Choroid Plexus Volume in Rural Chinese Older Adults: Distribution and Association With Cardiovascular Risk Factors and Cerebral Small Vessel Disease

Abstract

Background

The choroid plexus (CP) is involved in neurodegenerative diseases. However, the association of CP with cardiovascular risk factors and cerebral small vessel disease in older adults remains unclear.

Methods and Results

This population‐based study included 1263 participants (60 years and older) from the MIND‐China (Multimodal Interventions to Delay Dementia and Disability in Rural China) substudy (2018–2020), of which 111 individuals completed diffusion tensor imaging examination. CP volume was automatically segmented. White matter hyperintensities (WMHs), enlarged perivascular spaces (EPVS), cerebral microbleeds, and lacunes were assessed following the Standards for Reporting Vascular Changes on Neuroimaging 1. Peak width of skeletonized mean diffusivity and free water were derived from diffusion tensor imaging images. We used linear regression models to evaluate the association between CP volume and cardiovascular risk factors, WMH volumes, and diffusion tensor imaging metrics, and logistic regression models to examine the association between CP volume and EPVS, cerebral microbleeds, and lacunes. The CP volume increased with age (P<0.001). Men (β coefficient=0.47 [95% CI, 0.29–0.64]) and participants with diabetes (β coefficient=0.16 [95% CI, 0.01–0.31]) had larger CP volumes than women and individuals without diabetes, respectively (P<0.05). Greater CP volume was significantly associated with larger total and periventricular WMH volumes and moderate to severe EPVS in basal ganglia (P<0.05) but not with deep WMHs, EPVS in centrum semiovale, lacunes, or cerebral microbleeds. In the diffusion tensor imaging subsample, enlarged CP was significantly associated with higher peak width of skeletonized mean diffusivity and free water of periventricular and deep white matter (P<0.05).

Conclusions

An enlarged CP is associated with larger global and periventricular WMH volume and higher likelihoods of EPVS in basal ganglia and impaired white matter integrity, suggesting that an enlarged CP may represent a precursor of cerebral small vessel disease.

Nonstandard Abbreviations and Acronyms

BG

basal ganglia

CP

choroid plexus

CRF

cardiovascular risk factor

CSO

centrum semiovale

CSVD

cerebral small vessel disease

CMB

cerebral microbleed

DWMH

deep white matter hyperintensity

DTI

diffusion tensor imaging

EPVS

enlarged perivascular space

FW

free water

MIND‐China

Multimodal Interventions to Delay Dementia and Disability in Rural China

PSMD

peak width of skeletonized mean diffusivity

STRIVE‐1

Standards for Reporting Vascular Changes on Neuroimaging 1

WMH

white matter hyperintensity

WM

white matter

CLINICAL PERSPECTIVE.

What Is New?

• Older age, male sex, and diabetes are associated with larger choroid plexus volume.

• Enlarged choroid plexus volume is associated with increased global and periventricular white matter hyperintensity volumes, moderate to severe basal ganglia–enlarged perivascular space, and reduced white matter integrity.

• Enlarged choroid plexus may represent a precursor of cerebral small vessel disease in older adults.

What Are the Clinical Implications?

• Our findings imply that enlarged choroid plexus might represent a valuable precursor of cerebral small vessel disease.

The choroid plexus (CP) is a highly vascularized tissue located within the cerebral ventricular system. It consists of a monolayer of polarized cuboidal epithelial cells surrounding fenestrated capillaries and a connective stroma. ¹ The primary functions of CP are to secrete cerebrospinal fluid (CSF), form the blood‐CSF barrier, and mediate neuroinflammation. ² An enlarged CP generally indicates a deteriorative function of CP, such as decreased CSF production. ³ Small‐scale studies of healthy adults (sample size, n=≈100; age, 21–94 years) from the United States have suggested that enlarged CP volume was associated with older age, male sex, and higher adiposity indices. ⁴ , ⁵ , ⁶ In addition, a clinic‐based cross‐sectional study of patients with cognitive impairment from Korea showed that higher CP volume was related to hypertension, diabetes, and cardiovascular disease. ³ These studies suggest that a vascular pathway may be involved in CP enlargement in old age. However, large‐scale population‐based studies are imperative to further explore the association of CP volume with cardiovascular risk factors (CRFs) among ethnically, geographically, and socioeconomically diverse populations.

Cerebral small vessel disease (CSVD) refers to the age‐related pathological process of the arteries, arterioles, venules, and capillaries of the brain. ⁷ Conventional hallmarks of CSVD seen on structural magnetic resonance imaging (MRI) include white matter (WM) hyperintensities (WMHs), enlarged perivascular spaces (EPVS), cerebral microbleeds (CMBs), and lacunes. ⁸ In addition, the peak width of skeletonized mean diffusivity (PSMD) and free water (FW), which are 2 novel metrics derived from diffusion tensor imaging (DTI) that reflect microstructural damage of the WM, have been considered sensitive biomarkers of cerebral microvascular injury. ⁹ , ¹⁰ FW reflects the number of water molecules that are unrestricted within WM tissue, whereas PSMD represents the distribution of diffusivity along WM tracts. ¹⁰ Exploring the relationships of CP volume with conventional and DTI‐derived novel markers of CSVD may help understand the pathophysiological mechanisms underlying CSVD.

Previous studies have related glymphatic impairment with CSVD. ¹¹ , ¹² , ¹³ In addition, one of the primary functions of the CP is to produce CSF, which is crucial to maintain the functions of the glymphatic system. This suggests that CP is possibly involved in the pathophysiology of CSVD via glymphatic impairment. ¹⁴ Indeed, a clinic‐based study showed that CP enlargement was associated with WMH volume and its progression, in which impaired glymphatic function partially mediated this association. ¹⁵ However, large‐scale population‐based studies are warranted to investigate the association between CP volume and various CSVD markers.

Therefore, in this population‐based study of rural‐dwelling older adults in China, we sought to investigate the association of CP volume with CRFs and CSVD markers. We hypothesized that enlarged CP was associated with major CRFs and a higher CSVD burden, and that the association of enlarged CPs with CSVD burden might vary by demographic characteristics and major CRFs.

METHODS

Study Design and Participants

The data sets used and analyzed during the current study are available from the corresponding author upon reasonable request and approval by the MIND‐China (Multimodal Interventions to Delay Dementia and Disability in Rural China) Steering Committee.

This was a population‐based cross‐sectional study reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. The study participants were derived from the MIND‐China study, as previously described. ¹⁶ , ¹⁷ Briefly, the MIND‐China project included individuals who were 60 years and older by the end of 2017 and living in 52 rural communities (villages) of Yanlou Town, Yanggu County, western Shandong province, China. From March to September 2018, 5765 participants (74.9% of all eligible persons) were recruited for the MIND‐China trial. Of them, 1844 participants from 26 villages who randomly selected from all 52 villages were invited for the MRI substudy, and 1304 participants undertook structural brain MRI scans. The participants who underwent brain MRI scans were more likely to have hypertension compared with the remaining 540 who were selected but did not undergo the MRI procedure, but the 2 groups did not differ significantly in sociodemographics or other clinical conditions. Of those who undertwent brain MRI scans, 41 were excluded because of large lesions in the brain (n=23), missing data on CP volume (n=9), and other reasons (n=9), leaving 1263 participants for analyzing the association of CP volume with CRFs and conventional CSVD markers. Of these, data on DTI metrics were available in 111 individuals, a subsample used for analyzing the association of CP volume with PSMD and FW. The flowchart of the study participants is shown in Figure S1.

The MIND‐China project was approved by the ethics committee of Shandong Provincial Hospital in Jinan, Shandong, China. Written informed consent was obtained from all participants.

Data Collection and Assessment

Trained staff collected extensive data via face‐to‐face interviews, clinical examinations, and laboratory tests following a structured questionnaire. ¹⁶ , ¹⁷ The data included demographic factors (eg, age, sex, and education), cardiovascular risk factors (eg, smoking, alcohol consumption, hypertension, diabetes, hyperlipidemia, and obesity), medical history (eg, coronary heart disease and stroke), use of medications, and apolipoprotein E (APOE) genotype.

Educational level was categorized into no formal schooling, primary school, and middle school or above. Alcohol consumption and smoking status were categorized as never, former, and current drinking alcohol or smoking. ¹⁸ Body mass index, calculated as weight in kilograms divided by height in meters squared, was categorized into normal weight (<24 kg/m²), overweight (24–27.9 kg/m²), and obese (≥28 kg/m²). ¹⁸ Arterial blood pressure was measured on the right arm while the patient was in a sitting position after a 5‐minute rest using an electronic blood pressure monitor (Omron HEM‐7127J, Omron Corporation). ¹⁶ After an overnight fast, blood samples were taken in the morning. Fasting blood glucose and serum lipids were measured at the laboratory of Yanlou Town Hospital using an automatic biochemical analyzer (DIRUI CS‐600B, DIRUI Corporation). ¹⁶ Hypertension, diabetes, and dyslipidemia were defined as previously described. ¹⁸ APOE genotype was determined using multiple polymerase chain reaction and was dichotomized into carriers versus noncarriers of the ε4 allele.

MRI Acquisition Protocol

Structural brain MRI scans were acquired on either the Philips Ingenia 3.0T MR System in Southwestern Lu Hospital or the Philips Achieva 3.0T MR System in Liaocheng People's Hospital. The main MRI sequences included a sagittal 3‐dimensional sT1‐weighted, axial T2‐weighted, sagittal 3‐dimensional fluid‐attenuated inversion recovery images, and axial susceptibility weighted imaging. The detailed parameters of the core MRI sequences have been previously reported. ¹⁶ In addition, for participants who underwent structural brain MRI scans in Liaocheng People's Hospital, DTI images were also acquired using gradients applied in 16 directions with the following scan parameters: TR/TE=10 888/55 ms, voxel size=2×2×2 mm³, b‐value=0/1000 s/mm², total scan duration=331.7 s.

Assessment of CP Volume

Based on convolutional neural networks trained using the deep learning approach, we used AccuBrain (BrainNow Medical Technology Ltd.) to acquire the CP segmentation in the lateral ventricles. First, the CP was manually labeled on T1‐weighted images of 100 randomly selected participants at all slices where the CP was visible by a trained neurologist (S.L.) and verified by a radiologist (G.T.) using ITK‐SNAP software version 4.0.1 (www.itksnap.org). Then, these T1‐weighted images were used to train the neural networks, with the manually labeled CP serving as the reference standard. The network learns to recognize the unique features and patterns of the CP, allowing it to accurately segment this structure on new T1‐weighted images. Two trained neurologists (S.L. and L.C.) manually checked all of the automatically segmented CP volumes. We randomly selected an additional 20 participants to evaluate the accuracy of the automatically segmented CP volume against manual labels, which yielded an average Dice coefficient of 0.82 and a correlation coefficient of 0.93.

Evaluation of Conventional CSVD Markers

We assessed the conventional MRI markers for CSVD using structural brain MRI scans following the Standards for Reporting Vascular Changes on Neuroimaging 1 (STRIVE‐1). ⁸ CMBs were focal, rounded hypodense lesions <10 mm in diameter on susceptibility weighted imaging. ⁸ We quantitatively acquired CMBs by AccuBrain, as previously described. ¹⁹ Briefly, CMBs were identified on susceptibility weighted imaging using a fully connected network trained by the deep learning technique. For a given susceptibility weighted image, the network showed the CMBs locations by generating a probability map. Participants were stratified into presence or absence of CMBs based on the AccuBrain quantification. In addition, the visual assessments of lacunes and EPVS were performed by 2 well‐trained neurologists (J.W. for lacunes and M.Z. for EPVS) who were blinded to the clinical information, as previously described. ²⁰ , ²¹ In brief, lacunes were referred to rounded or ovoid CSF‐like lesions (3–15 mm in diameter), with a hyperintensity ring seen on fluid‐attenuated inversion recovery images. The trained rater (J.W.) counted lacunes in different anatomical brain regions in each hemisphere. Participants were stratified based on their visual count of lacunes. We defined EPVS as small (<3 mm in diameter) punctate (if perpendicular) or linear (if longitudinal to the plane of scan) hyperintensities on T2 images. The rater (M.Z.) first reviewed all MRI slices for EPVS in the areas of basal ganglia (BG) and centrum semiovale (CSO) and counted bilateral BG‐EPVS and CSO‐EPVS on all slices. Then, the EPVS count in BG and CSO was categorized according to the highest number of EPVS on the slice and the hemisphere, by following the scale: 0=none, 1=1 to 10 (mild), 2=11 to 20 (moderate), 3=21 to 40 (frequent), and 4=more than 40 EPVS (severe). ²² We further dichotomized BG‐EPVS and CSO‐EPVS into mild EPVS and moderate to severe EPVS, as previously reported. ²³ , ²⁴ We defined WMH as a hyperintense signal on T2 fluid‐attenuated inversion recovery images and acquired WMH volume using AccuBrain, as previously reported. ²⁵ In brief, T2 fluid‐attenuated inversion recovery images were used to calculate the signal contrast between normal brain tissue and WMH and then establish a signal threshold for WMH recognition, based on which WMHs were identified and extracted. Finally, the transformed T1‐weighted brain structure mask extracted from our study samples was used to refine and localize WMHs. AccuBrain further automatically classified WMH into periventricular WMH and deep WMH (DWMH) following the rule of “continuity to ventricle,” ie, WMH <10 mm distance from the ventricles was considered as periventricular WMH, otherwise it was DWMH. ²⁶

Six months after the initial assessment, the rater reassessed the MRI of 200 randomly selected participants for lacunes, which yielded a weighted ĸ statistic of 0.84. ²⁰ Similarly, 3 months after the initial assessment, EPVS on MRI of 30 randomly selected participants was reevaluated, which yielded a weighted ĸ statistic of 0.75 for BG‐EPVS and 0.74 for CSO‐EPVS. ²¹

DTI Processing and Analysis

We used MRtrix3 (www.mrtrix.org) to preprocess DTI data with the following steps: (1) image denoising; (2) removing Gibbs ringing artifacts; (3) motion and eddy current distortion correction; (4) bias field correction; and (5) skull stripping. ²⁷ PSMD was estimated following the PSMD marker script provided at www.psmd‐marker.com. Briefly, we used a tract‐based spatial statistics procedure to skeletonize DTI data. Normalized fractional anisotropy images were projected onto the skeleton template; then, the transformation was applied to the mean diffusivity images using the fractional anisotropy–derived projection parameters. After histogram analysis, the peak width of the mean diffusivity values within the skeleton was quantified. PSMD was defined as the difference between the 95th and 5th percentiles of the voxel‐based mean diffusivity values within the skeleton.

The pipelines for FW of periventricular WM and deep WM were based on our modified code using a single‐shell FW elimination diffusion tensor model. First, the preprocessed DTI data were fitted using the single‐shell FW elimination diffusion tensor model (https://github.com/sameerd/DiffusionTensorImaging). Briefly, we added spatial constraints by introducing a regularization term to encourage smooth transitions between neighboring voxels. Gradient descent computed the gradient of the objective function with respect to diffusion tensor parameters and took steps in the opposite direction of the gradient to reduce the overall cost. Please refer to mathematical documentation for detailed information (https://github.com/sameerd/DiffusionTensorImaging/blob/master/doc/SingleShellFreeWater.pdf). Second, the fitted FW images were normalized to the MNI152 space with a voxel size of 1×1×1 mm³, using the transform from individual DTI fractional anisotropy map to FMRIB58_FA_1mm template with ANTs toolbox. The periventricular and deep WM regions were defined according to the distance (10 mm) from the ventricles. ²⁶ Finally, the mean FW values of the periventricular and deep WM regions for each participant were extracted for statistical analysis. Figure S2 shows the example image of FW from a study participant.

Statistical Analysis

Categorical variables are presented as frequencies (percentages) and median (interquartile range [IQR]) for continuous variables. The Shapiro–Wilk test was used to assess whether continuous variables were normally distributed. We compared the characteristics of the study participants by the quartiles of CP volume using χ² test for categorical variables and Kruskal‐Wallis H test for continuous variables. We used linear regression models to examine the relationship of CP volume with demographics and CRFs, with CP volume as the dependent variable. Pearson correlation analysis was performed to evaluate the correlation between the CP volume and WMH volumes (ie, total WMH volume, DWMH volume, and periventricular WMH volume). Linear regression models were used to evaluate the association between CP volume and WMH volumes and DTI metrics, with WMH volumes and DTI metrics being the outcome variables, and we reported the estimated β coefficient and 95% CIs. Furthermore, binary logistic regression models were used to estimate the odds ratio (OR) and 95% CI of dichotomous CSVD markers (ie, presence of moderate to severe EPVS, lacunes, and CMBs) associated with CP volume, with CSVD markers being the outcome variables. Statistical interactions of CP volume with age groups, sex, and major CRFs on CSVD markers were assessed by simultaneously entering the independent variables and their cross‐product term into the same model. Once a statistical interaction was detected (P for interaction <0.05), stratified analysis was further performed to examine the direction and magnitude of the interaction. We presented the main results from 2 models: model 1 was adjusted for age, sex, education, and total intracranial volume. Model 2 was further adjusted for smoking, alcohol intake, body mass index, hypertension, diabetes, dyslipidemia, coronary heart disease, stroke, and APOE genotype.

All analyses were performed using SPSS Statistics for Windows version 26.0 (IBM) and Stata Statistical Software version 15.0 for Windows (StataCorp LLC). Two‐tailed P<0.05 was considered statistically significant.

RESULTS

Characteristics of Study Participants

The median age of the 1263 participants was 69 years (IQR, 66–72 years), 58.67% were women, and 35.47% had no formal schooling education. Compared with participants with a lower CP volume, those with a higher CP volume were older, more educated, and more likely to be male, smoke, and drink alcohol, and had a higher prevalence of diabetes (P<0.05) (Table 1). Participants across the quartiles of CP volume did not differ significantly in mean body mass index and the distribution of hypertension, dyslipidemia, coronary heart disease, stroke, and APOE ε4 allele (P>0.05) (Table 1).

Table 1.

Characteristics of Participants in the Total Sample and by Quartiles of CP Volume

Characteristics†	Total sample (N=1263)	CP volume, mL
		Quartile 1 (<2.85) (n=316)	Quartile 2 (2.85–3.61) (n=315)	Quartile 3 (3.62–4.52) (n=317)	Quartile 4 (>4.52) (n=315)	P value*
Age, y	69.0 (66.0–72.0)	68.0 (65.0–72.0)	69.0 (66.0–72.0)	70.0 (67.0–73.0)	70.0 (68.0–73.0)	<0.001
Women, n (%)	741 (58.67)	262 (82.91)	237 (75.24)	170 (53.63)	72 (22.86)	<0.001
Educational level, n (%)						<0.001
No formal education	448 (35.47)	150 (47.47)	143 (45.40)	102 (32.18)	53 (16.83)
Primary school	565 (44.73)	131 (41.46)	126 (40.00)	145 (45.74)	163 (51.75)
Middle school and above	250 (19.79)	35 (11.08)	46 (14.60)	70 (22.08)	99 (31.43)
Current smoking, n (%)	260 (20.59)	27 (8.54)	46 (14.60)	64 (20.19)	123 (39.05)	<0.001
Current drinking, n (%)	392 (31.26)	49 (15.56)	73 (23.40)	110 (35.14)	160 (50.96)	<0.001
Body mass index (kg/m2)	24.85 (22.50–27.10)	24.70 (22.20–27.30)	24.70 (22.50–27.00)	25.00 (22.90–27.60)	25.10 (22.60–26.90)	0.405
Hypertension, n (%)	849 (67.87)	207 (66.77)	204 (65.38)	209 (66.56)	229 (72.70)	0.198
Diabetes, n (%)	189 (14.96)	38 (12.03)	60 (19.05)	53 (16.72)	38 (12.06)	0.028
Dyslipidemia, n (%)	310 (24.54)	83 (26.27)	74 (23.49)	84 (26.50)	69 (21.90)	0.469
Coronary heart disease, n (%)	232 (18.37)	59 (18.67)	59 (18.73)	55 (17.35)	59 (18.73)	0.961
Stroke, n (%)	159 (12.59)	31 (9.81)	36 (11.43)	41 (12.93)	51 (16.19)	0.096
APOE ε4 allele, n (%)	188 (15.12)	50 (15.97)	39 (12.58)	52 (16.88)	47 (15.06)	0.479

Data are expressed as median (interquartile range) unless otherwise specified.

^† The number of participants with missing values was 9 for drinking alcohol, 5 for body mass index, 12 for hypertension, and 20 for apolipoprotein E (APOE) genotype.

^* P value is for the test of differences across the quartiles of choroid plexus (CP) volume.

Demographic Distributions of CP Volume

The CP volume increased with advancing age in both men and women, and men had a larger CP volume than women among all of the age groups (Figure 1). The linear regression analysis suggested that older age and male sex were significantly associated with larger CP volume (P<0.001) (Table 2). Education was not significantly associated with CP volume (P>0.05, Table 2).

Figure 1. Age‐ and sex‐specific distributions of (A) unadjusted choroid plexus (CP) volume and (B) the total intracranial volume (TIV)–adjusted CP volume.

Table 2.

Association of Demographics and Cardiovascular Risk Factors With CP Volume (n=1263)

Characteristics*	CP volume β Coefficient (95% CI), mL
	Model 1	P value	Model 2	P value
Age, y	0.06 (0.05–0.07)	<0.001	0.06 (0.05–0.07)	<0.001
Sex (female vs male)	−0.48 (−0.63 to −0.33)	<0.001	−0.47 (−0.64 to −0.29)	<0.001
Educational level, n (%)
No formal education	0.00 (reference)		0.00 (reference)
Primary school	0.04 (−0.09 to 0.17)	0.520	0.04 (−0.09 to 0.16)	0.572
Middle school or above	0.04 (−0.13 to 0.21)	0.659	0.03 (−0.14 to 0.20)	0.753
Smoking status
Never	0.00 (reference)		0.00 (reference)
Former	0.09 (−0.12 to 0.31)	0.392	0.08 (−0.14 to 0.30)	0.464
Current	0.15 (−0.05 to 0.35)	0.140	0.17 (−0.04 to 0.37)	0.107
Alcohol drinking
Never	0.00 (reference)		0.00 (reference)
Former	0.13 (−0.12 to 0.37)	0.322	0.15 (−0.10 to 0.40)	0.233
Current	0.06 (−0.10 to 0.22)	0.474	0.04 (−0.12 to 0.21)	0.605
Body mass index, kg/m2
Normal (<24)	0.00 (reference)		0.00 (reference)
Overweight (24–27.9)	0.09 (−0.02 to 0.21)	0.115	0.07 (−0.05 to 0.19)	0.256
Obesity (≥28.0)	0.10 (−0.04 to 0.25)	0.171	0.06 (−0.09 to 0.21)	0.420
Hypertension	0.09 (−0.02 to 0.21)	0.105	0.07 (−0.05 to 0.19)	0.228
Diabetes	0.18 (0.04–0.33)	0.015	0.16 (0.01–0.31)	0.035
Dyslipidemia	0.08 (−0.04 to 0.21)	0.194	0.04 (−0.09 to 0.16)	0.566
Coronary heart disease	0.01 (−0.13 to 0.14)	0.905	−0.02 (−0.16 to 0.12)	0.780
Stroke	0.17 (0.02–0.33)	0.031	0.15 (−0.01 to 0.31)	0.067
APOE ε4 allele	0.07 (−0.08 to 0.22)	0.372	0.07 (−0.08 to 0.22)	0.342

Model 1 was adjusted for age, sex, education, and total intracranial volume. Model 2 was further adjusted for smoking, alcohol intake, body mass index, hypertension, diabetes, dyslipidemia, coronary heart disease, stroke, and apolipoprotein E (APOE) genotype.

CP indicates choroid plexus.

^*The number of participants with missing values was 9 for drinking alcohol, 5 for body mass index, 12 for hypertension, and 20 for APOE genotype.

Associations of Cardiovascular Risk Factors With CP Volume

When controlling for sociodemographic factors and total intracranial volume in model 1, diabetes and stroke were both significantly associated with a larger CP volume (P<0.05) (Table 2), but the association with stroke became statistically marginal after additionally adjusting for multiple potential confounding factors in model 2 (P=0.067) (Table 2). The CP volume was not significantly associated with smoking, alcohol drinking, obesity, hypertension, dyslipidemia, coronary heart disease, and APOE genotype (Table 2).

Associations of CP Volume With CSVD Markers

The scatterplots showed significant positive correlations between WMH volumes and enlarged CP volume (Figure 2). Enlarged CP volume was correlated with higher total WMH volume (Pearson correlation coefficient r=0.23, P<0.001), DWMH volume (r=0.10, P<0.001), and periventricular WMH volume (r=0.25, P<0.001). In the multivariable‐adjusted model, greater CP volume was associated with higher total WMH volume (multivariable‐adjusted β coefficient, 0.06 [95% CI, 0.03–0.08]; P<0.001) and periventricular WMH volume (β coefficient, 0.06 [95% CI, 0.04–0.08]; P<0.001), whereas CP volume was not significantly correlated with DWMH volume (P>0.05). In addition, a greater CP volume was significantly associated with moderate to severe BG‐EPVS (multivariable‐adjusted OR, 1.17 [95% CI, 1.03–1.32]; P=0.019) (Table 3). Finally, CP volume was not significantly associated with CSO‐EPVS, lacunes, or CMBs (P>0.05).

Figure 2. Correlation of choroid plexus (CP) volume with white matter hyperintensity (WMH) volume.

The scatterplots show the correlations of CP with (A) total WMH volume, (B) deep WMH (DWMH) volume, and (C) periventricular WMH (PWMH) volume. The solid line represents the estimated β coefficient derived from the multivariable linear regression model that was adjusted for age, sex, education, total intracranial volume, smoking, alcohol intake, body mass index, hypertension, diabetes, dyslipidemia, coronary heart disease, stroke, and apolipoprotein E genotype. The shaded regions represent the 95% CI for the regression estimate. *The original data of WMH volume were natural log‐transformed.

Table 3.

Associations of CP Volume With Conventional CSVD Markers (n=1263)

CSVD markers	No. of patients	CSVD markers OR (95% CI)*
		Model 1	P value	Model 2	P value
BG‐EPVS
Mild	501	1.00 (reference)		1.00 (reference)
Moderate to severe	755	1.18 (1.04–1.33)	0.010	1.17 (1.03–1.32)	0.019
CSO‐EPVS
Mild	296	1.00 (reference)		1.00 (reference)
Moderate to severe	960	1.05 (0.91–1.21)	0.521	1.06 (0.92–1.22)	0.416
Lacunes
No	890	1.00 (reference)		1.00 (reference)
Yes	370	1.08 (0.95–1.23)	0.216	1.03 (0.91–1.18)	0.618
CMBs
No	961	1.00 (reference)		1.00 (reference)
Yes	245	1.00 (0.85–1.15)	0.883	0.97 (0.83–1.13)	0.668

Model 1 was adjusted for age, sex, education, and total intracranial volume. Model 2 was further adjusted for smoking, alcohol intake, body mass index, hypertension, diabetes, dyslipidemia, coronary heart disease, stroke, and apolipoprotein E genotype.

BG indicates basal ganglia; CMBs, cerebral microbleeds; CSO, centrum semiovale; and EPVS, enlarged perivascular spaces.

^*Odds ratio (OR; 95% CI) of different cerebral small vessel disease (CSVD) markers associated with a per 1‐mL increase in choroid plexus (CP) volume.

In the DTI subsample (n=111), the linear regression model showed that CP volume was associated with PSMD (multivariable‐adjusted β coefficient, 0.17 [95% CI, 0.03–0.32]; P=0.022), FW in deep WM (β coefficient, 0.08 [95% CI, 0.04–0.12]; P<0.001), and FW in periventricular WM (β coefficient, 0.28; [95% CI, 0.18–0.38]; P<0.001) (Table 4).

Table 4.

Associations of CP Volume With DTI‐Derived Metrics (n=111)

DTI‐derived metrics*	β Coefficient (95% CI) per 10 mL increase in CP volume, DTI‐derived metrics
	Model 1	P value	Model 2	P value
PSMD	0.16 (0.02–0.31)	0.027	0.17 (0.03–0.32)	0.022
FW in deep white matter	0.07 (0.04–0.11)	<0.001	0.08 (0.04–0.12)	<0.001
FW in periventricular white matter	0.26 (0.16–0.35)	<0.001	0.28 (0.18–0.38)	<0.001

Model 1 was adjusted for age, sex, education, and total intracranial volume. Model 2 was further adjusted for smoking, alcohol intake, body mass index, hypertension, diabetes, dyslipidemia, coronary heart disease, stroke, and apolipoprotein E (APOE) genotype.

CP indicates choroid plexus; DTI, diffusion tensor imaging; FW, free water; and PSMD, peak width of skeletonized mean diffusivity.

^*The original data of the DTI‐derived metrics were natural log‐transformed.

Interactions of CP Volume With Age, Sex, and Diabetes on CSVD Markers

There was no statistically significant interaction of CP volume with age, sex, and diabetes on any of the examined CSVD markers (all P for interaction >0.05).

DISCUSSION

In this population‐based study that targeted rural‐dwelling older adults, we found that: (1) older age, male sex, and diabetes were associated with greater CP volume; (2) a larger CP volume was associated with larger total and periventricular WMH volumes, moderate to severe BG‐EPVS, and increased PSMD and FW in periventricular and deep WM regions; and (3) there was no statistical interaction of CP volume with demographics and major CRFs on CSVD markers.

Our population‐based study showed that CP volume was higher among older individuals and that men had a larger CP volume than women among all age groups, even after correction for total intracranial volume. Along with increasing age, basement membrane thickening, infiltration of inflammatory cells, and an increased amount of connective tissue might contribute to CP enlargement. ²⁸ Given that CP expresses various receptors for sex hormones, morphology and function of CP could be modulated upon binding with the cognate hormones, ²⁹ which might partly explain the sex differences in CP volume.

Previous studies of healthy adult volunteers or patients with cognitive impairment have shown evidence that enlarged CP is associated with a higher body mass index, increased waist circumference, hypertension, diabetes, and cardiovascular disease. ³ , ⁶ However, our population‐based study of rural‐dwelling older adults showed an association of enlarged CP with only diabetes. The discrepancy may be partly attributed to differences in the study settings (eg, clinical setting versus the general population) and characteristics of the study populations (eg, age and ethnicity). For instance, the clinic‐based study from Korea enrolled patients with cognitive symptoms from a single medical center, ³ whereas the small‐scale cross‐sectional study included 123 cognitively unimpaired individuals derived from 2 population‐based studies in the United States where people had a wide age range of 21 to 94 years (mean age, 55 years). ⁶ Diabetes is known to be associated with both neurodegenerative and microvascular injury, which could be related to a higher CP volume. ³⁰

Our study showed that CP enlargement was associated with larger periventricular WMH volume and higher FW in both the deep WM and periventricular WM regions, but not with DWMH. Previous research of patients with CSVD found that the FW was increased in both WMH and normal‐appearing WM, suggesting that an increased FW could be detected before typical manifestations of CSVD and could capture microstructural injury or impaired integrity of WM invisible on conventional MRI. ³¹ In addition, we found that PSMD, which was considered a sensitive imaging biomarker of CSVD, ³² was associated with higher CP volume. Taken together, these results indicate that CP enlargement could be an early indicator or precursor of CSVD in older adults.

Glymphatic dysfunction has been implicated in CSVD. ¹¹ , ¹² , ¹³ The glymphatic system is a perivascular fluid transport system, where a considerable proportion of CSF inflows into the brain parenchyma via perivascular spaces, clears waste products from the brain parenchyma, and flows out along the perivenous spaces. ¹⁴ Thus, adequate CSF production by the CP could help maintain glymphatic function by providing a pressure gradient for fluid moving to the subarachnoid space and subsequently into the perivascular space. ¹⁴ However, CP enlargement is usually accompanied by epithelial atrophy, ²⁸ decreased CP perfusion, ⁵ and decreased CP permeability, ³ thus resulting in reduced CSF production and glymphatic malfunction. Indeed, a clinic‐based study of patients with WMH or with indications for lumbar puncture provided direct evidence that CP enlargement was associated with glymphatic dysfunction. ¹⁵ Therefore, it is possible that CP enlargement is involved in the pathophysiological processes of CSVD via the glymphatic pathway. In addition, impaired glymphatic clearance could lead to abnormal perivascular accumulation of metabolic waste, which triggers perivascular inflammation and further leads to EPVS. ¹² This was in accordance with our results showing that CP enlargement was associated with BG‐EPVS. Moreover, we did not find any association of CP volume with lacunes and CMBs. This was in contrast to the report from a clinic‐based study showing that CP enlargement was correlated with WMH volume, lacunes, and CMBs. ¹⁵ The discrepant results may partly be attributed to differences in the study settings; the clinic‐based study targeted patients with WMH or indications for lumbar puncture, among whom the prevalence rates of lacunes and CMBs were obviously higher than our study sample of older adults from the rural population.

Several potential mechanisms may underlie the associations of enlarged CP volume with CSVD. First, CP, a crucial component of the circadian clock, may indirectly affect the cardiac cycle and vascular tone by regulating the sleep–wake cycle, both of which are driving forces of glymphatic circulation. ³³ In addition, cumulative evidence indicates that CP enlargement is correlated with high levels of neuroinflammation in depression and neurodegenerative diseases. ³ , ³⁴ Furthermore, neuroinflammation is known to play a crucial role in CSVD pathogenesis ³⁵ and could result in increased extracellular FW in WMH and normal‐appearing WM. ³⁶

The major strengths of our MRI study included a large‐scale population‐based design that engaged rural‐dwelling older adults in China and the use of DTI‐derived novel metrics of CSVD in a subsample. However, several limitations deserve consideration in our study. First, the cross‐sectional design does not allow us to infer causality for any of the observed associations. Second, owing to the relatively small DTI sample, future large‐scale studies are warranted to further clarify the relationships of CP volume with the DTI‐derived novel metrics for CSVD. Third, we used single‐shell data to calculate FW, which has limitations compared with multishell data, although single‐shell data could provide plausible FW parameter maps. ³⁷ Finally, our sample was derived from only one rural region in western Shandong Province, which should be kept in mind when generalizing our findings to other populations.

CONCLUSION

This population‐based study of rural Chinese older adults indicates that older age, male sex, and diabetes are associated with larger CP volume. In addition, enlarged CP volume was associated with an increased burden of total and periventricular WMH, moderate to severe BG‐EPVS, and reduced WM integrity, which suggests that enlarged CP could be a valuable imaging marker for CSVD. These findings extend our understanding of the pathophysiology of enlarged CP and CSVD. Future large‐scale prospective cohort studies are warranted to elucidate the potential temporal relationships of enlarged CP volume with various markers for CSVD, as well as the pathophysiological mechanisms underlying these relationships.

Sources of Funding

The MIND‐China project was financially supported by the STI2030‐Major Project (grant numbers 2021ZD0201808 and 2022ZD0211600) and by additional grants from the National Key R&D Program of China (grant number 2017YFC1310100), the National Nature Science Foundation of China (grants numbers 82171175, 82 011 530 139, and 82 001 120), the Academic Promotion Program of Shandong First Medical University (grant numbers 2019QL020 and 2020RC009), and the Taishan Scholar Program of Shandong Province (grant numbers ts20190977 and Tsqn201909182). This work was further supported by additional grants from the Nature Science Foundation of Shandong Province (ZR2020QH098) and the Integrated Traditional Chinese and Western Medicine Program in Shandong Province (YXH2019ZXY008). J.W. received funds from the Weston Brain Institute in the same topic area of perivascular spaces but not directly funding this work, and from the Research Councils UK for the Dementia Research Institute and a bit of work time. C.Q. received grants from the Swedish Research Council (grant numbers 2017–05819 and 2020–01574) for the Sino‐Sweden Network and Research Projects, the Swedish Foundation for International Cooperation in Research and Higher Education (STINT) (grant number CH2019‐8320) for the Joint China‐Sweden Mobility program, and the Karolinska Institutet, Stockholm, Sweden. The funding agencies had no role in the study design, data collection and data analysis, writing of this article, or decision to submit the work for publication.

Disclosures

None.

Supporting information

Figures S1–S2