Aim
In this study, to investigate the pathophysiology of DESH, the volumes of DESH-related regions (ventricles, SF, and high-convexity and medial subarachnoid spaces) and brain parenchyma were quantitatively evaluated in a large sample of community-dwelling individuals aged ≥ 65 years. Our objectives were (ⅰ) to examine whether DESH is an age-related disorder or an accelerated aging stage; (ⅱ) to clarify whether brain atrophy is associated with the development of DESH-like morphology; (ⅲ) to explore the factors associated with the development of DESH-like morphology; and (ⅳ) to verify whether age-related DESH-like morphological changes are associated with clinical symptoms such as cognitive and gait dysfunctions.
Study Design And Participants
A total of 1,577 community-dwelling older residents in Arao City, Kumamoto Prefecture (southern Japan), were enrolled between November 2016 and March 2017. This cross-sectional analysis using baseline data was part of the Japan Prospective Studies Collaboration for Aging and Dementia, which is designed to enroll approximately 10,000 community-dwelling residents aged ≥ 65 years from eight sites in Japan to explore the genetic and environmental risk factors for dementia [17]. Participants were excluded if they had no or unsuitable magnetic resonance imaging (MRI) data, missing data, dementia (i.e., to examine the age-related changes in CSF dynamics), or severe gait disturbance (i.e., to exclude the effect of musculoskeletal disorders on gait assessment) [see Additional file 1: Table S1–4].
Standard Protocol Approvals, Registrations, And Patient Consents
Standard protocol approvals, registrations, and patient consents
This study was approved by the Research Ethics Committee of Kumamoto University (Kumamoto, Japan; approval number, GENOME-333). All participants provided written informed consent prior to data collection in accordance with the Declaration of Helsinki.
Procedures
Standardized approaches for questionnaires, blood tests, and dementia diagnosis were applied across all study sites, as previously described [17]. Cognitive function was assessed using the Mini-Mental State Examination (MMSE) [18]. Dementia [19] and its subtypes [17] were diagnosed based on standard criteria, and Petersen’s criteria were used to diagnose mild cognitive impairment (MCI) [20]. Individuals without dementia or MCI were considered cognitively normal in this study.
Gait was assessed using the Timed Up and Go (TUG) test [21]. The TUG test was performed twice, and the shorter time was used for the analysis. Gait disturbance was defined as a TUG time > 12.0 s [22], and severe gait disturbance as a TUG time > 16.7 s (three standard deviations above the mean of this cohort). Hypertension was defined as a blood pressure ≥ 140/90 mmHg and/or the use of antihypertensive agents. Diabetes was defined as fasting blood glucose ≥ 126 mg/dL, casual blood glucose ≥ 200 mg/dL, hemoglobin A1c ≥ 6.5%, and/or the use of glucose-lowering agents. Dyslipidemia was defined as low-density lipoprotein ≥ 140 mg/dL and/or high-density lipoprotein < 40 mg/dL, and/or taking medication for dyslipidemia. Body mass index (BMI) was calculated using body height and weight. The criteria for atrial fibrillation were based on self-reported questions and/or electrocardiographic evidence. Information on education level, history of coronary artery disease, heart failure, and smoking was obtained using self-reported questionnaires.
Imaging And Diagnoses
Brain MRI was conducted at the Arao Municipal Hospital (Kumamoto, Japan) and Omuta Tenryo Hospital (Fukuoka, Japan) using the 1.5-Tesla Ingenia CX dual scanner (Philips Healthcare, Best, Netherlands) or the 1.5-Tesla Signa HDxt Ver.23 scanner (GE Healthcare, Milwaukee, WI, USA). The Philips MRI scanning protocol consisted of a three-dimensional (3D) T1-weighted sequence (repetition time = 8.6 ms, echo time = 4.0 ms, flip angle = 9°, matrix = 192×192, slice thickness = 1.2 mm), a 3D T2-weighted sequence (repetition time = 5082.3 ms, echo time = 100.0 ms, flip angle = 90°, matrix = 356×248, slice thickness = 5.0 mm), a 3D fluid-attenuated inversion recovery (FLAIR) sequence (repetition time = 11000.0 ms, echo time = 120.0 ms, flip angle = 90°, matrix = 288×203, slice thickness = 5.0 mm), and a susceptibility-weighted imaging (SWI) sequence (repetition time = 78.4 ms, echo time = 41.4 ms, flip angle = 20°, matrix = 88×272, slice thickness = 2.4 mm). The GE Signa MRI scanning protocol consisted of a 3D T1-weighted sequence (repetition time = 8.3 ms, echo time = 3.4 ms, flip angle = 8°, matrix = 192 × 192, slice thickness = 1.2 mm), a 3D T2-weighted sequence (repetition time = 4517.0 ms, echo time = 92.6 ms, flip angle = 90°, matrix = 352 × 224, slice thickness = 5.0 mm), a 3D FLAIR sequence (repetition time = 10000.0 ms, echo time = 149.2 ms, flip angle = 90°, matrix = 288 × 193, slice thickness = 5.0 mm), and a 3D T2-Star weighted angiography sequence (repetition time = 75.2 ms, echo time = 57.9 ms, flip angle = 20°, matrix = 320×200, slice thickness = 3.0 mm).
FLAIR MRI was used to assess vascular diseases such as infarction and white matter hyperintensity (WMH) load. Participants with large vascular lesions (e.g., cortical infarction and hemorrhage), tumors, and artifacts were excluded from the analysis [see Additional file 1: Table S3]. The degree of WMH load was rated visually on axial FLAIR images by using the Fazekas scale (i.e., grade 1 [punctate], grade 2 [early confluent], or grade 3 [confluent]) in the periventricular and deep white matter (WM) regions [23]. The sum of the periventricular and deep WMH scores, ranging from 0 to 6, was used for the analysis. Lacunar infarction was defined as a CSF-like hypo-intensity with a diameter > 2 mm surrounded by a rim of hyperintensity on T2 FLAIR. Lacunar infarction was considered present if there was at least one visible lacunar infarction. Microbleeds were assessed using SWI or T2-Star software and recorded using the same method used for lacunar infarction. The Fazekas scale, lacunar infarction, and microbleeds were the covariates. All brain images were assessed by a neuroradiologist (N.T.) and two neuropsychiatrists (Y.H. and M.H.) who were blinded to the clinical data.
A visual rating scale was used to assess the DESH-related regions (Fig. 1a–c). The Evans index (EI) value was the ratio of the maximum diameter of the frontal horns of the lateral ventricles to the maximum inner diameter of the skull on the transverse section. The ventricular system (VS) was classified as “not dilated” if EI ≤ 0.3 or “dilated” if EI > 0.3 (Fig. 1a). Ventriculomegaly was defined as an EI > 0.3. The SF was assessed in transverse and coronal sections. SF enlargement was rated as ,0 normal; 1, mildly dilated; 2, moderately dilated; and 3, severely dilated (Fig. 1b). Scores of 2 or 3 indicated enlarged SF. The subarachnoid spaces at high convexity and midline (SHM) were evaluated using transverse and coronal section images, and their tightness was rated as follows: 0, not tight; 1, moderately tight; and 2, severely tight (Fig. 1c). Scores of 1 or 2 constituted tight SHM. Participants meeting all criteria for ventriculomegaly, enlarged SF, and tight SHM were diagnosed with DESH. “Possible iNPH with MRI support” was defined as DESH with MCI and/or gait disturbance. “Asymptomatic ventriculomegaly with features of iNPH on MRI (AVIM)” was defined as DESH with neither MCI nor gait disturbance.
To quantitatively assess the DESH-related regions, we utilized an automatic volumetric segmented brain image system, which was modified to evaluate iNPH [24, 25]. We prepared voxels of interest (VOI) templates for the intracranial volume, VS, SF, and SHM (Fig. 1d–f), as previously described [26]. Each regional VOI template was produced on a digital phantom Simulated Brain Database (http://www.bic.mni.mcgill.ca/brainweb/), according to the standard Montreal Neurological Institute space, with manual delineation of the contours of each structure. The SHM VOI template was produced manually in reference to the results of a previous study on voxel-based morphometry (VBM) in iNPH patients and normal controls [27]. For this process, the MRI for each participant was segmented into gray matter (GM), WM, and CSF by the SPM8 segmentation program (Wellcome Trust Centre for Neuroimaging, London, UK). The GM template image derived from the Simulated Brain Database was then spatially transformed into the GM image for each participant, and a normalization parameter was produced by using SPM8 and the Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra technique. This normalization parameter functions in the same manner as a reverse parameter produced in the anatomical normalization of an individual brain to a standard brain. With this parameter, the intracranial volume, VS, SF, and SHM VOI templates were transformed to each participant’s space. The intracranial volume was adjusted by using an image derived from the segmented GM, WM, and CSF images. The segmented images were derived by calculating the WM and GM areas with the voxels from the intracranial volume VOI template. CSF volumes of the VS, SF, and SHM were calculated using the transformed VS, SF, and SHM subarachnoid space VOI templates for each participant. Each regional volume was normalized to the total intracranial volume.
Volumetric segmentation was attained using FreeSurfer version 5.3 (http://surfer.nmr.mgh.harvard.edu/) on CentOS6. Using the Desikan–Killany Atlas, we measured 34 cortical regions, 7 subcortical regions, corpus callosum, cerebellum cortex, and vessels (represented perivascular space around the basal ganglia) in absolute volumes [see Additional file 1: Table S5] [28, 29]. The vessel volume was used as a covariate. Each regional volume was normalized to the total intracranial volume.
Statistical analysis
Statistical analysis was conducted using SPSS version 27.0 (IBM Corp., Armonk, NY, USA). To examine whether the volume of DESH-related regions showed continuity between normal aging and DESH, we analyzed the data in two groups: all individuals (n = 1,356) and the non-DESH group (n = 1,331).
Continuous variables were compared using t-tests. Differences in proportions were compared using the chi-square test. To explore the factors associated with DESH-related regions and examine correlations between brain structure volumes and age, we used multivariate linear regression while adjusting for the following possible confounders: sex, education, hypertension, diabetes mellitus, dyslipidemia, atrial fibrillation, coronary artery disease, heart failure, BMI, history of smoking, Fazekas score, lacunar infarction, microbleeds, and perivascular space. We used Pearson’s correlation coefficient (r) to determine the association between the VS, SF, and SHM volumes and total brain volume.
To identify the brain structures, including DESH-related regions (i.e., VS, SF, SHM), that affect cognitive function, we conducted hierarchical multiple regression analysis using the MMSE score as the dependent variable. After entering the independent variables (age, sex, education, hypertension, diabetes mellitus, dyslipidemia, atrial fibrillation, coronary artery disease, heart failure, BMI, history of smoking, MRI scanner, Fazekas score, lacunar infarction, microbleeds, and perivascular space) in Block 1, each brain structure was entered in Block 2. For each analysis, the standardized regression coefficient (βSTD) and the change in R2 (∆R2) were calculated. We ranked the effect of the brain structures on the MMSE score by descending ∆R2. The same analysis was conducted to evaluate the brain structures affecting gait function (i.e., TUG score).
For all multivariate analyses, we visually inspected normal Q-Q plot to check for normality of residuals. Collinearity was examined using the variance inflation factor (values > 10 were considered problematic). The Durbin–Watson statistic was used to identify autocorrelation (values < 1 and > 3 were considered problematic). Cook’s distance was calculated to check for influential outliers (values > 0.5 were considered problematic).
For all analyses, significance was set at P < 0.05. To correct for multiple comparisons, we employed Bonferroni corrections; for example, the significance level for the comparison of 46 brain structures (43 brain regions and three DESH-related regions) was P < 0.0011 (0.05/46).