Brain Connectivity Alterations Are Associated with the Development of Dementia in Parkinson's Disease

Abstract

Dementia affects a high proportion of Parkinson's disease (PD) patients and poses a burden on caregivers and healthcare services. Electroencephalography (EEG) is a common nonevasive and nonexpensive technique that can easily be used in clinical settings to identify brain functional abnormalities. Only few studies had identified EEG abnormalities that can predict PD patients at higher risk for dementia. Brain connectivity EEG measures, such as multiscale entropy (MSE) and phase-locking value (PLV) analyses, may be more informative and sensitive to brain alterations leading to dementia than previously used methods. This study followed 62 dementia-free PD patients for a mean of 3.4 years to identify cerebral alterations that are associated with dementia. Baseline resting state EEG of patients who developed dementia (N = 18) was compared to those of patients who remained dementia-free (N = 44) and of 37 healthy subjects. MSE and PLV analyses were performed. Partial least squares statistical analysis revealed group differences associated with the development of dementia. Patients who developed dementia showed higher signal complexity and lower PLVs in low frequencies (mainly in delta frequency) than patients who remained dementia-free and controls. Conversely, both patient groups showed lower signal variability and higher PLVs in high frequencies (mainly in gamma frequency) compared to controls, with the strongest effect in patients who developed dementia. These findings suggest that specific disruptions of brain communication can be measured before PD patients develop dementia, providing a new potential marker to identify patients at highest risk of developing dementia and who are the best candidates for neuroprotective trials.

Introduction

Although Parkinson's disease (PD) is defined mainly by motor symptoms, cognitive impairment is also very common. About 80% of PD patients will develop dementia in the course of the disease (Aarsland et al., 2003; Buter et al., 2008; Hely et al., 2008). Dementia strikingly affects the quality of life of PD patients and their caregivers, in addition to wielding a considerable socioeconomic impact (Johnson et al., 2013; Keränen et al., 2003; Vossius et al., 2011).

The cerebral mechanisms related to dementia development in PD remain poorly understood. Therapies to prevent or delay dementia in PD are limited, mainly because neuroprotective treatments may be effective only if administered before symptoms appear (Svenningsson et al., 2012). Identifying predictors of dementia and improving the understanding of how neurodegeneration develops are thus essential.

Previous studies have shown a slowing of resting state EEG spectral power in PD with dementia compared with PD patients without dementia (Bonanni et al., 2008; Ponsen et al., 2012), and this slowing was found to be potentially predictive of dementia development (Klassen et al., 2011). Recent studies have started to focus on changes in connectivity and brain dynamics to broaden our understanding of neurodegeneration. A common measure of connectivity is phase coherence analysis, which measures the synchrony of the signal phases between different brain regions. When phases are highly synchronized between two brain regions, these regions are traditionally said to be highly connected.

In PD, cross-sectional EEG and magnetoencephalographic (MEG) studies using synchronization likelihood and phase coherence analysis have reported mixed results (Babiloni et al., 2011; Bosboom et al., 2009; Ponsen et al., 2012; Stoffers et al., 2008), probably due to confounding variables (e.g., small sample sizes, different analyses, comparisons of patients at different PD stages, patients on vs. off medication). In a recent 4-year MEG follow-up study, reduced node clustering for all frequencies and loss of global network efficiency in alpha frequency were related to cognitive decline in PD (Olde Dubbelink et al., 2014a). However, this study included only three PD patients who converted to dementia. The development of brain changes leading to dementia and whether a predictive EEG marker can be identified therefore remain unclear.

An innovative approach to study brain dynamics is to measure brain signal variability, which considers what has generally been viewed as “noise” as part of the brain interactions. One way of measuring brain signal variability is through multiscale entropy (MSE) analysis, which informs about the predictability of the signal at different timescales. Previous studies using MSE have found that variability increases from childhood to adulthood and tends to decrease thereafter with aging, and that a higher variability is positively related to cognitive performance (Garrett et al., 2011; McIntosh et al., 2014). Greater information integration across brain networks could account for increased signal variability measured on the scalp.

EEG studies using different signal variability measures in early PD have found either no difference (Müller et al., 2001) or higher variability (Han et al., 2013; Pezard et al., 2001) compared to controls. An MEG study rather showed lower variability in PD (Gómez et al., 2011). To our knowledge, no study to date has considered changes related to cognitive decline.

This study aims to prospectively follow a cohort of PD patients and identify cerebral anomalies using EEG connectivity techniques in baseline resting state that are associated with the subsequent development of dementia.

Materials and Methods

Subjects

PD patients were recruited from the Department of Neurology of the Montreal General Hospital and the Unité des troubles du mouvement André-Barbeau of the Centre Hospitalier de l'Université de Montréal, Montreal, Canada. They were enrolled in our ongoing study on sleep and cognition in PD at the Centre for Advanced Research in Sleep Medicine, Hôpital du Sacré-Coeur de Montréal, Canada. PD patients were asked to participate in the study irrespective of the presence of a sleep disorder. Control subjects were recruited by word of mouth and through newspaper advertisements. The protocol was approved by the hospital's ethics committee, and all participants gave their written informed consent.

PD patients were diagnosed with idiopathic PD by a movement disorder specialist. Patients who developed dementia at follow-up were classified in the PDD group whereas those who remained dementia-free were classified in the PDnD group. Cognitive status was determined by consensus between the neuropsychologist and neurologist. Dementia diagnosis was established according to DSM-IV-TR criteria based on the neuropsychological assessment (American Psychiatric Association, 2000) and the recommendations of the Movement Disorder Society Task Force for PD patients (Dubois et al., 2007). More specifically, the following criteria were applied: (1) objective evidence of cognitive decline on the neuropsychological assessment (a score of 1.5 standard deviations below the standardized mean on at least two tests in the same cognitive domain) on at least two cognitive domains; and (2) significant functional impact of cognitive impairment on daily living activities. Impact on daily living activities was identified during the interview with patients and their caregiver when available, and it was defined as significant alterations in the ability to manage finances, perform chores, clean the house, prepare meals, do shopping, drive the car or use public transportation, or take medication.

If the patient could not perform the complete cognitive testing at follow-up, dementia diagnosis was based on the neurological assessment including the Mini-Mental State Examination and Montreal Cognitive Assessment (Folstein et al., 1975; Nasreddine et al., 2005). If patients were unable to be assessed in person at follow-up due to severe disability or dementia, a telephone interview was performed with the patient and/or spouse using the Telephone Interview for Cognitive Status (TICS) (Brandt et al., 1988), whenever possible. In some cases, the patient could not be interviewed, simply because they were too cognitively impaired to speak on the telephone. In those cases, we also corresponded directly with the treating physician. To assess whether a diagnostic bias could exist, analyses on PD patients who complete the neuropsychological assessment at follow-up were compared to analyses performed on the entire group.

Mild cognitive impairment (MCI) diagnosis was made according to published criteria (Gagnon et al., 2009; Litvan et al., 2012): (1) a subjective cognitive complaint by the patient or an informant during the structured interview; (2) objective evidence of cognitive decline on the neuropsychological assessment corresponding to a score of 1.5 standard deviations below the standardized mean on at least two tests in the same cognitive domain; (3) preserved activities of daily living based on previous and actual capacities; and (4) absence of medication use or other medical/psychiatric condition that explains cognitive deficits.

Exclusion criteria for both groups were as follows: (1) dementia at baseline; (2) a major psychiatric disorder such as bipolar disorder or schizophrenia, according to DSM-IV-TR criteria (American Psychiatric Association, 2000); (3) abnormal EEG features suggesting epilepsy; and/or (4) a history of stroke, head injury, or cerebrovascular disease. PD patients were on medication during the study (Table 1). Some patients were also taking other antiparkinsonian medication (i.e., amantadine, selegiline; 13 PDnD patients and 2 PDD patients), acetylcholinesterase inhibitors (2 PDD patients), antidepressants (22 PDnD patients and 2 PDD patients), and/or benzodiazepines (10 PDnD patients and 1 PDD patients). Controls were not taking any medication known to affect EEG signal during the study. Thirty-nine PDnD patients and 11 PDD patients also participated in a study on sleep spindles as a predictor of dementia in PD (Latreille et al., 2015).

Table 1.

Baseline Demographic and Clinical Characteristics of Participants

	Controls (n = 37), A	PDnD (n = 44), B	PDD (n = 18), C	p^a	Post hoc
Age, year	66.64 (8.90)	62.98 (7.93)	67.20 (5.95)	<0.001	B < C
Sex, % men	70	64	72	0.736^b
Education, year	13.91 (2.48)	14.86 (3.45)	14.12 (3.02)	0.363
Follow-up duration, year	—	3.26 (1.08)	3.87 (1.58)	0.163
Disease duration, year	—	3.51 (3.10)	4.06 (4.37)	0.596
Hoehn and Yahr stage	—	2.43 (0.94)	2.21 (0.77)	0.391
UPDRS part III “On”	—	23.85 (10.73)	20.76 (11.30)	0.329^b
% Tremor-dominant PD	—	52	35	0.234^b
Levodopa dosage, mg	—	499.73 (331.09)	394.06 (338.35)	0.294
Levodopa use, %	—	73	70	0.552^b
Dopamine agonist use, %	—	32	12	0.098^b
Mild cognitive impairment, %	0	43	82	0.007 ^b	A < B < C
Beck Depression Inventory	5.91 (6.39)	10.00 (6.74)	13.12 (9.86)	0.001 ^c	A < B, C
Beck Anxiety Inventory	3.80 (4.90)	9.90 (6.35)	13.87 (10.52)	<0.001 ^c	A < B, C
Epworth Sleepiness Scale	7.06 (4.10)	9.91 (5.17)	9.86 (4.79)	0.046 ^c	¥

Results are expressed as mean (standard deviation). Boldface indicates significant p values.

p Values for comparisons between Controls, PD patients who remained dementia-free at follow-up (PDnD) and PD patients who developed dementia at follow-up (PDD) or between PDnD and PDD groups.

Pearson's chi-square.

Kruskal–Wallis test with post hoc Mann–Whitney U tests and Bonferroni correction of p < 0.016.

¥: No post hoc analysis reached significance after Bonferroni correction.

UPDRS-III, Unified Parkinson's Disease Rating Scale part III; PD, Parkinson's disease.

Procedures

At baseline, all participants underwent neuropsychological assessment and resting-state EEG recording. PD patients also had a neurological assessment including Hoehn and Yahr scales and Unified PD Rating Scale part III to assess motor and disease severity (Fahn and Elton, 1987; Hoehn and Yahr, 1967). Participants also completed the Beck Depression Inventory, Beck Anxiety Inventory, and Epworth Sleepiness Scale questionnaires (Table 1). Follow-up duration was a minimum of 2 years after baseline assessment.

Cognitive assessment

The neuropsychological assessment included tests measuring five cognitive domains: attention (measured by the Digit Span subtest of the Wechsler Adult Intelligence Scale Third Edition [WAIS-III], the Trail Making Test part A, a modified version of the Stroop Color Word Test [interference condition]), executive functions (measured by the Trail Making Test part B, a modified version of the Stroop Color Word Test [flexibility condition], semantic verbal fluency [number of animals and fruits/vegetables in 1 min] and phonetic verbal fluency [number of words starting with the letters P, F, and L in the French version and F, A, and S in the English version in 1 min]), verbal episodic memory (measured by the Rey Auditory Verbal Learning Test–sum of trials 1–5, List B, immediate and delayed recalls, recognition), visuospatial abilities (measured by the copy of the Rey-O figure, the Block Design subtest of the WAIS-III, and the Bells test), and language (measured by the Boston Naming Test, the Vocabulary subtest of the WAIS-III and language items of the Mini-Mental State Examination). Details about cognitive tests, variables, normative data, and specific references are reported in (Latreille et al., 2015).

EEG recording and analysis

Waking EEG was performed at a minimum of 30 min after waking up in the morning. Participants were recorded for 10 min while lying in bed with eyes closed. To avoid drowsiness, participants were asked to open their eyes periodically. Recordings were performed with standard EEG leads placed according to the International 10/20 System for Electrode Placement, a bilateral electrooculogram, and a chin electromyogram. A Grass polygraph amplifier system (0.3–100 Hz bandpass filtered) was used. Data were acquired on a Stellate Harmonie software system at a sampling rate of 256 Hz. Four-second segments of visually inspected artefact-free data were selected from the recordings. Analyses were performed on the maximum number of 4-sec segments for each participant. Fourteen electrodes (F3, F4, F7, F8, C3, C4, P3, P4, T3, T4, T5, T6, O1, and O2) were retained for analysis.

MSE analysis

MSE analysis is a measure of the predictability of a time series at different timescales. First, each data segment was gradually down-sampled to create a time series corresponding to each timescale. Thus, for a timescale t, data points within nonoverlapping windows of length t were averaged to create a new time series. With averaging, finer timescales become more sensitive to higher frequencies and coarser timescales to lower frequencies. t ranged between 1 (no averaging) and 20. The predictability (sample entropy) of the amplitude between two data points (m = 2) was then estimated across all time series. The similarity criterion was set to r = 0.5 such that two consecutive data points were considered to have indistinguishable amplitude differences to two other consecutive data points if this difference ranged into ≤50% of time series standard deviation. Predictability for each timescale was then averaged across segments for each participant. Further details on MSE algorithms are available at www.physionet.org/physiotools/mse/and in Costa et al. (2002, 2005).

Phase coherence analysis

Phase-locking value (PLV) was chosen for phase coherence analysis since it is much less sensitive to conduction than other classical coherence analysis. The analysis was performed on Brainstorm (http://neuroimage.usc.edu/brainstorm). Each segment was first subjected to time-frequency analysis by removing the signal mean and tapering with a Hann window with 50% overlap, over frequencies 0–100 Hz. Segments were then concatenated within participants and PLVs were calculated for each frequency band. Five frequency bands were used: delta (2–4 Hz), theta (5–7 Hz), alpha (8–12 Hz), beta (15–29 Hz), and gamma (30–50 Hz). PLV represent phase differences between two brain regions and consistency of these phase differences can be estimated across time (Lachaux et al., 1999). PLV ranges from 0, when phase differences are randomly distributed, to 1, when phase differences are consistently the same. Thus, high PLV between two brain regions indicates high synchrony.

Statistical analysis

Demographic and clinical variables were compared between groups using one-way analyses of variance (ANOVAs), independent sample t-tests, and Pearson chi-square tests. Nonparametric equivalent tests were used for non-normally distributed variables. Shapiro–Wilk test and visual inspection of the data were used for normality testing. Partial least squares (PLS) analysis was used to assess MSE and PLV differences between controls, PDnD patients, and PDD patients. PLS is a multivariate analysis method, using a multiple comparison approach, used to extract patterns of maximal covariance between two datasets and assess whether one dataset predicts the other one, and vice versa (Krishnan et al., 2011). Latent variables (LVs) are assessed for significance using permutation tests. LVs were considered significant at <0.01. Using bootstrap resampling, estimates of LV confidence intervals were then calculated to assess relative contribution of each dataset to the model and for the determination of which channels or PLV were reliable (Kovacevic et al., 2013). For confidence intervals, bootstrap ratios were obtained from the ratio of individual weights to the estimated standard error. A minimum threshold of 95% confidence interval was used.

To identify the best predictors of dementia, a logistic regression was performed with dementia status at follow-up as the dependent variable and mean MSE values for finer scales (scales 1–3), mean MSE values for coarser scales (scales 12–14), mean PLV values in delta frequency, and mean PLV values in gamma frequency as independent variables.

Results

Sample characteristics

At baseline, 82 PD patients were enrolled. At follow-up, eight died, two refused reassessment, and one could not be contacted. Baseline waking EEG recordings were unusable for nine PD patients because of the presence of major artefacts on the EEG recording. Consequently, 62 PD patients (41 men, mean age ± SD: 65.60 ± 8.44 years) and 37 controls (26 men, mean age ± SD = 66.64 ± 8.90 years) were included in this study. At follow-up, 44 PD patients were dementia-free (PDnD) and 18 developed dementia (PDD). Thirty-seven PD patients (60%, 22% PDD and 78% PDnD) underwent at follow-up a complete neuropsychological assessment, 13 patients (21%, 38% PDD and 62% PDnD) underwent a neurological evaluation including office-based cognitive testing, and 12 (19%, 33% PDD and 67% PDnD) were followed up by telephone interview performed with the patients and/or spouses or with the patient's treating physician. At baseline, all groups were equivalent for demographic and clinical variables, except for age and presence of MCI (Table 1). Both PD groups reported to feel more depressed and anxious than controls (p ≤ 0.001).

MSE analysis

MSE and follow-up cognitive status

Figure 1A shows examples of MSE curves at two electrode sites. MSE patterns for each electrode significantly predicted group membership. Belonging to a specific group also significantly predicted MSE patterns (Fig. 1B). Post hoc analysis showed significant differences at finer (small) and coarser (large) timescales. At finer timescales (timescales ≤5), both PD groups had lower entropy than controls for all electrodes, with biggest differences between PDD patients and controls than PDnD patients (p < 0.01; Fig. 1B). When comparing PDD group to PDnD group, entropy differences were found at P3, O1, and T5, with lower entropy in those who developed dementia. In contrast, at coarser timescales (timescales ≥10), PDD patients showed higher entropy than PDnD patients and controls (p < 0.01; Fig. 1B). Entropy for controls and PDnD patients did not differ significantly over most electrodes, except at O1 and O2 for timescales 13–14 and 9–10, respectively.

FIG. 1.

Multiscale entropy (MSE) analysis. (A) Average MSE curves at F3 (left) and O1 (right) for controls, PDnD, and PDD patients. (B) Bootstrap ratio images of partial least squares (PLS) analyses showing significant group differences for each timescale and electrode, at minimal and maximal bootstrap thresholds of −2.3877 and 2.3877. Blue squares represent values lower than the minimal threshold and red squares represent values higher than the maximal threshold. Green squares represent no significant difference. PDD, patients with Parkinson's disease who developed dementia at follow-up; PDnD, patients with Parkinson's disease who remained dementia-free at follow-up.

MSE and MCI status

Since there was at baseline a higher proportion of patients with MCI in the PDD group than in other groups, further analysis were performed to assess whether the observed patterns could be attributed to the presence of MCI at baseline. PLS analysis between PD patients with MCI and PD patients without MCI at baseline, irrespective of the outcome measure (dementia at follow-up) was not significant (p > 0.05), meaning that the presence of MCI was not associated with a specific brain pattern. Moreover, PLS analysis between PDnD patients with MCI (n = 19) and PDnD patients without MCI (n = 25) was also not significant (p > 0.05), suggesting that the brain pattern observed was the same within groups, irrespective of the presence of MCI.

MSE and age

Age of PD patients was also not associated to a specific brain pattern (p > 0.05).

MSE and follow-up assessment type

Results were equivalent (p < 0.01) when performing analysis only on PD patients who performed the complete neuropsychological assessment at follow-up.

Phase coherence analysis

PLV and follow-up cognitive status

PLV significantly predicted which group a participant would belong to, and vice-versa. Correlations were significant for all post hoc comparisons. In delta frequency, PDD patients had lower PLVs than PDnD patients and controls (Fig. 2). Differences were more heterogeneous for theta and alpha frequencies. For beta and gamma frequencies, PDD patients showed higher PLVs compared with PDnD patients and controls (Fig. 2). PDnD patients also showed higher PLVs compared with controls for many electrode pairs in gamma frequency.

FIG. 2.

Phase coherence analysis. Significant partial least squares phase-locked electrode pairs for each frequency band at a minimal and maximal bootstrap thresholds of −2.3877 and 2.3877. From left to right, significant differences when Controls > PDD, when Controls > PDnD and when PDnD > PDD, are depicted in blue. Differences in the opposite direction are depicted in yellow. PDD, patients with Parkinson's disease who developed dementia at follow-up; PDnD, patients with Parkinson's disease who remained dementia-free at follow-up.

Increasing bootstrap thresholds revealed that strongest group differences in delta frequency were found mainly between posterior interhemispheric electrode pairs. In gamma frequency, strongest group differences were found between anterior–posterior pairs.

PLV and follow-up assessment type

Results were equivalent when performing analysis only with PD patients who performed the complete neuropsychological assessment at follow-up.

Predictors of dementia

The logistic regression was statistically significant (p < 0.001) with mean PLV values in delta frequency as the most significant predictor (p < 0.001), followed by mean PLV values in gamma frequency (p = 0.017), mean MSE values for finer scales (p = 0.022), and mean MSE values for coarser scales (p = 0.351). Namely, decreased phase coherence in delta frequency, higher phase coherence in gamma frequency, and lower variability in finer timescales were associated with a higher risk of developing dementia at follow-up.

Discussion

This study identified brain communication alterations among PD patient at highest risk of developing dementia in the next few years. First, significant changes in brain signal variability were found in PD patients who developed dementia at follow-up, with lower signal variability at timescales sensitive to higher frequencies and higher signal variability for timescales sensitive to lower frequencies. Lower phase synchrony in delta frequency with a shift toward higher phase synchrony in higher frequencies was also found in the PD patients who developed dementia.

A major finding of this study is that the combination of MSE and PLV analysis allowed for unique interpretations of the results. For instance, PD patients who developed dementia showed higher synchrony in high frequencies, which would traditionally be interpreted as higher connectivity between brain regions in that frequency band. However, combined with MSE results, higher synchrony and lower signal variability instead suggest a loss of information integration between brain networks in higher frequencies in PD patients who are at higher risk for developing dementia. Hypersynchrony has also been related to cognitive alterations and disrupted information processing, for instance, during epileptic seizures, using EEG (Schevon et al., 2007; Truccolo et al., 2014).

Moreover, in gamma frequency, the strongest phase coherence alterations were found between long-distance anterior–posterior brain regions. This finding is surprising because gamma is known to be involved in local cognitive processing (for reviews see Cannon et al., 2014, Knyazev, 2012). A neuroimaging study found decreased functional connectivity between the inferior frontal gyrus and the precuneus/posterior cingulate cortex, core regions of the default mode network, in PD patients with dementia compared with PD patients without dementia and controls (Rektorova et al., 2012). The long-distance gamma connectivity alterations found in this study could therefore be related to these functional changes. They may also be related to cognitive deterioration, specifically executive and visuospatial dysfunctions that are well known in PD and has been linked to white matter alterations using magnetic resonance imaging (Bertrand et al., 2012; Tröster, 2008). Gamma findings should, however, be interpreted with caution since they may be affected by muscle artefact.

On the other hand, higher complexity and lower phase synchrony were found for lower frequencies, predominantly in PD patients who developed dementia at follow-up. This could be interpreted as a compensation mechanism, whereby cerebral regions attempt to communicate more powerfully in low frequencies to compensate for alterations in high frequency dynamics. However, in a study using MEG, decreased low frequency phase coherence synchrony has been reported in PD patients with dementia compared with PD patients without dementia (Ponsen et al., 2012), where very limited compensation for cognitive decline occurs at this disease stage.

A second possible explanation in the case of dementia is that increased signal complexity is due to disrupted information integration such that networks gradually become randomly organized. This is supported by a longitudinal MEG study on PD progression, where cognitive decline was associated with more random network organization (Olde Dubbelink et al., 2014a). Thus, increased signal complexity combined to lower phase synchrony in lower frequencies may reflect loss of stable functional connectivity through randomization of network organization and may thus be a marker of neurodegeneration leading to dementia in PD.

The most reliable predictor of dementia was phase coherence alterations in delta frequency. Delta alterations were also found to be strongest in posterior brain regions. Recently, connectivity changes, hypometabolism, and white matter alterations in posterior brain regions has also been associated with cognitive decline in PD using neuroimaging techniques (Bertrand et al., 2012; Garcia-Garcia et al., 2012; Kamagata et al., 2013; Melzer et al., 2013; Olde Dubbelink et al., 2014b). These posterior changes in connectivity may also be related to visuoperceptual and visuospatial deficits or to default mode network dysfunctions reported in PD patients (Rektorova et al., 2012; Tröster, 2008; van Eimeren et al., 2009).

Although progressive dopaminergic denervation is the main neuropathological characteristic in PD, other neurotransmitter pathways are known to be affected. Cholinergic denervation appears to play a determinant role in cognitive decline and dementia in PD and is well known in Alzheimer's disease (Hall et al., 2014; Müller and Bohnen, 2013; Shinotoh et al., 1999). Interestingly, EEG and MEG studies using MSE analysis found the same pattern of variability changes in Alzheimer's patients as found here in PD patients who developed dementia, and this pattern correlated with cognitive decline (Gömez et al., 2007; Mizuno et al., 2010; Yang et al., 2013). This supports the possible involvement of cholinergic dysfunction in the pathogenesis of dementia in PD (Bohnen and Albin, 2011).

The shift of decreased variability/increased phase synchrony in high frequencies toward increased variability/decreased phase synchrony in low frequencies may be associated to the slowing of resting state EEG spectral power that has been frequently measured with linear analysis in PD patients with dementia compared with PD patients without dementia (Bonanni et al., 2008; Ponsen et al., 2012). This slowing has also been found to be potentially predictive of dementia in PD (Klassen et al., 2011). Interestingly, this could suggest that both linear and nonlinear mechanisms are involved in dementia development. Findings of this study further provide information about how neurodegeneration leading to dementia occurs by suggesting that brain connectivity alterations happen years before overt clinical dementia symptoms are identified in PD.

From a histopathological point of view, it can be suspected that PD patients who developed dementia at follow-up were at later Braak stages when they underwent EEG at baseline than the PD patients who remained dementia-free (Braak et al., 2003). It could thus be speculated that changes in entropy and phase coherence reflect higher cerebral development of alpha-synuclein aggregats and Lewy bodies in the cortex of PD patients who developed dementia, these neuropathological changes being well known in PD patients with dementia (Halliday et al., 2011).

One strength of this study is the prospective design used to identify predictors of cortical neurodegeneration in PD. Moreover, PD patients were at equivalent PD stage (same disease duration and motor impairment) at baseline, which strongly suggests that our findings are directly related to cortical neurodegeneration leading to cognitive decline.

The study nevertheless has some limitations. First, PD patients who developed dementia were older and were more cognitively impaired at baseline than PD patients who remained dementia-free. Although age and MCI are significant risk factor for dementia and disease duration was equivalent between groups, it cannot be ruled out that some of the effect could be attributable to normal brain aging or MCI at baseline. However, statistical analysis did not reveal significant relationships between age, presence of MCI, and findings. Moreover, PD patients who developed dementia were not significantly older than healthy controls, meaning that differences between the two groups were at least attributable to the risk of developing dementia. Another limitation is that not all patients completed a full neuropsychological assessment at follow-up. However, comparisons including only patients who underwent a neuropsychological assessment at follow-up revealed the same results. Additionally, it cannot be ruled out that the different medication used by PD patients had an impact on EEG recordings. Other studies using different recording and analysis methods would also help in generalizing and reproducing results.

In conclusion, increased variability/randomization of networks communication in low frequencies combined with hypersynchronization/loss of information processing in high frequencies were identified as potential predictive markers of dementia in PD. These markers may help identify PD patients at risk for dementia who are good candidates for neuroprotective trials. Further studies are needed to replicate the findings in other cohorts and to estimate sensitivity and specificity of markers.

Footnotes

Acknowledgments

This study was supported by grants from the Canadian Institutes of Health Research, Canada (MOP-84482; J.-F.G. and R.B.P.), the Fonds de Recherche du Québec–Santé, Canada (J.-F.G. and R.B.P.), and the Quebec Network for Research on Aging (J.-F.G., J.-A.B., R.B.P., and R.M.). J.-F.G. holds a Canada Research Chair in Cognitive Decline in Pathological Aging. J.-A. Bertrand and V.L. were supported by a scholarship from the Canadian Institutes of Health Research (CIHR).

Author Disclosure Statement

No competing financial interests exist.

References

Aarsland

, Andersen

, Larsen

, Lolk

, Kragh-Sørensen

. 2003. Prevalence and characteristics of dementia in Parkinson disease: an 8-year prospective study. Arch Neurol, 60:387–392.

American Psychiatric Association. 2000. Text Revised in Diagnostic and Statistical Manual of Mental Disorders. Washington, DC: American Psychiatric Press.

Babiloni

, De Pandis

, Vecchio

, Buffo

, Sorpresi

, Frisoni

, Rossini

. 2011. Cortical sources of resting state electroencephalographic rhythms in Parkinson's disease related dementia and Alzheimer's disease. Clin Neurophysiol, 122:2355–2364.

Bertrand

, Bedetti

, Postuma

, Monchi

, Génier Marchand

, Jubault

, Gagnon

. 2012. Color discrimination deficits in Parkinson's disease are related to cognitive impairment and white-matter alterations. Mov Disord, 27:1781–1788.

Bohnen

, Albin

. 2011. The cholinergic system and Parkinson disease. Behav Brain Res, 221:564–573.

Bonanni

, Thomas

, Tiraboschi

, Perfetti

, Varanese

, Onofrj

. 2008. EEG comparisons in early Alzheimer's disease, dementia with Lewy bodies and Parkinson's disease with dementia patients with a 2-year follow-up. Brain, 131:690–705.

Bosboom

, Stoffers

, Wolters

ECh

, Stam

, Berendse

. 2009. MEG resting state functional connectivity in Parkinson's disease related dementia. J Neural Transm, 116:193–202.

Braak

, Del Tredici

, Rüb

, de Vos

, Jansen Steur

, Braak

. 2003. Staging of brain pathology related to sporadic Parkinson's disease. Neurobiol Aging, 24:197–211.

Brandt

, Specter

, Folstein

. 1988. The telephone interview for cognitive status. Neuropsychiatry Neuropsychol Behav Neurol, 1:111–117.

10.

Buter

, van den Hout

, Matthews

, Larsen

, Brayne

, Aarsland

. 2008. Dementia and survival in Parkinson disease: a 12-year population study. Neurology, 70:1017–1022.

11.

Cannon

, McCarthy

, Lee

, Börgers

, Whittington

, Kopell

. 2014. Neurosystems: brain rhythms and cognitive processing. Eur J Neurosci, 39:705–719.

12.

Costa

, Goldberger

, Peng

. 2002. Multiscale entropy analysis of complex physiologic time series. Phys Rev Lett, 89:068102.

13.

Costa

, Goldberger

, Peng

. 2005. Multiscale entropy analysis of biological signals. Phys Rev E Stat Nonlin Soft Matter Phys, 71:021906.

14.

Dubois

, Burn

, Goetz

, Aarsland

, Brown

, Broe

, Dickson

, Duyckaerts

, Cummings

, Gauthier

, Korczyn

, Lees

, Levy

, Litvan

, Mizuno

, McKeith

, Olanow

, Poewe

, Sampaio

, Tolosa

, Emre

. 2007. Diagnostic procedures for Parkinson's disease dementia: recommendations from the movement disorder society task force. Mov Disord, 22:2314–2324.

15.

Fahn

, Elton

, UPDRS Program Members. Unified Parkinson's disease rating scale. In: Fahn

, Marsden

, Goldstein

, Calne

, editors. Recent developments in Parkinson's disease, Vol. 2. Florham Park, NJ: Macmillan Healthcare Information; 1987. p 153–163, 293–304.

16.

Folstein

, Folstein

, McHugh

. 1975. “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res, 12:189–198.

17.

Gagnon

, Vendette

, Postuma

, Desjardins

, Massicotte-Marquez

, Panisset

, Montplaisir

. 2009. Mild cognitive impairment in rapid eye movement sleep behavior disorder and Parkinson's disease. Ann Neurol, 66:39–47.

18.

Garcia-Garcia

, Clavero

, Gasca Salas

, Lamet

, Arbizu

, Gonzalez-Redondo

, Obeso

, Rodriguez-Oroz

. 2012. Posterior parietooccipital hypometabolism may differentiate mild cognitive impairment from dementia in Parkinson's disease. Eur J Nucl Med Mol Imaging, 39:1767–1777.

19.

Garrett

, Kovacevic

, McIntosh

, Grady

. 2011. The importance of being variable. J Neurosci, 31:4496–4503.

20.

Gómez

, Hornero

, Abásolo

, Fernández

, Escudero

. 2007. Analysis of MEG recordings from Alzheimer's disease patients with sample and multiscale entropies. Conf Proc IEEE Eng Med Biol Soc, 2007:6184–6187.

21.

Gómez

, Olde Dubbelink

, Stam

, Abásolo

, Berendse

, Hornero

. 2011. Complexity analysis of resting-state MEG activity in early-stage Parkinson's disease patients. Ann Biomed Eng, 39:2935–2944.

22.

Hall

, Reyes

, Landeck

, Bye

, Leanza

, Double

, Thompson

, Halliday

, Kirik

. 2014. Hippocampal Lewy pathology and cholinergic dysfunction are associated with dementia in Parkinson's disease. Brain, 137:2493–2508.

23.

Halliday

, Holton

, Revesz

, Dickson

. 2011. Neuropathology underlying clinical variability in patients with synucleinopathies. Acta Neuropathol, 122:187–204.

24.

Han

, Wang

, Yi

, Che

. 2013. Investigation of EEG abnormalities in the early stage of Parkinson's disease. Cogn Neurodyn, 7:351–359.

25.

Hely

, Reid

, Adena

, Halliday

, Morris

. 2008. The Sydney multicenter study of Parkinson's disease: the inevitability of dementia at 20 years. Mov Disord, 23:837–844.

26.

Hoehn

, Yahr

. 1967. Parkinsonism: onset, progression and mortality. Neurology, 17:427–442.

27.

Johnson

, Diener

, Kaltenboeck

, Birnbaum

, Siderowf

. 2013. An economic model of Parkinson's disease: implications for slowing progression in the United States. Mov Disord, 28:319–326.

28.

Kamagata

, Motoi

, Tomiyama

, Abe

, Ito

, Shimoji

, Suzuki

, Hori

, Nakanishi

, Sano

, Kuwatsuru

, Sasai

, Aoki

, Hattori

. 2013. Relationship between cognitive impairment and white-matter alteration in Parkinson's disease with dementia: tract-based spatial statistics and tract-specific analysis. Eur Radiol, 23:1946–1955.

29.

Keränen

, Kaakkola

, Sotaniemi

, Laulumaa

, Haapaniemi

, Jolma

, Kola

, Ylikoski

, Satomaa

, Kovanen

, Taimela

, Haapaniemi

, Turunen

, Takala

. 2003. Economic burden and quality of life impairment increase with severity of PD. Parkinsonism Relat Disord, 9:163–168.

30.

Klassen

, Hentz

, Shill

, Driver-Dunckley

, Evidente

, Sabbagh

, Adler

, Caviness

. 2011. Quantitative EEG as a predictive biomarker for Parkinson disease dementia. Neurology, 77:118–124.

31.

Knyazev

GG.

2012. EEG delta oscillations as a correlate of basic homeostatic and motivational processes. Neurosci Biobehav Rev, 36:677–695.

32.

Kovacevic

, Abdi

, Beaton

, for the Alzheimer's Disease Neuroimaging Initiative, McIntosh

. 2013. Revisiting PLS resampling: comparing significance vs reliability across a range of simulations. In: Abdi

, Chin

, Esposito Vinzi

, Russolillo

, Trinchera

(eds.) New Perspectives in Partial Least Squares and Related Methods. New York, NY: Springer-Verlag; pp. 159–171.

33.

Krishnan

, Williams

, McIntosh

, Abdi

. 2011. Partial Least Squares (PLS) methods for neuroimaging: a tutorial and review. Neuroimage, 56:455–475.

34.

Lachaux

, Rodriguez

, Martinerie

, Varela

. 1999. Measuring phase synchrony in brain signals. Hum Brain Mapp, 8:194–208.

35.

Latreille

, Carrier

, Lafortune

, Postuma

, Bertrand

, Panisset

, Chouinard

, Gagnon

. 2015. Sleep spindles in Parkinson's disease may predict the development of dementia. Neurobiol Aging, 36:1083–1090.

36.

Litvan

, Goldman

, Tröster

, Schmand

, Weintraub

, Petersen

, Mollenhauer

, Adler

, Marder

, Williams-Gray

, Aarsland

, Kulisevsky

, Rodriguez-Oroz

, Burn

, Barker

, Emre

. 2012. Diagnostic criteria for mild cognitive impairment in Parkinson's disease: Movement Disorder Society Task Force guidelines. Mov Disord, 27:349–356.

37.

McIntosh

, Vakorin

, Kovacevic

, Wang

, Diaconescu

, Protzner

. 2014. Spatiotemporal dependency of age-related changes in brain signal variability. Cereb Cortex, 24:1806–1817.

38.

Melzer

, Watts

, MacAskill

, Pitcher

, Livingston

, Keenan

, Dalrymple-Alford

, Anderson

. 2013. White matter microstructure deteriorates across cognitive stages in Parkinson disease. Neurology, 80:1841–1849.

39.

Mizuno

, Takahashi

, Cho

, Kikuchi

, Murata

, Takahashi

, Wada

. 2010. Assessment of EEG dynamical complexity in Alzheimer's disease using multiscale entropy. Clin Neurophysiol, 121:1438–1446.

40.

Müller

, Bohnen

. 2013. Cholinergic dysfunction in Parkinson's disease. Curr Neurol Neurosci Rep, 13:377.

41.

Müller

, Lutzenberger

, Pulvermüller

, Mohr

, Birbaumer

. 2001. Investigation of brain dynamics in Parkinson's disease by methods derived from nonlinear dynamics. Exp Brain Res, 137:103–110.

42.

Nasreddine

, Phillips

, Bédirian

, Charbonneau

, Whitehead

, Collin

, Cummings

, Chertkow

. 2005. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. J Am Geriatr Soc, 53:695–699.

43.

Olde Dubbelink

, Hillebrand

, Stoffers

, Deijen

, Twisk

, Stam

, Berendse

. 2014a. Disrupted brain network topology in Parkinson's disease: a longitudinal magnetoencephalography study. Brain, 137:197–207.

44.

Olde Dubbelink

, Schoonheim

, Deijen

, Twisk

, Barkhof

, Berendse

. 2014b. Functional connectivity and cognitive decline over 3 years in Parkinson disease. Neurology, 83:2046–2053.

45.

Pezard

, Jech

, Růzicka

. 2001. Investigation of non-linear properties of multichannel EEG in the early stages of Parkinson's disease. Clin Neurophysiol, 112:38–45.

46.

Ponsen

, Stam

, Bosboom

, Berendse

, Hillebrand

. 2012. A three dimensional anatomical view of oscillatory resting-state activity and functional connectivity in Parkinson's disease related dementia: an MEG study using atlas-based beamforming. Neuroimage Clin, 2:95–102.

47.

Rektorova

, Krajcovicova

, Marecek

, Mikl

. 2012. Default mode network and extrastriate visual resting state network in patients with Parkinson's disease dementia. Neurodegener Dis, 10:232–237.

48.

Schevon

, Cappell

, Emerson

, Isler

, Grieve

, Goodman

, McKhann

Jr. , Weiner

, Doyle

, Kuzniecky

, Devinsky

, Gilliam

2007. Cortical abnormalities in epilepsy revealed by local EEG synchrony. Neuroimage, 35:140–148.

49.

Shinotoh

, Namba

, Yamaguchi

, Fukushi

, Nagatsuka

, Iyo

, Asahina

, Hattori

, Tanada

, Irie

. 1999. Positron emission tomographic measurement of acetylcholinesterase activity reveals differential loss of ascending cholinergic systems in Parkinson's disease and progressive supranuclear palsy. Ann Neurol, 46:62–69.

50.

Stoffers

, Bosboom

, Deijen

, Wolters

ECh

, Stam

, Berendse

. 2008. Increased cortico-cortical functional connectivity in early-stage Parkinson's disease: an MEG study. Neuroimage, 41:212–222.

51.

Svenningsson

, Westman

, Ballard

, Aarsland

. 2012. Cognitive impairment in patients with Parkinson's disease: diagnosis, biomarkers, and treatment. Lancet Neurol, 11:697–707.

52.

Tröster

AI.

2008. Neuropsychological characteristics of dementia with Lewy bodies and Parkinson's disease with dementia: differentiation, early detection, and implications for “mild cognitive impairment” and biomarkers. Neuropsychol Rev, 18:103–119.

53.

Truccolo

, Ahmed

, Harrison

, Eskandar

, Cosgrove

, Madsen

, Blum

, Potter

, Hochberg

, Cash

. 2014. Neuronal ensemble synchrony during human focal seizures. J Neurosci, 34:9927–9944.

54.

van Eimeren

, Monchi

, Ballanger

, Strafella

. 2009. Dysfunction of the default mode network in Parkinson disease: a functional magnetic resonance imaging study. Arch Neurol, 66:877–883.

55.

Vossius

, Larsen

, Janvin

, Aarsland

. 2011. The economic impact of cognitive impairment in Parkinson's disease. Mov Disord, 26:1541–1544.

56.

Yang

, Wang

, Lai

, Tsai

, Yang

, Hwang

, Lo

, Huang

, Peng

, Fuh

. 2013. Cognitive and neuropsychiatric correlates of EEG dynamic complexity in patients with Alzheimer's disease. Prog Neuropsychopharmacol Biol Psychiatry, 47:52–61.