Type 2 diabetes is associated with cognitive decrements, accelerated cognitive decline, and increased risk for dementia. Patients with the metabolic syndrome, a major risk factor for diabetes, may display comparable cognitive decrements as seen in type 2 diabetes. Currently, the impact of diabetes and prediabetes on cognition and the underlying organization of functional brain networks still remain to be elucidated. This study investigated whether functional brain networks are affected in type 2 diabetes and prediabetes. Forty-seven participants with diabetes, 47 participants with prediabetes, and 45 control participants underwent detailed cognitive testing and 3-Tesla resting state functional MRI. Graph theoretical network analysis was performed to investigate alterations in functional cerebral networks. Participants with diabetes displayed altered network measures, characterized by a higher normalized cluster coefficient and higher local efficiency, compared with control participants. The network measures of the participants with prediabetes fell between those with diabetes and control participants. Lower processing speed was associated with shorter path length and higher global efficiency. Participants with type 2 diabetes have altered functional brain networks. This alteration is already apparent in the prediabetic stage to a somewhat lower level, hinting at functional reorganization of the cerebral networks as a compensatory mechanism for cognitive decrements.

The worldwide prevalence of diabetes is increasing rapidly, with the majority of patients having type 2 diabetes (1). Along with cardiovascular risk factors, type 2 diabetes is associated with cognitive decrements, accelerated cognitive decline, and an increased risk for dementia and Alzheimer disease (2,3). A broad range of cognitive domains are affected in type 2 diabetes, and one of the most commonly affected is processing speed (4,5). However, the underlying pathological mechanism is not yet clear.

The progression of normal glucose metabolism to type 2 diabetes is a gradual process in which insulin resistance plays a crucial role. Before the clinical presentation of type 2 diabetes, insulin resistance often is accompanied by other metabolic and vascular abnormalities. The cluster of these cardiovascular risk factors is referred to as the metabolic syndrome and is considered a major risk factor for diabetes (6). Patients with the metabolic syndrome have a high likelihood for diabetes and may display comparable cognitive decrements as seen in patients with type 2 diabetes (5). Furthermore, the cardiovascular risk factors are associated with an increased risk of late-life dementia (7).

Previous studies on cognitive decrements in diabetes using functional MRI (fMRI) have focused on changes in activation patterns, functional connectivity, and signal fluctuations. These studies have revealed associations among impaired cognition, altered activation, and decreased functional connectivity (814). Although functional connectivity was measured by correlating signal time series of various and a priori–selected cerebral regions, aberrant connections outside these predefined regions remain undiscovered. Moreover, the broad range of affected cognitive domains in type 2 diabetes suggests more widespread, global network disturbances rather than perturbations of merely localized processes or specific functional networks. Because several domains are affected in type 2 diabetes, the decreased cognitive performance in individuals with type 2 diabetes might be the result of a reduced efficiency or effectiveness of various processing resources (15). In this context, novel brain connectivity analyses that examine the integrity of networks in the entire brain are appropriate.

Disturbances in local and global network organization (or efficiency) can be assessed by graph theoretical analysis (1618). In general, the brain can be thought of as a graph, which is a collection of nodes (i.e., regions) and edges (i.e., functional connections between two regions). The organization of this graph can be characterized in terms of various so-called graph measures, including the cluster coefficient, local efficiency, characteristic path length, and global efficiency (Table 1). The cluster coefficient is a measure of local network connectivity and quantifies the extent that neighboring brain regions are also connected. A network with a high cluster coefficient contains densely connected local clusters. A measure that is closely related to the cluster coefficient is the local efficiency, which reflects the average efficiency of local clusters. The characteristic path length is a measure of global network connectivity and represents the shortest average number of connections between any two brain regions. A network is considered highly efficient if the path length is relatively short, and this is reflected in the graph measure global efficiency. It has been shown that the brain network organization typically behaves like a small-world network (18) and that this organization is disrupted in Alzheimer disease (19). Small-world networks are characterized by a high cluster coefficient and a low characteristic path length (16) and are more efficient in synchronizing neural activity between brain regions (20).

How cognitive decrements in diabetes reflect alterations in functional brain organization needs to be resolved. In particular, to what extent diabetes and prediabetes (i.e., the metabolic syndrome with impaired glucose tolerance) share the same alterations in functional brain networks or whether prediabetic alterations exist is unknown. Therefore, the current study investigated whether and to what extent global functional brain networks are affected in patients with type 2 diabetes and prediabetes.

Study Population and Design

We used data from The Maastricht Study, an ongoing observational, prospective, population-based cohort study. The rationale and methodology have been described previously (21). In brief, the study focuses on the etiology, pathophysiology, complications, and comorbidities of type 2 diabetes and is characterized by an extensive phenotyping approach. Eligible for participation were all individuals aged 40–75 years and living in the southern part of the Netherlands. Participants were recruited through mass media campaigns and from municipal registries and the regional Diabetes Patient Registry through mailings. Recruitment was stratified according to known type 2 diabetes status for reasons of efficiency. The present report is derived from cross-sectional data from the first 3,451 participants who completed the baseline survey between November 2010 and September 2013. The baseline surveys for each participant were performed within a 3-month time window. MRI was performed ∼12.7 ± 3.7 months later. The study was approved by the institutional medical ethical committee (NL31329.068.10) and the Netherlands Health Council under the Dutch Law for Population Studies (Permit 131088-105234-PG). All participants gave written informed consent.

Of the 3,451 participants, MRI measurements were available in 1,109 (MRI assessments started in December 2013 and are ongoing). From this group, we selected all participants with both the metabolic syndrome and impaired glucose tolerance (which was chosen to represent prediabetes). Forty-seven (out of 1,109) participants met both criteria; hence, we matched them one to one according to age, sex, and education level to the same number of participants with type 2 diabetes and participants without type 2 diabetes, the metabolic syndrome, or prediabetes. When this was not possible, we matched participants within the smallest age range (with a maximum of ±3 years) and/or within an education range of one level. This resulted in 47 participants with prediabetes with both the metabolic syndrome and impaired glucose tolerance, 47 participants with type 2 diabetes, and 47 participants without type 2 diabetes or the metabolic syndrome. No incidental findings at radiological examination were reported in the selected participants.

Glucose Metabolism Status

All participants (except those who used insulin) underwent a standardized oral glucose tolerance test. Participants were considered to have diabetes according to the World Health Organization 2006 criteria if they use diabetes medication or if they had a fasting blood glucose ≥7.0 mmol/L or a 2-h blood glucose ≥11.1 mmol/L. Participants were considered to have both an impaired glucose tolerance and the metabolic syndrome if they had a fasting blood glucose <7.0 mmol/L and a 2-h blood glucose ≥7.8 and <11.1 mmol/L, if they did not use any diabetes medication or did not have type 2 diabetes, and if they met two or more of the following criteria according to the adapted Adult Treatment Panel III criteria (22): waist circumference ≥88 cm for women and ≥102 cm for men, triglycerides ≥1.7 mmol/L, HDL cholesterol <1.3 mmol/L for women and <1.03 mmol/L for men, and blood pressure ≥130/85 mmHg or antihypertensive medication use. Participants without type 2 diabetes or prediabetes were defined by fasting blood glucose <6.1 mmol/L and a 2-h blood glucose <7.8 mmol/L and no use of diabetes medication.

MRI

For each participant, MRI data were acquired on a 3-T clinical magnetic resonance scanner (Magnetom Prisma Fit; Siemens Healthcare, Erlangen, Germany) by using a 64-element head/neck coil for parallel imaging with an acceleration factor of 2. The MRI protocol comprised structural scans for radiological evaluation (including T1-weighted, T2-weighted, and fluid attenuated inversion recovery sequences) and resting state fMRI scans. Resting state fMRI data were acquired using echoplanar imaging pulse sequence sensitive to blood oxygen level–dependent contrast (repetition time 2,000 ms, echo time 29 ms, flip angle of 90°, pixel size 2 × 2 mm, slice thickness 4 mm) and 195 volumes per acquisition. For anatomic references, a T1-weighted three-dimensional magnetization-prepared rapid acquisition gradient echo (repetition time 2,300 ms, echo time 2.98 ms, isotropic voxel size 1.00 mm, 176 continuous slices, matrix size 256 × 240, field of view 256 mm, inversion time 900 ms) was acquired.

Data Analysis

Image Preprocessing

The T1 images were automatically segmented to obtain cortical and subcortical gray and white matter areas and cerebrospinal fluid (CSF) by using the FreeSurfer software package (Martinos Center for Biomedical Imaging, Boston, MA) (23), and the segmentations were visually inspected. Data from two control participants were excluded due to image artifacts preventing reliable FreeSurfer segmentation; thus, data from 47 participants with type 2 diabetes, 47 participants with prediabetes, and 45 control participants were suitable for the final analysis.

Image preprocessing of the time series of the fMRI data were performed in Statistical Parametric Mapping 8 software (The Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College London, London, U.K.). fMRI images were corrected for slice timing and realigned to the mean image to correct for head movement. These images were then transformed to T1 space and spatially smoothed by using a 4-mm full-width half-maximum Gaussian kernel. Next, the individual anatomical T1 segmentations were registered to the fMRI images (24) to calculate the average time series for the cortical and subcortical gray and white matter areas and CSF. To reduce the effect of physiological noise (respiratory and cardiac artifacts), the fMRI data were filtered by using a band pass filter (0.01–0.1 Hz) and linear regression with the averaged time series of white matter, CSF, and movement parameters acquired in the previous realignment step as nuisance regressors (25,26). Finally, the averaged time series of the cortical and subcortical segmented areas (27,28) (a total of 82 areas, 41 for each hemisphere) was used to form a correlation matrix by calculating the Pearson correlation coefficients between all pairs of the segmented brain areas. Negative values were set to 0 (29). The correlation matrix stands for a brain graph in which each brain area represents a node and each connection between brain areas i and j represents an edge. For each participant, the brain graph was thresholded by a sparsity value (i.e., which percentage of the connections with the highest correlation coefficient were included while removing all other connections) based on the mean correlation matrix of the control participants. In addition, for a given sparsity value, the network (and, thus, the number of edges) was the same for each participant. Because no criteria are available for which sparsity value is the most biologically meaningful, a sparsity range of 0.1–0.9 with a step size of 0.01 was chosen, corresponding to 664–5,978 connections. A high sparsity value represents a relatively low number of connections with the highest correlation coefficients.

Network Parameters

For the global organization of large-scale networks, graph theoretical network measures (1618) were calculated from the brain graphs by using the Brain Connectivity Toolbox (30) in MATLAB (The MathWorks, Natick, MA). The following graph theoretical measures were calculated (Table 1): 1) the cluster coefficient, which is a measure of local network connectivity (i.e., a network with a high cluster coefficient is characterized by densely connected local clusters); 2) the characteristic path length, a measure of the average step length between two nodes (i.e., a network with low path length is characterized by short distances between two nodes); 3) global efficiency, a measure of the average inverse shortest path length and is inversely related to the path length; and 4) local efficiency, a measure of the mean global efficiency of subgraphs computed on the immediate neighbors of a node and related to the cluster coefficient. Furthermore, for each participant, the graph measures were normalized to comparable values from randomly generated networks (N = 100) (16,17) to evaluated whether the network had small-world properties. Small-world networks are characterized by a high normalized cluster coefficient >1 and a low normalized characteristic path length of ∼1.

Covariates

Educational level was assessed by interview and classified into eight levels commonly used in the Netherlands: 1) no education, 2) primary education, 3) lower vocational education, 4) intermediate general secondary education, 5) intermediate vocational education, 6) higher general secondary education, 7) higher vocational education, and 8) university degree. For this study, educational level was divided into three groups: low (levels 1–3), middle (levels 4–6), and high (levels 7 and 8). Office blood pressure was assessed three times on the right arm after a 10-min rest by using a noninvasive blood pressure monitor (Omron 705IT; Omron, Kyoto, Japan). A fourth measurement was performed when the difference between measurements two and three exceeded >10 mmHg. Here, we used the averaged blood pressure values over all the available measurements (21).

Cognition

Information processing speed was derived (z score over the 3,451 participants) from the Stroop Color Word Test (parts I and II) (31), Concept Shifting Test (parts A and B) (32), and the Letter-Digit Substitution Test (33) in which participants were instructed to match digits to letters as quickly as possible within 90 s. Global cognitive function was measured by the Mini-Mental State Examination (score range 0–30) (34). Memory function was derived from the immediate and delayed recall Verbal Word Learning Test (35) in which 15 words are presented in five subsequent trials followed by a recall phase immediately after each trial (score range 0–75) and a delayed recall phase after 20 min (score range 0–15). Executive functioning and attention was derived from the Stroop Color Word Test (part III) (31) and Concept Shifting Test (part C) (32).

Statistical Analysis

Descriptive participant characteristics are reported as mean ± SD. Group characteristics were tested using one-way ANOVA for continuous variables and Pearson χ2 test for categorical variables. Post hoc two-sided independent-samples t and Pearson χ2 tests were used for significant group differences. Linear regression analyses were performed to assess the association of the graph theoretical network measures with the type 2 diabetes, prediabetes, and control conditions (model 1). In addition, to evaluate whether the associations were robust, model 1 was extended with potential confounders (i.e., age, sex, education level, systolic blood pressure) (model 2). Finally, to assess whether the graph measures were associated with cognition, model 2 was further extended with information processing speed (model 3). Processing speed was chosen for this analysis because it is mediated by multiple brain regions globally distributed throughout the entire brain. Furthermore, model 2 was extended with the addition of alcohol consumption to assess the effect of alcohol on the graph measures (model 4). Additionally, linear regression analyses with continuous measures for fasting blood glucose levels, HbA1c levels, and insulin levels instead of the dichotomous prediabetes and diabetes status in model 1 were performed.

Table 2 shows the general characteristics of the participants. The groups were matched on age, sex, and education. As expected, the groups had different fasting blood glucose levels, HbA1c levels, BMI, waist circumference, systolic blood pressure, and cholesterol levels. Cognition scores were not significantly different; however, post hoc analysis showed that participants with type 2 diabetes had significantly lower scores on information processing speed than control participants.

Figure 1 displays the graph theoretical network measures over a wide range of sparsity values for the three groups. The networks of the three groups all exhibited a topology in the small-world regimen, with cluster coefficient >1 and characteristic path length close to 1 (Fig. 1A and B). Participants with type 2 diabetes displayed generally higher values for cluster coefficient and global and local efficiency over almost the entire sparsity range compared with control participants, and participants with prediabetes were between these two groups. Moreover, in participants with type 2 diabetes, these cluster coefficient and local efficiency values were significantly higher by sparsity values (0.77–0.88 and 0.90, and 0.62–0.90, respectively) compared with that in control participants. For participants with prediabetes, local efficiency was also significantly higher over a wide range of sparsity values (0.73–0.90) compared with control participants. For characteristic path length, the control group displayed higher values than the other two groups, which were significantly higher for a narrow range of sparsity values (0.83–0.86) compared with the prediabetes group. No significant differences in graph measures were found between participants with type 2 diabetes and prediabetes. Table 3 shows the results of the linear regression analysis performed for the graph measures corresponding to a sparsity value of 0.8. Supplementary Fig. 1 shows the relative (to the control participants) differences in graph measures for the type 2 diabetes and prediabetes groups. All results were intact after adjusting for age, sex, education level, and systolic blood pressure (model 2) (Supplementary Table 1).

Additional analyses incorporating information processing speed (model 3) revealed that increased characteristic path length was associated with improved processing speed, whereas an increased global efficiency was associated with lower processing speed (Supplementary Fig. 2). In addition, alcohol consumption was not significantly associated with any graph measures (model 4). Analyses with fasting blood glucose levels and HbA1c levels showed similar results for the local efficiency: High fasting blood glucose and high HbA1c levels were significantly associated with a high local efficiency. Both measures showed a trend for the cluster coefficient: High fasting blood glucose and high HbA1c levels were associated with a higher cluster coefficient. The other graph measures did not show a significant association. With respect to insulin levels, no consistent results were obtained because only a few scattered sparsity values displayed significant associations with characteristic path length and global efficiency.

This cross-sectional study is the first to evaluate global functional brain networks by exploring various graph measures in participants with type 2 diabetes, participants with prediabetes, and matched control participants in one study. Graph theoretical network measures were compared among these groups, and the relation with processing speed was investigated. First, participants with type 2 diabetes displayed altered network parameters, including a higher normalized cluster coefficient and a higher local efficiency than control participants. In addition, participants with prediabetes displayed higher normalized local efficiency than control participants, although these values were qualitatively lower than in participants with type 2 diabetes. Furthermore, no significant differences in network parameters were observed between prediabetes and type 2 diabetes, and all groups exhibited small-world organization of their networks indicative of an efficient cerebral topology. Cognitive performance was similar for the three groups, and lower information processing speed was associated with a lower normalized characteristic path length and a higher global efficiency.

In normal brain networks, the combination of a high cluster coefficient and a high local efficiency reflects a high local specialization of the brain in processing information. We observed this combination in participants with type 2 diabetes, suggesting that the brain is better organized than the cerebral networks of the control participants. In view of the known effect of diabetes on cognition, such as cognitive decrements and higher risk of dementia or Alzheimer disease, these results may seem contradictory because one could expect that the cerebral networks are less, rather than better, organized in diabetes. The diabetes population in this study is relatively healthy in terms of their mildly affected cognitive performance (in the range of normal performance) and in terms of good treatment control according to glucose levels. A better-organized cerebral network in diabetes can be attributed to an earlier stage of structural brain damage, where compensatory mechanisms, such as functional reorganization of networks, might play a role (36). In addition, all four network measures investigated in the current study showed the same direction, thus supporting an underlying compensatory mechanism. To further support the compensatory interpretation, additional analysis showed no effect of age (data not shown), which is known to have an impact on the network parameters (37,38). Of note, increased functional connectivity was previously observed in patients with type 1 diabetes without retinopathy compared with healthy control subjects (39). In addition, graph theoretical network analysis of diffusion data have already shown disrupted white matter networks in type 2 diabetes (40), and it will be interesting to investigate the relationship with the current findings.

The findings with respect to the participants with prediabetes, who showed intermediate topological values of the functional networks, also strengthen the explanation of an underlying compensatory mechanism. Therefore, a higher local efficiency might be an early marker for cognitive decrements that may eventually lead to brain degeneration. It was previously shown that patients with Alzheimer disease showed a decline of the small-world network organization characterized by a decreased normalized cluster coefficient and a decreased local efficiency, indicative of disrupted local network connectivity (19). Taken together, a hypothetical overview over time (Fig. 2) illustrates that before the onset of clinically manifested cognitive decrements, functional reorganization of the cerebral networks may already have started as a compensatory mechanism to counteract the slight decrements in cognitive performance in participants with prediabetes and type 2 diabetes. Then, when the functional reorganization fails, functional networks become disrupted, and the first signs of cognitive decrements will be recognizable in the participants with diabetes. Finally, this might possibly explain the increased risk to develop mild cognitive impairment and Alzheimer disease in a later stage. It remains to be explored how the graph theoretical network measures behave in patients with diabetes with more severe cognitive decrements or mild cognitive impairment (3) (Fig. 2, gray area) because those results could support the presumed underlying compensatory mechanism.

In normal brain networks, the combination of a shorter characteristic path length and a higher global efficiency reflects the great ability of the brain to globally integrate information. We observed that this combination was associated with less processing speed and that participants with type 2 diabetes scored lower on processing speed than control participants (Table 2). These results also fully complement the hypothetical underlying compensatory mechanism (Fig. 2).

Supplementary Fig. 3 plots local efficiency as a function of fasting blood glucose level. It can be seen that local efficiency increases with blood glucose level. As the results indicate, the effect of a high blood glucose level on local efficiency is already apparent in the prediabetic stage. Moreover, additional analysis with continuous fasting blood glucose and HbA1c levels instead of the dichotomous diabetes and prediabetes status showed that higher local efficiency was significantly associated with higher fasting blood glucose and HbA1c levels. Similar analyses with insulin did not yield comparable results because participants with diabetes had excellent control and displayed even lower insulin levels than participants with prediabetes (Table 2). This is consistent with other studies suggesting that hyperinsulinemia alone may not be a strong determinant of cardiovascular disease (41), which might possibly also hold for cerebral alterations. Because improved glycemic control in type 2 diabetes is associated with improved cognition (42), good blood glucose management in the prediabetes stage seems essential to reduce its impact on the brain. In this way, the onset of cognitive decrements may be postponed or, hopefully, halted. We should not neglect, however, the additional role of the other metabolic syndrome factors on local efficiency. A post hoc analysis revealed that participants with type 2 diabetes who also meet the criteria for the metabolic syndrome displayed even higher local efficiency values (data not shown). Moreover, the effect of the medication on the graph theoretical network measures in diabetes should be investigated in future studies.

The blood oxygen level–dependent signal is likely affected in type 2 diabetes because of an impaired vascular reactivity as a consequence of an altered hemodynamic response and changes in neurovascular coupling (43). How exactly changes in the hemodynamic response translate to changes in functional connectivity and network efficiency remains to be elucidated.

This study has several strengths. First, to our knowledge, the study is the first to investigate graph theoretical network measures in participants with diabetes and prediabetes compared with matched control participants. Second, the participants were extensively characterized. Because correction for age, sex, and blood pressure did not change the significant results, we are confident that these factors do not underlie the observed differences among groups.

The study also has some limitations. First, it had a cross-sectional design; therefore, the results should be interpreted cautiously in terms of causality. In addition, the time span between the biochemical measurements and MRI acquisition of an average 12.7 months might limit the validity of the participant characteristics, although it is expected that the number of substantial biochemical alterations would be low (44). Nevertheless, the first results provide new insights and prompt new questions for future studies to investigate the behavior of network measures over time. Second, because of the inclusion of a relatively healthy study population, we were not able to investigate the full range of the hypothetical curve because participants with severe diabetes complications did not participate in the study. Therefore, an investigation of the functional network properties of participants with diabetes and severe complications would be interesting. Third, numerous other factors of potential influence that deserve attention could not be investigated, for instance, the effects of sleep, fasting, estrogen/hormone medication, therapy compliance, and disease duration (4548). Finally, the effect of diabetes medication within the diabetes group (treated vs. untreated) should be investigated in future studies.

To conclude, this study reveals altered functional brain networks in patients with type 2 diabetes as well as in patients with prediabetes, indicating a more-efficient cerebral organization. This was observed in patients with mild cognitive decrements, which are still in the range of normal cognitive functioning. The functional network measures of the prediabetes group were intermediate between the type 2 diabetes and control groups, hinting at compensation for cognitive decrements in terms of functional reorganization of the cerebral networks. Furthermore, the study shows that functional graph theoretical network measures provide new insights and may serve as an early MRI biomarker for subtle cerebral alterations in relation to cognitive decline.

Acknowledgments. The authors thank Marc Geerlings and Jos Slenter (Radiology and Nuclear Medicine, Maastricht University Medical Center, Maastricht, the Netherlands) for continuous hardware and software support.

Funding. J.F.A.J. was funded by Veni research from the Netherlands Organisation for Scientific Research and the Netherlands Organisation for Health Research and Development (grant 916.11.059). Additionally, this work was supported by Stichting de Weijerhorst (Maastricht, the Netherlands). The Maastricht Study was supported by the European Regional Development Fund via OP-Zuid, the Province of Limburg, the Ministry of Economic Affairs, the Netherlands (grant 31O.041), Stichting de Weijerhorst, Pearl String Initiative Diabetes (Amsterdam, the Netherlands), Cardiovascular Center (Maastricht), Cardiovascular Research Institute Maastricht School for Cardiovascular Diseases (Maastricht), CAPHRI School for Public Health and Primary Care (Maastricht), NUTRIM School for Nutrition and Translational Research in Metabolism (Maastricht), Stichting Annadal (Maastricht), Health Foundation Limburg (Maastricht), and unrestricted grants from Janssen-Cilag B.V. (Tilburg, the Netherlands), Novo Nordisk Farma B.V. (Alphen aan den Rijn, the Netherlands), and Sanofi Netherlands B.V. (Gouda, the Netherlands).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. F.C.G.v.B., W.H.B., and J.F.A.J. researched data and contributed to the discussion and writing, review, editing, and final approval of the manuscript. T.M.v.V., P.A.M.H., M.P.J.v.B., M.T.S., S.J.S.S., P.C.D., N.S., C.D.A.S., and J.E.W. contributed to the discussion and critical review and editing and final approval of the manuscript. F.C.G.v.B. and J.F.A.J. are the guarantors of this work, and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Polonsky
KS
.
The past 200 years in diabetes
.
N Engl J Med
2012
;
367
:
1332
1340
[PubMed]
2.
Spauwen
PJ
,
Köhler
S
,
Verhey
FR
,
Stehouwer
CD
,
van Boxtel
MP
.
Effects of type 2 diabetes on 12-year cognitive change: results from the Maastricht Aging Study
.
Diabetes Care
2013
;
36
:
1554
1561
[PubMed]
3.
Biessels
GJ
,
Staekenborg
S
,
Brunner
E
,
Brayne
C
,
Scheltens
P
.
Risk of dementia in diabetes mellitus: a systematic review
.
Lancet Neurol
2006
;
5
:
64
74
[PubMed]
4.
Messier
C
.
Impact of impaired glucose tolerance and type 2 diabetes on cognitive aging
.
Neurobiol Aging
2005
;
26
(
Suppl. 1
):
26
30
[PubMed]
5.
van den Berg
E
,
Dekker
JM
,
Nijpels
G
, et al
.
Cognitive functioning in elderly persons with type 2 diabetes and metabolic syndrome: the Hoorn study
.
Dement Geriatr Cogn Disord
2008
;
26
:
261
269
[PubMed]
6.
Lorenzo
C
,
Okoloise
M
,
Williams
K
,
Stern
MP
,
Haffner
SM
;
San Antonio Heart Study
.
The metabolic syndrome as predictor of type 2 diabetes: the San Antonio Heart Study
.
Diabetes Care
2003
;
26
:
3153
3159
[PubMed]
7.
Whitmer
RA
,
Sidney
S
,
Selby
J
,
Johnston
SC
,
Yaffe
K
.
Midlife cardiovascular risk factors and risk of dementia in late life
.
Neurology
2005
;
64
:
277
281
[PubMed]
8.
Chen
Y
,
Liu
Z
,
Zhang
J
, et al
.
Altered brain activation patterns under different working memory loads in patients with type 2 diabetes
.
Diabetes Care
2014
;
37
:
3157
3163
[PubMed]
9.
Cui
Y
,
Jiao
Y
,
Chen
YC
, et al
.
Altered spontaneous brain activity in type 2 diabetes: a resting-state functional MRI study
.
Diabetes
2014
;
63
:
749
760
[PubMed]
10.
Xia
W
,
Wang
S
,
Sun
Z
, et al
.
Altered baseline brain activity in type 2 diabetes: a resting-state fMRI study
.
Psychoneuroendocrinology
2013
;
38
:
2493
2501
[PubMed]
11.
Zhou
X
,
Zhang
J
,
Chen
Y
, et al
.
Aggravated cognitive and brain functional impairment in mild cognitive impairment patients with type 2 diabetes: a resting-state functional MRI study
.
J Alzheimers Dis
2014
;
41
:
925
935
[PubMed]
12.
Zhang
H
,
Hao
Y
,
Manor
B
, et al
.
Intranasal insulin enhanced resting-state functional connectivity of hippocampal regions in type 2 diabetes
.
Diabetes
2015
;
64
:
1025
1034
[PubMed]
13.
Zhou
H
,
Lu
W
,
Shi
Y
, et al
.
Impairments in cognition and resting-state connectivity of the hippocampus in elderly subjects with type 2 diabetes
.
Neurosci Lett
2010
;
473
:
5
10
[PubMed]
14.
Musen
G
,
Jacobson
AM
,
Bolo
NR
, et al
.
Resting-state brain functional connectivity is altered in type 2 diabetes
.
Diabetes
2012
;
61
:
2375
2379
[PubMed]
15.
Koekkoek
PS
,
Kappelle
LJ
,
van den Berg
E
,
Rutten
GE
,
Biessels
GJ
.
Cognitive function in patients with diabetes mellitus: guidance for daily care
.
Lancet Neurol
2015
;
14
:
329
340
[PubMed]
16.
Watts
DJ
,
Strogatz
SH
.
Collective dynamics of ‘small-world’ networks
.
Nature
1998
;
393
:
440
442
[PubMed]
17.
Stam
CJ
,
Reijneveld
JC
.
Graph theoretical analysis of complex networks in the brain
.
Nonlinear Biomed Phys
2007
;
1
:
3
[PubMed]
18.
Bullmore
E
,
Sporns
O
.
Complex brain networks: graph theoretical analysis of structural and functional systems
.
Nat Rev Neurosci
2009
;
10
:
186
198
[PubMed]
19.
Supekar
K
,
Menon
V
,
Rubin
D
,
Musen
M
,
Greicius
MD
.
Network analysis of intrinsic functional brain connectivity in Alzheimer’s disease
.
PLOS Comput Biol
2008
;
4
:
e1000100
[PubMed]
20.
Latora
V
,
Marchiori
M
.
Efficient behavior of small-world networks
.
Phys Rev Lett
2001
;
87
:
198701
[PubMed]
21.
Schram
MT
,
Sep
SJ
,
van der Kallen
CJ
, et al
.
The Maastricht Study: an extensive phenotyping study on determinants of type 2 diabetes, its complications and its comorbidities
.
Eur J Epidemiol
2014
;
29
:
439
451
[PubMed]
22.
Grundy
SM
,
Brewer
HB
 Jr
,
Cleeman
JI
,
Smith
SC
 Jr
,
Lenfant
C
;
American Heart Association
;
National Heart, Lung, and Blood Institute
.
Definition of metabolic syndrome: report of the National Heart, Lung, and Blood Institute/American Heart Association conference on scientific issues related to definition
.
Circulation
2004
;
109
:
433
438
[PubMed]
23.
Fischl
B
,
Salat
DH
,
Busa
E
, et al
.
Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain
.
Neuron
2002
;
33
:
341
355
[PubMed]
24.
Jenkinson
M
,
Smith
S
.
A global optimisation method for robust affine registration of brain images
.
Med Image Anal
2001
;
5
:
143
156
[PubMed]
25.
Van Dijk
KR
,
Hedden
T
,
Venkataraman
A
,
Evans
KC
,
Lazar
SW
,
Buckner
RL
.
Intrinsic functional connectivity as a tool for human connectomics: theory, properties, and optimization
.
J Neurophysiol
2010
;
103
:
297
321
[PubMed]
26.
Vlooswijk
MC
,
Vaessen
MJ
,
Jansen
JF
, et al
.
Loss of network efficiency associated with cognitive decline in chronic epilepsy
.
Neurology
2011
;
77
:
938
944
[PubMed]
27.
Desikan
RS
,
Ségonne
F
,
Fischl
B
, et al
.
An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest
.
Neuroimage
2006
;
31
:
968
980
[PubMed]
28.
Cappellani
R
,
Bergsland
N
,
Weinstock-Guttman
B
, et al
.
Subcortical deep gray matter pathology in patients with multiple sclerosis is associated with white matter lesion burden and atrophy but not with cortical atrophy: a diffusion tensor MRI study
.
AJNR Am J Neuroradiol
2014
;
35
:
912
919
[PubMed]
29.
Smith
SM
,
Miller
KL
,
Salimi-Khorshidi
G
, et al
.
Network modelling methods for FMRI
.
Neuroimage
2011
;
54
:
875
891
[PubMed]
30.
Rubinov
M
,
Sporns
O
.
Complex network measures of brain connectivity: uses and interpretations
.
Neuroimage
2010
;
52
:
1059
1069
[PubMed]
31.
Van der Elst
W
,
Van Boxtel
MP
,
Van Breukelen
GJ
,
Jolles
J
.
The Stroop color-word test: influence of age, sex, and education; and normative data for a large sample across the adult age range
.
Assessment
2006
;
13
:
62
79
[PubMed]
32.
Van der Elst
W
,
Van Boxtel
MP
,
Van Breukelen
GJ
,
Jolles
J
.
The Concept Shifting Test: adult normative data
.
Psychol Assess
2006
;
18
:
424
432
[PubMed]
33.
van der Elst
W
,
van Boxtel
MP
,
van Breukelen
GJ
,
Jolles
J
.
The Letter Digit Substitution Test: normative data for 1,858 healthy participants aged 24-81 from the Maastricht Aging Study (MAAS): influence of age, education, and sex
.
J Clin Exp Neuropsychol
2006
;
28
:
998
1009
[PubMed]
34.
Folstein
MF
,
Folstein
SE
,
McHugh
PR
.
“Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician
.
J Psychiatr Res
1975
;
12
:
189
198
[PubMed]
35.
Van der Elst
W
,
van Boxtel
MP
,
van Breukelen
GJ
,
Jolles
J
.
Rey’s verbal learning test: normative data for 1855 healthy participants aged 24-81 years and the influence of age, sex, education, and mode of presentation
.
J Int Neuropsychol Soc
2005
;
11
:
290
302
[PubMed]
36.
Schoonheim
MM
,
Geurts
JJ
,
Barkhof
F
.
The limits of functional reorganization in multiple sclerosis
.
Neurology
2010
;
74
:
1246
1247
[PubMed]
37.
Grady
CL
.
Cognitive neuroscience of aging
.
Ann N Y Acad Sci
2008
;
1124
:
127
144
[PubMed]
38.
Cao
M
,
Wang
JH
,
Dai
ZJ
, et al
.
Topological organization of the human brain functional connectome across the lifespan
.
Dev Cogn Neurosci
2014
;
7
:
76
93
[PubMed]
39.
van Duinkerken
E
,
Schoonheim
MM
,
Sanz-Arigita
EJ
, et al
.
Resting-state brain networks in type 1 diabetic patients with and without microangiopathy and their relation to cognitive functions and disease variables
.
Diabetes
2012
;
61
:
1814
1821
[PubMed]
40.
Reijmer
YD
,
Leemans
A
,
Brundel
M
,
Kappelle
LJ
,
Biessels
GJ
;
Utrecht Vascular Cognitive Impairment Study Group
.
Disruption of the cerebral white matter network is related to slowing of information processing speed in patients with type 2 diabetes
.
Diabetes
2013
;
62
:
2112
2115
[PubMed]
41.
Gerstein
HC
,
Pais
P
,
Pogue
J
,
Yusuf
S
.
Relationship of glucose and insulin levels to the risk of myocardial infarction: a case-control study
.
J Am Coll Cardiol
1999
;
33
:
612
619
[PubMed]
42.
Ryan
CM
,
Freed
MI
,
Rood
JA
,
Cobitz
AR
,
Waterhouse
BR
,
Strachan
MW
.
Improving metabolic control leads to better working memory in adults with type 2 diabetes
.
Diabetes Care
2006
;
29
:
345
351
[PubMed]
43.
Duarte
JV
,
Pereira
JM
,
Quendera
B
, et al
.
Early disrupted neurovascular coupling and changed event level hemodynamic response function in type 2 diabetes: an fMRI study
.
J Cereb Blood Flow Metab
2015
;
35
:
1671
1680
[PubMed]
44.
Tabák
AG
,
Herder
C
,
Rathmann
W
,
Brunner
EJ
,
Kivimäki
M
.
Prediabetes: a high-risk state for diabetes development
.
Lancet
2012
;
379
:
2279
2290
[PubMed]
45.
Pallayova
M
,
Donic
V
,
Gresova
S
,
Peregrim
I
,
Tomori
Z
.
Do differences in sleep architecture exist between persons with type 2 diabetes and nondiabetic controls
?
J Diabetes Sci Technol
2010
;
4
:
344
352
[PubMed]
46.
Khalsa
S
,
Mayhew
SD
,
Przezdzik
I
, et al
.
Variability in cumulative habitual sleep duration predicts waking functional connectivity
.
Sleep
2016
;
39
:
87
95
[PubMed]
47.
Lisofsky
N
,
Mårtensson
J
,
Eckert
A
,
Lindenberger
U
,
Gallinat
J
,
Kühn
S
.
Hippocampal volume and functional connectivity changes during the female menstrual cycle
.
Neuroimage
2015
;
118
:
154
162
[PubMed]
48.
Arélin
K
,
Mueller
K
,
Barth
C
, et al
.
Progesterone mediates brain functional connectivity changes during the menstrual cycle-a pilot resting state MRI study
.
Front Neurosci
2015
;
9
:
44
[PubMed]

Supplementary data