Association of Glymphatic Dysfunction with Cognitive Performance in Type 2 Diabetes Mellitus and Prediabetes

Abstract

Introduction

The glymphatic system (GS) is a brain-wide clearance pathway whose dysfunction has been implicated in cognitive decline. This study was designed to assess GS alterations in prediabetes and type 2 diabetes mellitus (T2DM) using multiple MRI-derived metrics and to examine their associations with cognitive performance.

Methods

This cross-sectional study was conducted from October 2023 to June 2024 at Nanfang Hospital, a tertiary care center. A total of 147 participants were recruited, including 52 patients with T2DM, 44 with prediabetes, and 51 healthy controls (HC). All participants underwent brain MRI, laboratory examinations, and cognitive function assessments. GS function was evaluated using MRI-derived metrics, including choroid plexus (CP) volume, perivascular space (PVS) volume, free water (FW) index, and the diffusion tensor imaging analysis along the PVS (DTI-ALPS) index. Group differences were tested with general linear models adjusted for sex, age, and education, and associations with clinical variables were evaluated using partial correlations.

Results

Compared with HC, prediabetes showed increased CP volume and FW index, without significant changes in DTI-ALPS index or PVS volume. T2DM showed reduced DTI-ALPS index, increased CP volume, and higher FW index, with unchanged PVS volume. In both groups, lower DTI-ALPS index and higher CP volume or FW index correlated with poorer cognitive performance.

Conclusion

Prediabetes and T2DM exhibit GS alterations, including reduced DTI-ALPS index and increased CP volume and FW index, which are associated with poorer cognition. GS metrics may serve as potential neuroimaging biomarkers for cognitive impairment in this population.

Introduction

Diabetes mellitus is a prevalent metabolic disorder with a rapidly increasing global burden. Type 2 diabetes mellitus (T2DM) and prediabetes together constitute a major global health challenge, with a substantial proportion of the adult population affected by dysglycemia worldwide [1]. Both conditions have been linked to accelerated brain aging, cognitive decline [2, 3], and an increased risk of Alzheimer’s disease (AD) [4].

The glymphatic system (GS), recently identified as a macroscopic waste clearance pathway in the central nervous system, is formed by the perivascular network and facilitates cerebrospinal fluid (CSF) influx, CSF-interstitial fluid exchange, and removal of metabolic waste. These processes, which can be indirectly assessed by MRI, play a key role in eliminating brain waste products such as β-amyloid (Aβ) and in distributing glucose, lipids, amino acids, and neuromodulators [5–8]. Impaired glymphatic function is often associated with disrupted polarization of aquaporin-4 (AQP4) in astrocytic endfeet and has been implicated in the pathogenesis of neurodegenerative diseases, including AD and Parkinson’s disease [9, 10]. Recent evidence shows that selective inhibition of protein kinase RNA-like endoplasmic reticulum kinase in astrocytes restores AQP4 localization, enhances clearance of β-amyloid and tau, and markedly improves cognitive function in an AD mouse model [11].

Glymphatic dysfunction has also been implicated in the pathophysiology of diabetic encephalopathy. In T2DM, expression of AQP4, a key glymphatic component, is reduced, resulting in impaired glymphatic transport [12, 13]. A previous study using intrathecal gadolinium-based contrast agent-enhanced MRI demonstrated reduced glymphatic clearance in T2DM rats compared with controls with normal glucose metabolism [14]. However, this approach is invasive and may cause severe neurotoxic complications when the dose exceeds a certain threshold [15], limiting its application in humans. To overcome this, Taoka et al. [16] proposed a noninvasive imaging alternative by quantifying perivascular space (PVS) diffusivity using diffusion tensor imaging (DTI) to derive the DTI-ALPS index, which serves as a proxy for glymphatic function. Studies using DTI-ALPS have reported significantly lower ALPS indices in T2DM and prediabetes patients, with reductions correlating with clinical parameters such as fasting plasma glucose (FPG), HbA1c, high-density lipoprotein levels, and cognitive decline [12, 13, 16, 17]. Nevertheless, these investigations examined only a single glymphatic-related metric, whereas multiple MRI-based approaches have recently been proposed. Because no single parameter fully captures the complexity of the GS, a multimodal assessment is warranted. In addition to DTI-ALPS, several other MRI-derived metrics have been proposed to characterize different aspects of glymphatic function. PVS volume reflects macroscopic structural changes of perivascular compartments and has been linked to impaired interstitial fluid drainage in aging and neurodegenerative diseases [18]. The fractional volume of free water in white matter (FW-WM) is thought to capture alterations in extracellular water content, which may increase in the context of glymphatic dysfunction and neuroinflammation [19]. Choroid plexus (CP) volume has emerged as a potential marker of CSF production and immune surveillance within the CSF, both of which are integral to glymphatic circulation [20]. In this study, we employed noninvasive MRI to comprehensively evaluate glymphatic function in individuals with varying glucose metabolism status using multiple metrics, including PVS volume, DTI-ALPS index, FW-WM index, and CP volume, and examined their associations with cognitive performance.

Methods

Participants and Clinical Data Collection

A total of 147 participants were recruited, including 52 patients with T2DM, 44 patients with prediabetes, and 51 healthy controls (HCs). The diagnostic criteria for prediabetes and T2DM were based on the American Diabetes Association (ADA) guidelines [21]. Prediabetes was defined as FPG 5.6–6.9 mmol/L, 2-h plasma glucose during an oral glucose tolerance test (2 h OGTT) 7.8–11.0 mmol/L, or hemoglobin A1c (HbA1c) 5.7–6.4%. T2DM was diagnosed in individuals with a self-reported physician diagnosis of diabetes, current antidiabetic medication use, FPG ≥7.0 mmol/L, 2 h OGTT plasma glucose ≥11.1 mmol/L, or HbA1c≥ 6.5%.

For the prediabetes group, inclusion criteria were as follows: (1) absence of contraindications to MRI; (2) right-handedness; and (3) fulfillment of the diagnostic criteria for prediabetes. Exclusion criteria were (1) a history of neurological or psychiatric disorders such as traumatic brain injury, intracranial hemorrhage, cerebral infarction, schizophrenia, depression, or anxiety; (2) thyroid dysfunction, cardiovascular or cerebrovascular disease, or a history of corticosteroid therapy; and (3) incomplete MRI data. For the T2DM group, inclusion criteria were identical to items (1) and (2) for the prediabetes group, with the additional requirement of meeting the diagnostic criteria for T2DM. Exclusion criteria were the same as those applied to the prediabetes group. During the same period, HCs were recruited from the community. Eligibility required both FPG <6.1 mmol/L and HbA1c <6.5%. Exclusion criteria were identical to those applied to the patient groups.

All participants underwent general clinical data collection and completed the Montreal Cognitive Assessment (MoCA) prior to MRI scanning. After fasting for more than 8 h overnight, venous blood samples were obtained in the morning to measure FPG, HbA1c, 2 h OGTT glucose, total cholesterol, triglycerides, low-density lipoprotein cholesterol, and high-density lipoprotein cholesterol. The MoCA was administered in a quiet room by a physician trained in psychological assessment to evaluate participants’ cognitive function.

MRI Acquisition

For each subject, 3D 1.0-mm-isotropic T1-and T2-weighted (T1w, T2w), and diffusion tensor imaging (DTI), were acquired on a 3.0T scanner (Signa Architect 3.0T, GE, USA). DTI parameters included 32 non-collinear directions with b = 1,000 s/mm², a repetition time of 7,000 ms, an echo time of 81.6 ms, and a slice thickness of 2.0 mm without gap.

The main data processing steps are illustrated in Figure 1. Whole-brain segmentation was performed on T1w images using the FastSurfer toolbox [22]. DTI data were preprocessed with denoising, eddy current, and head motion correction. PVS Segmentation: T2w images were registered to T1w, and enhanced PVS images were calculated as the T1w/T2w ratio. PVS masks were segmented using a deep learning-based method, named as SHIVA_PVS [23]. CP segmentation: ventricular and CP masks from FastSurfer were merged, and a Gaussian Mixture Model clustered voxels based on T1w intensity to segment the choroid plexus [24]. Total, left, and right choroid plexus volumes (CP_T, CP_L, CP_R) were extracted. ALPS index: Diffusion data were fitted to a DTI model and normalized to MNI space by registering T1w images to a template. ROIs were automatically extracted from the registered images, and diffusivities within ROIs were used to calculate the ALPS index [25]. FW-WM Index: Diffusion data were fitted to a regularized bi-tensor model using the Dipy toolbox (https://dipy.org/). The mean FW fraction within the white matter mask was extracted as the FW-WM index [26].

Fig. 1.

The main data processing steps to calculated GS indicators.

Statistical Analysis

Statistical analyses of clinical data and MoCA scores among the three groups were performed using SPSS Statistics version 26.0 (IBM Corp., Armonk, NY, USA). The normality of distributions and homogeneity of variances were assessed using the Shapiro-Wilk test and Levene’s test, respectively. For continuous variables with a normal distribution, one-way analysis of variance was applied, and results were expressed as mean ± standard deviation. Post hoc multiple comparisons were conducted using Tukey’s test when the assumption of homogeneity of variance was met, and the Games-Howell test when it was not. For continuous variables with a non-normal distribution, the Kruskal-Wallis rank-sum test was used, and results were expressed as median (first quartile-third quartile), i.e., M (Q1, Q3). Post hoc pairwise comparisons were performed using Dunn’s test. Categorical variables were analyzed using Fisher’s exact test, and results were presented as frequencies and percentages. These post hoc procedures were applied to clinical variables and MoCA scores.

All GS indices were compared between HC, prediabetes, and T2DM groups using general linear models, with sex, age, and years of education included as covariates. Post hoc pairwise comparisons for GS indices were adjusted using Bonferroni correction to account for multiple comparisons. Correlations with clinical indices were assessed using Pearson’s or Spearman’s partial correlation analysis. All statistical tests were two-tailed, and a p value <0.05 was considered statistically significant.

Results

Participant Characteristics and Clinical Characteristics

Table 1 summarizes the demographic, clinical, and cognitive characteristics of the participants. A total of 147 individuals were enrolled, comprising 51 HCs, 52 patients with T2DM, and 44 with prediabetes. The mean age was 33.3 ± 7.9 years, and 48.3% were female. Regarding cognitive performance, both the prediabetes and T2DM groups exhibited lower MoCA scores compared with the HC group (p < 0.05), whereas no significant difference was observed between the prediabetes and T2DM groups (p = 0.057).

Table 1.

Demographic and clinical characteristics in the PDM, T2DM, and HC groups

​	HC (n = 51)	PDM (n = 44)	T2DM (n = 52)	F/t/χ2	p values
Age, years	30.0 (25.0, 35.5)	32.5 (28.0, 38.0)	35.0 (28.8, 40.0)	6.952	0.031
Gender, M/F	20/31	19/25	37/15	12.343	0.002
Education, years	16.0 (15.0, 16.0)	15.0 (12.0, 16.0)	15.0 (12.0, 16.0)	11.342	0.003
BMI, kg/m2	23.1 (21.4, 27.0)	30.6 (29.0, 32.8)	30.1 (27.3, 34.1)	56.572	<0.001
WHR	0.85 (0.80, 0.90)	0.92 (0.87, 0.96)	0.98 (0.95, 1.01)	53.916	<0.001
FPG, mmol/L	5.11 (4.84, 5.27)	5.01 (4.88, 5.38)	7.21 (5.98, 9.57)	73.972	<0.001
HbA1c, %	5.50 (5.35, 5.59)	5.97 (5.87, 6.13)	9.22 (6.69, 11.2)	105.775	<0.001
TC, mmol/L	4.89 (4.23, 5.45)	5.20 (4.45, 5.58)	5.21 (4.79, 5.76)	6.238	0.044
TG, mmol/L	1.11 (0.84, 1.79)	1.75 (1.42, 2.57)	2.17 (1.79, 3.42)	34.338	<0.001
HDL, mmol/L	1.41 (1.10, 1.56)	1.17 (1.01, 1.25)	0.99 (0.86, 1.08)	35.943	<0.001
LDL, mmol/L	3.16±0.85	3.54±0.94	3.52±0.80	2.971	0.054
MoCA	28.0 (27.5, 29.0)	27.0 (25.0, 28.0)	26.5 (26.0, 28.0)	27.331	<0.001

Continuous variable data that conform to a normal distribution are expressed as mean ± standard deviation, while continuous variable data that do not conform to a normal distribution are expressed as median (first quartile, third quartile). Categorical variables are expressed as frequencies.

BMI, body mass index; FPG, fasting plasma glucose; HbA1c, hemoglobin A1c; HC, healthy control group; HDL, high-density lipoprotein; LDL, low-density lipoprotein; MoCA, Montreal Cognitive Assessment; OGTT, oral glucose tolerance test; PDM, prediabetes mellitus; T2DM, type 2 diabetes mellitus; TC, total cholesterol; TG, triglycerides; WHR, waist-to-hip ratio.

GS Indices Measurements

As shown in Figure 2 and Table 2, significant differences in GS indices were observed. After Bonferroni correction, no significant differences in left ALPS (ALPS_L), right ALPS (ALPS_R), or mean bilateral ALPS (ALPS_M) values were found between the HC and prediabetes groups (all p > 0.05). The T2DM group showed significantly lower ALPS_L, ALPS_R, and ALPS_M values compared with both the HC and prediabetes groups (all p < 0.001; Fig. 2a–c). Both the prediabetes group (all p < 0.001) and the T2DM group (P_{CP_L} = 0.001, P_{CP_R} = 0.002, P_{CP_T} <0.001) had significantly larger volumes of the left choroid plexus (CP_L), right choroid plexus (CP_R), and total CP (CP_T) than the HC group (Fig. 2f), with no significant differences between the prediabetes and T2DM groups (all p > 0.05). FW-WM values were higher in both the prediabetes (p < 0.001) and T2DM (p = 0.008) groups than in HCs, whereas no significant difference was found between prediabetes and T2DM groups (p = 0.886; Fig. 2h). No significant group differences were observed in PVS indices (Fig. 2i).

Fig. 2.

Comparison of glymphatic system (GS) indicators among the prediabetes group, T2DM group, and HC group. a Left analysis along the perivascular space(ALPS_L). b Right analysis along the perivascular space(ALPS_R). c Mean analysis along the perivascular space (ALPS_M). d Left choroid plexus volume (CP_L). e Right choroid plexus volume (CP_R). f Total choroid plexus volume (CP_T). g Fractional water content in white matter (FW-WM). Normally distributed data are shown as bar plots (mean ± SD), and non-normally distributed data as violin plots (median, Q1–Q3). SD, standard deviation.

Table 2.

GS indexes in the PDM, T2DM, and HC groups

​	HC	PDM	T2DM	FWE-corrected p value
				HC vs. PDM	HC vs. T2DM	PDM vs. T2DM
ALPS_L	1.66 (1.56, 1.73)	1.58 (1.54, 1.67)	1.50 (1.43, 1.59)	0.498	<0.001	<0.001
ALPS_R	1.54 (1.47, 1.59)	1.53 (1.48, 1.57)	1.40 (1.34, 1.48)	1.000	<0.001	<0.001
ALPS_M	1.59 (1.53, 1.66)	1.55 (1.53, 1.60)	1.46 (1.40, 1.52)	1.000	<0.001	<0.001
CP_L, mm3	651±142	782±168	847±163	<0.001	0.155	0.001
CP_R, mm3	655±130	769±167	828±149	<0.001	0.002	0.398
CP_T, mm3	1,336 (1,178, 1,449)	1,505 (1,293, 1727)	1,649 (1,500, 1,858)	<0.001	<0.001	0.106
FW-WM	0.15 (0.14, 0.15)	0.15 (0.15, 0.16)	0.15 (0.15, 0.16)	<0.001	0.008	0.886
PVS, mm3	989 (532, 2,148)	891 (480, 1,807)	948 (651, 1,796)	0.969	1.000	0.188

Continuous variable data that conform to a normal distribution are expressed as mean ± standard deviation, while continuous variable data that do not conform to a normal distribution are expressed as median (first quartile, third quartile).

ALPS_L, left analysis along the perivascular space; ALPS_R, right analysis along the perivascular space; ALPS_M, mean analysis along the perivascular space; CP_L, left choroid plexus volume; CP_R, right choroid plexus volume; CP_T, total choroid plexus volume; FW-WM, fractional volume of free water in white matter; FWE-corrected p, family-wise error-corrected p value; HC, healthy control; PDM, prediabetes mellitus; PVS, perivascular space volume; T2DM, type 2 diabetes mellitus.

Correlation between the GS Indices and Clinical Characteristics

To investigate whether alterations in GS indices were associated with cognitive performance, partial correlation analyses were conducted between GS indices and MoCA scores, adjusting for age, sex, and years of education. As shown in Figure 3, within the prediabetes group, both ALPS_L and ALPS_M were positively correlated with MoCA scores (ALPS_L: r = 0.380, p = 0.016; ALPS_M: r = 0.361, p = 0.022), whereas ALPS_R showed no significant association (r = 0.155, p = 0.340) (Fig. 3a–c). Similarly, in the T2DM group, ALPS_L and ALPS_M were positively correlated with MoCA scores (ALPS_L: r = 0.580, p < 0.001; ALPS_M: r = 0.413, p = 0.004), while ALPS_R remained nonsignificant (r = 0.080, p = 0.591) (Fig. 3d–f).

Fig. 3.

Associations between glymphatic system (GS) indicators and MoCA in prediabetes and T2DM. a–c Correlations of ALPS_L, ALPS_R, and ALPS_M with MoCA scores in the prediabetes group. d–f Correlations of ALPS_L, ALPS_R, and ALPS_M with MoCA scores in the T2DM group. g–i Correlations of CP_L, CP_R, and CP_T volumes with MoCA scores in the prediabetes group. j–l Correlations of CP_L, CP_R, and CP_T volumes with MoCA scores in the T2DM group. m Correlation of FW-WM with MoCA scores in the prediabetes group. n Correlation of FW-WM with MoCA scores in the T2DM group.

For CP volume, in the prediabetes group, CP_L and CP_T volumes were inversely correlated with MoCA scores (CP_L: r = −0.510, p < 0.001; CP_T: r = −0.393, p = 0.011), whereas CP_R was not significantly associated (r = −0.051, p = 0.750) (Fig. 3g–i). In the T2DM group, CP_T volume showed an inverse correlation with MoCA scores (r = −0.295, p = 0.039), while CP_L and CP_R were not significantly related (CP_L: r = −0.262, p = 0.069; CP_R: r = −0.195, p = 0.179) (Fig. 3j–l).

For FW-WM, no significant correlation with MoCA scores was observed in the prediabetes group (r = −0.151, p = 0.345) (Fig. 3m). In contrast, in the T2DM group, FW-WM volume was inversely correlated with MoCA scores (r = −0.436, p = 0.002) (Fig. 3n).

Discussion

In this study, we employed noninvasive MRI techniques to comprehensively assess GS function in individuals with prediabetes and T2DM compared with HCs, using four complementary measures: the DTI-ALPS index, PVS volume, CP volume, and FW-WM index. We further examined the associations between these GS indices and cognitive function. The principal findings were as follows: (1) both prediabetes and T2DM groups exhibited significantly lower MoCA scores compared with HCs, indicating poorer cognitive performance in these populations; (2) GS indices in both prediabetes and T2DM groups showed significant differences from HCs, indicating GS alterations associated with dysglycemia; and (3) in both prediabetes and T2DM groups, reduced ALPS indices together with increased CP volume and FW-WM were associated with poorer cognitive performance.

We observed that T2DM patients had significantly reduced ALPS indices (left, right, and mean) compared with both HCs and prediabetes patients. This finding is consistent with prior work by Yang et al. [27] and Tuerxun et al. [12], who also reported reduced ALPS indices in T2DM patients using DTI-ALPS. Experimental evidence further supports these findings: Jiang et al. [14] demonstrated that diabetic rodent models exhibited greater perivascular tracer retention, particularly within the hippocampus, accompanied by higher residual tracer concentrations and delayed clearance, indicative of impaired glymphatic function. Our results align with these observations, supporting the notion that T2DM adversely impacts glymphatic function. In contrast, no significant differences in ALPS indices were detected between prediabetes and HC groups, consistent with Tian et al. [13], though differing from the findings of Tuerxun et al. [12]. The discrepancy may be attributable to differences in participant age, as their prediabetes cohort (mean age 72 years) was substantially older than ours (mean age 33 years). Since age has been shown to affect ALPS indices [28], this factor may partly explain the inconsistency. Furthermore, compared with these two studies, our study included a larger sample size. In addition, both of those studies relied on manually placed ROIs. Given that the DTI-ALPS index is influenced by the imaging plane [29], manual placement may introduce inter-subject variability. In the present study, we refined the approach by defining ROIs on a standard template. Compared with the manual method, this approach demonstrated higher inter- and intra-rater intraclass correlation coefficients, thereby enhancing the reliability of our findings. Taoka et al. [16] have noted that ALPS evaluation confined to specific regions may not fully capture global GS function, as the technique reflects only a portion of the system’s diverse activities.

To provide a more comprehensive assessment, we additionally evaluated CP volume, FW-WM, and PVS volume, indices that have been applied in other conditions such as obstructive sleep apnea [30] and chronic neurodegenerative diseases [26], though rarely in prediabetes or T2DM. Our results showed that both prediabetes and T2DM groups exhibited increased CP volume and FW-WM compared with HCs, with no significant differences between prediabetes and T2DM. These alterations may reflect the cumulative effects of chronic hyperglycemia, systemic low-grade inflammation, and insulin resistance [31, 32]. By contrast, no significant group differences were observed in PVS volume. This finding may be attributable to several factors. PVS enlargement is generally considered a cumulative and structural alteration that becomes detectable on conventional MRI only when pathological changes reach a sufficient magnitude. Methodologically, PVS volume represents a relatively coarse macroscopic measure of PVS dilation. Compared with FW-WM and CP volume, which capture more dynamic changes related to extracellular water content and CSF-associated processes, PVS volumetric measures are inherently less sensitive to subtle pathological changes. Furthermore, large-scale studies across the AD spectrum have demonstrated that PVS alterations exhibit marked regional specificity, with abnormalities predominantly observed in the centrum semiovale rather than in the basal ganglia or hippocampus [33]. This pronounced spatial heterogeneity suggests that global PVS volumetric measures may obscure subtle, regionally confined disease-related changes, thereby reducing their sensitivity in group-level comparisons.

Correlation analyses further revealed that in both prediabetes and T2DM groups, ALPS_L and ALPS_M were positively correlated with MoCA scores, whereas ALPS_R was not. Previous studies have yielded mixed findings regarding the association between DTI-ALPS and cognition: one study [17] reported a significant positive relationship in T2DM, whereas another [13] did not, possibly due to limited sample size (HC = 22, prediabetes = 20, T2DM = 22). However, a recent study focusing specifically on patients with T2DM [34] reported a significantly reduced DTI-ALPS index and demonstrated a robust association between ALPS-derived glymphatic dysfunction and cognitive performance. Importantly, this relationship was shown to be modulated by gray matter integrity, providing convergent evidence that impaired glymphatic transport is closely linked to diabetes-related cognitive decline. Similar associations between lower ALPS indices and cognitive impairment have been described in Parkinson’s disease and AD [9, 35–38]. Given that T2DM and AD share pathological features such as chronic oxidative stress, insulin resistance, and neuroinflammation [31, 32], these mechanisms may underlie the observed cognitive impairments. The lack of association between ALPS_R and MoCA in our study may reflect functional hemispheric asymmetry, potentially influenced by handedness, as all participants were right-handed. At the group level, right-handedness is generally associated with left-hemisphere dominance. Prior neuroimaging studies have demonstrated that neurovascular coupling during cognitive processing exhibits hemispheric lateralization, with more pronounced vascular regulation and dynamic responses in the dominant hemisphere [39]. Given that glymphatic transport is tightly coupled to vascular pulsatility and neurovascular regulation [40], such lateralized neurovascular mechanisms may lead to hemisphere-specific sensitivity of the DTI-ALPS index to cognitive function. Consequently, ALPS measures in the left hemisphere, which likely represents the presumed dominant hemisphere at the group level, may be more closely linked to cognition-related glymphatic alterations than those in the contralateral hemisphere.

We also found that CP_T volume and FW-WM were negatively correlated with MoCA scores in both prediabetes and T2DM groups. Enlargement of the CP volume and increased FW-WM both indicate dysfunction of the GS. Previous studies have shown that GS impairment often reduces the efficiency of brain metabolite clearance, with the accumulation of metabolic waste closely linked to cognitive decline. These findings are consistent with studies in AD and Parkinson’s disease [41, 42], where enlargement of the CP has been associated with disease progression, particularly cognitive impairment. Similarly, elevated FW-WM has also been strongly associated with cognitive decline in other neurological disorders [43–45]. Taken together, these results further support the potential role of GS dysfunction in cognitive impairment among patients with prediabetes and T2DM, suggesting that restoring or improving CSF system function may help slow or ameliorate cognitive deterioration in these populations.

This study has several limitations. First, the sample size was relatively small and drawn from a single center. Larger, multicenter studies are needed to increase generalizability and statistical power. Second, the study population was relatively young, with a mean age of approximately 33 years. Although this facilitated the detection of early glymphatic alterations associated with dysglycemia and may provide insight into early-stage pathophysiological changes, it may limit the generalizability of the findings to middle-aged and older populations, in whom diabetes-related cognitive impairment is more prevalent. Future studies with a broader age range are warranted to validate and extend our findings. Third, the cross-sectional design of this study precludes causal inference regarding the relationships among glucose metabolism, GS alterations, and cognitive performance. Future longitudinal, prospective cohort studies are, therefore, required to clarify temporal ordering and causal pathways. Fourth, cognitive function was assessed primarily using the MoCA, which provides a global index of cognition but lacks sensitivity to specific cognitive domains. Future studies incorporating more comprehensive, domain-specific neuropsychological assessments may better delineate the relationships between GS abnormalities and distinct cognitive deficits. Finally, MRI scans were performed during the daytime while participants were awake. As glymphatic function is modulated by circadian rhythms, future studies should carefully consider scan timing.

Conclusion

Patients with prediabetes and T2DM exhibited glymphatic dysfunction, as reflected by altered GS indices. Reduced ALPS indices and increased CP and FW-WM were linked to poorer cognitive performance. These findings suggest that GS indices may serve as biomarkers of glymphatic impairment and potential targets for preserving brain health in prediabetes and T2DM.

Acknowledgments

The authors thank all volunteers for their participation in this study as well as the clinicians, nurses, and technicians from Nanfang Hospital, Southern Medical University, for their invaluable contributions and dedicated support to this research.

Statement of Ethics

This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Nanfang Hospital, Southern Medical University (Approval No. NFEC-2024-409). Written informed consent was obtained from all participants.

Conflict of Interest Statement

The authors declare no conflicts of interest.

Funding Sources

This study has received funding by the National Natural Science Foundation of China (82172012, 82572175). The funder had no role in the design, data collection, data analysis, and reporting of this study.

Author Contributions

Liya Gong: conceptualization, methodology, formal analysis, data curation, and writing – original draft, review, and editing. Jing He and Ziqi Wu: validation, data curation, formal analysis, visualization, and writing – review and editing. Mingxuan Gao: investigation and writing – review and editing. Yanmei Zeng: resources and writing – review and editing. Wei Cui and Junyan Wen: software and writing – review and editing. Jingwen Luo: data curation and writing – review and editing. Ge Wen: resources, supervision, project administration, funding acquisition, and writing – review and editing.

Funding Statement

This study has received funding by the National Natural Science Foundation of China (82172012, 82572175). The funder had no role in the design, data collection, data analysis, and reporting of this study.

Data Availability Statement

Due to privacy restrictions, the data supporting the findings of this study are not publicly available but are available from the corresponding author on reasonable request.