2.1 Participants
This study involved 106 participants who were over 60 years of age. Patients were recruited from local memory services if they had a clinical diagnosis of MCI and reported additional clinical symptoms suggestive of Lewy body disease (e.g. mood changes, sleep disturbances, or autonomic symptoms) or any core DLB features (visual hallucinations, cognitive fluctuations, Parkinsonism, and REM sleep behavior disorder). After obtaining written informed consent, participants underwent a baseline clinical assessment including theAddenbrooke’s Cognitive Examination – Revised (ACE-R), from which the Mini Mental State Examination (MMSE) score was derived, the Unified Parkinson’s Disease Rating Scale motor sub-score (UPDRS-III), the Epworth Sleepiness Scale (ESS), and the Geriatric Depression Scale (GDS). Additionally, the Instrumental Activities of Daily Living (IADL) scale, the Clinician Assessment of Fluctuations (CAF), the Dementia Cognitive Fluctuations Scale (DCFS), the North-East Visual Hallucinations Interview (NEVHI), the Neuropsychiatric Inventory (NPI), and the Mayo Sleep Questionnaire (MSQ) were administered to informants and the Clinical Dementia Rating (CDR) scale and the Cumulative Illness Rating scale for Geriatrics (CIRS-G) were completed based on clinical history and research assessments.In addition to a detailed clinical assessment, participants had already undergone dopaminergic imaging with 123I-N-fluoropropyl-2β-carbomethoxy-3β-(4-iodophenyl) single-photon emission computed tomography (FP-CIT SPECT) and 123iodine-metaiodobenzylguanidine (MIBG) myocardial scintigraphy through their involvement in an ongoing study investigating the diagnostic accuracy of imaging biomarkers in MCI and this information was used to apply diagnostic criteria (see below).
After the initial baseline visit, participants were followed annually for up to three years with a mean follow-up time of 16.5 months (standard deviation=6.5 months). Patients who were taking dopaminergic medication were assessed in the “ON” motor state.
2.2 Diagnosis
Diagnoses were based on all information available at the end of the recruitment period (December 2019) including any baseline and follow-up visits. MCI diagnoses were made independently by a consensus panel of three experienced old-age psychiatrists (AJT, PCD, JPT) in accordance with NIA-AA criteria [2], i.e. subjective and objective cognitive impairment with maintained independence of function with minimal aids or assistance and a CDR of 0 or 0.5. Patients with a diagnosis of dementia were excluded from the study. Furthermore, patients with possible contributing frontotemporal or vascular etiologies or with a history of Parkinsonism of more than one year prior to the onset of cognitive impairment were excluded. The presence or absence of the core Lewy body symptoms was rated by the panel utilizing the rating scales and all information from the clinical assessments[6]. Findings from the FP-CIT and MIBG scans were used for diagnosis (see below), but the clinical MCI diagnoses as well as the rating of presence/absence of core DLB symptoms wereperformed blind to these imaging findings.
A diagnosis of MCI with probable Alzheimer’s disease (MCI-AD) was given to patients who had no core Lewy body symptoms, negative FP-CIT and MIBG findings, and evidence of cognitive decline that was characteristic of AD, i.e. they met the additional NIA-AA criterion for “etiology of MCI consistent with AD pathophysiologic process” [2].
Probable MCI with Lewy bodies (MCI-LB) was diagnosed if a patient had two or more core Lewy body symptoms or one core symptom in addition to a positive FP-CIT or MIBG scan[5].
Out of 103 MCI participants who were included in the study, 20had only one core Lewy body symptom with negative FP-CIT and MIBG scans or no core symptoms in addition to a positive FP-CIT or MIBG scan, i.e. they did not meet criteria for either MCI-AD or probable MCI-LB and were therefore not included in the present analysis (Figure 1). Additionally, eight MCI participants did not have usable EEG data available. This study therefore included 39participants who were diagnosed with probable MCI-LB and 36 who were diagnosed with MCI-AD.Healthy control participants (N=31) were recruited from relatives and friends of patients and from a local research register. Control participants had the same assessment as the patients and had no history of psychiatric or neurological illness and no evidence of any cognitive decline. They also had normal structural MR imaging.
The study was approved by the local ethics committee and written informed consent was obtained from all participants.
2.3 EEG acquisition and pre-processing
Resting state high-density EEG recordings were acquired from all participants using Waveguard caps (ANT Neuro, The Netherlands) comprising 128 sintered Ag/AgCl electrodes that were placed according to the 10-5 system. Participants were seated during the recording and instructed to remain awake. Electrode impedance was kept below 5kΩ and continuous EEG data were recorded at a sampling frequency of 1024 Hz. 300 seconds of eyes-closed data were recorded from each participant. Participants were supervised by the EEG technician during the recording to monitor adherence to the protocol. The ground electrode was attached to the right clavicle and all EEG channels were referenced to Fz during recording.
Pre-processing of eyes-closed EEG data was performed using the EEGLAB toolbox (version 14) in Matlab (R2017a) [8] and was blind to group membership. First, EEG data were bandpass-filtered between 0.3 and 54 Hz using a second order Butterworth filter and split into non-overlapping epochs of 2 seconds. Subsequently, EEG recordings were visually inspected to identify noisy channels and noisy epochswhich were excluded prior to applying independent component analysis for further artefact removal. The resulting components were visually inspected and components representing muscular, cardiac, ocular, or electrical (50 Hz line noise) artefacts were rejected. The previously excluded channels were then replaced using spherical spline interpolation and data were recomputed against the average reference. For each participant, the first 45 2-second long artefact-free epochs were selected for further analysis.
2.4 Frequency analysis
For each 2-second epoch, the power spectral density (PSD) was estimated using Bartlett’s method in Matlab(R2017a) with a frequency resolution of 0.25 Hz and a Hamming window across the power spectrum from 2 to 30 Hz, for each electrode separately.To compensate for inter-individual variability in brain neurophysiology, anatomy, and physical tissue properties, the PSD was normalized by the total power across the power spectrum[9].
For each electrode separately, the mean power across all included epochs was estimated for different standard EEG frequency bands including delta (2-4 Hz), theta (4-5.5 Hz), pre-alpha (5.5-8 Hz), alpha (8-13 Hz), and beta (13-30 Hz). Higher frequencies were excluded because these are particularly affected by muscular artefacts [10]. The dominant frequency was calculated as the frequency with the highest power between 4 and 15 Hz (averaged across epochs). Dominant frequency was calculated for all electrodes as well as from occipital electrodes only (PO9, PO7, POO9h, PO5, O1, PO3, POO3h, OI1h, POz, Oz, PO4, POO4h, PO6, O2, OI2h, PO8, POO10h, PO10). Dominant frequency variability was defined as the standard deviation of dominant frequency across epochs[11].
2.5 Statistics
Statistical analyses were performed in SPSS and R (https://www.r-project.org/). Relative power within the different frequency bands was compared between the three groups using a multivariate ANOVAwith a within-subject factor of frequency band and a between-subject factor of diagnosis, followed by univariate ANOVAs and post-hoc tests, Bonferroni-corrected for multiple comparisons. Theta/alpha ratio, dominant frequency, anddominant frequency variabilitywere compared between the groups using univariate ANOVAs followed by post-hoc tests (Bonferroni-corrected). To account for differences in the number of male and female participants in the three groups, sex was included as a covariate in all analyses.
A receiver operating characteristics (ROC) analysis was conducted in R to assess sensitivity and specificity of the different EEG measures to distinguish between MCI-AD and MCI-LB patients. The sensitivity/specificity cut-off was determined using Youden’s index.
To assess the association between the range of Lewy body symptomatology and EEG abnormalities, we performed an exploratory analysis in whichwe compared quantitative EEG measures between MCI-LB patients who had two core symptoms or one core symptom and one abnormal biomarker (N=13) with those patients who had more than two core symptoms/abnormal biomarkers (N=26) using two-sample t-tests. Furthermore, Spearman’s correlations between symptom/biomarker count (ranging from 2 to 6) and the different EEG measures were computed in the MCI-LB group.
The association between quantitative EEG measures and overall cognitive impairment was assessed using Spearman’s correlations, in the MCI-AD and MCI-LB groups separately. P-values were FDR-corrected for multiple comparisons.
Additionally, we investigated the association between EEG characteristics and the core Lewy body symptoms of visual hallucinations and cognitive fluctuations which have been shown to be related to EEG abnormalities in dementia patients [12–14]. To this end,two-sample t-tests were performed,dichotomizing the MCI-LB group according to the presence/absence of visual hallucinations and cognitive fluctuations.
Given previous reports of aneffect of acetylcholinesterase inhibitors on the EEG signal [15,16], a two-sample t-test was performed in the MCI-LB group to compare EEG characteristics between patients who were taking acetylcholinesterase inhibitors and those patients not taking these medications.