Changes in Electrical Activities of the Brain Associated with Cognitive Functions in Type 2 Diabetes Mellitus: A Systematic Review

Abstract

Introduction: Electroencephalogram (EEG) has the potentials to decipher the neural underpinnings of cognitive processes in clinical and healthy populations. Objective: The current systematic review is intended to examine the functional brain changes underlying cognitive dysfunctions in T2DM patients. Methods: The review was conducted on studies published in the PubMed, WebofScience, Cochrane, PsycInfo database till June 2021. The keywords used were electroencephalogram, T2DM, cognitive impairment/dysfunction. We considered studies using resting-state EEG and ERP. The preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines were followed to compile the studies. Results: The search yielded a total of 2384 studies. Finally, 16 independent studies were included. There was a pattern of a shift in EEG power observed from higher to lower frequencies in T2DM patients, though to a lesser degree than Alzheimer's disease patients. P300 latency was increased in T2DM patients mainly over frontal, parietal, and posterior regions. P300 and N100 amplitudes were decreased in T2DM patients than in healthy controls. Conclusion: The results indicate that T2DM has consequences for cognitive functions, and it finds a place in the continuum of healthy cognition to dementia.

Keywords

type 2 diabetes (T2DM)EEG ERP cognitive function cognitive decline

Introduction

Diabetes Mellitus (DM) is a significant global health concern. It affected approximately 463 million people in 2019 and is estimated to reach 700.2 million by 2045.¹ Around 90% of these cases are diagnosed as Type 2 Diabetes Mellitus (T2DM),^2,3 which is a late-onset most common type of diabetes characterized by the reduced capacity of peripheral tissue to regulate glucose homeostasis in response to insulin.⁴ T2DM is known for its long-term macrovascular and microvascular complications, ultimately affecting brain health that may lead to Alzheimer's disease and vascular dementia.⁵ To be specific, epidemiological studies suggest that individuals with T2DM have an increased risk of cognitive decline (1.5%) and dementia (1.6%) than healthy controls.⁶

Identifying mechanisms behind this association is crucial to tackling cognitive complications of T2DM. Recognizing these mechanisms will help address the complications at an earlier stage where treatment is more efficient. However, the mechanisms responsible for cognitive impairments in T2DM patients are still poorly understood. Some factors identified so far are hypoglycemia, compromised glycemic control, impaired insulin signalling and the most important, hyperglycemia.⁷ Along with these, advanced glycation end (AGE) production (mediated by hyperglycemia) coupled with oxidative stress can degenerate neurons, damage vascular endothelium, and lead to cognitive impairment.⁸

The information gathered from structural and functional brain changes also enables the early detection of cognitive problems. Brain structural changes investigated in T2DM patients via MRI (Magnetic Resonance Imaging) studies, reported the presence of regional atrophy in the hippocampus, basal ganglia, orbitofrontal and occipital lobes.⁹ An increased presence of subcortical infarcts and large vessel infarcts were also reported.¹⁰ T2DM patients also exhibited the presence of white matter lesions as compared to healthy controls.^10,11 Diffusion Tensor Imaging (DTI) metrics provide indices of white matter axonal integrity, tract anatomy, and connectivity between brain regions. It demonstrated that microstructural white matter abnormalities might contribute to deficits in brain structure and function in adults with T2DM.^12–14 Collectively, it can be implicated that structural MRI indices provide evidence of localized and widespread brain abnormalities in T2DM adults.

Electroencephalogram (EEG) is another approach to study the functional brain changes through the brain's cortical activity. It measures the electrical activities of the brain and represents the aggregate post-synaptic currents of neurons in the brain. EEG has been used to identify different types and severities of cognitive impairments. Moreover, it is a cost-effective and non-invasive method, accessible in most countries across the globe.¹⁵ Hence EEG could be a promising tool to extract the characteristics of cortical functional brain connective related to MCI in T2DM.

The current systematic review sought to examine the cortical activity of the brain underlying cognitive dysfunction in T2DM. This review focuses on both the task-based EEG, ie, event-related potentials (ERPs) and resting-state EEG (rsEEG) studies, without limiting to any particular kind of analysis technique adopted (power spectrum analysis, synchronization, coherence etc) in the studies. To the best of our knowledge, this is the first review ever presented, consolidating the pieces of evidence of both EEG and ERP studies investigating cognitive processes in T2DM patients.

Methods

Search Strategy and Study Selection

Database search was conducted in PubMed, Cochrane, Science Direct, and Wiley Online Library from the earliest records till Jan 2020. Search terms were entered as follows: Type 2 Diabetes Mellitus (T2DM) and cognitive dysfunction, T2DM and EEG, T2DM and event-related potentials (ERPs), neurocognitive changes in T2DM, T2DM, and brain functions. Studies on humans, published in the English language and used only EEG as an assessment tool, were considered for the study. Exclusion criteria included animal studies, patients with type1 diabetes, and imaging modalities other than EEG.

The duplicate articles were deleted as the first step of study selection. The remaining returns were then evaluated based on the title and the abstract and were included only if they: 1) were original and empirical studies, 2) analysed cognitive functions in T2DM, 3) used only EEG or ERP method to analyse the cognitive processes. Studies surviving this step of evaluation were then searched for full article.

Figure 1 provides the summary of the search process.

Figure 1.

Result of systematic search.

Quality Assessment and Extracted Information

NIH Quality assessment tool for case-controlled and cross-sectional studies was used to rate the quality of the included studies. Information extracted from the studies consists of the publication year, country, imaging modality, analysis, and participant demographics, including the number of participants.

Results

The search yielded a total of 90 studies. Sixteen studies were selected after excluding the irrelevant studies. No overlapping of participant samples was seen in any of the studies. The studies were of fair to good quality as assessed by the NIH quality assessment tool.

Study Demographics and Details

The studies were conducted across many different countries including Sweden (n = 1), USA (n = 2), Japan (n = 2), India (n = 2), Brazil (n = 1), Finland (n = 1) and China (n = 7) being the highest. Not all studies reported the average age of the participants. However, the age of the participants ranged from 32 to 84 years. The disease duration ranged from 1 to 28 years. Only three studies reported whether the participants were using insulin.^16–18 Most of the studies reported groups matched for age, gender, and education. Six studies performed ERP analysis, and eight studies performed frequency analysis. Besides, two studies performed both ERP and frequency analysis.

In summary, eight of the sixteen studies used ERP analysis. While four studies performed power spectral analysis, two studies did coherence analysis. Synchronization and coupling analysis were carried out by one study each.

The supplementary tables provide study demographics details (Tables 1 and 2).

Table 1.

Studies of Resting State EEG in Type 2 Diabetes Mellitus.

Study	Participants	T2DM duration	Years of education	HbA1c, insulin use, HOMA-IR	Country	EEG measures	Results	EEG correlated with neuropsychological test performance
Benwell et al (2019)	T2DM, n = 27 (12 females, 50-78 years) AD, n = 18, (11 females, 52-86 years) Heathy Control, n = 27, (13 females, 50-77 years)	Not reported	Not reported	HbA1c < 10	USA	Power density (absolute and relative power, spectral power ratio), peak frequency	Relative d + θ power higher for AD compared to T2DM and HC. Relative α + β power were lower for AD compared to T2DM and HC. Higher relative power in lower gamma band in AD compared to HC. Relative alpha was lower in T2DM in frontal, temporal, and posterior regions. Lower alpha in temporal region in T2DM	An overall relationship between the composite neuropsychological scores and the Spectral power ratio (SPR). Positive association with learning and memory, and EFs with SPR. Not related with T2DM.
Bian et al (2014)	(T2DM) aMCI, n = 16 (5 males, 11 females mean age 69.7 ± 8.4 years, range 52 to 84 years Control, n = 12 (6 males, 6 females mean age 73.3 ± 4.6 years, range 63 to 80 years	aMCI = 9.3 ± 2.4 years, range 1 to 20 years Controls = 14.0 ± 3.1 years, range 1 to 30 years	aMCI = 12.9 ± 1.8 years, range 6 to 16 years Control = 13.8 ± 3.0 years, range 9 to 19 years	Not reported	China	Relative power and coherence	Relative power not different across the groups. Higher ratios of P(theta)/P(alpha) in the frontal and left temporal regions in aMCI. Lower alpha coherence in posterior region (intra region) for aMCI group. Higher inter-hemispheric coherence in aMCI in delta, a lower coherence in the theta.	P(theta)/P(alpha) negatively correlated to MoCA in the frontal and left temporal regions. No significant correlation between ratios of power and neuropsychological tests. Alpha, theta, and delta coherences correlated to MoCA scores.
Cooray et al (2011)	T2DM, n = 28 Age range 50 to 70 years (T2DMi) n = 15 (T2DMr) n = 13 Controls, n = 21	T2DMi = 8.6 years (1-22) T2DMr = 11.2 years (2-28)	12 (7-20) years	(HbA1c) T2DMi = 8% (5.8-12.9) T2DMr = 8% (5.1-10.0) (Insulin) T2DMi = 80% T2DMr = 69%	Sweden	Power, phase lag index	Reduced power in the beta band in T2DM. Most pronounced reduction over central and posterior regions. No differences in PLI in the theta, alpha or beta band between T2DM and controls. Increased alpha-power and connectivity in alpha and theta across hemispheres in T2DMi. Increase in beta band connectivity over central region in T2DMi.	Not reported
Cui et al (2016)	(T2DM), n = 46 aMCI, n = 26, (10 male, 16 female) Controls, n = 20, (12 male, 8 female	Not reported	Not reported	Not reported	China	Synchronization	Decreased synchronization in all regions in aMCI. Particularly low in central and occipital regions. NWPMI and SNWPMI are effective index to estimate the synchronization strength.	SNWPMI correlated with MMSE and semantic fluency significantly.
Cui et al (2018)	(T2DM), n = 35 (14 males, 21 females) Age 69.43 ± 8.79, range 43 to 84 years aMCI, n = 18, 6 M, 12 F Controls, n = 17, 8 M, 9 F	Not reported	aMCI = 13.44 ± 3.09 Controls = 13.44 ± 3.09	Not Reported	China	Magnitude squared coherence for synchronization	Larger theta and delta coherence in aMCI group. Lower Alpha coherence in aMCI.	Coherence negatively correlated with WAIS-DST, AVLT, BNT, MoCA. aMCI coherence lower in left hemisphere, positively correlated with MoCA.
Cui et al (2014)	(T2DM) aMCI = 8, 5 M, 3F Age 70 ± 10.784 years Control = 11, 5 M, 6 F Age 74.27 ± 3.349 years	Not reported	aMCI = 13.88 ± 3.441 control = 13.64 ± 2.541	Not Reported	China	Synchronization	Decreased synchronization in aMCI in all bands. Frontal S, SI, and GSI values of aMCI group were not significantly different. Three synchronization values in parietal and temporal regions all were different in delta and beta2 bands between aMCI and control group. Occipital theta and beta2 synchronization more significantly different between the two groups, especially in the beta2 band.	Only temporal theta S values, Occipital theta S values, and Boston Name Testing were strictly correlated significantly.
Cui et al (2016)	(T2DM) n = 32, 15 M, 17 F aMCI, n = 18, 8 M, 10 F Age 70.28 ± 8.5 NC, n = 14, 9 M, 5 F Age 63.79 ± 16.9	Not reported	aMCI = 13.72 ± 2.87 NC = 12.57 ± 3.23	Not reported	China	Determinism	Determinism of SRP (SRP_DET) of aMCI has higher values, indicating increase of EEG regularity in the temporal and occipital regions. SRP_DET is more effective than PRP_DET in finding the correlated EEG deterministic characteristic of aMCI in T2DM.	SRP_DET is strictly negatively correlated to MoCA and memory functions at electrode P4.
Mooradian et al (1988)	(NIDDM) n = 43, Age 66.3 ± 0.8 years HC n = 41, Age 65.3 ± 0.3 years	13.3 ± 1.8 years	Not reported	HbA1c = 109 ± 36 g/L Insulin users = 20 Hypoglycemic agents = 40	USA	Frequency power	Slowing of EEG rhythms over the central cortex. Significantly higher % of theta activity at Cz in diabetics. Significantly less alpha power at Pz in diabetics. More power distribution in diabetics in slower delta and theta EEG bands.	Not reported.
Wen et al (2016)	(T2DM), n = 39 Age 68.95 ± 8.95, range 43 to 84 years aMCI, n = 19, 12 F, 7 M NC, n = 20, 11 F, 9 M	aMCI = 9.19 ± 6.29, Range 1 to 20 years NC = 13.60 ± 8.59 years Range 1 to 30 years	aMCI = 13.00 ± 2.94 NC = 12.70 ± 3.40	Not reported	China	Coupling strength and directionality (permutation conditional mutual information)	Difference in coupling strength and directionality of EEG signals between aMCI, T2DM and NC in different brain regions. PCMI can effectively calculate the coupling strength and directionality.	Coupling strength and directionality correlated with MoCA but not for all electrode pairs.
Zeng et al (2015)	[T2DM] aMCI, n = 16, 5 F, 11M Age = 69.7 ± 8.4 Controls, n = 12, 6 F, 6 M Age = 73.3 ± 4.6	aMCI = 9.3 ± 2.4 Controls = 14.0 ± 3.1	aMCI = 12.9 ± 1.8 Controls = 13.8 ± 3.0	NA	China	Synchronization (PLI), graph analysis	Decreased global mean PLI in lower alpha, upper alpha, and beta bands in aMCI. Lower functional connection at a short and long intra-hemispheric distance on left hemisphere. Clustering coefficient lower in aMCI group, and the path length significantly increased.	Cognitive status measured by MoCA had a significant positive correlation with cluster coefficient and negative correlation with path length in lower alpha band.

Table 2.

Studies of Task Based EEG (ERP) in Type 2 Diabetes.

Study	Participants	T2DM duration	Years of education	HbA1c, insulin use, HOMA-IR	Country	Results	Neurocognitive task and Cognitive function assessed
Kurita et al. (1996)	(NIDDM) n = 60 50.7 ± 8.9 years, range 32 to 67 years, 39 men, 21 women (Neurologically healthy) n = 20, mean age 49.8 ± 9.8 years, range 32 to 67 years,	110 ± 7.3 years, range 1 to 28 years	Not reported	HbA1c = 7.4 to 15.9%, mean 10.9 ± 2.0% Insulin users = 26 Hypoglycaemic agents = 18	Japan	The slope of the latency/age regression line was 1.51 ms/year. Longer P300 values in diabetics. Mean P3 latency in order of decreasing length were HbA1c ≥ 10%, HbA1c < 10%, and control subjects. No between P300 values and HbA1c levels.	Auditory Oddball
Mooradian et al, (1988)	(NIDDM) n = 43, Age 66.3 ± 0.8 years HC n = 41, Age 65.3 ± 0.3 years	13.3 ± 1.8 years	Not reported	HbA1c = 109 ± 36 g/L Insulin users = 20 Hypoglycemic agents = 40	USA	No increased P100 latency in controls or diabetics at Cz. P3 latency significantly different for diabetics and controls. Trend towards longer latencies in diabetics at Fz and Cz. The intravenous administration of glucose did not alter P300 latency.	Auditory and visual oddball (checkerboard vs diamond pattern), as an index of attention and information processing time.
Cooray et al (2011)	(T2DM) n = 28 (HC) n = 21 Age range 50 to 70 years (T2DMi) n = 15 (T2DMr) n = 13	T2DMi = 8.6 years (1-22) T2DMr = 11.2 years (2-28)	12 (7-20) years	(HbA1c) T2DMi = 8% (5.8-12.9) T2DMr = 8% (5.1-10.0) (Insulin) T2DMi = 80% T2DMr = 69%	Sweden	Lowered N100 amplitude in T2DM. Prolonged P300 latency in T2DM patients No difference in N100 peak latency and P300 peak amplitude between patients and healthy controls.	Auditory oddball
Tondon et al (1998)	(NIDDM) n = 30 Age 43.6 ± 9 years, Controls n = 30 Age 36.73 ± 13.3 years	2 to 10 years	Not reported	HbA1c = 9.9 ± 1.0%	India	Higher N200 and P300 peak latencies diabetics. No significant correlation between latencies of N2 & P3 with height, weight, Glucose levels, and diabetes duration.	Auditory oddball
Mochizuki et al (1998)	(NIDDM) n = 24, mean age 56.7 years, Controls n = 16, mean age 54.9	Not reported	Not reported	HbA1 = 9.7%	Japan	Longer N200 & P300 latencies in the diabetic group. Smaller N200 & P300 amplitudes in diabetes patients. Shorter P300 latency after treatment than before.	Auditory oddball
Hazari et al (2011)	(T2DM) Group 1, n = 11, age 52 ± 6 Group 2, n = 17, age 53 ± 6 Healthy controls N = 18, age 50 ± 7	(T2DM) Group 1 = 3 ± 2 years Group 2 = 10 ± 4 years	Not reported	Insulin users Group 1 = 0 Group 2 = 24%	India	Patients with over 5 years of duration had much prolonged P300 latencies in contrast to patients with 5 years or less disease duration and showed trends similar to that of controls. P300 did not differ significantly among the groups. Hypertensive diabetics showed much prolonged P3 latencies compared to normotensive diabetics.	Novel three stimulus oddball paradigms
Hissa et al (2002)	NIDDM n = 44, 12M, 32 F Age 58.84 ± 8.4, Range 38 to 75 HC n = 17, 2M, 15F Age 56.53 ± 8.09, Range 43 to 69	10.52 ± 7.97 years, Range 2 to 41 years	Not Reported	NIDDM HbA1c 8.28 ± 2.09 Min-Max 5.1 to 13.5 HC HbA1c 5.74 ± 0.95 Min-Max 4.3 to 8.0	Brazil	Higher P300 latency in diabetics. Retinopathy did not influence P300 latency. Insignificant trend for lower values of the P300 latency in patients with no hypoglycemic episodes.	Acoustic ball paradigm
Vanhanen et al (1996)	NIDDM n = 9, 3M, 6 F Age 72.7 ± 2.5 years HC n = 9, 4 M, 5 F Age 74.6 ± 1.8 years	8.2 ± 4.8 years, range 1 to 15 years	NIDDM 6.1 ± 1.6 years HC 7.1 ± 1.5 years	Not reported	Finland	Differences in N100 were spread over a wide range of electrodes. Differences in MMN most significant in fronto-central areas. Habituation of N1 is decreased in diabetics. Shorter N100 latency in central brain areas in diabetics. MMN area, was smaller in diabetics. No difference in N2b and P300 components in both groups. No differences in latencies or amplitudes between the groups. N100 correlated with delayed recall of the wordlist in both the groups. MMN did not correlate with any of the neuropsychological tests. FBG or insulin levels did not correlate with any ERPs within the groups.	Oddball in three separate behavioural condition.

Resting-State EEG (rsEEG)

Ten of sixteen studies have analysed the EEG frequencies to ascertain the cognitive relevance of the cortical activity in T2DM patients. Four studies assessed the spectral power, eight studies evaluated the synchronization levels, and one study assessed determinism.

Benwell et al (2020) compared the EEG frequency characteristics between Alzheimer's Disease patients (AD), T2DM, and healthy controls (HC). A pattern of a shift in EEG power was observed from higher to lower frequencies in T2DM patients, particularly over temporal regions. AD patients showed higher relative (δ + θ) power than T2DM and HC, whereas relative (α + β) power was lower for AD than T2DM and HC.¹⁹ T2DM patients show slowing of EEG rhythms with the reduction in alpha and beta band power over the parietal and central and posterior regions, respectively.^18,20 Also, T2DM patients had a significantly higher % of θ activity at Cz and less power in α at Pz simultaneously. People with diabetes tended to have more power consistently distributed in the slower δ and θ EEG bands at all the three recording sites, although it was not significant.¹⁸ Furthermore, Bian et al reported that the ratios of power of theta versus the power of alpha [P(θ)/P(α)] in the frontal and left temporal region were significantly higher in the T2DM patients having an amnestic mild cognitive impairment (aMCI) as compared to the patients without aMCI.²¹

The synchronization values tend to decrease in T2DM patients with amnestic mild cognitive impairment (aMCI) compared to cognitively healthy T2DM patients, particularly over central and occipital regions.²² The decrease was observed in all the EEG frequency bands.¹⁹ Similar to the power analysis observations, the aMCI group had larger coherence values in θ and δ frequency bands.²³ In addition, α frequency coherence values were lower in fronto-posterior, right-temporo posterior regions.^21,23 Phase lag index (PLI) as a measure of synchronization also shows that the global mean PLI in lower α, upper α, and β bands were significantly decreased in T2DM patients with aMCI.²⁰ However, T2DM patients do not show any difference in PLI than healthy controls.³

Task-Based EEG Studies (Event-Related Potentials)

There were a total of 8 ERP studies. Majority of the studies focused on the P300 component. Six of the studies assessed the P300 ERP component, while three studies assessed the N100 ERP component. The P100 & N200 were evaluated only by one study.

Interestingly, all the studies used oddball tasks as a measure of ERP. Five of the eight studies have used the classical auditory oddball task.^{17,20,24–26} Other remaining studies have used a novel three stimuli auditory oddball task.¹⁶ Finally, one study has employed auditory oddball tasks in three different behavioural conditions.²⁷

Studies using the classic auditory oddball task showed that the T2DM patients had longer P300 latencies than healthy controls.^16,18 Studies employing an altered version of the oddball task also observed similar results.^18,20 The prolonged P300 latency was visible primarily over the brain's frontal, central, and posterior regions.¹⁶ Besides, P100 and N200 latencies were also increased in diabetic patients than in healthy controls.^18,25 At the same time, the N100 component has a heterogeneous outcome from showing no difference to significantly increased latencies in T2DM patients.^20,26

Kurita et al found mean P300 latencies in the order of decreasing length for those with retinopathy, without retinopathy, and control subjects.¹⁷ In terms of the HbA1c group, P300 latencies in the order of decreasing length were HbA1c ≥ 10%, HbA1c < 10%, and controls. The P300 latency was also much delayed in hypertensive T2DM patients.²⁴ On the contrary, it was found to be unaffected by the presence of retinopathy in T2DM patients¹⁶ also, the latencies of P300 and N100 were unchanged with the duration of T2DM. Interestingly, only three studies reported the P300 amplitude, and it tends to decrease in T2DM patients compared to the control group. T2DM patients had lower N100 amplitude than controls, mainly over the central and posterior regions.²⁰ Likewise, N200 and P300 amplitudes were also decreased in T2DM patients than in controls.²⁵ However, only four studies reported measuring the amplitude of ERP components.

Neurocognitive Functions

There were a total of 7 studies assessing various cognitive functions. Patients with AD performed worse in all tests than the T2DM and healthy control.¹⁹ For scores on the Digit Symbol Substitution Test (DSST), Ray Auditory Verbal Learning Test (RAVLT) learning and delayed recognition trials, logical memory immediate and delayed recall trials. In semantic fluency, TMT time, TMT errors, Digit Span backward, RAVLT delayed recall, Boston Naming Test, and GDS, the AD patients performed worse than the HC and T2DM groups. However, the T2DM and HC groups did not differ from each other. Only on the Digit Span forward test was there a difference between HC and AD, while T2DM was no different from either HC or AD.^20,28

The T2DM patients with amnestic Mild Cognitive Impairment (aMCI) tend to have reduced global cognition compared to the T2DM patients with normal cognitive functions.^15,20,23,28

Discussion

The current review aims to understand the neurophysiological changes associated with cognitive functions in T2DM patients as observed by the rsEEG and ERP studies. Overall, T2DM patients show some EEG and ERP characteristics that indicate towards cognitive impairment or future cognitive decline. However, the findings of the previous studies on cognitive impairment in T2DM patients should be carefully interpreted because of the diversity of the study design, sample size and characteristics, and analysis techniques adopted. We intend to discuss the relevant points of the findings hereafter.

Spectral Power Analysis

Spectral power analysis is a very well-known method in EEG signal processing. It is used for the quantification of the spontaneous electrical activities of the brain. Furthermore, with the neuropsychological correlations, the power analysis of the EEG series provides valuable information to distinguish healthy and impaired brain functions.¹⁹

The power analysis of frequency bands reveal that the T2DM patients exhibit a dominance of lower brain frequencies over the higher frequencies, which is similar to the characteristics observed in MCI and AD. Benwell et al, (2020) compared ratios of [P(α + β)/P(δ + θ)] in AD and T2DM. The Spectral power ratio demonstrated a shifting pattern from higher oscillating frequency to lower frequency in AD. Apparently, this shifting pattern was also present in T2DM patients, but to a lesser extent. T2DM patients having aMCI show higher ratio of [P(θ)/P(α)] in frontal and left temporal regions as compared to cognitively healthy T2DM patients.²¹ The present observations reiterate the findings of the previous studies that have demonstrated the dominance of slower brain frequencies in AD and MCI.^29–36 Moreover, the cortical rhythm correlates to the grey matter volume, a candidate biomarker of MCI and AD patients. The higher δ sources and lower α sources are related to the decreased cortical grey matter volume in MCI and AD.³⁷ It means, better cognitive function or better scores cognitive tests are directly proportional to the increased grey matter volume. Hence, the findings of the current review in concurrence with the earlier findings, suggest that the rsEEG activity observed in T2DM is strictly a pathophysiological phenomenon.

The alpha power is reported to be linked with impairments in learning and memory in AD patients and it is also correlated with the hippocampal volume.³⁸ It is well known that hippocampal atrophy is associated with cognitive impairment in MCI as well as in AD. Recently, the measurement of normalized hippocampal atrophy has been introduced in the guidelines for assessing early AD.³⁹ The decreased magnitude of alpha frequency is found to be correlated with progressive hippocampal atrophy in the parietal, occipital and temporal areas in MCI and AD. Hence, future studies should incorporate structural and functional neuroimaging techniques to find and identify the specific characteristics in T2DM that are indicative of cognitive processes. Interestingly, alpha power was shown to be increased in a subset of T2DM patients with aMCI after receiving a 2-month glycemic control treatment. The increased alpha power was associated with improvements in visuospatial and semantic memory performance.²⁰ The increased alpha power after the intervention suggests that the improved glycemic control and early intervention could improve the cognitive performance. Notably, poor glycemic control is one of the mechanisms hypothesized to be responsible for cognitive dysfunction in T2DM.⁷ However, more such investigations are required to investigate the relationship between the glucose levels and oscillatory alpha activity.

Coherence Analysis

The nature of EEG signal characteristics is very complex. Therefore, different methods have been used to analyse the EEG signals from different perspectives. Some of these methods include coherence analysis, coupling analysis, mutual information, and synchronization analysis.

The EEG coherence analysis has been used earlier to evaluate the functionality of cortical connections and provide information about the synchronization of the regional cortical activity.²¹ The coherence analysis used in T2DM patients with aMCI showed a reduction in alpha and theta bands compared to cognitively healthy T2DM patients. The lower alpha band was observed in posterior, fronto-right temporal/fronto-posterior/right temporo-posterior regions. On the other hand, the theta band was reduced in the left and right sides of the central and parietal regions of the brain. In addition, the inter-hemispheric coherence reported increased delta band connectivity in left and right temporal areas as observed in aMCI patients.²¹

Similar changes in alpha and delta frequency bands were also reported in previous studies.^40–42 It is suggested that the increased inter-hemispheric coherence in temporal region is linked to hippocampal atrophy. In contrast the decreased coherence in fronto-parietal region is linked to the subcortical CVD.⁴⁰ Notably, hippocampal atrophy and CVD also are associated with the cognitive decline.^43,44 These outcomes are notable, as the microvascular and macrovascular complications are one of the hypothesized mechanisms for the cognitive impairment in T2DM. In summary, findings indicate that coherence analysis can be used to deduce some EEG characteristics to identify the occurrence and severity of the cognitive impairment in T2DM patients.

Event-Related Potentials

P300 Component

The P300 component is associated with detecting novel stimuli, updating working memory, inhibitory control, and selective attentional processes.⁴⁵ Additionally, P300 is also characterized by a large amplitude (μV) wave and smaller latency (ms) generated by discrimination and attentional neural processes.⁴⁵ The characteristics of P300 component affect differently in healthy and clinical conditions. The findings of P300 latency in T2DM patients have been consistent across all studies as compared to healthy controls.^{16–18,20,24–27} It was also positively correlated with the age, duration, and severity of T2DM.^16,17,24 The findings are suggestive of a possible contribution of microangiopathy or metabolic derangement in a small part. However, the influence of disease duration on P300 latency might be because of the test novelty, which increases the workload of the cognitive task by presenting stimuli at a higher rate.¹⁶ Notably, only P300 latency could differentiate the groups for their cognitive performance. Most of the studies did not find any difference in P300 amplitude among the groups. However, in another study, P300 amplitude was best highlighted only with executive functions tasks, while the latency was highlighted even with the oddball task.⁴⁶ The studies included in the current systematic review used only the oddball paradigm, indeed with some variations. The relative uniformity in the tasks used and heterogeneity in the samples and study designs may have led to the indifferences in the P300 amplitude.

N100 Component

The N100 amplitudes were decreased in both the studies. The decrease in amplitude might be reflecting impaired arousal and probably slight impairment in the ability to automatically redirect the attention. Cooray et al, (2008) hypothesized that the N100 amplitude reduction in Type 1 Diabetes Mellitus (T1DM) patients could be caused by a loss of nerve impulse synchrony in auditory tracts in the white matter. This hypothesis is again supported by the evidence of white matter lesions obtained from MRI studies in both T1DM and T2DM.^11,47,48 Interestingly Vanhanen et al, (1996) found shorter latency in diabetic patients as compared to controls. The auditory N100 ERP component is suggested to signal the detection of acoustic change in the environment. Such acoustic changes cause widespread cerebral activation as part of orienting reaction.⁴⁹ The shorter N100 latency in T2DM might be due to loss of a non-specific arousal component, which emerges slightly later than N100 generated at auditory cortical areas. An alternate explanation could suggest tonically maintained attention to auditory stimuli and an inability to release underlying processes. Both explanations suggest possible impairment in the automatic ability to allocate attentional resources.

Limitations and Future Considerations

The current review involved studies that investigated neurophysiological changes associated with cognitive functions in T2DM patients. There are some limitations in the current study that must be considered. The studies involved in the review varied in design, sample characteristics, and methods of assessments. Majority of the studies were cross-sectional, and only one study used an intervention protocol. Some studies compared T2DM patients with HC, MCI, and AD, while some attempted to understand the difference in cortical activity within a subgroup of T2DM having aMCI. With the varied study designs and analysis methods, it is difficult to reach to a common understanding of the results. Some rsEEG studies were aimed to explore the significance of the analysis techniques in identifying the cognitive characteristics of T2DM patients. Hence, the lack of reproducibility of the findings remains a challenge.

The information pertaining to the demographic characteristics of the samples, like the glucose levels, education years, and the disease duration were also missing in some studies. It is unclear whether the studies failed to report the information or not gathered at all in the process of the investigation. The age range of the samples was excessively stretched from young adult to elderly population. Therefore, it would be difficult to ascertain whether the changes in cortical activities related to cognition are associated with natural aging or pathological condition of T2DM.

The interventional studies assessing cortical activity are rare to find. The current review found only one study that provided intensified glycemic control treatment and concluded with a plausible improvement in cognitive performance as a consequence of improved glycemic control. Again, reliability of the findings remains a challenge as no other interventional studies have assessed the association between glycemic control and cortical activity in T2DM patients.

Hence, we suggest that the future studies should try to assess the cortical activity along with MRI. Cortical activity accompanied with MRI assessments will provide a comprehensive understanding of the cognitive characteristics in T2DM. Moreover, it will help to enhance our knowledge about the cortical activity related to cognitive functions. For example, knowing whether an observed cortical activity provides domain-specific information of cognitive functions or indicates the severity of the cognitive impairment. Neuropsychological studies show that memory, executive functions and information processing are mainly affected in T2DM patients. Hence, unlike the studies of the current review that used only oddball task, future ERP studies may explore the domain-specific cognitive processes in T2DM patients.

Previous studies have demonstrated the possibility of reducing the risk of cognitive decline by providing early interventions to T2DM patients. Hence, developing and identifying interventions to prevent or reduce the risk of cognitive decline in T2DM patients is equally important. Interventions corresponding to alternative and complementary medicines have received growing attention because of their holistic approach, and yoga therapy is one of the most widely accepted. Yoga therapy has shown to be beneficial to improve glycemic control and reduce stress levels among diabetic patients. Unfortunately, no studies were found that assessed the effect of yoga on cognitive processes in T2DM patients. Hence, it will be good to explore the effects of yoga practices on cortical activities related to cognitive functions in the T2DM population.

Conclusion

With the current review, the EEG emerges as a promising tool to investigate the cortical activities associated with cognitive functions in T2DM patients. The rsEEG studies demonstrated that the T2DM patients show some functional alterations in the brain compared to their healthy cohort. These alterations are similar to the characteristics of EEG activity in MCI and AD or Dementia. The dominance of low frequency power, and prolonged latencies and decreased amplitudes of ERP components observed in T2DM patients suggest problems in the domains of attention, memory, and executive functions, which may have cognitive functioning consequences.

Footnotes

Acknowledgments

This study was primarily funded by Ministry of AYUSH, govt. of India. (Sanction number - Z.28015/209/2015HPC(EMR)-AYUSH). The authors express deep gratitude to the research fellows and Anvesana Research Laboratories for their consistent support to accomplish this project.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Ethical Approval

Not applicable, because this article does not contain any studies with human or animal subjects.

ORCID iDs

Amit Kanthi

Deepeshwar Singh

References

International Diabetes Federation. IDF Diabetes Atlas; 2019.

Macpherson

Formica

Harris

Daly

. Brain functional alterations in type 2 diabetes – A systematic review of fMRI studies. Front Neuroendocrinol. 2017;47:34-46. doi:10.1016/j.yfrne.2017.07.001

Zeng

Wang

Ouyang

Bian

Wang

. Complex network analysis of resting state EEG in amnestic mild cognitive impairment patients with type 2 diabetes. Front Comput Neurosci. 2015;9. doi:10.3389/fncom.2015.00133

Roberto

Crisafulli

. Consequence of type 1 and 2 diabetes mellitus on the cardiovascular regulation during exercise: a brief review. Curr Diabetes Rev. 2017;13(6):560-565. doi:10.2174/1573399812666160614123226

Cheng

Huang

Deng

Wang

. Diabetes as a risk factor for dementia and mild cognitive impairment: a meta-analysis of longitudinal studies. Intern Med J. 2012;42(5):484-491. doi:10.1111/j.1445–5994.2012.02758.x

Cukierman

Gerstein

Williamson

. Cognitive decline and dementia in diabetes – systematic overview of prospective observational studies. Diabetologia. 2005;48(12):2460-2469. doi:10.1007/s00125-005-0023-4

Dik

Jonker

Comijs

, et al. Contribution of metabolic syndrome components to cognition in older individuals. Diabetes Care. 2007;30(10):2655-2660. doi:10.2337/dc06-1190

Yamagishi

S-I

Imaizumi

. Diabetic vascular complications: pathophysiology, biochemical basis and potential therapeutic strategy. Cur Phar Des. 2005;11(18):2279-2299.

Moulton

Costafreda

Horton

Ismail

CHY

. Meta-analyses of structural regional cerebral effects in type 1 and type 2 diabetes. Brain Imaging Behav. 2015;9(4):651-662. doi:10.1007/s11682-014-9348-2

10.

Van Harten

Oosterman

Potter Van Loon

Scheltens

Weinstein

. Brain lesions on MRI in elderly patients with type 2 diabetes mellitus. Eur Neurol. 2007;57(2):70-74. doi:10.1159/000098054

11.

Van Harten

de Leeuw

F-E

Weinstein

Scheltens

Biessels

. Brain imaging in patients with diabetes: a systematic review. Diabetes Care. 2006;29(11):2539-2548. doi:10.2337/dc-06-1637

12.

Reijmer

Brundel

Bresser

J de

, et al. Microstructural white matter abnormalities and cognitive functioning in type 2 diabetes: a diffusion tensor imaging study. Diabetes Care. 2013;36(1):137. doi:10.2337/DC12-0493

13.

Hsu

J-L

Chen

Y-L

Leu

J-G

, et al. Microstructural white matter abnormalities in type 2 diabetes mellitus: a diffusion tensor imaging study. NeuroImage. 2012;59(2):1098-1105. doi:10.1016/j.neuroimage.2011.09.041

14.

Hoogenboom

Marder

Flores

, et al. Cerebral white matter integrity and resting-state functional connectivity in middle-aged patients with type 2 diabetes. Diabetes. 2014;63(2):728-738. doi:10.2337/db13-1219

15.

Cui

Liu

Bian

Wang

. Cortical source multivariate EEG synchronization analysis on amnestic mild cognitive impairment in type 2 diabetes. Sci World J. 2014;2014:523216. doi:10.1155/2014/523216

16.

Hazari

MAH

Ram Reddy

Uzma

Santhosh Kumar

. Cognitive impairment in type 2 diabetes mellitus. Int J Diab Melli. 2015;3(1):19-24. doi:10.1016/j.ijdm.2011.01.001

17.

Kurita

Katayama

Mochio

. Neurophysiological evidence for altered higher brain functions in NIPPM. Diabetes Care. 1996;19(4):361-364.

18.

Mooradian

Perryman

Fitten

Kavonian

Morley

. Cortical function in elderly non-insulin dependent diabetic patients: behavioral and electrophysiologic studies. Arch Int Med. 1988;148(11):2369-2372. http://archinte.jamanetwork.com/

19.

Benwell

CSY

Davila-Pérez

Fried

, et al. EEG Spectral power abnormalities and their relationship with cognitive dysfunction in patients with Alzheimer’s disease and type 2 diabetes. Neurobiol Aging. 2020;85:83-95. doi:10.1016/j.neurobiolaging.2019.10.004

20.

Cooray

Nilsson

Wahlin

Laukka

Brismar

. Effects of intensified metabolic control on CNS function in type 2 diabetes. Psychoneuroendocrinology. 2011;36(1):77-86. doi:10.1016/j.psyneuen.2010.06.009

21.

Bian

Wang

Yin

. Relative power and coherence of EEG series are related to amnestic mild cognitive impairment in diabetes. Front Aging Neurosci. 2014;6(11):11. doi:10.3389/fnagi.2014.00011

22.

Cui

Liu

, et al. A new EEG synchronization strength analysis method: s-estimator based normalized weighted-permutation mutual information. Neural Netw. 2016;82:30-38. doi:10.1016/j.neunet.2016.06.004

23.

Cui

, et al. Magnitude squared coherence method based on weighted canonical correlation analysis for EEG synchronization analysis in amnestic mild cognitive impairment of diabetes mellitus. IEEE Trans Neur Sys and Rehab Eng. 2018;26(10):1908-1917. doi:10.1109/TNSRE.2018.2862396

24.

Hissa

Artur Costa D’almeida

Cremasco

de Bruin

VMS

. Event related P300 potential in NIDDM patients without cognitive impairment and its relationship with previous hypoglycemic episodes. Neuroendocrinol Lett. 2002;23(3):226-230. https://www.researchgate.net/publication/11295416

25.

Mochizuki

Oishi

Hayakawa

Matsuzakl

Takasu

. Improvement of P300 latency by treatment in non-insulin-dependent diabetes Mellitus. Clin Electroencephalogr. 1998;29(4):194-196.

26.

Tandon

Verma

Ram

. Cognitive dysfunction in NIDDM: p3 event related evoked potential study. Indian J Physiol Pharmacol. 1999;43(3):383-388.

27.

Vanhanen

Karhu

Partanen

Laakso

Riekkinen

. ERPs reveal dificits in automatic cerebral stimulus processing in patients with NIDDM. Neuroreport. 1996;7(15–17):2767-2771.

28.

Wen

Bian

Wang

. Resting-state EEG coupling analysis of amnestic mild cognitive impairment with type 2 diabetes mellitus by using permutation conditional mutual information. Clin Neurophysiol. 2016;127(1):335-348. doi:10.1016/j.clinph.2015.05.016

29.

Babiloni

Frisoni

Steriade

, et al. Frontal white matter volume and delta EEG sources negatively correlate in awake subjects with mild cognitive impairment and Alzheimer’s disease. Clin Neurophysiol. 2006;117(5):1113-1129. doi:10.1016/j.clinph.2006.01.020

30.

Fraga

Falk

Kanda

PAM

Anghinah

. Characterizing Alzheimer’s disease severity via resting-awake EEG amplitude modulation analysis. PLoS ONE. 2013;8(8):e72240. doi:10.1371/journal.pone.0072240

31.

Babiloni

Lizio

Marzano

, et al. Brain neural synchronization and functional coupling in Alzheimer’s disease as revealed by resting state EEG rhythms. Int J Psycho. 2016;103:88-102. doi:10.1016/j.ijpsycho.2015.02.008

32.

Babiloni

Carducci

, et al. Movement-related electroencephalographic reactivity in Alzheimer disease. NeuroImage. 2000;12(2):139-146. doi:10.1006/nimg.2000.0602

33.

Bennys

Rondouin

Vergnes

Touchon

. Diagnostic value of quantitative EEG in Alzheimer’s disease. Clin Neurophysiol. 2001;31(3):153-160.

34.

Johnson

. The conversational brain: fronto-hippocampal interaction and disconnection. Med Hypotheses. 2006;67(4):759-764. doi:10.1016/j.mehy.2006.04.031

35.

Jelic

Johansson

S-E

Almkvist

, et al. Quantitative electroencephalography in mild cognitive impairment: longitudinal changes and possible prediction of Alzheimer’s disease. Neurobiol Aging. 2000;21(4):533-540

36.

Neto

Biessmann

Aurlien

Nordby

Eichele

. Regularized linear discriminant analysis of EEG features in dementia patients. Front Aging Neurosci. 2016;8:273. doi:10.3389/FNAGI.2016.00273

37.

Babiloni

Carducci

Lizio

, et al. Resting state cortical electroencephalographic rhythms are related to gray matter volume in subjects with mild cognitive impairment and Alzheimer's disease. Hum Brain Mapp. 2013;34(6):1427-1446. doi:10.1002/hbm.22005

38.

Babiloni

Frisoni

Pievani

, et al. Hippocampal volume and cortical sources of EEG alpha rhythms in mild cognitive impairment and Alzheimer disease. NeuroImage. 2009;44(1):123-135. doi:10.1016/j.neuroimage.2008.08.005

39.

Dubois

Feldman

Jacova

, et al. Revising the NINCDS-ADRDA criteria. Lancet Neurol. 2007;6(8):734-746. doi:10.1016/S1474

40.

Moretti

Frisoni

, … MP-J group. Cerebrovascular disease and hippocampal atrophy are differently linked to functional coupling of brain areas: an EEG coherence study in MCI subjects. J Alzheimer’s Disease. 2008;14(3):285-299.

41.

Jeong

. EEG Dynamics in patients with Alzheimer’s disease. Clin Neurophysiol. 2004;115(7):1490-1505. doi:10.1016/j.clinph.2004.01.001

42.

Babiloni

Frisoni

Pievani

, et al. White-matter vascular lesions correlate with alpha EEG sources in mild cognitive impairment. Neuropsychologia. 2008;46(6):1707-1720. doi:10.1016/j.neuropsychologia.2008.03.021

43.

Lanzino

Wang

. Cerebrovascular disease and cognitive dysfunction. Neurological Report. 2012;24(4):331-336.

44.

Mueller

Schuff

Yaffe

Madison

Miller

Weiner

. Hippocampal atrophy patterns in mild cognitive impairment and Alzheimer’s disease. Hum Brain Mapp. 2010;31(9):1339-1347. doi:10.1002/hbm.20934

45.

Patel

Azzam

. Characterization of N200 and P300: selected studies of the event-related potential. Int J Med Sci. 2005;2(4):147-154.

46.

Paitel

Samii

Nielson

. A systematic review of cognitive event-related potentials in mild cognitive impairment and Alzheimer’s disease. Behav Brain Res. 2021;396:112904. doi:10.1016/j.bbr.2020.112904

47.

Kodl

Seaquist

. Cognitive dysfunction and diabetes Mellitus. Endocr Rev. 2008;29(4):494-511. doi:10.1210/er.2007

48.

Jongen

van der Grond

Kappelle

Biessels

Viergever

Pluim

JPW

. Automated measurement of brain and white matter lesion volume in type 2 diabetes mellitus. Diabetologia. 2007;50(7):1509-1516. doi:10.1007/s00125-007-0688-y

49.

Remijn

Hasuo

Fujihira

Morimoto

. An introduction to the measurement of auditory event-related potentials (ERPs). Acoust Sci Technol. 2014;35(5):225-242. doi:10.1250/ast.35.229