The impaired white-matter structural network mediates the relationship between the glymphatic system and emotional abnormalities in patients with chronic kidney disease

Abstract

Background

Patients with chronic kidney disease (CKD) tend to have white-matter network and glymphatic system dysfunction, and may experience cognitive decline and psychiatric abnormalities. This study explored the relationships between graph theory-based white-matter networks and diffusion tensor image analysis along the perivascular space (DTI-ALPS) and the choroid plexus volume (CPV).

Methods

Sixty-one CKD patients, including 22 non-dialysis dependent (NDD) and 39 dialysis-dependent (DD) patients, and 43 age-, gender-, and education-matched healthy control (HC) participants were recruited for the study. All participants underwent blood biochemical tests, neuropsychological assessments, and magnetic resonance imaging scans. Preprocessed images were used to construct fractional anisotropy networks, and calculate the DTI-ALPS index and CPV. Intergroup comparisons were performed for all measurements, and correlations were analyzed.

Results

Compared to the HCs, the DD-CKD patients had elevated small-worldness (σ) and normalized clustering coefficient (P<0.01) values, while no significant statistical differences were observed between NDD-CKD patients and the HCs. The clustering coefficient (P<0.05), and global and local efficiency (P<0.01) values were decreased, while the characteristic path length (P<0.01) was increased in the CKD patients, especially those who were DD. The CKD patients also had decreased DTI-ALPS index (P=0.001) and increased CPV (P<0.001) values. These indicators were correlated with verbal memory, anxiety, and depression levels (all P<0.05). The mediation analysis revealed that σ partially mediated the effect of the DTI-ALPS index on depression levels (mediation effect: −5.0404), σ fully mediated the effect of the CPV on depression levels (mediation effect: 0.0013), and σ fully mediated the effect of the DTI-ALPS index on anxiety levels (mediation effect: −5.9291).

Conclusions

CKD patients exhibit impaired glymphatic system function and abnormal white-matter network topology, both of which are correlated with cognition, verbal memory, and emotional states. The structural integrity and organizational efficiency of white-matter networks may be a critical factor linking glymphatic dysfunction to emotional health in CKD patients.

Introduction

Patients with chronic kidney disease (CKD) are at a significantly higher risk of cognitive impairment, including confusion, encephalopathy, and dementia, than the general population (1). Depending on the stage of CKD and the assessment methods used, the prevalence of cognitive impairment has been reported to range from 10–40% (2,3). Cognitive impairment becomes particularly severe in end-stage patients. Further, hemodialysis-induced hemodynamic disturbances significantly increase the risk of various adverse outcomes (4,5). There may be multiple mechanisms of association between CKD and cognitive impairment. Complications of CKD, such as uremia, other metabolic abnormalities, and anemia, are risk factors for cognitive impairment (6). It has also been suggested that the accumulation of uremic neurotoxins may be more crucial in the pathogenesis of cognitive impairment in CKD than hemodynamic factors or lipid metabolism disorders (7).

Cerebrospinal fluid (CSF) enters the brain extracellular space through the perivascular space (PVS) and is exchanged with the interstitial fluid to remove waste products (8). This clearance pathway, known as the glymphatic system, is critical for maintaining homeostasis in brain function (9,10). There are several methods to assess the function of the glymphatic system. Of these, non-invasive magnetic resonance imaging (MRI) is more widely used on human subjects compared to those invasive methods. The diffusion tensor image analysis along the perivascular space (DTI-ALPS) technique uses a non-invasive diffusion tensor imaging (DTI) sequence to assess the water diffusion capacity of the PVS (11). Lower diffusivity suggests glymphatic system impairment. A previous study demonstrated the robustness of the DTI-ALPS results using the fixed imaging method (12). In addition, PVS ratings, choroid plexus volume (CPV) measurements, and free water analyses are also used to assess the glymphatic system (13). Recently, numerous studies have found that an abnormal DTI-ALPS index is associated with various neurological disorders, such as Alzheimer’s disease, epilepsy, dementia, and cerebrovascular disease (14-17). Given the influence of factors such as uremic neurotoxin accumulation, hemodynamic alterations, and lipid metabolism disorders, it can be hypothesized that glymphatic system dysfunction exists in CKD patients and contributes to abnormal cognitive and emotional changes. Previous studies have identified abnormalities in the DTI-ALPS index in both early and end-stage CKD (18,19).

The human brain is a connectome with complex functions and structures. Recently, graph-theoretic analysis has been widely used to explore the structure of human brain white-matter networks and any alterations in patients with diseases. Important local and global organization parameters for human brain structure can be quantified by graph-theoretic analysis (20). The small-world network is an economical structure with low wiring costs, and efficient information isolation and integration (21). White-matter microstructural lesions in CKD patients may lead to impaired small-world network topology (22). Studies have shown that CKD patients exhibit abnormal small-world functional network properties and reduced structural network efficiency compared to normal subjects (23,24). Additionally, CKD patients have also shown abnormal spontaneous functional activity in multiple brain regions associated with the default mode network (25,26). However, structural network abnormalities in CKD have been less studied, and the differences in structural networks between maintenance hemodialysis and non-dialysis CKD patients remain to be explored. To the best of our knowledge, to date, no studies on the structural network properties and indicators for the assessment of the glymphatic system in CKD patients have been conducted.

In this study, we hypothesized that patients with CKD, especially dialysis-dependent (DD) patients, have impaired glymphatic systems and white-matter networks, and that these impairments are associated with executive function, verbal memory, anxiety, and depression levels. We evaluated the glymphatic system non-invasively using the DTI-ALPS method and CPV measurements, analyzed the topological properties of the cerebral white-matter structural network using graph theory, and explored their associations using correlation, regression, and mediation analyses. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-24-1692/rc).

Methods

Participants

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of the First Affiliated Hospital of Dalian Medical University (No. PJ-KS-KY-2021-121), and informed consent was obtained from all individual participants. CKD patients were included in the study if they met the following inclusion criteria: (I) had been confirmed by a nephrologist to meet the Kidney Disease Outcomes Quality Initiative diagnostic criteria (27); (II) were aged >18 years; and (III) were right-handed. The CKD patients were further divided into the non-dialysis dependent chronic kidney disease (NDD-CKD) and dialysis-dependent chronic kidney disease (DD-CKD) groups based on whether they received hemodialysis treatment (3–4 times per week for at least 3 months) or not. Patients were excluded from the study if they met any of the following exclusion criteria: (I) had a previous traumatic brain injury, psychiatric disease, or other neurological disorder; (II) had received a renal transplant; and/or (III) had a contraindication to MRI examination. Ultimately, 61 patients were prospectively recruited for this study from April 2020 to October 2023, of whom 22 had NDD-CKD and 39 had DD-CKD. Additionally, 43 healthy control (HC) participants, who were right-handed and compatible with patients in terms of age, gender and education level, were recruited from local communities. The exclusion criteria for the HCs were the same as those for the patient group.

Laboratory and neuropsychological assessments

All participants underwent neuropsychological assessments by a well-trained neurologist before MRI data acquisition, including the Beijing version of Montreal Cognitive Assessment (MoCA) (28) for cognitive valuation, Rey’s auditory verbal learning test (AVLT) (29) for verbal memory evaluation, the digital symbol substitution test (DSST) (30) for attention and visual-motor coordination evaluation, and the trail making test (TMT) parts A and B for executive ability evaluation (31). The Hamilton anxiety scale (HAM-A) (32) and Hamilton depression scale (HAM-D) (33) were also administered. All the CKD patients underwent clinical data collection and blood biochemical tests, including course of disease, hemoglobin, albumin, serum creatinine, serum urea, uric acid, cholesterol, triglyceride, serum sodium, serum calcium, and serum phosphorus tests.

MRI acquisition

Each participant underwent a brain MRI scan, using a 3.0 T MRI scanner (Ingenia CX, Philips Healthcare, Best, the Netherlands) equipped with a 32-channel phased-array head coil. The MRI protocol included the following sequences for the (I) high-resolution three-dimensional T1-weighted multi-shot turbo field echo sequence: echo time (TE) =3.0 ms, repetition time (TR) =6.6 ms, flip angle =12°, slices =188, field of view (FOV) =256×256 mm², matrix size =256×256, and thickness =1.0 mm; and (II) diffusion-weighted single-shot echo planar imaging sequence: TE =92 ms, TR =6,000 ms, flip angle =90°, voxel size =2×2×2 mm³, FOV =256×256 mm², matrix size =128×128, axial slice =68, and thickness =2 mm (to cover the whole brain without gap). Each DTI dataset included 64 non-collinear spatial directions with a b value of 1,000 s/mm², and one baseline image with a b value of 0 s/mm². In addition, conventional whole-brain axial fluid-attenuated inversion recovery scans were used to detect the presence of other lesions in the participant’s brain. DTI and T1-weighted images were used for subsequent processing and analysis processes as shown in Figure 1.

Figure 1.

Processing and analysis procedures. AAL, anatomical automatic labeling; DTI-ALPS, diffusion tensor image analysis with the perivascular space; Dxx, diffusivity along the x-axis; Dyy, diffusivity along the y-axis; Dzz, diffusivity along the z-axis; FA, fractional anisotropy.

MRI analysis

Structure network construction

The DTI data were preprocessed using PANDA toolbox (34) based on the Functional MRI of the Brain (FMRIB) Software Library (FSL) version 5.0 (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki) and the Diffusion Toolkit (http://www.trackvis.org/dtk/). The preprocessing procedure included eddy current-induced distortion correction, motion artifact correction, skull stripping, and an image quality check. The deterministic tractography was subsequently applied using the Fiber Assignment by Continuous Tracking algorithm with parameters of a fractional anisotropy (FA) value threshold of 0.2 and a turning angle threshold of 45° (35).

The whole brain was segmented into 90 brain regions using the anatomical automatic labeling parcellation. The average FA value of each connection between each two brain regions was calculated and saved as a weighted connectivity edge. To reduce the risk of spurious connections due to noise and minimize the number of spurious connections in the network, a fiber number ≥3 between brain regions was set as the criterion for the FA network matrix construction (36). The following graph-theoretic topological features of each participant’s weighted FA network matrix were then computed using the GRETNA toolbox (37): (I) small-world parameters, including the clustering coefficient (Cp) and normalized clustering coefficient (γ), which were used to quantify the local interconnectivity of the network, characteristic path length (Lp) and normalized characteristic path length λ, which were used to quantify the overall routing efficiency of the network, and small-worldness (σ) (the ratio of γ and λ), which measures the σ of a network; and (II) network efficiency, including global efficiency (Eg), which was used to measure the ability of parallel information transmission over the network, and local efficiency (Eloc), which was used to measure the fault tolerance of the network. We generated 500 random networks matching real networks for the calculation of the small-world parameters. For detailed calculations and explanations of these network measures, please refer to the work of Rubinov and Sporns (38).

CPV segmentation and quantization

The T1-weighted images of all participants were automatically segmented using FreeSurfer version 7.2 (http://surfer.nmr.mgh.harvard.edu/) to obtain the left and right CPV for each participant. The total intracranial volume was also obtained using the FreeSurfer function for estimating the total intracranial volume.

Quantization of the DTI-ALPS index

Diffusion maps in the x-axis (Dxx, right-left), y-axis (Dyy, anterior-posterior), and z-axis (Dzz, inferior-superior) directions were obtained for each participant using the FSL toolbox. Using standard color-coded FA maps, regions of interest (ROIs) of 5 mm in diameter were placed on the left side of the lateral ventricular body on the projection, association, and subcortical fibers by two radiologists independently (Figure 1). Next, the diffusivity of the ROIs was extracted in the three directions of the x, y, and z axes. The averaged ALPS index reflects the ratio of the water diffusivity along the peripapillary space of the medullary veins to the diffusivity along the other non-fibrous running directions. The DTI-ALPS index was calculated using the following formula (11):

$$DTI\text{-}ALPSindex=\frac{mean\left(Dxx\text{\_}proj,Dxx\text{\_}associ\right)}{mean\left(Dyy\text{\_}proj,Dzz\text{\_}associ\right)}$$ [1]

where Dxx_proj and Dxx_associ refer to the diffusivity of the projection and association fibers along the x-axis, respectively; Dyy_proj refers to the diffusivity of the projection fiber along the y-axis; and Dzz_associ refers to the diffusivity of the association fiber along the z-axis.

Statistical analysis

All the statistical analyses were performed using IBM SPSS Statistics version 27.0 (IBMCorp., Armonk, NY, USA) and R (version 4.2.1). The demographic and clinical factors were compared using the chi-squared test, one-way analysis of variance (ANOVA), or non-parametric test depending on the distribution of the data. Structural network measures, diffusivities, and the DTI-ALPS index were compared using ANOVA. Post-hoc analysis comparisons were performed using the least significant difference (LSD) method. The Spearman correlation test and multiple linear regression were used to explore the associations between the indicators. Mediation analyses were carried out to explore the mediation effects among the glymphatic system evaluation indicators, brain white-matter network topography measures, and clinical scales using PROCESS version 3.5 (http://www.processmacro.org/). A P value <0.05 was considered statistically significant.

Results

Demographic and clinical characteristics

Table 1 summarizes the demographic and clinical characteristics that were statistically different among groups. There were no significant differences between the groups in terms of gender, age, and years of education (P>0.05). The DD-CKD patients had increased serum urea, serum creatinine, and serum calcium levels, and decreased serum sodium levels compared to the NDD-CKD patients. Compared to the HCs, the DD-CKD patients had decreased MoCA attention and DSST scores, and increased TMT, HAM-A, and HAM-D scores. All the CKD patients had decreased cholesterol, total MoCA scores, and MoCA subscores, including visuospatial/executive and language subscores, compared with the HCs. Compared to the HCs, the DD-CKD patients had decreased MoCA attention scores, while no significant statistical differences were observed between NDD-CKD patients and the HCs. Additionally, there were differences in albumin, hemoglobin, and the estimated glomerular filtration rate (eGFR) among the three groups. The MoCA delayed recall of the NDD-CKD patients was lower than that of the HCs and DD-CKD patients (all P<0.05, LSD corrected). Comparative results for all data are presented in the Supplementary Material (Table S1).

Table 1. Demographics and clinical characteristics.

Characteristics	DD-CKD group (n=39)	NDD-CKD group (n=22)	HC group (n=43)	t/F/χ2/H	P value
Demographics
Age, years	60.00 (51.00, 66.00)	55.50 (42.00, 64.50)	57.00 (50.00, 65.00)	1.367	0.505
Gender (male/female), n	22/17	14/8	20/23	1.882	0.390
Education, years	12 (9, 14)	9 (8, 12)	12 (9, 15)	5.962	0.051
Hypertension (%)	89.74	81.82	–	0.775	0.379
Diabetes mellitus (%)	30.77	40.91	–	0.641	0.423
Laboratory tests
Hemoglobin, g/L	112.97±11.27	99.33±20.26	143.98±14.16	83.219	<0.001*abc
Albumin, g/L	39.08±3.37	37.64±4.42	44.60±2.51	48.791	<0.001*abc
eGFR, mL/min/1.73 m2	4.13 (3.50, 4.86)	16.00 (8.59, 28.15)	99.61 (93.27, 105.02)	86.779	<0.001*abc
Total cholesterol, mmol/L	4.41±1.19	4.66±1.63	5.54±0.98	9.124	<0.001*ab
Serum urea, mmol/L	25.81±6.69	20.95±8.95	–	−2.405	0.019*
Serum creatinine, μmol/L	937.00 (853.00, 1,033.00)	317.50 (211.50, 529.50)	–	33.265	<0.001*
Serum sodium, mmol/L	138.97±2.70	141.05±1.99	–	3.145	0.003*
Serum calcium, mmol/L	2.27 (2.17, 2.41)	2.06 (1.90, 2.17)	–	21.068	<0.001*
Neuropsychological assessments
MoCA	26.00 (23.00, 27.00)	24.50 (19.50, 26.00)	27.00 (25.00, 28.00)	16.091	<0.001*ab
MoCA_Visuospatial/Executive	4.00 (3.00, 5.00)	4.00 (2.75, 5.00)	5.00 (5.00, 5.00)	25.059	<0.001*ab
MoCA_Attention	5 (4, 6)	6 (4, 6)	6 (5, 6)	9.971	0.007*a
MoCA_Language	2 (2, 3)	2 (2, 2)	3 (2, 3)	13.602	0.001*ab
MoCA_Delayed Recall	4.00 (3.00, 5.00)	2.00 (0.00, 3.25)	3.00 (2.00, 5.00)	12.961	0.002*bc
TMT part A	56.85 (45.4, 73.12)	49.00 (30.45, 75.84)	43.87 (33.55, 52.31)	7.963	0.019*a
TMT part B	87.88 (61.73, 135.35)	80.11 (52.40, 116.40)	58.59 (45.31, 90.96)	9.317	0.009*a
DSST	32.00 (27.00, 42.25)	37.50 (28.25, 53.00)	43.00 (31.00, 58.00)	7.575	0.023*a
HAM-A	9.00 (4.00, 16.00)	8.00 (3.75, 14.00)	4.00 (2.00, 7.00)	10.567	0.005*a
HAM-D	9.00 (6.00, 15.00)	8.50 (5.00, 11.25)	6.00 (4.00, 11.00)	8.885	0.012*a

Values are reported as the mean ± standard deviation or median (interquartile range) for the quantitative variables. *, P<0.05. ^a, ^b, and ^c denote statistically significant differences between DD vs. HC, NDD vs. HC, and NDD vs. DD, respectively, with least significant difference adjustment. CKD, chronic kidney disease; DD, dialysis-dependent; DSST, digital symbol substitution test; eGFR, estimated glomerular filtration rate; HAM-A, Hamilton anxiety scale; HAM-D, Hamilton depression scale; HC, healthy control; MoCA, Montreal Cognitive Assessment (Beijing version); NDD, non-dialysis dependent; TMT, trail making test.

Comparison of structure network measures

The σ values of the white-matter networks in the CKD patients and HCs were much larger than 1, suggesting that they all exhibited a typical small-world topology (i.e., the white-matter networks of all participants had larger Cps and almost the same characteristic Lps compared to the matched random networks) (Figure 2). However, the results revealed significant differences in the Cp, characteristic Lp, σ, γ, Eg, and Eloc indicators among the three groups (P<0.05). The post-hoc comparisons showed that compared to the HCs, the CKD patients had significantly decreased Cp, Eg, and Eloc, and increased characteristic Lp. Decreased Cp, Eg, and Eloc, and increased Lp were also observed in the DD-CKD patients compared to the NDD-CKD patients. In addition, significantly elevated γ and σ were found in the DD-CKD patients compared to the HCs (P<0.05, LSD corrected).

Figure 2.

Comparison of network topology metrics. *, P<0.05; **, P<0.01, least significant difference corrected. CKD, chronic kidney disease; DD, dialysis-dependent; HC, healthy control; NDD, non-dialysis dependent.

Comparison of the DTI-ALPS index and CPV

The comparison results for the DTI-ALPS index and CPV are shown in Table 2 and Figure 3. Interobserver agreement was excellent for the DTI-ALPS index (interclass correlation coefficient: 0.880). The ANOVA results revealed significant differences in the DTI-ALPS index among the three groups. The post-hoc comparisons revealed that the DTI-ALPS index was significantly decreased in the NDD-CKD and DD-CKD patients compared to the HCs (P<0.05, LSD corrected). No difference was observed in the DTI-ALPS index between the NDD-CKD and DD-CKD patients. Additionally, when age was included as a covariate in the statistical analysis, the results remained consistent. Conversely, apart from the Dxx_proj and the diffusivity of the subcortical fibers along the x-axis (Dxx_subcortical), the diffusivity measures of the CKD patients were higher than those of the HCs (P<0.05, LSD corrected). The CPV of the CKD patients was higher than that of the HCs (P<0.05, LSD corrected). However, no such statistically significant difference was observed between the NDD-CKD and DD-CKD patients.

Table 2. Between-group comparisons of glymphatic system indicators.

Parameter	DD-CKD group (n=39)	NDD-CKD group (n=22)	HC group (n=43)	P value
Projection fiber
Dxx	0.604 (0.515–0.715)	0.609 (0.458–0.683)	0.602 (0.497–0.674)	0.784
Dyy	0.543 (0.419–0.787)	0.590 (0.449–0.708)	0.499 (0.381–0.610)	<0.001*
Dzz	1.091 (0.907–1.398)	0.973 (0.770–1.070)	0.953 (0.793–1.192)	<0.001*
Association fiber
Dxx	0.662 (0.518–0.837)	0.655 (0.550–0.743)	0.618 (0.494–0.750)	0.002*
Dyy	1.161 (1.020–1.345)	1.121 (1.037–1.405)	1.112 (0.979–1.311)	0.009*
Dzz	0.365 (0.238–0.534)	0.339 (0.233–0.425)	0.316 (0.250–0.437)	<0.001*
Subcortical fiber
Dxx	1.081 (0.947–1.260)	1.072 (0.964–1.377)	1.054 (0.842–1.247)	0.282
Dyy	0.799 (0.640–1.023)	0.761 (0.652–0.924)	0.659 (0.436–1.036)	<0.001*
Dzz	0.673 (0.419–0.964)	0.583 (0.493–0.757)	0.567 (0.452–0.885)	<0.001*
DTI-ALPS index	1.416 (1.142–1.685)	1.329 (1.056–1.805)	1.491 (1.240–1.922)	0.001*
CPV	1.396 (0.817–2.855)	1.413 (0.713–2.792)	0.912 (0.417–2.065)	<0.001*

Data are presented as the median (range). Diffusivities are presented as the apparent diffusion coefficients (×10⁻³ mm²/s). *, P<0.05. CKD, chronic kidney disease; CPV, choroid plexus volume; DD, dialysis-dependent; DTI-ALPS, diffusion tensor image analysis with the perivascular space; Dxx, diffusivity along the x-axis; Dyy, diffusivity along the y-axis; Dzz, diffusivity along the z-axis; HC, healthy control; NDD, non-dialysis dependent.

Figure 3.

Comparison of the DTI-ALPS index and choroid plexus volume. *, P<0.05; **, P<0.01, least significant difference corrected. CKD, chronic kidney disease; DD, dialysis-dependent; DTI-ALPS, diffusion tensor image analysis with the perivascular space; HC, healthy control; NDD, non-dialysis dependent.

Analysis of associations between indicators

The correlation analysis results (Figure 4) showed that in both the NDD-CKD and DD-CKD patients, the DTI-ALPS index and CPV were correlated with multiple white-matter network topological metrics. The white-matter network topological metrics were also correlated with the accuracy of Rey’s AVLT in the patient group, MoCA abstraction scores in the NDD-CKD group, and the DSST and TMT part B scores in the DD-CKD group. Additionally, the DTI-ALPS index (r=−0.426, P=0.048) was negatively correlated with cholesterol. Eloc (r=0.317, P=0.049) was positively correlated with creatinine. Additionally, we analyzed the correlations within the HC group. The results indicated that while there were correlations among glymphatic system measures, white-matter network metrics, cognitive levels, and emotional states, these correlations were not particularly strong. A heat map of the correlation results is provided in the Supplementary Material (Figure S1).

Figure 4.

Correlation analysis of the DTI-ALPS index, choroid plexus volume, network indicators, and laboratory and neuropsychological scales in the NDD-CKD and DD-CKD patients. *, P<0.05. γ, normalized clustering coefficient; AVLT, auditory verbal learning test; CKD, chronic kidney disease; DD, dialysis-dependent; DSST, digital symbol substitution test; DTI-ALPS, diffusion tensor image analysis with the perivascular space; MoCA, Montreal Cognitive Assessment (Beijing version); NDD, non-dialysis dependent; TMT, trail making test.

To further explore the associations, we included scales and correlation results in multiple linear regression analyses, controlling for the variables of age, gender, and education years. The results revealed that σ (β=0.494, R²=0.131, P=0.004) and Eg (β=0.337, R²=0.131, P=0.048) were independent influencing factors on the HAM-A scores, and σ (β=0.365, R²=0.139, P=0.007) and the DTI-ALPS index (β=0.282, R²=0.139, P=0.035) were independent influencing factors on the HAM-D scores. In addition, we observed that Eg (β=0.274, R²=0.075, P=0.032) was an independent influencing factor on Rey’s AVLT accuracy, and Eloc (β=0.467, R²=0.218, P<0.001) was associated with the DSST scores (Table 3). The results of the mediation analysis (Figure 5) revealed that σ partially mediated the effect of the DTI-ALPS index on the HAM-D scores (mediation effect: −5.0404); σ fully mediated the effect of the CPV on the HAM-D scores (mediation effect: 0.0013); and σ fully mediated the effect of the DTI-ALPS index on the HAM-A scores (mediation effect: −5.9291).

Table 3. The results of the multiple linear regression analysis.

Independent vs. dependent variable	R2	β	B (95% CI)	P value
σ vs. HAM-A	0.131	0.494	10.655 (3.451, 17.859)	0.004
Eg vs. HAM-A	0.131	0.337	199.314 (1.487, 397.141)	0.048
σ vs. HAM-D	0.139	0.365	6.307 (1.787, 10.827)	0.007
DTI-ALPS index vs. HAM-D	0.139	0.282	10.813 (0.781, 20.844)	0.035
Eg vs. AVLT accuracy	0.075	0.274	220.217 (19.079, 421.356)	0.032
Eloc vs. DSST	0.218	0.467	276.502 (133.666, 419.339)	<0.001

AVLT, auditory verbal learning test; B, unstandardized regression coefficient; β, standardized regression coefficient; CI, confidence interval; DSST, digital symbol substitution test; DTI-ALPS, diffusion tensor image analysis with the perivascular space; Eg, global efficiency; Eloc, local efficiency; HAM-A, Hamilton anxiety scale; HAM-D, Hamilton depression scale.

Figure 5.

Results of mediation analyses. Small-worldness (σ) partially mediated the effect of the DTI-ALPS index on the HAM-D (mediation effect: −5.0404) (A); σ fully mediated the effect of choroid plexus volume on the HAM-D (mediation effect: 0.0013) (B); σ fully mediated the effect of the DTI-ALPS index on the HAM-A (mediation effect: −5.9291) (C). DTI-ALPS, diffusion tensor image analysis with the perivascular space; HAM-A, Hamilton anxiety scale; HAM-D, Hamilton depression scale.

Discussion

This study compared the differences in the structural network topology properties, DTI-ALPS index, and CPV among DD-CKD, NDD-CKD, and HC groups. The main findings revealed abnormal segregation and integration in the structural network of the CKD patients compared to that of the HCs, with more severe impairments in the DD-CKD patients than the NDD-CKD patients. The CKD patients had a significantly lower DTI-ALPS index and a significantly higher CPV than the HCs, indicating impaired glymphatic system function. The correlation and regression analyses showed that these indicators were associated with the patients’ levels of cognition, language, anxiety, and depression. The mediation analysis revealed that σ mediated the effect of the DTI-ALPS index and the CPV on anxiety and depression levels.

The dysfunction of the glymphatic system in CKD may have various causes. In the adult population, vascular factors and uremic neurotoxins associated with CKD may have a large effect on neurological abnormalities (7). CKD patients commonly have a high burden of vascular disease, including atherosclerosis and endothelial dysfunction (39), which affects the permeability of the blood-brain barrier (BBB), which in turn leads to glymphatic system dysfunction. The impaired clearance of uremic toxins, such as high levels of inflammatory factors like uric acid, may lead to neurotoxicity and apoptosis (40-42). Some uremic toxins cross the BBB relatively easily, and have deleterious effects on endothelial cells and blood vessels (7). In addition, hypertension and diabetes are both common causes and complications of CKD (43-45). In the present study, the percentage of CKD patients with hypertension and diabetes mellitus was quite high. Elevated systolic blood pressure and blood pressure variability are independent risk factors for PVS dysfunction (46). As mentioned above, the PVS is an important conduit for the composition of the glymphatic system. It is thus likely that hypertension leads to abnormal PVS function, which in turn affects glymphatic system function. High blood glucose concentrations can lead to increased levels of growth factors, angiotensin II, and endothelin, which can lead to endothelial dysfunction (47), resulting in the disruption of the BBB and leading to glymphatic system dysfunction.

An arterial spin-labeling MRI study showed increased overall cerebral blood flow in CKD patients compared to HCs (48). This may alter the way in which substances are exchanged between blood and neurons, leading to impairments in the BBB and the glymphatic system. Additionally, sleep disorders are common in CKD patients (49). The glymphatic clearance of waste products occurs primarily during sleep (50). Thus, this is a possible explanation for the presence of glymphatic disorder in CKD. Alternately, the function of the glymphatic system may be related to aquaporin 4 (AQP-4). Aquaporins are a class of transmembrane channels that can transport water across cell membranes (51). A study has shown that aquaporins are associated with acute kidney injury and multiple CKDs (52). Moreover, renal AQP-4 expression is reduced in mice with hydronephrosis (53). These factors may lead to a decrease in CSF transport, thereby affecting the glymphatic system.

In this study, we found that the DTI-ALPS index was abnormal in both the NDD-CKD and DD-CKD patients, even after excluding neurological diseases. This indicates that glymphatic system disorders in CKD may appear before obvious abnormalities in imaging and clinical manifestations. Notably, no significant difference was found in the DTI-ALPS index between the NDD-CKD and DD-CKD patients in our study. Lee et al. also found no difference in the glymphatic system of CKD patients before and after starting dialysis (54). This result may be related to the fact that patients on hemodialysis treatment show more complex, rapid, and variable hemodynamic changes (7). It may be that regular hemodialysis treatment slows down the further decline of glymphatic system function to some extent. In addition, while we found higher diffusivity measures in the CKD patients than the HCs, this may stem from pathological changes like inflammation or edema in the CKD patients. The lower diffusivities in the HCs suggest optimal white-matter microstructure and efficient glymphatic function, while the increased diffusivities in the CKD patients could indicate axonal damage or neuroinflammation (7), contributing to the reduced DTI-ALPS index. Our findings align with previous research, such as Heo et al.’s study (19) in which the early CKD patients exhibited elevated Dyy_proj and Dzz_associ compared to the HCs, but had a reduced DTI-ASLPS index. Similarly, Liang et al. (55) found that Alzheimer’s disease patients with a decreased DTI-ASLPS index exhibited higher Dyy_proj and Dzz_associ than the HCs.

In terms of the laboratory indicators, we observed that the DD-CKD patients had lower eGFR levels than the NDD-CKD patients. This finding is consistent with the pattern of disease progression, whereby the eGFR declines as renal function deteriorates. Albumin levels serve as an indicator of the body’s nutritional status, liver function, and chronic inflammatory state. CKD patients often experience malnutrition, chronic inflammation, and protein metabolism disorders due to renal failure, leading to decreased albumin levels (56), which aligns with our findings. Interestingly, the DD-CKD patients had higher albumin levels than the NDD-CKD patients. This may be because the dialysis treatment reduces toxin accumulation and improves nutritional status, and such dialysis patients receive more meticulous clinical care and daily nutritional management than non-dialysis patients. Further, we found that the hemoglobin levels of the CKD patients were lower than those of the HCs. Anemia is highly prevalent among CKD patients and increases in severity with the progression of CKD (57). Erythropoietin (EPO)-producing cells sense tissue hypoxia and respond by increasing EPO production to counteract the hypoxic condition. Chronic renal failure disrupts this process, resulting in EPO deficiency and subsequently causing anemia, characterized by reduced circulating hemoglobin levels (58). In our study, the NDD-CKD patients had lower hemoglobin levels than the DD-CKD patients. This difference was anticipated, as NDD-CKD patients may not receive regular follow ups and may be unaware of their anemia, leading to lower treatment rates. This conclusion is also supported by observations from other studies (58,59).

There is a growing number of studies exploring neurological alterations in disease through graph theory, a completely data-driven approach. Unlike conventional seed-point based studies, network analysis focuses on local specialization and global integration in the human brain (60). The topology of small-world networks is characterized by dense local connections and fewer distant connections but shorter Lps between distant nodes. This topology simultaneously provides high information transfer efficiency at a low wiring cost, both functionally and structurally, enabling the brain to achieve an optimal balance between separated and integrated information processing. It is more widely accepted that CKD is a disease associated with brain network disruption, both structurally and functionally. Our finding that the brain networks of all the participants had small-world properties suggest that the human brain, regardless of disease state, is organized into a small-world topology to support efficient information processing.

Although the white-matter networks of the CKD patients and HCs had an economic small-world topology, the network organization was compromised in the CKD patients compared to the HCs. Cp is a measure that tests the local connectivity of the network. Eg and Eloc are quantitative measures of the local and overall efficiency of information distribution through the network, respectively. Both the CKD subgroups in our study had lower Cp, Eg, and Eloc metrics than the HCs, and these metrics were lower in the DD subgroup than the NDD subgroup, suggesting more extensive network impairment in the CKD patients, especially in those who were DD. Besides, the CKD patients, especially those in the DD subgroup, had longer characteristic Lp, suggesting a slower transmission of information through the CKD network (38). Our findings are also consistent with recent CKD network results (24,61-63). However, it is noteworthy that both the σ and γ were higher in the DD-CKD patients than the HCs, and Eloc was positively correlated with the serum creatinine levels. Considering the effects of hemodialysis treatment and the existence of compensatory mechanisms in the nervous system of the human brain, it is reasonable to speculate that this result may be attributed to the reconstruction of local nerve fibers in response to brain impairment.

The choroid plexus is involved in CSF production and may be a driver of the glymphatic system (64). Our results demonstrated that compared to the HCs, all the CKD patients had an elevated CPV. We hypothesize that this phenomenon may represent a compensatory effect in response to glymphatic system dysfunction. Specifically, the increased CPV might enhance CSF production by increasing the number of functional units, potentially offsetting the dysfunction of the glymphatic system and enhancing waste clearance capacity (65,66). Moreover, a correlation between the CPV and white-matter network metrics was found in both the NDD-CKD and DD-CKD patients, similar to the DTI-ALPS index results. We also found abnormally elevated cholesterol levels in the CKD patients, and a negative correlation between the cholesterol levels and DTI-ALPS index in the NDD-CKD patients. CKD is highly correlated with dyslipidemia (67). Liu et al. (68) used mouse models with long-term alcohol exposure and high-fat diet, and found that dyslipidemia can cause damage to the glymphatic system, which is consistent with our findings.

In our study, we found significant differences between the CKD groups and the HC group on the MoCA, TMT, and DSST scales. The regression analysis indicated that the Eloc and Eg of the white-matter network were influencing factors on the DSST and Rey’s AVLT, respectively. The MoCA was used to assess the participants’ cognitive status, while the AVLT, TMT, and DSST were used to assess the participants’ visual memory, attention, and information processing speed. Compared to the general population, CKD patients are at higher risk of developing cognitive impairment, even in the early stages of CKD, which is characterized by deficits in visual-spatial executive function, memory, and attention (7). Anxiety and depression are also very common mental health problems in CKD patients (69). Our further mediation analysis revealed that the DTI-ALPS index and CPV mediated the HAM-D scores through σ, and the DTI-ALPS index mediated HAM-A scores through σ. These findings suggest that glymphatic system dysfunction contributes to depression and anxiety in CKD patients, partly through abnormalities in the σ of brain white-matter networks. While σ is a network measure rather than a direct biological process, it reflects the brain’s organizational efficiency. Our findings imply that the structural organization of white-matter networks could influence how glymphatic dysfunction affects emotional outcomes. We hypothesize that glymphatic dysfunction may impair the segregation and integration of white-matter networks, leading to the accumulation of toxins related to depression (7) and anxiety, thereby contributing to mental disorders. Additionally, depressive states may further inhibit glymphatic clearance (8), creating a vicious cycle of cerebral white-matter damage and emotional impairment. However, the relationship between σ and glymphatic dysfunction is likely bidirectional. σ could both mediate and be affected by glymphatic dysfunction. This complex interplay suggests that the structural integrity of white-matter networks and glymphatic function jointly influence the emotional health of CKD patients and needs to be further investigated.

This study had some limitations. First, this prospective study was conducted with a relatively small sample; thus, further studies with relatively large populations need to be conducted. Second, the radial asymmetry prevalent in white-matter tracts changes with age and neurodegeneration (70), which may affect the accuracy of the ALPS index in reflecting the function of the glymphatic system. The use of age as a covariate in our study might have somewhat attenuated this effect; however, a more comprehensive assessment approach is needed in the future. Third, we did not assess patients’ neuroinflammatory indicators, sleep status, or other potential confounders (e.g., vascular risk factors, comorbid psychiatric disorders, or medication effects), which might have implications for neurologic impairment and glymphatic system assessment. Fourth, our study was only cross-sectional, and future longitudinal studies are still needed to explore the causal relationships inherent to alterations in the glymphatic system and white-matter networks. Finally, not all participants completed the cognitive assessment due to participant compliance, which might have had some effect on the results of interest.

Conclusions

We found that compared to the HCs, the CKD patients had a decreased DTI-ALPS index, and abnormal CPV and white-matter network topology. The evidence of this abnormal network topology was more pronounced in the DD-CKD patients compared to NDD-CKD patients. In addition, the patients’ DTI-ALPS index, CPV, and white-matter network parameters were correlated with cognition, language, anxiety, and depression levels. σ mediated the effect of the DTI-ALPS index and CPV on anxiety and depression levels. These findings suggest that glymphatic system dysfunction and white-matter structural damage are potential pathophysiological mechanisms leading to emotional abnormalities in CKD; however, these results require further verification in follow-up studies.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of the First Affiliated Hospital of Dalian Medical University (No. PJ-KS-KY-2021-121), and informed consent was obtained from all individual participants.

Data Sharing Statement

Available at https://qims.amegroups.com/article/view/10.21037/qims-24-1692/dss

qims-15-07-6016-dss.pdf ^{(72KB, pdf)}

DOI: 10.21037/qims-24-1692