Quantitative EEG findings in patients with acute, brief depression combined with other fluctuating psychiatric symptoms: a controlled study from an acute psychiatric department

The following criteria had to be fulfilled for inclusion in the study group (AUDS patients): A history of a rapidly developing psychiatric condition starting within the last 14 days. Within these 14 days the patient had at different times shown symptoms that met criteria for at least two diagnoses in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) Axis 1 categories (with exception of the time criterion) [16]. One of these new DSM-IV Axis 1 categories had to be a depressive episode (with exception of the time criterion) defined as Montgomery and Aasberg Depression Rating Scale (MADRS) ≥ 20 [17]. Patients with one of the following conditions were excluded: a psychiatric condition due to direct effects of acute intoxication, dementia or mental disabilities to such a degree that informed consent could not be obtained, unstable personality disorder with identical symptoms at former admissions, or unable to speak English or Norwegian. From a total of 1984 admittances we evaluated 1038 persons for inclusion. Twenty-eight fulfilled criteria for inclusion in the study group. Twelve were not included (not willing to participate, informed consent not possible to obtain, mentally disabled to an extensive degree, or not speaking Norwegian or English). Sixteen patients entered the study group (AUDS). Mean age was 32.1 years (Table 1). Their co-diagnoses (with exception of the time criterion) were DSM-IV 298.8 "brief psychotic disorder" (nine patients), DSM-IV 300.1 "panic disorder" (four patients) and DSM-IV 296.0 "single manic episode" (three patients).

The control group consisted of 16 acutely admitted sex- and age (+/- 5 years), matched patients meeting criteria (MADRS ≥ 20) for current Axis 1 major depressive episode (MDE). Mean age was 32.8 years (Table 1). The first patient meeting such criteria after inclusion of a study group patient was recruited to the control group. Written consent was obtained from all the patients prior to inclusion. The study was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki. The Regional Ethical Committee approved of the study.

We examined the patients on day 2, day 4–6, day 14–16 and after 3 months. Alcohol use was assessed with The Alcohol Use Disorders Identification Test (AUDIT) [18]. Illicit drug use was assessed with the Structural Clinical Interview for DSM-IV (SCID-1) [19], and urine samples were screened for drugs at admittance. Current medication concentrations were analyzed in blood samples at admittance.

We obtained thirty minutes of eyes-closed EEG-videometry on day 2, day 4–5, and day 8–10 after admittance. Three tests were performed in order to increase the probability of detecting epileptiform activity as well as to evaluate time-related effects caused by e.g. drug use. They were all done at the same time of day (+/- 1.5 hours). We used digital equipment (Nervus 2.4 or 3.0) with 21 EEG electrodes placed according to the 10/20 international system, electrocardiography (ECG) and eye movement detection channels. Standard photic stimulation (1 to 60 Hz) and 3 min of hyperventilation were performed. Data loss occurred in one AUDS patient in test 2, and in 3 MDE patients in test 3. An experienced EEG physician blinded for the diagnosis interpreted and performed a clinical EEG analysis with classification of alpha frequency, epileptic spike activity and focal-and generalized slow activity. Eye opening and closing were performed every minute in order to control vigilance fluctuations. An experienced EEG technician noted presence of drowsiness in the recording. The technician, the physician and others doing the analyses and statistical work were blinded regarding the patient group.

We selected artefact free epochs of 4 seconds duration. Epochs with eyeblinks, muscle artefacts, electrode/movement artefacts, epileptiform activity and drowsiness were excluded. The total duration of such epochs was between 20 s and 60 s. We used the average reference montage, a band pass filter of 0.5 Hz – 70 Hz and a 50 Hz notch filter. Sixteen channels (F3, F4, F7, F8, T3, T4, T5, T6, O1, O2, C3, C4, P3, P4, T1 and T2) were exported to MATLAB™. The time/amplitude series had a sampling frequency of 256 Hz. For each selected EEG-segment we eliminated the start/end-discontinuities by using a cosine window. Fast Fourier Transform (FFT) was performed on four second samples (N = 1024 points), with two seconds overlap. This transform computes the power spectrum (EEG-power in μV² on the y-axis as a function of frequency in Hz in the x-axis). The average amplitude spectra (square root of power) across all the 4 seconds long samples were thereafter calculated and smoothed. Band amplitudes were calculated as the sum of spectral amplitudes for all bins in the frequency band and used in the statistical comparisons. The frequency bands were delta (1–3.75 Hz), theta (4–7.75 Hz), and alpha (8–12.75 Hz).

For regional analysis we averaged mean frontocentral, (F3, F4, F7, F8, C3, C4), temporal (T3, T4, T5, T6, T1, T2) and parietooccipital (O1, O2, P3, P4) amplitudes.

We calculated alpha peak frequencies at the occipital electrodes (O1 and O2). In addition we calculated the crude hemispheric asymmetry (absolute value of the difference between right side and left side: ABS (right-left) for all amplitude and peak frequency measures.

Categorical group differences were tested with Fisher's exact test. For the main QEEG analysis we used mean values (test one + test two + test three/3) and analyzed group differences with Mann-Whitney U tests for amplitude, frequency and asymmetry. As we used non-parametric tests, no log transformations were necessary. We performed subgroup analyses on AUDS patients using anti-epileptic treatment (compared to those who did not and the MDE group) with Kruskall-Wallis tests. Possible significant outcomes were tested with post-hoc Mann-Whitney U tests to identify the deviating group. In order to estimate possible drug or abstinence influence on the EEG, we performed an additional analysis on the third test (Possible abstinence or illegal drug effects would minimally influence the test 10 days after admittance). We studied multiple predefined hypotheses; hence we did not apply any correction for multiple significance testing. These corrections are not generally appropriate in such studies [20, 21]. Two sided p- values ≤ 0.05 are reported as significant. A two-sample Student's t-test with 32 subjects has 72% power to detect an effect = 90% of group SD.

Eight (50%) of the AUDS patients and one (6%) of the MDE patients had two or more abnormal features after three registrations (p = 0.016, Table 2). More AUDS than MDE patients had abnormal EEGs: In the first test 38% of AUDS patients and 25% of MDE patients had abnormal EEG. Epileptiform sharp activity, focal slow activity, and generalized slow activity were all found in more AUDS patients than MDE patients, but differences did not reach statistical significance.

We found significantly higher temporal delta band amplitude in AUDS patients- than in MDE patients (Table 3, Figure 1). Delta activity was most prominent in the right hemisphere (Figure 2). Delta asymmetry was larger in AUDS than MDE in all regions (Table 4, Figure 3). Twelve of 16 AUDS patients had temporal delta amplitude above 8.9 μV as compared to 3 of 16 MDE patients (sensitivity 76%, specificity 81% for this cut-off). Three MDE – and 11 AUDS patients had absolute temporal delta amplitude side difference above 0.1 μV (sensitivity 69%, specificity 81% for this cut-off). These cutoffs were selected by trial and error after inspection of the box-and-whisker plots as those that yielded the optimal separation between the two groups.

Although the AUDS patients had slower occipital alpha peak frequency than the MDE patients, neither this difference nor the peak frequency asymmetry was significantly different between groups (Table 3).

Drowsiness and AUDIT scores did not differ between groups (Table 1). Comparing band amplitudes in the third test only (in order to eliminate possible effects of drug or alcohol abstinence) did confirm our main result: increased temporal delta amplitude was found in AUDS patients compared to the MDE patients (10.3 μV vs. 8.5 μV, p = 0.04). More AUDS patients than MDE patients used anti-epileptic medication (Table 1). Some patients used more than one medicament: seven patients were on lamotrigine, one patient used valproate, and three patients used carbamazepine, while four patients and one control used clonazepam. However, subgroup analyses showed that antiepileptic drug use did not influence any of the significant results in Table 3 and 4. Indeed the AUDS subgroup not using anti-epileptic drugs did also have significantly more temporal delta activity than MDE patients (mean (SD) 9.4 (0.8) μV vs. 8.5 (1.6) μV, p = 0.04).