A population neuroscience approach to the study of cerebral small vessel disease in midlife and late life: an invited review

Abstract

Aging in later life engenders numerous changes to the cerebral microvasculature. Such changes can remain clinically silent but are associated with greater risk for negative health outcomes over time. Knowledge is limited about the pathogenesis, prevention, and treatment of potentially detrimental changes in the cerebral microvasculature that occur with advancing age. In this review, we summarize literature on aging of the cerebral microvasculature, and we propose a conceptual framework to fill existing research gaps and advance future work on this heterogeneous phenomenon. We propose that the major gaps in this area are attributable to an incomplete characterization of cerebrovascular pathology, the populations being studied, and the temporality of exposure to risk factors. Specifically, currently available measures of age-related cerebral microvasculature changes are indirect, primarily related to parenchymal damage rather than direct quantification of small vessel damage, limiting the understanding of cerebral small vessel disease (cSVD) itself. Moreover, studies seldom account for variability in the health-related conditions or interactions with risk factors, which are likely determinants of cSVD pathogenesis. Finally, study designs are predominantly cross-sectional and/or have relied on single time point measures, leaving no clear evidence of time trajectories of risk factors or of change in cerebral microvasculature. We argue that more resources should be invested in 1) developing methodological approaches and basic science models to better understand the pathogenic and etiological nature of age-related brain microvascular diseases and 2) implementing state-of-the-science population study designs that account for the temporal evolution of cerebral microvascular changes in diverse populations across the lifespan.

INTRODUCTION

Small vessels in the brain undergo remodeling and exhibit signs of damage that accumulate with older age. Common findings on histopathological examination of postmortem brains of older individuals (i.e., those 60 yr and older) include tortuosity of venules and arterioles, capillary rarefaction, small infarcts, microhemorrhages, and enlarged perivascular spaces. These features are taken as manifestations of aging microcirculation. With the advent of magnetic resonance imaging (MRI), a wider range of cerebral age-related changes has been identified. These include parenchymal changes, such as infarcts, microhemorrhages, lacunes, periventricular spaces and also abnormalities of cerebral blood flow (CBF), white matter hyperintensities (WMHs), and reduced fiber alignment (via diffusion tensor imaging). Although studies relating MRI and histopathological features of age-related cerebral microcirculation are sparse, the above MRI measures are nonetheless considered lesions of presumed vascular origin and are grouped together and referred to as “neuroradiological markers of cerebral small vessel disease” (cSVD), a convention that we hereby adopt for simplicity and for consistency with prior reports (149).

The neuroradiological features of cSVD have become the focus of intense study. Numerous reports, reviews, and meta-analyses have been published on the relationships between the radiological markers of cSVD and negative health outcomes, including stroke, dementia, disability, and mortality. Although major goals in this field encompass developing a precise understanding of cSVD pathogenesis and identifying potential prevention and treatment targets, many pressing barriers to these goals remain unaddressed.

For example, there is evidence that blood pressure and other cardiovascular parameters are associated with at least some of the many radiological manifestations of cSVD; however, older adults with unremarkable cardiovascular history also commonly present with cSVD. This dissociation indicates the existence of multifactorial causal pathways that are more complex than of purely cardiovascular origin. Additionally, unlike other pathologies that progress over time, several of the neuroradiological measures of cSVD appear stable over time and occasionally even “spontaneously” regress. Finally, and likely related to the above, treatments targeting cardiovascular risk factors have yielded only modest and inconsistent reductions in radiological markers of cSVD for vascular patient populations (59).

In view of these and related issues, the aims of this review are to summarize key features and barriers to existing work and to propose a framework to answer open questions. The review focuses on age-related cSVD, namely, cSVD occurring in middle- and older-aged adults living in the community rather than cardiovascular patients per se.

These aims and foci are timely. First, intense technological advancement could revolutionize our understanding of the brain and brain aging. Recent technological advancements in neuroradiology can potentially capture direct changes affecting the small vessels earlier in the evolution of the pathological processes. Second, adults 60 yr and older are rapidly growing not only in number but also in terms of heterogeneity of racial and health characteristics. Increasing numbers of older adults from racial/ethnic minority backgrounds with diverse cardiovascular exposure histories are likely to add to the complex determinants of cSVD and its prediction, prevention, and treatment.

This review is articulated into four sections. neuroradiological markers and histological correlates of pathology of csvd provides a review of in vivo imaging measures and histopathological correlates of age-related cSVD. associations with risk factors addresses cSVD relationships with risk factors in the context of demographic and health-related measures. pathogenesis and clinical consequences of age-related csvd provides an overview of the pathophysiology of cSVD and its clinical manifestations. discussion: unanswered questions and research priorities concludes with a conceptual perspective on a novel research framework to advance the future study of cSVD.

METHODS

This review adopts the STRIVE definition of cSVD (149). By this definition, cSVD is thought to comprise a syndrome of radiological manifestations affecting brain areas with predominantly poor collateral vascularization (e.g., watershed areas, frontosubcortical regions, etc.). According to this scheme, radiological markers of cSVD include WMHs, microbleeds, silent brain infarcts (SBIs), and lacunes. Microbleeds were recently covered in an excellent review published in the American Journal of Physiology-Heart and Circulatory Physiology (137); hence, they are not examined here. We also do not discuss the parenchymal markers of cSVD included in the STRIVE definition, gray matter atrophy and diffusion-related indices (e.g., fractional anisotropy and mean diffusivity), because they appear to be removed from actual vascular pathology.

To account for the most recent methodological developments to characterize brain circulation, we include more novel markers of cSVD, which appear to be more direct measures of vasculature integrity. For example, endothelial dysfunction, an early stage marker of cSVD, can be measured by CBF at rest and changes in response to specific challenges (e.g., breath holding or hypercapnia) (149). We also discuss morphological measures (e.g., tortuosity, density) of arterioles via time-of-flight MRI and of small veins via ultrahigh field susceptibility-weighted MRI.

We carried out two literature searches to examine 1) correlations of histopathological and neuroradiological markers of cSVD and 2) risk factors for neuroradiological markers of cSVD. First, we carried out a PubMed search of all neuropathological studies where comparisons were made between histopathological and neuroradiological findings performed pre- or postmortem. Separately, we carried out a survey of the literature to understand the existing state of research and knowledge surrounding cSVD in otherwise healthy individuals. The search method to examine associations with risk factors is shown in Fig. 1, and it follows the PRISMA guidelines (http://www.prisma-statement.org/). We searched PubMed for reviews and original articles examining neuroradiological markers of cSVD (see Table 1 for definitions). When available, meta-analyses were used in lieu of the original articles. Studies that were cited by the articles thus included were further reviewed and included if appropriate. Vessel morphology articles were not found using these search terms, and, therefore, a hand search was carried out. We included studies examining community-dwelling, neurologically healthy individuals. Exclusion criteria were 1) hospitalized populations or disease state-only population without a control group, 2) sample size < 50, and 3) narrative reviews. For example, if a study dealt with recent stroke patients, that study was excluded. If a study was carried out only in individuals with diabetes and no control subjects, that study was excluded.

Fig. 1.

PRISMA flow diagram for the search completed on June 20, 2017. The process of identification for articles is included in the risk factor review.

Table 1.

Neuroradiological markers of cSVD

Commonly used neuroradiological markers

White matter hyperintensities

Silent brain infarcts

Lacunes (small sharply defined regions of tissue loss most frequently associated with deep gray and white matter)

Novel markers

CBF

CVR

Morphology of small arteries and veins (e.g. tortuosity and density)

cSVD, cerebral small vessel disease; CBF, cerebral blood flow (measured at rest); CVR, cerebrovascular reactivity (changes in CBF after stimulation).

NEURORADIOLOGICAL MARKERS AND HISTOLOGICAL CORRELATES OF PATHOLOGY OF cSVD

Here, we describe the histopathological correlates of WMHs and SBIs/lacunes, the most common radiological signs included under the umbrella term cSVD. We also outline recent neuroimaging technologies that have the potential to characterize the cerebral microcirculation but that have not yet been widely used.

White Matter Hyperintensities

WMHs are broadly defined as areas that appear as bright or “hyperintense” on T2 and T2* proton density and fluid-attenuated inversion recovery MRI. They may also appear as hypointense on volumetric T1-weighted MRI (Fig. 2). It is generally accepted that the variations in MR contrast that contribute to the visualization of WMHs correspond to vastly different parenchymal changes, as captured by histopathology. The purpose of this section is to review the neuropathological correlates of WMHs.

Fig. 2.

White matter hyperintensities. White matter hyperintensities vary greatly from incomplete infarcts with no tissue loss in patchy areas visible as subcortical punctate white matter hyperintensities (red) to periventricular smooth halos or caps (blue). Images were provided by Dr. Howard Aizenstein (University of Pittsburgh).

White matter abnormalities were detected in early CT scans of aged human brains. However, when MRI was developed a few years later, it quickly proved to be more sensitive. Once MRI became economically tractable, its routine use substantially heightened our awareness of these age-associated lesions. But what are they? Beginning in the late 1980s and proceeding up to today, there have been numerous studies attempting to define the neuropathological correlates of WMHs (4, 6, 12, 15, 16, 39, 41, 50, 60, 84, 85, 90, 99, 105, 114, 119, 124, 127–130, 156). These correlative studies used either premortem or postmortem MRI or some combination of techniques. To a reasonable degree of consensus, we can draw the following conclusions.

“Periventricular hyperintense signal” (also known as “rims” or “caps”).

These lesions that outline the edge of ventricles tend to be nonprogressive and have the greatest consensus regarding their neuropathological correlates. They consist of a denudation of ependyma from ventricular lining with variable subependymal gliosis (Fig. 3). They may also reflect periventricular necrosis and axonal degeneration with or without increased intraventricular pressure. Some have found reactive astrocytosis with remarkably well-preserved underlying myelin.

Fig. 3.

Deep white matter pathology of age associated with white matter hyperintensities. A and B: low-power paraffin section of the midfrontal cortex from the brain of an 80 yr old. A: hematoxylin and eosin-stained section shows unremarkable gray matter (GM) with adjacent white matter (WM) containing numerous dilated vascular spaces. Overall, the deep white matter staining is more pale than the U-fibers (UF) immediately underlying the gray matter. B: successive section stained with luxol fast blue (LFB), where the gray matter appears pale and white matter a deep blue. Numerous dilated perivascular spaces are noted in the white matter. C and D: high-power micrograph of individual blood vessels within deep white matter. C: the blood vessel lumen (L) is surrounded by mildly thickened media. The adventitia (AD) surrounding the vessel contains macrophages filled with golden hemosiderin. D: special stain for CD68 identifies the hemosiderin laden cells as macrophages (brown cells labeled with arrows). E: medium power micrograph of deep white matter showing a “moth-eaten” pattern imparted by the numerous dilated perivascular spaces. F: the lumen of blood vessels is surrounded by a thickly hyalinized vessel wall. G: some blood vessels have convoluted and collapsed lumina. H: ependymal surface (arrows) of ventricular wall overlay sclerotic vessels. *The ventricular surface focally demonstrates ependymal denudation. Images were provided by Dr. Clayton Wiley (University of Pittsburgh).

“Punctate and confluent deep WMHs.”

These lesions can progress over time from punctate to clustered to confluent. Punctate forms show a variety of small vessel pathologies (Fig. 3). Accentuation of perivascular (Virchow-Robin) spaces is associated with a variety of vascular pathologies, including convoluted vessels (arterioles vs. venules), sclerotic vessels (with or without amyloidosis), arteriolosclerosis, hyalinosis, or collagenosis.

When WMHs are associated with susceptibility artifacts, as detected on sequences sensitive to the magnetic properties of the iron in hemoglobin (e.g., gradient-recalled echo or susceptibility-weighted MRI; Fig. 3, C and D), these lesions are thought to involve microhemorrhage with extravasation of red blood cells, pericyte erythrophagocytosis, hemoglobin degradation, and hemosiderin deposition (57). How such changes might reflect “vascular integrity” or the blood-brain barrier remains to be fully elucidated.

Some WMHs are associated with complete or incomplete infarctions [hematoxylin and eosin (H&E); Fig. 4], including conspicuous lacunar type infarcts or less obvious microinfarcts. Given the limitations of sampling microinfarcts (i.e., infarctions not visible to the naked eye), it has been suggested that histopathology grossly underestimates the frequency of microinfarctions (e.g., for every microinfarct noted during routine brain sampling 500 microinfarcts are present in vivo; see Ref. 150). Reactive gliosis (astrocytic vs. microglial) is the sine qua non of central nervous system damage and repair (Fig. 4). Its associations with WMHs and perivascular changes are definitive proof that the histological changes are not fixation or processing artifacts but are instead representative of in vivo pathology.

Fig. 4.

Lacunar infarction of basal ganglia. A: coronal section through the frontal lobe of the brain of an elderly patient. Ventricles (VEN) are mildly dilated, and there is a lacunar infarct in the head of the caudate on the right side (between arrows). B: hematoxylin and eosin-stained section of the lacune demonstrates a central region of infarction (INF) with tissue loss. C: immunostain for glial fibrillary acidic protein (GFAP) (brown) shows prominent peri-infarct astrocytosis. D: higher-power micrograph of lacunar infarct showing intact ependyma overlying VEN. Tissue surrounding the infarct contains numerous sclerotic blood vessels with dilated perivascular spaces. E: immunostain for GFAP demonstrates pronounced astrocytosis. Images were provided by Dr. Clayton Wiley (University of Pittsburgh).

Myelin pallor (rarefaction) is a frequently described pathological finding associated with WMHs. Despite the pathological term being vague and nonspecific, it is frequently misquoted as “demyelination” or as other more specific but inaccurate descriptors. This has resulted in substantial confusion and inappropriate assumptions about the pathogenesis of myelin pallor. Histologically, it is manifest as decreased staining of white matter using any of a number of stains (e.g., H&E, luxol fast blue, staining for myelin basic protein, etc.). Like the MRI finding of WMHs, this histological finding of myelin pallor can be broadly interpreted as altered tissue water content with numerous potential etiologies. Simple brain edema (vascular or cytotoxic) can expand the extra- and intracellular space, leading to decreased stain density per square millimeter. Decreased numbers of axons (e.g., loss of fibers projecting through a region of rarefaction secondary to distal stroke) can decrease the tissue density and be associated with staining pallor. Demyelination could lead to decreased staining, but despite frequent use of this term in the literature, there is no convincing evidence that primary demyelination (that is, selective loss of myelin sheaths as opposed to loss of myelin sheaths secondary to loss of axons) is a true feature of WMHs. Based on measurements of select myelin proteins and alterations in vasoconstrictive (endothelin 1) and vascular growth factors (VEGF), Love and Miners (81) have recently put forward the hypothesis that white matter rarefaction is the result of decreased perfusion. Although an enticing explanation of white matter dysfunction, there is no formal proof for demyelination in aged white matter associated with WMHs.

Testing the hypothesis that WMH (or at least some of its components) is the result of demyelination is technically quite difficult. Love and Miners (81) have attempted to compare ratios of myelin-associated glycoprotein (MAG) to proteolipid protein (PLP) as a surrogate marker of demyelination. Although this is a convenient approach, there are multiple reasons other than demyelination that could lead to changes in the MAG-to-PLP ratio. Axons in the central nervous system become myelinated once they achieve a diameter of 0.5 μm. Because that is at the resolution limit of light microscopy, ultrastructural analysis [i.e., electron microscopy (EM)] is required to sensitively assess myelination. Most EM studies require perfusion-fixed tissue, which of course is essentially impossible to get in humans. Thus, given the ultrastructural artifacts introduced by postmortem studies, the optimal human study would require brain biopsy followed by glutaraldehyde fixation, a perhaps unattainable goal in humans. Therefore, testing the hypothesized role of demyelination in WMHs awaits the development of a good animal model that would permit in vivo perfusion and EM confirmation.

Lacunes and SBIs

Penetrating vessels supplying the deep white matter and basal ganglia lack collateral networks. Vasculopathy related to uncontrolled hypertension and diabetes, systemic inflammatory and/or hypercoagulable states, and other cardio- and neurovascular risk factors may result in spontaneous thrombosis of these terminal vessels, which supply small patchwork regions of parenchyma, resulting in lacunar type infarcts. These infarcts differ from territorial infarcts more typically associated with atherosclerosis and cardioembolic disease and manifest as small lake-like regions of parenchymal volume loss, or lacunes. Although these infarcts can frequently result in substantial neurological deficit, typically due to involvement of the corticospinal tracts and deep gray nuclei, smaller insults may be subclinical or only manifest as the overall burden increases. These SBIs of the white matter are frequently associated with WMHs that are typically described as chronic small vessel ischemic changes.

Cerebral Blood Flow

CBF and blood volume are substantially different between the gray and white matter of the brain, with 5- to 10-fold greater vasculature evident in the cortical and deep gray matter. Measures of perfusion have historically focused on cortical supply; white matter perfusion has received less attention. Methods of measuring blood flow include PET (H₂¹⁵O), considered the gold standard, contrast CT perfusion, contrast MRI methods, including dynamic susceptibility contrast, and noncontrast MRI arterial spin labeling. Global and regional blood flow can be altered by systemic or acute disease states as well as by normal cerebral function through neurovascular coupling. Compensation for systemic diseases such as carotid stenosis or heart failure may result in expanded blood volume and reduced blood flow to optimize oxygen extraction and delivery to tissue. Atherosclerosis, or hardening of the arteries, may impair this compensation and decrease neurovascular reserve. A loss of neurovascular reserve may render the supplied parenchyma vulnerable to both acute and chronic ischemic injury. Although ischemic injury to cortical gray matter may directly affect neurons, ischemic injury to white matter may primarily affect supportive glial cells, such as oligodendrocytes, and may result in impaired axonal function and/or axonal loss. It is unclear how the physiology of neurovascular compensation differentially impacts cortical gray and white matter supply.

Small Vessel Morphological Characteristics

In postmortem specimens with cSVD, small vessels appear as tortuous pencil-like structures with a thick basal lamina. The differentiation of small arteries from small veins is challenging in postmortem specimens, and it is based on a combination of methods, including 1) negative staining of smooth muscle cells, which are more abundant on the arteriolar than on the venular side, where they acquire distinct morphological characteristics with long cytoplasmic elongations wrapped around the small veins; 2) noninflammatory, periventricular venulopathy with concentric collagen deposition, which causes intramural thickening, stenosis, and ultimately luminal occlusion; 3) localization, where small veins are more visible in the posterior horns of the ventricle and where arteries are less present due to the watershed characteristics of these regions; and 4) the number of branches and amount of capillary free space surrounding the vessels (12, 17, 18, 40, 87, 88).

In vivo, small penetrating arteries of the white matter and basal ganglia and small veins in the periventricular areas are typically poorly visualized and below the resolution of most noninvasive techniques such as CT and MRI. However, with ultra-high field time-of-flight MR angiography and susceptibility-weighted imaging, these penetrating small arteries and veins become more conspicuous. Small vessels usually manifest as thin, smooth, and relatively straight vasculature with moderate branching. In frank vasculopathies, such as distal internal carotid artery stenosis/occlusion, these penetrating vessels may be recruited as collateral pathways and greatly increase in number, with a more tortuous and aggressively branching pattern, and are prone to hemorrhage. Time-of-flight MR angiography can be used to visualize small arteries as thin thread-like areas of flow-related contrast (Fig. 5A). Using this technique, features of cerebrovascular diseasing include reduced vessel number, and increases in tortuosity can be examined (21). The morphology of small veins in vivo is best visualized using susceptibility-weighted imaging (Fig. 5B); this approach is very sensitive to motion artifacts; hence, utilization of ultra-high field magnets that reduce time of acquisition typically yield images with higher spatial resolution. Number of veins, vein length, tortuosity, and vein number or length per area (density) are measures that can be obtained from this modality (13, 69, 123).

Fig. 5.

Time-of-flight angiography of arterial tree and susceptibility-weighted image (SWI) of medullary small veins. A: time-of-flight angiography: resolution is 0.32 × 0.32 × 0.32 mm³. This method can be used to visualize small arteries that appear as thin thread-like areas of flow-related contrast. B: SWI. This method can be used to visualize small veins. This is shown here as a thin, smooth, and relatively straight vasculature with moderate branching darker compared with surrounding parenchyma. Resolution is 0.2 × 0.2 × 1.5 mm³. All images are acquired without the use of any contrast agents at 7T. Images are acquired using the Tic Tac Toe Radiofrequency Coil System (http://rf-research-facility.engineering.pitt.edu/). Images were provided by Dr. Tamer Ibrahim (University of Pittsburgh).

In summary, the neuroradiological methods for characterizing cSVD in older age have relied mostly on parenchymal damage, such as WMHs, rather than on the pathological events directly affecting the small vessels. A reason for this limitation is that most brain imaging measures prevent differentiation between the cellular and noncellular volumetric components in vivo. Postmortem approaches can distinguish the cellular components, but they typically occur in cases of advanced pathology. The advancement of MRI technology holds promise to further our understanding of cSVD, because it captures a greater range of physical properties of the tissue, but histopathological correlates of these most recent radiological methods have not yet been studied. Although the studies reviewed here are from aging populations, these observations alone do not explain how older chronological age drives such pathological changes. In the following sections, we review the studies relating demographic and health-related characteristics with neuroradiological makers of cSVD (associations with risk factors) and propose an explanation of the pathway linking older age with cSVD (pathogenesis and clinical consequences of age-related csvd).

ASSOCIATIONS WITH RISK FACTORS

The main findings of studies examining the associations of cSVD with chronological age and age-related cardiovascular factors are described in this section and shown in Figs. 6 and 7. In Fig. 6, we highlight the direction of association between the risk factor and the outcome and the study design, with specific attention to the frequency of measurements of risk factors and outcomes. We found a considerable representation of studies with longitudinal designs especially for blood pressure parameters, but few had reliable measures of duration of exposure to the risk factors and/or longitudinal changes in both predictors and outcomes. Length of followup of longitudinal studies was mostly limited to 5 yr, with few extending over a period of time long enough to accurately estimate age-related changes. In Fig. 7, we highlight sample size and mean age of the populations for whom associations were examined with antihypertensive medication, hypertension, inflammation, race, diabetes, and vascular stiffness. The majority of studies examined populations age 70 yr and older, with fewer capturing earlier stages of the aging process, and many had large sample sizes.

Fig. 6.

Associations between risk factors, white matter hyperintensities, and silent brain infarct/lacunes. ⌀, Not a significant association; ↓, significant inverse association; ↑, significant positive association. Red font, <60 yr; black font, ≥60 yr. *Meta-analysis. Studies are cross-sectional unless otherwise noted. L, longitudinal; L¹, predictor precedes outcome, each measured once; L², repeated measures of predictor, outcome measured only at end of followup; L³, predictor measured only once at baseline, repeated measures of the outcome; L⁴, repeated measures of both predictor and outcome.

Fig. 7.

Associations between risk factors and white matter hyperintensities by sample size, study design, and age group. Each circle indicates one study, and the number inside each circle reports the sample size. y-Axis: sample size, scale varies for each risk factor; x-axis: chronological age of the sample. Green, significant association (expected direction, unless otherwise noted); red, not a significant association; square, inverse association of expected direction; solid fill, longitudinal design; no fill, cross-sectional. Lines, case control. *Meta-analysis.

Chronological Age

Although the association of WMHs and SBIs/lacunes with chronological age is well established in neuroradiological and postmortem studies, precise estimates are not yet available; we found a wide variability in the rate of accrual of WMHs per year, ranging from 5% to 50%/yr, with some studies also showing apparent reduction of lesions over time (23, 25, 42, 45, 49, 89, 110, 117, 126, 138, 141, 152). Reasons for such discrepancies may include an overreliance on cross-sectional designs of adults with a wide age range, rather than repeated measurement of within-person changes over time. Among the studies reporting within-person changes, most had short (<5 yr) followup times, which may not be long enough to capture changes. We also noted that smaller (∼5%) accruals were found for the oldest old (>85 yr) (23, 126) compared with adults 65–75 yr old, averaging at ∼12% across studies (42, 117, 138, 139); such discrepancies raise the possibility that the oldest adults more likely to return for repeated serial scans are also more likely to be healthier, hence biasing the results toward lower accrual.

Isolating the effect of chronological age on cSVD is complicated by the well-known association between chronological age and those conditions that in turn can affect cerebrovascular pathology, including multimorbidities and dysregulation of biological factors, such as genetic and epigenetic factors (5, 62, 80). Thus, accurate analysis of the age-related effects on cSVD must also have a comprehensive characterization of risk associated with these other factors and test for interactions of age or conduct age-stratified analyses. Among the studies we reviewed, the effects of chronological and cardiovascular factors appeared independent of each other to some extent. For example, hypertension and blood pressure parameters remained significantly associated with cSVD in models adjusted for chronological age (30, 32, 65, 75, 100); in age-stratified analysis, associations appeared to have similar effect sizes for the oldest and the youngest subgroups of participants (66, 75). The effects of risk factors on cSVD could also be time dependent; that is, the effects could vary depending on the age at time of exposure (when), not only the duration (for how long). For example, exposure to high blood pressure during middle age has been linked repeatedly to increases in WMHs (30, 36, 38) and risk of vascular dementia (155). However, with advanced age, high blood pressure often has smaller or nonsignificant associations with WMHs (37). The reasons for such differences are not entirely clear. One possibility is that long-term exposure of the brain to high blood pressure may irreversibly alter the physiological structure of the arterioles and capillaries, leading to a less modifiable state (96). If confirmed in other studies, these findings would have repercussions on the success of intervention strategies, as they may be more effective for younger but not older adults.

Among the newer measures of cSVD, associations between chronological age and lower CBF or cerebrovascular reactivity have also yielded conflicting results. We are aware of one report on chronological age and small arteries’ morphological characteristics measured with time of flight showing an association between older age and greater tortuosity (22). Tortuosity was not found to be significantly associated with age in a restricted age sample of community-dwelling older adults (ages 70–89 yr) (123). Relationships of age with venous metrics remain to be fully examined.

Cardiovascular, metabolic, and inflammatory factors

Figures 6 and 7 show that the evidence for an association of risk factors with both WMHs and SBIs/lacunes appears most consistent for hypertension and blood pressure parameters. In contrast, descriptive studies examining the associations of antihypertensive medications with WMHs or SBIs/lacunes have yielded mixed findings. We did not find antihypertensive intervention trials in older adults without prior stroke.

We found that measures of large vessel disease (stiffness, distensibility, intima media thickness, plaque, and heart disease) appeared to be more strongly related to SBIs/lacunes than to WMHs. This is consistent with a prior report (63) suggesting different risk profile patterns for WMHs compared with SBIs/lacunes. A recent study of the Lothian Birth Cohort found that vascular risk factors contributed <2% to the prevalence of WMHs (147), underscoring the need to examine nonvascular pathways leading to cSVD.

Associations of cSVD with inflammation, dyslipidemia, or obesity were not consistent across studies. The state of the evidence is strongest for a positive association of inflammatory factors with WMH. Both C-reactive protein and IL-6 have been found to be related to WMHs (92, 112), but overall study results are mixed (3, 83, 92, 112, 118, 120, 140). Cumulative exposure, longer duration of exposure, and variability over time in the levels of these factors all appear to be important, but few studies have examined these characteristics concurrently. We found no strong evidence to support an association between inflammatory factors and SBIs/lacunes. No studies evaluated inflammatory factors in relation to CBF or cerebrovascular reactivity, and no studies evaluated these associations in those <60 yr of age. Future studies should focus on those areas and attempt to account for cumulative exposure to inflammatory factors.

Overall, the evidence does not support a strong link between diabetes and cSVD (34, 36, 53, 65, 66, 74, 78, 95, 101, 118, 141, 144). Diabetes may interact with other risk factors such as sleep duration (101) and systolic blood pressure (66), and these interactions should be evaluated in future studies. It has been suggested that insulin resistance may be a more sensitive risk factor for SBIs/lacunes than diabetes, as both cross-sectional (74) and longitudinal associations (34) have been reported.

Among the newer measures of cSVD, Jennings, et al. (58) found that reduced CBF was associated with increased cardiometabolic risk and worse intima media thickness.

Sex

There is consistent evidence among adults >60 yr of age for an association between the female sex and having higher SBIs/lacunes (78, 141, 144) as well as for progression of WMHs (141). However, we also found results in studies with a younger mean age that the female sex was not associated with SBIs/lacunes. This raises the possibility that the sex-SBIs/lacune relationship could differ by age, with sex-related effects on SBIs stronger for older compared with younger subgroups. This possible interaction remains to be confirmed within future study samples. If confirmed, this would appear to mirror the age × sex interaction found in stroke patients, whereby younger women are at lower risk of stroke than men, but older postmenopausal women are at greater risk of stroke than men (76). The reasons for such sex-related differences and the interaction with age are not clear, but they could be due to differences in hormonal profiles and their changes over time and/or in overall disease burden in vascular districts outside the brain. Of note, few studies were designed with the primary aim of evaluating WMHs or SBIs/lacunes by sex or examined sex × age interactions; hence, they may have been underpowered to examine sex-related differences.

Associations of sex with newer radiological markers of cSVD (CBF, time of flight, susceptibility-weighted imaging) hint at sex differences as well. Jennings et al. (58) found that male subjects had lower CBF than female subjects in one sample (n = 576) but not the other (n = 76) reported in the same report. This discrepancy in findings may be due to greater power to detect a difference in the larger study. In addition, a small study reported an association of the male sex with greater venous tortuosity, which did not survive adjustment for confounders (123). We did not find studies reporting sex differences in time of flight or CVR.

Black Race

Several studies have shown that black adults older than 60 yr have a higher probability of having WMHs and SBIs/lacunes compared with white adults of similar age. Studies have found associations between black race and increased burden of WMHs cross-sectionally and longitudinally (47, 48) and a higher prevalence of SBIs cross-sectionally (47, 53). Some studies did not find racial differences in WMHs (1, 77) or SBIs (7, 20). It has been theorized that such seemingly premature appearance of cSVD among black subjects could be due to the higher burden of cardiovascular risk factors and diseases in black subjects compared with white subjects. However, substantial evidence in support of this theory is lacking, as few reports of racial differences in neuroradiological markers of cSVD also had comprehensive measures of health-related characteristics (48, 100). Among the studies examining the role of health-related characteristics as explanatory factors of racial differences, some have found that adjustment for cardiovascular risk factors (smoking, hypertension, diabetes, hypercholesterolemia, and cardiac disease) (100) attenuated racial differences, whereas others did not (47, 48). As noted above, time of exposure and of measurement of the outcomes as well as duration of exposure all need to be taken into account. We noticed that those studies that failed to find explanatory factors of racial differences in cSVD were conducted in older subgroups (48, 95). Perhaps the explanatory effect of cardiovascular risk factors is stronger earlier in age compared with later in life, when the process of cSVD accumulation has already been unfolding for a long time. Another possibility is that the older subgroups of black subjects who are participating in MRI studies are exceptionally resilient. Black subjects are known to have a higher risk of death and morbidity compared with white subjects; hence, studies with older subgroups of black subjects may be vulnerable to survival bias.

With regard to more novel measures of cSVD, we are aware of only one study measuring CBF conducted in younger middle-aged participants (58), where black subjects had no significant differences in CBF compared with white subjects.

Allele Apolipoprotein E4

A recent meta-analysis (115) has indicated that overall apolipoprotein E4 (APOE4) carrier status is not associated with WMHs cross-sectionally or longitudinally but that these associations differ depending on the presence of other comorbidities. One study included in this meta-analysis indicated that the association was stronger in the presence of hypertension (32); a related study published after the meta-analysis also indicated that the association between APOE4 and WMHs was stronger in the presence of total cholesterol < 200 mg/dl (151). We found only one study that assessed the relationship of APOE4 carrier status with incident SBIs/lacunes (141), with nonsignificant results. In our study of small veins (123), we found a strong association between APOE4 carrier status and higher tortuosity of small veins. We did not find studies reporting associations with other novel markers of cSVD.

Sleep

Cross-sectional studies evaluating sleep and cSVD have suggested that both longer and shorter than average sleep duration are associated with WMHs, with associations stronger in the presence of diabetes (101) and in the absence of chronic kidney disease (93). Sleep duration has not been found to be associated with SBIs/lacunes; however, obstructive sleep apnea has among those more than 65 yr of age but not among those under 65 (26, 93).

Other Factors

Lifestyle factors such as smoking (36, 37, 46, 65, 95, 141) and sedentary behavior (122) have consistently been implicated in higher probability of having WMHs and more recently with reduced CVR in periventricular white matter (43).

Social factors in childhood have been recently assessed in relation to later-life WMH in a meta-analysis of 30 studies (8), with lower childhood IQ, socioeconomic status, and education levels exhibiting small but significant associations with greater WMHs in later life.

In the past few years, β-amyloid, the protein associated with Alzheimer’s disease plaques, has emerged as a potential risk factor for cSVD. Lower levels of plasma β-amyloid have been reported to be associated with greater WMH progression (61), and lower cerebrospinal fluid levels of β-amyloid have been associated cross-sectionally with WMHs (132, 154). A recent systematic review found that 10 of 13 studies reported no association between β-amyloid measured via PET imaging and WMHs, whereas 2 studies reported that increased β-amyloid was associated with increased WMHs and 1 of 13 studies reported the opposite (108).

Results regarding homocysteine and cSVD have been mixed, with some results reporting a nonsignificant relationship (79) and others reporting significant relationships (145, 153, 154). Higher levels of circulating proteins relating to glial cells and endothelial dysfunction have been associated with cSVD (141), as have higher levels of serum alkaline phosphatase (73). Other circulating factors, such as cystatin C (121, 146) and creatinine (78), have also been found to be associated with cSVD. We found that the angiogenic factor VEGF was inversely associated with tortuosity of small veins, although this result did not survive adjustment for confounders (123). Autoantibodies against oxidized low-density lipoprotein were not associated with traditional cSVD markers (109).

Studies that have evaluated the genetic risk for cSVD found associations with genes important in glial and inflammatory pathways (143), oxidative phosphorylation pathways (136), chromosome 6p25.3 (25a), the gene encoding transforming growth factor-β₁ (133), four single-nucleotide polymorphisms in the NOTCH3 gene (in subjects with hypertension only) (116), and the MTHFR genotype (68). The CHARGE consortium found several novel loci that were suggestive of associations with cSVD but did not reach genome-wide significance. Specific fibrinogen haplotypes have been related to SBIs/lacunes and periventricular WMHs (142), whereas C-reactive protein haplotypes were not associated with WMHs or SBIs/lacunes (104). Overall, there seems to be agreement that genes explain little of the variance in late-life cSVD (143).

In summary, there is agreement that blood pressure parameters are related to neuroradiological manifestations of cSVD both cross-sectionally and longitudinally. However, the actual proportion of the variance of cSVD that is explained by blood pressure and related vascular risk factors appears to be very modest, thus raising the possibility that nonvascular pathways are at play. Moreover, relationships between risk factors and neuroradiological manifestations of cSVD appear to follow distinct patterns. Specifically, the strength of the association between blood pressure parameters and other risk factors with cSVD appeared to vary depending on the presence of other health-related and demographic characteristics, as illustrated by the significant interactions found among risk factors and/or with age, race, and sex. Examining such interactions is important, as they suggest mechanisms by which cSVD may occur and give information that can be used for precision medicine approaches for both risk stratification and intervention. In the next section, we build on such associations to derive clues into the pathogenesis of age-related cSVD.

PATHOGENESIS AND CLINICAL CONSEQUENCES OF AGE-RELATED cSVD

Our review of radiological and histopathological characteristics of cSVD (neuroradiological markers and histological correlates of pathology of csvd) suggests that knowledge is primarily limited to abnormalities of brain parenchyma, specifically white matter, rather than actual small vessel pathology. It is generally assumed that such parenchymal abnormalities are secondary to small vessel disease because of the robust associations with blood pressure parameters, as reviewed in associations with risk factors. Based on this assumption, damage that begins in the small vessels would propagate to the parenchyma over time. However, the process could begin primarily in the parenchyma and subsequently propagate to involve the vessels. Both pathways are plausible; the components of the vessel wall and those of the surrounding tissue are so tightly interconnected that affecting one would likely affect the other. In this section, we maintain this assumption and examine candidate molecular pathways that trigger damage in the vessel wall first. We propose an additional clue into the pathogenesis of cSVD, that the distribution of damage in different compartments of the vasculature (e.g., arteries vs. veins) and in different lobes and areas of the brain (e.g., cortical or subcortical) can help understand the pathogenesis of cSVD and that this has implications for its clinical manifestations.

Key Molecular Regulators of cSVD

Under physiologically normal conditions, CBF autoregulation relies on a number of molecular pathways that protect the microcirculation from extreme variations in pressure. Chronological aging of the vessel wall components and/or exposure to risk factors typical of aging (hypertension, hyperglycemia, and dyslipidemia) can compromise the structure and functionality of cerebral vessels of both small and middle/large caliber (Fig. 8, red and blue arrows, respectively).

Fig. 8.

Age-related damage to the components of the vessel wall can affect small arterioles and capillaries (red arrows) and/or middle/large caliber arteries (blue arrows). Damage affects endothelial as well as smooth muscle cells and pericytes; these, along with hyalinosis of the basal lamina, can compromise the robustness of the wall, with possible rupture and/or impairment of vasoregulatory ability, resulting in abnormalities in flow and pressure. As they unfold, these events can trigger inflammatory reactions with extravasation of leukocytes, activation of astrocytes, and gliosis. Over time, such events will propagate to other cellular components, affecting oligodendrocytes, neurons, and glial cells.

It has been proposed that mechanical insults to the smooth muscle cells located in the wall of small arterioles may trigger oxidative stress (Fig. 8; red arrows), which, in turn, disrupt their vasoconstriction capacity (red solid arrow). Hyperglycemia is also known to trigger inflammatory and oxidative pathways, including dysregulation of nitric oxide production and less bioactivity, and production of reactive oxygen species (ROS) and advanced glycation end products (91), with detrimental effects on the vessel walls. Another proposed mechanism is through impairments in the regulation of the transient receptor potential cation channel (TRPC). Under physiologically healthy conditions, vascular walls respond to high pressure by upregulating TRPC, which increases intracellular Ca²⁺ and triggers vasoconstriction, resulting in increased vascular resistance and ultimately efficient CBF autoregulation (134). Both aging and deficiencies in insulin-like growth factor 1 (IGF-1) in the presence of hypertension have been shown to impair upregulation of TRPC (134, 135). It has been hypothesized that the loss of these protective autoregulatory processes leaves the microcirculation vulnerable to the damage caused by high blood pressure (134, 135). As cerebral resistance decreases, the pressure wave penetrates deep into the microcirculation, thus disrupting the structure and function of arterioles, capillaries, and venules, which, in turn, may lead to increased blood-brain barrier leakage and neuroinflammation and ultimately drive cognitive impairment (96, 134, 135).

A more mechanical hypothesis has been suggested whereby the primary pathogenic event is the accumulation of collagen patches in the internal elastic lamina/intima over the full duration of adult life in response to decades of small vessel wall's exposure to shear stress. Normally, with each pulse, a bolus of blood travels through the small arterioles, the capillaries, and the small venules, with the vessels operating like a peristaltic pump. To accommodate the bolus and recover normal anatomy, the small vessels need to be perfectly elastic. However, with aging there is a gradual accumulation of multiple molecular fractures of the intima and internal elastic lamina, with each typically repaired with elastin and collagen. The accumulation of such patches eventually transforms the elastic lamina from a fully elastic, homogeneous structure to a stiffened and friable wall, which causes a consequent gradual reduction in elasticity/compliance. Small vessels can no longer accept the volume of each pulse bolus by circumferential stretching only and thus rely on linear stretching. Like a peristaltic pump, the vessel flattens a bit between pulses and then becomes circular during the pulse, providing a greater potential volume to receive the pulse of blood such that the bolus of each pulse is accepted within the vessel by a combination of length and circumferential changes. This process not only results in the tortuous appearance of the vessel but also distorts the perivascular space, which distends, generating widened Virchow-Robin spaces that the pathologist sees in standard microscopic sections around tortuous arterioles. If this is the predominant pathway, primary prevention of cSVD may depend on slowing the development of the molecular changes that cause the declines in vascular elasticity.

The resulting array of structural degenerative events in the vessel wall, including death of endothelial cells, basal membrane thickening, and atrophy of the smooth muscle cells (19), can all lead to rupture (microbleeds), vessel walls’ collapse (microinfarcts), and/or ectasia (reduced blood perfusion) and loss of tight junctions with reduced blood brain-barrier integrity. In turn, greater permeability of the blood-brain barrier can lead to interstitial accumulation of water, gliosis, and further worsening of inflammatory/oxidative stress events that damage the gray and white matter parenchyma. Oligodendrocytes are extremely vulnerable to ischemic and toxic insults, and it has been suggested that the damage spreads to the gray matter after it reaches the white matter through reactive inflammatory processes. Nonhuman primate research has demonstrated a loss of myelinated nerve fibers, a reduction in their density, dystrophic myelin sheath changes, and inefficient remyelination in both the cingulate bundle and genu of the corpus callosum (14). In addition to a role for vascular injury in initiating damage, vascular damage may also be implicated in poor repair [for a review, see Kohama et al. (67)] Work in rhesus monkeys has demonstrated both increased microglial activation and phagocytic microglial activation with age (125). Phagocytic microglia should clear damaged myelin and allow repair to proceed. However, given that phagocytic microglial activation correlated with poor cognitive performance, the investigators posit that this may represent poor phagocytosis due to a chronic inflammatory state with aging (125). This chronic inflammatory state with aging acts to prevent remyelination [for a review, see Rawji et al. (103)].

Loss of functionality of aged endothelial cells can also occur in the presence of relatively normal or moderately high levels of blood pressure because of their intrinsically (age-related) reduced capacity to produce vasoregulatory molecules, including endothelia nitric oxide synthase. It has also been postulated that higher levels of parenchymal β-amyloid may have a role in inducing production of contractile proteins that render the vasculature smooth muscle cells hypercontractile (64).

A role for stiffness of cerebral arteries of middle caliber to initiate these events has also been proposed. According to this pathway (Fig. 8, blue arrows), age-related risk factors cause stiffness of cerebral arteries of middle caliber, which can lower cerebral perfusion pressure. Reduced cerebral blood perfusion in turn compromises parenchyma integrity and triggers inflammatory reaction with the release of vasoactive peptides, leading to both functional and structural abnormalities of the small vessels.

The above pathways are not contradictory, with each likely playing a role with the others in leading to cSVD as well as in inciting each other in a vicious cycle. Such interrelated phenomena can explain why intervention trials selectively targeting oxidative or hypertensive pathways have not yielded significant beneficial effects on reducing WMHs or SBIs (11). It is likely that a combined therapy would be needed to address the pleiotropic effects of the key molecular regulators of cSVD-related pathways. Although human observational studies have also supported a multifactorial pathogenesis of cSVD, they also suggested that older chronological age may override the influence of cardiovascular factors on the vasculature after long periods of time (e.g., among the oldest old). It is likely that once these processes have been triggered and have been unfolding for a long period of time, the progression of damage may be self-sustaining. This could explain why associations between certain risk factors and cSVD are less strong among the oldest old.

Spatial Distribution of cSVD Within the Microcirculation

It is generally assumed that damage in the small vessels is confined to small arteries and capillary beds. The engagement of the venular side of the microcirculation has received less attention in relation to cSVD. Venules are vulnerable to damage to a greater extent than arterioles because they have thinner, more permeable walls. Venules are also exposed to low pulsatile and slow flow (about half of what is observed in the arteriolar/capillary side). The slow flow and low pressure in the venules, combined with their thin and permeable walls, facilitate adhesion of inflammatory cells to venular endothelium, and their extravasation on the parenchymal side. Higher pulsatile flow, as observed in hypertension, can damage the venular wall, causing collagenosis, and both venular dilation and lumen narrowing. These changes can cause microischemic events, increase vascular resistance, and disrupt upstream hemodynamics, with damage propagating upstream to capillaries and arterioles, leading to abnormalities in blood flow. Another implication of venular damage is venous insufficiency and consequent vessel leakage (i.e., vasogenic edema) compromising interstitial fluid circulation and causing perivenular space abnormalities. Damage to perivenular spaces has been implicated in disruption of perivascular clearance of β-amyloid. A small 7-T neuroimaging study of older adults found a trend for an association with pulse pressure, although it did not find strong associations with other blood pressure parameters (123). To date, evidence for a role of venular damage in cSVD is derived primarily from animal studies (72, 94) and postmortem studies in humans (12, 40, 87, 98), and thus evidence for associations with risk factors is currently too sparse to reach conclusions.

Spatial Distribution of cSVD Within the Brain and Clinical Manifestations

The regions at the greatest risk of displaying neuroradiological and histopathological manifestations of cSVD are likely to be those where energy utilization is highest and collateral circulation is least and in circumstances when opportunity to recover is limited, for example, the watershed areas where collateral vascularization is intrinsically lower or areas where there is β-amyloid accumulation or other preexisting pathologies. In these regions, microinfarcts would appear first, and neuronal and synaptic losses would be greatest. Disturbances of cerebral perfusion preferentially affect regions at the outskirts of individual nutritional areas, such as interterritorial borders of nonanastomosing watershed arterial systems in the frontal lobes, the deep striatocapsular areas, and deep white matter, with subsequent ischemia, infarcts, and lacunes in these areas. The particularly low cerebral perfusion pressure of the distal fields between the deep and the superficial arterial system and in the watershed areas make the frontal cortical regions, the basal ganglia GM, and the deep WM particularly susceptible to repeated episodes of hypotension and/or to other toxic/inflammatory insults. Longitudinal studies following changes in radiological markers of cSVD would be critical to clarify the spatiotemporal evolution of damage and provide further clues into the pathogenesis of cSVD.

The spatial distribution of cSVD can also explain its clinical manifestations. SBIs and lacunes in the frontosubcortical areas is of concern because of the role of these areas in executive function (29, 71, 148) and decision making (86), both of which are impaired in older age. Likewise, WMHs in the periventricular areas, where long connecting corticosubcortical tracts gather, can explain reductions in speed of processing. Executive dysfunction is a hallmark cognitive feature of aging, comprising deficits in information-processing speed, psychomotor efficiency, attention, cognitive flexibility, and visuospatial perception. Dysfunction in these domains can impair planning and problem solving (44, 111), which are of concern in this population because they can affect independence with activities of daily living. Moreover, executive dysfunction is a well-established risk factor for dementia, depression, disability, and mortality (9, 35, 70, 71, 102, 106, 107, 113).

DISCUSSION: UNANSWERED QUESTIONS AND RESEARCH PRIORITIES

The number of older adults at risk of cSVD is projected to dramatically increase over the next several years, and neither preventive nor disease-modifying therapies are available. Therefore, capturing the earliest stages of cSVD and slowing its accrual over time are important public health problems that urgently require novel solutions. To date, our understanding of cSVD consists of a heterogeneous constellation of risk factors, signs, and symptoms. Heterogeneity and complexity are often seen as the evil of science, hampering progress toward targeted treatments. We hereby propose that the heterogeneity of cSVD can instead hold the secret to a more complete understanding of the process, and therefore, it should be leveraged. We propose that a study of cSVD should be at the systems level rather than reductionist. Accordingly, we call for an integrated approach that synergizes and harmonizes principles of 1) neuroscience, to directly characterize cerebrovascular pathology directly, rather than indirectly, via its late-stage parenchymal manifestations; 2) population science, to capture the populations (and its subgroups) at greater risk of developing cSVD; and 3) geroscience, to carefully characterize the molecular mechanisms underlying cSVD in translational models that bridge basic and clinical models. Such an approach could potentially actualize the goals of precision medicine and its foundational principle that different treatments are needed for different people.

Directly Characterizing Small Vessel Pathology and Mixed-Disease States

In vivo and postmortem findings of age-related small vessel disease cover mixed pathologies, including vascular, parenchymal, and amyloid-related pathologies. Different combinations of such mixed pathologies may have similar clinical manifestations. To explain the variability of cSVD, it is of paramount importance to accurately identify mixed lesion states. An important question is whether neuroimaging can distinguish and quantify mixed pathogenic features with such accuracy that it could produce biomarkers of onset, evolution, and response of cSVD to therapeutic interventions. As described above in neuroradiological markers and histological correlates of pathology of csvd, neuroimaging techniques that appear most readily applicable to the study of small vessel pathology include time-of-flight (Fig. 5A) and ultra-high-field susceptibility-weighted neuroimaging (Fig. 5B) to discern structural characteristics of small vessels. Accurate studies of cerebral blood flow, especially in the presence of chemical hypoxic stimuli, are also very valuable to capture endothelial dysfunction. Further study of the characteristics of the small vessels can have implications for treatment. For example, studying the primary pathogenic events of vascular tortuosity can uncover novel targets for primary prevention, such as determinants of protocollagen and protoelastin generation and transit as extracellular matrix proteins to sites of fracture and efficiency of elastin repair.

Who Should We Study, When, and For How Long?

Our review found important interactions with age, sex, and race on prevalence and severity of cSVD, thus highlighting the need to conduct carefully designed studies in diverse populations of older adults, with well-characterized health-related measures. Novel treatments of many chronic conditions are leading to an increase in the number of individuals who will survive longer with the initial condition and will also be exposed to morbidities typical of older chronological age (prevalence = incidence × duration). Such changes in demographic and health characteristics are contributing to the development of a “novel” aging phenotype and must be considered to understand age-related brain vascular pathologies, especially those affecting the oldest old segment of the population. Our review showed that, although each factor alone is an important contributor to cSVD, the interactions among these factors are likely more complex than expected. Thus, there are many novel questions that cSVD researchers must answer. How rapidly will brain integrity decline for these “new” older adults? Will rapidity of decline vary for different age, sex, ethnic subgroups, or patterns of comorbid conditions? Will these factors affect decline in different aspects of brain integrity (e.g., blood flow, atrophy) or behavior (e.g., attention, memory)? What factors could delay or accelerate the progression of brain abnormalities, and will they differ across age, racial, or patient subgroups? If different underlying mechanisms are at play for different age groups, different prevention/intervention strategies may be needed in different age groups.

Harmonization of Clinical, Translational, and Molecular Biology Methods

There is a need to understand the molecular mechanisms linking risk factors, small-vessel pathology, and radiological markers. Basic science models of clinical cSVD have been sought to improve our understanding of causal factors, pathophysiological natural history, and potential intervention and treatment targets. However, to date, animal models of cSVD have not described the full spectrum of phenomena observed in human cSVD (52). Many models attempt to recapitulate cSVD through a single causative initiating factor examined in isolation from the others, such as chronic hypoperfusion, hyperhomocysteinemia, or chronic hypertension (for a review, see Ref. 51). Such reductionist approaches do not capture the range of phenomena seen in cSVD resulting from complex and interwoven pathological processes. A systems-level approach would be more adept at identifying root causes and mechanisms of cSVD. Additionally, our review underscores the importance of measuring both onset and duration of exposure, how frequently the exposure and outcome are measured, and the time at which the outcome is measured (e.g., age). Thus, it is our hope that this review would serve as an invitation to the basic science community to design animal models of cSVD that instantiate a systems-level and life course approach. New animal models would serve as a catalyst for novel translational research bridging molecular characterization of pathophysiological pathways through experimental methods and clinical manifestations of cSVD. We further hope this review would serve as an invitation to the human subjects research community to incorporate systems-level and life course approaches and as an invitation for more collaboration between basic scientists and human subjects researchers.

What Would an Ideal Study Look Like?

To address the above questions, there is a need to build a population-based study in parallel with basic research equivalents of age-related cSVD to translate results from animal studies to human studies and vice versa. Specifically, we recommend recruiting a racially diverse cohort of men and women in their early adulthood or perhaps in childhood, when exposure to early-life determinants of midlife and late-life health first occurs and is modifiable. A life course approach would increase our understanding of the natural history of cSVD pathophysiology and identify potential time-dependent processes and critical/vulnerable periods. Participants would agree to donate their brains and be eligible and willing to undergo serial exams, until as close as possible to death, for multimodal neuroimaging scans, medical examination, interviews on lifestyle, blood draws, and behavioral assessment. Primary measures of interest would include neuroimaging, serum biomarkers, genotype, behavior and function, postmortem neuroimaging, and histopathology. Neuroimaging measures should be integrative and include those listed in neuroradiological markers and histological correlates of pathology of csvd. Integrating modalities could yield new insights, and such integration is beginning to occur [see Bangen et al. (10) for an example of such a study]. Assessing the functional and clinical relevance of such changes would be achieved by including measures of cognitive and physical functioning and following participants for critical end points such as cognitive decline, dementia, mobility disability, and death. The size of the samples must be large enough to conduct well-powered sex- and race-stratified analysis of data thus obtained, with a focus on interactions between multimorbidities.

A basic science approach would be critical to conduct mechanistic studies of cSVD. For example, the causative association between cSVD and neuroinflammation could be studied both in animal models and in humans using selected radioligands via PET and nanomolecular imaging approaches (24, 56). Similarly, this method could be used to study sex-related differences in cSVD manifestations.

Analysis of data should aim to identify patterns of risk factor profiles and explain the variance of the neuroradiological, clinical, and postmortem manifestations of cSVD. Specifically, statistical analysis could discriminate among the multiple combinations of events that ultimately lead to cSVD. For example, pattern analyses correlating risk factors with characteristics of cSVD that account for different methods of cSVD ascertainment and population characteristics may uncover distinct patterns of associations or pathways. Tools to handle large amounts of data exist. For example, machine learning and decision tree algorithms can combine high dimensional neuroimaging data, demographics, and clinical variables obtained at repeated time points to produce predictive models of clinically relevant outcomes (27). We propose that this approach is effective when integrated with an understanding of the underlying conditions and principles of rigorous study design.

Adding some of these exams to existing neuroimaging studies would be the most feasible approach. For example, the current human connectome study (54), existing developmental studies, and new National Institutes of Health developmental imaging initiatives (55) could be feasible options and open new venues for collaborative efforts. The United Kingdom Biobank would also be a promising study on which to build due to its prospective nature, large sample size (n = 500,000), and wealth of variables measured. However, this would need to be supplemented by other racially and ethnically diverse samples due to its small proportion of black participants.

In addition to providing novel knowledge of the pathophysiology, descriptive mechanistic studies are needed to plan and design randomized controlled trials targeting cSVD. Such studies could have rippling effects on reducing distal consequences of cSVD and could potentially prevent future strokes (131), Alzheimer’s disease, and related dementias (28).

GRANTS

This work was supported by National Institute of Health Grant R01-AG-044474.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

D.R.J., C.E.S., C.A.W., and C.R. conceived and designed research; D.R.J., C.E.S., C.A.W., and C.R. interpreted results of experiments; D.R.J., C.E.S., C.A.W., and C.R. prepared figures; D.R.J., C.E.S., C.A.W., P.J.G., J.M., and C.R. drafted manuscript; D.R.J., C.E.S., C.A.W., P.J.G., J.M., and C.R. edited and revised manuscript; D.R.J., C.E.S., C.A.W., P.J.G., J.M., and C.R. approved final version of manuscript.