Impact of glycemic control on brain microstructure in type 2 diabetes mellitus: insights from diffusion tensor imaging

Abstract

Background

Type 2 diabetes mellitus (T2DM) has been associated with brain microstructural alterations, potentially contributing to cognitive decline and neurodegeneration. Diffusion tensor imaging (DTI) provides a non-invasive method to assess these changes. However, the relationship between glycemic control and brain microstructural integrity remains unclear. This study aims to investigate the association between glycemic control and brain microstructural changes in T2DM using DTI.

Methods

This retrospective study included 90 participants (30 healthy controls, 60 T2DM patients) who underwent 1.5T MRI DTI at The Affiliated Shunde Hospital of Jinan University between January 2023 and May 2024. T2DM patients were categorized into well-controlled (HbA1c < 7%, n = 30) and poorly controlled (HbA1c ≥ 7%, n = 30) groups. Fractional anisotropy (FA) and apparent diffusion coefficient (ADC) values were analyzed across multiple white matter regions. Pearson’s correlation was used to assess associations between HbA1c and DTI metrics, while group differences were evaluated using Bayesian effect size estimation.

Results

HbA1c negatively correlated with ADC values in the right hippocampus (r = -0.33, p = 0.0013), suggesting a relationship between poor glycemic control and increased tissue diffusivity. A weak but significant positive correlation between HbA1c and FA in the right hippocampus (r = 0.23, p = 0.03) was observed. ADC values were higher in the poorly controlled T2DM group, indicating potential diabetes-related microstructural changes. No significant FA or ADC differences were found in other brain regions (p > 0.05).

Conclusions

Poor glycemic control in T2DM is associated with microstructural alterations in the right hippocampus, potentially reflecting early neurodegenerative processes. Longitudinal studies are needed to further investigate these findings.

Highlights

Poor glycemic control in T2DM is associated with microstructural alterations in the right hippocampus, as evidenced by significantly increased apparent diffusion coefficient (ADC) values, suggesting potential neurodegenerative changes.

HbA1c levels correlate negatively with ADC values and positively with fractional anisotropy (FA) in the right hippocampus, indicating that chronic hyperglycemia may impact white matter integrity in this region.

Diffusion tensor imaging (DTI) provides valuable insights into diabetes-related brain changes, highlighting the need for early neuroimaging assessment in T2DM patients to monitor potential microstructural damage associated with poor metabolic control.

Introduction

Type 2 diabetes mellitus (T2DM) is a prevalent metabolic disorder characterized by chronic hyperglycemia resulting from insulin resistance and/or impaired insulin secretion [1, 2]. Beyond its well-known systemic complications, T2DM has been increasingly associated with central nervous system alterations, including cognitive decline and an elevated risk of brain diseases [3]. These neurological impairments are believed to be linked to microstructural changes in the brain’s white and gray matter [4, 5].

Advancements in neuroimaging techniques, particularly diffusion tensor imaging (DTI), have provided a valuable means for in vivo assessment of brain microstructure. DTI measures the diffusion of water molecules along white matter tracts, offering key metrics such as fractional anisotropy (FA) and apparent diffusion coefficient (ADC). FA reflects the integrity and organization of neural pathways [6], while ADC quantifies the overall magnitude of water diffusion [7], providing insights into tissue microstructural integrity [8]. Studies utilizing DTI have identified widespread alterations in both FA and ADC values in individuals with T2DM, suggesting disruptions in neural connectivity and microstructural damage [9, 10]. Specifically, reduced FA and elevated ADC values have been observed in several brain regions, indicating potential white matter degeneration and increased extracellular water content, which may contribute to the cognitive deficits commonly reported in this population [11]. Notably, DTI has also been successfully employed to evaluate peripheral nerve abnormalities, demonstrating its broader clinical utility in T2DM patients beyond the central nervous system [12].

Glycemic control, typically assessed by glycated hemoglobin (HbA1c) levels, plays a crucial role in T2DM management and the prevention of peripheral complications [13, 14]. However, its impact on brain microstructure remains insufficiently explored. Chronic hyperglycemia, characteristic of poor glycemic control, may accelerate neurodegenerative processes through mechanisms such as oxidative stress, neuroinflammation, and microvascular complications, which can result in neuronal axon and myelin integrity damage [15,16,17]. While some studies have suggested that poor glycemic control exacerbates diabetes-related brain alterations, potentially accelerating neurodegeneration [18, 19], other research has found no significant associations [20]. Further exploration is required to elucidate these relationships comprehensively.

This study aims to further elucidate the relationship between glycemic control and brain microstructural integrity using DTI, providing insights into the potential neurological consequences of poor glucose regulation in T2DM patients.

Methods

Study design and participants

This retrospective study was conducted at The Affiliated Shunde Hospital of Jinan University from January 2023 to May 2024. A total of 90 participants were recruited, comprising 60 individuals diagnosed with T2DM and 30 healthy controls. Participants were eligible for inclusion if they met the following criteria: 1) Aged between 18 and 70 years. 2) No history of other endocrine disorders. 3) No other intracranial diseases that could lead to white matter alterations. And the exclusion criteria included the presence of any of the following: 1) History of known cerebrovascular diseases. 2) Severe diabetic complications. 3) Diagnosed hypertension or transient ischemic attack within past two years. 4) History of major medical conditions. 5) Severe substance abuse. 6) Contraindications for MRI examination. Participants from both groups were recruited independently based on the predefined inclusion and exclusion criteria. Although no formal individual matching procedure was employed, demographic characteristics such as age and sex were statistically compared between the two groups after recruitment. Since no statistically significant differences emerged, the groups were considered demographically comparable, reducing the likelihood of confounding due to these factors. The retrospective study was approved by the local institutional review board. Due to the retrospective study design, the need for informed consent was waived and all data were anonymized.

Table 1 Clinical characteristics of 90 participants included

The T2DM group was further stratified based on glycemic control, as determined by HbA1c levels. HbA1c levels were measured using ion-exchange high-performance liquid chromatography with the Bio-Rad D-10 Hemoglobin Analyzer (Bio-Rad Laboratories, Inc.), following standard laboratory procedures. Patients with HbA1c < 7% were classified as the well-controlled group (n = 30), whereas those with HbA1c ≥ 7% were designated as the poorly controlled group (n = 30) [21]. The control group (n = 30) consisted of healthy individuals with no history of diabetes and normal blood glucose and lipid levels. A comprehensive flowchart detailing participant enrollment, exclusion criteria, and subgroup classification is illustrated in Fig. 1.

Fig. 1

Flowchart depicting the participant enrollment and grouping process for the study

MRI data acquisition

MRI scans were performed using a Philips Multiva 1.5T superconducting MRI system (Philips Healthcare, Netherlands) equipped with a standard 16-channel head coil to ensure high-resolution imaging of brain microstructure. Prior to scanning, participants were thoroughly briefed on the examination procedure, estimated duration, and safety protocols to alleviate anxiety and ensure compliance. To minimize motion artifacts, earplugs were provided to reduce scanner noise, and foam padding was placed between the participant’s head and the head coil to restrict involuntary movements. Participants were instructed to remain still throughout the scan and to stay awake to maintain image quality.

DTI was conducted using a single-shot spin-echo echo-planar imaging sequence, optimized for assessing white matter microstructural integrity. The acquisition parameters were set as follows: repetition time (TR) = 4700 ms, echo time (TE) = 87 ms, flip angle = 90°, and b-values = 0 and 1000 s/mm² to enhance diffusion contrast. The imaging field of view (FOV) was 224 mm × 224 mm, with a matrix size of 256 × 256, a slice thickness of 4 mm, and no inter-slice gap (0 mm). Diffusion-weighted images were acquired along 15 non-collinear gradient directions, ensuring robust diffusion anisotropy calculations for FA and ADC analysis. These imaging parameters were chosen to optimize signal-to-noise ratio and facilitate accurate assessment of white matter alterations in T2DM patients.

DTI image processing

DTI data were processed using the Philips Extended Workstation with the Function Tool software for post-processing. Following automatic system transmission of the raw DTI images to the workstation, several preprocessing steps were applied to enhance data quality. These included echo-planar imaging (EPI) distortion correction, artifact removal, and brain extraction to exclude non-brain tissues while ensuring full-brain coverage. The post-processed data were then used to generate FA grayscale maps and FA color-coded maps, which facilitated further analysis of white matter microstructural integrity.

For quantitative analysis, region of interest (ROI) placement was performed on the FA color-coded maps to extract FA and ADC values from multiple predefined brain regions. ROIs were manually placed on bilateral centrum semiovale, frontal white matter, parietal white matter, temporal white matter, occipital white matter, hippocampal white matter, internal capsule, and the genu and splenium of the corpus callosum. Each ROI was carefully positioned based on anatomical landmarks, ensuring that its placement was within the designated white matter structures while avoiding contamination from adjacent gray matter or cerebrospinal fluid to minimize partial volume effects. The standard ROI size was approximately (20 ± 2) mm², with minor shape adjustments applied to accommodate the orientation of fiber tracts.

FA and ADC values were automatically calculated by the system for each selected region. To improve measurement reliability, three independent measurements were taken for each region, and the final values were expressed as the mean value.

Statistical analysis

Continuous variables were reported as means ± SD or medians with IQR, depending on data distribution, while categorical variables were summarized as frequencies and percentages. Group comparisons were performed using Student’s t-test, Wilcoxon signed-rank test, or chi-square/Fisher’s exact test, as appropriate.

Associations between FA and ADC values with HbA1c levels were assessed using Pearson’s correlation analysis, with results reported as correlation coefficients (r), 95% CIs, and p values. Bayesian factor analysis (BF₀₁) and highest density intervals (HDI) provided additional probabilistic estimates. For group comparisons of ADC values, Bayesian effect size estimation was used, reporting the posterior difference (δ) with 95% CI and expected a priori evidence (ETI) based on a Cauchy prior distribution.

All tests were two-sided, with p < 0.05 considered statistically significant. Analyses were conducted using R software (version 4.3.1, R Foundation for Statistical Computing, Vienna, Austria; http://www.R-project.org).

Results

Baseline characteristics

As shown in Table 1, the mean age of participants was 60.04 ± 14.29 years, with no significant difference among groups (p = 0.112). The proportion of female participants was higher in the poorly controlled T2DM group (66.67%) compared to the well-controlled (60.00%) and control groups (43.33%), but this difference was not statistically significant (p = 0.171).

As expected, HbA1c levels differed significantly across groups (p < 0.001), with medians of 5.55% [5.03–6.10] in the control group, 6.10% [5.82–6.47] in the well-controlled T2DM group, and 8.50% [7.60–9.40] in the poorly controlled group. Similarly, fasting plasma glucose (FPG) levels were significantly higher in the poorly controlled group (8.65 [7.82–10.73] mmol/L) than in the well-controlled (5.55 [5.20–6.10] mmol/L) and control groups (5.40 [4.80–6.02] mmol/L, p < 0.001).

Regarding fractional anisotropy (FA) and apparent diffusion coefficient (ADC) values, no significant differences were observed among the groups in most brain regions (p > 0.05), except for ADC values in the right hippocampus, which were significantly higher in the poorly controlled T2DM group (p = 0.048).

Correlation between DTI imaging and HbA1c

Pearson’s correlation analysis was conducted to examine the relationships between FA and ADC values with HbA1c levels across multiple brain regions (Figs. 2 and 3). Additionally, subgroup-specific analyses were performed to further illustrate these associations, with detailed results provided in supplementary visual aids (Figures S1 to S6). Specifically, Figures S1 and S2 present the correlations between HbA1c levels and FA and ADC values, respectively, within the normal control subgroup. Figures S3 and S4 demonstrate these relationships within the well-controlled T2DM subgroup, whereas Figures S5 and S6 depict the correlations observed in the poorly controlled T2DM subgroup. These supplementary figures provide clear visualizations, helping elucidate how different degrees of glycemic control relate distinctly to brain microstructural integrity across various regions.

Fig. 2

Correlation analysis between brain FA values and HbA1c levels in various brain regions. (A-P) depict results from the left and right centrum semiovale, parietal lobe, internal capsule, corpus callosum (genu and splenium), occipital lobe, frontal lobe, temporal lobe, and hippocampus. The correlation was assessed using Pearson’s correlation coefficient (r) with corresponding 95% CI. Bayesian factor analysis (BF₀₁) and highest density intervals (HDI) were also calculated to support statistical inference. Among all regions, a significant correlation was observed in the right hippocampus (P; r = 0.23, p = 0.03), while other regions did not show statistically significant associations

Fig. 3

Correlation analysis between brain ADC values and HbA1c levels in various brain regions. (A-P) illustrate the relationships in the left and right centrum semiovale, parietal lobe, internal capsule, corpus callosum (genu and splenium), occipital lobe, frontal lobe, temporal lobe, and hippocampus. Pearson’s correlation coefficient (r) with 95% CI was computed to assess associations. BF₀₁ and HDI were also utilized for statistical inference. Notably, a significant negative correlation was found in the right hippocampus (P; r = -0.33, p = 0.0013), suggesting an association between elevated HbA1c levels and increased ADC values in this region. No significant correlations were observed in other brain regions

For FA values shown in Fig. 2, no significant correlations were observed between HbA1c and FA in most regions (p > 0.05). However, a weak but significant positive correlation was found in the right hippocampus (r = 0.23, p = 0.03), suggesting a potential association between glycemic control and microstructural integrity in this region. This observation was consistent across the three subgroups, with Figures S1, S3, and S5 highlighting that the positive relationship was most pronounced in the well-controlled group, where better glycemic regulation might help maintain white matter integrity. In contrast, the normal subgroup (Figure S1) showed no significant correlations, and the poorly-controlled subgroup (Figure S5) exhibited weaker associations.

Regarding ADC values, shown in Fig. 3, a significant negative correlation was observed in the right hippocampus (r = -0.33, p = 0.0013), indicating that higher HbA1c levels were associated with increased ADC, which may reflect microstructural alterations due to prolonged hyperglycemia. This trend was consistent across the subgroups, with Figures S2, S4, and S6 illustrating that poorly controlled T2DM participants exhibited the strongest negative correlation in the right hippocampus, indicating more pronounced neurodegenerative changes. In contrast, the well-controlled group showed a weaker but still significant correlation, while the normal subgroup did not demonstrate a substantial relationship.

Comparison across groups

To further explore the impact of glycemic control on brain microstructure, ADC values in the right hippocampus were compared across the three study groups: normal controls (n = 30), well-controlled T2DM (n = 30), and poorly controlled T2DM (n = 30) (Fig. 4).

Fig. 4

Group comparison of ADC in the right hippocampus among different groups of participants. The normal and well-controlled T2DM groups (n = 60) are compared against the poorly controlled T2DM group (n = 30). BF₀₁ and effect size estimation were performed, with the posterior difference (δ) and 95% CI displayed. The results indicate a trend toward increased ADC values in the poorly controlled T2DM group compared to the normal and well-controlled groups

Bayesian estimation of effect sizes revealed a trend toward increased ADC values in the poorly controlled T2DM group compared to the normal and well-controlled groups, suggesting greater microstructural alterations in individuals with inadequate glycemic control. The posterior difference (δ) with 95% CI and BF₀₁ further supported this trend. While the differences did not reach conventional statistical significance, the observed pattern indicates that prolonged hyperglycemia may contribute to increased diffusivity in the right hippocampus, potentially reflecting neurodegenerative changes in T2DM patients with poor metabolic control.

Discussion

This study investigated the impact of glycemic control on brain microstructure in individuals with T2DM using DTI. Our findings revealed that poorly controlled T2DM is associated with microstructural alterations in specific brain regions, particularly the right hippocampus, as evidenced by increased ADC values.

The observed elevation in ADC values in the right hippocampus of poorly controlled T2DM patients suggests increased water diffusivity, which may reflect underlying microstructural degeneration such as axonal loss, demyelination, or increased extracellular space. The hippocampus is known to be particularly vulnerable to metabolic insults, given its high energy demands and dense vascular supply. Previous studies have reported both volumetric reductions and microstructural impairments in the hippocampus among individuals with T2DM, correlating with deficits in memory and learning [22, 23]. For instance, Xiong et al. [24] identified white matter microstructural abnormalities in T2DM patients, highlighting the vulnerability of certain brain regions to diabetes-related changes.

Interestingly, our study observed a weak but statistically significant positive correlation between HbA1c levels and FA values in the right hippocampus. While this finding appears paradoxical—given that elevated HbA1c is typically associated with neurodegeneration—it may reflect more nuanced pathophysiological processes. One possible explanation is that mildly elevated glucose levels, particularly in early or compensated stages of T2DM, could sustain neuronal metabolism and preserve microstructural integrity in select brain regions [18]. This aligns with prior work suggesting that overly intensive glycemic control may be detrimental in certain contexts: for example, the ACCORD trial found that aggressive HbA1c reduction in patients with cardiovascular risk factors was associated with increased mortality and cardiovascular events [25]. Moreover, emerging evidence indicates that lower HbA1c levels may impair microvascular perfusion in susceptible patients. For instance, a recent study demonstrated that lower HbA1c was associated with reduced microcirculatory function in diabetic polyneuropathy, suggesting that tighter glucose control is not universally beneficial and may, in some cases, compromise tissue oxygenation [26]. In contrast, ADC could be a more consistently sensitive marker for detecting early microstructural changes associated with poor glycemic control [27]. Such an interpretation is consistent with previous studies, including Jing et al. [14], who reported subtle yet significant white matter microstructural abnormalities in adults with T2DM.

The significant correlations between HbA1c levels and ADC values underscore the importance of maintaining optimal glycemic control to preserve brain health in T2DM patients. Chronic hyperglycemia may contribute to neurodegenerative processes through mechanisms such as oxidative stress, inflammation, and microvascular complications, ultimately leading to structural brain changes [28, 29]. This is supported by studies like that of Yoon et al. [30], which found brain changes in overweight/obese and normal-weight adults with T2DM, suggesting that metabolic factors play a crucial role in brain health.

Our findings are consistent with previous research indicating that T2DM is associated with alterations in brain microstructure. For example, a meta-analysis by Huang et al. [31] reported widespread white matter abnormalities in T2DM patients, emphasizing the impact of the disease on brain integrity. However, it is important to note that not all studies have found significant associations between glycemic control and brain microstructure. Some research has reported no significant differences in DTI metrics between T2DM patients and controls, suggesting that factors such as disease duration, severity, and the presence of complications may influence the extent of brain alterations.

Limitations

This study has several limitations. First, its cross-sectional design prevents the establishment of causal relationships between glycemic control and brain microstructural alterations. Longitudinal studies are needed to assess the progression of these changes over time. Second, the relatively small sample size may limit the generalizability of our findings, particularly regarding subgroup comparisons. Larger, multicenter studies are necessary to validate these results in more diverse populations. Additionally, this study did not include cognitive assessments, which would have provided a more comprehensive evaluation of the clinical implications of the observed brain microstructural changes. Future research integrating neuropsychological testing with DTI analysis could clarify the functional consequences of white matter alterations in T2DM. Lastly, while DTI is a powerful tool for assessing microstructural integrity, it does not provide direct histopathological evidence of neuronal or axonal damage. Future longitudinal studies with larger cohorts, multimodal imaging approaches, and cognitive assessments are warranted to better elucidate the impact of glycemic control on brain microstructure in T2DM.

Conclusions

In conclusion, our study suggests that poor glycemic control in T2DM is associated with microstructural alterations in the brain, particularly in the right hippocampus. These findings highlight the importance of stringent glycemic management to potentially mitigate diabetes-related brain changes and preserve neurological health.