Abstract
Background:
Type 2 diabetes mellitus (T2DM) accelerates cognitive decline, which is believed to be triggered by aberrant neural activity.
Objective:
To explore how glucose fluctuations impact brain functional architecture and cognition in T2DM patients.
Methods:
T2DM patients were divided according to glycemic variability, forming two categories: patients with fluctuating glucose levels and patients with stable glucose levels. Degree centrality (DC) was calculated within the cerebral gray matter of each participant and was compared among the two patient groups and a healthy control group. The relationships between glucose fluctuations and aberrant DC and cognitive performance, as well as the relationship between aberrant DC and cognitive performance, were further explored.
Results:
Compared with T2DM patients with stable glucose levels, T2DM patients with fluctuating glucose levels exhibited significantly worse performance on the Montreal Cognitive Assessment, Trail Making Test-B (TMT-B), and verbal fluency test (VFT), as well as significant decreases in DC in certain regions, most of which were within the default mode network. In the combined T2DM group, the mean amplitude of glycemic excursions (MAGE) was positively correlated with TMT-B scores and negatively correlated with VFT scores. Moreover, the MAGE was negatively correlated with DC in the left medial prefrontal cortex (mPFC). In addition, TMT-B scores were negatively correlated with reduced DC in the left mPFC.
Conclusion:
These findings further contribute to the mounting evidence of the effects of glycemic variability on the diabetic brain. Tightened control of glucose fluctuations might prevent cognitive decline and changes in brain functional architecture in T2DM individuals.
INTRODUCTION
Type 2 diabetes mellitus (T2DM) affects over 400 million people worldwide, and its prevalence continues to increase with changing lifestyles and the aging of the population [1]. Long-term cognitive dysfunction has been increasingly recognized over the past few decades as a clinical complication of T2DM. Patients with T2DM are more likely to experience cognitive decline than people without T2DM, and cognitive impairment tends to proceed more quickly in the former cohort as well [2].
Glycemic disorders consist of two components, including the duration and magnitude of chronic sustained hyperglycemia and the fluctuations in glucose over a daily period [3]. A number of studies have shown that hemoglobin A1c (HbA1c), which is regarded as an indicator of a long-term hyperglycemic state, is an independent risk factor for the decline in cognitive performance in T2DM patients [4]. A recent longitudinal study suggested that the management of blood glucose was significantly associated with the rate of decline in global cognition as well as in memory and executive function [5]. Likewise, based on recent clinical evidence, glycemic fluctuations should be considered when planning glucose control strategies to reduce the burden of diabetes-related cognitive decline. Rizzo et al. showed for the first time that poor cognitive performance in T2DM patients was associated with daily acute glucose fluctuations [6]. A 20-year longitudinal study indicated that glucose peaks are a risk factor for dementia and the exacerbation of cognitive decline among diabetes patients [7]. Furthermore, according to an anatomical magnetic resonance imaging (MRI) study, glycemic fluctuations are independently associated with brain atrophy and reduced cognitive performance [8]. However, previous investigations are based only on cognitive performance and structural MRI, which are not able to track the very early progression of brain functional alterations. No study has examined the functional brain alterations related to glucose fluctuation or their relationship with cognitive impairment, which may provide valuable insight into the neural mechanisms underlying the physiological and psychological changes that occur during glucose excursions.
The implementation of functional MRI (fMRI) techniques has advanced our ability to study brain alterations and T2DM-related cognitive impairment. Functional connectivity analyses based on fMRI can identify the mechanisms that underlie cognitive differences between individuals [9]. Although seed-based functional connectivity analysis measures the signal synchrony of low-frequency fluctuation activity among different brain areas, it does not provide information on whole-brain functional architecture. Aiming to study the effects of glycemic variability on brain functional architecture and cognition in T2DM patients, we calculated degree centrality (DC) to reflect the interneuronal connections for a full structural exploration of the diabetic brain. DC, a graph theory metric based on a functional network analysis, is commonly employed to measure the number of direct connections of a given node and to quantify the connectivity strength between each node and the rest of the brain in a comprehensive connectivity matrix of the brain (i.e., the functional connectome); this measure indicates how much the node affects the entire brain and integrates information across functionally segregated brain regions [10, 11]. A high DC value implies that a given node has a crucial role in the functional brain network and a broad potential influence within and beyond its own network through its connections. This algorithm has been demonstrated to have high test-retest reliability [12], and has been applied to elucidate pathological mechanisms of neuropsychological diseases, such as Alzheimer’s disease, alcohol dependence, and bipolar disorder [13–15]. We hypothesized 1) that T2DM patients with fluctuating glucose levels would exhibit aberrant connectivity strength and reduced cognitive function compared with T2DM patients with relatively stable glucose levels, and 2) that the altered connectivity strength in certain brain regions would be associated with cognitive performance and the degree of glycemic variability.
METHODS
Subjects
One hundred fifty-one right handed participants, consisting of 97 diabetic patients from the Department of Endocrinology, Nanjing First Hospital, and 50 age-matched healthy subjects, were recruited into the study from June 2016 to January 2018. The healthy controls were recruited during the same period through community health screening or online advertisements and were matched for sex, age, and education. The inclusion criteria were as follows: 1) participants were between 45 and 70 years old, and 2) they had more than 6 years of education. The exclusion criteria included the following: 1) previous history of brain disease; 2) neurological or psychiatric illness; 3) alcohol or drug abuse; and 4) major medical illness, such as severe anemia, thyroid dysfunction, cancer, severe heart disease, and damaged liver or kidney function. The protocol and informed consent document were approved by the research ethics committees of the Nanjing First Hospital, Nanjing Medical University. The T2DM subjects were using oral hypoglycemic agents, with or without the addition of insulin. Patients who used only insulin were excluded. All individuals gave written informed consent before their participation in the study protocol.
Clinical and neurological test information
Clinical data were collected at a preliminary examination. A fasting blood draw was obtained to measure plasma glucose, HbA1c, and lipids. Neurological assessments were then performed. The Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA) were used to test for possible dementia. In addition, the Auditory Verbal Learning Test (AVLT), Rey-Osterrieth Complex Figure Test (CFT), Digit Span Test (DST), Trail Making Test-A and B (TMT-A and TMT-B), and Clock Drawing Test (CDT) were administered in a fixed order to evaluate memory, attention, executive function, and visuospatial function.
MRI acquisition and image preprocessing
All subjects were scanned using a 3.0-T MRI scanner (Ingenia, Philips Medical Systems, Netherlands) using a receive-only 8-channel head array coil. The subjects lay in a supine position with their heads fixed to minimize head motion. Earplugs were used to attenuate scanner noise. The subjects were instructed to keep their eyes closed, to refrain from thinking about anything in particular, and to minimize head motion during the scan.
Blood oxygenation level-dependent (BOLD) images of the whole brain were obtained in 36 axial slices using a gradient echo-planar imaging sequence (repetition time (TR) = 2000 ms, echo time (TE) = 30 ms, slices = 36, thickness = 4 mm, gap = 0 mm, field of view (FOV) = 240 mm×240 mm, acquisition matrix = 64×64, flip angle (FA) = 90°, voxel size = 3.75×3.75×4.0 mm3). Structural images were obtained using a rapid acquisition gradient echo three-dimensional T1-weighted sequence (TR/TE = 8.1/3.7 ms, slices = 170, thickness = 1 mm, gap = 0 mm, FA = 8°, acquisition matrix = 256×256, and FOV = 256 mm×256 mm).
fMRI data preprocessing was performed using Data Processing & Analysis for (Resting-State) Brain Imaging (DPABI_V2.3_170105) based on statistical parametric mapping (SPM8) and rs-fMRI data analysis toolkits (REST) [16]. Briefly, the first 10 volumes were discarded to avoid the magnetization equilibrium of the initial MR signals. Slice-timing correction, realignments for head-motion correction, spatial normalization to the Montreal Neurological Institute template (resampling voxel size = 3×3×3 mm3), and spatial smoothing with an isotropic Gaussian kernel (full width at half maximum (FWHM) = 6 mm) were then performed on the remaining 230 consecutive volumes. The BOLD signal was then processed to eliminate linear trends and bandpass filtered (0.01–0.08 Hz) to reduce the effects of low-frequency drift and high-frequency physiological noise. Global mean signals, white matter signals, cerebrospinal fluid signals and head-motion parameters were regressed out of the rs-fMRI data as nuisance covariates. Finally, the mean time series of 90 regions of interest (ROIs) defined by the Automated Anatomical Labeling atlas were obtained for each individual by averaging the rs-fMRI time series over all voxels in each ROI. The time course of head motion was obtained by estimating the translation in each direction and the rotation in angular motion around each axis for all 240 consecutive volumes. Four subjects (3 T2DM patients and 1 healthy control) with head motion >2.0 mm translation or 2.0° rotation in any direction were excluded. Thus, further analysis was conducted on 97 T2DM patients and 50 healthy controls.
The voxelwise centrality analyses were restricted to a predefined gray matter mask that included tissue with gray matter probabilities greater than 20%, as previously described [17]. The individual network centrality maps were generated in a voxelwise fashion in the mask. First, the preprocessed functional runs were subjected to voxel-based whole-brain correlation analysis. The time course of each voxel from each participant was correlated with the time course of every other voxel, which resulted in a correlation matrix. An undirected adjacency matrix was then obtained by thresholding each correlation at r > 0.25 [10, 18]. We chose a high threshold to prevent voxels with low temporal correlations due to signal noise from being counted. Then, we calculated the binary DC value of the whole-brain network. A graph of the DC values for each GM voxel of each participant was obtained. Finally, these voxelwise DC maps were converted into a z-score map for group comparisons.
We first estimated the spatial distribution of the mean DC in each group. The individual z values were entered into the SPM12 software for a random-effect one-sample t-test in a voxelwise manner to show the average DC maps within each group. We applied a threshold of p < 0.01, corrected for multiple comparisons by the false discovery rate (FDR) method. This correction was confined to the aforementioned GM mask.
Continuous glucose monitoring
Continuous glucose monitoring (CGM) is an important component in diabetes management and a useful method to visualize glucose excursions. Each patient was equipped with a CGM sensor (Medtronic Incorporated, Northridge, Minnesota, USA) for 3 days. Specifically, the CGM sensor was subcutaneously embedded on day 0 at approximately 16 : 00–17 : 00. Subjects were instructed to keep the sensor fixed in place and to protect it from water. The specialist nurse measured the subjects’ glucose four times daily using the finger stick method for calibration. The patients were instructed to maintain their treatment and usual diet over the period with CGM, and they were not aware of the CGM readings during data collection. The subjects had the sensors removed at approximately 16 : 00–17 : 00 on day 3, and the CGM data were downloaded to a computer. In order to avoid bias due to insertion and removal of the sensor, the characteristic glucose pattern of each individual was calculated by averaging the profiles obtained on study days 1 and 2. Several plasma glucose fluctuation parameters, such as the mean amplitude of glycemic excursions (MAGE), the mean blood glucose (MBG), and the standard deviation (SD) of the MBG, were calculated with software provided by Medtronic Incorporated.
According to Zhou et al., a MAGE < 3.9 mmol/L was recommended as the normal glycemic variability range for Chinese adults [19]. Therefore, the patients with MAGE less than 3.9 mmol/L were placed into the group of T2DM patients with stable glucose levels, and the patients with MAGE equal to or greater than 3.9 were placed into the group of T2DM patients with fluctuating glucose levels. Post hoc analyses were conducted among the above two diabetic groups and healthy controls.
Statistical analysis
For the analysis of demographic and clinical variables, comparisons between groups were made using the Mann-Whitney U test for asymmetrically distributed variables, Student’s t-test for normally distributed variables, and χ2 test for categorical variables. A p-value <0.05 was considered to indicate statistical significance. For all T2DM individuals, Pearson’s correlation was computed between the MAGE and each cognitive score, corrected for age, sex, education, and MBG. p < 0.05 was considered statistically significant.
A two-tailed two-sample t-test was then conducted to investigate the differences in the DC maps between groups. Between-group comparisons of the DC maps were performed in SPM12 software using general linear model (GLM) analysis, with age, sex, education, and MBG included as nuisance covariates. A correction for multiple comparisons was performed by the FDR method with a corrected threshold of p < 0.01. For between-group analysis, a mask was created by combining the significant clusters in both groups, which were obtained from one-sample t-test results.
The mean Z values of each brain area that showed significant DC group differences were extracted. Pearson’s correlation was computed between the glucose variability parameters and the mean DC values as well as each cognitive score (uncorrected for multiple comparisons) in the combined T2DM group. p < 0.05 was considered statistically significant, and the analyses were corrected for age, sex, and education. Given the wide age range of the subjects in the study, we also performed multiple regression in SPM8 to investigate the brain regions potentially altered with increased age in each group as well as in the combined group.
RESULTS
The clinical characteristics of the participants are presented in Table 1. The three groups did not significantly differ in age, sex, education level, or blood pressure. No significant difference was observed between the group of T2DM patients with fluctuating glucose levels and those with stable glucose levels in terms of diabetes duration (6.27±5.09 years versus 7.59±5.97 years). The T2DM patients with fluctuating glucose levels had higher MAGE (6.31±2.23 versus 2.78±0.72 mmol/L, p < 0.001) than the subjects in the group of T2DM patients with stable glucose levels (Fig. 1). As shown in Table 2, the T2DM patients performed more poorly on most neuropsychological tests than the healthy controls. The fluctuating glucose group exhibited the worst performance. The fluctuating glucose group showed significantly worse performance than the stable glucose group on the MoCA, TMT-B, and VFT (p < 0.05).
Demographic and clinical characteristics of study subjects
Data are represented as Mean±SD. *Indicated significant differences among the three groups (p < 0.05). aIndicated significant differences between T2DM patients with fluctuating glucose level and T2DM patients with stable glucose level (p < 0.05). bIndicated significant differences between T2DM patients with fluctuating glucose level and healthy controls. cIndicated significant differences between T2DM patients with stable glucose level and healthy controls. BMI, body mass index; BP, blood pressure; LDL-C, low density lipoprotein cholesterol; HDL-C, high density lipoprotein cholesterol.
The average blood glucose concentrations by hour in T2DM patients.
Neuropsychological test information in all subjects
Data are represented as Mean±SD. *Indicated significant differences among the three groups (p < 0.05). aIndicated significant differences between T2DM patients with fluctuating glucose level and T2DM patients with stable glucose level (p < 0.05). bIndicated significant differences between T2DM patients with fluctuating glucose level and healthy controls. MMSE, Mini Mental State Exam; MoCA, Montreal Cognitive Assessment; AVLT, auditory verbal learning test; CFT, complex figure test; TMT, trail making test; CDT, clock drawing test; DST, digit span test; VFT, verbal fluency test.
In the healthy control group (Fig. 2A), stable glucose group (Fig. 2B), and fluctuating glucose group (Fig. 2C), the weighted DC was mostly localized in the prefrontal, temporal, and occipital cortex. Specifically, after two-sample t-test analysis, the stable glucose group, compared with the healthy control group, exhibited significantly reduced DC in the left middle frontal gyrus (Fig. 3A, Table 3). Moreover, relative to the stable glucose group, the fluctuating glucose group showed significantly decreased DC in the left medial prefrontal cortex (mPFC), right middle temporal gyrus, left precuneus, and right cuneus (Fig. 3B, Table 3) (p < 0.01, FDR corrected).
Within-group DC maps for each group. Within-group DC maps for the healthy control group (A), T2DM with stable glucose group (B), and T2DM with fluctuating glucose group (C).
A) Regions showing significant DC changes between T2DM patients with fluctuating glucose and healthy controls. B) Regions showing significant DC changes between T2DM patients with fluctuating glucose levels and T2DM patients with stable glucose levels. Significance thresholds were corrected using the FDR and set at p < 0.01. Note that the left side of the image corresponds to the right hemisphere.
Regions showing significant differences of degree centrality between T2DM patients with fluctuating glucose and stable glucose and healthy controls
Significant thresholds were corrected using FDR criterion and set at p < 0.01. BA, Brodmann’s area; MNI, Montreal Neurological Institute; L, left; R, right.
The results of the correlation analyses are displayed in Fig. 4. In the combined T2DM group, MAGE was positively correlated with TMT-B scores (r = 0.375, p < 0.001) and negatively correlated with VFT scores (r = –0.251, p = 0.015). Moreover, MAGE was negatively correlated with DC in the left mPFC (r = –0.400, p < 0.001). In addition, TMT-B scores were correlated with reduced DC in the left mPFC (r = –0.384, p < 0.001). These correlations were corrected for age, sex, and education. No other correlations between disrupted DC values and other clinical characteristics or neuropsychological test results were detected. We also did not detect any significant brain regions (p < 0.05, corrected) affected by increasing age.
Correlations among glucose fluctuations, neuropsychological test results, and aberrant DC in the left medial prefrontal cortex in T2DM patients. A) Correlation between MAGE and TMT-B scores. B) Correlation between MAGE and VFT scores. C) Correlation between MAGE and DC in the left mPFC. D) Correlation between TMT-B scores and DC in the left mPFC.
DISCUSSION
Since the neural mechanisms underlying T2DM-related cognitive impairment are poorly understood and multiple brain systems are affected, we used DC to analyze functional connectivity strength within the whole-brain network. In the current study, considering that very little research has focused on the effect of glucose excursions on the diabetic brain, we closely attended to impaired cognitive function and abnormalities in resting-state functional connectivity strength in T2DM patients with fluctuating glucose levels. Moreover, correlations between glucose fluctuations and decreased DC and neuropsychological tests, as well as a correlation between aberrant DC and neuropsychological tests in T2DM patients, were observed, which indicated that disrupted brain functional architecture might lead to the glucose fluctuation-related decline in cognitive function in T2DM patients.
Consistent with previous studies supporting the presence of multidimensional cognitive abnormalities in T2DM patients, the evidence from this study suggested that T2DM patients exhibited cognitive decrements across widespread domains of function. The concept of chronic hyperglycemia encompasses both persistent and fluctuating hyperglycemia. In clinical practice, HbA1c levels are taken as the standard measure to reflect the long-term blood glucose control of diabetic patients. However, glycemic fluctuation is an aspect of hyperglycemia that is not well captured by HbA1c levels. The effects of blood glucose fluctuations on chronic complications of diabetes mellitus cannot be neglected. The damage to endothelial cells caused by blood glucose fluctuations might be more severe than the damage caused by constant high glucose levels [20, 21]. Plenty of evidence has shown that fluctuations in blood glucose levels can increase oxidative stress and fluctuations, which leads to vascular endothelial dysfunction in T2DM [22]. In this pathway, glucose fluctuations likely lead to the onset and progression of dementia [6].
Clinical studies have also shown that fluctuations in blood glucose levels are one of the independent causes of cognitive impairment [23]. The glycoalbumin/HbA1c ratio, a marker of fluctuations in blood glucose levels, has been demonstrated to be an independent factor contributing to the decrease in cognitive function in elderly subjects with T2DM [24]. Rizzo et al. [6] showed that the MAGE observed over a daily period was associated with cognitive impairment independent of HbA1c among aged T2DM patients. The MAGE, which was designed by Service et al. [25], has commonly been used for assessing blood glucose fluctuations. It was obtained by measuring the arithmetic mean of the blood glucose changes from blood glucose peaks to nadirs or vice versa, when both ascending and descending segments exceeded the value of one SD of the blood glucose over 24 hours. In the current study, T2DM patients with fluctuating glucose levels, whose MAGE was less than 3.9 mmol/L, displayed significantly worse performance than patients with stable glucose on three neurocognitive tests (the MoCA, TMT-B, and VFT) that are dependent on global cognitive function. Verbal and executive capabilities were compromised compared to those of patients with relatively stable glucose. In addition, MAGE was correlated with TMT-B and VFT scores, consistent with previous studies, supporting the relationship between glucose fluctuation and cognitive function, especially executive and verbal ability. A previous study observed that reduced performance in attention and executive functioning tests, which are putatively associated with the frontal lobes, are often detected in the early stage of cognitive decline [26–28]. Therefore, our data imply that fluctuating glucose could aggravate cognitive impairment in T2DM patients to some extent.
We observed that T2DM patients with stable glucose levels had limited decreases in DC relative to healthy controls, while T2DM patients with fluctuating glucose levels showed abnormally decreased DC across widespread regions, most of which were within the default mode network (DMN). Changes in the DMN are the most consistent finding of recent studies focusing on brain alterations in T2DM patients [29]. Interestingly, DMN regions, such as the mPFC, middle temporal gyrus, and precuneus, displayed decreased DC in T2DM patients with fluctuating glucose levels compared to their counterparts with stable glucose levels. The main region affected by glucose excursion is the mPFC, which is a predominant region within the DMN and is also a key area for executive functions, encompassing multiple higher-level control processes and organizing other cognitive operations [30]. In the analyses restricted to T2DM patients with fluctuating glucose levels, decreased DC in the mPFC was correlated with the MAGE and TMT-B scores, suggesting that deterioration of brain architecture might mediate glucose excursion and cognitive impairment. Despite the large age range of the subjects in the study, the potential effect of age is undetectable. Although the effect of age on cognitive function has been demonstrated [31, 32], it did not significantly influence the brain in the current cohort. Hence, enlarging the sample size and reanalyzing the data will be necessary for future studies.
The visual cortex is linked to the processing of vision-related information and the encoding of visual memories [33]. In our previous work and other fMRI studies [34, 35] focused on T2DM patients, the visual cortex displayed significant abnormal neural activity, and correlations between abnormal neural activity in the visual cortex and cognitive performance involving visual memory had also been reported. We speculate that altered activity in the visual cortex might play an important role in vision-dependent tasks in patients with T2DM. Nevertheless, in the current study, we did not find any decreased cognitive performance in terms of visual memories in patients with fluctuating glucose. This outcome might be explained by potential alterations in functional brain architecture prior to the cognitive performance changes. However, future studies are warranted to clarify the underlying neuropathology of visual cortex functioning in T2DM patients and its relationship with glucose fluctuation. Several limitations of this study should be noted. First, this is a cross-sectional, observational study of a relatively small cohort. Longitudinal studies with larger samples are therefore needed to further elucidate whether reducing glycemic variability might prevent cognitive decline. Second, we collected the CGM data over a period of only 48 hours to calculate the glycemic variability measures, guaranteeing the insufficient stabilization of the monitoring system and the accuracy of the data. A longer period of glucose monitoring data could be more helpful for displaying an individual’s glucose excursions. Third, microangiopathy is potentially related to cognitive and brain alterations; however, we cannot avoid confounders, given that there is no “gold standard” to assess the existence of microangiopathy. Finally, only patients aged 45–70 years were included in the present study to minimize the diversity of vascular burdens among the participants, so our study results may not be entirely generalizable to all adults with T2DM. Studies focusing on relatively younger T2DM individuals will be further needed.
In summary, we observed that excessive glucose excursions are associated with disruption of executive, verbal, and attentional functioning, and deficits in attention and executive function were the most prominent characteristics of T2DM patients with fluctuating glucose levels. Disruption of the functional architecture of the brain, especially the frontal lobe, plays a vital role in the process. Our results further support the view that glucose variability should be regarded as one of the targets of treatment for T2DM patients and imply that therapy aimed at controlling glucose fluctuation in T2DM patients may help prevent the onset and development of cognitive impairments. In addition, once dementia has developed, it becomes very difficult to control glycemic fluctuations. T2DM patients’ quality of life could be improved by breaking this vicious cycle.
Footnotes
ACKNOWLEDGMENTS
We would like to express our heartfelt gratitude to all the staff of the Department of Endocrinology and the Department of Radiology, Nanjing First Hospital, Nanjing Medical University, for their selfless help and valuable assistance.
This work was supported by the National Natural Science Foundation of China (No. 81600638), Medical Science and Technology Development Foundation of Nanjing Department of Health (No. YKK16140), and National Key R&D Program of China (No.2018YFC1314100). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
