Participants
All participants were recruited from the community via online advertisements or word of mouth. Participants included 21 healthy controls (14 female, 7 male; MAGE = 41.4 years, SDAGE = 13.2) and 14 participants who self-reported having received a diagnosis of MDD (9 female, 5 male; MAGE = 32.4 years, SDAGE = 15.2). The HC and MDD groups did not differ significantly in age or education. The MDD group had significantly higher scores on D-FIS (p = .002), HADS-A (p < .001), and HADS-D (p < .001) scores (see Table 1).
All participants were right-handed and had normal or corrected to normal vision and hearing. Participants were excluded if they had a diagnosis of any psychiatric illness (with the exclusion of depression for the MDD group), diagnosis of a learning disorder, a brain injury resulting in concussion or loss of consciousness less than 6 months prior, electroconvulsive therapy less than 1 year prior, or a diagnosis of any neurological illness.
Table 1. Participant Demographics and Symptom Scale Scores.
|
HC (n = 21)
|
MDD (n = 14)
|
p-value
|
Sex (M/F)
|
7/14
|
5/9
|
|
Age
|
41.43 (±13.15)
|
32.36 (±15.16)
|
.069
|
Education (in years)
|
15.38 (±2.10)
|
14.08 (±1.55)
|
.063
|
D-FIS
|
3.95 (±6.54)
|
11.29 (±6.04)
|
.002**
|
HADS-A
|
5.00 (±3.13)
|
10.92 (±3.73)
|
< .001**
|
HADS-D
|
2.00 (±2.55)
|
7.93 (±3.52)
|
< .001**
|
Medication Status
|
|
12 medicated (85.7%)
|
|
Note. The above table displays the group mean values for each demographic and clinical variable collected. ** indicates significant differences in group means at the p < .01 level.
Procedure
Test sessions were scheduled between 9:00 am and 11:00 am to ensure uniformity between participants and control for time of day effects (Hines, 2004). Participants were required to abstain from illicit drugs, alcohol, marijuana, and caffeine from midnight the night before their session. On arrival, verbal confirmation of abstinence was obtained; if the participant did not comply, the session was rescheduled. Participants then completed informed consent procedures and were given a series of questionnaires to complete. When finished, EEG set-up was completed, and the neurophysiological recordings were obtained. Participants were instructed to rest for 3 minutes with their eyes closed. All procedures and analyses were cleared by the Nova Scotia Health Authority Research Ethics Board (NSHA REB #1020019), the University Research Ethics Board of Mount Saint Vincent University (REB # 2015-031), and the St. Francis Xavier University Research Ethics Board (REB #22595).
Questionnaires
Hospital Anxiety and Depression Scale (HADS)
The HADS was used to measure symptoms of depression and anxiety. It is composed of seven self-report items measuring anxiety (eg, “I feel tense or wound up” and “I get sudden feelings of panic”) and seven items measuring depression (eg, "I feel as if I am slowed down" and "I still enjoy the things I used to enjoy"). Correlating with the Structured Clinical Interview for DSM-IV diagnostic criteria for MDD, the HADS has a sensitivity of 90% and a specificity of 87.3%, with a score of 8 or higher on the depression subscale (Honarmand & Feinstein, 2009).
Daily Fatigue Impact Scale (D-FIS)
The D-FIS was used to assess fatigue on the day of testing. It is an 8-item self-report questionnaire that is a valid and reliable measure of fatigue in a clinical population (Fisk & Doble, 2002). Examples of items included are “because of fatigue I feel slowed down in my thinking” and “because of fatigue I have to reduce my workload or responsibilities.”
EEG Parameters
EEG data were recorded from an electrode cap with Ag+/Ag+-Cl− electrodes at 19 scalp sites according to the 10-20 system of electrode placement. This included electrode placement at the following sites: Fp1, Fp2, F7, F3, Fz, F4, F8, T7, T8, C3, Cz, C4, P7, P3, Pz, P4, P8, O1, and O2. Electrodes were placed on bilateral mastoids to serve as a reference, and mid-forehead to serve as a ground. Bipolar recordings of horizontal (HEOG) and vertical (VEOG) electrooculogram activity were taken from supra-/sub-orbital and external canthi sites, respectively. All electrode impedances were kept below 10 kΩ at the time of recording. Electrical activity was recorded with an amplifier bandpass of DC-100Hz, digitized at 500 Hz, and stored on a hard disk for later offline analysis.
Microstate Analysis
Offline data processing included applying filters from 2-20 Hz with a notch filter at 60 Hz, segmentation into 2-second epochs (with no overlap), and artifact rejection of any epochs with electrical activity exceeding ±50µV. Microstate data was then analyzed using the microstates plug-in created for EEGlab by Thomas König (https://www.thomaskoenig.ch/index.php/software/microstates-in-eeglab/identifying-microstates-on-the-level-of-the-individual-eeg-first-level-clustering; Delorme & Makeig, 2004; Poulsen et al., 2018). First, individual microstate maps were identified by k-means spatial cluster analysis while disregarding the polarity of the maps. In accordance with recent reports of microstates in MDD (Li et al., 2021; Murphy et al., 2020), we selected five microstates (clusters) as the optimal cluster analysis. This decision was further validated by the significant increase of variance accounted for in the data by the 5-cluster model (M = 80.2%, SD = 7.7%) compared to a 4-cluster model (M = 79.4%, SD = 7.8%) determined by a paired samples t-test (t[34] = -9.96, p < .000). The individual microstate maps were then sorted according to distributions of scalp potentials from previously established maps (Koenig et al., 1999) and class labeled according to the Norms NI202 published template. Finally, the resulting class-labelled individual model maps were exported for statistical analysis. Within each participant, each microstate class yielded the following parameters:
- Duration: the mean length of that microstate class in seconds (s).
- Occurrence: the mean amount of observations of that specific microstate class each second.
- Contribution: the proportion (%) of total time spent in that specific microstate class while recording.
- Transitional probabilities: both the observed number of transitions from one microstate class to another, and the expected amount of transitions based on the observed occurrences in the dataset.
Topographic maps for each microstate class were averaged independently in each group and exported for visual inspection to ensure relative uniformity among distribution (see Figure 1).
Statistical Analysis
All data were analyzed using the Statistical Packages for Social Sciences (SPSS; IBM Corp, Armonk, NY). First, we completed independent samples t-tests to explore any significant differences between groups in the demographic variables of education and age, as well as symptoms scale scores. To examine group differences between the duration, occurrence, and contribution values for each microstate class (A, B, C, D, and E), separate multivariate general linear models (GLM) were completed where group (HC vs. MDD) served as a between-subject factor and Bonferroni adjustment for multiple comparisons was used. To examine differences between the amount of observed and expected transitional probabilities for each microstate transition pair, separate paired-samples t-tests were completed for all 20 transitional pairs in the HC and MDD groups separately. Finally, to examine microstate transitional abnormalities in both groups, the expected transition value was subtracted from the observed transition value and then transformed to an absolute value to derive a ∆transition value for all 20 microstate pairs. This ∆transition value was then compared between the HC and MDD groups using a multivariate GLM where group served as a between-subject factor, and Bonferroni adjustment for multiple comparisons was used.
To examine the relationship between the HADS and D-FIS symptom scale scores and microstate parameters, Spearman's bivariate correlations were completed between the duration, occurrence, and contribution of each microstate class and total scores on the HADS-A, HADS-D, and D-FIS in both HC and MDD groups. Additionally, to examine the relationship between these symptom scales and microstate transitional probabilities, Spearman’s bivariate correlations between total symptom scale scores, observed transitions, and ∆transition values were completed for each microstate pair in each group.