Cognitive impairment and dementia after stroke: design and rationale for the DISCOVERY study

Abstract

Stroke is a leading cause of the adult disability epidemic in the US, with a major contribution from post-stroke cognitive impairment and dementia (PSCID), the rates of which are disproportionally high among the health disparity populations. Despite the PSCID’s overwhelming impact on public health, a knowledge gap exists with regard to the complex interaction between the acute stroke event and highly prevalent pre-existing brain pathology related to cerebrovascular and Alzheimer’s disease, or related dementia. Understanding the factors that modulate PSCID risk in relation to index stroke event is critically important for developing personalized prognostication of PSCID, targeted interventions to prevent it, and for informing future clinical trial design. The Determinants of Incident Stroke Cognitive Outcomes and Vascular Effects on RecoverY (DISCOVERY), a collaborative network of thirty clinical performance clinical sites with access to acute stroke populations and the expertise and capacity for systematic assessment of PSCID will address this critical challenge. DISCOVERY is a prospective, multi-center, observational, nested-cohort study of 8,000 nondemented ischemic and hemorrhagic stroke patients enrolled at the time of index stroke and followed for a minimum of 2 years, with serial cognitive evaluations and assessments of functional outcome, with subsets undergoing research MRI and PET and comprehensive genetic/genomic and fluid biomarker testing. The overall scientific objective of this study is to elucidate mechanisms of brain resilience and susceptibility to PSCID in diverse US populations based on complex interplay between life-course exposure to multiple vascular risk factors, pre-existing burden of microvascular and neurodegenerative pathology, the effect of strategic acute stroke lesions, and the mediating effect of genomic and epigenomic variation.

Post-stroke Cognitive Impairment and Dementia: the Scope of the Problem

Stroke results in an epidemic of adult disability in the United States (US).^1,2 Cognitive impairment and dementia after stroke are highly prevalent (15 to 70%)^3–8 and disabling^9–11, and the US populations experiencing health disparities carry a disproportionate burden of this disability with distinct phenotypes.^4,12,13 Therefore, understanding the mechanisms and determinants of post-stroke cognitive impairment and dementia (PSCID) is of utmost clinical importance, is a critical public health priority, and key to delivering equitable health care to diverse US populations.

Cerebrovascular disease and Alzheimer’s pathologies are key contributors to the overall burden of dementia, and the need to understand their relationship is a national research priority¹⁴. As traditionally defined, vascular dementia accounts for ≥20% of all dementia diagnoses.¹⁵ This figure, however, substantially underestimates the vascular contributions to cognitive impairment and dementia (VCID), particularly the high burden of concomitant Alzheimer disease (AD) and AD-related dementia (AD/ADRD) pathologies.^16–18 There is strong evidence of additive or possibly synergistic interactions between cerebrovascular and AD pathology predisposing to dementia, such as reports that: 1) patients with mixed versus single pathology are three times more likely to be demented;¹⁸ 2) vascular pathology lowers the threshold of AD pathology required for dementia;^19,20 and 3) the presence of AD pathology²¹ contributes to cognitive decline in VCID.²² An analysis of 1079 individuals in the clinical-pathologic Religious Orders Study and Memory and Aging Project found isolated AD in only 9% versus 40% with AD plus an advanced vascular pathology (macroscopic infarcts, cerebral amyloid angiopathy, atherosclerosis, or arteriolosclerosis) and 44% with AD, vascular, and another neurodegenerative pathology.²³ These studies indicate that overlapping vascular and neurodegenerative pathologies are the rule rather than the exception in age-related dementia.

The differential and relative contributions of cerebrovascular and neurodegenerative disease to PSCID is currently unknown. The US National Institutes of Health (NIH) created the MarkVCID consortium to validate biomarkers for VCID due to cerebral small vessel disease (SVD)^24,25; similarly, a number of large-scale international collaborations including STROKOG (Stroke and Cognition Consortium), SVDs@target (Small Vessel Diseases-at-target), and the HBC (Heart-Brain Connection) have focused on the mechanisms of post-stroke VCID.²⁶ Among these, STROKOG consortium represents a truly global effort reporting 32 member studies with a total of 18,000 individuals and representing 18 countries.²⁷ In a recent decade, the global consortia and independent studies have made progress in the conceptualization and definitions of VCID,^28,29 including the risk, trajectory of PSCID^30–32 and exploratory associations with imaging and fluid biomarkers.^33–35 However, no biomarkers have yet been demonstrated to reliably differentiate between neurodegenerative and cerebrovascular disease. Thus, understanding the relationships between the two remains a high priority to enable targeting of underlying mechanisms and reducing the overall burden of cognitive impairment and dementia.^21,22

An acute stroke insult to a brain with pre-existing microvascular (SVD) and neurodegenerative changes likely initiates a series of pathological events leading to PSCID. The features of stroke, in combination with pre-existing cerebrovascular and AD/ADRD pathologies predispose to PSCID; however, the mechanisms and interactions of these effects are unclear. A critical limiting factor in this is lack of consensus on how and when to define PSCID, given: 1) no clear pre-stroke cognitive status data in most studies and 2) the time course of cognitive changes in the early months after stroke, which leads to inconsistent estimates of post-stroke dementia occurrence. Individuals who experience a first-ever stroke have worse cognitive performance even before the stroke, as compared to their counterparts without stroke.³⁶ In most studies, however, cognitive change cannot be evaluated, so estimates of PSCID may not adequately account for any cognitive dysfunction which was actually present pre-stroke, or may not account for any decline which was underway before the stroke onset. Despite this challenge, in those studies in which cognition has been evaluated pre-stroke, such as the REGARDS study, stroke patients do appear to experience steeper decline after a stroke.³⁶

The chronology of when to evaluate PSCID is a second major challenge. If cognition is considered in the very early period post-stroke, estimates of impairment can seem remarkably high (91.5% at two weeks, in one series).³⁷ Although early cognitive decline may be common, most individuals improve within the first six months post-stroke,³⁸ and therefore true post-stroke dementia may be an inappropriate label for these early fluctuations in cognition that are so common to stroke patients.

Despite these limitations in the field, however, several studies have converged with evidence for patterns which might increase the likelihood of developing PSCID. The size and location of the stroke have been implicated in the development of PSCID, with greater likelihood of PSCID for individuals experiencing larger strokes or those in supratentorial regions, or, specifically, the anterior circulation.^39–41 Beyond lesion topography, the acute injury from stroke may induce secondary neurodegenerative changes in anatomically linked regions of cortex, presenting a plausible mechanism for PSCID.⁴² Regarding pre-existing burden of disease, several different studies of non-demented, stroke-free adults have shown an association between chronic microvascular injury and global cognition and performance in specific cognitive domains,^43–45 and chronic microvascular disease and its progression are strongly associated with an increased risk of incident stroke.^46,47 Further, radiographic markers of SVD such as peak width skeletonized mean diffusivity and free water modeling on diffusion tensor imaging (DTI), have been linked to white matter (WM) structural integrity, extracellular edema, and cognitive status in stroke-free adults with SVD.^48,49 Thus, in the setting of acute stroke, the burden and location of pre-existing SVD disease is an important risk factor for development of PSCID.^50,51 Similarly, evidence of AD pathology,²¹ or the presence of the APOE ε4 allele,^52,53 further increases PSCID risk. In total, these findings suggest that combined analyses of acute stroke features and pre-existing SVD and neurodegeneration holds significant potential to elucidate the mechanisms of early and late PSCID.

Despite the established relationships between stroke and dementia, PSCID is vastly understudied because of the lack of infrastructure and combined expertise to characterize this complex phenotype with sufficient statistical power and to identify its determinants. Although clinical evidence has linked acute stroke and pre-existing brain pathology to development of PSCID, a large-scale, multicenter, prospective study with systematic ascertainment of clinical and radiographic variables accompanied by longitudinal neuropsychological evaluations is essential in order to elucidate the specific stroke characteristics that cause PSCID in distinct populations. The Determinants of Incident Stroke Cognitive Outcomes and Vascular Effects on RecoverY (DISCOVERY) funded jointly by the National Institute of Neurological Disorders and Stroke and National Institute on Aging is designed to fill this knowledge gap.

PSCID in Populations Experiencing Health Disparities

Cognitive decline after stroke disproportionally affects populations experiencing health disparities in the US. Stroke is a disease of disparities with Blacks followed by LatinX populations experiencing significantly greater incidence, and poorer outcomes after stroke.⁵⁴ A body of literature linking social determinants of health to stroke disparities has been growing and multiple social determinants including lack of social support, low educational status, health access, community level poverty, discrimination and stigma have likewise been associated with stroke disparities^55,56. Little has been reported about race-ethnic disparities, social determinants and PSCID; however, studies show that stroke is more frequently associated with dementia five years post-stroke in African-Americans versus Whites, despite a younger age of incident stroke,⁵⁷ and also with greater cognitive decline post-stroke, even after accounting for any additional interim strokes.⁵⁸ Furthermore, cognition before stroke, as stated above, is associated with greater risk of stroke, and this association is even more pronounced in Black versus White patients.⁵⁹ In addition, individuals born in a high stroke mortality state are at greater risk for dementia than those born elsewhere, and this association was strongest for Black versus White individuals (with greatest dementia risk among Blacks born in a high stroke mortality state).⁶⁰

Acute Stroke Trials with Cognitive Outcomes – Experience to Date

DISCOVERY is occurring in the context of a greater appreciation of the clinical importance of cognition as an outcome in trials focused on stroke prevention. Cognitive outcomes have been part of phase III stroke prevention trials since at least the Vitamin Intervention for Stroke Prevention (VISP) trial, which began enrolling patients in September 1996.^61,62 In VISP, sites administered the Mini Mental State Exam (MMSE) locally to participants and no significant difference was found in the rates of ≥2 point decline in MMSE at 1 year between treatment groups (17.7% vs. 19.1%), consistent with the absence of an effect on stroke events. Since then, investigators have developed a better appreciation for and different approaches to assess PSCID. Harmonized standards have been put forth to evaluate vascular cognitive impairment, in hopes of reducing risk of false negative results and facilitating cross-study comparisons.⁶³

Trials of blood pressure lowering have chosen to use cognitive testing to measure incidence rates of mild cognitive impairment and dementia. The HYVET (Hypertension in the Very Elderly)-COG trial had local research staff assess MMSE and a fall to <24 points or >3 points in 1 year triggered further evaluations to assess for dementia.⁶⁴ HYVET-COG found that antihypertensive treatment in elderly patients did not significantly reduce incident dementia. The SPRINT-MIND trial⁶⁵ was ancillary to the Systolic Blood Pressure Intervention Trial) SPRINT (NCT01206062).⁶⁶ Participants were randomized to a systolic blood pressure goal of either less than 120 mm Hg (intensive treatment) or less than 140 mm Hg (standard treatment). SPRINT demonstrated that intensive blood pressure control was associated with significant reductions in the primary composite endpoint of myocardial infarction, other coronary syndromes, stroke, heart failure, or death from cardiovascular cause. This led to stopping the trial after a median follow-up of 3.26 years. In SPRINT-MIND, 8563 patients completed at least one cognitive follow-up. Cognitive testing (Montreal Cognitive Assessment [MOCA] and additional tests) was performed by local research staff. SPRINT-MIND did not find a significant therapeutic effect on dementia, but did find that intensive blood pressure control significantly reduced the risk of mild cognitive impairment (HR, 0.81). These two large blood pressure lowering studies illustrate the insensitivity of the dementia endpoint relative to milder cognitive decline.

Some trials have opted to administer cognitive testing centrally by telephone. This has the advantage of reducing measurement variability and local research coordinator burden. ARCADIA CSI, a sub-study of ARCADIA (NCT03192215), illustrates this approach. In the parent study, a target of 1100 patients with cryptogenic stroke and atriopathy are randomized to either apixaban or aspirin and study for secondary prevention of stroke. ARCADIA CSI plans to randomize 500 patients. The patients have to be >3 months post index stroke, have no history of pre-existing dementia and have at least an 8th grade education. Cognition is assessed using a centrally administered telephone battery estimated to take <30 minutes, and the cognitive outcome will assess change over 3 years.

Stroke trials with cognitive outcomes are not limited to those with pharmacological interventions. CREST-2 is an ongoing trial of revascularization plus intensive medical therapy versus intensive medical therapy alone.⁶⁷ The risk factor targets of centrally directed intensive medical therapy are LDL cholesterol <70 mg/dl and systolic blood pressure <140 mmHg. The primary outcome is the composite of stroke and death within 44 days following randomization and stroke ipsilateral to the target vessel thereafter, up to four years. Change in cognition is a secondary outcome. As with ARCADIA-CSI, cognitive assessments are performed centrally and by telephone. CREST-H (NCT03121209)⁶⁸ a sub-study of CREST-2, tests the hypothesis that the magnitude of treatment differences (revascularization vs. medical management alone) with regard to cognition will differ between those with flow impairment compared to those without flow impairment identified and characterized by CT or MR perfusion studies.

Overall, early attempts to incorporate cognitive outcomes or cognitively based diagnostic outcomes (e.g., dementia) have been mixed or are ongoing and have focused on homogeneous stroke or at-risk populations. DISCOVERY takes advantage of clinic-based assessments for performing detailed cognitive testing and the advantage of telephone-based assessments for scalability, site-burden, and standardization. As such, DISCOVERY will provide essential information on cognitive trajectories for the global stroke population, and this should greatly help in planning future trials targeting cognitive impairment.

Etiologic Stroke Subtype and Risk of Post-Stroke Cognitive Decline

Prior studies have suggested that cortical, as opposed to subcortical strokes, are more frequently associated with cognitive impairment,⁴¹ with support from studies demonstrating higher rates of impairment in individuals with large-vessel or cardioembolic strokes.³⁷ However, few longitudinal studies have systematically examined the association of stroke subtype with PSCID, and even fewer studies studied ethnically diverse US populations¹³ or the differential effect of stroke subtype on the domains of cognitive impairment.⁶⁹ Of all the ischemic stroke subtypes, lacunar (SVD) has been the most extensively studied. In the Secondary Prevention of Small Subcortical Strokes (SPS3) Trial, neuropsychological testing was performed at study entry, 2 weeks to 6 months following the index stroke. Among the 1636 who received testing in English, 47 % were classified as having mild cognitive impairment (MCI).⁷⁰ Among the 117 who received testing in Spanish, 40 % were classified as having mild cognitive impairment.¹³ A meta-analysis of 24 studies with 7575 patients found the prevalence of MCI or dementia post-stroke to be similar between lacunar (29%) and non-lacunar stroke (24%) (OR 0.72; 95% CI 0.43 to 1.20), but data were limited by short follow-up, subtype classification methods and confounding and there was substantial heterogeneity between studies.⁷¹ Some^72,73 but not all studies found PSCID to be more common in Oxfordshire subtype categories of total or partial anterior circulation infarction compared with lacunar or posterior circulation infarction. Few studies have compared cognitive outcomes across Trial of ORG 10172 in Acute Stroke Treatment (TOAST) subtype categories. A study in China found an increased risk of PSCID at 3 months post-stroke for large artery compared to lacunar stroke (OR 1.42; 95% CI 1.06-1.90).⁷³ A study in Norway found no differences in cognitive impairment at 3 and 18 months among TOAST subtypes in a global statistical test, but the risk was also lowest in lacunar stroke.⁶⁹

Hemorrhagic stroke (HS) is the most severe form of stroke, accounting for ~50% of stroke related morbidity and mortality.^74,75 Intracerebral hemorrhage (ICH) and aneurysmal subarachnoid hemorrhage (aSAH) account for over 90% of all HS events and lead to clinically significant cognitive decline in up to 50% of HS survivors.^76–79 Despite their impact on post-stroke cognitive outcomes, HS subtypes have been the topic of limited dedicated PSCID investigations to date. As previously mentioned, stroke anatomical location is likely to affect risk of PSCID. Because ICH/aSAH anatomical location is not defined by arterial or venous vascular territories (as in ischemic stroke), larger hematomas may well affect multiple brain area/networks relevant to PSCID. These considerations prompted incorporation of dedicated methodologies in DISCOVERY HS sub-study. In addition, specific pathophysiology mechanisms are likely to be relevant to PSCID risk after ICH or aSAH. Because of the pathophysiological role played by cerebral SVD in primary ICH, imaging and neuropathology markers of microvascular disease presence / severity have been associated with dementia risk after ICH.^76,77 However, the exact mechanisms linking SVD to PSCID after primary ICH remains poorly understood. In aSAH, delayed cerebral ischemia (DCI) has been identified as a key predictor of PSCID, but the biological mechanisms accounting for these findings are poorly understood at best.⁷⁹ DISCOVERY offers novel, targeted strategies to identify HS survivors at high risk for PSCID, explore relevant biological mechanisms, and therefore guide future investigative efforts aimed at improving long-term cognitive outcomes after HS.

Pathophysiology of Post-Stroke Cognitive Impairment/Dementia and Gaps in Knowledge

There is a consensus that cognitive impairment or dementia resulting from a vascular insult of any type constitutes vascular cognitive impairment (VCI) or vascular dementia.⁸⁰ Whether caused by progressive, insidious accumulation of subcortical microvascular pathology burden (including “silent” brain infarcts, diffuse white matter disintegration detected on T2-MRI as white matter hyperintensities (WMH), cerebral microbleeds, enlarged perivascular spaces, and cortical microinfarcts) or caused by a single stroke event that affects brain structures critical for cognition, vascular contribution to cognitive impairment and dementia is now broadly recognized as a highly prevalent yet significantly understudied entity.⁸¹ Further, since VCID may be modifiable, there is a great impetus for rapid advancement in knowledge in this field.⁸² Despite the established associations between stroke, neurodegenerative pathology, and dementia, there is however a lack of knowledge regarding the specific stroke subtypes (ischemic versus hemorrhagic, vessel types and sizes, anatomical locations) and magnitude of stroke events which cause PSCID (Figure 1).^83,84 Further, the complexity of interaction between the pre-existing pathology in the brain and an acute stroke insult serving as a trigger or accelerator of cognitive decline in the vulnerable brain requires the scope, size, and a systematic approach offered by DISCOVERY.⁸⁵

Figure 1. Brain susceptibility to post-stroke cognitive impairment and dementia.

A complex interaction between life-course exposures to multiple vascular risks and socio-demographic factors, pre-existing burden of cerebral small vessel disease and Alzheimer’s and related neurodegenerative pathology in the brain, the effect of strategic acute stroke lesions, and the mediating effect of genomic and epigenomic variation contribute to brain susceptibility (or resilience) to cognitive decline after stroke. An acute stroke event, whether ischemic or hemorrhagic, likely initiates or exacerbates/accelerates a series of pathological events leading to post-stroke cognitive impairment or dementia.

At the center of any brain dysfunction related to vascular insult is the neurovascular unit (NVU), the smallest building block of the brain parenchyma that emphasizes the unique, interdependent relationship between the brain cells and the cerebral vasculature.^86–88 The NVU’s ability to withstand or succumb to the insults related to cerebrovascular dysfunction, metabolic disarray, proteinopathy, or inflammatory responses constitutes what has been recently referred to as “brain health.”⁸⁹ Mechanisms of disease in this case includes the NVU as the common final pathway, but the specific etiologies define potentially identifiable targets for interventions to prevent brain’s susceptibility to dementia, stroke and other prevalent ways of health failures.⁹⁰

Several cell types including neurons, glia (astrocytes, microglia and oligodendrocytes), and vascular cells (endothelial cells, pericytes and vascular smooth muscle cells) comprise NVU.⁹¹ Some of the most critical roles of the NVU are to regulate cerebral blood flow in response to increasing functional demands and to ensure intact brain parenchymal milieu serving as the building block of the blood-brain barrier (BBB).^92,93 Bidirectional molecular transport across the BBB is highly regulated with exception of the essential deliveries (e.g., diffusion of oxygen and carbon dioxide, small liposoluble particles <400kDa). Associated with increased BBB permeability, microvascular dysfunction contributes to impaired mechanisms of clearance as well as “unauthorized entry” by neurotoxic molecules from the blood stream, which ultimately leads to neurodegeneration.^88,89 Specifically, cerebrovascular damage – such as that from the impact of life-course exposure to multiple vascular risk factors like hypertension, diabetes mellitus, or smoking – serves as the primary insult to the NVU and the BBB, which is both necessary and sufficient to initiate neuronal injury and neurodegeneration through resting CBF reductions and accumulation of neurotoxic plasma proteins (fibrinogen, plasminogen, thrombin etc).^94,95 However, it may also facilitate accumulation of the Alzheimer’s toxin, amyloid-β peptide (Aβ), in the brain through diminished clearance and increased production of Aβ.^96,97 These two pathways referred to respectively as Aβ-independent and Aβ-dependent, predictably interact in the brain and a largely considered to be synergistic in the brain that ultimately succumbs to cognitive impairment/dementia.^88,89

While there is general agreement that specific stroke events may either accelerate processes of AD/ADRD and microvascular dysfunction in a vulnerable brain, or unmask their pre-existing burden, leading to PSCID, the mechanisms of susceptibility to cognitive dysfunction after stroke remain unclear.^83,84 Conversely, data are lacking on the protective factors that constitute brain’s ability to withstand the effects of acute insult or allow the brain to recover fully and rapidly to its pre-stroke level of function, i.e. the mechanisms of brain resilience in stroke. We identified a number of important questions to be addressed systematically by DISCOVERY to bridge the existing gap in knowledge in the field of PSCID (Table 1).

Table 1.

Knowledge gaps in post-stroke cognitive decline to be addressed by DISCOVERY.

Knowledge gap	Plans to address in DISCOVERY
The role of an acute insult on outcomes after stroke
1.The effects of acute stroke type (ischemic and hemorrhagic), lesion size, and location	Detailed acute imaging characterization including vascular territory, vessel size, vascularity status assessment, infarct and hematoma characteristics including cortical-subcortical, lobar-deep, multiplicity of lesions, brain structures affected, hemorrhagic transformation, edema, vascular occlusions, joint multi-modal modeling using machine learning and radiomics analysis
2. Etiologic stroke subtype and the combined effect of vascular risk profile, antecedent co-morbidities, and functional baseline	Analysis of age, sex, race, ethnicity and other sociodemographic determinants, vascular risk factors, other co-morbidities, baseline functional status, social support structure
3. The effects of acute stroke magnitude (clinical stroke severity) and post-stroke course (hospitalization and long-term)	Include acute treatments, complications, discharge destination, recurrent events, social supports, depression, functional recovery.
4. The interaction with pre-existing burden of disease	MRI and CT markers of white matter disease and other SVD, brain volume and atrophy, ventricular size, prior silent brain infarcts, SVD score, vascularity
5. The effect of genetic susceptibility	Whole exome sequencing at baseline, DNA methylation at Follow-up 1 & 2 (genome-wide and epigenome-wide association studies)
6.The role of blood biomarkers of neurodegeneration, inflammation, neuroaxonal and vascular injury	Blood draw at baseline, and at 3-6 and 18 months after stroke
The mechanisms underlying PSCID
1.Synergistic effects of AD features (Aβ and tau)	PET and plasma biomarkers with acute stroke features
2.Progression of chronic neurodegeneration and SVD pathology in relation to PSCID	3T MRI with morphometric, structural and functional connectivity, and perfusion MRI assessment in relation to acute stroke features, stroke subtypes, patient characteristics, and personalized “multi-omics” signature
Personalized approach to prediction and prevention of PSCID
1.A practical “bedside” clinical tool	Development of the DISCOVERY Score, a clinical prediction tool
2.A tool for selection of the patients into clinical trials	Developing a sophisticated, granular characterization of disease subtype using multimodal personalized biomarker data

Abbreviations: AD – Alzheimer’s disease, CT – computer tomography, MRI – magnetic resonance imaging, PET – positron emission tomography, PSCID – post-stroke cognitive impairment and dementia, SVD – small vessel disease

Scientific Objectives & Specific Aims of DISCOVERY

The central hypothesis in DISCOVERY is that a stroke event (including acute ischemic stroke, AIS; ICH; or aSAH), in an already vulnerable brain, precipitates cascading pathological events ultimately resulting in PSCID. To address this hypothesis, we proposed the DISCOVERY Network, a collaborative of 30 clinical sites with access to large diverse acute stroke populations and expertise and capacity to systematically assess for PSCID (Figure 2). This is the first time a project of such scale has brought together the components of a multi-center acute stroke clinical trial (including screening and enrollment of patients at the time of index stroke hospital admission) and a longitudinal nested-cohort study with cognitive and functional assessments and multidimensional/multimodal research tools, including blood and imaging biomarkers. Developing the scientific framework of DISCOVERY required expertise in acute stroke, SVD, VCID, AD/ADRD, health disparities, and multi-center research.

Figure 2. The DISCOVERY Network.

The network is comprised of four study cores (Administrative, Recruitment and Retention, Statistical, and Repository) and thirty clinical performance sites across the United States.

Mandated by the National Institutes of Health, the Network’s overarching goals are to determine: 1) which specific subsets of stroke events cause (or do not cause) PSCID and 2) which additional demographic (sex, race, and ethnicity), clinical factors and comorbidities that synergize with the acute stroke to result in or prevent PSCID. Expanding well beyond these goals, the overall scientific objective of DISCOVERY is to elucidate mechanisms of brain resilience and/or susceptibility to PSCID in diverse US populations based on complex interplay between life-course exposure to multiple vascular risk factors, pre-existing burden of SVD and AD/ADRD pathology, the effect of strategic acute stroke lesions, and the mediating effect of genomic and epigenomic variation. Specifically, the DISCOVERY aims to: 1) identify the independent and combined effects of the acute stroke lesion, pre-existing SVD and AD/ADRD burden, and baseline markers of brain resilience on PSCID, 2) determine the role of acute stroke as a critical factor in accelerating AD/ADRD and SVD pathology leading to PSCID, and 3) develop generalizable personalized-medicine models for individual patient outcome prediction and patient selection for clinical trials.

Overall Study Design and Timeline

DISCOVERY is a prospective, multi-center, observational, cohort study that will enroll 8,000 non-demented race/ethnically diverse patients with incident AIS, ICH, or aSAH admitted across 30 high-volume US stroke centers and will follow the participants for ≥2 years, with serial cognitive and functional evaluations, with subgroups undergoing detailed MR and PET imaging (Figure 3). Recruitment projections and statistical power considerations in DISCOVERY are included in supplemental material (Figure I). All patients will have standard-of-care (SOC) stroke evaluation and treatment on admission, and will undergo pre-stroke dementia screening before enrollment. During the inpatient phase, subjects will undergo baseline cognitive and functional outcomes assessments (Table 2). Baseline SOC parenchymal and vascular imaging, clinical variables and blood specimens for biomarker studies will be entered in the study database. In the outpatient phase, participants will be followed for ≥2 years through a series of in-person and telephone visits. Every participant who initially failed cognitive screen during their hospital admission will be re-screened by telephone within 6 weeks of the index stroke and will re-enter the study in the outpatient phase, if eligible. During the outpatient phase, all surviving subjects will be seen in two in-person clinic follow-up visits (at 3 to 6 months and 18 months). All clinic follow-up visits will occur at the DISCOVERY Network Clinical Performance Sites (CPS) that will be designated as Tier 1, 2, or 3 (Figure 2) based on infrastructure, resources, and proficiency to perform in-depth neurocognitive assessments, advanced 3T protocol MRI scans, and Aβ and tau PET imaging. Every DISCOVERY participant will undergo serial centralized telephone-based cognitive and functional outcomes assessments conducted by the Telephonic Assessment Center (TAC) team at the Mayo Clinic (Jacksonville, FL) at 12 months post-stroke, and annually thereafter for the study’s duration. DISCOVERY participants who are unable to attend in-person assessments at 3 to 6 or 18 months will be assessed via Virtual Visit (teleneurology), or by phone (TAC), in order to compile the minimum necessary PSCID dataset for analysis. The DISCOVERY study design offers solutions for a number of key methodological challenges often faced by the previous studies of post-stroke cognitive impairment and dementia (Table I).

Figure 3. The DISCOVERY Study schema.

Patients diagnosed with acute ischemic stroke, intracranial hemorrhage, and aneurysmal subarachnoid hemorrhage will be enrolled during the index stroke admission (Inpatient Phase), undergo baseline cognitive and functional outcomes assessments, and will be followed for ≥2 years through a series of in-person and phone research visits (Outpatient Phase). During the outpatient phase, all surviving subjects will be seen in two clinic follow-up visits (at 3-6 months and 18 months). All clinic follow-up visits will occur at the DISCOVERY Network Clinical Sites that will be designated as Tier 1, 2, or 3 based on infrastructure, resources, and proficiency to perform in-depth neurocognitive assessments, advanced 3T protocol MRI scans, and PET imaging. Every DISCOVERY subject will have serial centralized phone-based cognitive and functional outcomes assessments at 12 months post-stroke, and annually thereafter for the study’s duration.

Table 2.

Cognitive and functional assessments in DISCOVERY.

	Baseline		Clinic Follow-Up 1 & 2		Telephonic Assessment
	Cognitive	Functional	Cognitive	Functional	Cognitive	Functional
Tier 1	DISCOVERY t-MoCA,* Line cancellation test	Pre-stroke and discharge mRS, BI, NIHSS, PROMIS-10 Global Health,	DISCOVERY t-MoCA*	NIHSS, mRS, BI, PROMIS-10 Global Health, SSQ-short, PHQ-9	DISCOVERY t-MoCA,* AD8 (informant)	T-NIHSS, mRS, PROMIS-10 Global Health, SMF
Tier 2/3			DISCOVERY full cognitive battery,** ORRT

^*DISCOVERY t-MoCA Battery: MoCA (domain: global); CERAD Word List Memory Task (domain: immediate, delayed memory); Number Span Test - Forward/Backward (domain: attention).

^**DISCOVERY Full Cognitive Battery: MoCA (domain: global), CERAD Word List Memory Task – Immediate/Delayed (domain: memory); Benson Complex Figure Copy – Immediate/Delayed (domain: visuospatial); Trail Making Test – A/B (domain: executive function/processing speed); Letter Digit Substitution (domain: : executive function/processing speed); Number Span Test - Forward/Backward (domain: attention); Craft 21 Story Recall – Immediate/Delayed (domain: memory); Verbal Fluency – Phonemic (domain: executive function/processing speed); Category Fluency (domain: language); CES-D (domain: depression); CDR (informant) (domain: cognitive status); MBI-C (informant)(domain: neuropsychiatric status).

Abbreviations: AD – Alzheimer’s Dementia, BI – Barthel index, CDR – Clinical Dementia Rating scale, CES-D – Centers for Epidemiologic Studies Depression scale, MBI-C – Mild Behavioral Impairment Checklist, MoCA – Montreal Cognitive Assessment scale, NIHSS – National Institutes of Health Stroke Scale, ORRT – Oral Reading Recognition Test, PHQ – Patient Health Questionnaire, PROMIS – Patient-Reported Outcomes Measurement Information System, SMF – Subjective memory Form, SSQ – Social Support Questionnaire, t-MoCA – telephonic MoCA, T-NIHSS – telephonic NIHSS.

Approach to Cognitive Assessment in DISCOVERY

Evaluation of cognition in the post-stroke setting must consider: 1) the role and value of repeated cognitive assessment; 2) the timing of the assessment and the pre-stroke cognitive status, and 3) the problem of attrition, in general, cognitive change, as can be analyzed with repeated cognitive assessment, allows less confounded measurement of cognitive performance, due to less reliance on sociodemographic factors than when cognition is measured at a single time point. Cognitive performance at a single time point is related to social and demographic factors, including educational level⁹⁸ and childhood cognitive function;⁹⁹ this relationship is especially pronounced in Blacks (vs Whites), in whom the benefit in performance from education is especially large.⁸ Risk of dementia is >60% higher in the Atherosclerosis Risk in Community (ARIC)’s least vs most educated participants,¹⁰⁰ but when change in cognition is considered, education does not impact 20-year cognitive decline.¹⁰¹ Consideration of cognitive change, however, is not possible for pre to post-stroke evaluation in a study that recruits at the time of the incident stroke, such as in DISCOVERY, but repeated assessments over the years post-stroke will allow consideration of these trajectories. In addition, it is critical to consider appropriate normative data, to more accurately adjudicate PSCID in relevant disparities populations.

To account for pre-stroke cognitive status, because DISCOVERY is a study of post-stroke dementia, it is critical to exclude any individuals who already have dementia at the time of stroke, even if not formally diagnosed by a physician. The use of the IQCODE¹⁰² allows to exclude those individuals with probable pre-existing dementia, and truly measure incident PSCID after index stroke. As data are collected on individuals throughout the course of the study, their cognitive performance will be tracked and adjudicated, using standardized definitions of MCI and dementia whenever possible, with the understanding that the post-stroke setting may require distinct consideration of time course and functional impairment, as many patients experience functional impairment for non-cognitive reasons post-stroke. The first year post-stroke may represent a time when early cognitive impairment is identified but when we cannot confidently give a diagnosis of dementia, and in DISCOVERY, we hope to understand the link between these early changes and a persistent dementia diagnosis beyond that time period.

Finally, attrition is a highly important consideration in any longitudinal study of cognition. Studies evaluating long-term cognitive trajectories are potentially susceptible to attrition, from death as well as dropout. Importantly, cognitive impairment is a risk factor for attrition,¹⁰³ as is stroke, even among enrollees in clinical trials.¹⁰⁴ Thus, the potential risk of attrition in the stroke population is especially high, and if those individuals at greatest risk for a poor outcome are less likely to attend visits, dementia rates could appear lower, as shown in the OXVASC study, where 17% of individuals with TIA and stroke who came to full clinic follow-ups were diagnosed with dementia, vs 45% in the TIA and stroke population without complete in-person assessment.¹⁰⁵ Other studies report similar discrepancies, with higher rates of dementia among the non-examined. ^106,107 Thus, evaluation of cognitive outcomes in a stroke cohort must include not only a robust retention plan, but also appropriate statistical methods to account for attrition, as some of it may be unavoidable (e.g. due to death or other competing illness). Such methods include inverse probability of attrition weighting (IPAW), which up-weights a study population to its original makeup.^108,109 Multiple Imputation with Chained Equations (MICE), another approach to addressing attrition, reduces the impact of missingness (from attrition or death) by imputing results, including outcome, for missing data points.

Building a Data Management System for Future Success

The management and sharing of multi-modal data generated by multi-site projects is a challenging endeavor. Making data useful within the project means creating a data management infrastructure that integrates, rather than silos, data through meticulously documenting, collecting and making available metadata from multiple project domains. Interest in data sharing has driven the construction of repositories and tools that recognize the challenges of integration and the importance of well-documented metadata and has led to guiding principles such as FAIR from Force11 that address these issues.¹¹⁰ FAIR stands for “findable”, “accessible”, “interoperable”, and “reusable”. The DISCOVERY Repository follows the FAIR principles and we expect that adherence to those principles will maximize utilization of this unique data set.

The DISCOVERY Repository system provides a web portal that serves as the main entry point into the project’s data management capabilities. The portal provides project information, access to project documents, as well as modules to register the acquisition of imaging and biofluid data and record clinical and neuropsychological assessment instrument data. The Repository is built to provide fully integrated access to all the data generated in DISCOVERY. The DISCOVERY Repository leverages and integrates the Entrepôt image management system, NeuroBank electronic data capture system and Virtual Biorepository (Figure II), with the infrastructure and experience of the Biospecimens repository developed to build an integrated management infrastructure that can store, move, and share data from all DISCOVERY domains. In addition, DISCOVERY uses the Globus data transfer service for movement of project data, both for the collection of data from the Clinical Performing Sites and for the distribution to the various Cores and analysis centers. We also deploy an extensive suite of tools to for quality control for data, biospecimens and images as well as tools and workflows for integrity checking of the integrated data.

The scientific aims of the DISCOVERY Repository are to: 1) create an integrated data management infrastructure that follows the best practices for data management established by MarkVCID consortium and NIH guidance documents, 2) develop a repository infrastructure that will allow data to be “findable” and “accessible”, and 3) create a metadata infrastructure that will allow data to be “interoperable” and “reusable”. In this last aim, the DISCOVERY repository leverages work by other NIH repositories and ontology-development groups to create metadata that maximizes interoperability, using well-established terms and common data elements wherever possible, so that data consumers are able to understand and successfully reuse DISCOVERY data. The rich set of metadata will be used not only to aid in the development of processing and analysis pipelines, but will also be used by both the Statistics and Recruitment and Retention Cores to generate operational reports and track participants.

Neuroimaging Biomarkers of Post-Stroke Cognitive Impairment/Dementia

Neuroimaging is a key research tool in DISCOVERY (Figure III), with neuroimage analysis strategy built upon the existing body of evidence regarding the effects of pre-existing MRI-detectable brain disease and its role in functional and cognitive outcomes following a superimposed acute stroke event.

Acute image analysis.

The acute images obtained during hospitalization of every eligible stroke patient in DISCOVERY will serve as the largest to-date, prospective clinical stroke imaging repository linked to the rich clinical and detailed outcome assessment data. An acute infarct on DWI MRI will be confirmed during enrollment of AIS patients, while hemorrhagic stroke patients are expected to have an ICH or aSAH on computer tomography (CT)/CT angiogram (CTA). Sequential images obtained during the index admission or imaging including additional modalities (MR angiography, digital subtraction angiography, CT or MR perfusion images) will be also collected, if available. These data will be critical in several lines of analytical inquiries within DISCOVERY including: (1) the effect of size and topography of acute stroke lesion on stroke severity and outcomes, (2) the effect of pre-existing brain pathology on stroke severity and outcomes, (3) the contribution of imaging to stroke subtyping and disease classification in DISCOVERY, and (4) the interaction between clinical and neuroimaging phenotypes on outcomes in complex models of novel constructs such as effective brain reserve, rich-club organization, lesion network mapping, vascularity assessment, and brain health assessment. These data will be analyzed using multi-modal, automated, big-data approaches to large, multi-center imaging repositories comprised of 1.5T MRI and CT scans obtained as part of hospital care and will utilize novel computational approached developed to address the challenges of this type of high-volume, low-quality imaging data.

Longitudinal MRI analysis and PET imaging.

Cerebrovascular and AD/ADRD pathologies frequently coexist, ¹¹¹ altering cognitive trajectories of aging adults¹¹². Data are however lacking on the impact of acute stroke event on either accelerating or unmasking – the pathologic cascades in the brain leading to PSCID. Using research MRI and PET protocols, we will focus on discerning the underlying brain mechanisms that lead to PSCID.

3T MRI imaging.

Longitudinal analyses of 3T research protocol MRI for a subset of DISCOVERY participants (n=2000) with representative stroke subtypes and health disparity populations at two timepoints (3-6months and 18 months post-stroke) presents an experimental framework for assessment of the MRI markers of cognitive decline. We will acquire high quality imaging data using multi-site implementation of MR protocols that will leverage cutting edge image acquisition (e.g. multi-band acceleration) and also employ reproducible processing methodologies that can measure subtle changes across the two time-points. Specifically, we will acquire the following data: structural MRI (sMRI) for brain structure changes, FLAIR MRI and SWI for vascular assessments, diffusion MRI for microstructural changes and structural connectivity changes (using traditional DTI models as well as advanced biophysical models such as neurite orientation and dispersion and density imaging, NODDI), resting-state functional (rsfMRI) for functional connectivity changes, and arterial spin labeling perfusion (ASL) methods for perfusion changes. These detailed structural and functional changes on 3T MRI scans will play an important role for mapping post-stroke brain changes and identifying key features predictive of PSCID.

PET imaging.

Leveraging recent standardization efforts^15,113–116 and using the emerging plasma as well as PET biomarkers for AD, we have a unique opportunity to investigate how AD pathophysiology and the acute stroke impact cognitive decline and neurodegenerative processes. In a subset of 350 subjects selected from the nested cohort of 2,000 stroke survivors at Tier 3 Network sites, we will conduct PET imaging at two time points (3-6 and 18 months) to assess for synergistic effects of acute stoke features and AD biomarkers (Aβ and tau) on cognitive decline and burden of AD/ADRD pathology progression. This longitudinal PET substudy is complementary to the multi-tier approach to testing the hypothesis regarding acute stroke events accelerating, or unmasking, the burden of AD/RD pathology in the brain and would allow for a rare mechanistic evaluation of underlying biology of PSCID. In addition to testing the effect of baseline plasma amyloid and tau on cognitive decline in the entire DISCOVERY cohort (Tier 1), and the relationship between plasma amyloid and tau and structural/functional MRI markers of SVD/ND progression at Tier 2 studies, we will also use composite metrics for baseline amyloid and tau levels on PET and evaluate the association of the PET biomarkers with stroke event in predicting future PSCID, trajectories of cognitive decline, and association with MRI disease progression (Tier 3). This study will also test if the acute stroke event accelerates the accumulation of amyloid and tau longitudinally. The availability of both plasma biomarkers and PET scans in the same subjects will also allow us to correlate the PET composite metrics for amyloid and tau with plasma amyloid and tau measures to validate the plasma biomarkers we use in the first two tiers of hypothesis testing.

DISCOVERY Study Fluid Biomarker Strategy

Fluid-based biomarkers are readily accessible to a greater subset of the healthcare environment and offer a promise of personalized medicine¹¹⁷, when tested as a targeted, hypothesis-driven tool for screening and validation of underlying disease biology in stroke populations. The DISCOVERY study’s fluid biomarker strategy is designed to interrogate the contribution of multiple potential pathways that may converge to drive PSCID (Figure 4):

Figure 4. The DISCOVERY Network Fluid Biomarker Strategy.

All enrolled subjects (Tier 1) will undergo a baseline measure of circulating biomarkers in the proposed pathogenic categories influencing post-stroke cognitive impairment/dementia. Tier 2 subjects will participate in longitudinal biomarker assessment using emerging biomarkers including several biomarker kits established by the MarkVCID Consortium.

Inflammation.^118,119

Emerging data from MarkVCID and ARIC suggests that an aggregate serum measure of peripheral inflammatory markers is associated with WMH and silent cerebrovascular injury.^120,121 Therefore, baseline measurements of inflammation will be critical to identifying individuals at risk for PSCID. Moreover, the degree of inflammation appears to vary among racial and ethnic groups;^122,123 thus, measures of inflammation may serve an valuable role in evaluating differences in PSCID among racial and ethnic groups.

Neurodegeneration.

Concurrence of AD/ADRD and cerebrovascular (SVD) pathologies on post-mortem studies and neuroimaging of patients with stroke and dementia^124–126 compels our team to study the burden of ND changes at stroke onset and how it modifies the development of PSCID. CSF measures of neurodegeneration are quite reliable ¹²⁷, however, obtaining CSF in post-stroke subjects is burdensome, potentially dangerous (given the use of anticoagulants/dual antiplatelet agents), and likely to introduce a selection bias for less-affected individuals. Existing data suggest that low plasma β-amyloid ratios (42:40)^128,129 and plasma total and phosphorylated isoforms of tau (p-tau181, p-tau217)^130–132 are predictive of cognitive decline in cognitively normal or MCI subjects.^133,134

Neuro-axonal injury.

Plasma neurofilament light chain (pNfl) level is emerging as a sensitive indicator for neuro-axonal injury in acute stroke, SVD¹³⁵ and AD/ADRD^136,137. While pNfl lacks specificity, its utility in evaluating PSCID will be to quantitatively define the baseline burden of SVD and/or AD/ADRD pathologies while also helping to determine if stroke alone triggers a progressive axonal degeneration that identify those at increased risk for PSCID.

Vascular injury.

Individuals presenting with stroke often have an underlying risk profile that favors ongoing cerebrovascular injury due to hypertension, diabetes, other VRFs, also linked with poor cognition.⁶⁵ In addition, endothelial-specific (CD31+/CD146+) exosomes (EDEs) shed with higher levels of brain-specific cargo proteins are enriched in individuals with WMH¹³⁸ and EDEs are also undergoing biomarker validation in MarkVCID. Understanding the contribution of ongoing endothelial injury can drive therapeutic interventions aimed at tightening risk factor control thereby minimizing the risk for PSCID.

A targeted fluid-based biomarker strategy offers a novel, thorough, hypothesis-driven approach to insight into pathophysiologic mechanisms of PSCID and building personalized outcome prediction models.

Genetics and Genomics Strategy in DISCOVERY

Mechanisms of brain resilience to PSCID are based on the complex interplay between genetic susceptibility and environmental factors (life-course exposure to multiple vascular risks (VRFs), or use of protective lifestyle factors) that can be interrogated using a systematic and informed “omics” and fluid biomarker strategy. The DISCOVERY investigators have been instrumental in defining the landscape of genetic discovery in large-scale, population-based studies of stroke¹³⁹, VCID-related phenotypes^140,141, and dementia, including multiethnic genome-wide association study (GWAS)^142–146, and a recent GWAS highlighting the role of common variation in post-stroke outcomes^147,148.

GWAS and next-generation (whole exome sequencing, WES) studies have made substantial progress towards identifying the genetic loci that influence risk of stroke and its subtypes¹⁴⁹, AD¹⁵⁰, MRI-defined endophenotypes of VCID^143,151, and their VRFs. Certain genes such as APOE¹⁵², FOXF2¹⁴⁶ and ATP5H¹⁵³ have been independently related to both the risks of stroke and AD. Such genetic information provides powerful means to examine mechanisms of PSCID risk and trajectories and their link to VRFs, stroke subtypes, and AD/ADRD and MRI-related endophenotypes.

Injury and recovery after stroke are dependent on a set of processes that involve altered orchestrated gene expression.

The epigenome describes the set of dynamic molecular modifications that regulate gene expression independent of gene sequence.¹⁵⁴ DNA methylation (DNAm) is the most studied epigenetic modification, is “plastic” and highly sensitive to interventions and changes over time.^155,156 Using epigenome-wide DNAm profiles in African-American ARIC participants, a unique DNAm signature (accelerated epigenetic aging) has been identified in association with increased WMH burden, an MRI-phenotype highly relevant to VCID and post-stroke dementia.¹⁴⁵ Changes in global DNAm following a stroke have been described and suggest that DNAm influences cell death post-stroke through silencing of neuroprotective genes, but is also implicated in repair mechanisms such as synaptic plasticity and neurogenesis.¹⁵⁷ Whether these global changes in DNAm influences trajectories of PSCID has not been investigated, nor have specific epigenetic signatures of the trajectories of PSCID been identified.

In DISCOVERY, we will examine whether PSCID risk after specific stroke subtype is linked to genetic susceptibility (detected via WES, GWAS, and epigenome-wide association studies). We will undertake genetic and multi-omic studies in DISCOVERY to: (a) identify the genetics and biological pathways that determine resilience and recovery from an acute stroke, thus potentially identifying novel biomarkers and drug targets; (b) utilize polygenic risk scores as prognostic indicators for risk stratification in clinical trials and as aggregate measures of genetic propensity to explore gene-environment interactions; (c) use Mendelian randomization strategies, wherein genetic determinants of a putative risk factor or biomarker (related to stroke outcomes) are used as instrumental variables to examine whether observed association is causal. Identifying the genetics of recovery from stroke, DNAm signatures of PSCID trajectories, and integrating the “omics” data has the potential to uncover novel molecular targets that reflect the contribution of genetic and non-genetic risk factors to post-stroke outcomes.

Summary: The Promise of DISCOVERY

The vision of DISCOVERY is to dramatically reduce the rates on post-stroke disability by preventing cognitive decline and dementia in high-risk US populations based on their stroke events type or the underlying vulnerability to PSCID. The limitations to consider for this study in achieving its aims include potential selection bias in enrollment and follow-up, limited statistical power for subgroup analyses (e.g., PET imaging), the constraints of cognitive assessment in aphasic patients, and as of recent, the potential impact of the COVID-19 pandemic on the diverse study populations. However, the fundamental success of this project will be ensured by the key innovations and central elements of the study including the collective experience and expertise of the study scientific leadership team, the geographically and populationally diverse high-volume network clinical performance site structure, and the innovative two-pronged approach of combining a large-scale prospective cohort study with multi-tiered highly mechanistic sub-studies nested within. Within the framework of DISCOVERY, a rigorous, multimodal, and systematic characterization of stroke events and comorbid factors that increase susceptibility to PSCID will serve as the starting point for future work developing personalized post-stroke interventions for PSCID, including prevention, treatment and rehabilitation. Further, a large-scale study such as DISOCVERY will allow consideration of race and ethnicity in PSCID and address key disparities in rates of not only stroke but also dementia.

Non-standard Abbreviations and Acronyms

AD/ADRD

Alzheimer disease/Alzheimer disease-related dementia

AIS

acute ischemic stroke

ARIC

the atherosclerosis risk in community study

ARCADIA-CSI

the atrial cardiopathy and antithrombotic drugs in prevention after cryptogenic stroke – cognition and silent infarcts trial

ASL

arterial spin labeling

aSAH

aneurysmal subarachnoid hemorrhage

BBB

blood-brain barrier

CPS

clinical performance site

CREST

the carotid revascularization endarterectomy versus stenting trial

CSF

cerebrospinal fluid

CT/A

computer tomography/angiography

DISCOVERY

the determinants of incident stroke cognitive outcomes and vascular effects on recovery study

DNAm

deoxyribonucleic acid methylation

DTI

diffusion tensor imaging

DWI

diffusion weighted imaging

EDE

endothelial-specific endosomes

FLAIR

fluid attenuated inversion recovery

GWAS

genome-wide association study

HS

hemorrhagic stroke

ICH

intracerebral hemorrhage

IPAW

inverse probability of attrition weighting

IQCODE

the informant questionnaire on cognitive decline in the elderly

HYVET-COG

the hypertension in the very elderly cognition trial

MCI

mild cognitive impairment

MICE

multiple imputation with chained equations

MoCA

the Montreal cognitive assessment test

MMSE

mini-mental state exam

MRI

magnetic resonance imaging

ND

neurodegeneration

NIH

National Institutes of Health

NODDI

neurite orientation and dispersion and density imaging

NVU

neurovascular unit

OXVASC

the Oxford vascular study

PET

positron emission tomography

pNFL

plasma neurofilament light chain

PSCID

post-stroke cognitive impairment and dementia

SOC

standard of care

SPRINT-MIND

the systolic blood pressure intervention trial

SVD

small vessel disease

SWI

susceptibility weighted imaging

TAC

the telephonic assessment center

TIA

transient ischemic attack

TOAST

the trial of ORG 10172 in acute stroke treatment

VCID

vascular cognitive impairment/dementia

VISP

the vitamin intervention for stroke prevention trial

VRF

vascular risk factor

WES

whole exome sequencing

WMH

white matter hyperintensity

US

the United States

APPENDIX (Study Group): The DISCOVERY Investigators

Kevin Barrett, MD (Mayo Clinic Jacksonville);

Alessandro Biffi, MD (Massachusetts General Hospital);

Bernadette Boden-Albala, PhD (UC Irvine);

Myriam Fornage, PhD (University of Texas Houston);

Mark Etherton MD, PhD (Massachusetts General Hospital);

Polina Golland, PhD (Massachusetts Institute of Technology);

Jonathan Graff-Radford, MD (Mayo Clinic Rochester);

Jason Hinman, MD, PhD (University of California Los Angeles);

Cliff Jack Jr, MD (Mayo Clinic Rochester);

Jayashree Kalpathy-Cramer, PhD (MGH Martinos Center for Biomedical Imaging);

David Knopman, MD (Mayo Clinic Rochester);

Steven Kittner, MD (U of Maryland);

Val Lowe, MD (Mayo Clinic Rochester);

Jennifer Manly, PhD (Columbia University);

Thomas Mosley, PhD (University of Mississippi);

Ronald Petersen, MD (Mayo Clinic Rochester);

Robert Rissman, PhD (University of California San Diego);

Markus Schirmer, PhD (Massachusetts General Hospital);

Kristin Schwab (Massachusetts General Hospital);

Sudha Seshadri, MD (University of Texas San Antonio);

Alexander Sherman (Massachusetts General Hospital);

Prashanthi Vemuri, PhD (Mayo Clinic Rochester);

Anand Viswanathan, MD, PhD (Massachusetts General Hospital).