multisensory contrasts for each region of interest extracted from task-based fMRI; and (4) Fisher z-transformed resting-state functional connectivities between pairs of regions extracted from resting-state fMRI.

Covariates identified in our prior studies (Holtzer et al., 2007; Verghese et al., 2007b,2008a; Mahoney and Verghese, 2018, 2019, 2020), including but not limited to age, gender, ethnicity, medical comorbidities (including cardiovascular disease), total intracranial volume, and attentional capacity will be selected to account for their influence on VSI and association with outcomes. Participants will be categorized into two groups based on preclinical AD diagnosis at baseline. All statistical approaches will be supervised by our study statistician, Dr. Wang.

SAS 9.4 (Cary, NC) will be used for the analyses. We will conduct multivariate mixed effects models for imaging outcomes for the following a priori selected regions of interest including: dorsal lateral PFC, rostral middle frontal, and superior frontal gyrus regions, superior temporal sulcus, motor cortex, thalamus, basal ganglia, hippocampal, and cerebellum (one per outcome, with group factors as necessary). These regions are selected based on preliminary findings in 100 older adults (unpublished data) which reveal significant (p < 0.05) associations between magnitude of VSI and measures of structural integrity (defined here as either volume or cortical thickness) in the following regions: parahippocampal (memory); caudal middle-frontal dorsal lateral prefrontal cortex (DLPFC: cognitive functions - especially executive & attention); superior temporal sulcus (STS; multisensory); precentral (motor), postcentral (somatosensory), and lateral occipital (visual). Preliminary fMRI findings in 56 healthy older adults (ages 65–92; unpublished data) further supports inclusion of these regions given significant associations between VSI magnitude and blood-oxygen-level-dependent (BOLD) responses in known multisensory (middle temporal), motor (basal ganglia), and cognitive areas (PFC including DLPFC). The outcomes in these models include measures of neural integrity, structural (volume and thickness), and/or functional (BOLD) activation isolated by region, and the predictor of interest is the magnitude of VSI. Additional models will further include cognitive status (normal or preclinical AD) and its interaction with magnitude of VSI. The effect of cognitive status (normal or preclinical AD) on magnitude of VSI will be evaluated using similar mixed effects models. The hypothesis-driven analyses will be limited to BOLD activation in the aforementioned regions of interest. Models will be run unadjusted and then adjusted for confounders.

Here, we propose that reduced VSI activation in specific regions of PFC (fMRI BOLD responses), will be associated with worse balance (unipedal stance), slower gait (worse Pace scores), and increased risk of incident falls. Additionally, we propose that preclinical AD will reduce VSI activation in PFC regions of interest, which will in turn adversely affect mobility measures. Our longitudinal design will allow us to identify the impact of functional changes in (multi)sensory, motor, and cognitive processes (and their interactions) on the progression of AD that result in mobility decline.

The VSI task and multimodal neuroimaging procedures for this aim have been described above in Aim 1. For Aim 2, the VSI task will be run in the magnet to obtain fMRI task-related BOLD responses with concurrent psychophysical data. The following aim-related mobility procedures [established tests that have been validated and utilized in our center for over two decades (Verghese et al., 2002a,2007b,2008a,2009, 2012; Holtzer et al., 2007, 2012, 2014; Verghese and Xue, 2010, 2011; Ayers and Verghese, 2014; Mahoney et al., 2014, 2017, Holtzer et al., 2014; Mahoney and Verghese, 2018, 2020)] will be included in Aim 2:

Balance (∼2 min) will be assessed using unipedal stance time, which requires individuals to balance their body weight with foot on the ground for a maximum of 30 s (Hurvitz et al., 2000, 2001). Unipedal stance time is a widely used clinical test that is listed under NIH’s toolbox. Poor scores on this test have been associated with presence of neuropathy (Hurvitz et al., 2001), and predict falls in older adults (Hurvitz et al., 2000). This test will be administered twice during each study visit and maximum unipedal stance time (sec) will serve as the outcome measure.

Quantitative Gait (∼5 min) will be assessed on a 28-foot instrumented walkway (PKMAS system; Zenometrics LLC) with embedded pressure sensors that provides spatial and temporal gait parameters including: gait velocity, stride length, percentage of double support, stride time, stance time, cadence, stride length variability, and swing time variability. Participants will be assessed twice while walking on the mat at their everyday pace. Gait velocity, as well as the Pace Factor score comprised of gait velocity, stride length and percentage of double support, will serve as dependent measures.

History of falls (∼5 min) in the past 1 year, number of incident falls over a 3 year longitudinal study-period, and fall information such as type, injury and location will be tracked at yearly in-house interviews and during bimonthly telephone interviews using established criteria and standardized questionnaires (Tinetti et al., 1994). Falls are defined as sudden, unintentional, unprovoked changes in body posture, not due to a major intrinsic event (stroke) or overwhelming hazard. Dichotomous ratings of fall-history over the past 1 year (0, 1), presence of incident fall over study period (0, 1) and time to fall/censor will serve as outcome measures.

The association of functional VSI activation in specific PFC regions of interest with mobility measures of balance and quantitative gait will be examined cross-sectionally at baseline (Year 1), using linear regression models. Linear mixed effects models (LMEM) will be used to examine the association of baseline functional VSI activation in the PFC on the changes in the longitudinal balance and quantitative gait performance. The predictors will be examined as continuous variables to facilitate clinical translation of results. Adjustments for multiple comparisons will be made. Additional LMEMs will be employed to examine interplay and time course of multisensory, cognitive, and motor functioning. Cox proportional hazard model will be used to evaluate the association of magnitude of VSI with the risk of incident falls (Mahoney et al., 2019) and hazard ratios (HR) with 95% confidence intervals (CI) will be reported. Time to fall will be recorded as number of days from baseline study date to the interview date when the fall was recorded. If the participant does not report a fall, the follow-up time will be defined as the number of days from the baseline in-house visit to the last date of contact. Repeated incident falls will be examined using Andersen-Gil extension of Cox model (Anderson and Gill, 1982) and Poisson models. Robust sandwich covariance estimates account for correlations among multiple events within the same participant. Cox models will be adjusted for potential confounders. Proportional hazards assumptions of all models will be tested graphically and analytically. We will also apply mediation analysis using product of coefficients methods to evaluate whether cognitive status (normal vs. preclinical AD; independent variable) causes variation in PFC-related VSI activation (mediator), which in turn causes variation in specific mobility measures (dependent variables) using separate mediation models (balance and pace). Mediation analyses will be run using IBM’s Statistical Package for the Social Sciences (SPSS-28) and Hayes’ PROCESS package (Hayes, 2018). Confidence intervals that do not include 0 for the mediator will be defined as mediation.

Alzheimer’s disease is associated with build-up of specific proteins (i.e., biological markers) in the brain, namely Amyloid-βeta (Aβ) in the form of plaques (Thal et al., 2002) and tau (T) in the form of neurofibrillary tangles. Common brain imaging techniques such as MRI or Computerized Tomography (CT) do not afford assessment of amyloid plaques and neurofibrillary tangles. However, molecular imaging procedures like Positron Emission Tomography (PET) imaging directly visualize these characteristic features of Alzheimer’s disease. Aβ accumulates in sensory association areas well before higher-order cognitive areas like the PFC (Thal et al., 2002). In Aim 3, we predict that AD pathology (i.e., accumulation of Aβ) will be associated with decreased magnitude of VSI in preclinical Alzheimer’s disease participants. Here, presence of Aβ will be measured in blood at baseline using plasma-based assays and established cut-scores (Bateman et al., 2019). Individuals that are Aβ+ on plasma-based tests at baseline, will receive amyloid PET imaging in study year 2. The combined use of conventional MRI with these techniques will also contribute to the early identification of Alzheimer’s disease.

The experimental design for this aim has been described above. Beyond the VSI task the following specific Aβ procedures will be implemented to determine its association with magnitude of VSI, and ultimately its use as a novel and early biomarker for preclinical AD.

Plasma-Based Blood Testing will be conducted during baseline visits for each participant. Aβ (40 and 42) and Apolipoprotein E (ApoE) will be assessed by C₂N Diagnostics lab using novel multiplexed assays (Jack et al., 2018). PrecivityAD™ accuracy for determining amyloid positive versus negative status was 86%. Blood samples will be collected and placed in Einstein’s biorepository. Frozen samples will be subsequently shipped to C₂N Diagnostics to be processed. Results from plasma-based testing conducted on bloods drawn at baseline, in conjunction with neuropsychological performance, will be critical for study group assignment, as well as disentangling AD symptomology from AD pathology.

Amyloid PET Imaging will only be conducted in study Year 2 for participants enrolled in the preclinical AD group (n = 104). All PET scans will be conducted at Montefiore Medical Center by Dr. Valdivia and her team. Positron emission tomography (PET) is expensive and involves the use of an imaging device (scanner) and a radiotracer that is injected into the patient’s bloodstream. The radiotracer used to estimate Aβ neuritic plaque density for this specific aim is called Neuraceq (florbetaben F-18) and is manufactured by Piramal Imaging. PET imaging after the radiotracer is injected, will afford quantification of the distribution of Aβ in the brain, where affected brain regions containing Aβ will be tabulated.

The association of magnitude of VSI with presence of Aβ protein levels in both plasma-based Aβ levels (continuous measure) and amyloid PET scans (dichotomous ± rating by brain region) will be examined using logistic and linear regression models, respectively, while adjusting for potential confounders.