EEG spectral analysis as a putative early prognostic biomarker in nondemented, amyloid positive subjects

Neurobiology of Aging 57 (2017) 133e142

Contents lists available at ScienceDirect

Neurobiology of Aging journal homepage: www.elsevier.com/locate/neuaging

EEG spectral analysis as a putative early prognostic biomarker in nondemented, amyloid positive subjects Alida A. Gouw a, b, *, Astrid M. Alsema a, b, Betty M. Tijms a, Andreas Borta c, Philip Scheltens a, Cornelis J. Stam b, Wiesje M. van der Flier a, d a

Alzheimer Center and Department of Neurology, Neuroscience Campus Amsterdam, VU University Medical Center, Amsterdam, the Netherlands Department of Clinical Neurophysiology and MEG Center, Neuroscience Campus Amsterdam, VU University Medical Center, Amsterdam, the Netherlands c Boehringer Ingelheim Pharma GmbH Co KG, Ingelheim am Rhein, Germany d Department of Epidemiology and Biostatistics, Neuroscience Campus Amsterdam, VU University Medical Center, Amsterdam, the Netherlands b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 5 May 2016 Received in revised form 15 May 2017 Accepted 23 May 2017 Available online 1 June 2017

We studied whether electroencephalography (EEG)-derived measures of brain oscillatory activity are related to clinical progression in nondemented, amyloid positive subjects. We included 205 nondemented amyloid positive subjects (63 subjective cognitive decline [SCD]; 142 mild cognitive impairment [MCI]) with a baseline resting-state EEG data and
1-year follow-up. Peak frequency and relative power of 4 frequency bands were calculated. Relationships between normalized EEG measures and time to clinical progression (conversion from SCD to MCI/dementia or from MCI to dementia) were analyzed using Cox proportional hazard models. One hundred eight (53%) subjects clinically progressed after 2.1 (IQR 1.3e3.0) years. In the total sample, none of the EEG spectral measures were significant predictors. Stratified for baseline diagnosis, we found that in SCD patients higher delta and theta power (HR [95% CI] ¼ 1.7 [1.0e2.7] resp. 2.3 [1.2e4.4]), and lower alpha power and peak frequency (HR [95% CI] ¼ 0.5 [0.3 e1.0] resp. 0.6 [0.4e1.0]) were associated with clinical progression over time. In amyloid positive subjects with normal cognition, slowing of oscillatory brain activity is related to clinical progression. Ó 2017 Elsevier Inc. All rights reserved.

Keywords: Electroencephalography Prognostic biomarker Clinical progression Alzheimer’s disease Amyloid beta

1. Introduction Alzheimer’s disease (AD) develops gradually over the course of 15e20 years. One of the first pathological changes of the disease is the accumulation of amyloid beta in the brain, which starts many years before the appearance of first symptoms of cognitive decline (Jack et al., 2013). Identifying subjects in the earliest stages of the disease offers the opportunity to apply potential preventive measures, before neurodegeneration and synapse loss are irreversible. Recent research criteria have taken amyloid-b 1-42 concentration in CSF and amyloid PET imaging into account to support the diagnosis of AD in subjects with and without dementia (Albert et al., 2011; Dubois et al., 2014; McKhann et al., 2011; Sperling et al., 2011). Synaptic dysfunction resulting from synaptic toxicity of amyloid beta supposedly occurs early in the cascade of events eventually

* Corresponding author at: Alzheimer Center and Department of Neurology, VU University Medical Center, Department of Clinical Neurophysiology and MEG Center, VU University Medical Center, PO Box 7057, 1007 MB Amsterdam, the Netherlands. Tel.: þ31 20 444 0722; fax: þ31 20 444 4816. E-mail address: [email protected] (A.A. Gouw). 0197-4580/$ e see front matter Ó 2017 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.neurobiolaging.2017.05.017

leading to cognitive decline and dementia (Palop and Mucke, 2010; Selkoe, 2002; Sperling et al., 2013). The most clinically relevant method to capture in vivo synaptic functioning is electroencephalography (EEG) that directly measures postsynaptic dendritic currents of synchronized cortical neurons. Previous EEG studies in patients with dementia due to AD show a gradual diffuse slowing of brain electrical activity reflected by theta power increases and beta power decreases, followed in later stages by a decrease in alpha power and increase in delta power (de Haan et al., 2008; Jeong, 2004; van Straaten et al., 2014). At the mild cognitive impairment (MCI) stage, EEG abnormalities are intermediate between healthy controls and dementia patients (Kwak, 2006; van der Hiele et al., 2007). Several longitudinal studies in MCI have suggested that EEG measures are associated with incident clinical progression over 1e3 years (Huang et al., 2000; Jelic et al., 2000; Luckhaus et al., 2008). In subjects with subjective cognitive decline, only one study has been performed and reported prediction of decline to MCI by several spectral and covariance measures, predominantly in the theta band (Prichep et al., 2006). However, these studies did not take into account the underlying pathology, in particular amyloid status, in their study populations, so it remains unclear if these

134

A.A. Gouw et al. / Neurobiology of Aging 57 (2017) 133e142

subjects belonged to the AD pathophysiological continuum at all. To answer questions about prognosis in the nondementia phases of AD, it is therefore crucial to select subjects who have biomarker proof of underlying Alzheimer’s pathology (Giannakopoulos et al., 2009; Sperling et al., 2011). Although it has been demonstrated that subjects with positive amyloid markers in the preclinical and MCI stages have an increased risk to develop dementia compared with amyloid negative subjects (Van Harten et al., 2013), relationships with severity of impairment are modest and its prognostic value for predicting time to dementia is quite limited (Prestia et al., 2015; van Rossum et al., 2012b). Additional markers sensitive enough to predict cognitive decline are required. For the effective design of prevention trials in AD that are targeted against amyloid pathology, these prognostic markers and the selection of subjects with proven amyloid pathology are crucial (Hampel et al., 2011). Here, we studied in nondemented subjects with an amyloid positive biomarker status, whether EEG-derived measures of brain oscillatory activity are associated with clinical progression. We hypothesized that diffuse slowing of oscillatory activity, reflected by increased relative power in the lower frequency bands (theta and delta) and decreased relative power in higher frequency bands (alpha and beta), is related to clinical progression. 2. Methods 2.1. Subjects We included 205 nondemented, amyloid positive subjects from the Amsterdam Dementia Cohort (van der Flier et al., 2014). All subjects were referred to the Alzheimer Center between February 2001 and January 2014. They underwent a standardized screening including medical history, informant based history, physical and neurological examination, neuropsychological evaluation, EEG, magnetic resonance imaging, laboratory tests, and lumbar puncture. All diagnoses were made in a multidisciplinary consensus meeting and were based on the full standardized diagnostic work-up. Followup visits were generally performed annually. The large majority of these visits consisted of a standardized neuropsychological assessment and a neurological evaluation and diagnoses were re-evaluated in a multidisciplinary meeting. Inclusion criteria for the present study were: (1) amyloid positivity defined as CSF amyloid-b 1-42 <640 pg/mL (Zwan et al., 2014); (2) diagnosis of subjective cognitive decline (SCD) or mild cognitive impairment (MCI) using the standard diagnostic criteria (Albert et al., 2011; Jessen et al., 2014); (3) at least 1 year follow-up; and (4) availability of a 20-minute resting-state EEG at baseline. Exclusion criteria were: a medical history of other significant neurological disorders (e.g., current epilepsy, lobar infarcts/hemorrhages, and severe brain trauma) or psychiatric disorders (e.g., autism, schizophrenia), current use of acetylcholineesterase inhibitors, antipsychotic drugs, lithium, anti-epileptic drugs, or neuropathic pain medication. Primary end point was clinical progression, defined as a conversion from SCD at baseline to MCI or dementia at follow-up or conversion from MCI at baseline to dementia at follow-up. We also included a group of amyloid negative patients (CSF amyloid-b 1-42
640 pg/mL) with similar inclusion and exclusion criteria. All subjects gave written informed consent for the storage of their examinations in a local database and for use of their data for research purposes. The ethical review board of the VU University Medical Center approved the study. 2.2. CSF analysis CSF samples were collected by lumbar puncture between the L3/L4, L4/L5, or L5/S1 intervertebral space by a 25-gauge needle

and syringe and collected in polypropylene tubes. CSF biomarker analyses were performed at the Neurochemistry laboratory of the department of Clinical Chemistry of the VUmc. Amyloid-b 1-42 (Ab42), total tau, and tau phosphorylated at threonine 181 (p-tau) concentrations are measured with sandwich ELISAs (Innotest, beta-amyloid1-42, Innotest hTAU-Ag and Innotest PhosphoTAU181p, Innogenetics, Belgium). 2.3. EEG At baseline, a 20-minute resting-state EEG was recorded at 21 electrode positions of the 10e20 system. Three EEG-systems were used over the years: Nihon Kohden digital EEG equipment (EEG 2100; Nihon Kohden, Tokyo, Japan), and 2 versions of OSG digital equipment (BrainLab and BrainRT, OSG b.v., Rumst, Belgium). EEGs were recorded against an average reference including all electrodes, with the following order of channels: Fp2/Fp1, F8/F7, F4/ F3, A2/A1, T4/T3, C4/C3, T6/T5, P4/P3, O2/O1, Fz, Cz, Pz. Sample frequency of these EEG recordings was 200 Hz (Nihon Kohden) or 500 Hz (BrainLab and BrainRT). Filter settings were: time constant 1 second (Nihon Kohden and BrainLab) or 0.6 seconds (BrainRT), low pass filter 70 Hz (Nihon Kohden and BrainLab) or 100 Hz (BrainRT) and no notch filter (all). Analog to digital conversion precision was 12 bit (Nihon Kohden and BrainLab) or 20 bit (BrainRT). Electrode impedance was kept below 5 kOhm. Patients were seated in a slightly reclined chair in a sound attenuated but fully lit room and were instructed to keep their eyes closed and stay awake. EEG technicians monitored the recording carefully and alerted the patients by sound stimuli at first signs of drowsiness. Based on the knowledge that at least 4 epochs per subject are needed to obtain stable values of quantitative EEG measures (Engels et al., 2015; Nuwer, 1988; van Diessen et al., 2015), we selected 5 epochs per subject (2048 samples; 10.2 seconds per epoch [Nihon Kohden] or 4096 samples; 8.2 seconds per epoch [BrainLab and BrainRT]). All epochs were selected by visual inspection by a trained EEG researcher (AA), based on the presence of a minimum of artifacts (e.g., excessive muscle activity, eye blinks) and drowsiness and were rated on quality: score 1 ¼ no eye movement, muscle, signs of drowsiness or other artifacts; 2 ¼ minimal presence of artifacts; 3 ¼ moderate presence of artifacts; 4 ¼ strong presence of artifacts. All epochs with a score 3 and 4 were evaluated by another rater (AG). If no consensus on sufficient quality was reached, the epochs were replaced by other epochs or the EEG was excluded from analyses. During this standard epoch selection procedure (Jobert et al., 2012; van Diessen et al., 2015), the investigator was blinded for baseline diagnosis and follow-up status. 2.4. EEG spectral analyses EEG spectral analyses were performed with open-access software BrainWave (version 0.9.151.5, developed by CS; available at http://home.kpn.nl/stam7883/brainwave.html). Relative power in 5 standard frequency bands (delta: 0.5e4 Hz, theta: 4e8 Hz, alpha: 8e13 Hz, beta: 13e30 Hz, gamma: 30e48 Hz) and peak frequency (Hz; dominant frequency between 4e13 Hz) were calculated at each electrode using Fast Fourier Transformation. The gamma band was excluded for further analysis because EEG signal in this band is significantly contaminated with muscle artifacts (Whitham et al., 2007). EEG values of the 5 epochs per subject were averaged to obtain values at subject level. Global EEG measures were calculated by averaging values of all 21 electrodes. EEG measures at lobar level were obtained by averaging FP1, FP2, F3, F4, F7, F8, and Fz for the frontal lobes, C3, C4, and Cz for the central region, P3, P4, Pz for the parietal lobes, T3, T4, T5, T6, A1, and A2 for the temporal lobes, and O1 and O2 for the occipital lobes.

A.A. Gouw et al. / Neurobiology of Aging 57 (2017) 133e142

2.5. Statistical analyses Statistical software package SPSS version 22.0 (SPSS Inc, Chicago, IL, USA) and R Studio version 0.99.484 for Windows were used for statistical analyses. Demographic characteristics and global and lobar EEG spectral measures of stable and clinically progressive subjects were compared with Mann-Whitney U tests, independent samples t test, or c2 tests where appropriate. For survival analyses, all EEG measures were Z-transformed to make the interpretation of the results comparable across frequency bands. For each predictor, we used Cox proportional hazard models to test whether global EEG measures (independent variables) were associated with time to clinical progression (dependent variable), adjusting for age, gender, and EEG system (model 1). All hazard ratios can be interpreted as the increased risk per SD increase in EEG relative power or peak frequency. We repeated these Cox proportional hazard models with the addition of MTA-scores (model 2; Scheltens et al., 1995) and CSF-tau levels (model 3) as covariates. For regional analyses, group comparisons and survival analyses (adjusting for age, gender, and EEG system) were repeated with EEG spectral measures per electrode as predictors. Statistical significance of the regional tests was determined with permutation testing (Noble, 2009). Briefly, group labels (stable vs. clinical progression) were randomly assigned 10.000 times and the maximum/minimum statistics (t-statistic for group comparisons or b-coefficient for survival analyses) were used to obtain permutation distributions. The p-value (denoted as ppermuted) is the proportion of the permutation distribution that is greater or equal to the calculated regional EEG measure statistic. This procedure provides a strong control for Type I errors (Singh et al., 2003). Statistical significance was determined at p < 0.05 for whole-brain analyses and analyses at the level of brain lobes, and ppermuted <0.05 for electrode level analyses. All analyses were repeated after stratification for baseline diagnosis (SCD or MCI). For comparison of EEG spectral measures between amyloid positive and negative patients, independent samples t tests were used. To explore the potential influence of artifacts on the main results, epoch quality scores were compared between stable and clinically progressive subjects using c2 test, and survival analyses (model 1) were repeated with the exclusion of epochs with quality score ¼ 4. 3. Results 3.1. Baseline characteristics In total, 205 patients (mean age 67.6  7.7 years; 103 females) were included and followed during a median period of 2.2 years (Table 1). Of these patients, 63 had a diagnosis of SCD and 142 had a diagnosis of MCI at baseline. Subjects with MCI were more often tau-positive with higher CSF tau and p-tau levels than SCD patients. Groups did not differ with regard to age, gender, level of education, APOE e4 status, medial temporal lobe atrophy (MTA) score (Scheltens et al., 1995), CSF amyloid-b level, follow-up time, and EEG system. As expected, MCI patients performed worse than SCD patients on most neuropsychological tests across cognitive domains. EEG spectral measures were indicative of global slowing of oscillatory activity in the MCI patient group compared with the SCD patient group, characterized by a higher theta power, lower alpha power, and lower global peak frequency. 3.2. EEG predictors in total study sample Table 2 shows the demographic and EEG characteristics of the study sample by clinical follow-up status. In total, 108 (53%) of the 205 subjects showed clinical progression during follow-up. These

135

subjects had lower baseline MMSE scores and were more often taupositive with higher levels of CSF tau and p-tau than subjects who remained stable (all p < 0.001). Mean age, level of education, use of comedication, APOE e4 status, MTA-score, CSF amyloid-b levels, follow-up duration, and EEG system were comparable between stable and progressive patients. In the total study sample, none of the global EEG measures significantly differed between subjects who remained stable and those who showed clinical progression. Only a trend was found for lower relative alpha power and lower peak frequency in subjects that progressed. Survival analyses for global EEG predictors, adjusted for age, gender and EEG system, showed that none of the EEG measures was a significant predictor for clinical progression (Table 3). In model 2, when MTA-score was entered as an additional covariate, a trend was found for higher theta power. Fig. 1 shows the relative power of each frequency band and peak frequency at the 21 electrode positions for stable and clinically progressive subjects and the absolute differences between those groups. Survival analyses were repeated with age, gender, and EEG system as covariates and correction for multiple testing. For the total study sample, no significant regional EEG predictors were found. 3.3. EEG predictors after stratification for baseline diagnosis All analyses were repeated after stratification for baseline diagnosis. Table 2 shows the demographic and EEG characteristics of the study sample by clinical follow-up status, stratified for baseline diagnosis. Of the 63 subjects with SCD at baseline, 25 showed clinical progression (N ¼ 19 progression to MCI; N ¼ 6 progression to dementia). Of the SCD patients that progressed to dementia, 4 patients developed AD type of dementia, and 2 patients developed another type of dementia (n ¼ 1 vascular dementia and n ¼ 1 Lewy Body dementia [DLB]). Patients who clinically progressed had higher CSF tau and p-tau levels than those who remained stable (p < 0.05 and p < 0.01, respectively). Of the 142 subjects with a baseline diagnosis of MCI, n ¼ 83 showed clinical progression over time. The majority developed AD type dementia (n ¼ 79), whereas 3 patients developed another type of dementia (4%; n ¼ 2 vascular dementia and n ¼ 1 frontotemporal lobe dementia). MCI patients that progressed had a significantly lower MMSE (p < 0.01) and were more often tau-positive with higher CSF tau and p-tau levels than stable subjects (both p < 0.001). We found that SCD patients who progressed showed a significantly higher theta power (mean 0.13 [SD 0.05]) than SCD patients who did not progress (mean 0.10 [SD 0.03]; p < 0.01). In addition, they showed a trend toward a lower alpha power and peak frequency (p < 0.10). In the subsample with MCI at baseline, there were no differences between groups. Cox proportional hazard models (Table 4), adjusted for the covariates age, gender, and EEG system (model 1), showed that the following EEG parameters were associated with time to progression in the SCD group: a higher global delta power, a higher global theta power, a lower global alpha power, and a higher global peak frequency. The strongest predictor was global theta power with an HR of 2.33, indicating that for each 4% (¼1 SD) increase in theta power the risk for clinical progression over time multiplies with a factor of 2.33. After additional adjustment for MTA-score (model 2), relative theta power remained a significant predictor for clinical progression in this subsample with a HR (CI-95%) of 2.36 (1.15e4.81). Relative delta power, relative alpha power, and global peak frequency, lost significance and were only predictors at trend level. When CSF-tau was added as an additional covariate to model 1, no significant predictors were found. Only relative alpha power was a predictor at trend level. In the MCI, subsample none of the EEG measures were related to clinical progression.

136

A.A. Gouw et al. / Neurobiology of Aging 57 (2017) 133e142

Table 1 Clinical and neuropsychologic characteristics of study sample

Demographics N (%) Female, N (%) Age y Education, median (IQR) Comedication, N (%) APOE e4 positive, N (%) CSF Ab42 (pg/mL), median (IQR) CSF tau (pg/mL), median (IQR) Tau-positivity (>375 mg/L); n (%) CSF p-tau (pg/mL), median (IQR) MTA-score Follow-up time y, median (IQR) Neuropsychological characteristics MMSE, median (IQR) Digit span forward, n ¼ 202 Digit span backward, n ¼ 202 VAT memory, n ¼ 185 VAT naming, n ¼ 192 TMT A (s, n ¼ 200), median (IQR) TMT B (s, n ¼ 197), median (IQR) Stroop 1 (s, n ¼ 173), median (IQR) Stroop 2 (s, n ¼ 162), median (IQR) Stroop 3 (s, n ¼ 161), median (IQR) 15 word test memory (n ¼ 173) EEG spectral measures Relative delta power Relative theta power Relative alpha power Relative beta power Peak frequency (Hz)

Total sample

SCD

MCI

205 103 67.6 5 35 131 496 467 129 71 0.66 2.2

(50%) (7.7) (4e6) (17%) (70%) (398e571) (321e673) (63%) (50e97) (0.70) (1.3e3.1)

63 32 66.2 6 13 38 504 345 26 56 0.55 2.1

(51%) (8.2) (5e7) (21%) (64%) (422e573) (224e485) (41%) (44e72) (0.66) (1.1e3.0)

142 70 68.3 5 22 93 489 528 103 78 0.71 2.2

(49%) (7.4) (4e6) (16%) (73%) (394e570) (361e771)*** (73%)*** (58e104)*** (0.72) (1.4e3.1)

28 12.3 8.7 9.9 11.9 44 112 44 62 115 32.6

(26e29) (2.6) (2.4) (2.8) (0.3) (34e54) (85e144) (41e50) (54e74) (97e137) (8.8)

28 12.5 9.3 11.6 12.0 38 89 44 62 113 39.3

(27e29) (2.6) (2.6) (0.9) (0) (30e53) (76e111) (41e49) (54e73) (92e129) (7.7)

27 12.2 8.4 9.2 11.9 45 126 45 62 115 29.7

(25e28)*** (2.6) (2.2)** (3.0)*** (0.4)*** (36e56)** (100e158)*** (41e50) (54e74) (98e141) (7.6)***

0.31 0.13 0.30 0.20 8.92

(0.11) (0.06) (0.13) (0.08) (0.97)

0.30 0.12 0.34 0.20 9.28

(0.11) (0.04) (0.13) (0.07) (1.06)

0.31 0.14 0.29 0.20 8.77

(0.11) (0.07)* (0.12)* (0.08) (0.89)***

Values are means (SD), unless stated otherwise. Education level (range 1e7) was available for n ¼ 204; APOE e4 status was available for n ¼ 187; MTA-score (range 1e4; mean of left and right MTA-score) was available for n ¼ 180. Comedication: current use of benzodiazepines, antidepressants, neuropathic analgesics. Differences between SCD and MCI subjects were tested with c2 tests, t tests, or Mann-Whitney U tests. * p < 0.05, **p < 0.01, ***p < 0.001. Key: EEG, electroencephalography; MCI, mild cognitive impairment; MMSE, miniemental state examination; MTA, medial temporal lobe atrophy; SCD, subjective cognitive decline; TMT, trail making test; VAT, visual association test (range 1e12).

Fig. 1B and C show the EEG spectral measures at 21 electrode positions for patients who remain stable and those who clinically progress and the absolute differences between those groups, for SCD and MCI subsamples respectively. In the SCD group (Fig. 1B), subjects who progressed showed slowing of the posterior oscillatory activity mainly reflected by a higher delta and theta power, lower alpha power and a lower peak frequency in the parietal-temporal-occipital regions. In addition, an increase in fronto-central theta power, and decreases in frontal alpha power and peak frequency and left central beta power were found. These regional group differences, corrected for multiple comparisons, were only significant for relative theta power (Fz, Cz, C3, C4, Pz, P3, T5 and O1: all ppermuted <0.05). Survival analyses at electrode level, corrected for covariates and for multiple comparisons, showed that relative theta power was predictive for progression in SCD patients specifically at the frontal-central midline (Fz, Cz) and left posterior temporal region (T5; all ppermuted <0.05). For the MCI subgroup (Fig. 1C), no significant group differences or regional predictors were found.

peak frequency in the frontal lobe. In the subsample of MCI patients, higher relative theta power in the parietal and occipital lobes, and lower relative alpha power in the parietal, temporal, and occipital lobes were found in progressive patients. 3.5. Relation between amyloid status and EEG spectral measures To explore the relation between amyloid status and EEG spectral measures, we compared EEG spectral measures of our amyloid positive sample with an amyloid negative sample that had similar inclusion and exclusion criteria. This amyloid negative sample consisted of 281 patients, of which 180 patients had a baseline diagnosis of SCD and 101 had MCI. Mean age (SD) of the group was 61.4 (8.77) years and 36% were female (n ¼ 100). Median MMSE-score (IQR) was 28 (27e29). We found that amyloid positive patients had more global slowing of oscillatory brain activity than amyloid negative patients, indicated by a higher relative theta power, and a lower relative alpha and beta power, and that these differences were driven by the MCI subsample (see Supplementary Material).

3.4. Lobar EEG spectral measures 3.6. Possible influence of EEG artifacts on results For illustration purposes, lobar EEG spectral measures for stable and clinically progressive patients are depicted in the Supplementary Figure, both for the total study sample and stratified for baseline diagnosis. In the total sample, we found that clinically progressive patients had a significantly lower frontal peak frequency than patients who remained stable. In the subsample of patients with SCD at baseline, progressive patients had significantly higher theta power in all lobes, and lower relative alpha power and

To investigate the possible influence of EEG artifacts on the main results, we compared epoch quality scores between patients who remained stable with those who progressed and found no differences. Further, survival analyses with global EEG predictors (model 1) were repeated excluding patients with epoch’s quality score ¼ 4, and findings remained essentially unchanged (see Supplementary Material).

A.A. Gouw et al. / Neurobiology of Aging 57 (2017) 133e142

137

Table 2 Clinical and EEG characteristics by clinical follow-up status

Demographics N (%)

Total sample (n ¼ 205)

Baseline SCD (n ¼ 63)

Baseline MCI (n ¼ 142)

Stable

Stable

Stable

Progression

97 (47%)

Female, N (%) Age y Education, median (IQR) Comedication, N (%) MMSE, median (IQR) APOE e4 positive, N (%) CSF Ab42 (pg/mL), median (IQR) CSF tau (pg/mL), median (IQR) Tau-positivity (>375 mg/L); n (%) CSF p-tau (pg/mL), median (IQR) MTA-score Follow-up time y, median (IQR) EEG system N (%) Brainlab BrainRT Nihon Kohden EEG spectral measures Relative delta power Relative theta power Relative alpha power Relative beta power Peak frequency (Hz)

45 67 5 16 28 66 514 360 45 55 0.56 2.3

(46%) (8) (4e6) (16%) (27e28) (72%) (400e586) (226e571) (46%) (39e78) (0.66) (1.3e3.1)

74 (76%) 15 (16%) 8 (8%) 0.30 0.13 0.32 0.20 9.04

(0.11) (0.07) (0.14) (0.08) (0.99)

108 (53%) 57 68 5 19 27 65 487 554 84 79 0.76 2.1

(53%) (8) (4e6) (18%) (25e26)*** (68%) (396e553) (424e800)*** (78%)*** (65e114)*** (0.73) (1.3e3)

91 (84%) 7 (7%) 10 (9%) 0.31 0.14 0.29 0.21 8.82

(0.11) (0.06) (0.11)# (0.08) (0.96)#

38 (60%) 18 66 5 6 29 24 510 320 13 49 0.60 2.3

(47%) (9) (4e6) (16%) (28e28) (67%) (445e582) (200e424) (34%) (36e72) (0.73) (1.1e3.1)

28 (74%) 4 (10%) 6 (16%) 0.28 0.10 0.36 0.21 9.47

(0.10) (0.03) (0.14) (0.08) (0.97)

Progression 25 (40%); 19 to MCI, 6 to dementia 14 (56%) 67 (7) 6 (5e7) 7 (28%) 28 (27e28) 14 (61%) 496 (380e561) 377 (320e574)* 13 (52%) 68 (52e98)** 0.48 (0.54) 1.6 (1.1e2.8) 23 (92%) 1 (4%) 1 (4%) 0.32 0.13 0.30 0.19 8.98

(0.11) (0.05)** (0.11)# (0.07) (1.15)#

59 (42%) 27 68 5 10 28 42 519 420 32 62 0.54 2.2

(46%) (7) (4e6) (17%) (26e27) (75%) (369e588) (274e620) (54%) (42e91) (0.62) (1.3e3.1)

46 (78%) 11 (19%) 2 (3%) 0.31 0.14 0.29 0.20 8.77

(0.11) (0.08) (0.14) (0.09) (0.90)

Progression 83 (58%) 43 68 5 12 26 51 487 625 71 84 0.84 2.2

(52%) (8) (4e6) (14%) (24e26)** (71%) (397e552) (467e810)*** (86%)*** (69e119)*** (0.76)* (1.4e3)

68 (82%)* 6 (7%) 9 (11%) 0.31 0.14 0.28 0.21 8.77

(0.10) (0.06) (0.11) (0.08) (0.89)

Values are means (SD), unless stated otherwise. Education level (range 1e7) was available for n ¼ 204; APOE e4 status was available for n ¼ 187; MTA-score (range 1e4; mean of left and right MTA-score) was available for n ¼ 180. Comedication: current use of benzodiazepines, antidepressants, neuropathic analgesics. Differences between stable and progressive subjects were tested with c2 tests, t tests, or Mann-Whitney U tests. * p < 0.05, **p < 0.01, ***p < 0.001, #p < 0.10. Key: EEG, electroencephalography; MCI, mild cognitive impairment; MMSE, miniemental state examination; MTA, medial temporal lobe atrophy; SCD, subjective cognitive decline.

4. Discussion The main finding of our study is that EEG-derived measures of brain oscillatory activity were related to clinical progression at very early stages of AD, when objective cognitive impairment has not yet occurred. The SCD patients included in this study visited our memory clinic with cognitive complaints but had no objective cognitive impairment. Experiencing subjective cognitive complaints is a risk factor for developing AD, especially combined with biomarker evidence for amyloid positivity (Jessen et al., 2014; Van Harten et al., 2013). Despite an increased risk of clinical decline, amyloid concentrations are not indicative of the rate of decline, or duration until dementia onset (Prestia et al., 2015; van Rossum et al., 2012a). In the present study, we found that diffuse slowing of EEG oscillatory activitydindicated by higher relative power in the lower frequency

Table 3 EEG predictors of clinical progression in amyloid positive subjects: total sample

Relative delta power Relative theta power Relative alpha power Relative beta power Peak frequency

Model 1

Model 2

Model 3

1.11 1.11 0.88 0.96 0.88

0.17 1.18 0.84 0.92 0.85

1.00 1.03 0.95 1.03 0.98

(0.90e1.37) (0.94e1.30) (0.71e1.10) (0.80e1.16) (0.71e1.08)

(0.94e1.47) (0.99e1.42)# (0.66e1.07) (0.75e1.12) (0.68e1.06)

(0.81e1.25) (0.86e1.22) (0.77e1.19) (0.85e1.23) (0.78e1.22)

Values in model 1 are hazard ratios (95% CI) of Cox proportional hazard models adjusted for age, gender, and EEG system. Values in model 2 are hazard ratios (95% CI) of model 1 with the additional adjustment for medial temporal lobe atrophy score. Values in model 3 are hazard ratios (95% CI) of model 1 with the additional adjustment for CSF-tau level. All predictors were z-transformed before regression. The hazard ratios can be interpreted as the increased risk per SD increase in EEG relative power or peak frequency. # p < 0.10. Key: EEG, electroencephalography.

bands (delta 0.5e4 Hz and theta 4e8 Hz), lower relative power in a higher frequency band (alpha 8e13 Hz), and lower global peak frequency were significant predictors for time to clinical progression in individuals with biomarker proven preclinical AD. The results support our hypothesis and correspond to the general view of a pathophysiological continuum of the disease, as these EEG findings reflect the diffuse slowing that is commonly seen in patients in more advanced stages of AD (de Waal et al., 2012). This gradual change in EEG abnormalities in the course of developing AD is also supported by our finding that subjects with MCI had more severe EEG abnormalities compared with subjects with SCD at baseline. Moreover, progressive subjects with SCD had EEG spectral values intermediate between stable subjects with SCD and subjects with MCI. This study is the first to analyze the prognostic value of EEG spectral measures in cognitively normal subjects with biomarker proven AD. Even in this highly specific patient group, which is already at increased risk for cognitive decline, we could find an additional effect of EEG spectral measures. Relative theta power appeared to be the strongest prognostic marker. As several other variables are also known to be associated with clinical progression, including CSF-tau and MTA (Eckerström et al., 2015; Sierra-Rio et al., 2016; Stomrud et al., 2010; Vos et al., 2013), we repeated the survival analyses with additional correction for these measures. We found that relative theta power is a prognostic marker that has additive value over MTA score, but not over CSF tau levels. Therefore, EEG spectral measures cannot be regarded as predictors that are completely independent of other known predictive variables but can be used as a widely available, easy to apply and patient friendly alternative modality to aid in the prediction of future cognitive decline. Our findings extend on the results of an earlier published EEG study that also focused on the prediction of cognitive decline in SCD subjects. This study analyzed 44 SCD subjects with

138

A.A. Gouw et al. / Neurobiology of Aging 57 (2017) 133e142

Fig. 1. Head plots of distribution of mean relative power per frequency band and mean peak frequency are shown by follow-up status. A warmer color indicates a higher relative power (scaled from minimum to maximum value of total group). (A) Amyloid positive nondemented subjects, total group, and stratified for baseline diagnosis (B: SCD patients; C: MCI patients). Significant hazard ratios, corrected covariated and for multiple comparison testing, are depicted by white asterixes. Abbreviations: MCI, mild cognitive impairment; SCD, subjective cognitive decline. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

A.A. Gouw et al. / Neurobiology of Aging 57 (2017) 133e142

139

Table 4 EEG predictors of clinical progression in amyloid positive subjects, stratified for baseline diagnosis Model 1 SCD Relative delta power Relative theta power Relative alpha power Relative beta power Peak frequency

1.66 2.33 0.53 0.92 0.59

Model 2 MCI

(1.01e2.74)* (1.24e4.40)* (0.30e0.95)* (0.59e1.43) (0.36e0.97)*

1.08 1.02 1.00 0.93 0.98

Model 3

SCD (0.84e1.38) (0.84e1.23) (0.78e1.29) (0.76e1.15) (0.76e1.27)

1.61 2.36 0.57 0.83 0.64

MCI (0.96e2.68)# (1.15e4.81)* (0.32e1.03)# (0.52e1.34) (0.39e1.07)#

1.15 1.05 0.97 0.90 0.95

SCD (0.89e1.50) (0.84e1.32) (0.72e1.29) (0.72e1.12) (0.71e1.27)

1.52 2.06 0.56 1.07 0.67

MCI (0.90e2.56) (0.89e4.77) (0.31e1.02)# (0.70e1.64) (0.39e1.14)

0.94 0.96 1.09 1.00 1.10

(0.74e1.20) (0.79e1.17) (0.85e1.40) (0.81e1.23) (0.84e1.44)

Values in model 1 are hazard ratios (95% CI) of Cox proportional hazard models adjusted for age, gender, and EEG system. Values in model 2 are hazard ratios (95% CI) of model 1 with the additional adjustment for medial temporal lobe atrophy score. Values in model 3 are hazard ratios (95% CI) of model 1 with the additional adjustment for CSF-tau level. All predictors were z-transformed before regression. The hazard ratios can be interpreted as the increased risk per SD increase in EEG relative power or peak frequency. * p < 0.05, #p < 0.10. Key: EEG, electroencephalography; MCI, mild cognitive impairment; SCD, subjective cognitive decline.

unknown amyloid status and reported increases in theta power, decreases in peak frequency and changes in covariance in right hemispheric regions in subjects who declined compared with stable subjects (Prichep et al., 2006). In cognitively normal patients with Parkinson’s disease, EEG spectral measures were also found to be strong prognostic markers for future dementia (Klassen et al., 2011). Contrary to our hypothesis, we did not find any EEG predictors for time to clinical progression in subjects with a baseline MCI diagnosis. As cross-sectional studies that had compared SCD, MCI, and AD patients found repeatedly that EEG findings in MCI group are intermediate to SCD and AD (Babiloni et al., 2011b; Jeong, 2004; Rossini et al., 2006), we hypothesized that progressive MCI patients would have higher theta and delta power and lower alpha and beta power than stable MCI patients. However, also in previous longitudinal prediction studies with MCI patients without proof of underlying AD pathology, the reported findings lack consistency. On one hand, several studies indeed found global slowing of resting-state rhythms (mainly increase in theta power and decrease in alpha power; Huang et al., 2000; Jelic et al., 2000; Luckhaus et al., 2008) or a more regional increase in delta power in the temporal lobe (Rossini et al., 2006) as predictors of clinical progression. In a study focusing on magnetic instead of electrical brain activity using magneto-encephalography, these findings were confirmed. More specifically, a higher theta power in a multivariate model with left hippocampal volume and a neuropsychological test discriminated progressive MCI from stable MCI patients (López et al., 2016). On the other hand, counterintuitive findings that contradict abovementioned studies have also been reported. An increase in power in higher frequency bands, reflected by an increase in alpha 3 [10.9e12.9 Hz]/alpha 2 [8.9e10.9 Hz] ratio in one study (Moretti et al., 2011) and increases of alpha1 power in the parieto-temporal and central-limbic areas in another study, has been reported to be predictive for clinical progression to dementia (Rossini et al., 2006). Moreover, a decrease of the posterior alpha peak amplitude was related to both improvement and worsening of cognitive functioning at follow-up (Babiloni et al., 2011a). A study that undertook a data mining approach using 35 EEG biomarkers extracted from several signal processing domains found that a combination of mainly beta activity-related biomarkers classified stable versus progressive MCI patients more accurately than individual biomarkers alone (Poil et al., 2013). Finally, it has also been reported that EEG theta power had additional value to a prediction model with neuropsychological variables, but EEG alone did not have a high prognostic value (van der Hiele et al., 2008). The negative findings in our MCI sample should therefore be seen in the context of these heterogeneous previous results. Moreover, unlike earlier studies, we preselected the study sample based on amyloid positivity, and it has been described in earlier studies that CSF biomarkers

are related to EEG in both cognitively normal subjects and memory clinic patients (Kramberger et al., 2013; Stomrud et al., 2010). To explore whether our selection of amyloid positivity influenced our results, we investigated whether EEG spectral measures differed between amyloid positive and negative patients and found that amyloid positive MCI patients were more severely affected with regard to cortical function than amyloid negative patients (Supplementary Material). We may hypothesize that MCI patients who are amyloid positive form a subgroup in whom functional damage, reflected by EEG spectral measures, occurs in an earlier stage of the disease and only evolves in a slower rate, thereby decreasing their predictive value at this stage of the disease. In cross-sectional studies comparing controls, MCI patients and AD patients, the left posterior brain region has repeatedly been described to be the best discriminator between diagnosis groups (Besga et al., 2010; Jelic et al., 2000; Luckhaus et al., 2008; Rice et al., 1991; Ueda et al., 2013). In a study analyzing magnetoencephalography dipole density across the cognitive spectrum of AD, left and right temporo-occipital regions could discriminate between MCI and controls, and differences in the left parietal region were found between MCI and AD patients (Besga et al., 2010). In our longitudinal study, regional analyses at electrode level confirmed that the left posterior temporal region is important in the prediction of clinical progression in patients with SCD at baseline. In addition, theta power at the fronto-central midline was also predictive for clinical progression. This finding may be related to the earlier reported anterior shift of resting-state oscillatory activity sources toward frontal regions in progressive MCI and AD patients, reflecting a relative decrease in activity of the posterior/occipital generators (Babiloni et al., 2011a; Engels et al., 2015; Huang et al., 2000; Osipova et al., 2005). Our specific findings in the SCD group indicate that synaptic dysfunction may be a process that starts very early in the pathophysiological cascade of AD, before cognitive decline had started. Our results support the model proposed by the National Institute on Aging-Alzheimer’s Association in which synaptic dysfunction is one of the earliest pathophysiological changes starting before Taumediated neuronal injury occurs (Sperling et al., 2011). EEG oscillations reflect the coupling of local groups of cortical inhibitory neurons and cortical excitatory pyramidal neurons. It can be hypothesized that less activation of these neurons by acetylcholine or synaptic damage to these cortical inhibitoryeexcitatory feedback loops leads to a lower alpha peak frequency, a shift from alpha to theta power, and higher amplitudes of the oscillations. Moreover, it has been suggested that toxic soluble amyloid beta oligomers interfere specifically with synaptic function (Nimmrich and Ebert, 2009), a process that is not captured by measurements in CSF as these methods primarily reflect fibrillary amyloid deposits (Cheng et al., 2014; Shankar et al., 2008). This provides support for our

140

A.A. Gouw et al. / Neurobiology of Aging 57 (2017) 133e142

findings that EEG is a prognostic marker additive to amyloid positivity in CSF. The strengths of this study are the use of a large, wellcharacterized clinical cohort with standardized diagnostic workup of patients. In addition, the availability of CSF amyloid enabled us to select a relatively large cohort of nondemented, amyloid positive subjects. We were therefore able to specifically investigate the predictive value of EEG in subjects with proven underlying AD pathology. This improves on previous studies, in which predictive effects might have been overestimated due to a heterogeneous mix of underlying pathology. To define our patient population, we used a CSF amyloid-b level of 640 pg/mL as a cut-off value for amyloid positivity. Generally, cut-offs are used for the diagnosis of AD patients and are determined by comparison of center-specific AD versus control groups, which introduces variability in CSF cut-offs (Dumurgier et al., 2013; Verwey and Flier, 2009). Moreover, older healthy controls can harbor abnormal amyloid and may therefore introduce bias. For this study, we used a cut-off based on the high concordance with amyloid PET imaging in patients ranging from cognitively normal to dementia (Zwan et al., 2014). It has been shown that this cut-off more accurately reflects amyloid deposition in the brain than clinically based cut-off values (Zwan et al., 2016) and is therefore more appropriate in our study in nondemented patients than clinically based cut-offs. However, it is still possible that patients who were borderline amyloid negative, actually have underlying AD pathology and were unjustly excluded from this cohort. On the other hand, it is known that abnormal amyloid deposition also exists in other neurodegenerative diseases, such as DLB (Halliday et al., 2011). EEG changes occur early in the course of DLB, and it has been described that EEG in the MCI stage of DLB is predictive for conversion to dementia (Bonanni et al., 2015). Of the clinically progressive patients of our cohort, only a small proportion (5 patients; 4.6%) developed another diagnosis than MCI or AD type of dementia: 1 of the patients converted to DLB, 3 patients were diagnosed with vascular dementia at follow-up, and 1 patient developed frontotemporal lobe dementia. It is therefore likely that the contribution of DLB in our sample of nondemented patients is limited. However, for the SCD patients who progressed to MCI, it is still unclear what the underlying cause is. These considerations should be kept in mind, as repeated analyses using other cut-offs, that is, more or less stringent values or combinations of amyloidb and tau (Dubois et al., 2014; Mulder et al., 2010), potentially have an impact on the results. Another potential limitation is the use of 3 EEG systems over the years, which, although we have corrected for this potential confounder, may have introduced noise into the analyses. However, this can also be considered to be a strength of the study as it suggests that the prognostic measures we found are robust for the use of different systems and have the potential to be used as biomarkers in multicenter clinical trials. Finally, we followed standard epoch selection procedures based on visual inspection of the EEG recording, resulting in EEG data that contained a minimum, but not complete absence, of artifacts. Although supplementary analyses support that our results are not explained by trivial group differences in the artifact load, it would still be recommended to reproduce our findings in future independent cohorts. In this study, the prognostic value of EEG appeared to be of modest strength, which implies that EEG spectral analysis alone will not be sufficient to make individual predictions for clinical progression. Future studies may focus on (a combination with) other EEG measures that not only take into account local coupling of cortical neurons but also reflect the functional interactions between brain regions and their organization in a functional network. Normal cognition depends on those processes and alterations in functional connectivity and in brain network organization have

been described in AD and MCI patients (de Haan et al., 2012; Stam et al., 2007; Tijms et al., 2013; Vecchio et al., 2014). Possibly, the combination of EEG with other modalities, such as neuropsychological tests, structural/functional MRI or neurodegenerative markers is eventually required to reach a prognostic accuracy that has clinical use. Promising for future studies are analytical methods based on machine learning, as they are especially powerful when various potential predictors are combined (Deo, 2015). For the design of prevention trials in preclinical AD, EEG may be an additive tool for selecting subjects that have a high risk of conversion to dementia. Although not very sensitive, it is widely available, cheap, and relatively easy to apply. Alternatively, EEG might be a useful surrogate outcome biomarker in clinical trials, which will be further explored in future studies (de Waal et al., 2014; van Straaten et al., 2014). 5. Conclusions In conclusion, EEG spectral measures were related to clinical progression in nondemented subjects with proven amyloid pathology, specifically in the very early stages of the disease, when cognitive disturbances have not yet occurred. Disclosure statement Ph. Scheltens has received grant support (for the institution) from GE Healthcare, Danone Research, Piramal and Merck. In the past 2 years, the author has received consultancy/speaker fees (paid to the institution) from Lilly, GE Healthcare, Novartis, Forum, Sanofi, Nutricia, Probiodrug and EIP Pharma. Research programs of W. van der Flier have been funded by ZonMW, NWO, EU-FP7, Alzheimer Nederland, CardioVascular Onderzoek Nederland, stichting Dioraphte, Gieskes-Strijbis fonds, Boehringer Ingelheim, Piramal Neuroimaging, Roche BV, Janssen Stellar, Combinostics. All funding is paid to her institution. Wiesje M. van der Flier has been an invited speaker at Boehringer Ingelheim. The remaining authors declare no conflict of interest. Acknowledgements Research of the VUmc Alzheimer Center is part of the neurodegeneration research program of the Neuroscience Campus Amsterdam. The VUmc Alzheimer Center is supported by Stichting Alzheimer Nederland and Stichting VUmc fonds. The clinical database structure was developed with funding from Stichting Dioraphte. This study was funded by a research grant of Boehringer Ingelheim Pharma GmbH Co KG, Germany. The authors acknowledge Meichen Yu for his help with the figure. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.neurobiolaging. 2017.05.017. References Albert, M.S., DeKosky, S.T., Dickson, D., Dubois, B., Feldman, H.H., Fox, N.C., Gamst, A., Holtzman, D.M., Jagust, W.J., Petersen, R.C., Snyder, P.J., Carrillo, M.C., Thies, B., Phelps, C.H., 2011. The diagnosis of mild cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on AgingAlzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers. Dement 7, 270e279. Babiloni, C., Frisoni, G.B., Vecchio, F., Lizio, R., Pievani, M., Cristina, G., Fracassi, C., Vernieri, F., Rodriguez, G., Nobili, F., Ferri, R., Rossini, P.M., 2011a. Stability of clinical condition in mild cognitive impairment is related to cortical sources of

A.A. Gouw et al. / Neurobiology of Aging 57 (2017) 133e142 alpha rhythms: an electroencephalographic study. Hum. Brain Mapp. 32, 1916e1931. Babiloni, C., Vecchio, F., Lizio, R., Ferri, R., Rodriguez, G., Marzano, N., Frisoni, G.B., Rossini, P.M., 2011b. Resting state cortical rhythms in mild cognitive impairment and Alzheimer’s disease: electroencephalographic evidence. J. Alzheimers. Dis. 26 (Suppl 3), 201e214. Besga, A., Ortiz, L., Fernández, A., Maestu, F., Arrazola, J., Gil-Gregorio, P., Fuentes, M., Ortiz, T., 2010. Structural and functional patterns in healthy aging, mild cognitive impairment, and Alzheimer disease. Alzheimer Dis. Assoc. Disord. 24, 1e10. Bonanni, L., Perfetti, B., Bifolchetti, S., Taylor, J., Franciotti, R., Parnetti, L., Thomas, A., Onofrj, M., 2015. Neurobiology of Aging Quantitative electroencephalogram utility in predicting conversion of mild cognitive impairment to dementia with Lewy bodies. Neurobiol. Aging 36, 434e445. Cheng, X., Wu, J., Geng, M., Xiong, J., 2014. The role of synaptic activity in the regulation of amyloid beta levels in Alzheimer’s disease. Neurobiol. Aging 35, 1217e1232. de Haan, W., Stam, C.J., Jones, B.F., Zuiderwijk, I.M., van Dijk, B.W., Scheltens, P., 2008. Resting-state oscillatory brain dynamics in Alzheimer disease. J. Clin. Neurophysiol. 25, 187e193. de Haan, W., van der Flier, W.M., Koene, T., Smits, L.L., Scheltens, P., Stam, C.J., 2012. Disrupted modular brain dynamics reflect cognitive dysfunction in Alzheimer’s disease. Neuroimage 59, 3085e3093. de Waal, H., Stam, C.J., de Haan, W., van Straaten, E.C.W., Scheltens, P., van der Flier, W.M., 2012. Young Alzheimer patients show distinct regional changes of oscillatory brain dynamics. Neurobiol. Aging 33, 1008.e25e1008.e31. de Waal, H., Stam, C.J., Lansbergen, M.M., Wieggers, R.L., Kamphuis, P.J.G.H., Scheltens, P., Maestú, F., van Straaten, E.C.W., 2014. The effect of souvenaid on functional brain network organisation in patients with mild Alzheimer’s disease: a randomised controlled study. PLoS One 9, e86558. Deo, R.C., 2015. Machine learning in medicine. Circulation 132, 1920e1930. Dubois, B., Feldman, H.H., Jacova, C., Hampel, H., Molinuevo, J.L., Blennow, K., DeKosky, S.T., Gauthier, S., Selkoe, D., Bateman, R., Cappa, S., Crutch, S., Engelborghs, S., Frisoni, G.B., Fox, N.C., Galasko, D., Habert, M.-O., Jicha, G.A., Nordberg, A., Pasquier, F., Rabinovici, G., Robert, P., Rowe, C., Salloway, S., Sarazin, M., Epelbaum, S., de Souza, L.C., Vellas, B., Visser, P.J., Schneider, L., Stern, Y., Scheltens, P., Cummings, J.L., 2014. Advancing research diagnostic criteria for Alzheimer’s disease: the IWG-2 criteria. Lancet Neurol. 13, 614e629. Dumurgier, J., Vercruysse, O., Paquet, C., Bombois, S., Chaulet, C., Laplanche, J.L., Peoc’h, K., Schraen, S., Pasquier, F., Touchon, J., Hugon, J., Lehmann, S., Gabelle, A., 2013. Intersite variability of CSF Alzheimer’s disease biomarkers in clinical setting. Alzheimer’s Dement 9, 406e413. Eckerström, C., Olsson, E., Klasson, N., Berge, J., Nordlund, A., Bjerke, M., Wallin, A., 2015. Multimodal prediction of dementia with up to 10 years follow up: the Gothenburg MCI study. J. Alzheimer’s Dis. 44, 205e214. Engels, M.M.A., Stam, C.J., van der Flier, W.M., Scheltens, P., de Waal, H., van Straaten, E.C.W., 2015. Declining functional connectivity and changing hub locations in Alzheimer’s disease: an EEG study. BMC Neurol. 15, 145. Giannakopoulos, P., Missionier, P., Kovari, E., Gold, G., Michon, A., 2009. Electrophysiological markers of rapid cognitive decline in mild. Front Neurol. Neurosci. 24, 39e46. Halliday, G.M., Holton, J.L., Revesz, T., Dickson, D.W., 2011. Neuropathology underlying clinical variability in patients with synucleinopathies. Acta Neuropathol. 122, 187e204. Hampel, H., Wilcock, G., Andrieu, S., Aisen, P., Blennow, K., Broich, K., Carrillo, M., Fox, N.C., Frisoni, G.B., Isaac, M., Lovestone, S., Nordberg, A., Prvulovic, D., Sampaio, C., Scheltens, P., Weiner, M., Winblad, B., Coley, N., Vellas, B., 2011. Biomarkers for Alzheimer’s disease therapeutic trials. Prog. Neurobiol. 95, 579e593. Huang, C., Wahlund, L.-O., Dierks, T., Julin, P., Winblad, B., Jelic, V., 2000. Discrimination of Alzheimer’s disease and mild cognitive impairment by equivalent EEG sources: a cross-sectional and longitudinal study. Clin. Neurophysiol. 111, 1961e1967. Jack, C.R., Knopman, D.S., Jagust, W.J., Petersen, R.C., Weiner, M.W., Aisen, P.S., Shaw, L.M., Vemuri, P., Wiste, H.J., Weigand, S.D., Lesnick, T.G., Pankratz, V.S., Donohue, M.C., Trojanowski, J.Q., 2013. Tracking pathophysiological processes in Alzheimer’s disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol. 12, 207e216. Jelic, V., Johansson, S.-E., Almkvist, O., Shigeta, M., Julin, P., Nordberg, A., Winblad, B., Wahlund, L.-O., 2000. Quantitative electroencephalography in mild cognitive impairment: longitudinal changes and possible prediction of Alzheimer’s disease. Neurobiol. Aging 21, 533e540. Jeong, J., 2004. EEG dynamics in patients with Alzheimer’s disease. Clin. Neurophysiol. 115, 1490e1505. Jessen, F., Amariglio, R.E., van Boxtel, M., Breteler, M., Ceccaldi, M., Chételat, G., Dubois, B., Dufouil, C., Ellis, K.A., van der Flier, W.M., Glodzik, L., van Harten, A.C., de Leon, M.J., McHugh, P., Mielke, M.M., Molinuevo, J.L., Mosconi, L., Osorio, R.S., Perrotin, A., Petersen, R.C., Rabin, L.A., Rami, L., Reisberg, B., Rentz, D.M., Sachdev, P.S., de la Sayette, V., Saykin, A.J., Scheltens, P., Shulman, M.B., Slavin, M.J., Sperling, R.A., Stewart, R., Uspenskaya, O., Vellas, B., Visser, P.J., Wagner, M., 2014. A conceptual framework for research on subjective cognitive decline in preclinical Alzheimer’s disease. Alzheimers. Dement 10, 844e852. Jobert, M., Wilson, F.J., Ruigt, G.S.F., Brunovsky, M., Prichep, L.S., Drinkenburg, W.H.I.M., 2012. Guidelines for the recording and evaluation of pharmaco-EEG data in man: the International Pharmaco-EEG Society (IPEG). Neuropsychobiology 66, 201e220.

141

Klassen, B.T., Hentz, J.G., Shill, H.A., Driver-Dunckley, E., Evidente, V.G.H., Sabbagh, M.N., Adler, C.H., Caviness, J.N., 2011. Quantitative EEG as a predictive biomarker for Parkinson disease dementia. Neurology 77, 118e124. Kramberger, M.G., Kåreholt, I., Andersson, T., Winblad, B., Eriksdotter, M., Jelic, V., 2013. Association between EEG abnormalities and CSF biomarkers in a memory clinic cohort. Dement. Geriatr. Cogn. Disord. 36, 319e328. Kwak, Y.T., 2006. Quantitative EEG findings in different stages of Alzheimer’s disease. J. Clin. Neurophysiol. 23, 456e461. López, M.E., Turrero, A., Cuesta, P., López-Sanz, D., Bruña, R., Marcos, A., Gil, P., Yus, M., Barabash, A., Cabranes, J.A., Maestú, F., Fernández, A., 2016. Searching for primary predictors of conversion from mild cognitive impairment to Alzheimer’s disease: a multivariate follow-up study. J. Alzheimers. Dis. 52, 133e143. Luckhaus, C., Grass-kapanke, B., Blaeser, I., Ihl, R., Supprian, T., 2008. Quantitative EEG in progressing vs stable mild cognitive impairment ( MCI ): results of a 1year follow-up study. Int. J. Geriatr. Psychiatry 23, 1148e1155. McKhann, G.M., Knopman, D.S., Chertkow, H., Hyman, B.T., Jack, C.R., Kawas, C.H., Klunk, W.E., Koroshetz, W.J., Manly, J.J., Mayeux, R., Mohs, R.C., Morris, J.C., Rossor, M.N., Scheltens, P., Carrillo, M.C., Thies, B., Weintraub, S., Phelps, C.H., 2011. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers. Dement 7, 263e269. Moretti, D.V., Frisoni, G.B., Fracassi, C., Pievani, M., Geroldi, C., Binetti, G., Rossini, P.M., Zanetti, O., 2011. MCI patients’ EEGs show group differences between those who progress and those who do not progress to AD. Neurobiol. Aging 32, 563e571. Mulder, C., Verwey, N.A., van der Flier, W.M., Bouwman, F.H., Kok, A., van Elk, E.J., Scheltens, P., Blankenstein, M.A., 2010. Amyloid-beta(1-42), total tau, and phosphorylated tau as cerebrospinal fluid biomarkers for the diagnosis of Alzheimer disease. Clin. Chem. 56, 248e253. Nimmrich, V., Ebert, U., 2009. Is Alzheimers disease a result of presynaptic failure? Synaptic dysfunctions induced by oligomeric beta-amyloid. Rev. Neurosci. 20, 1e12. Noble, W.S., 2009. How does multiple testing correction work? Nat. Biotechnol. 27, 1135e1137. Nuwer, M.R., 1988. Quantitative EEG: I. Techniques and problems of frequency analysis and topographic mapping. J. Clin. Neurophysiol. 5, 1e43. Osipova, D., Ahveninen, J., Jensen, O., Ylikoski, A., Pekkonen, E., 2005. Altered generation of spontaneous oscillations in Alzheimer’s disease. Neuroimage 27, 835e841. Palop, J.J., Mucke, L., 2010. Amyloid-beta-induced neuronal dysfunction in Alzheimer’s disease: from synapses toward neural networks. Nat. Neurosci. 13, 812e818. Poil, S.-S., de Haan, W., van der Flier, W.M., Mansvelder, H.D., Scheltens, P., Linkenkaer-Hansen, K., 2013. Integrative EEG biomarkers predict progression to Alzheimer’s disease at the MCI stage. Front. Aging Neurosci. 5, 58. Prestia, A., Caroli, A., Wade, S.K., van der Flier, W.M., Ossenkoppele, R., Van Berckel, B., Barkhof, F., Teunissen, C.E., Wall, A., Carter, S.F., Schöll, M., Choo, I.H., Nordberg, A., Scheltens, P., Frisoni, G.B., 2015. Prediction of AD dementia by biomarkers following the NIA-AA and IWG diagnostic criteria in MCI patients from three European memory clinics. Alzheimers. Dement 1, 1552e5260. Prichep, L.S., John, E.R., Ferris, S.H., Rausch, L., Fang, Z., Cancro, R., Torossian, C., Reisberg, B., 2006. Prediction of longitudinal cognitive decline in normal elderly with subjective complaints using electrophysiological imaging. Neurobiol. Aging 27, 471e481. Rice, D.M., Buchsbaum, M.S., Hardy, D., Burgwald, L., 1991. Focal left temporal slow EEG activity is related to a verbal recent memory deficit in a non-demented elderly population. J. Gerontol. 46, P144eP151. Rossini, P.M., Del Percio, C., Pasqualetti, P., Cassetta, E., Binetti, G., Dal Forno, G., Ferreri, F., Frisoni, G., Chiovenda, P., Miniussi, C., Parisi, L., Tombini, M., Vecchio, F., Babiloni, C., 2006. Conversion from mild cognitive impairment to Alzheimer’s disease is predicted by sources and coherence of brain electroencephalography rhythms. Neuroscience 143, 793e803. Scheltens, P., Launer, L.J., Barkhof, F., Weinstein, H.C., van Gool, W.A., 1995. Visual assessment of medial temporal lobe atrophy on magnetic resonance imaging: interobserver reliability. J. Neurol. 242, 557e560. Selkoe, D.J., 2002. Alzheimer’s disease is a synaptic failure. Science 80, 789e792. Shankar, G.M., Li, S., Mehta, T.H., Garcia-Munoz, A., Shepardson, N.E., Smith, I., Brett, F.M., Farrell, M.A., Rowan, M.J., Lemere, C.A., Regan, C.M., Walsh, D.M., Sabatini, B.L., Selkoe, D.J., 2008. Amyloid-beta protein dimers isolated directly from Alzheimer’s brains impair synaptic plasticity and memory. Nat. Med. 14, 837e842. Sierra-Rio, A., Balasa, M., Olives, J., Antonell, A., Iranzo, A., Castellví, M., Bosch, B., Grau-Rivera, O., Fernandez-Villullas, G., Rami, L., Lladó, A., Sánchez-Valle, R., Molinuevo, J.L., 2016. Cerebrospinal fluid biomarkers predict clinical evolution in patients with subjective cognitive decline and mild cognitive impairment. Neurodegener. Dis. 16, 69e76. Singh, K.D., Barnes, G.R., Hillebrand, A., 2003. Group imaging of task-related changes in cortical synchronisation using nonparametric permutation testing. Neuroimage 19, 1589e1601. Sperling, R.A., Aisen, P.S., Beckett, L.A., Bennett, D.A., Craft, S., Fagan, A.M., Iwatsubo, T., Jack, C.R., Kaye, J., Montine, T.J., Park, D.C., Reiman, E.M., Rowe, C.C., Siemers, E., Stern, Y., Yaffe, K., Carrillo, M.C., Thies, B., Morrison-Bogorad, M., Wagster, M.V., Phelps, C.H., 2011. Toward defining the preclinical stages of Alzheimer’s disease: recommendations from the National Institute on

142

A.A. Gouw et al. / Neurobiology of Aging 57 (2017) 133e142

Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers. Dement 7, 280e292. Sperling, R.A., Karlawish, J., Johnson, K.A., 2013. Preclinical Alzheimer disease-the challenges ahead. Nat. Rev. Neurol. 9, 54e58. Stam, C.J., Jones, B.F., Nolte, G., Breakspear, M., Scheltens, P., 2007. Small-world networks and functional connectivity in Alzheimer’s disease. Cereb. Cortex 17, 92e99. Stomrud, E., Hansson, O., Minthon, L., Blennow, K., Rosén, I., Londos, E., 2010. Slowing of EEG correlates with CSF biomarkers and reduced cognitive speed in elderly with normal cognition over 4 years. Neurobiol. Aging 31, 215e223. Tijms, B.M., Wink, A.M., de Haan, W., van der Flier, W.M., Stam, C.J., Scheltens, P., Barkhof, F., 2013. Alzheimer’s disease: connecting findings from graph theoretical studies of brain networks. Neurobiol. Aging 34, 2023e2036. Ueda, T., Musha, T., Yagi, T., 2013. Diagnosis method of mild cognitive impairment based on power variance of EEG. In: Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, pp. 6003e6006. van der Flier, W.M., Pijnenburg, Y.A.L., Prins, N., Lemstra, A.W., Bouwman, F.H., Teunissen, C.E., van Berckel, B.N.M., Stam, C.J., Barkhof, F., Visser, P.J., van Egmond, E., Scheltens, P., 2014. Optimizing patient care and research: the Amsterdam dementia cohort. J. Alzheimers. Dis. 41, 313e327. van der Hiele, K., Bollen, E.L.E.M., Vein, A.A., Reijntjes, R.H.A.M., Westendorp, R.G.J., Van Buchem, M.A., Middelkoop, H.A.M., Van Dijk, J.G., 2008. EEG markers of future cognitive performance in the elderly. J. Clin. Neurophysiol. 25, 83e89. van der Hiele, K., Vein, A.A., Reijntjes, R.H.A.M., Westendorp, R.G.J., Bollen, E.L.E.M., van Buchem, M.A., van Dijk, J.G., Middelkoop, H.A.M., 2007. EEG correlates in the spectrum of cognitive decline. Clin. Neurophysiol. 118, 1931e1939. van Diessen, E., Numan, T., van Dellen, E., van der Kooi, A.W., Boersma, M., Hofman, D., van Lutterveld, R., van Dijk, B.W., van Straaten, E.C.W., Hillebrand, A., Stam, C.J., 2015. Opportunities and methodological challenges in EEG and MEG resting state functional brain network research. Clin. Neurophysiol. 126, 1468e1481. Van Harten, A.C., Visser, P.J., Pijnenburg, Y.A.L., Teunissen, C.E., Blankenstein, M.A., Scheltens, P., Van Der Flier, W.M., 2013. Cerebrospinal fluid AB42 is the best predictor of clinical progression in patients with subjective complaints. Alzheimer’s Dement 9, 481e487. van Rossum, I.A., Visser, P.J., Knol, D.L., van der Flier, W.M., Teunissen, C.E., Barkhof, F., Blankenstein, M.A., Scheltens, P., 2012a. Injury markers but not

amyloid markers are associated with rapid progression from mild cognitive impairment to dementia in Alzheimer’s disease. J. Alzheimers. Dis. 29, 319e327. van Rossum, I.A., Vos, S.J.B., Burns, L., Knol, D.L., Scheltens, P., Soininen, H., Wahlund, L.-O., Hampel, H., Tsolaki, M., Minthon, L., L’italien, G., van der Flier, W.M., Teunissen, C.E., Blennow, K., Barkhof, F., Rueckert, D., Wolz, R., Verhey, F., Visser, P.J., 2012b. Injury markers predict time to dementia in subjects with MCI and amyloid pathology. Neurology 79, 1809e1816. van Straaten, E.C., Scheltens, P., Gouw, A.A., Stam, C.J., 2014. Eyes-closed task-free electroencephalography in clinical trials for Alzheimer’s disease: an emerging method based upon brain dynamics. Alzheimers. Res. Ther. 6, 86. Vecchio, F., Miraglia, F., Marra, C., Quaranta, D., Vita, M.G., Bramanti, P., Rossini, P.M., 2014. Human brain networks in cognitive decline: a graph theoretical analysis of cortical connectivity from EEG data. J. Alzheimers. Dis. 41, 113e127. Verwey, N.A., Van Der Flier, W.M., Blennow, K., Clark, C., Sokolow, S., de Deyn, P.P., Galasko, D., Hampel, H., Hartmann, T., Kapaki, E., Lannfelt, L., Mehta, P.D., Parnetti, L., Petzold, A., Pirttila, T., Saleh, L., Skinningsrud, A., van Swieten, J.C., Verbeek, M.M., Wiltfang, J., Younkin, S., Scheltens, P., Blankenstein, M.A., 2009. A worldwide multicentre comparison of assays for cerebrospinal fluid biomarkers in Alzheimer’s disease. Ann. Clin. Biochem. 181, 235e240. Vos, S.J.B., van Rossum, I.A., Verhey, F., Knol, D.L., Soininen, H., Wahlund, L.-O., Hampel, H., Tsolaki, M., Minthon, L., Frisoni, G.B., Froelich, L., Nobili, F., van der Flier, W., Blennow, K., Wolz, R., Scheltens, P., Visser, P.J., 2013. Prediction of Alzheimer disease in subjects with amnestic and nonamnestic MCI. Neurology 80, 1124e1132. Whitham, E.M., Pope, K.J., Fitzgibbon, S.P., Lewis, T., Clark, C.R., Loveless, S., Broberg, M., Wallace, A., DeLosAngeles, D., Lillie, P., Hardy, A., Fronsko, R., Pulbrook, A., Willoughby, J.O., 2007. Scalp electrical recording during paralysis: quantitative evidence that EEG frequencies above 20 Hz are contaminated by EMG. Clin. Neurophysiol. 118, 1877e1888. Zwan, M., van Harten, A., Ossenkoppele, R., Bouwman, F., Teunissen, C., Adriaansea, S., Lammertsma, A., Scheltens, P., van Berckel, B., van der Flier, W.M., 2014. Concordance between cerebrospinal fluid biomarkers and [C]PIB PET in a memory clinic cohort. J. Alzheimer’s Dis. 41, 801e807. Zwan, M.D., Rinne, J.O., Steen, G., Nordberg, A., Lleo, A., Herukka, S.-K., Soininen, H., Law, I., Bahl, J.M.C., Carter, S.F., Fortea, J., Blesa, R., Teunissen, C.E., Bouwman, F.H., van Berckel, B.N.M., Visser, P.J., 2016. Use of amyloid-PET to determine cutpoints for CSF markers A multicenter study. Neurology 86, 50e58.