This study was approved by the ethics committee of The First Affiliated Hospital of Guangxi Medical University (approval No. 2023-K118-01) and has been registered in the Chinese Clinical Trials Registry (registration No. Chi-CTR2300075515).

The power analyses were conducted using the pwr package in R software, factoring a significance level of P = 0.05, a target power of 0.8, and an effect size was of r = 0.4[30,31]. The power analysis for the analysis of variance indicated a minimum sample size of 64 participants to achieve a power of 80%. In this study, 64 participants who qualified the inclusion criteria were recruited, including 22 with normal glucose metabolism (NGM) as controls, 20 with prediabetes, and 22 with T2DM. The three groups were matched for sex and age. The inclusion criteria were: (1) Right-handed; (2) Number of years of education: ≥ 6; (3) No history of thyroid, heart, liver, or kidney diseases; (4) No history of neurological or psychiatric diseases such as PD and epilepsy; (5) No history of brain diseases (tumor, hemorrhage, infarction) or traumatic brain injury; and (6) No contraindications to MRI.

Each subject provided a medical history and underwent a physical examination, during which clinical data were recorded or measured by standard laboratory tests. These variables included sex, age, years of education, height, weight, body mass index (BMI), waist circumference, hip circumference, waist-to-hip ratio, fasting plasma glucose (FPG), 2-hour postprandial blood glucose (2hPG), glycosylated hemoglobin (HbA1c), fasting insulin (FINS), blood pressure, total cholesterol (TC), triglycerides, high-density lipoprotein (HDL), and low-density lipoprotein (LDL). Educational level was recorded in terms of years of education: Six years for elementary school graduation, nine years for junior high school graduation, twelve years for high school graduation or vocational high school, fifteen years for college students, sixteen years for bachelor’s degree, nineteen years for master’s degree, and twenty-two years for doctoral degree. BMI was calculated as BMI = weight (kg)/[height (m)]². Insulin resistance score was determined using insulin resistance index (HOMA-IR), which was calculated as follows: HOMA-IR = (FPG, mmol/L) × (FINS, pmol/L)/135. Normal individuals have a HOMA-IR index of 1, which may be higher than 1 as the level of insulin resistance increases.

Since the DM diagnosis and the glucose metabolism status classification were based on the 1999 World Health Organization diagnostic criteria, plasma glucose analysis was conducted only at two time points of FPG and 2hPG. The subjects were divided according to the status of glucose metabolism and the history of DM: (1) T2DM was diagnosed using established criteria based on medical history, medication use, FPG ≥ 7.0 mmol/L and 2hPG ≥ 11.1, the first diagnosis of T2DM; (2) Prediabetes group: Impaired fasting glucose (6.1 mmol/L ≤ FPG < 7.0 mmol/L and 2hPG < 7.8 mmol/L) and impaired glucose tolerance (FPG < 7.0 mmol/L and 7.8 mmol/L ≤ 2hPG < 11.1 mmol/L); (3) FPG < 6.1 mmol/L and 2hPG < 7.8 mmol/L.

The mini-mental state examination (MMSE) was used to evaluate the overall cognitive function; the MMSE score ranged from 0 to 30 points.

A 3.0T MRI scanner (MAGNETOM Prisma, Siemens Healthineers, Germany) with a 64-channel combined head and neck coil was used for the imaging of all volunteers. Conventional magnetic resonance sequences, including T1-weighted imaging, T2-weighted imaging, T2-fluid attenuated inversion recovery, and diffusion-weighted imaging (DWI), were employed to detect and characterize brain tumors, cerebral hemorrhage, and cerebral infarction. Diffusion spectrum imaging (DSI) scans were acquired using an echo planar imaging sequence, incorporating multiple b-values (TR = 3700 milliseconds, TE = 72 milliseconds, FOV = 220 mm × 220 mm, matrix = 110 × 110, slice thickness = 2 mm, number of slices = 66, and in-plane resolution = 2 mm). One hundred DWI volumes were obtained using a multiband sequence (Simultaneous Multi-Slice = 2, GRAPPA = 2). These volumes encompassed 12 b-values from b = 0 to b = 3000 seconds/mm² and 99 diverse diffusion encoding orientations.

Diffusion data with b-values ranging from 0 to 1000 seconds/mm² were separated from the acquired DSI data using a custom MATLAB code (MATLAB 2016b, The Math Works, Inc.). Within this study’s complete set of DSI data, those with b-values of 0-1000 seconds/mm² provided sufficient diffusion directions to reconstruct the DTI model. The FSL software (version 6.0.6.3, https://fsl.fmrib.ox.ac.uk/fsl/fslwiki) was used for eddy current correction, motion correction, and brain mask extraction. Additionally, component denoising and Gibbs-unringing[32] were performed using MRtrix3 (http://www.mrtrix3.org). The fractional anisotropy (FA) maps produced using FSL were subsequently converted into FA color maps using AFNI (https://afni.nimh.nih.gov/) to facilitate the direct delineation of ALPS regions of interest (ROIs) using tools such as ITK SNAP (http://www.itksnap.org). The above process was implemented using a custom Linux shell script, which functioned as a batch processor for the data of all subjects.

Figure 1 shows the comparison of the dispersion of lateral ventricle horizontal projection fibers and association fibers along the PVS. The PVS’s direction was perpendicular to the ventricle wall at the lateral ventricle level, resulting in its predominant alignment along the left and right directions (x-axis) in the axial plane. This direction is also perpendicular to the projection fibers (mainly on the z-axis) and the associated fibers (mainly on the y-axis; Figure 1A). Thus, in regions with projected/associated fibers, the diffuse nature along the x-axis represents at least partially the diffusion pattern along the perivascular space. A spherical ROI of 32 mm³ was placed in the projection fibers (blue in Figure 1B) and association fibers (green in Figure 1B)[11]. The diffusivity for each ROI was calculated for both groups along the x, y, and z axes. In the case of projection fibers, the diffusivity was observed along the ROI x-axis, denoted as Dx-projection. In contrast, it was referred to as Dx-association in the association fiber regions. The diffusion coefficients of the projection fiber y-axis (Dy-projection) and association fiber z-axis (Dz-association) were also obtained.

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Schematic illustration of the diffusion tensor image analysis method along the perivascular space. A: Schematic diagram showing the relationship between the direction of the perivascular space (gray cylinder) and the direction of the projection fibers (blue) and association fibers (green). Note that the orientation of the perivascular space is perpendicular to the projection and association fibers; B: Diffusion tensor image superimposed color shows the distribution of projection fibers (z-axis: Blue) and association fibers (y-axis: Green). Four regions of interest were placed on both sides of the projection fibers (projection region) and the association fibers (association region) to measure the diffusion coefficients in the three directions (x, y, and z axes).

The DTI-ALPS index, as proposed by Taoka et al[11], was determined by calculating the ratio of the average values of projection fibers (Dx-projection) and association fibers (Dx-association) along the x-axis to the average value of projection fibers (Dy-projection) along the y-axis and association fibers (Dz-association) along the z-axis. Mathematically, the DTI-ALPS index can be expressed as the DTI-ALPS index = mean (Dx-projection, Dx-association)/mean (Dy-projection, Dz-association).

Analysis of demographic data, inter-group comparisons, and correlation analysis were conducted using the R statistic software (version 4.2.3) (https://cran.r-project.org/). Pairwise comparisons for normally distributed variables were performed using the LSD method, while the Wilcoxon rank-sum test was employed for pairwise comparisons of non-normally distributed variables. Differences between the prediabetes group, the T2DM group, and the NGM group were assessed using one-way analysis of variance. The paired t-test was used to analyze the lateralization effect on DTI-ALPS indices within each group. Pearson’s or Spearman’s partial correlation analysis was used to assess the correlation between the DTI-ALPS index and clinical features. P values < 0.05 were considered indicative of statistical significance.

The demographic and clinical characteristics of participants in the three groups are summarized in Table 1. There were significant differences between the three groups with respect to FPG, 2hPG, HOMA-IR, and HbA1c levels (P < 0.01 after Bonferroni correction). However, there were no significant differences with respect to the other clinical features.

There was no significant difference in the left-side (P = 0.680; Figure 2A), right-side (P = 0.614; Figure 2B), and mean (P = 0.585; Figure 2C) DTI-ALPS index between the prediabetes group and the NGM group. Compared with the NGM group, the left-side (P = 0.025, 95%CI: 1.646 to 1.918 vs 1.691 to 1.951; Figure 2A), right-side (P = 0.002, 95%CI: 1.655 to 1.943 vs 1.712 to 1.987; Figure 2B), mean (P = 0.001, 95%CI: 1.672 to 1.909 vs 1.722 to 1.949; Figure 2C) DTI-ALPS index were significantly reduced in the T2DM group. Compared with the prediabetes group, there was no significant difference in the left-side DTI-ALPS index of the T2DM group (P = 0.072; Figure 2A), but the right-side (P = 0.011, 95%CI: 1.401 to 1.676 vs 1.655 to 1.943; Figure 2B), and mean (P = 0.010, 95%CI: 1.461 to 1.687 vs 1.672 to 1.909; Figure 2C) DTI-ALPS index were significantly reduced.

Open in New Tab   Full Size Figure   Download Figure

Figure 2 The left-side\right-side\mean diffusion tensor image analysis method along the perivascular space index was evaluated. A: The diffusion tensor image analysis method along the perivascular space (DTI-ALPS) index of the left-side cerebral hemisphere in patients with type 2 diabetes mellitus (T2DM) was significantly lower than in the normal glucose metabolism (NGM) group; B: The DTI-ALPS index of the right cerebral hemisphere in patients with T2DM was significantly lower than that in the prediabetes group and the NGM group; C: The mean DTI-ALPS index of both cerebral hemispheres in patients with T2DM was significantly higher in the prediabetes and NGM groups. T2DM: Type 2 diabetes mellitus; NGM: Normal glucose metabolism; DTI-ALPS: Diffusion tensor image analysis method along the perivascular space; L-DTI-ALPS: Diffusion tensor image analysis method along the perivascular space index of the left cerebral hemisphere in prediabetes and type 2 diabetes mellitus; R-DTI-ALPS: Diffusion tensor image analysis method along the perivascular space index of the right cerebral hemisphere of prediabetes and type 2 diabetes mellitus; Mean-DTI-ALPS: Mean Diffusion tensor image analysis method along the perivascular space index in both cerebral hemispheres.

There was no significant difference in DTI-ALPS index between the left-side and right-side basal ganglia in the NGM group (P = 0.704), prediabetes group (P = 0.849), and T2DM group (P = 0.224).

To examine the existence of cognitive impairment in the prediabetes and T2DM groups, we compared the neuropsychological scores between the three groups (Figure 3). There was no significant difference in the MMSE score between the prediabetes group and the NGM group (P = 0.956). However, the MMSE score of the T2DM group was significantly lower than that of the NGM group (P = 0.020, Z = -2.334) and prediabetes group (P = 0.025, Z = -2.254). After P-value correction, there was no significant decrease in MMSE score in the T2DM group compared with the NGM (P = 0.061) and prediabetes (P = 0.061) groups.

Open in New Tab   Full Size Figure   Download Figure

Figure 3 The mini-mental state examination score in the type 2 diabetes mellitus group was significantly lower than that in the normal glucose metabolism and prediabetic groups. T2DM: Type 2 diabetes mellitus; NGM: Normal glucose metabolism; MMSE: The mini-mental state examination.

To investigate whether changes in the DTI-ALPS index were associated with clinical features, we assessed the correlation between the DTI-ALPS index and clinical features in the prediabetic group and T2DM group (Table 2).

After adjusting for sex, there was no significant correlation between the left-side (r = -0.197, P = 0.418), right-side (r = -0.336, P = 0.160), and mean DTI-ALPS (r = -0.350, P = 0.142) index and the MMSE score in the prediabetes group, and there was no significant correlation between the left-side (r = -0.117, P = 0.614), right-side (r = 0.106, P = 0.646), and mean DTI-ALPS (r = 0.065, P = 0.781) index and the MMSE score in the T2DM group. However, in the prediabetes group, after adjusting for sex, there was a positive correlation between the left-side DTI-ALPS index and 2hPG (r = 0.582, P = 0.009, 95%CI: 0.192 to 0.830; Figure 4A). In contrast, negative correlations were observed with TC (r = -0.567, P = 0.011, 95%CI: -0.831 to -0.152; Figure 4B) and LDL (r = -0.545, P = 0.016, 95%CI: -0.720 to 0.235; Figure 4C). In the prediabetes group, the right-side DTI-ALPS index exhibited a negative correlation with waist-to-hip ratio (r = -0.495, P = 0.031, 95%CI: -0.784 to 0.017; Figure 4D) and HbA1c level (r = -0.662, P = 0.002, 95%CI: -0.867 to -0.122; Figure 4E). Additionally, the mean DTI-ALPS index in the prediabetes group showed a negative correlations with waist-to-hip ratio (r = -0.467, P = 0.044, 95%CI: -0.770 to 0.067; Figure 4F) and HbA1c level (r = -0.725, P = 0.0004, 95%CI: -0.886 to -0.398; Figure 4G), as well as positive correlation with 2hPG (r = 0.477, P = 0.039, 95%CI: -0.103 to 0.797; Figure 4H). In the T2DM group, the left-side DTI-ALPS index exhibited a positive correlation with height (r = 0.553, P = 0.009, 95%CI: -0.142 to 0.822; Figure 4I), while it demonstrated a negative correlation with HDL level (r = -0.548, P = 0.010, 95%CI: -0.792 to -0.109; Figure 4J). Similarly, the right DTI-ALPS index in the T2DM group displayed a positive correlation with height (r = 0.621, P = 0.003, 95%CI: 0.096 to 0.828; Figure 4K). Furthermore, the mean DTI-ALPS index of the T2DM group was positively correlated with height (r = 0.651, P = 0.001, 95%CI: 0.016 to 0.853; Figure 4L) and negatively correlated with HDL (r = -0.544, P = 0.010, 95%CI: -0.802 to -0.013; Figure 4M).

Open in New Tab   Full Size Figure   Download Figure

Figure 4 The correlation between the diffusion tensor image analysis method along the perivascular space index and other clinical features. A: The left-side diffusion tensor image analysis method along the perivascular space (DTI-ALPS) index in the prediabetes group was positively correlated with 2-hour postprandial glucose (2hPG) (r = 0. 582, P = 0.009, 95%CI: 0.192 to 0.830); B and C:The left-side DTI-ALPS index in the prediabetes group was negatively correlated with total cholesterol (r = -0.567, P = 0.011, 95%CI: -0.831 to -0.152) and low-density lipoprotein (r = -0.545, P = 0.016, 95%CI: -0.720 to 0.235); D and E: The right-side DTI-ALPS index in the prediabetes group was negatively correlated with waist-to-hip ratio (r = -0. 495, P = 0.031, 95%CI: -0.784 to 0.017) and glycosylated hemoglobin (HbA1c) (r = -0.662, P = 0.002, 95%CI: -0.867 to -0.122); F and G: The mean DTI-ALPS index in the prediabetes group negatively correlated with waist-to-hip ratio (r = -0.467, P = 0.044, 95%CI: -0.770 to 0.067) and HbA1c (r = -0.725, P = 0.0004, 95%CI: -0.886 to -0.398); H:The mean DTI-ALPS index in the prediabetes group was positively correlated with 2hPG (r = 0.477, P = 0.039, 95%CI: -0.103 to 0.797); I: The left-side DTI-ALPS index in the type 2 diabetes mellitus (T2DM) group was positively correlated with height (r = 0.553, P = 0.009, 95%CI: -0.142 to 0.822); J: The left-side DTI-ALPS index in the T2DM group was negatively correlated with high-density lipoprotein (HDL) (r = -0.548, P = 0.010, 95%CI: -0.792 to -0.109); K: The right-side DTI-ALPS index in the T2DM group was positively correlated with height (r = 0.621, P = 0.003, 95%CI: 0.096 to 0.828); L: The mean DTI-ALPS index of the T2DM group was positively correlated with height (r = 0.651, P = 0.001, 95%CI: 0.016 to 0.853); M: The mean DTI-ALPS index in the T2DM group negatively correlated with HDL (r = -0.544, P = 0.010, 95%CI: -0.802 to -0.013). T2DM: Type 2 diabetes mellitus; NGM: Normal glucose metabolism; DTI-ALPS: Diffusion tensor image analysis method along the perivascular space; L-DTI-ALPS: Diffusion tensor image analysis method along the perivascular space index of the left cerebral hemisphere in prediabetes and type 2 diabetes mellitus; R-DTI-ALPS: Diffusion tensor image analysis method along the perivascular space index of the right cerebral hemisphere of prediabetes and type 2 diabetes mellitus; Mean-DTI-ALPS: Mean Diffusion tensor image analysis method along the perivascular space index in both cerebral hemispheres; 2hPG: 2-hour postprandial glucose; TC: Total cholesterol; LDL: Low-density lipoprotein; HbA1c: Glycosylated hemoglobin; HDL: High-density lipoprotein.

Chronic hyperglycemia can potentially modify the water diffusion characteristics in brain tissue. Furthermore, chronic hyperglycemia over an extended period can impact synapse formation, intensify the body’s response to oxidative stress, and lead to the accumulation of advanced glycation end products. Additionally, it can expedite the development of atherosclerotic plaques, which in turn can cause neuronal and vascular injury. These effects increase the incidence of peripheral vascular complications[17,34].

Furthermore, the deposition and elimination of Aβ and tau proteins can potentially impact the properties of tissue water diffusion[35]. This phenomenon may potentially account for the alteration in the DTI-ALPS index induced by chronic hyperglycemia. Notably, we found no significant decrease in the DTI-ALPS index among individuals with prediabetes compared to those with NGM. However, the DTI-ALPS index in the T2DM group was significantly decreased compared to the NGM group. Additionally, the right-side and mean DTI-ALPS index in the T2DM group were significantly decreased compared to the prediabetic group, which is consistent with the findings of Yang et al[19]. The decline in the DTI-ALPS index may be ascribed to the association between T2DM and atherosclerosis[36-38], which promotes the aggregation of Aβ and tau proteins. Hence, glymphatic dysfunction primarily manifests in individuals with T2DM across various glucose metabolism states. Nevertheless, the precise mechanisms responsible for glymphatic impairment in different glucose metabolic states are not clear. Potential factors contributing to this impairment include reduced arterial pulses, heightened PVS, modified expression and distribution of AQP4, or astrocyte swelling[19,39]. Additional research is required to ascertain the impact of various physiological lymphokines, including meningeal glymphatic and venous outflow and CSF production rates, on glymphatic function following an ischemic stroke (IS)[40]. Moreover, further investigations are required to identify potential glymphatic impairments in different stages of impaired glucose metabolism and establish therapeutic interventions for addressing such impairments.

In contrast to prior investigations, the present study did not observe exponential lateralization of DTI-ALPS among the three groups[17,19,41]. Due to the limited sample sizes in various studies and the heterogeneity of the diseases under examination, ascertaining the lateralizing effects in the glymphatic system of the human brain is challenging. Fur-thermore, the thickness of the superior longitudinal fasciculus is influenced by chirality and dominant cerebral flanks[42]. These factors can potentially impact the measurement of the DTI-ALPS index, thereby compromising the accuracy of the obtained results. Due caution should be exercised when addressing present manifestations of laterality. Further studies are required to clarify the presence and underlying factors of laterality within the glymphatic system.

This study observed a notable decline in MMSE scores in the T2DM group compared to the NGM and prediabetes groups. Although the difference was insignificant after post hoc correction of the P values, it showed a trend. While diabetes is widely recognized as a risk factor for dementia, the potential effect of prediabetes in causing cognitive decline has been debated[7,43,44]. This study did not identify any significant difference between the prediabetic and NGM groups in this respect. However, given the limited sample size and the cross-sectional study design, a multicenter study with a larger sample size is required to validate the findings regarding the potential decline in cognitive function among individuals with prediabetes. It is worth mentioning that our findings align with previous research in the T2DM phase[45-47]. Glucose dysregulation is implicated in the onset and advancement of neurological and cognitive complications associated with diabetes[48]. Nevertheless, various longitudinal studies investigating the impact of diabetes-related cognitive decline have yielded inconclusive results[44,49,50]. These studies were limited by their relatively short follow-up periods and incomplete prediabetes or glycemic control evaluation, potentially accounting for the discrepancies observed in the outcomes above.

Previous studies have demonstrated a positive correlation between the DTI-ALPS index and MMSE scores[17,18,51]. In the present study, although MMSE scores were significantly lower in the T2DM group than in the NGM and prediabetes groups, no correlation was found between the left-side, right-side, and mean DTI-ALPS indices of the prediabetes and T2DM groups and MMSE scores. This may be attributed to this small sample size and cross-sectional study design. Multicenter studies with a larger sample size are required to validate the correlation between the DTI-ALPS index and MMSE among diabetic patients.

The DTI-ALPS index is an approximate gauge for evaluating glymphatic function, although factors other than glymphatic dysfunction can influence its variations. In this study, the main variables impacting the DTI-ALPS index in individuals with prediabetes were 2hPG, TC, LDL, waist-to-hip ratio, and HbA1c level. Conversely, the key factors influencing the DTI-ALPS index in patients with T2DM were height and HDL level. Of note, a positive correlation was observed between the left-side DTI-ALPS index and 2hPG in the prediabetes group. This association could be attributed to the fact that the glucose concentration in CSF depends on the glucose levels in the peripheral blood[52]. The CSF is present within the perivascular brain tissue, specifically within the PVS. In individuals with prediabetes, the glymphatic system, which is responsible for clearing waste from the brain, remains functional to manage increased blood glucose levels within the perivascular inter fluid in the basal ganglia. While the DTI-ALPS index in the prediabetes group showed a positive correlation with 2hPG levels, it exhibited a negative correlation with HbA1c levels. HbA1c is an indirect measure of the average blood glucose concentration over the past three months and does not reflect short-term glycemic control[53]. Hence, despite the inverse association between the prediabetic DTI-ALPS index and 2hPG, a positive correlation was found with HbA1c. Numerous studies have indicated that elevated levels of LDL and TC increase the susceptibility to cardiovascular disease[54,55] and the development of atherosclerosis. In the prediabetes group, a negative correlation of DTI-ALPS index with LDL and TC was observed, potentially indicating a link to LDL and TC-induced arteriosclerosis. The waist-to-hip ratio is a highly sensitive indicator of fat accumulation and cardiovascular disease, surpassing the BMI’s predictive capacity[56,57]. In this study, prediabetic individuals with a greater waist-to-hip ratio exhibited a diminished DTI-ALPS index, indirectly suggesting an increased likelihood of developing glymphatic dysfunction. Furthermore, within the T2DM group, a higher DTI-ALPS index was associated with lower levels of HDL, a biomarker known for its anti-inflammatory properties and consequential health benefits[58]. However, several studies have shown that comorbidities may potentially alter the composition and functionality of HDL particles, impeding their beneficial effects and potentially leading to harmful consequences[59-63]. The present study found a positive correlation between height and the DTI-ALPS index. Given the scarcity of comparable studies and the limited sample size in this study, the association between height and the DTI-ALPS index needs to be investigated in a larger multicenter study.

One notable aspect of this study was its utilization of the DTI-ALPS index to assess the functioning of the brain’s glymphatic system among individuals with different glucose metabolism states while also investigating the clinical characteristics that impact the DTI-ALPS index. Our findings indicate that glymphatic dysfunction primarily manifests during T2DM. Based on the aforementioned preliminary findings, we posit that the DTI-ALPS index may serve as a useful marker for assessing the condition of the glymphatic system in individuals with different glucose metabolism states. Nevertheless, more robust studies with larger sample sizes are imperative to substantiate this preliminary assertion.

Some limitations of this study should be acknowledged. Due to the small sample size, the findings should be interpreted with caution. Secondly, due to multiple b-value acquisitions, the DSI acquisition method may be deemed the most optimal DTI-ALPS scanning protocol. This assertion is supported by a recent study[41,64], which demonstrated a strong correlation among DTI-ALPS values obtained through varying scan parameters, affirming the feasibility of conducting this study using our research protocol. Additionally, this study only collected 25 clinical features for correlation analysis. Further study is required to investigate the impact of drugs on the DTI-ALPS index. Moreover, future studies should include a broader range of clinical features to investigate their potential impact on DTI-ALPS.