White Matter Free Water Outperforms Cerebral Small Vessel Disease Total Score in Predicting Cognitive Decline in Persons with Mild Cognitive Impairment

Abstract

Background:

Vascular pathology is an important partner of Alzheimer’s disease (AD). Both total cerebral small vessel disease (CSVD) score and white matter free water (FW) are useful markers that could reflect cerebral vascular injury.

Objective:

We aim to investigate the efficacy of these two metrics in predicting cognitive declines in patients with mild cognitive impairment (MCI).

Methods:

We enrolled 126 MCI subjects with 3D T1-weighted images, fluid-attenuated inversion recovery images, T2^* images, diffusion tensor imaging images, cerebrospinal fluid biomarkers and neuropsychological tests from the Alzheimer’s Disease Neuroimaging Initiative database. The total CSVD score and FW values were calculated. Simple and multiple linear regression analyses were applied to explore the association between vascular and cognitive impairments. Linear mixed effect models were constructed to investigate the efficacy of total CSVD score and FW on predicting cognitive decline.

Results:

FW was associated with baseline cognition and could predict the decline of executive and language functions in MCI subjects, while no association was found between total CSVD score and cognitive declines.

Conclusion:

FW is a promising imaging marker for investigating the effect of CSVD on AD progression.

Keywords

Cerebral small vessel disease cognitive decline diffusion tensor imaging mild cognitive impairment

INTRODUCTION

Alzheimer’s disease (AD) is a common progressive neurodegenerative disease in the aged [1]. It is now widely accepted that vascular pathologies contribute to AD progression [2, 3], and it is possible to delay AD occurrence and progression by preventing stroke [4] and cerebral small vessel disease (CSVD) [5]. Strategic plans have been proposed and clinical trials are on the road. At this time, reliable and sensitive markers are needed for predicting cognitive decline and monitoring treatment effects.

While many previous studies found that individual CSVD imaging features including dilated perivascular spaces (dPVS), white matter hyperintensities (WMH), cerebral microbleeds (CMBs), and lacunes [6] were associated with AD progression [7, 8], these markers may present in one patient at the same time. Recently, the total CSVD score was proposed incorporating these four markers to reflect brain vascular burden [9, 10]. Studies revealed that the total CSVD score was associated with cognitive impairment and could predict cognitive decline either in community cohort [11 –13], patients with lacunar stroke [14, 15], mild cognitive impairment (MCI) [16], or neurodegeneration [17].

In addition, advanced diffusion models may provide sensitive markers [18]. Derived from the bi-tensor model, the free water (FW) fraction index can reflect mild brain injuries associated with vascular pathologies in relatively healthy adults, CSVD patients, and AD patients [19, 20]. Interestingly, white matter FW was not associated with AD pathological markers but associated with vascular damages [21]. Therefore, it could be used a surrogate marker for CSVD in AD studies. Likewise, FW is also associated with cognitive performance [22 –24] and can predict cognitive decline [25, 26].

Since both imaging metrics are useful vascular burden markers, it is of clinical interest to compare their performances. While assessment of the total CSVD score is relatively simple, it requires multi-modal images and may suffer from inter-rater variability. FW can be automatically calculated but it is not readily available. Investigating their efficacy in predicting cognitive decline may point the direction for future improvement. Yet to our best knowledge, no study has been performed to clarify this issue in MCI subjects.

In this research, we aim to investigate the potential of total CSVD score and FW in predicting the progression of cognitive declines in patients with MCI using longitudinal clinical data. We hypothesize that: 1) Both total CSVD score and FW are associated with baseline cognitive performance and can predict longitudinal cognitive declines in MCI subjects; 2) Compared with total CSVD score, FW may be relatively more sensitive.

MATERIALS AND METHODS

Data used in the preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (https://adni.loni.usc.edu). The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early AD. For up-to-date information, please see https://www.adni-info.org. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from all participants and/or authorized representatives and the study partners before any protocol-specific procedures were carried out in the ADNI study.

Subjects

We searched the ADNI database and selected 134 MCI subjects with available CSF biomarkers, 3D T1-weighted images, fluid-attenuated inversion recovery (FLAIR) images, T2 star images, DTI images (there were many different DTI acquisition directions, therefore we chose the direction with the largest number of subjects and longest follow-up years), and neuropsychological tests at baseline, as well as neuropsychological tests at per-year follow-up visits (varied from 0 to 10 years), including Mini-Mental State Exam, Montreal Cognitive Assessment, Trail Making Test-A (TMT-A), Trail Making Test-B (TMT-B), Rey Auditory Verbal Learning Test-Delayed Recall, Logical Memory-Delayed Recall, Category Fluency-Animals, and Boston Naming Test. Eight subjects were excluded because of converting to cognitively normal (CN) or AD during follow-up visits according to the “Diagnostic Summary [ADNI1,GO,2,3]” document (https://ida.loni.usc.edu/pages/access/studyData.jsp). After screening, the dataset contained 126 subjects from ADNIGO and ADNI2.

CSF analyses

CSF data were downloaded from the ADNI database. Baseline CSF samples were collected in the morning after an overnight fast and then processed, amyloid-β (Aβ) and phosphorylated tau (p-tau) were measured subsequently as previously described [27, 28]. Briefly, CSF was collected into syringes or collection tubes provided to each site, then transferred into polypropylene transfer tubes within 1 h after collection followed by freezing on dry ice and transported overnight on dry ice to the ADNI Biomarker Core laboratory at the University of Pennsylvania Medical Center. These samples were thawed at room temperature (1 h) and gently mixed to prepare aliquots (0.5 ml). These aliquots were stored at –80°C in bar code-labeled polypropylene vials. Aβ and p-tau were measured using the multiplex xMAP Luminex platform (Luminex Corp, Austin, TX) with Innogenetics (INNO-BIA AlzBio3; Ghent, Belgium; for research use-only reagents) immunoassay kit-based reagents.

Imaging acquisition

All subjects underwent 3T MRI scans using 3T GE Medical Systems scanners. Sequences were acquired as follows: 1) 3D T1 sequence (Flip Angle = 11.0 degree; Matrix X = 256 pixels; Matrix Y = 256 pixels; Matrix Z = 196 pixels; voxel size = 1.0^*1.0^*1.2 mm³; TE = 2.8∼3.2 ms; TI = 400 ms; TR = 7.0∼7.7 ms); 2) 2D DTI sequence (Flip Angle = 90.0 degree; Gradient Directions = 41.0; Matrix X = 128 pixels; Matrix Y = 128 pixels; voxel size = 2.7^*2.7^*2.7 mm³ reconstructed to 1.4^*1.4^*2.7 mm³; TE = minimum; TR = minimum); 3) 2D T2 FLAIR sequence (Flip Angle = 90.0/125.0 degree; Matrix X = 256 pixels; Matrix Y = 256 pixels; voxel size = 0.9^*0.9^*5.0 mm³; TE = 147.9∼154.0 ms; TI = 2250.0 ms; TR =11000.0 ms); 4) 2D T2 star sequence (Flip Angle = 20.0 degree; Matrix X = 256 pixels; Matrix Y = 256 pixels; voxel size = 0.8^*0.8^*4.0 mm³; TE =20.0 ms; TI = 0.0 ms; TR = 350.0/640.0/650.0 ms). Please note that ADNI is a multi-center project and that many different scanners had been used to acquire brain imaging data. Therefore, acquisition parameters may vary among different sites. Nevertheless, the acquisition protocols had been optimized to minimize inter-site differences. More details can be found at: https://adni.loni.usc.edu/methods/documents/mri-protocols.

CSVD visual rating

The CSVD visual rating was assessed by a neuroradiologist at first (ZZ). Perivascular spaces (PVS) were defined as CSF-like signal lesions (low signal on T1 images and high signal on T2 images), which were round, oval, or linear, with a maximum diameter of less than 3mm and a smooth outline. The severity degrees of PVS were rated in the slice containing the maximum number of PVS in basal ganglia (BG) on T1 images and were defined as follows: degree 1 (< 5 PVS); degree 2 (≥5 and≤10 PVS); degree 3 (> 10 PVS and still numerable); and degree 4 (innumerable number of PVS and cribriform change in BG) [29]. The periventricular and deep WMH were scored according to the Fazekas standard, ranged from 0 to 3 [30]. Lacunes were referred to rounded or oval lesions, in the centrum semiovale, basal ganglia, internal capsule, or brainstem, with diameters between 3 mm and 20 mm, of CSF signal intensity on FLAIR images and high signal intensity on the rim, with no high signal intensity on DWI. Cerebral microbleeds were referred to small (< 5 mm), homogeneous, round foci of hypointense on paramagnetic-sensitive MR sequences like T2 star-weighted gradient-recalled echo images in cortico-subcortical junction, BG, brainstem, cerebellum, and deep grey matter or white matter (WM) [9]. To evaluate inter-observer consistency, another neuroradiologist (RZ) performed evaluation on randomly selected 50 cases. Weighted kappa was used to assess consistency between the two observers.

Total CSVD score

The total CSVD score for each participant was calculated by counting the score of four CSVD features as follows: 1 for PVS severity degrees in the basal ganglia (grade 2–4), 1 for WMHs severity degrees (periventricular WMH: Fazekas score 3 and/or deep WMH: Fazekas score 2 or 3), 1 for any existence of lacunes (≥1 lesion), and 1 for any existence of microbleeds (≥1 lesion) [9]. The total score was varied from 0 to 4.

DTI processing

The preprocessing of diffusion-weighted images was performed using MRtrix3 (https://www.mrtrix.org/), and the steps included denoising, removing Gibbs artifact, and eddy current correction. The preprocessed data were analyzed using scripts (https://github.com/mvgolub/FW-DTI-Beltrami) based on DIPY [31]. Briefly, in each voxel, the signal was fitted to a two-compartment model, including a FW compartment (isotropic tensor) and a tissue compartment (FW-corrected tensor). The estimated parameters were the fractional volume of the FW compartment (i.e., the FW measure) and the tensor of the tissue compartment. The FW measure expresses the relative contribution of FW in each voxel, ranging from 0 to 1. The tensor of the tissue compartment reflects the tissue microstructure after removing the signal contributed by FW. The generated FW maps were coregistered to structural images through b0 images using elastic registration, and mean WM FW was extracted using each subject’s shrunken WM mask (derived by brain segmentation, eroded by 2 voxels) to avoid signal contamination from other brain tissues. Visual inspection of the raw data, reconstructed maps, and co-registered images were performed to ensure data quality.

Statistical analyses

Demographic and clinical characteristics were summarized using descriptive statistical analyses (SPSS, v.26.0). Annual decline rates were also calculated ((last follow-up test score –baseline test score) / interval time). Partial correlation between the total CSVD score and FW were investigated, taking age, sex, and education as covariates. Association between the two imaging metrics and AD pathologies were also explored using the same covariates.

Firstly, we examined associations between the two imaging metrics and baseline cognitive performance using simple linear regression models. Each of the baseline cognitive scores were entered as dependent variables, while age, sex, and education were entered as independent variables (Model 1). Then we added FW (Model 2) or total CSVD score (Model 3) into the initial model. Next, multiple linear regression models with variable selection were constructed to evaluate the relative importance of the two imaging metrics on baseline cognitive performance. FW and total CSVD score (Model 4) were both added to the initial model and variable selection was performed using the step-wise method. Bonferroni correction for multiple comparisons was applied (p < 0.05/8). Explained variances (adjusted R²), standardized betas (std β) and p-values of each model were calculated.

To test whether imaging metrics could predict longitudinal cognitive decline, linear mixed effect (LME) model analyses (R software, v.4.0.0; lme4 package, v1.1-27) were conducted with the following settings:

Test Score_ij =β0 +β1^*Time_ij +β2^*Age_i +β3^*Sex_i+β4^*Education_i +β5^*FW_i + b0_i + b1_i^*Time_ij +ɛ_ij (Model 5)

Test Score_ij =β0 +β1^*Time_ij +β2^*Age_i +β3^*Sex_i+β4^*Education_i +β5^*Total CSVD score_i + b0_i +b1_i^*Time_ij +ɛ_ij (Model 6)

Test Score_ij represented the neuropsychological test score of participant i at time j, and time was the interval from the baseline imaging time to each follow-up time point (in years, baseline = 0). Baseline age, sex, and education were added as covariates. β represented estimates for the fixed effects, and b represented estimates for the subject specific random effects with individual intercept and slope of time modeled in.

Additionally, we repeated all the previous statistical analyses (Model 7-Model 12), adding apoli-poprotein E (APOE) ɛ4 status, Aβ, and p-tau to each previous model. All the results with 2-tailed p < 0.05 were defined as statistically significant.

RESULTS

Demographic and clinical characteristics

Demographic and clinical characteristics are shown in Table 1. There were 50 females (39.7%). The mean age was 73.3, ranging from 55.2 to 93.6. The mean education years was 15.9, ranging from 11 to 20. The mean FW value was 0.202, ranging from 0.101 to 0.404. The median CSVD score was 1. The weighted kappa of CSVD score demonstrated good inter-observer consistency (weighted kappa = 0.694). There was a moderate correlation between total CSVD score and FW (r = 0.373, p < 0.001). None of the imaging metrics were associated with AD pathologies (Supplementary Table 1).

Table 1

Demographic and Clinical Characteristics

	Sample number	Mean±SD or %
Sociodemographic Parameters
Male/Female	126	76/50
Age	126	73.3±7.6
Education	126	15.9±2.7
Pathology Parameters
APOE ɛ4 (Non-Carriers/Carriers)	116	52/64
Aβ	91	170.3±51.1
P-tau	91	44.9±28.8
Imaging Metrics
FW	126	0.202±0.047
Total CSVD Score (0/1/2/3/4)	126	0/72/39/12/3
Global Cognitive Function
MMSE Baseline	126	27.7±1.7
MMSE Annual Decline	106	–0.025±0.049
MoCA Baseline	119	22.9±3.0
MoCA Annual Decline	106	–0.038±0.068
Executive Function
TMT-A Baseline	119	41.8±18.4
TMT-A Annual Decline	106	0.089±0.252
TMT-B Baseline	119	113.7±62.3
TMT-B Annual Decline	104	0.116±0.237
Memory Function
RAVLT-Delayed Recall Baseline	119	3.9±3.6
RAVLT-Delayed Recall Annual Decline	79	–0.160±0.276
Logical Memory-Delayed Recall Baseline	126	7.3±3.9
Logical Memory-Delayed Recall Annual Decline	96	–0.086±0.229
Language Function
Category Fluency-Animals Baseline	119	16.6±5.2
Category Fluency-Animals Annual Decline	106	–0.025±0.125
BNT Baseline	118	26.5±3.2
BNT Annual Decline	105	–0.011±0.050

APOE, apolipoprotein E; Aβ, amyloid-β; p-tau, phosphorylated tau; FW, free water; CSVD, cerebral small vessel disease; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; TMT-A, Trail Making Test A; TMT-B, Trail Making Test B; RAVLT, Rey Auditory Verbal Learning Test; BNT, Boston Naming Test.

Relationship between imaging metrics and baseline cognitive performance

In simple linear regression analysis, FW was linked to baseline TMT-A (std β=0.344, p = 0.001), baseline TMT-B (std β= 0.230, p = 0.028) and baseline Category Fluency-Animals (std β= –0.241, p = 0.019), while the effects of FW predicting the other five baseline cognitive scores and total CSVD score predicting all eight baseline cognitive scores did not meet significant statistical differences (Table 2). In multiple linear regression analyses, total CSVD score was not associated with any of the cognitive domains, but associations between FW and TMT-A, TMT-B and Category Fluency-Animals were still significant (Table 3).

Table 2

Simple linear regression analyses for baseline cognitive performance (Covariates: age, sex and education)

Neuropsychological Tests		Model 1: age + sex + education	Model 2: Model 1 + FW		Model 3: Model 1 + Total CSVD Score
		R²	R²	Std β (p)	R²	Std β (p)
Global Cognitive Function	MMSE	0.047	0.055	–0.147 (0.154)	0.041	0.045 (0.633)
	MoCA	0.149	0.146	–0.077 (0.446)	0.158	0.136 (0.134)
Executive Function	TMT-A	0.058	0.138	0.344 (0.001) ^**	0.055	0.075 (0.431)
	TMT-B	0.061	0.092	0.230 (0.028) ^*	0.066	0.123 (0.196)
Memory Function	RAVLT-Delayed Recall	–0.015	–0.021	–0.063 (0.564)	–0.022	0.036 (0.718)
	Logical Memory-Delayed Recall	0.041	0.041	0.104 (0.316)	0.033	–0.005 (0.960)
Language Function	Category Fluency-Animals	0.091	0.126	–0.241 (0.019)^*	0.085	0.045 (0.632)
	BNT	0.043	0.044	–0.115 (0.283)	0.035	–0.013 (0.894)

These linear regression analyses are served to identify FW or total CSVD score are associated with baseline cognitive performance. First of all, age, sex and education were entered as independent variables into simple linear regression models for each neuropsychological assessment (Model 1). Next, FW (Model 2) and total CSVD score (Model 3) were independently added to the initial model. Explained variances (R²), standardized betas (std β) and p-values of each model were calculated. Bonferroni correction for multiple comparisons was applied, statistically significant results before and after Bonferroni correction for 8 comparisons (p < 0.05/8) were shown in bold and labeled ^* and ^** separately. FW, free water; CSVD, cerebral small vessel disease; R², adjusted R²; std β, standardized betas; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; TMT-A, Trail Making Test A; TMT-B, Trail Making Test B; RAVLT, Rey Auditory Verbal Learning Test; BNT, Boston Naming Test.

Table 3

Multiple linear regression analyses for baseline cognitive performance (Covariates: age, sex, and education)

Model 4: Model 1 + FW + Total CSVD Score	Global Cognitive Function		Executive Function		Memory Function		Language Function
	MMSE	MoCA	TMT-A	TMT-B	RAVLT-Delayed Recall	Logical Memory-Delayed Recall	Category Fluency-Animals	BNT
	R² = 0.047	R² = 0.149	R² = 0.138	R² = 0.092	R² = –0.015	R² = 0.041	R² = 0.126	R² = 0.043
Age	–0.238 (0.008)^*	–0.261 (0.003)^**	0.100 (0.319)	0.177 (0.087)	–0.093 (0.322)	–0.038 (0.670)	0.013 (0.899)	–0.173 (0.060)
Sex	–0.015 (0.873)	–0.173 (0.052)	0.120 (0.179)	0.056 (0.538)	–0.044 (0.647)	0.026 (0.778)	0.017 (0.849)	0.032 (0.737)
Education	0.123 (0.177)	0.331 (< 0.001) ^**	–0.134 (0.138)	–0.053 (0.568)	0.051 (0.594)	0.243 (0.008) ^*	0.354 (< 0.001) ^**	0.178 (0.059)
FW	/	/	0.344 (0.001) ^**	0.230 (0.028) ^*	/	/	–0.241 (0.019)^*	/
Total CSVD Score	/	/	/	/	/	/	/	/

Values show standardized regression coefficients: standardized β (p-value) for predictor variables in regression models, controlling for effects of age, sex, and education. Bonferroni correction for multiple comparisons was applied, statistically significant results before and after Bonferroni correction for 8 comparisons (p < 0.05/8) were shown in bold and labeled ^* and ^** separately. R² refers to adjusted R² (explained variances). FW, free water; CSVD, cerebral small vessel disease; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; TMT-A, Trail Making Test A; TMT-B, Trail Making Test B; RAVLT, Rey Auditory Verbal Learning Test; BNT, Boston Naming Test.

Taking age, sex, education, APOE ɛ4, Aβ, and p-tau as covariates, we detected fewer results in simple and multiple linear regression models. In simple linear regression models, FW was associated with baseline TMT-A (std β= 0.334, p = 0.003) and baseline TMT-B (std β= 0.221, p = 0.044) status, while the effects of FW was not associated with the other six baseline cognitive scores and total CSVD score was not associated with all eight baseline cognitive scores (Supplementary Table 2). The total CSVD score did not survive in all multiple regression analyses, but FW remained significantly associated with executive function (Supplementary Table 3).

Relationship between imaging metrics and longitudinal cognitive changes

FW could predict the prolongation of TMT-A (std β= 0.255, p < 0.001) and TMT-B (std β= 0.202, p = 0.024) completion time, and the decline of Category Fluency-Animals numbers (std β= –0.199, p = 0.016, Table 4). The total CSVD score did not survive in all LME analyses (Table 5).

Table 4

Linear mixed effect model analyses for baseline FW predicting follow-up cognitive declines (Covariates: age, sex, and education)

Model 5	Global Cognitive Function		Executive Function		Memory Function		Language Function
	MMSE	MoCA	TMT-A	TMT-B	RAVLT-Delayed Recall	Logical Memory-Delayed Recall	Category Fluency-Animals	BNT
Time	–0.430 (< 0.001) ^**	–0.376 (< 0.001) ^**	0.268 (< 0.001) ^**	0.218 (< 0.001) ^**	–0.138 (0.001) ^**	–0.119 (0.002) ^**	–0.187 (< 0.001) ^**	–0.121 (< 0.001) ^**
Age	–0.125 (0.016) ^*	–0.134 (0.033) ^*	0.071 (0.327)	0.217 (0.019) ^*	–0.133 (0.144)	–0.028 (0.724)	–0.011 (0.893)	–0.092 (0.250)
Sex	–0.029 (0.527)	–0.075 (0.183)	0.092 (0.166)	0.083 (0.321)	–0.117 (0.157)	–0.021 (0.777)	–0.001 (0.989)	0.006 (0.934)
Education	0.167 (< 0.001) ^**	0.220 (< 0.001) ^**	–0.110 (0.101)	–0.107 (0.196)	0.140 (0.092)	0.188 (0.012) ^*	0.284 (< 0.001) ^**	0.181 (0.014) ^*
FW	–0.026 (0.602)	–0.054 (0.366)	0.255 (< 0.001) ^**	0.202 (0.024) ^*	–0.080 (0.366)	0.021 (0.794)	–0.199 (0.016) ^*	–0.066 (0.398)

Values show standardized regression coefficients: standardized β (p-value) for predictor variables in regression models, controlling for effects of age, sex, and education. Bonferroni correction for multiple comparisons was applied, statistically significant results before and after Bonferroni correction for 8 comparisons (p < 0.05/8) were shown in bold and labeled ^* and ^** separately. FW, free water; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; TMT-A, Trail Making Test A; TMT-B, Trail Making Test B; RAVLT, Rey Auditory Verbal Learning Test; BNT, Boston Naming Test.

Table 5

Linear mixed effect model analyses for baseline total CSVD score predicting follow-up cognitive declines (Covariates: age, sex and education)

Model 6	Global Cognitive Function		Executive Function		Memory Function		Language Function
	MMSE	MoCA	TMT-A	TMT-B	RAVLT-Delayed Recall	Logical Memory-Delayed Recall	Category Fluency-Animals	BNT
Time	–0.430 (< 0.001) ^**	–0.376 (< 0.001) ^**	0.274 (< 0.001) ^**	0.215 (< 0.001) ^**	–0.136 (0.001) ^**	–0.119 (0.002) ^**	–0.184 (< 0.001) ^**	–0.121 (< 0.001) ^**
Age	–0.138 (0.004) ^**	–0.138 (0.016) ^*	0.182 (0.011) ^*	0.277 (0.001) ^**	–0.171 (0.042) ^*	–0.017 (0.816)	–0.117 (0.142)	–0.120 (0.104)
Sex	–0.029 (0.537)	–0.062 (0.267)	0.080 (0.262)	0.059 (0.483)	–0.114 (0.174)	–0.020 (0.786)	0.002 (0.976)	0.009 (0.901)
Education	0.163 (< 0.001) ^**	0.217 (< 0.001) ^**	–0.068 (0.327)	–0.089 (0.277)	0.126 (0.125)	0.192 (0.009) ^*	0.248 (0.002) ^**	0.172 (0.019) ^*
Total CSVD Score	< 0.001 (0.995)	–0.078 (0.172)	0.062 (0.390)	0.160 (0.063)	–0.007 (0.929)	–0.003 (0.964)	–0.011 (0.894)	–0.014 (0.852)

Values show standardized regression coefficients: standardized β (p-value) for predictor variables in regression models, controlling for effects of age, sex, and education. Bonferroni correction for multiple comparisons was applied, statistically significant results before and after Bonferroni correction for 8 comparisons (p < 0.05/8) were shown in bold and labeled ^* and ^** separately. CSVD, cerebral small vessel disease; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; TMT-A, Trail Making Test A; TMT-B, Trail Making Test B, RAVLT, Rey Auditory Verbal Learning Test; BNT, Boston Naming Test.

Taking age, sex, education, APOE ɛ4, Aβ, and p-tau as covariates, results were similar. FW could predict the prolongation of TMT-A (std β= 0.255, p < 0.001) and TMT-B (std β= 0.202, p = 0.024) completion time, and the decline of Category Fluency-Animals (std β= –0.199, p = 0.016) numbers (Supplementary Table 4). The total CSVD score did not survive in all LME analyses (Supplementary Table 5).

DISCUSSION

The findings in the present study supported our hypothesis that FW is a more sensitive marker than total CSVD score in predicting longitudinal cognitive declines in MCI subjects. To be specific, FW was associated with baseline and follow-up executive and language functions when taking age, sex, and education as covariates. After APOE ɛ4, Aβ, and p-tau were introduced as covariates, FW was still independently associated with executive performance. This study underscores the potential of FW in future clinical applications.

Our major finding is that higher FW was associated with poorer cognitive performance and faster cognitive decline among MCI subjects, especially in the executive and language domains. Cognitive impairments in these two domains are common in AD patients [32], and may appear at very early stages [33]. Interestingly, recent studies pointed out that these impairments might be attributed to vascular damages [34, 35]. In a cross-sectional investigation of 115 subjects including AD and vascular dementia, there was a strong correlation between WM FW and executive/language function [34]. Another two studies also showed that FW within the whole WM or some specific fiber tracts were associated with executive function [19, 36]. Additionally, WM FW was able to reflect longitudinal changes of executive function [19], which is important in the progression from MCI to AD [37]. Regarded as one of the core features of AD, the impairment of language capability has a close relationship with AD severity [38]. Several studies suggested that the whole WM and left hemispheric WM FW were related to language deficits [26, 34]. Moreover, it was observed in a longitudinal study that verbal fluency performance was associated with FW within the cingulum [39]. Our findings basically coincide well with previous studies and further consolidate the link between FW and certain cognitive domains.

Here FW rather than total CSVD score was found related to cognitive performance and cognitive decline. Indeed, water diffusion is sensitive to pathological alterations. Diffusion changes were found more strongly correlated to cognitive impairment than traditional MRI results [40 –42]. Compared with other imaging metrics, FW had the strongest association with clinical deficits [20] and was the strongest predictor of cognitive decline [19].

While it is easy to imagine that microstructural changes (FW) may be more sensitive than macro level imaging signs (CSVD score), we would like to point out that they also had different weightings for vascular-related brain changes. To better reflect total vascular burden, several kinds of total CSVD scores have been proposed in recent years [43, 44], with different weightings assigned to different CSVD markers. Interestingly, as most of the CSVD markers contribute to WM FW, it can be treated as another way for weighting vascular damages. In our previous study, WM FW was found associated with dPVS, WMH and lacunes, but not CMBs [45]. Since lacunes were relatively rare in MCI subjects, here FW might be mainly associated with other vascular markers, including the observable dPVS and WMH, and unobservable changes such as blood barrier disruption. From this point of view, because CMBs were closely associated with AD pathologies and lacunes were rare, the CSVD score cannot provide additional pathological information compared to FW. On the other hand, FW can detect abnormalities in a wider range (less ceiling or floor effects) and may incorporate other vascular pathology information. Together, these reasons may give rise to the superiority of FW.

When AD pathological factors were added, the associations between baseline FW and executive and language functions were still significant in longitudinal analyses. Additionally, among all the models in our study, AD pathologies were mainly related to baseline and follow-up memory function, rarely in execution and language domains. Although both vascular and AD pathologies could cause impairments in most cognitive domains [46 –48], the degree of their influences might be different. Here our sample size is not quite large, therefore some associations with small effect sizes were not revealed, showing a relatively disassociation pattern. Moreover, none of the imaging metrics were related to AD pathologies, further confirming the independent prediction ability of vascular-related metrics.

Our study has some limitations. First, due to the essential of the inclusion criteria, our data sample is relatively small, the absence of follow-up data is also common, which may result in a slight selection deviation within the original sample. Second, as longitudinal CSF and blood pathological markers including Aβ and p-tau are relatively scarce (most subjects had only baseline and 2-year follow-up data), it is difficult to investigate the effect of vascular factors on the progression of AD pathological markers.

FW rather than total CSVD score can predict the progress of cognitive decline in MCI subjects. FW may become a promising imaging biomarker for monitoring and predicting disease progression.

Footnotes

ACKNOWLEDGMENTS

Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.

This study was supported by the National Natural Science Foundation of China (Grant Nos. 81771820, 82101987, 81901707 and 82001766), the Health Foundation for Creative Talents in Zhejiang Province, China (No: 2016) and the Zhejiang province Postdoctoral Science Foundation.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Lane

, Hardy

, Schott

(2018) Alzheimer’s disease. Eur J Neurol 25, 59–70.

Montagne

, Nation

, Pa

, Sweeney

, Toga

, Zlokovic

(2016) Brain imaging of neurovascular dysfunction in Alzheimer’s disease. Acta Neuropathol 131, 687–707.

Zlokovic

(2011) Neurovascular pathways to neurodegeneration in Alzheimer’s disease and other disorders. Nat Rev Neurosci 12, 723–738.

Hachinski

, Einhaupl

, Ganten

, Alladi

, Brayne

, Stephan

BCM

, Sweeney

, Zlokovic

, Iturria-Medina

, Iadecola

, Nishimura

, Schaffer

, Whitehead

, Black

, Ostergaard

, Wardlaw

, Greenberg

, Friberg

, Norrving

, Rowe

, Joanette

, Hacke

, Kuller

, Dichgans

, Endres

, Khachaturian

(2019) Preventing dementia by preventing stroke: The Berlin Manifesto. Alzheimers Dement 15, 961–984.

Blair

, Appleton

, Law

, Doubal

, Flaherty

, Dooley

, Shuler

, Richardson

, Hamilton

, Shi

, Stringer

, Boyd

, Thrippleton

, Sprigg

, Bath

, Wardlaw

(2018) Preventing cognitive decline and dementia from cerebral small vessel disease: The LACI-1 Trial. Protocol and statistical analysis plan of a phase IIa dose escalation trial testing tolerability, safety and effect on intermediary endpoints of isosorbide mononitrate and cilostazol, separately and in combination. Int J Stroke 13, 530–538.

Wardlaw

, Smith

, Biessels

, Cordonnier

, Fazekas

, Frayne

, Lindley

, O’Brien

, Barkhof

, Benavente

, Black

, Brayne

, Breteler

, Chabriat

, Decarli

, de Leeuw

, Doubal

, Duering

, Fox

, Greenberg

, Hachinski

, Kilimann

, Mok

, Oostenbrugge

, Pantoni

, Speck

, Stephan

, Teipel

, Viswanathan

, Werring

, Chen

, Smith

, van Buchem

, Norrving

, Gorelick

, Dichgans

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

Stefaniak

, Su

, Mak

, Sheikh-Bahaei

, Wells

, Ritchie

, Waldman

, Ritchie

, O’Brien

(2018) Cerebral small vessel disease in middle age and genetic predisposition to late-onset Alzheimer’s disease. Alzheimers Dement 14, 253–258.

Liu

, Braidy

, Poljak

, Chan

DKY

, Sachdev

(2018) Cerebral small vessel disease and the risk of Alzheimer’s disease: A systematic review. Ageing Res Rev 47, 41–48.

Staals

, Makin

, Doubal

, Dennis

, Wardlaw

(2014) Stroke subtype, vascular risk factors, and total MRI brain small-vessel disease burden. Neurology 83, 1228–1234.

10.

Klarenbeek

, van Oostenbrugge

, Rouhl

, Knottnerus

, Staals

(2013) Ambulatory blood pressure in patients with lacunar stroke: Association with total MRI burden of cerebral small vessel disease. Stroke 44, 2995–2999.

11.

Del Brutto

, Ortiz

, Del Brutto

, Mera

, Zambrano

, Biller

(2018) Total cerebral small vessel disease score and cognitive performance in community-dwelling older adults. Results from the Atahualpa Project. Int J Geriatr Psychiatry 33, 325–331.

12.

Staals

, Booth

, Morris

, Bastin

, Gow

, Corley

, Redmond

, Starr

, Deary

, Wardlaw

(2015) Total MRI load of cerebral small vessel disease and cognitive ability in older people. Neurobiol Aging 36, 2806–2811.

13.

Jiang

, Wang

, Yuan

, Xu

, Zhang

, Zhu

, Li

, Suo

, Tian

, Fan

, Jin

, Ye

, Dong

, Cui

, Chen

(2019) Total cerebral small vessel disease burden is related to worse performance on the Mini-Mental State Examination and incident dementia: A prospective 5-year follow-up. J Alzheimers Dis 69, 253–262.

14.

Huijts

, Duits

, van Oostenbrugge

, Kroon

, de Leeuw

, Staals

(2013) Accumulation of MRI markers of cerebral small vessel disease is associated with decreased cognitive function. A study in first-ever lacunar stroke and hypertensive patients. Front Aging Neurosci 5, 72.

15.

Uiterwijk

, van Oostenbrugge

, Huijts

, De Leeuw

, Kroon

, Staals

(2016) Total cerebral small vessel disease MRI score is associated with cognitive decline in executive function in patients with hypertension. Front Aging Neurosci 8, 301.

16.

, Yuan

, Qin

, Yang

, Li

, Hu

(2021) Higher total cerebral small vessel disease burden was associated with mild cognitive impairment and overall cognitive dysfunction: A propensity score-matched case-control study. Front Aging Neurosci 13, 695732.

17.

Fan

, Shen

, Huo

, Gao

, Li

, Zheng

, Zhang

(2021) Total cerebral small vessel disease burden on MRI correlates with medial temporal lobe atrophy and cognitive performance in patients of a memory clinic. Front Aging Neurosci 13, 698035.

18.

Douaud

, Menke

, Gass

, Monsch

, Rao

, Whitcher

, Zamboni

, Matthews

, Sollberger

, Smith

(2013) Brain microstructure reveals early abnormalities more than two years prior to clinical progression from mild cognitive impairment to Alzheimer’s disease. J Neurosci 33, 2147–2155.

19.

Maillard

, Fletcher

, Singh

, Martinez

, Johnson

, Olichney

, Farias

, DeCarli

(2019) Cerebral white matter free water: A sensitive biomarker of cognition and function. Neurology 92, e2221–e2231.

20.

Duering

, Finsterwalder

, Baykara

, Tuladhar

, Gesierich

, Konieczny

, Malik

, Franzmeier

, Ewers

, Jouvent

, Biessels

, Schmidt

, de Leeuw

, Pasternak

, Dichgans

(2018) Free water determines diffusion alterations and clinical status in cerebral small vessel disease. Alzheimers Dement 14, 764–774.

21.

Finsterwalder

, Vlegels

, Gesierich

, Araque Caballero

, Weaver

, Franzmeier

, Georgakis

, Konieczny

, Koek

, Dominantly Inherited Alzheimer Network, Karch

, Graff-Radford

, Salloway

, Oh

, Allegri

, Chhatwal

, DELCODE study group, Jessen

, Duzel

, Dobisch

, Metzger

, Peters

, Incesoy

, Priller

, Spruth

, Schneider

, Fliessbach

, Buerger

, Janowitz

, Teipel

, Kilimann

, Laske

, Buchmann

, Heneka

, Brosseron

, Spottke

, Roy

, Ertl-Wagner

, Scheffler

, Alzheimer’s Disease Neuroimaging Initiative, Utrecht VCI study group, Seo

, Kim

, Na

, Kim

, Jang

, Ewers

, Levin

, Schmidt

, Pasternak

, Dichgans

, Biessels

, Duering

(2020) Small vessel disease more than Alzheimer’s disease determines diffusion MRI alterations in memory clinic patients. Alzheimers Dement 16, 1504–1514.

22.

Archer

, Coombes

, McFarland

, DeKosky

, Vaillancourt

(2019) Development of a transcallosal tractography template and its application to dementia. Neuroimage 200, 302–312.

23.

, Pasternak

, Ng

, Chong

JSX

, Liu

, Zhang

, Shim

, Loke

, Tan

, Venketasubramanian

, Chen

, Zhou

(2019) White matter microstructural abnormalities and default network degeneration are associated with early memory deficit in Alzheimer’s disease continuum. Sci Rep 9, 4749.

24.

Ofori

, DeKosky

, Febo

, Colon-Perez

, Chakrabarty

, Duara

, Adjouadi

, Golde

, Vaillancourt

, Alzheimer’s Disease Neuroimaging Initiative (2019) Free-water imaging of the hippocampus is a sensitive marker of Alzheimer’s disease. Neuroimage Clin 24, 101985.

25.

Reas

, Hagler

Jr , White

, Kuperman

, Bartsch

, Wierenga

, Galasko

, Brewer

, Dale

, McEvoy

(2018) Microstructural brain changes track cognitive decline in mild cognitive impairment. Neuroimage Clin 20, 883–891.

26.

Reas

, Hagler

, Kuperman

, Wierenga

, Galasko

, White

, Dale

, Banks

, McEvoy

, Brewer

(2020) Associations between microstructure, amyloid, and cognition in amnestic mild cognitive impairment and dementia. J Alzheimers Dis 73, 347–357.

27.

Shaw

, Vanderstichele

, Knapik-Czajka

, Clark

, Aisen

, Petersen

, Blennow

, Soares

, Simon

, Lewczuk

, Dean

, Siemers

, Potter

, Lee

, Trojanowski

, Alzheimer’s Disease Neuroimaging Initiative (2009) Cerebrospinal fluid biomarker signature in Alzheimer’s disease neuroimaging initiative subjects. Ann Neurol 65, 403–413.

28.

Shaw

, Vanderstichele

, Knapik-Czajka

, Figurski

, Coart

, Blennow

, Soares

, Simon

, Lewczuk

, Dean

, Siemers

, Potter

, Lee

, Trojanowski

, Alzheimer’s Disease Neuroimaging Initiative (2011) Qualification of the analytical and clinical performance of CSF biomarker analyses in ADNI. Acta Neuropathol 121, 597–609.

29.

Zhu

, Tzourio

, Soumare

, Mazoyer

, Dufouil

, Chabriat

(2010) Severity of dilated Virchow-Robin spaces is associated with age, blood pressure, and MRI markers of small vessel disease: A population-based study. Stroke 41, 2483–2490.

30.

Fazekas

, Chawluk

, Alavi

, Hurtig

, Zimmerman

(1987) MR signal abnormalities at 1.5 T in Alzheimer’s dementia and normal aging. AJR Am J Roentgenol 149, 351–356.

31.

Golub

, Neto Henriques

, Gouveia Nunes

(2021) Free-water DTI estimates from single b-value data might seem plausible but must be interpreted with care. Magn Reson Med 85, 2537–2551.

32.

Teitsdottir

, Jonsdottir

, Lund

, Darreh-Shori

, Snaedal

, Petersen

(2020) Association of glial and neuronal degeneration markers with Alzheimer’s disease cerebrospinal fluid profile and cognitive functions. Alzheimers Res Ther 12, 92.

33.

Valech

, Tort-Merino

, Coll-Padros

, Olives

, Leon

, Rami

, Molinuevo

(2018) Executive and language subjective cognitive decline complaints discriminate preclinical Alzheimer’s disease from normal aging. J Alzheimers Dis 61, 689–703.

34.

, Pasternak

, Liu

, Loke

, Choo

, Hilal

, Xu

, Ikram

, Venketasubramanian

, Chen

, Zhou

(2017) Distinct white matter microstructural abnormalities and extracellular water increases relate to cognitive impairment in Alzheimer’s disease with and without cerebrovascular disease. Alzheimers Res Ther 9, 63.

35.

Camerino

, Sierpowska

, Reid

, Meyer

, Tuladhar

, Kessels

RPC

, de Leeuw

, Piai

(2021) White matter hyperintensities at critical crossroads for executive function and verbal abilities in small vessel disease. Hum Brain Mapp 42, 993–1002.

36.

Archer

, Moore

, Shashikumar

, Dumitrescu

, Pechman

, Landman

, Gifford

, Jefferson

, Hohman

(2020) Free-water metrics in medial temporal lobe white matter tract projections relate to longitudinal cognitive decline. Neurobiol Aging 94, 15–23.

37.

Espinosa

, Alegret

, Boada

, Vinyes

, Valero

, Martinez-Lage

, Pena-Casanova

, Becker

, Wilson

, Tarraga

(2009) Ecological assessment of executive functions in mild cognitive impairment and mild Alzheimer’s disease. J Int Neuropsychol Soc 15, 751–757.

38.

Yeung

, Iaboni

, Rochon

, Lavoie

, Santiago

, Yancheva

, Novikova

, Xu

, Robin

, Kaufman

, Mostafa

(2021) Correlating natural language processing and automated speech analysis with clinician assessment to quantify speech-language changes in mild cognitive impairment and Alzheimer’s dementia. Alzheimers Res Ther 13, 109.

39.

Edde

, Theaud

, Rheault

, Dilharreguy

, Helmer

, Dartigues

, Amieva

, Allard

, Descoteaux

, Catheline

(2020) Free water: A marker of age-related modifications of the cingulum white matter and its association with cognitive decline. PLoS One 15, e0242696.

40.

, Zhou

, Li

, Cao

, Lin

, Pan

, Chen

(2010) Diffusion tensor imaging changes correlate with cognition better than conventional MRI findings in patients with subcortical ischemic vascular disease. Dement Geriatr Cogn Disord 30, 317–326.

41.

Hoy

, Ly

, Carlsson

, Okonkwo

, Zetterberg

, Blennow

, Sager

, Asthana

, Johnson

, Alexander

, Bendlin

(2017) Microstructural white matter alterations in preclinical Alzheimer’s disease detected using free water elimination diffusion tensor imaging. PLoS One 12, e0173982.

42.

Jokinen

, Schmidt

, Ropele

, Fazekas

, Gouw

, Barkhof

, Scheltens

, Madureira

, Verdelho

, Ferro

, Wallin

, Poggesi

, Inzitari

, Pantoni

, Erkinjuntti

, LADIS Study Group (2013) Diffusion changes predict cognitive and functional outcome: The LADIS study. Ann Neurol 73, 576–583.

43.

Jokinen

, Koikkalainen

, Laakso

, Melkas

, Nieminen

, Brander

, Korvenoja

, Rueckert

, Barkhof

, Scheltens

(2020) Global burden of small vessel disease–related brain changes on MRI predicts cognitive and functional decline. Stroke 51, 170–178.

44.

Al Olama

, Wason

, Tuladhar

, van Leijsen

, Koini

, Hofer

, Morris

, Schmidt

, de Leeuw

F-E

, Markus

(2020) Simple MRI score aids prediction of dementia in cerebral small vessel disease. Neurology 94, e1294–e1302.

45.

Huang

, Zhang

, Jiaerken

, Wang

, Hong

, Yu

, Lian

, Li

, Zeng

, Luo

, Yu

, Wu

, Xu

, Zhang

(2022) White matter free water is a composite marker of cerebral small vessel degeneration. Transl Stroke Res 13, 56–64.

46.

Hamilton

OKL

, Backhouse

, Janssen

, Jochems

ACC

, Maher

, Ritakari

, Stevenson

, Xia

, Deary

, Wardlaw

(2021) Cognitive impairment in sporadic cerebral small vessel disease: A systematic review and meta-analysis. Alzheimers Dement 17, 665–685.

47.

Marchant

, Reed

, Sanossian

, Madison

, Kriger

, Dhada

, Mack

, DeCarli

, Weiner

, Mungas

, Chui

, Jagust

(2013) The aging brain and cognition: Contribution of vascular injury and abeta to mild cognitive dysfunction. JAMA Neurol 70, 488–495.

48.

Insel

, Donohue

, Sperling

, Hansson

, Mattsson-Carlgren

(2020) The A4 study: Beta-amyloid and cognition in 4432 cognitively unimpaired adults. Ann Clin Transl Neurol 7, 776–785.