White matter network damage mediates association between cerebrovascular disease and cognition

Abstract

To determine whether white matter network disruption mediates the association between MRI markers of cerebrovascular disease (CeVD) and cognitive impairment. Participants (n = 253, aged ≥60 years) from the Epidemiology of Dementia in Singapore study underwent neuropsychological assessments and MRI. CeVD markers were defined as lacunes, white matter hyperintensities (WMH), microbleeds, cortical microinfarcts, cortical infarcts and intracranial stenosis (ICS). White matter microstructure damage was measured as fractional anisotropy and mean diffusivity by tract based spatial statistics from diffusion tensor imaging. Cognitive function was summarized as domain-specific Z-scores.

Lacunar counts, WMH volume and ICS were associated with worse performance in executive function, attention, language, verbal and visual memory. These three CeVD markers were also associated with white matter microstructural damage in the projection, commissural, association, and limbic fibers. Path analyses showed that lacunar counts, higher WMH volume and ICS were associated with executive and verbal memory impairment via white matter disruption in commissural fibers whereas impairment in the attention, visual memory and language were mediated through projection fibers.

Our study shows that the abnormalities in white matter connectivity may underlie the relationship between CeVD and cognition. Further longitudinal studies are needed to understand the cause-effect relationship between CeVD, white matter damage and cognition.

Keywords

Cerebrovascular disease white matter microstructure magnetic resonance imaging population-based cognition

Introduction

Cerebrovascular disease (CeVD) is the major contributor to cognitive impairment and dementia in the elderly.¹ MRI markers of CeVD include small vessel (lacunes, white matter hyperintensities (WMH), cerebral microinfarcts, and cerebral microbleeds) and large vessel disease (cortical infarcts and intracranial stenosis (ICS)), which are identified as focal lesions (macro-structural).² These CeVD markers have been linked with multiple clinical outcomes including cognitive decline, conversion to mild cognitive impairment (MCI) and incident dementia.^3,4 However, it remains largely inconclusive, which factors contribute to the vulnerability of cognitive decline in individuals with CeVD.

One possible factor that explains the impact of CeVD on cognitive function and dementia risk is the disruption of white matter microstructure based on the hypothesis that CeVD disrupts normal structural connectivity of the brain.⁵ The damage caused by CeVD extends beyond what is visible on the conventional neuroimaging, where normal appearing white matter may show microstructural abnormalities. Diffusion Tensor Imaging (DTI) quantifies subtle changes in white matter microstructure and detects structural dysconnectivity between neural networks.⁶ Alterations in white matter networks have been associated with cognitive impairment.⁷ However, unequivocal evidence of how CeVD impacts cognition via white matter disruption is lacking. Previous studies are limited by specific cognitive tasks which do not reflect the multiple domains of cognitive ability commonly affected in MCI and Alzheimer’s disease (AD).^8,9 Moreover, previous studies are mainly focused on examining the association of either cerebral small or large vessel disease markers with cognition^10–14 and/or clinical outcomes^4,10,11,15 and do not take into consideration microstructural changes as mediators. There is a paucity of data that has evaluated the mediating effects of white matter network disruption between cerebral small vessel disease and cognition.^16,17 More work is needed that incorporates all imaging markers of CeVD including both small and large vessel disease and determine their joint effect on cognition in the presence of white matter network disruption.

We, therefore, investigated whether specific white matter disruption is the key link between CeVD burden and worse performance on different cognitive domains in a non-demented elderly population.

Materials and methods

Participants

Participants were drawn from the Epidemiology of Dementia In Singapore (EDIS) study, which is part of the Singapore Epidemiology of Eye Disease study and includes the three major ethnicities from Singapore (Chinese, Malay and Indians).¹⁸ The present cross-sectional analysis was restricted to the Chinese component of the EDIS study, in which DTI data was available. The screening and assessment of the Chinese participants were performed from 12 August 2010 to 10 February 2012. Participants aged ≥60 years were screened using the Abbreviated Mental Test (AMT) and a self-report of progressive forgetfulness. Screen-positives were defined as AMT score ≤6 for those with ≤6 years of formal education, or AMT score ≤8 for those with >6 years of formal education; or if the participant or caregiver reported progressive forgetfulness.¹⁹ Screen-positive subjects were invited to take part in the second phase of this study, which included an extensive neuropsychological test battery and brain MRI.¹⁸ Of the total 300 participants, 18 did not have MRI scans whereas another 22 scans did not pass through the quality control of DTI data. Of the remaining 260 participants, 7 were diagnosed with dementia and hence were excluded from the analysis. The final sample consisted of 253 individuals.

Ethical approval for EDIS was obtained from the National Healthcare Group Domain-Specific Review Board (Singapore) and the Singapore Eye Research Institute Institutional Review Board (Singapore), respectively. The study was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained for all participants prior to recruitment.

Neuropsychological assessment

Trained research psychologists administered a formal neuropsychological battery locally validated for older Singaporeans.¹⁸ This battery assesses seven cognitive domains: (1) Executive function with the frontal assessment battery²⁰ and maze task;²¹ (2) attention with digit-span, visual memory span²² and auditory detection tests;²³ (3) language with the Boston naming test²⁴ and verbal fluency;²⁵ (4) visuomotor speed with the symbol digit modality test²⁶ and digit cancellation;²⁷ (5) visuoconstruction with the Weschler memory scale-revised (WMS-R) visual reproduction copy task,²² clock drawing²⁸ and Weschler adult intelligence scale-revised subtest of block design²⁹; (6) verbal memory with word list recall³⁰ and story recall; (7) visual memory with picture recall and WMS-R visual reproduction.²²

Domain-specific z-scores were obtained by averaging the z-scores of the subtests of each domain of the present sample. For each participant, raw scores from each individual test within a domain were first transformed to standardized z-scores using the mean and standard deviation (SD) of that test. Subsequently, for each participant a mean z-score for each domain was calculated by averaging the z-scores of all the individual tests within that domain. These mean z-scores of each domain were then standardized using the mean and SD of that domain-specific z-score.³¹

Clinical diagnoses

Diagnoses of cognitive impairment were made at weekly consensus meetings attended by study investigators. ‘No cognitive impairment (NCI)’ was defined as normal cognitive functioning on the neuropsychological test battery. A diagnosis of ‘Cognitive impairment no dementia’ (CIND) was based on no significant loss of independence in daily activities and impairment in at least one domain of the neuropsychological test battery. Participants failed a test if they scored 1.5 SD below education-adjusted cut-off values on individual tests. Impairment in a domain was defined as failure in at least half of the tests in that domain.¹⁸

Neuroimaging data

MRI was performed on a 3 T Siemens Magnetom Trio Tim scanner, using a 32-channel head coil, at the Clinical Imaging Research Centre of the National University of Singapore. MRI protocol included T1-weighted, T2-weighted, Fluid Attenuated Inversion Recovery (FLAIR), Susceptibility Weighted Images (SWI), DTI and Magnetic Resonance Angiography (MRA).³²

DTI scans were obtained using a diffusion-weighted echo-planar imaging sequence (61 non-collinear diffusion gradient directions at b = 1150 s/mm², seven volumes of b = 0 s/mm², repetition time (TR)/time to echo (TE)=6800/85 ms, 48 contiguous slices, and voxel size = 3.1 × 3.1 × 3.0 mm³). High-resolution T1-weighted structural MRI was acquired using magnetization-prepared rapid gradient echo (MPRAGE) sequence (192 continuous sagittal slices, TR/TE/TI = 2300/1.9/900 ms, flip angle = 9°, isotropic voxel size = 1.0 mm³).

Markers of CeVD

CeVD markers were assessed by two clinicians on 60 randomly selected scans. Any disagreement on CeVD markers gradings were further discussed in the weekly consensus meetings to make a final decision. The CeVD markers were graded using the following STRIVE criteria:

Lacunes were subcortical lesions, 3–15 mm in diameter, with low signal on T1-weighted image and FLAIR, a high signal on T2-weighted image, and a hyperintense rim with a center following CSF intensity on FLAIR.³³ The intra- and inter-rater reliability as expressed by k statistic was 0.80 and 0.75.

Cortical cerebral microinfarcts were graded on T1, T2-weighted and FLAIR sequences and were defined as hypointense lesions on T1-weighted images, hyperintense on T2 and FLAIR images with a size <5 mm in diameter, restricted to cortex and perpendicular to the cortical surface.² The intra- and inter-rater reliability showed good to excellent agreement (k = 0.83 and k = 0.80).

Cerebral microbleeds were defined as focal, rounded lesions of hypointensity graded on SWI using the Brain Observer Microbleed Scale (BOMBS).^34,35 The intra-and inter-rater reliability of microbleeds was k = 0.78 and 0.74.

Cortical infarcts were identified as focal lesions involving cortical gray matter with hyperintense rim on FLAIR images and center following CSF intensity and size >5 mm. This is further aided by tissue loss of variable magnitude, with prominent adjacent sulci and ipsilateral ventricular enlargement.² The intra-and inter-rater reliability was k = 0.85 and k = 0.80.

ICS was graded on MRA and was defined as stenosis ≥50% in any of the intracranial arteries.³⁶ The intra-rater reliability expressed as k statistic was 0.79 whereas the inter-rater reliability was 0.69.

WMH volume of the whole brain was quantified by automatic segmentation using the FLAIR sequence (see Supplementary Figure 1 for group WMH maps) whereas hippocampus volume was derived using a model-based automated procedure (Free Surfer, v.5.1.0) on T1-weighted images, in collaboration with Erasmus University Medical Center, Netherlands.³⁷

White matter microstructure derivation

The DTI data were preprocessed using FSL v5.0 (http://www.fmrib.ox.ac.uk/fsl) following our previous approach.³⁸ Eddy current distortion and head movement were corrected through affine registration of diffusion-weighted images to the first b0 volume. Participants with more than 3 mm of absolute maximum motion were excluded from further analysis. Diffusion gradients were rotated to improve consistency with the motion parameters. Individual maps were visually inspected for signal dropout, artifacts, and additional motion. Fractional anisotropy (FA) and mean diffusivity (MD) images were created by fitting a diffusion tensor model to the diffusion data at each voxel (DTIFIT, FSL). We then applied Tract-Based Spatial Statistics (TBSS)³⁹ to analyze the FA data in the major brain WM tracts. FA images were first non-linearly registered using FNIRT (FMRIB's Nonlinear Registration Tool) and transformed to the high resolution (1 mm³) FMRIB58_FA image. The FA images were skeletonized and represented the center of the WM tracts in the brain. The MD images were projected to the mean FA skeleton to obtain skeletonized MD images.

Statistical analysis

Data normality was assessed by Kolmogorov–Smirnov Test and Q-Q plots. Baseline characteristics (including demographic, vascular risk factors (yes vs. no), and CeVD markers) were compared across diagnostic groups using Analysis of Variance (ANOVA) followed by post-hoc analyses. CeVD markers were treated as count data and a logarithmical transformation was performed to achieve a normal distribution. The association between CeVD markers (counts) and cognitive domains were constructed using multiple linear regression adjusting for age and gender. Residual normalities were evaluated using Kolmogorov–Smirnov Tests. We additionally adjusted the model for education, vascular risk factors, hippocampus volume, and diagnosis. Because of the multiple testing performed within seven cognitive tests, Bonferroni correction was applied to obtain a revised significance level of 0.05/7∼0.0071.

Relationship between white matter microstructure and CeVD markers

To identify the white matter regions where the microstructure was associated with the CeVD markers, we built voxel-wise general linear models for FA and MD separately. We included the skeletonized FA or MD images as the dependent variable, the six CeVD markers as the covariates of interest and age, gender, and handedness as the nuisance variables. Handedness was included as a covariate in addition to age and gender because of the potential association between lateralisation of white matter network and handedness.⁴⁰ We identified the voxels where FA was negatively associated with each CeVD marker, using permutation-based non-parametric test (FSL RANDOMISE). Results were thresholded at p < 0.05 with voxel-wise threshold-free cluster enhancement (TFCE) family-wise error (FWE) correction.⁴¹ The same methods were used to identify voxels where MD was positively related to each CeVD marker.

Mediation analyses

To evaluate whether and how white matter microstructural changes mediate the effects of CeVD markers on cognition, we performed path analyses by including CeVD markers significantly associated with worse cognition as predictors, white matter microstructure as mediators, cognitive domain z-scores as outcomes, controlling for age and gender using structural equation modelling method (R (v 3.3.1) packages Lavaan (0.5–20)). To represent white matter microstructure, we created brain masks containing only the voxels that were significantly correlated with all the cognition-related CeVD markers (one brain mask based on the FA general linear model analyses and another similar one on the MD). We grouped these significant voxels into white matter regions of interests (ROIs) according to the white matter atlas ICBM-152.⁴² ROIs with less than 20 significant voxels were excluded from further calculation. ROIs were further categorized according to their fiber types (projection, association, commissural, limbic system and brainstem fibers). Mean FA and MD from ROIs of the same fiber types were averaged to obtain the overall FA and MD estimates of each fiber type. FA and MD of each fibre type were entered into their respective models as mediators. We created one FA model and one MD model for each of the cognitive domains.

We first created a just-identified model where all the variances and covariances of the variables (CeVD markers and fiber types) and all the direct effects were estimated. To test whether the effect of predictors on the cognitive outcomes could be explained by the mediation effect of one fiber type, we created single-fiber model where only one path between the fiber type mediators and the cognitive function was to be estimated. For example, we allowed the path from the projection fiber mediator to the executive function to be freely estimated while fixing the paths from the remaining fiber type mediators to executive function to zero. The number of single-fiber models that were created equals the number of fiber type mediators in the just-identified models (i.e. with four fiber type mediators, we created four single-fiber models for each cognitive domain).

We evaluated the model fits of the single-fiber model using Chi-square test, Root Mean Square Error of Approximation (RMSEA), Standardized Root Mean Square Residual (SRMR) and Comparative Fit Index (CFI). We defined a satisfactory fitting on the following criteria:^43,44 (1) The models should not be significantly different from the just-identified model based on the Chi-square tests; (2) RMSEA should not be >0.05 with upper boundary of the confidence interval (CI) <1; (3) SRMR should be <0.08; (4) CFI should be >0.95. If single-fiber models did not yield satisfactory fitting, we allowed two paths from the mediators to the cognitive function to be non-zero. The selection of the path to be freed was based on both modification index (MI) and the standardized expected parameter changes (SEPC).⁴⁵ The path was only added if the MI was larger than 6.635 (the Chi-square value for 1 degree of freedom at p = 0.01 was chosen to minimize the probability of type-I error) and the SEPC was >0.2.⁴⁶ For the final model of each cognitive domain, the non-significant paths were further eliminated. We also eliminated the non-significant covariance estimation. In all previous steps, we allowed all the covariance to be estimated. We only fixed the covariance to be zero in the final model when the covariance was not significant in previous steps. The significant covariance was still estimated in the final models. We also tested the indirect effect of each predictor based on the final model.

Three validation analyses followed after we identified the final models. First, considering the potential effect of education on cognition, we added education as a predictor to each of the final models. We examined the model fit again after keeping the significant but removing non-significant education-related paths. Second, given that the AD is the most common cause of cognitive impairment in elderly, we added hippocampus volume as a mediator to each of the final models to evaluate the disease impact. Third, we introduced diagnosis group as an additional factor to all final models to see whether the relationships described in the final models are applicable to both NCI and CIND. We compared two models: model (1), we fixed all regression coefficients to be the same between the two groups; and model (2), we allowed all regression coefficients to be freely estimated. Comparing model (1) and (2), a non-significant finding would indicate that the same relationships are applicable to both groups.

Results

Table 1 provides characteristics of the included participants according to the two diagnostic groups. Persons diagnosed with CIND were older, consisted of more women, had lower education and increased burden of lacunes, cortical infarcts, ICS and WMH compared to NCI.

Table 1.

Characteristics of study participants.

	NCI (n = 114)	CIND (n = 139)	P-values
Demographics
Age, years	67.4 (4.8)	72.5 (6.1)	<0.001
Women, No. (%)	48 (42.1)	81 (58.3)	0.010
Education, mean (SD)	7.9 (4.8)	3.9 (4.1)	<0.001
Handedness, Right, No. (%)	111 (97.4)	135 (97.1)	0.407
Vascular risk factors
Hypertension, No. (%)	85 (74.6)	115 (82.7)	0.112
Diabetes, No. (%)	24 (21.1)	42 (30.2)	0.099
Hyperlipidemia, No. (%)	60 (52.6)	87 (59.2)	0.110
CeVD markers
Presence of lacunes, No. (%)	5 (4.4)	36 (25.9)	<0.001
Presence of cortical CMIs, No. (%)	2 (1.8)	9 (6.5)	0.067
Presence of microbleeds, No. (%)	32 (28.1)	49 (35.3)	0.223
Presence of cortical infarcts, No. (%)	0	6 (4.3)	0.025
Presence of Intracranial stenosis, No. (%)	4 (3.5)	22 (15.8)	0.001
WMH, ml, median (IQR)	1.24 (3.2)	2.78 (5.9)	0.001
Hippocampal volume, ml, mean (SD)	7.2 (0.92)	7.1 (0.89)	0.343

P values indicate the differences between variables based on two-sample t-test or Pearson’s χ² test or Mann-Whitley U test (p < 0.05).

No. and percentages refer to the presence of respective characteristic in the study participants.

CIND: cognitive impairment no dementia; CMIs: cortical cerebral microinfarcts; IQR: interquartile range; ml: milliliters; NCI: no cognitive impairment; no: number; SD: Standard deviation.Note: Significant statistical test results are marked in bold.

Association of CeVD markers with cognition

Lacunes, WMH and ICS were consistently associated with impaired performance in the domains of executive function, language, attention, verbal and visual memory after controlling for age and gender (Table 2). Most of these associations remained after further controlling for education, vascular risk factors, hippocampus volume, and diagnosis. After applying Bonferroni correction, lacunes remained significantly associated with visual memory whereas ICS was associated with executive function. No associations were observed between cerebral microbleeds, cerebral microinfarcts and cortical infarcts with cognition. Residual normality tests found no significant deviation from normal distribution. In view of these findings, lacunes, WMH and ICS and the five domains were subsequently used in path analyses.

Table 2.

Age and gender adjusted associations of cerebrovascular disease markers with cognition.

	Executive function Mean difference (95%CI)	Attention Mean difference (95%CI)	Language Mean difference (95%CI)	Visuomotor speed Mean difference (95%CI)	Visuoconstruction Mean difference (95%CI)	Verbal memory Mean difference (95%CI)	Visual memory Mean difference (95%CI)
Lacunes	–0.28 (–0.54; –0.03)	–0.26 (–0.51; –0.01)	–0.26 (–0.52; 0.00)	–0.31 (–0.53; –0.09)	–0.16 (–0.40; 0.07)	–0.26 (–0.51; –0.02)	–0.39 (–0.62; –0.15)
	P = 0.029	P = 0.043	P = 0.052	P = 0.006^a	P = 0.179	P = 0.038	P = 0.001^a,b
WMH	–0.35 (–0.65; –0.05)	–0.39 (–0.69; –0.09)	–0.35 (–0.66; –0.04)	–0.25 (–0.52; 0.02)	–0.41 (–0.69; –0.13)	–0.39 (–0.69; –0.05)	–0.29 (–0.57; –0.02)
	P = 0.024^b	P = 0.011^b	P = 0.028^b	P = 0.066	P = 0.005^a,b	P = 0.008^b	P = 0.038^b
CMB	–0.02 (–0.06; 0.01)	–0.00 (–0.04; 0.03)	0.00 (–0.04; 0.04)	–0.00 (–0.04; 0.029)	–0.03 (–0.06; 0.00)	0.00 (–0.04; 0.04)	–0.01 (–0.05; 0.02)
	P = 0.211	P = 0.924	P = 0.916	P = 0.857	P = 0.073	P = 0.988	P = 0.417
CMI	–0.23 (–0.61; 0.15)	0.18 (–0.19; 0.56)	0.17 (–0.22; 0.56)	0.038 (–0.29; 0.37)	0.28 (–0.07; 0.64)	0.03 (–0.33; 0.40)	0.03 (–0.32; 0.38)
	P = 0.234	P = 0.342	P = 0.399	P = 0.820	P = 0.116	P = 0.857	P = 0.870
Cortical infarcts	0.29 (–0.34; 0.91)	–0.29 (–0.91; 0.34)	–0.27 (–0.91; 0.38)	–0.12 (–0.67; 0.43)	–0.16 (–0.75; 0.43)	–0.17 (–0.78; 0.44)	–0.59 (–1.17; –0.01)
	P = 0.369	P = 0.366	P = 0.418	P = 0.668	P = 0.597	P = 0.584	P = 0.045
ICS	–0.23 (–0.38; –0.07)	–0.16 (–0.31; –0.00)	–0.19 (–0.35; –0.04)	–0.19 (–0.33; –0.06)	–0.23 (–0.37; –0.08)	–0.19 (–0.35; –0.05)	–0.19 (–0.33; –0.05)
	P = 0.004^a,b	P = 0.045	P = 0.016	P = 0.004^a,b	P = 0.002^a	P = 0.009	P = 0.008

^aIndicates significant after Bonferroni correction at p < 0.05 (0.05/7 = 0.0071).

^bIndicates significant after adjusting for vascular risk factors and diagnosis.

CI: confidence interval; CMB: cerebral microbleeds; CMI: cerebral microinfarcts; ICS: intracranial stenosis; WMH: white matter hyperintensities.Note: Significant statistical test results are marked in bold.

Association of CeVD markers with white matter microstructure deterioration

Increasing CeVD markers were associated with lower FA and higher MD in widespread brain regions across all participants (Supplementary Tables 1 and 2). Interestingly, among the three cognition-related CeVD markers, lacunes and WMH, markers of small vessel disease, were associated with lower FA and higher MD in all five fiber types (projection, association, commissural, limbic and brainstem) (TFCE FWE corrected p < 0.05) (Figure 1). Only four fiber types including projection, association, limbic and commissural fibers were significantly linked with ICS, a marker of large vessel disease (TFCE FWE corrected p < 0.05). After combining the findings from these three CeVD markers, we focused on FA and MD of four fiber types in subsequent path analyses. Among the remaining three CeVD markers, lower FA and higher MD in all five fiber types were related to cerebral microbleeds. But cerebral microinfarcts and cortical infarcts were not significantly linked to FA values in any regions and only related to MD values in limited regions.

Figure 1.

CeVD markers were associated with lower fractional anisotropy and higher mean diffusivity in projection, association, commissural, and limbic fibers. We built a voxel-wise general linear model based on TBSS with the skeletonized FA or MD images as the dependent variable, the six CeVD markers (WMH, cortical infarct, lacunar infarct, cerebral microinfarct, cerebral microbleeds and ICS) as the covariates of interest, and age, gender, and handedness as the nuisance variables. Voxels (highlighted in blue) where lower FA was related to each CeVD marker and voxels (highlighted in red) where higher MD was linked to each CeVD marker were identified (TFCE FWE corrected p < 0.05). The top row also showed the overlap of voxels that were linked to the three cognition-related CeVD markers (WMH, lacunar infarct and ICS). We grouped these significant voxels into regions (labeled in yellow) according to the white matter atlas ICBM-152³² which could be categorized into projection, commissural, association, and limbic fibers (Supplementary Tables 1 and 2). The group-level skeletonized mean FA (in green) was thresholded to reflect major white matter tracts in the brain and displayed against the mean FA image as the background.

Path analyses

We found that decreased mean FA of the commissural fibers mediated the effects of increasing lacunar counts (p = 0.015), higher WMH volume (p = 0.011), and ICS (p = 0.022) on executive function (Figure 2a, Table 3). In addition, lacunar counts and ICS contributed directly to executive dysfunction. FA of the commissural fibers mediated the effect of lacunar counts (p = 0.012), WMH volume (p = 0.009) and ICS (p = 0.019) on verbal memory. WMH and ICS were also associated with visual memory directly (Figure 2d, Table 3). The FA of projection fibers mediated the effects of lacunar counts (p < 0.001), WMH volume (p < 0.001) and ICS (p = 0.001) on attention. Since all the direct paths from CeVD markers to attention were not significant, it suggested a possible complete mediation effect of the white matter microstructure in projection fibers (Figure 2b, Table 3). The FA of the projection fibers also mediated the effect of lacunar counts (p = 0.005), WMH volume (p = 0.003) and ICS (p = 0.009) on visual memory. Lacunar counts and ICS were also directly associated with visual memory independent of microstructural changes (Figure 2c, Table 3).

Figure 2.

White matter microstructural changes were mediators between CeVD markers and cognition. Schematic diagram of the path analyses using lacunar counts, white matter hyperintensities volume and number of intracranial stenosis as predictors, mean fractional anisotropy and mean diffusivity of commissural, projection, association and limbic fibers as mediators and cognition as outcomes. Age, sex and education were added as covariates. Paths with (a) Executive function, (b) Attention, (c) Visual memory, (d) Verbal memory, and (e) Language uses FA as a mediator whereas paths with (f) Executive function and (g) Visual memory uses MD as a mediator. Numbers on the paths indicate standardized coefficients that were statistically significant.

Table 3.

Effects of CeVD markers (lacunes, white matter hyperintensities, intracranial stenosis) on cognition (executive function, attention, visual memory, verbal memory and language) through mediators (commissural, projection, association and limbic fibers).

Commissural

Projection

Association

Limbic

Executive function

Path analysis A

Lacunes

–0.477

0.106

<0.001

–0.502

0.097

<0.001

–0.367

0.122

0.003

–0.302

0.118

0.011

–0.322

0.119

0.007

WMH

–0.686

0.132

<0.001

–0.856

0.122

<0.001

–0.553

0.149

<0.001

–0.485

0.148

0.001

ICS

–0.250

0.067

<0.001

–0.264

0.061

<0.001

–0.233

0.077

0.003

–0.207

0.075

0.006

–0.172

0.075

0.021

Commissural

0.190

0.066

0.004

Model fit

(χ² (degree of freedom (df) = 14) = 19.054, p = 0.163, CFI = 0.994, RMSEA (CI) = 0.038 (0.000, 0.076), and SRMR = 0.042)

Path analysis B

Commissural

Projection

Association

Limbic

Attention

Lacunes

–0.453

0.105

<0.001

–0.514

0.097

<0.001

–0.368

0.122

0.003

–0.306

0.118

0.010

WMH

–0.797

0.117

<0.001

–0.797

0.117

<0.001

–0.548

0.149

<0.001

–0.465

0.148

0.002

ICS

–0.250

0.067

<0.001

–0.263

0.061

<0.001

–0.233

0.077

0.003

–0.207

0.075

0.006

Projection

0.319

0.060

<0.001

Model fit

(χ² (df = 17) = 21.872, p = 0.190, CFI = 0.994, RMSEA (CI) = 0.034 (0.000, 0.070), and SRMR = 0.042)

Path analysis C

Commissural

Projection

Association

Limbic

Visual memory

Lacunes

–0.477

0.106

<0.001

–0.502

0.097

<0.001

–0.367

0.122

0.003

–0.302

0.118

0.011

–0.315

0.112

0.005

WMH

–0.686

0.132

<0.001

–0.856

0.121

<0.001

–0.553

0.149

<0.001

–0.485

0.148

0.001

ICS

–0.250

0.067

<0.001

–0.264

0.062

<0.001

–0.233

0.077

0.003

–0.207

0.075

0.006

–0.169

0.069

0.015

Projection

0.212

0.064

0.001

Model fit

(χ² (df = 13) = 14.391, p = 0.347, CFI = 0.998, RMSEA (CI) = 0.021 (0.000, 0.067) and SRMR = 0.032)

Path analysis D

Commissural

Projection

Association

Limbic

Verbal memory

Lacunes

–0.477

0.106

<0.001

–0.502

0.097

<0.001

–0.367

0.122

0.003

–0.302

0.118

0.011

WMH

–0.686

0.132

<0.001

–0.856

0.121

<0.001

–0.553

0.149

<0.001

–0.485

0.148

0.001

–0.335

0.144

0.020

ICS

–0.250

0.067

<0.001

–0.264

0.062

<0.001

–0.233

0.077

0.003

–0.207

0.075

0.006

–0.180

0.072

0.012

Commissural

0.191

0.063

0.002

Model fit

(χ² (df = 13) = 10.555, p = 0.648, CFI = 1.000, RMSEA (CI) = 0.000 (0.000, 0.052) and SRMR = 0.028)

Path analysis E

Commissural

Projection

Association

Limbic

Language

Lacunes

–0.477

0.106

<0.001

–0.502

0.097

<0.001

–0.367

0.122

0.003

–0.302

0.118

0.011

WMH

–0.686

0.132

<0.001

–0.856

0.121

<0.001

–0.553

0.149

<0.001

–0.485

0.148

0.001

ICS

–0.250

0.067

<0.001

–0.264

0.062

<0.001

–0.233

0.077

0.003

–0.207

0.075

0.006

–0.152

0.077

0.047

Projection

0.328

0.078

<0.001

Limbic

–0.150

0.073

0.039

Model fit

(χ² (df = 14) = 16.341, p = 0.293, CFI = 0.997, RMSEA (CI) = 0.026 (0.000, 0.069), and SRMR = 0.036)

Path analysis F

Commissural

Projection

Association

Limbic

Executive function

Lacunes

0.419

0.111

<0.001

0.484

0.088

<0.001

0.283

0.102

0.006

0.396

0.105

<0.001

–0.296

0.118

0.013

WMH

0.586

0.139

<0.001

1.333

0.111

<0.001

1.098

0.128

<0.001

0.825

0.132

<0.001

ICS

0.160

0.070

0.022

0.107

0.056

0.177

0.064

0.006

0.154

0.066

0.020

–0.195

0.073

0.008

Commissural

–0.162

0.065

0.013

Model fit

(χ2 (df = 12) = 16.090, p = 0.187, CFI = 0.997, RMSEA (CI) = 0.037 (0.000, 0.079), and SRMR = 0.042)

Path analysis G

Commissural

Projection

Association

Limbic

Visual memory

Lacunes

0.476

0.109

<0.001

0.522

0.087

<0.001

0.324

0.101

0.001

0.451

0.103

<0.001

–0.308

0.111

0.005

WMH

0.584

0.140

<0.001

1.332

0.111

<0.001

1.097

0.128

<0.001

0.823

0.132

<0.001

–0.277

0.138

0.045

ICS

0.062

0.036

0.083

–0.199

0.068

0.003

Commissural

–0.170

0.060

0.005

Model fit

(χ2 (df = 15) = 21.479, p = 0.122, CFI = 0.995, RMSEA (CI) = 0.041 (0.000, 0.078), and SRMR = 0.051)

Values represent the result of the path analyses with different cognitive domains. Path analyses A-E used mean FA of the fiber type as the white matter tract metrics. Path analysis A shows that the effects of lacunes, WMH and ICS on executive function is mediated by commissural fibers. Path analysis B shows the effects of lacunes, WMH and ICS on attention were mediated by projection fibers. Path analysis C shows the effects of lacunes, WMH and ICS on visual memory mediated by projection fibers. Path analysis D shows the mediation effects of commissural fibers between lacunes, WMH and ICS and verbal memory. Path analysis E shows the effects of lacunes, WMH and ICS on language which is mediated by projection and limbic fiber. Path analyses F-G used mean MD of the fiber type as the white matter tract metrics. Path analysis F shows that the effects of lacunes, WMH and ICS on executive function is mediated by commissural fibers. Path analysis G shows that the effects of lacunes, WMH and ICS on visual memory is mediated by commissural fibers.

ICS: intracranial stenosis; SE: standard error; WMH: white matter hyperintensities; β: regression coefficient.

For the language domain, none of the single-fiber models yielded satisfying model fit. Therefore, we included two mediators in the model. Only the path from the projection fiber FA to the language scores in the limbic-fiber model fulfilled the modification requirements (MI >6.635 and SEPC >0.2). Allowing paths from the projection fiber mediator and the limbic fiber mediator to the language scores to be freely estimated yielded satisfactory model fitting (Table 3 and Figure 2e). The FA of projection fibers mediated the effect of lacunar counts (p = 0.001), WMH volume (p < 0.001) and ICS (p = 0.003) on language. Though the FA of limbic fibers was associated with both the language scores and all three CeVD markers, the mediation effect of the limbic fiber FA was not significant. Additionally, ICS was independently associated with impaired language domain without being mediated by white matter microstructural changes.

The MD in the commissural fibers mediated the effect of lacunar counts (executive function: p = 0.039; visual memory: p = 0.017) and WMH volume (executive function: p = 0.033; visual memory: p = 0.019) on the executive function and the visual memory domain (Table 3F–G). The direct effect of ICS on executive function was significant (p = 0.010) but not the indirect effect via the commissural fiber MD (p = 0.069). No effects of the CeVD markers on the other cognitive domain were mediated by the white matter fiber MD.

In all models, the associations between CeVD markers and white matter microstructure were estimated simultaneously. Therefore, the coefficient of the lacune-FA path, for example, reflected the effect of lacune on FA when controlling for the effects of WMH volumes and ICS. For the covariance among CeVD markers, we found that the lacunes were significantly correlated with both ICS (beta = 0.084, p < 0.001) and WMH volumes (beta = 0.062, p < 0.001), while the association between WMH volumes and ICS was not significant.

Considering the potential effect of education, we added this as a predictor to all the final models. Education was significantly related to all five cognitive functions. Education was also associated with limbic FA, but not with FA or MD of other fibers. Among the covariance, older age participants have lower education. Importantly, all the mediation effects remained significant after education was added to the models. Upon adding hippocampus volume as a mediator to all the final models, no significant association was observed between hippocampal volume and other variables and hence, all paths were eventually pruned away following this procedure.

When we applied our final models to NCI and CIND separately, with regression coefficients either fixed to be the same or freely estimated, the model fits were not significantly different for executive function, language, verbal memory and visual memory domains. It suggested that the relationship among the CeVD markers, the white matter microstructure, and the cognitive outcomes did not differ significantly between NCI and CIND. Estimating the coefficients separately for each group was preferred for the attention domain. Applying the same final model to the CIND group alone indicated good model fit (χ² (df = 17) = 19.836, p = 0.283, CFI = 0.994, RMSEA (CI) = 0.035 (0.000, 0.088), and SRMR = 0.062), suggesting the relationship remained applicable in CIND. For NCI, by applying the just-identified model and repeating the pruning procedure revealed that projection (p = 0.020) and association (p = 0.023) fiber FA together fully mediated the effect of WMH volume on attention.

Discussion

The present study demonstrates that CeVD markers have both direct and indirect effects (via disruption of white matter microstructure) on cognitive performance in elderly with normal cognition or MCI. Path analyses revealed that CeVD markers, specifically increased lacunar counts, higher WMH volume, and ICS were associated with worse cognition through disruption of white matter network in commissural and projection fibers. A complete mediation effect was observed with attention. Moreover, these three CeVD markers were directly related to altered cognitive performance particularly executive function, visual memory, verbal memory and language without being fully mediated by white matter network disruption. No association was observed between cerebral microbleeds and cerebral microinfarcts with cognition. Taken together, our findings provide new insights into specific white matter pathophysiological mechanisms underlying CeVD burden and how these CeVD markers can give rise to worse performance in several cognitive domains.

Several studies have shown associations between CeVD markers and microstructural changes in white matter.^47–49 It is also reported that WMH were linked with cerebral hypometabolism and cognitive dysfunction regardless of their location.⁵⁰ Furthermore, both lacunes and WMH disrupt white matter network which leads to frontal/temporoparietal atrophy and frontal-executive/memory dysfunctions.^17,51 Till today no study has yet examined the effects of the large vessel disease on white matter network and its eventual effect on cognition. Compared to the widespread disruption of white matter microstructure by small vessel disease (lacunes, WMH, cerebral microbleeds), we showed that the effects of large vessel disease on white matter microstructure disruption were relatively restricted. We did not observe association between brain stem fibre type and ICS. Cortical infarcts were neither linked to white matter FA nor cognition. Nevertheless, our data add further evidence to support that ICS in addition to lacunes and WMH may induce cerebral hypoperfusion and ischemic changes in the white matter causing demyelination and axonal loss of the white matter network.^36,52 The relatively meagre blood supply of the white matter tracts may increase their vulnerability to hypoxia and to the effects of both small and large vessel damage. Injury to the axons leads to degeneration of the neuron both proximally and distally (Wallerian degeneration) resulting in cognitive dysfunction.^53,54

Our data supported that the CeVD markers (lacunes, WMH and ICS) induced fiber specific microstructural changes which lead to impairment in specific cognitive domains. All three CeVD markers were related to impairment in executive function and verbal memory via damage to commissural fibers microstructure (reflected by lower FA and increased MD) suggesting robust findings. Emerging evidence has suggested that commissural fibers which constitute the major tracts of corpus callosum may play an important role in memory processing due to inter-hemispheric interaction.⁵⁵ It is reported that the corpus callosum facilitates more efficient learning and recall for verbal information. Moreover, the corpus callosum is known to connect regions within intrinsic connectivity networks such as the default mode network and executive control network, providing structural highways for functional network processing.^55,56 Damage to the corpus callosum by subcortical ischemic damage may thus contribute to network dysconnectivity by disrupting the inter-hemispheric information transfer leading to memory and executive impairment.

In this study, CeVD markers were also related to attention and visual memory deficits via microstructural damage to projection fibers (i.e. lower FA). A recent study has shown that projection fibers (consisting of corticospinal, corticobulbar and thalamic radiation) functionally connect frontal, parietal and occipital cortical regions and thus create a pattern of convergent connectivity that could integrate working memory, executive control, and spatial attention.⁵⁷ Moreover, disruption of connectivity in prefrontal-subcortical (e.g. thalamus to frontal cortex connections) and pre-frontal-parietal (superior longitudinal fasciculus) circuits are associated with age-related memory dysfunction as well as attention.⁵⁸ We propose that damage to this circuits may reflect a potential mechanism for deficits observed in attention and visual memory. Interestingly, higher MD in commissural fibers could also explain the effects of increased lacunes and higher WMH volumes on visual memory impairment. It further emphasized the important role of the commissural fibers in bridging the CeVD markers and cognitive functions. However, when comparing the model fit measures between projection FA and commissural MD to estimate white matter microstructure, the model using projection FA performs better than the model using commissural MD in explaining the association between CeVD markers and visual memory impairment. Furthermore, the association between ICS and visual memory impairment could not be explained by increased MD in the commissural fibers, thus supporting the crucial mediation of the projection fibers. To ensure that the MD values entering the MD path models were from voxels associated with the CeVD markers, we identified the voxels using a separate MD model instead of calculating the MD from the voxels obtained from the FA model. As a consequence, FA and MD values in the path models were not from exactly the same voxels. But substantial numbers of ROIs were included in both the FA and MD path models (Supplementary Tables 1 and 2). Nevertheless, comparison between FA and MD path models should be interpreted with caution. In terms of the language domain, besides projection fibers, limbic fibers were related to both CeVD markers and language dysfunction. But the mediation effect between CeVD markers and language was shouldered by projection, not limbic fibers. The complex trajectory of the limbic fibers involves tracts from the temporal lobe, the superior portion of which is involved in the speech perception. Ischemic processes may thus affect white matter in both projection and limbic fibers leading to neuronal degeneration and cognitive dysfunction. However, it is important to note that language impairment in our study might also be due to lexical retrieval difficulties or semantic memory loss in participants that involve multiple pathways.

Moreover, lacunes, WMH, and ICS have direct effects upon cognitive performance in four different domains. However, the direct association pattern between these CeVD markers and cognitive performance was different suggesting different pathophysiology underlying these imaging features of CeVD. The direct association of lacunes with executive function and visual memory and that of WMH with verbal memory supports the findings that lacunes represent the extreme end of the CeVD spectrum and may very well have a greater impact on complex cognitive domains, i.e. memory, executive function, and perceptual speed compared to WMH. Furthermore, as our study was a subsample of a population-based setting, it is still conceivable that the individuals in our study did not have significant burden of WMH required for wide spread disruption. On the other hand, ICS affected all four cognitive domains including executive function, visual memory, verbal memory and language suggesting that large vessel disease impairs overall cognitive performance through diffuse ischemia. Cerebral hypoperfusion caused by carotid stenosis disrupts tasks requiring cortical integration (i.e., those that rely on “distributed systems”) and may be particularly susceptible to disconnection and cognitive dysfunction.^17,36 Our findings are thus in line with a recent study that showed that CeVD burden score explained a substantial proportion of the variance in cognitive function on top of the markers of atrophy and PiB amyloid.¹⁶

The strengths of our study include: a population-based study, availability of a wide range of CeVD markers to reflect vascular brain damage, and an extensive neuropsychological battery to provide comprehensive information on multiple cognitive domains. Limitations of our study include: First, as the current study is of cross-sectional design, we can only suggest associations. Here, we presumed that microstructural changes precede CeVD markers. However, the causative link between CeVD markers, DTI and cognition cannot be assessed and our findings do not exclude the possibility that white matter microstructural changes may take place independent of or even before these CeVD markers. Second, as AD is the most common cause of cognitive impairment in elderly, we adjusted for the surrogate marker of AD i.e. hippocampal volume in our models where most of the associations remained. As our sample consists of controls (no cognitive impairment) and CIND, we postulate that hippocampal atrophy may not be pronounced and contribute to the path model. On the other hand, it is also possible that the effects of AD is not fully excluded by simply regressing out the effects of hippocampal volume which is a crude and non-specific marker of AD. Third, though a differential pattern of relationships was observed across cognitive domains (specifically executive function and visual memory) and white matter microstructures, this evidence can be better corroborated in larger studies with longitudinal design and with a higher prevalence of CeVD markers. Lastly, enlarged perivascular spaces and cerebral perfusion should be considered in future CeVD studies.

In conclusion, our study shows that differential patterns of white matter microstructure abnormalities underlie the relationship between various types of CeVD burden and domain-specific cognitive impairment in older adults. Further longitudinal studies are needed to understand the sequential order and cause–effect relationship between CeVD, white matter damage, and cognitive decline.

Supplemental Material

sj-pdf-1-jcb-10.1177_0271678X21990980 - Supplemental material for White matter network damage mediates association between cerebrovascular disease and cognition

Supplemental material, sj-pdf-1-jcb-10.1177_0271678X21990980 for White matter network damage mediates association between cerebrovascular disease and cognition by Saima Hilal, Siwei Liu, Tien Yin Wong, Henri Vrooman, Ching-Yu Cheng, Narayanaswamy Venketasubramanian, Christopher LH Chen and Juan Helen Zhou in Journal of Cerebral Blood Flow & Metabolism

Footnotes

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Epidemiology of Dementia in Singapore study is supported by the National Medical Research Council (NMRC), Singapore (NMRC/CG/NUHS/2010 [Grant no: R-184-006-184-511]), (NMRC/CSA/038/2013) and Bight Focus Foundation [R‐608‐000‐248‐597]. The research conducted in this study is also supported by the National Medical Research Council, Singapore (NMRC0088/2015), the Duke-NUS Medical School Signature Research Program funded by Ministry of Health, Singapore, and Center for Sleep and Cognition funded by Yong Loo Lin School of Medicine, National University of Singapore.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

SH is responsible for study concept and design, participated in data acquisition, performed statistical analysis, drafting and revising the manuscript and obtaining funding. SL is responsible for study concept and design, participated in data acquisition, performed statistical analysis, drafting and revising the manuscript. TYW, HV, CYC, NV and CC participated in data acquisition and revising the manuscript for intellectual content. JHZ was responsible for study concept and design, data interpretation, obtaining funding and revising the manuscript. De-identified data are available upon reasonable requests from the corresponding authors.

ORCID iDs

Saima Hilal

Siwei Liu

Juan Helen Zhou

Supplementary material

Supplemental material for this article is available online.

References

Hilal

Mok

Youn

, et al. Prevalence, risk factors and consequences of cerebral small vessel diseases: data from three Asian countries. J Neurol Neurosurg Psychiatry 2017; 88: 669–674.

Hilal

Sikking

Shaik

, et al. Cortical cerebral microinfarcts on 3T MRI: a novel marker of cerebrovascular disease. Neurology 2016; 87: 1583–1590.

Liu

Braidy

Poljak

, et al. Cerebral small vessel disease and the risk of Alzheimer's disease: a systematic review. Ageing Res Rev 2018; 47: 41–48.

Rensma

van Sloten

Launer

, et al. Cerebral small vessel disease and risk of incident stroke, dementia and depression, and all-cause mortality: a systematic review and meta-analysis. Neurosci Biobehav Rev 2018; 90: 164–173.

Pasi

van Uden

Tuladhar

, et al. White matter microstructural damage on diffusion tensor imaging in cerebral small vessel disease: clinical consequences. Stroke 2016; 47: 1679–1684.

Tae

Ham

Pyun

, et al. Current clinical applications of diffusion-tensor imaging in neurological disorders. J Clin Neurol 2018; 14: 129–140.

Charlton

Landau

Schiavone

, et al. A structural equation modeling investigation of age-related variance in executive function and DTI measured white matter damage. Neurobiol Aging 2008; 29: 1547–1555.

Head

Kennedy

Rodrigue

, et al. Age differences in perseveration: cognitive and neuroanatomical mediators of performance on the Wisconsin card sorting test. Neuropsychologia 2009; 47: 1200–1203.

Head

Rodrigue

Kennedy

, et al. Neuroanatomical and cognitive mediators of age-related differences in episodic memory. Neuropsychology 2008; 22: 491–507.

10.

Vermeer

Longstreth

Jr. Koudstaal

PJ.

Silent brain infarcts: a systematic review. Lancet Neurol 2007; 6: 611–619.

11.

Debette

Markus

HS.

The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: systematic review and meta-analysis. BMJ 2010; 341: c3666.

12.

Lei

Lin

Tao

, et al. Association between cerebral microbleeds and cognitive function: a systematic review. J Neurol Neurosurg Psychiatry 2013; 84: 693–697.

13.

Edwards

Jacova

Sepehry

, et al. A quantitative systematic review of domain-specific cognitive impairment in lacunar stroke. Neurology 2013; 80: 315–322.

14.

van Sloten

Protogerou

Henry

, et al. Association between arterial stiffness, cerebral small vessel disease and cognitive impairment: a systematic review and meta-analysis. Neurosci Biobehav Rev 2015; 53: 121–130.

15.

Debette

Schilling

Duperron

, et al. Clinical significance of magnetic resonance imaging markers of vascular brain injury: a systematic review and meta-analysis. JAMA Neurol 2019; 76: 81–94.

16.

Banerjee

Jang

Kim

, et al. Total MRI small vessel disease burden correlates with cognitive performance, cortical atrophy, and network measures in a memory clinic population. J Alzheimers Dis 2018; 63: 1485–1497.

17.

Kim

Kwon

, et al. Clinical effect of white matter network disruption related to amyloid and small vessel disease. Neurology 2015; 85: 63–70.

18.

Hilal

Ikram

Saini

, et al. Prevalence of cognitive impairment in Chinese: epidemiology of dementia in Singapore study. J Neurol Neurosurg Psychiatry 2013; 84: 686–692.

19.

Sahadevan

Lim

Tan

, et al. Diagnostic performance of two mental status tests in the older Chinese: influence of education and age on cut-off values. Int J Geriat Psychiatry 2000; 15: 234–241.

20.

Dubois

Slachevsky

Litvan

, et al. The FAB: a frontal assessment battery at bedside. Neurology 2000; 55: 1621–1626.

21.

Porteus

SD.

The maze test and clinical psychology. Palo Alto: Pacific Books, 1959.

22.

Wechsler

Wechsler memory scale – revised. 3rd ed. San Antonio: Jovanovich, 1997.

23.

Lewis

Rennick

PM.

Manual for the repeatable cognitive perceptual-motor battery. Clinton Township: Axon, 1979.

24.

Mack

Freed

Williams

, et al. Boston naming test: shortened versions for use in Alzheimer's disease. J Gerontol 1992; 47: 154–158.

25.

Isaacs

Kennie

AT.

The set test as an aid to the detection of dementia in old people. Br J Psychiatry 1973; 123: 467–470.

26.

Smith

Symbol digit modalities test. Western Psychological Services, Los Angeles, 1973.

27.

Diller

Ben-Yishay

Gerstman

LJ.

Studies in cognition and rehabilitation in hemiplegia. New York: New York University Medical Center Institute of Rehabilitation Medicine, 1974.

28.

Sunderland

Hill

Mellow

, et al. Clock drawing in Alzheimer's disease. A novel measure of dementia severity. J Am Geriatr Soc 1989; 37: 725–729.

29.

Wechsler

Wechsler adult intelligence scale - Revised. San Antonio: Harcourt Brace Jovanovich, 1981.

30.

Sahadevan

Tan

, et al. Cognitive testing of elderly Chinese people in Singapore: influence of education and age on normative scores. Age Ageing 1997; 26: 481–486.

31.

Hilal

Xin

Ang

, et al. Risk factors and consequences of cortical thickness in an Asian population. Medicine 2015; 94: e852.

32.

Hilal

Saini

Tan

, et al. Ankle-brachial index, cognitive impairment and cerebrovascular disease in a Chinese population. Neuroepidemiology 2014; 42: 131–138.

33.

Wardlaw

Smith

Biessels

, STandards for ReportIng Vascular changes on nEuroimaging (STRIVE v1)et al. Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol 2013; 12: 822–838.

34.

Cordonnier

Potter

Jackson

, et al. Improving interrater agreement about brain microbleeds: development of the brain observer MicroBleed scale (BOMBS). Stroke 2009; 40: 94–99.

35.

Hilal

Saini

Tan

, et al. Cerebral microbleeds and cognition: the epidemiology of dementia in Singapore study. Alzheimer Dis Assoc Disord 2014; 28: 106–112.

36.

Hilal

Saini

Tan

, et al. Intracranial stenosis, cerebrovascular diseases, and cognitive impairment in Chinese. Alzheimer Dis Assoc Disord 2015; 29: 12–17.

37.

Vrooman

Cocosco

van der Lijn

, et al. Multi-spectral brain tissue segmentation using automatically trained k-nearest-neighbor classification. Neuroimage 2007; 37: 71–81.

38.

Liu

Ong

Hilal

, et al. The association between retinal neuronal layer and brain structure is disrupted in patients with cognitive impairment and Alzheimer's disease. J Alzheimers Dis 2016; 54: 585–595.

39.

Smith

Jenkinson

Johansen-Berg

, et al. Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage 2006; 31: 1487–1505.

40.

Chen

Wang

, et al. Handedness- and hemisphere-related differences in small-world brain networks: a diffusion tensor imaging tractography study. Brain Connect 2014; 4: 145–156.

41.

Smith

Nichols

TE.

Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage 2009; 44: 83–98.

42.

Mori

Oishi

Jiang

, et al. Stereotaxic white matter atlas based on diffusion tensor imaging in an ICBM template. Neuroimage 2008; 40: 570–582.

43.

Bentler

PM.

Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equ Modeling 1999; 6: 1–55.

44.

Kline

RB.

Principles and practice of structural equation modeling. 3rd ed. New York: The Guilford Press, 2004.

45.

Whittaker

TA.

Using the modification index and standardized expected parameter change for model modification. J Exp Educ 2012; 80: 26–44.

46.

Saris

Satorra

van der Veld

WM.

Testing structural equation models or detection of misspecifications?

Struct Equ Modeling 2009; 16: 561–582.

47.

Chao

Decarli

Kriger

, et al. Associations between white matter hyperintensities and beta amyloid on integrity of projection, association, and limbic fiber tracts measured with diffusion tensor MRI. PloS One 2013; 8: e65175.

48.

Lee

Fletcher

Martinez

, et al. Regional pattern of white matter microstructural changes in normal aging, MCI, and AD. Neurology 2009; 73: 1722–1728.

49.

Kim

Kwon

, et al. Effects of amyloid and small vessel disease on white matter network disruption. J Alzheimers Dis 2015; 44: 963–975.

50.

Tullberg

Fletcher

DeCarli

, et al. White matter lesions impair frontal lobe function regardless of their location. Neurology 2004; 63: 246–253.

51.

Chong

JSX

Liu

Loke

, et al. Influence of cerebrovascular disease on brain networks in prodromal and clinical Alzheimer's disease. Brain 2017; 140: 3012–3022.

52.

Park

Seo

Kim

, et al. Effects of cerebrovascular disease and amyloid beta burden on cognition in subjects with subcortical vascular cognitive impairment. Neurobiol Aging 2014; 35: 254–260.

53.

Ginsberg

Martin

LJ.

Axonal transection in adult rat brain induces transsynaptic apoptosis and persistent atrophy of target neurons. J Neurotrauma 2002; 19: 99–109.

54.

Schallert

Jones

Lindner

MD.

Multilevel transneuronal degeneration after brain damage. Behavioral events and effects of anticonvulsant gamma-aminobutyric acid-related drugs. Stroke 1990; 21: III143–III146.

55.

Qiu

Liu

Hilal

, et al. Inter-hemispheric functional dysconnectivity mediates the association of corpus callosum degeneration with memory impairment in AD and amnestic MCI. Sci Rep 2016; 6: 32573.

56.

van den Heuvel

Mandl

Kahn

, et al. Functionally linked resting-state networks reflect the underlying structural connectivity architecture of the human brain. Hum Brain Mapp 2009; 30: 3127–3141.

57.

Jarbo

Verstynen

TD.

Converging structural and functional connectivity of orbitofrontal, dorsolateral prefrontal, and posterior parietal cortex in the human striatum. J Neurosci 2015; 35: 3865–3878.

58.

Lockhart

Mayda

Roach

, et al. Episodic memory function is associated with multiple measures of white matter integrity in cognitive aging. Front Hum Neurosci 2012; 6: 56.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.17 MB