Cognitive reserve against vascular contributions to cognitive impairment and dementia: A scoping review

Abstract

Vascular contributions to cognitive impairment and dementia (VCID) are common in older adults and are typically associated with cerebral small vessel disease (cSVD). However, little remains known about non‐pharmacological, modifiable lifestyle variables that may build cognitive reserve against cSVD/VCID. To address this gap, we conducted a scoping review of the existing literature on this topic, following Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines. We identified and reviewed over 1000 articles that included keywords associated with cSVD/VCID and cognitive reserve. Only a small number of these articles met our inclusion criteria for testing moderation with CR. Our review comprehensively evaluates and synthesizes results from these studies, which produce a pattern of mixed findings. We discuss the factors that may have contributed to these findings and conclude with a discussion of current gaps in the literature and recommendations for future research to address these knowledge gaps.

Highlights

• A scoping review of the literature identified studies that explored if cognitive reserve protects cognitive function from cerebral small vessel disease (cSVD).

• Results were generally inconsistent with some studies reporting positive cognitive reserve effects and others reporting null findings.

• Recommendations for future research are provided and include longitudinal designs and standardization of neuroimaging methods.

• Future longitudinal work should also identify which lifestyle variables (such as diet, physical activity) best mitigate cognitive deficits associated with cSVD.

1. INTRODUCTION

Cerebrovascular disease (CVD) is a major contributor to age‐related cognitive decline and dementia. ¹ , ² , ³ , ⁴ Specifically, vascular dementia is the second most prevalent type of dementia after Alzheimer's disease (AD), affecting upward of 2.7 million Americans, either alone or with other pathologies. ⁵ , ⁶ However, findings from in vivo neuroimaging and postmortem studies indicate that CVD is present in the majority of older adults, to varying degrees, even in those without significant cognitive impairment. ⁷ , ⁸ This suggests that CVD is common to aging and exists on a continuum instead of a discrete category such as the presence or absence of vascular dementia.

In keeping with this view, the concept of vascular contributions to cognitive impairment and dementia (VCID) has gained increased recognition and attention as an umbrella term used to describe a broad spectrum of cognitive impairment (i.e. mild cognitive impairment through dementia), associated with CVD. ² , ⁹ , ¹⁰ VCID is thought to primarily result from cerebral small vessel disease (cSVD) which is characterized by damage to the brain's blood vessels (arterioles, capillaries, and veins) and can occur in the absence of a detectable stroke or significant clinical symptoms. ¹¹ , ¹² The established STandards for ReportIng Vascular changes on nEuroimaging (STRIVE) guidelines for magnetic resonance imaging (MRI) markers of cSVD include white matter hyperintensities (WMHs), infarcts, lacunes, enlarged perivascular spaces, cerebral microbleeds, and cortical superficial siderosis. ¹³ , ¹⁴

Given the prevalence of cSVD in older adults, it is of practical importance to identify non‐pharmacological, modifiable lifestyle factors that may mitigate its effects on cognitive function. Cognitive reserve (CR) provides a useful framework for understanding how modifiable lifestyle factors may mitigate the effects of cSVD on cognitive function. CR refers to the phenomenon that similar amounts of brain pathology/damage can result in different cognitive profiles, with some individuals showing cognitive impairment but others remaining cognitively normal. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ CR is generally thought to be built up over years through the adoption of lifestyle variables such as healthy diet, physical exercise, cognitive or social stimulation, educational or occupational attainment, and effectiveness of stress‐coping strategies, among other potential factors. ¹⁹ , ²⁰ , ²¹

This scoping review seeks to identify and summarize studies that have explored CR against cSVD/VCID. A schematic illustration representing reserve factors which may protect cognitive function from cSVD is provided in Figure 1. In a previous review, Pinter et al. ²² summarized six studies investigating CR in the context of cSVD. Our current work uses a start date of 2013 for study inclusion to coincide with the established STRIVE guidelines for MRI markers of cSVD. ¹⁴ In addition, this review uses updated Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines for scoping reviews. ²³ Finally, we adopt recent consensus guidelines for testing of CR, which require that a putative CR variable be a statistical moderator in the relationship between a measure of brain pathology/damage and cognitive function. ¹⁵ Using these criteria, this review assesses the current state of the literature on CR variables that may protect cognitive function in the presence of cSVD.

FIGURE 1.

A schematic illustration of lifestyle factors that may contribute to cognitive reserve against VCID. The illustration depicts some of the lifestyle factors (images in circles) discussed in this scoping review that have been suggested to contribute to the development of cognitive reserve against cSVD and VCID. cSVD, cerebral small vessel disease; VCID, vascular contributions to cognitive impairment and dementia.

2. METHODS

2.1. Information sources & search strategy

All study methodology followed the 2020 updated PRISMA guidelines ²⁴ and the Prisma Extension for Scoping Reviews guidelines. ²³ Published peer reviewed articles were searched in the following three databases: PubMed, Embase, and Web of Science. The databases were searched from August 1, 2013, to October 1, 2024. The start date of August 1, 2013, was selected to coincide with the first publication of the STRIVE consensus guidelines to identify cSVD. ¹⁴ Search terms were selected to retrieve articles which included the following broad categories: cSVD, MRI markers of cSVD, neuropsychological tests, and cognitive reserve. The specific search terms used for each database are listed in File S1. The vendor and search period covered by each database is additionally described in Table S1.

2.2. Eligibility criteria

Eligible studies had to be written in English, include adults age 40+, and include adults without the presence of identified neurological diseases or other disease states (e.g., multiple sclerosis, Parkinson's disease, Huntington's disease, Down syndrome, diabetes, heart disease). Individuals with diabetes and heart disease have an elevated risk of VCID. ²⁵ , ²⁶ However, as a first scoping review in this area, our motivation was to summarize the state of the literature in relatively healthy older adults prior to exploring patterns of reserve against cSVD in groups with significant metabolic and cardiovascular conditions.

Eligible studies also had to include at least one CR proxy as a predictor variable, at least one MRI biomarker of VCID as a predictor variable, and at least one cognitive test as the outcome measure. Cognitive reserve proxies could include education, occupation, leisure activities, physical activity, dietary‐intake, childhood IQ/cognitive performance, socioeconomic status, marital status, and residual‐based metrics but could not be solely age or sex/gender. MRI markers of VCID could include white matter hyperintensities, small subcortical infarcts, lacunes of presumed vascular origin, enlarged perivascular spaces, cerebral microbleeds, and diffusion metrics (i.e., diffusion tensor imaging, free water, and/or peak width of skeletonized mean diffusivity). Cognitive measures had to be based on neuropsychological testing rather than on fMRI based‐tasks. A final, but key, inclusion criterion was that reserve had to be tested via statistical moderation (CR proxy × MRI‐based VCID biomarker predicting cognition), as recommended by several reserve workgroups. ¹⁵ , ²⁰ We focused on articles reporting empirical findings rather than reviews, conference proceedings, or meta‐analyses.

2.3. Selection process

Abstracts from articles were screened for compliance with eligibility criteria by two investigators (C.E.B. and C.P.) working independently, with any potential discrepancies settled by the principal investigator (B.T.G.).

2.4. Data collection process and items

Abstracts meeting eligibility criteria proceeded to full text article review by two investigators (C.E.B. and C.P.). The entire articles were then reviewed to ensure that eligibility criteria were met. A data‐extraction form was created to allow for charting of the data. Both investigators independently reviewed five articles using the data‐extraction form for calibration. Following this, any changes to the form were discussed and updated accordingly.

From each eligible article, the study design (cross‐sectional or longitudinal), sample size, exclusion criteria, age, sex/gender ratio, education level, and cognitive status were recorded. The type of MRI‐based VCID biomarker (along with MRI field strength and sequences used), the specific CR proxy/proxies, and the cognitive domains explored were similarly noted. Finally, statistical moderation tests, results, and interpretations from each study were reported. Additional pertinent study‐specific details were recorded as notes. Subsequent figures were created in BioRender and R version 4.5.2 (ggplot2 and ggpubr).

2.4.1. Supplemental abstract review

After the completion of our main scoping review, we conducted a supplemental review of any abstracts that were excluded for not indicating moderation or interaction effects within the abstract (Sections 2.2 and 2.3). This was done to both help prevent the exclusion of any articles using moderation analyses that may have been missed and to report results of articles that used stratified analyses (i.e. reporting a significant finding in one subgroup but not another) providing evidence for the “weak” form of cognitive reserve. ¹⁵ Re‐reviewed abstracts had to meet the original inclusion criteria (minus indicating ’moderation’ in the abstract) and mention “reserve” to be eligible for full text review for this supplemental review. The full‐text articles were then reviewed by two investigators (C.E.B. and C.P.) to ensure eligibility criteria were met and to investigate whether the article explored reserve either using moderation or stratified groups analyses (the latter of which one was found ²⁷ ) and is reported in Table S2.

3. RESULTS

3.1. Search results

A total of 1265 article abstracts were screened for eligibility following duplicate removal. Of those 1265 reports, 38 empirical articles were selected for full text review. Twenty‐three articles did not meet inclusion criteria due to other health conditions being studied (n = 2), age of the participants (n = 2), absence of moderation with the CR variable (n = 15), and absence of neuropsychological tests as the dependent variable (n = 4). This resulted in a total of 15 eligible articles for our scoping review. Further details about identification, screening, and inclusion are reported in Figure 2.

FIGURE 2.

PRISMA flowchart for selection of studies included in this scoping review. The figure summarizes the major steps involved in assessing articles for possible eligibility in this scoping review. A total of 1265 abstracts were screened for potential eligibility, with 16 articles ultimately meeting our scoping review inclusion criteria. DV, dependent variable; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta‐Analyses.

3.1.1. Supplemental search results

A total of 246 abstracts were re‐reviewed as part of our supplemental abstract review. Of those, 231 abstracts had been rejected for lack of moderation noted in their abstract and 15 full‐text articles rejected for the same reason in Section 3.1. Only one article meeting inclusion criteria investigated results using the “weak” form (Table S2). This study reported that lifelong occupation and leisure time was positively correlated with cognition in healthy controls, but that this association was not present in those with cSVD pathology (WMHs).

Using the criterion from the supplementary abstract review, we were able to find an additional article that investigates reserve using the recommended (moderation) analysis. Since this scoping review is focused on reporting moderation findings in the context of cSVD, this study was added to our pool of 15 eligible studies reported above (section 3.1) to bring the total to 16 eligible studies.

3.2. Study demographics and characteristics

The study design and participant characteristics varied across the 16 articles selected for review and are reported in Table 1. Slightly less than half of the studies used a longitudinal design while the others were cross‐sectional. The average age of participants ranged from the early 50s to early 80s. There tended to be more female than male participants enrolled in these studies, with the exception of three studies where the majority were men ²⁸ , ²⁹ , ³⁰ and one study that only included men due to the population of interest being examined (Vietnam War Veterans). ³¹ The mean years of education ranged from 7 to 17. Levels of education varied by cognitive status and racial group. Additionally, there was one study where a small portion of participants did not obtain any formal education. ²⁹ The majority of participants across studies did not have dementia and were most often classified as being cognitively unimpaired or having mild cognitive impairment (MCI).

TABLE 1.

Sample characteristics

Authors	Study design	Sample size (n)	Age (years)	Gender (female)	Education	Cognitive status
Arola et al. 2021	Cross‐sectional	152	70.6 (2.9)	62.5% (95)	13.0 (4.5)	No neurological disease MoCA: 23.5 (3.3)
Avila et al. 2021	Longitudinal	Overall: 1553	73.8 (5.7)	63.7% (990)	11.4 (3.8)
		White: 416	73.9 (6.5)	56% (232)	14.7 (3.4)	CU: 84% (350);MCI: 16% (66)
		Black: 547	73.7 (5.7)	66% (361)	12.9 (3.5)	CU: 79% (432); MCI: 21% (115)
		Hispanic: 590	73.7 (5.2)	67% (397)	7.7 (4.4)	CU: 74% (436);MCI: 19% (110)
Casaletto et al. 2020	Longitudinal	Overall: 829	75.5 (7.2)	58% (481)	15.6 (3.2)
		UCSF sample: 344	73.5 (7.3)	55.8% (192)	17.5 (2.0)	Dementia free MMSE median: 29 [28, 30]
		UCD sample: 485	76.9 (7.1)	59.6% (289)	14.3 (4.1)	Dementia free MMSE median: 28 [26, 29]
Chen et al. 2022	Cross‐sectional	Overall: 232	63.4 (6.3)	52.6%
		WMH group + high CR: 68	63.3 (8.7)	54.4% (37)	‐	MMSE: 28.5 (1.5) MoCA: 25.4 (1.5)
		WMH group + low CR: 69	67.3 (6.8)	56.5% (39)	‐	MMSE: 27.4 (2.2) MoCA: 21.1 (2.6)
		Control group + high CR: 47	55.0 (3.5)	55.3% (26)	‐	MMSE: 28.3 (2.2) MoCA: 24.7 (4.4)
		Control group + low CR: 48	66.0 (4.9)	41.7% (20)	‐	MMSE: 27.9 (2.7) MoCA: 23.8 (4.1)
Durrani et al. 2022	Cross‐sectional	200	72.8 (6.8)	48% (96)	–	Vascular MCI, MCI, or SCI MoCA: 24.5 (3.3)
Gallo et al. 2021	Longitudinal	430	70.4 (8.9)	58.6% (252)	12.7 (4.3)	Dementia free MMSE: 29.1 (1.0)
Jansen et al. 2021	Cross‐sectional	Overall: 543		60% (326)		MMSE: 25.6
		SCD: 108	Median: 71 [66, 78]	57% (61)	Median: 5 [3, 5] a	MMSE: 28.1 (2.0)
		MCI: 190	Median: 78 [73, 82]	56% (106)	Median: 4 [4, 5] a	MMSE: 26.3 (2.5)
		Dementia due to AD: 245	Median: 80 [76, 84]	65% (159)	Median: 4 [3, 5] a	MMSE: 22.3 (3.8)
Jokinen et al. 2016	Longitudinal	615	73.6 (5.1)	54.8% (337)	9.6 (3.8)	Mixed cognitive status MMSE: 27.4 (2.4)
Li et al. 2023	Cross‐sectional	179	69.3 (3.9)	39.7% (71)	No formal: 23.5%Elementary: 43.6%Middle or above: 33.0%	Dementia free MCI: 26.3% (47)MMSE: 25.0 (4.8)
Rosso et al. 2016	Longitudinal	Overall: 3241				Mixed cognitive status
		White LT: 917	72.1 (4.7)	57.5%	DNC HS: 35.6%; HS: 30%; HS+: 34.4%	3MS: 92.1 (6.8)
		White MT: 893	71.8 (4.9)	59.8%	DNC HS: 21.3%; HS: 33.5%; HS+: 45.2%	3MS: 92.1 (6.5)
		White HT: 917	71.9 (4.8)	53.7%	DNC HS: 9.8%; HS: 21.5%; HS+: 68.7%	3MS: 93.4 (5.6)
		Black LT: 176	73.4 (5.6)	67.1%	DNC HS: 47.2%; HS: 22.7%; HS+: 30.1%	3MS: 82.6 (10.8)
		Black MT: 167	72.5 (5.4)	65.9%	DNC HS: 37.1%; HS: 27.6%; HS+: 35.3%	3MS: 86.8 (8.2)
		Black HT: 171	71.8 (5.1)	56.1%	DNC HS: 26.9%; HS: 21.6%; HS+: 51.5%	3MS: 88.0 (8.0)
Sanderson‐Cimino et al. 2021	Cross‐sectional	242	61.9 (2.6)	0% (all males)	13.9 (2.0)	Not provided: Middle‐aged participants
Soldan et al. 2020	Longitudinal	271	57.9 (10.5)	59.4% (161)	17.1 (2.3)	All CU at baseline MMSE: 29.7 (0.7) Dementia at follow‐up: 15.9% (43)
Turcotte et al. 2022	Cross‐sectional	1628	73.5 (7.2)	47.4% (771)	16.1 (2.8)	CU: 39.3% MCI: 43.5% AD: 17.3%
Yang et al. 2024	Longitudinal	34,041	Low CR: 54.5 (7.4)	55.3%	‐	Dementia free
			Mid CR: 54.8 (7.4)	62.4%	‐	Global cognitive score Low CR: 0.03 (0.67)
			High CR: 53.4 (7.3)	44.6%	‐	Mid CR: 0.15 (0.65) High CR: 0.19 (0.66)
Zahodne et al. 2013	Longitudinal	Overall: 703	80.1 (5.5)	67% (471)	10.8 (4.8)
		No MCI: 553	79.8 (5.9)	66.7% (369)	11.1 (4.7)	No MCI: 75.8 %
		Amnestic MCI: 97	81.1 (5.9)	72.2% (70)	11.4 (4.3)	Amnestic MCI: 13.8%
		Non‐amnestic MCI: 73	80.6 (5.7)	67.1% (49)	7.9 (5.0)	Non‐amnestic MCI: 10.4%
Zahodne et al. 2019	Cross‐sectional	Overall: 1136	77.2 (5.6)	62.8%	11.7 (4.6)	Dementia free
		2005 Cohort: 623	80.0 (5.6)	67.7% (422)	10.8 (4.7)
		2011 Cohort: 513	73.9 (5.7)	56.7% (291)	12.8 (4.5)

Abbreviations: CR, cognitive reserve; CU, cognitively unimpaired; DNC HS, did not complete high school; HS, high school; HS +, schooling beyond high school; HT, highest tertile; LT, lower tertile; MCI, mild cognitive impairment; MMSE, Mini Mental State Exam; MoCA, Montreal Cognitive Assessment; MT, middle terile; SCD, subjective cognitive decline; UCD, University of California, Davis; UCSF, University of California, San Francisco; 3MS, Modified Mini‐Mental State Examination. Education is reported in years unless otherwise noted.

^aEducation values represent the following: incomplete lower‐level secondary education (3), vocational training (4), and advanced vocational training or lower professional degree (5). All values represent means and standard deviations or percentage and number unless otherwise noted. In the case where summary statistics for the overall sample were not included in a given manuscript, the pooled mean and standard deviation were calculated by the authors.

3.3. Neuroimaging modalities

The MRI field strength, sequences, and MRI‐based VCID biomarkers used for each study are summarized in Table 2 and Figure 3A. The majority of studies (62.5%) had 3 Tesla scanners available for at least some portion of the study, while the remainder had at least 1.5 Tesla scanners available. As the vast majority of the reviewed studies (93.8%) used WMHs as a VCID biomarker, T2‐weighted fluid‐attenuated inversion recovery scans were frequently implemented (81.3%). Other frequently used MRI‐based VCID biomarkers included “brain infarcts”, “lacunar infarcts”, or “lacunes” (31.3%). A single study ³² used diffusion tensor imaging (fractional anisotropy) as an MRI‐based VCID biomarker, and none of the studies used advanced diffusion metrics such as free water (FW), peak width of skeletonized mean diffusivity (PSMD), or neurite orientation dispersion and density imaging (NODDI).

TABLE 2.

Summary of independent and dependent variables.

Authors	MRI strength	MRI sequence	MRI measure	CR measure	Cognitive Domains
Arola et al. 2021	3T	T2‐weighted FLAIR T1‐weighted sequence	Total WMH volume	Years of education mCRS Composite of education and mCRS	Processing speed Executive function Verbal memory Visuospatial Verbal reasoning
Avila et al. 2021	1.5T (‘92‐‘99 cohort) or 3T (‘09 cohort)	T2‐weighted FLAIR	Total WMH volume (reverse scored)	Years of education	MemoryLanguage
Casaletto et al. 2020	3T (UCSF)	Diffusion weighted	Whole brain fractional anisotropy	Physical activity scale for the elderly Cognitive Activity Scale	Global cognition Episodic memory Executive function Spatial processing speed Verbal processing speed Semantic processing
	3T (UCD)	T2‐weighted FLAIR	Total WMH volume	LEAF physical activity subscale LEAF cognitive activity subscale	Global cognition Episodic memory Executive function Spatial processing speed Semantic processing
Chen et al. 2022	3T	T2‐weighted FLAIR T1‐weighted sequence	WMH group vs. controls	Cognitive reserve index questionnaire	Global cognition Memory Executive function Visuospatial function Information processing speed
Durrani et al. 2022	3T	T2‐weighted FLAIR	Vascular lesion burden (2 or more brain infarcts or confluent WMHs)	Composite of education level, occupational attainment, marital status, social activities, household income, number of languages spoken, and frequency of physical activity	Global cognition Memory Executive function Processing speed
Gallo et al. 2021	1.5T	T2‐weighted FLAIR and T1‐weighted fastfield‐echo sequence	Brain integrity index (combines WMH volume, whole brain GMV, HCV LVV, number of lacunes, and PVS score	Residual‐based CR from episodic memory Residual‐based CR from global cognition Activity‐based CR (composite of early life education, midlife work complexity, late‐life leisure activities, and late‐life social network)	Global cognition Perceptual speed Episodic memory Semantic memory Letter fluency Category fluency
Jansen et al. 2021	1.5T	T2‐weighted FLAIR and T1‐weighted sequence Axial T2‐weighted sequence	MRI‐based neuropathy (medial temporal atrophy, global atrophy, and rated WMHs)	Education level (low, average, and high)	Global cognition Episodic memory Executive function
Jokinen et al. 2016	0.5T or 1.5T	T2‐weighted FLAIR T2‐weighted fast spin echo sequence T1‐weighted sequence	WMH volume Presence of lacunar infarcts	Years of education Occupational attainment (high vs. low)	Global cognition Processing speed Selective attention Inhibition and flexible Set shifting Initiation and executive control Memory
Li et al. 2023	1.5T or 3T	T2‐weighted FLAIR Axial T2‐weighted sequence	WMH volume Number of lacunes	Composite of early life education level, occupational attainment, marital status, late‐life physical activity, late‐life social support, and late‐life social activity	Global cognition Episodic memory Language Executive function Attention Visuospatial ability
Rosso et al. 2016	1.5T	Sagittal T1‐weighted Spin Echo Axial spin density Axial T2‐weighted sequence	Rated WMHs Lacunar infarcts	Composite of neighborhood SES status	Global cognition Processing speed
Sanderson‐Cimino et al. 2021	3T	T1‐weighted sequence T2‐weighted sequence Proton‐density‐weighted sequence	Abnormal white matter volume (global, deep, and periventricular)	Armed forces qualification test as young adult (GCA)	Visuospatial ability Executive function Episodic memory Processing speed Verbal fluency
Soldan et al. 2020	1.5T	T2‐weighted FLAIR	Global WMH volume	Composite of years of education, baseline NART scores, and baseline scores on the vocabulary subtest of the WAIS‐R	Global cognition Verbal episodic memory Executive function
Turcotte et al. 2022	1.5T or 3T	T1‐weighted sequence	Global WMH volume	Composite of education level, occupational complexity, and verbal IQ.	Verbal episodic memory Language and semantic memory Attention Executive function
Yang et al. 2024	3T	T2‐weighted FLAIR	Global WMH volume	Latent class analysis using education level, occupational attainment, TV viewing time, frequency of confiding, frequency of social connection, and number of leisure activities	Global cognition Numerical memory Prospective memory Pairs matching Fluid intelligence Reaction time
Zahodne et al. 2013	1.5T	T2‐weighted FLAIR T1‐weighted sequence	Brain composite score (MemB; total brain volume, total cranial volume, WMH volume, and hippocampal volume)	MemR: Residual variance when looking at association between brain composite score and memory.	Language Memory
Zahodne et al. 2019	1.5T (2005 cohort)	T2‐weighted FLAIR	Global WMH volume	Years of education	Episodic memory
	3T (2011 cohort)	T2‐weighted FLAIR	Global WMH volume	Years of education	Episodic memory

Note: Bold values signify the cognitive domains used as an outcome variable for moderation analyses.

Abbreviations: CR, cognitive reserve; FLAIR, fluid‐attenuated inversion recovery; GMV, gray matter volume; HCV, hippocampal volume; LEAF, life experiences assessment form; LVV, lateral ventricular volume; mCRS, modified Cognitive Reserve Scale; MRI, magnetic resonance imaging; NART, National Adult Reading Test; PVS, periventricular spaces; SES, socioeconomic status; UCD, University of California, Davis; UCSF, University of California, San Francisco; WAIS‐R, Wechsler Adult Intelligence Scale‐Revised; WMH, white matter hyperintensities.

FIGURE 3.

Magnetic resonance imaging, cognitive reserve, and cognitive variables reported across studies. Depicted above are horizontal bar charts summarizing the different neuroimaging (A), cognitive reserve (B), and cognitive (C) measures used in the 16 studies selected for the review. The x‐axis refers to the number of studies that report using each measure. In the case of the cognitive reserve variables, we report the overall composites as well as the individual components (i.e. education, occupation).

3.4. Cognitive reserve measures

A detailed description of the CR measures utilized can be found in Table 2 and Figure 3B. The majority of studies (75.0%) used education alone or in combination with other metrics to measure CR. Education was often used in composite scores. Approximately 56.3% of studies created a composite score, which could include measures of occupational complexity, physical activity, social activity, and leisure activity in addition to education. One study ³³ created a composite score using variables related to neighborhood socioeconomic status.

Physical and cognitive activity were also used as independent proxies of CR in one study ³² and general cognitive abilities during young adulthood were used in another. ³¹ There were two studies that used a residual‐based method to calculate reserve. ³⁴ , ³⁵ In general, this approach uses the residualized values from demographic and MRI‐based brain integrity measures on episodic memory performance. The resulting values indicate differences between observed and predicted performance for a given participant. Taken together, a number of different measures are used to encompass CR, with education being the most common proxy.

3.5. Neuropsychological test measures

The cognitive domains explored in each study are displayed in Table 2, Table 3, and Figure 3C. Memory or episodic memory was the domain most often explored (87.5%) in the reviewed studies. Executive function was examined in 62.5% of the reviewed studies whereas processing speed was evaluated slightly less frequently (50.0%). Several studies also explored language abilities (37.5%). Finally, global cognition was examined in 62.5% of the reviewed studies. The majority of studies used composites of neuropsychological tests to assess cognitive performance for each cognitive domain. If only a single neuropsychological test was used to assess cognition in moderation analyses, that test was specified in Table 3 under “cognitive outcome”.

TABLE 3.

Moderation analyses.

Authors	Predictor (MRI or CR)	Moderator (MRI or CR)	Cognitive outcome	Results
Arola et al. 2021	WMH volume	Composite of education and mCRS	Processing speed	Not significant
	WMH volume	Composite of education and mCRS	Executive function	Not significant
	WMH volume	Composite of education and mCRS	Verbal memory	Not significant
	WMH volume	Composite of education and mCRS	Visuospatial	Not significant
	WMH volume	Composite of education and mCRS	Verbal reasoning	Not significant
Avila et al. 2021	WMH volume	Education	Memory	White individuals: Significant; higher CR was protective against the relationship between WMH volume and memory decline. Black individuals: Not significant Hispanic individuals: Not significant
	WMH volume	Education	Language	White individuals: Significant; higher CR was protective against the relationship between WMH volume and language decline. Black individuals: Not significant Hispanic individuals: Not significant
Casaletto et al. 2020	Global DTI FA	Physical activity	Global cognition	Not significant
	Global DTI FA	Cognitive activity	Global cognition	Significant; individuals with low FA had poorer cognitive performance if they also had low CR. Higher CR protected individuals with low FA from poorer cognitive performance.
	Total WMH volume	Physical activity	Global cognition	Significant; higher CR was protective against the relationship between higher WMH volume and poorer cognition.
	Total WMH volume	Cognitive activity	Global cognition	Not significant
Chen et al. 2022	WMH group	CRIq	MMSE	Not significant
	WMH group	CRIq	MoCA	Significant; high CR led to better cognition vs. low CR for the WMH group. No interaction observed with the control group.
	WMH group	CRIq	MoCA: Memory	Not significant
	WMH group	CRIq	MoCA: Executive function	Not significant
	WMH group	CRIq	MoCA: Visuospatial ability	Not significant
	WMH group	CRIq	MoCA: Processing speed	Not significant
Durrani et al. 2022	Vascular lesion burden	CR categorical composite	Global cognition	Not significant a
	Vascular lesion burden	CR categorical composite	Memory	Not significant a
	Vascular lesion burden	CR categorical composite	Executive function	Not significant a
	Vascular lesion burden	CR categorical composite	Processing speed	Not significant a
	Vascular lesion burden	CR categorical composite (not dichotomized)	Global cognition	Not significant
	Vascular lesion burden	CR categorical composite (not dichotomized)	Memory	Not significant
	Vascular lesion burden	CR categorical composite (not dichotomized)	Executive function	Not significant
	Vascular lesion burden	CR categorical composite (not dichotomized)	Processing speed	Not significant
	Vascular lesion burden	Education (categorical)	Global cognition	Not significant
	Vascular lesion burden	Education (categorical)	Memory	Not significant
	Vascular lesion burden	Education (categorical)	Executive function	Not significant
	Vascular lesion burden	Education (categorical)	Processing speed	Not significant
	Vascular lesion burden	CR categorical composite (without marital status)	Global cognition	Not significant
	Vascular lesion burden	CR categorical composite (without marital status)	Memory	Not significant
	Vascular lesion burden	CR categorical composite (without marital status)	Executive function	Not significant
	Vascular lesion burden	CR categorical composite (without marital status)	Processing speed	Not significant
Gallo et al. 2021	Brain integrity index	Residual based CR	Global cognition	Significant; higher CR is protective against the effects of poor brain integrity on cognitive decline.
	Brain integrity index	Residual based CR (categorical)	Global cognition	Not significant
	Brain integrity index	Activity based CR	Global cognition	Not significant
	Brain integrity index	Activity based CR (categorical)	Global cognition	Not significant
	Brain integrity index	Residual based CR	MMSE	Significant; higher CR is protective against the effects of poor brain integrity on cognitive decline.
	Brain integrity index	Residual based CR (categorical)	MMSE	Not significant
	Brain integrity index	Activity based CR	MMSE	Not significant
	Brain integrity index	Activity based CR (categorical)	MMSE	Not significant
	Brain integrity index	Cognitive composite residual based CR	MMSE	Significant; higher CR is protective against the effects of poor brain integrity on cognitive decline.
	Brain integrity index	Cognitive composite residual based CR (categorical)	MMSE	Significant; higher and mid tertile are protective against the effects of poor brain integrity on cognitive decline.
Jansen et al. 2021	Education	MRI‐based neuropathy	Global cognition	SCD: Not significant MCI: Not significant AD: Not significant
	Education	MRI‐based neuropathy	Episodic memory	SCD: Not significant following multiple comparison correction. MCI: Not significant following multiple comparison correction. AD: Not significant
	Education	MRI‐based neuropathy	Executive functioning	SCD: Not significant MCI: Not significant AD: Not significant
	Education	WMH	Global cognition	SCD: Not significant MCI: Not significant AD: Not significant
	Education	WMH	Episodic memory	SCD: Not significant MCI: Not significant AD: Not significant
	Education	WMH	Executive functioning	SCD: Not significant MCI: Not significant AD: Not significant
	Education	MRI‐based neuropathy	Global cognition: sensitivity normed	SCD: Not significant MCI: Not significant AD: Not significant
	Education	MRI‐based neuropathy	Episodic Memory: sensitivity normed	SCD: Not significant following multiple comparison correction. MCI: Not significant AD: Not significant
	Education	MRI‐based neuropathy	Executive functioning: sensitivity normed	SCD: Not significant MCI: Not significant AD: Not significant
Jokinen et al. 2016	Education	WMH volume	Time on Stroop I	Significant; higher education was protective against higher WMH volume for time to complete the Stroop I.
	Education	WMH volume	Time on Stroop II	Significant; higher education was protective against higher WMH volume for time to complete the Stroop II.
	Education	WMH volume	TMTA	Not significant
	Education	WMH volume	Symbol digit modalities	Not significant
	Education	WMH volume	Digit cancellation	Not significant
	Education	WMH volume	Trail B minus A	Not significant
	Education	WMH volume	Stroop subtraction	Not significant
	Education	WMH volume	Verbal fluency	Not significant
	Education	WMH volume	Digit span backwards	Not significant
	Education	WMH volume	Immediate word recall	Not significant
	Education	WMH volume	Delayed word recall	Not significant
	Education	WMH volume	TICS b	Significant; higher education was protective against higher WMH volume for decline in TICS.
	Education	Lacunar infarcts	Time on Stroop I	Not significant
	Education	Lacunar infarcts	Time on Stroop II	Not significant
	Education	Lacunar infarcts	TMTA	Not significant
	Education	Lacunar infarcts	Symbol digit modalities	Not significant
	Education	Lacunar infarcts	Digit cancellation	Not significant
	Education	Lacunar infarcts	Trail B minus A	Not significant
	Education	Lacunar infarcts	Stroop subtraction	Not significant
	Education	Lacunar Infarcts	Verbal fluency	Not significant
	Education	Lacunar infarcts	Digit span backwards	Not significant
	Education	Lacunar infarcts	Immediate word recall	Not significant
	Education	Lacunar infarcts	Delayed word recall	Significant; Higher education was protective against greater number of lacunar infarcts for delayed word recall decline.
	Education	Lacunar infarcts	TICS b	Not significant
	Occupation	WMH volume	Time on Stroop I	Not significant
	Occupation	WMH volume	Time on Stroop II	Not significant
	Occupation	WMH volume	TMTA	Not significant
	Occupation	WMH volume	Symbol digit modalities	Not significant
	Occupation	WMH volume	Digit cancellation	Not significant
	Occupation	WMH volume	Trail B minus A	Not significant
	Occupation	WMH volume	Stroop subtraction	Not significant
	Occupation	WMH volume	Verbal fluency	Not significant
	Occupation	WMH Volume	Digit span backwards	Not significant
	Occupation	WMH volume	Immediate word recall	Significant; Higher occupational complexity was protective against higher WMH volume for decline in immediate word recall.
	Occupation	WMH volume	Delayed word recall	Not significant
	Occupation	WMH volume	TICS b	Not significant
	Occupation	Lacunar infarcts	Time on Stroop I	Not significant
	Occupation	Lacunar infarcts	Time on Stroop II	Not significant
	Occupation	Lacunar infarcts	TMTA	Not significant
	Occupation	Lacunar infarcts	Symbol digit modalities	Not significant
	Occupation	Lacunar infarcts	Digit cancellation	Not significant
	Occupation	Lacunar Infarcts	Trail B minus A	Not significant
	Occupation	Lacunar infarcts	Stroop subtraction	Not significant
	Occupation	Lacunar infarcts	Verbal fluency	Not significant
	Occupation	Lacunar infarcts	Digit span backwards	Not significant
	Occupation	Lacunar infarcts	Immediate word recall	Significant; higher occupational complexity was protective against greater lacunar infarcts for decline in immediate word recall.
	Occupation	Lacunar infarcts	Delayed word recall	Significant; higher occupational complexity was protective against greater lacunar infarcts for decline in delayed word recall.
	Occupation	Lacunar infarcts	TICS b	Not significant
Li et al. 2023	WMH volume	Cognitive reserve composite (categorical)	MMSE c	Not significant
	Lacunes	Cognitive reserve composite (categorical)	MMSE c	Not significant
Rosso et al. 2016	Neighborhood SES	WMH rating	3MS	White individuals: Not significant Black individuals: Not significant
	Neighborhood SES	WMH rating	DSST	White individuals: Not significant Black individuals: Not significant
	Neighborhood SES	Lacunar infarcts	3MS	White individuals: Not significant Black individuals: Not significant
	Neighborhood NSES	Lacunar infarcts	DSST	White individuals: Not significant Black individuals: Not significant
Sanderson‐Cimino et al. 2021	Total abnormal white matter	Young adult GCA	Visual spatial	Not significant
	Deep abnormal white matter	Young adult GCA	Visual spatial	Not significant
	Periventricular abnormal white matter	Young adult GCA	Visual spatial	Not significant
	Total abnormal white matter	Young adult GCA	Executive function and working memory	Not significant
	Deep abnormal white matter	Young adult GCA	Executive function and working memory	Not significant
	Periventricular abnormal white matter	Young adult GCA	Executive function and working memory	Not significant
	Total abnormal white matter	Young adult GCA	Episodic memory	Not significant
	Deep abnormal white matter	Young adult GCA	Episodic memory	Not significant
	Periventricular abnormal white matter	Young adult GCA	Episodic memory	Not significant
	Total abnormal white matter	Young adult GCA	Processing speed	Not significant
	Deep abnormal white matter	Young adult GCA	Processing speed	Not significant
	Periventricular abnormal white matter	Young adult GCA	Processing speed	Not significant
	Total abnormal white matter	Young adult GCA	Verbal fluency	Not significant
	Deep abnormal white matter	Young adult GCA	Verbal fluency	Not significant
	Periventricular abnormal white matter	Young adult GCA	Verbal fluency	Not significant
Soldan et al. 2020	WMH volume	CR composite score	Global cognition	Not significant
	WMH volume	CR composite score	Verbal episodic memory	Not significant
	WMH volume	CR composite score	Executive function and speed of processing	Not significant
Turcotte et al. 2022	WMH volume	CR composite score	Verbal episodic memory	Not significant
	WMH volume	CR composite score	Language and semantic memory	Not significant
	WMH volume	CR composite score	Attention	Not significant
	WMH volume	CR composite score	Executive function	Not significant
Yang et al. 2024	CR classification	WMH volume	Global cognition	Not significant
Zahodne et al. 2013	Brain composite score (MemB)	Residualized memory‐based CR	Language	Significant; Higher CR was protective against the relationship between brain poorer brain health and decline in language abilities.
Zahodne et al. 2019	WMH volume	Education	Episodic memory	2005 Cohort: Significant; stronger negative relationship between WMHs and episodic memory for those with higher levels of education. 2011 Cohort: Significant; weaker negative relationship between WMHs and episodic memory for those with higher levels of education.

Abbreviations: AD, Alzheimer's disease; CR, cognitive reserve; CU, cognitively unimpaired; DSST, Digit Symbol Substitution Test; FA, fractional anisotropy; GCA, general cognitive ability; MCI, mild cognitive impairment; mCRS, modified Cognitive Reserve Scale; MMSE, Mini‐Mental State Exam; MoCA, Montreal Cognitive Assessment; MRI, magnetic resonance imaging; SCD, subjective cognitive decline; SES, socioeconomic status; TICS, Telephone Interview for Cognitive Status; TMTA, Trail Making Time A; WMH, white matter hyperintensities.

^a Separate models were tested controlling for age and sex or age, sex, and cognitive status. Results remained similar and are reported together.

^b Data was collected approximately seven years from baseline as compared to three years for all other cognitive measures.

^c Separate models were tested controlling for age and sex, vascular risk factors, and brain integrity. Results remained similar and are reported together.

3.6. Moderation analyses

Moderation results across the reviewed studies are displayed in Table 3. Nine out of 16 studies (56.3%) had null findings for moderation testing. The remaining seven (43.8%) had at least one finding where the CR proxy significantly moderated the association between the MRI‐based VCID biomarker and cognition. The significant moderation effects were reported with a variety of different CR proxies and cognitive domains. In general, greater CR was protective against the negative effects of VCID‐related MRI pathology on cognitive abilities.

Three of the seven studies found a moderation effect between CR and WMH volume for global cognition. ³² , ³⁶ , ³⁷ CR proxies for these studies included measures such as physical activity, Cognitive Reserve Index questionnaire (CRIq), and residual‐based calculations. An additional two studies found moderation effects between CR and WMHs for memory. ³⁸ , ³⁹ Both studies used education as the CR metric. Finally, two studies used a brain integrity composite which included WMHs and residual‐based CR. ³⁴ , ³⁵ In these studies, language and global cognition were impacted by a moderation effect. It is important to note, however, that while most moderation effects were reported using WMHs as the VCID marker, the majority of findings with WMHs were not significant. ²⁹ , ³⁰ , ³¹ , ³³ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ ,  

Interactions were also observed with other MRI metrics of VCID and CR. For instance, moderation was observed between DTI FA (measured globally in the brain) and participation in cognitive activity when examining global cognition. ³² Interactions were present between lacunar infarcts and occupational complexity for memory; a similar relationship was observed with education and memory when considering brain infarcts. ³⁷ Conversely, another study found no interactions with lacunar infarcts and CR measures. ³³ Finally, no moderation was present between a composite measure of vascular lesion burden (based on WMHs and infarcts) and CR. ²⁸ A summary of all moderation results are displayed via balloon plot in Figure 4.

FIGURE 4.

Summary of moderation findings in selected articles. Depicted above is a balloon plot demonstrating null versus significant moderation findings across the 16 articles meeting inclusion criteria. The cognitive reserve measures are displayed on the x‐axis while the neuroimaging measures are displayed on the y‐axis. Each row represents one of the cognitive domains studied, which are listed on the right. The color and size of the circle represent the number of studies that correspond to the cognitive reserve by neuroimaging interaction for the given cognitive domain. The small black circle indicates that the given interaction was tested but not found. For the interactions listed as significant, a positive effect was observed between the MRI variable and the reserve variable for the cognitive domain reported (i.e. less pathology and more reserve were associated with better cognitive performance). DTI, diffusion tensor imaging; NSES, neighborhood socioeconomic status; WMH, white matter hyperintensities.

4. DISCUSSION

The present scoping review assessed the state of the literature on cognitive reserve (CR) against VCID. Over one thousand abstracts were initially identified and reviewed, with only a small number meeting recent consensus guidelines for statistical testing of CR, among other criteria. The majority of the studies reviewed used MRI‐derived VCID biomarkers of white matter hyperintensities (WMHs), with some using lacunes. Results were generally conflicting, with some reporting positive CR effects and others reporting null findings. Overall, our review demonstrates that the field of CR against VCID is still in its infancy and much more work is needed in this area in order to identify protective factors that may slow or mitigate VCID. In what follows, we discuss some potential reasons for the paucity of work in the field so far, the mixed pattern of reported findings, and some potential factors that likely contribute to the mixed findings. We conclude with recommendations for future studies in this field.

4.1. Recent emergence of VCID as a clinical category

We carefully reviewed 1265 abstracts containing a broad set of keywords related to CR and VCID/cSVD. Of these, only 16 published articles met our full inclusion criteria, which included meeting consensus guidelines for statistical testing of CR. ¹⁵ One reason for the paucity of work on reserve against VCID likely relates to the relatively recent formulation of VCID as a clinical category. ⁹ The term “vascular dementia” has been used for decades, with many studies exploring this disease. ⁴⁴ , ⁴⁵ , ⁴⁶ However, the conceptualization of VCID, which can include subtle cognitive changes accompanying cSVD in the absence of a detectable stroke, has been formalized more recently. ² , ⁹ In contrast, AD has been recognized as a form of dementia for over 100 years. ⁴⁷ , ⁴⁸ , ⁴⁹ As such, it is expected that the literature on CR against AD would be significantly more developed (see ⁵⁰ , ⁵¹ , ⁵² ) than that of CR against VCID. 

4.2. Inherent differences in the etiologies of VCID and AD

Another reason the published CR literature may differ between AD and VCID relates to the etiologies of the diseases themselves. AD is defined by specific underlying neuropathologies of extracellular amyloid plaques (amyloid‐beta deposits) and intracellular neurofibrillary tangles (hyperphosphorylated tau accumulation), both of which can be identified postmortem or in vivo using PET scans. In contrast, VCID is broadly defined as any disease process damaging the brain's blood vessels that negatively influences cognitive function with no specific, universally agreed upon biomarker (although WMHs are the most commonly used). Despite the less specific etiology of VCID, progress in the development of additional VCID biomarkers beyond WMHs is accelerating rapidly, with potential MRI and biofluid markers being developed and tested by the MarkVCID consortium ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ , the Diverse VCID consortium ⁵⁸ , the Heart‐Brain study ⁵⁹ , and the DISCOVERY consortium ⁶⁰ , among other ongoing efforts. ⁶¹

4.3. Specific MRI markers of cSVD/VCID

The MRI markers of VCID used in the reviewed studies were fairly homogenous in their use of macrostructural MRI measures. In particular, almost all of the studies used WMHs as an MRI marker of VCID, whether through the use of quantitative WMH volumes or qualitative visual ratings (usually Fazekas score). This is not surprising given that WMHs are the most frequently used biomarker of cSVD/VCID. ⁶² , ⁶³ About a third of the reviewed studies also used lacunes as VCID markers. This was also somewhat expected as lacunes seen on MRI are another commonly used biomarker of cSVD, reflecting ischemic strokes damaging small arteries supplying subcortical structures such as white matter, the basal ganglia, and to a lesser extent the hippocampus and/or pons. ¹² , ¹⁴

4.4. Methods used to compute MRI markers of cSVD/VCID

Despite the use of generally agreed upon MRI markers of VCID, how these markers were assessed varied significantly between studies. When considering WMH measures only, methodological differences included the use of quantitative measurement versus visual ratings, different MRI pulse sequences to assess WMHs (mostly T2‐weighted but a few T1‐weighted images), different segmentation methods, varying thresholds for WMHs, and using WMHs as a stand‐alone marker versus being part of composites. Criteria for rating lacunes and brain infarcts also varied in that some studies rated these lesions dichotomously (i.e., present/not present), while others quantified them by number. These measurement differences may have contributed to the mixed evidence for CR against VCID reported in this review. 

4.5. Emerging MRI methods for assessing cSVD/VCID

While WMHs and brain infarcts are clearly visible on standard MRI sequences, and reflect macrostructural tissue damage, ² , ⁶² , ⁶³ , ⁶⁴ they may not be as sensitive to early‐stage VCID as newer, diffusion MRI‐based microstructural measures. ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁶⁵ , ⁶⁶ Diffusion MRI‐based microstructural measures include fractional anisotropy, mean diffusivity, peak width of skeletonized mean diffusivity (PSMD), free water (FW), and neurite orientation dispersion and density imaging (NODDI) among other measures. Typically, these diffusion measures are assessed within the brain's white matter (WM), providing estimates of the extent to which molecular water movement is hindered by boundaries such as myelin and axons. In general, more isotropic diffusion suggests WM injury, which is thought to be due in part to damage to long, penetrating arterioles that supply deep WM regions. Only one study reviewed used a diffusion MRI metric, reporting that the effect of low whole‐brain FA on cognitive performance was moderated by physical activity level. ³²

4.6. Heterogeneity of CR variables and their measurement

Education was the most commonly used CR proxy in the studies reviewed, either alone or as part of a composite measure including other variables such as occupational attainment. Although education level is among the earliest proposed CR variables, it is a static variable that may not reflect an older adult's current health‐related lifestyle choices. Education level has also failed to show CR effects against AD neuropathology in a number of studies with moderate to large sample sizes. ¹⁷ , ⁶⁷ , ⁶⁸ , ⁶⁹ Further, as a proxy measure, CR effects associated with education level may be less informative about targetable intervention(s) than more specific variables such as dietary intake or physical exercise. For these reasons, it is unclear if education level represents an ideal measure of CR. We suspect that the continued use of education in CR research is likely in part due to its historical use over 30+ years in this field ¹⁶ , ¹⁷ and to it being a measure of convenience that is obtained in essentially all studies.

In addition to education, other CR variables in the studies reviewed included cognitive and/or physical activity scales for older adults and the CRIq used as single proxy measures, ³¹ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ composites, ²⁸ , ²⁹ , ³⁰ , ³² , ³³ , ³⁶ , ⁴¹ , ⁴² , ⁴³ or residual‐based measures. ³⁴ , ³⁵ These variations in the CR measures may have contributed to the relatively inconsistent results reported in this review.

4.7. Heterogeneity of neuropsychological tests

Cognitive domains such as executive function, processing speed, attention, episodic memory, and language were used as outcome measures in the studies reviewed. Memory was tested more often than other cognitive domains, either as a composite or alone. Memory may have been selected most frequently because CR is often studied in the context of AD, where episodic memory deficits represent a hallmark feature of the disease. ⁷⁰ , ⁷¹ However, given that executive function and processing speed may be more affected by cSVD, ⁷² , ⁷³ these cognitive domains should be more thoroughly explored as the outcome measures in CR research against VCID.

Global measures of cognition were also examined in a majority of the studies. In most instances, global cognition was assessed by creating a composite score from all of the cognitive domains measured in the given study. In other cases, it was assessed by cognitive screening tests such as the Mini‐Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), or the Telephone Interview for Cognitive Status (TICS). While screening tools such as the MoCA are reflective of overall cognitive function, they may not fully capture subtle age‐related changes in cognition for cognitively unimpaired individuals when used alone. ⁷⁴

4.8. Limitations

There are several limitations of this scoping review. While our review summarizes the literature on CR against VCID, it does not report on studies testing other reserve constructs, namely brain maintenance (BM) and brain reserve (BR). We chose to focus upon CR because it was one of the first reserve constructs proposed ¹⁷ , ⁷⁵ , ⁷⁶ and remains the most commonly investigated reserve construct. Further, at least some degree of cSVD is common in healthy older adults, making it important to understand variables that promote CR against vascular pathology. Nevertheless, we acknowledge that our review is a starting point and that future reviews will be needed to explore the potential effects of BM and BR against VCID as those literatures continue to grow.

Additionally, our review only reports results from observational studies of CR against VCID. Clinical trials did not meet our inclusion criteria as they explore different questions. For instance, among interventions designed to improve vascular health (e.g. SPRINT‐MIND, ACCORDION‐MIND, and HOPE‐3 clinical trials ⁷⁷ , ⁷⁸ , ⁷⁹ ), few if any studies have formally tested statistical moderation effects between CR (e.g. intervention effect) and VCID neuroimaging markers on cognitive function. In addition, clinical trials differ from CR research in that they are typically conducted over a short duration, which is unlikely to parallel how CR is thought to develop—over decades rather than weeks or months.

Nevertheless, identification of short‐term interventions that may boost cognitive and brain health is of obvious importance to the field. Encouragingly, results from clinical trials that have incorporated combined interventions including diet, physical activity, and cognitive engagement have shown promise in improving cognition over time. ⁸⁰ , ⁸¹ In addition, clinical trials have the ability to test/establish the length of time needed for a putative CR intervention to be successful. Ultimately, more cross‐over between observational and clinical trial research designs could be beneficial in addressing these issues and moving the CR field forward.

5. CONCLUSIONS AND FUTURE DIRECTIONS

Several conclusions can be drawn from our scoping review examining the effects of CR against VCID. First, relatively few studies have been published that follow consensus guidelines for assessment of CR effects. Second, the majority of studies we reviewed used WMH volume as the independent variable and did not consider newer MRI metrics of vascular health. We recommend that “classic” markers of cSVD such as WMHs still be used to quantify VCID‐related pathology, but suggest that potentially more sensitive diffusion MRI markers such as FW and PSMD also be incorporated. Standardization of methods used to assess MRI markers of VCID should also be a priority and is currently underway at a number of consortia, as described above.

Third, most studies of CR against VCID conducted thus far have used educational level as a putative CR variable which, as discussed in this review, appears to be a sub‐optimal measure of CR. Going forward, we recommend that more dynamic and modifiable measures such as physical activity, cognitive activity, and dietary‐intake be tested as possible CR variables against VCID. Fourth, multiple domains of cognition should continue to be studied, but an emphasis should be placed on executive function and processing speed because these domains appear to be more impaired by VCID than other domains such as memory. Finally, like all review articles, our review is limited to articles identified using a specific set of search criteria during a particular range of dates.

Overall, our review shows that findings related to CR against VCID are mixed, with inconsistent results concerning protective versus null effects. We conclude by calling for more research on this important topic as we believe that pursuing this line of research is critical to untangling the observed discrepancies and identifying the most consistent and effective forms of CR that may mitigate VCID. More specifically, future work should implement more advanced MRI metrics, determine which CR variables may be most relevant/specific to VCID, and prioritize longitudinal investigations. By continuing to study reserve in cSVD we can expand our knowledge of resilience outside the relatively narrow confines of AD‐specific pathology that has dominated the field of reserve.

AUTHOR CONTRIBUTION

Christopher E. Bauer: Conceptualization, data curation, methodology, writing—original draft, writing—review and editing. Brian T. Gold: Conceptualization, methodology, writing—original draft, writing—review and editing, funding acquisition, project administration, resources, supervision. Colleen Pappas: Conceptualization, data curation, methodology, writing—original draft, writing—review and editing.

CONFLICT OF INTEREST STATEMENT

All authors report no conflicts of interest. Author disclosures are available in the Supporting Information.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Supporting Information