Frontal Alpha Asymmetry Correlates with Suicidal Behavior in Major Depressive Disorder.

OBJECTIVE: Based on the constant associations made between major depressive disorder (MDD) and alpha asymmetry, and MDD and suicide, this study aimed to examine the relationship between frontal alpha asymmetry and suicide in MDD patients.
METHODS: Sixty-six MDD patients, of whom fifteen were male and fifty-one were female, were recruited. Independent groups were created based on the median score of frontal alpha asymmetry: the left dominant (LD) group and the right dominant (RD) group. The alpha band (8‒12 Hz) and its sub-bands (i.e., low alpha band: 8‒10 Hz; high alpha band: 10‒12 Hz) were of interest. Source level alpha asymmetry was calculated as well.
RESULTS: Suicidal behavior was positively correlated with the asymmetry indices of the low alpha band and the alpha band in the LD group and that of the high alpha band in the RD group. Source level analysis revealed positive correlations between suicidal behavior and the asymmetry index of the low alpha band in the LD group.
CONCLUSION: Frontal alpha asymmetry, especially that of the low alpha band, might reflect the cognitive deficits associated with suicidal behaviors in MDD patients.

INTRODUCTION

Among the various psychiatric disorders that have been associated with suicidal behavior, depression has been one of the most consistent and robust predictors of suicide.1,2) Even after controlling the effects of other co-morbid disorders, depression is still a significant predictor of suicidal behavior, especially suicide ideation.3) It is thus crucial to assess suicide risk when treating patients with depression.

In general, abnormal alpha band activity has been suggested as an indicator of depression.4,5) Perhaps, one of the most promising biomarkers of depression might be asymmetrical electroencephalographic (EEG) activity during the resting state. Characterized by asymmetrical alpha band activity (8–12 Hz) between the left and the right frontal regions (i.e., alpha asymmetry),6) alpha asymmetry distinguishes individuals with depression from individuals without depression.7–9) In particular, depressive individuals typically display a greater alpha activity in their left hemisphere than in their right hemisphere. The enhanced alpha power is thought to reflect a reduction in cortical activity in the corresponding region, and hypoactivation in the left frontal hemisphere has been reported in individuals with depression.10–12)

In addition, frontal alpha asymmetry has been examined in relations to specific cortical regions including the dorsolateral prefrontal cortex (DLPFC), the ventromedial prefrontal cortex (vmPFC), the anterior cingulate cortex (ACC), the rostral ACC (rACC), and the subgenual ACC (sgACC), has been extensively examined.13–16) Impairments in the DLPFC and the vmPFC have been associated with deficits in negative emotional control and automated processing of emotionally relevant stimuli, respectively.17,18) Moreover, patients with major depressive disorder (MDD) have shown reduced perfusion or metabolic activity in the DLPFC, particularly in the left hemisphere,19–21) as well as hypoperfusion or hypometabolism in the ACC, which is thought to be a byproduct of reduced cortical volume of the left sgACC.22–24) Furthermore, disturbed functioning as well as reduced volumes of the orbitofrontal cortex (OFC) have been reported to be associated with increased impulsivity and aggression, which are known as risk factors of suicidal behaviors.23,25–28)

Despite the strong evidence supporting the relationships between alpha asymmetry and depression, and depression and suicide, the association between alpha asymmetry and suicidal behaviors has been relatively unexplored. The only study that has directly examined the relationship between alpha asymmetry and suicide was conducted in female adolescents.29) Results indicated that depressive suicidal adolescents display a greater alpha activity in the left than the right hemisphere. Given that suicidal individuals are often characterized by cognitive rigidity and attentional biases that are associated with impaired frontal lobe functioning,30–33) it is worthwhile to examine alpha asymmetry and suicidal behavior in adult patients with MDD.

Meanwhile, studies on EEG alpha power suggest the need to distinguish the alpha band into two sub-bands: the low alpha (8–10 Hz) and the high alpha (10–12 Hz).34,35) Low alpha is thought to represent a diffused attentional and brain-state phenomenon, whereas high alpha is thought to represent more localized and task-specific cognition, such as semantic memory.36) In a group of ADHD patients, alpha asymmetry of the low alpha band indicated the existence of clinically significant symptoms.37) The role of the low and high alpha bands in depression, however, remains equivocal and, to our knowledge, their relationship to suicidal behavior has not been previously examined.

The purpose of the present study was to explore the relationship between frontal alpha asymmetry and suicidal behavior. In addition, we aimed to examine how the cortical level alpha asymmetries would correlate with suicidal risk. We hypothesized that suicidal behaviors would correlate with a greater relative alpha power in the left hemisphere as compared to the right hemisphere. More specifically, we hypothesized that the low alpha band, which is thought to be associated with attention, would be related to suicidality, because suicidal individuals often display cognitive deficits.38,39) For example, attentional bias has been discovered in suicidal individuals in that they often fixate on suicidal cues.40,41)

METHODS

Participants

Seventy-eight participants with MDD were included in the study. Among those, 12 were excluded due to an insufficient number of epochs in their EEG data. A total of 66 subjects, of whom 15 were male (22.7%) and 51 were female (77.3%), were included in the initial analyses. The mean age was 44.77 years (standard deviation [SD], 13.62 years; range, 19–65 years) and the mean years of education was 12.77 years (SD, 6.41 years). When asked about prior suicide attempt, 24 participants (36.4%) reported one or more lifetime attempts, whereas 42 (63.6%) reported no prior history. All participants were right-handed and those with brain damage or histories of psychotic symptoms or bipolar disorder were excluded from the analyses. Although our participants were all diagnosed with MDD, some exhibited greater relative alpha power in their right hemisphere than in their left. This pattern contradicted the previous findings identified in MDD patients (i.e., greater relative alpha power in the left hemisphere).

Groups were created based on the median scores of the asymmetry indices of the alpha bands at each region of interest (ROI): FP1–FP2 (low alpha, 1.80; high alpha, 1.13; alpha, 1.24), F3–F4 (low alpha, 2.79; high alpha, 2.49; alpha, 2.66), and F7–F8 (low alpha, 5.22; high alpha, 4.14; alpha, 4.52). All participants were matched to either the left-dominant (LD) group or the right-dominant (RD) group for all three band frequencies at each site. Only those who fell into the same group for all three band frequencies at each site were included in the final analyses (n = 56). Participants signed an informed consent approved by the Institutional Review Board at Inje University Ilsan Paik Hospital (2015-03-212) prior to their participation.

As a part of their treatment, most of the participants (n = 52) received antidepressants. The most common type of antidepressant in use was escitalopram followed by venlafaxine, paxil, tianeptine, and sertraline. In addition, most (n = 45) were treated with benzodiazepines such as lorazepam, clonazepam, alprazolam, and etizolam.

Psychological Measures

Cognitive Flexibility Inventory (CFI)

Cognitive flexibility was measured with the Korean version of the CFI.42) The scale consists of two subscales, the control subscale and the alternative subscale. The control subscale measures the extent to which individuals believe they are capable of controlling problematic situations; the alternative subscale measures how much individuals believe in their ability to find alternative solutions in unfavorable situations. A total of nineteen questions are measured on a 7-point Likert scale, in which higher total scores indicate greater cognitive flexibility. In the present study, the internal consistencies (Cronbach’s alpha) for the alternative subscale, the control subscale, and the entire inventory were 0.87, 0.90, and 0.91, respectively.

Rumination Reflection Questionnaire (RRQ)

Focused attention was assessed with the Korean version of the RRQ.43) Rumination is defined as fixation on negative thoughts, whereas reflection is defined as introspection that promotes self-growth. The RRQ can measure both traits (RRQT) and states (RRQS). The Rumination and Reflection Questionnaires, each consists of 12 items that are measured on a 5-point Likert scale. In the present study, both trait and state rumination and reflection were assessed. Rumination (α = 0.82), reflection (α = 0.81) and the entire questionnaire (α = 0.86) all exhibited high internal consistencies.

Beck Depression Inventory (BDI) II

Depression was assessed with the Korean version of the Beck Depression Inventory (i.e., BDI-II). The BDI-II is consisted of 21 items that are rated on a 4-point scale. The Korean version of the BDI-II has shown a high internal consistency (α = 0.85) and test-retest reliability (r = 0.75).44) Cronbach’s alpha for the current sample was 0.95.

Beck Anxiety Inventory (BAI)

Anxiety was measured with the Korean version of the BAI, a self-reported questionnaire composed of 21 items. Its internal consistency and test-retest reliability were 0.93, and 0.84, respectively.45) Cronbach’s alpha for the current sample was 0.94.

Suicidal Behavior Questionnaire-Revised (SBQ-R)

Suicidal behavior was measured through the Korean translation of the SBQ-R.46) The SBQ-R is composed of 4 items that measure different dimensions of suicidality; 1) lifetime suicide ideation and suicide attempt, 2) the frequency of suicidal ideation over the past 12 months, 3) the threat of suicidal behavior, and 4) the likelihood of suicidal behavior by self-report.47) The SBQ-R, which has exhibited excellent reliability,48) was positively correlated with the Korean version of the Beck Scale for Suicide Ideation (r = 0.62).49) The total scores of the SBQ-R range from 3 to 18. The Cronbach’s alpha for the Korean version of the SBQ-R in the original study was 0.93, and 0.85 for the current sample.

EEG Recordings and Quantitative EEG Analysis

Resting state EEG data was collected in a sound attenuated room for three minutes each during two conditions: the eyes-closed condition and the eyes-opened condition. The EEG signal was obtained using a 62 channel (FP1, FPZ, FP2, AF3, AF4, F7, F5, F3, F1, FZ, F2, F4, F6, F8, FT7, FC5, FC3, FC1, FCZ, FC2, FC4, FC6, FT8, T7, C5, C3, C1, CZ, C2, C4, C6, T8, TP7, CP5, CP3, CP1, CPZ, CP2, CP4, CP6, TP8, P7, P5, P3, P1, PZ, P2, P4, P6, P8, PO7, PO5, PO3, POZ, PO4, PO6, PO8, CB1, O1, OZ, O2, and CB2) Quik Cap and a NeuroScan SynAmps amplifier (Compumedics USA, El Paso, TX, USA). An international 10–20 placement scheme was applied. The ground electrode was located on the forehead and Cz was used as the reference electrode. We attached the vertical electro-oculogram (EOG) electrodes above and below the left eye, and the horizontal EOG on the outer canthus of both eyes. The EEG data was recorded once the impedance fell below 5 kΩ. A 0.1 to 100 Hz band pass filter at a sampling rate of 1,000 Hz was applied to process the all of the data.

SCAN 4.3 software (Advanced Medical Equipment Ltd., Horsham, UK) was used to preprocess the EEG data. An experienced technician with no prior knowledge of the origin of the data screened artifacts such as eye movements. The current study used eyes-opened data for its analyses. A total of 30 epochs of which the length of each epoch was configured as 2 seconds (2,048 points), were prepared for each participant. Epochs with signals exceeding ± 100 μV on any channel were rejected from further analysis. A fast Fourier transformation (FFT) was performed on each epoch for 62 electrode channels to analyze the spectral powers. The FFT window size was set to match the length of each epoch (2,048 points) with the hamming window, and no overlap was allowed. The frequency bin size was ~0.488 Hz (1,000/2,048) with a total of 1,025 bins. After FFT was conducted on the 30 epochs, the resultant spectral power was averaged. Band powers of three frequency bands (i.e., low alpha: 8–10 Hz; high alpha: 10–12 Hz; alpha: 8–12 Hz) were then calculated.50) The closest frequencies corresponding to each band were selected to determine the bin size.

Frontal alpha asymmetry was measured on three paired sites: pre-frontal (FP1–FP2), mid-frontal (F3–F4), and lateral-frontal (F7–F8).51–53) To measure alpha asymmetry, we compared the corresponding frequency band percentages of the left and right hemispheres. More specifically, the difference between the two hemispheres was divided by their sum (i.e., alpha asymmetry = [Pleft–Pright]/[Pleft+ Pright] × 100; Pleft and Pright are absolute powers). An asymmetry value of zero indicates that alpha power and brain activity are equal in both hemispheres; a positive asymmetry value indicates greater alpha power and hypoactivation of the left hemisphere; a negative asymmetry value indicates greater alpha power and hypoactivation of the right hemisphere. The asymmetry indices were calculated for the three alpha bands of interest.

Source Activity Analysis

The cortical distribution of the standardized source current density of brain sources in the resting state for the frequency domain was calculated through standardized low-resolution brain electromagnetic tomography (sLORETA). sLORETA is a source imaging method that is advantageous of solving the EEG inverse problem.54) Based on the assumption that the source activation of a voxel is similar to that of the surrounding voxels, it applies an appropriate standardization of the current density. The lead field matrix was computed using a realistic head model that was segmented based on the Montreal Neurological Institute (MNI) 152 standard template, in which the three-dimensional solution space was restricted to only the cortical gray matter and hippocampus.55) The solution space was comprised of 6,239 voxels with 5-mm resolution. Anatomical labels, such as the Brodmann areas, were provided with an appropriate transformation from MNI to Talairach space.56)

The ROIs related to frontal alpha asymmetry were selected according to the results of previous neuroimaging and EEG source localization studies. They include the DLPFC, vmPFC, ACC, rACC, and sgACC.13–16) In addition, we examined the OFC, which has been associated with cognitive deficits (e.g., rigidity, impulsivity, decision making) in suicidal individuals.57) The source activities of the ROIs in low, high, and total alpha frequency bands were then extracted. For further analyses, we calculated the alpha asymmetry indices for each ROI using source activities from both hemispheres.

Statistical Analysis

Spearman’s correlation analysis was conducted to examine the associations between the asymmetry indices and the psychological measures. The specific method was selected because our data did not demonstrate a normal distribution.58) In addition, a 5,000 bootstrap resampling technique was applied to address issues related to multiple comparisons.59) Further correlation analysis was executed between the asymmetry indices and the psychological measures based on the groups created by the median split. Within those groups, the correlation between the SBQ-R and alpha asymmetry indices (i.e., low alpha asymmetry, high alpha asymmetry, and alpha asymmetry) was examined in each ROI. Considering that the majority of our participants were females, sex was controlled for in all correlation analysis.

Afterwards, independent t tests were performed to examine if scores on cognitive flexibility and trait and state rumination differed depending on history of prior suicide attempt and scores on the SBQ-R.

RESULTS

Descriptive Statistics

Descriptive statistics of the psychological measures and the asymmetry indices of alpha, low alpha, and high alpha in the frontal region are presented in Table 1. The mean and standard deviation for CFI, RRQS, RRQT, BDI-II, BAI, and SBQ-R were 81.36 ± 19.04, 70.73 ± 12.34, 75.94 ± 12.80, 22.21 ± 15.30, 41.20 ± 13.92, and 8.97 ± 4.75, respectively. The mean and the standard deviation of the asymmetry indices at each frontal site were as follows; FP1–FP2 (alpha, 0.63 ± 6.57; low alpha, 0.91 ± 7.68; high alpha, 0.48 ± 6.32), F3–F4 (alpha, 1.91 ± 4.66; low alpha, 2.47 ± 5.98; high alpha, 1.54 ± 4.83), F7–F8 (alpha, 4.33 ± 7.40; low alpha, 4.96 ± 9.29; high alpha, 3.88 ± 6.98). The asymmetry indices were indicative of a higher relative alpha activity in the left hemisphere, which aligns with previous findings in patients with depression.

Table 1

Demographic information of the participants including means and standard deviations

Variable	Participant (n = 66)	LD group (n = 28)	RD group (n = 28)	p value
Sex				0.001
Male	15 (22.7)	11 (39.3)	1 (30.6)
Female	51 (77.3)	17 (60.7)	27 (960.4)
Age (yr)	44.77 ± 13.62	46.39 ± 11.54	40.86 ± 15.65	0.139
Education (yr)	12.77 ± 6.41	13.93 ± 9.05	11.78 ± 3.47	0.255
Past suicide attempt				0.406
Yes	24 (36.4)	8 (28.6)	11 (39.3)
No	42 (63.6)	20 (71.4)	17 (60.7)
CFI	81.36 ± 19.04	82.68 ± 16.27	80.25 ± 20.31	0.623
RRQS	70.73 ± 12.34	71.11 ± 11.10	76.86 ± 15.40	0.842
RRQT	75.94 ± 12.80	75.07 ± 11.45	70.39 ± 15.19	0.624
BDI-II	22.21 ± 15.30	19.64 ± 13.40	23.39 ± 16.43	0.354
HAM-D	14.65 ± 6.80	13.43 ± 7.13	15.57 ± 5.92	0.228
BAI	41.20 ± 13.92	39.29 ± 12.77	41.96 ± 14.70	0.469
SBQ-R	8.97 ± 4.75	8.21 ± 4.43	9.54 ± 4.83	0.290
Alpha asymmetry index of FP1–FP2
Low alpha band	0.91 ± 7.68	3.17 ± 5.01	−0.81 ± 8.55	0.038
High alpha band	0.48 ± 6.32	2.13 ± 3.30	−0.33 ± 7.92	0.134
Alpha band	0.63 ± 6.57	2.52 ± 3.57	−0.56 ± 7.89	0.068
Alpha asymmetry index of F3–F4
Low alpha band	2.47 ± 5.98	5.59 ± 4.58	−0.33 ± 5.63	< 0.001
High alpha band	1.54 ± 4.83	3.52 ± 4.22	0.02 ± 5.04	0.007
Alpha band	1.91 ± 4.66	4.32 ± 3.51	−0.02 ± 4.74	< 0.001
Alpha asymmetry index of F7–F8
Low alpha band	4.96 ± 9.29	12.27 ± 6.95	−2.68 ± 6.17	< 0.001
High alpha band	3.88 ± 6.98	9.56 ± 3.76	−2.01 ± 5.37	< 0.001
Alpha band	4.33 ± 7.40	10.64 ± 4.24	−2.19 ± 5.25	< 0.001

Values are presented as number (%) or mean ± standard deviation.

LD group, left-dominant group; RD group, right-dominant group; CFI, Cognitive Flexibility Inventory; RRQS, Rumination Reflection Questionnaire-State; RRQT, Rumination Reflection Questionnaire-Trait; BDI-II, Beck Depression Inventory II; HAM-D, Hamilton Rating Scale for Depression; BAI, Beck Anxiety Inventory; SBQ-R, Suicidal Behavior Questionnaire-Revised.

Alpha asymmetry was measured at three frontal regions (i.e., FP1–FP2, F3–F4, F7–F8) for the low alpha band, the high alpha band, and the alpha band.

Sensory Level Correlations

Correlations between asymmetry indices and suicidal behavior were examined based on the groups that were created with the median scores of the asymmetry indices. No correlations were found at FP1–FP2 and F3–F4. Hence, these channels were disregarded from further analyses. Significant correlations were observed at F7–F8. In those who had greater alpha activity in their left frontal region, the asymmetry index of the low alpha band (r = 0.494, p = 0.009) and that of the entire alpha band (r = 0.385, p = 0.047) was positively correlated with suicidal behavior. In those who had greater alpha activity in their right frontal region, the asymmetry index of the high alpha band was positively correlated with suicidal behavior (r = 0.460, p = 0.016; Fig. 1).

Fig. 1

Alpha asymmetry indices associated with suicidal behavior at F7–F8. In the left-dominant group (A), the asymmetry indices of the low alpha band and the alpha band were positively correlated with suicidal behavior. In the right-dominant group (B), the asymmetry index of the high alpha band was positively correlated with suicidal behavior.

SBQ-R, Suicidal Behavior Questionnaire-Revised.

Source Level Correlations

ROIs of source level analysis is depicted in Figure 2; correlations between suicidal behavior and the alpha asymmetry indices of each ROI in the LD and RD groups in F7–F8 are presented in Table 2. No significant correlations were found in the RD group. On the other hand, suicidal behavior was positively correlated with the asymmetry index of the low alpha band (r = 0.445, p = 0.020), the high alpha band (r = 0.468, p = 0.014), and the alpha band (r = 0.463, p = 0.015) at the rACC, as well as the asymmetry index of the low alpha band (r = 0.508, p = 0.007), the high alpha band (r = 0.436, p = 0.023) and the alpha band (r = 0.455, p = 0.017) at the DLPFC in the LD group. Furthermore, suicidal behavior was also positively correlated with the asymmetry index of the low alpha band at the vmPFC (r = 0.405, p = 0.036) and the asymmetry index of the low alpha band (r = 0.386, p = 0.047) at the OFC. There were no significant correlations between suicidal behavior and the alpha asymmetry indices at the ACC and the sgACC.

Fig. 2

Regions of interest for examination of the relationship between asymmetry of the alpha band source activity and suicidal behavior.

Table 2

Correlations coefficients between suicidal behavior and alpha asymmetry at cortical regions of interest (n = 56)

Variable	LD group (n = 28)		RD group (n = 28)
	SBQ-R	p value	SBQ-R	p value
Past suicide attempt	−0.591**	0.001	−0.478*	0.010
Alpha asymmetry index
Low alpha	0.494**	0.009	−0.026	0.897
High alpha	0.344	0.079	0.460*	0.016
Alpha	0.385*	0.047	0.241	0.226
Anterior cingulate cortex (ACC)
Low alpha asymmetry	0.362	0.064	−0.051	0.802
High alpha asymmetry	0.317	0.107	−0.116	0.565
Alpha asymmetry	0.352	0.072	−0.052	0.795
Rostral ACC
Low alpha asymmetry	0.445*	0.020	−0.009	0.963
High alpha asymmetry	0.468*	0.014	−0.139	0.490
Alpha asymmetry	0.463*	0.015	−0.043	0.830
Subgenual ACC
Low alpha asymmetry	0.290	0.143	−0.080	0.691
High alpha asymmetry	0.172	0.392	0.181	0.367
Alpha asymmetry	0.194	0.331	0.049	0.807
Dorsolateral prefrontal cortex
Low alpha asymmetry	0.508**	0.007	0.017	0.935
High alpha asymmetry	0.436*	0.023	0.072	0.722
Alpha asymmetry	0.455*	0.017	0.047	0.818
Ventromedial prefrontal cortex
Low alpha asymmetry	0.405*	0.036	0.104	0.605
High alpha asymmetry	0.255	0.200	0.144	0.475
Alpha asymmetry	0.343	0.080	0.110	0.583
Orbitofrontal cortex
Low alpha asymmetry	0.386*	0.047	0.139	0.488
High alpha asymmetry	0.146	0.467	0.180	0.370
Alpha asymmetry	0.239	0.239	0.142	0.481

Groups were divided based on the median score (4.37) of the asymmetry index of the alpha band at F7–F8.

LD group, left-dominant group; RD group, right-dominant group; SBQ-R, Suicidal Behavior Questionnaire-Revised.

p < 0.05,

p < 0.01.

T Tests

Figure 3 shows comparisons between cognitive flexibility and ruminative tendencies and prior suicide attempts and scores on the SBQ-R. Those who had made prior suicide attempts had significantly higher scores on trait rumination (t = −3.010, p = 0.004), state rumination (t = −2.496, p = 0.015), and the overall RRQT (t = −2.390, p = 0.020), and significantly lower scores on the control subscale of the CFI (t = 2.866, p = 0.006) and the entire CFI (t = 2.452, p = 0.017) compared to those who had never attempted suicide. The difference was more pronounced when the groups in comparison were divided by their scores on the SBQ-R; those who scored higher on the SBQ-R had significantly higher scores on trait rumination (t = −3.418, p < 0.001), state rumination (t = −3.673, p < 0.001), overall RRQT (t = −2.543, p = 0.013), and RRQS (t = −2.264, p = 0.027), and significantly lower scores on the control subscale (t = 3.673, p < 0.001), the alternative subscale (t = 2.193, p = 0.032) and the entire CFI (t = 3.502, p < 0.001) than those who scored lower on the SBQ-R.

Fig. 3

Rumination and cognitive flexibility scores by group. (A) Comparisons are made between individuals with and without a history of prior suicide attempt. (B) Comparisons are made between individuals who scored high and individuals who scored low on the Suicidal Behavior Questionnaire-Revised (i.e., SBQ-R).

RRQT, Rumination Reflection Questionnaire-Trait; RRQS, Rumination Reflection Questionnaire-State; CFI, Cognitive Flexibility Scale; Control, control subscale of CFI; Alternative, alternative subscale of CFI.

*p<0.05, **p<0.01, ***p<0.001.

DISCUSSION

While attempts to identify biological correlates of suicidal behaviors have increased,60,61) the primary objective of our study was to examine the association between alpha asymmetry and suicidal behavior. The main findings are as follows: 1) sensory level analysis revealed a significant positive correlation between the alpha asymmetry indices of the low alpha band, the alpha band and suicidal behavior in the LD group, and a significant positive correlation between the alpha asymmetry index of the high alpha band and suicidal behavior in the RD group; 2) source level analysis revealed significant positive correlations between indices of alpha asymmetry and suicidal behavior in the LD group but not in the RD group; and 3) suicidal individuals scored higher on measures of rumination and inflexible cognitive style than non-suicidal individuals.

In the LD group, the alpha asymmetry indices of the low alpha band and the alpha band were positively correlated with suicidal behavior. In contrast, the alpha asymmetry index of the high alpha band was positively correlated with suicidal behavior in the RD group. In our study, the LD group showed greater relative alpha power in the left frontal hemisphere than in the right, which is thought to be normative in depression.9,62) According to Graae et al.,29) adolescents with depression who had previously attempted suicide showed a greater alpha in their left frontal hemisphere than in their right. Our study has replicated the findings in adults with MDD.

Previous research has suggested that the desynchronization of the lower frequency alpha band is associated with attention.34,63) That is, suppression of the lower frequency alpha band occurs as attentional demands increase. Suicidal individuals often experience difficulty in controlling attention (i.e., deciding which stimuli to focus on)32) or have attentional biases toward suicide-related stimuli.30) In this case, alterations of the lower frequency alpha band may provide profound insight into suicidal behaviors of patients with depression. Little to no research has been done to examine the specific functions of the high alpha band and its relations to suicidal behavior in patients with depression. To our knowledge, the only prior studies of the high alpha band reported that it may be related to semantic or task specific effects.34,35) Therefore, future investigations of the high alpha band are warranted.

Our results revealed significant associations between the alpha asymmetry indices of the DLPFC, vmPFC, rACC, and OFC and suicidal behaviors in the LD group, but not in the RD group. The most consistent relationships were observed between the alpha asymmetry indices of the low alpha band and suicidal behavior. These findings indicate that suicide risk increases in coherence with the relative power of the lower frequency alpha band in the left hemispheres of each region. In particular, reduced activity in the left DLPFC has been shown to be associated with a tendency to negatively interpret emotional cues in MDD patients. According to Disner et al.,64) negative emotional interpretation could be a byproduct of ruminative tendencies that are reflected by hypoactivation of the left DLPFC. The study also claimed that abnormal cortical activity of the rACC might lead to ruminative tendencies in individuals with depression by inhibiting positive stimuli and simultaneously impeding the inhibition of negative stimuli. Moreover, significant reductions in gray matter in the left DLPFC and the rACC have been identified in suicidal individuals,65,66) as has cortical thickness in the left DLPFC and the ACC.67) These results suggest that such regions may play a significant role in suicidal behavior. In another study, Jollant et al.68) found reduced activity in the left OFC when individuals, who had previously attempted suicide, performed risky tasks. They suggested that reduced activity of the left OFC may be related with impaired decision making. In addition, deficits in the OFC have been associated with greater impulsivity,25,26) which is thought to indirectly increase suicide risk.69)

The reason that a relationship between alpha asymmetry of the cortical ROIs and suicidal behavior was only seen in the LD group remains ambiguous. A possible explanation is that the RD group might represent individuals with atypical depression, which has discrete biological foundations.70) It would thus be worthwhile for future studies to examine differences in alpha asymmetry among various subtypes of depression.

Lastly, we found that suicidal patients with depression were more likely to have ruminative tendencies and cognitive inflexibility than non-suicidal patients with depression. That is, suicidal individuals tend to fixate in their suicidal thoughts and are often unable to think of alternative solutions during mental sufferings. Such results are consistent with previous findings of cognitive rigidity in suicidal individuals.71,72)

Despite the significant findings, there are some limitations to our study. The main limitation is that we did not control for comorbidities, which may affect alpha asymmetry. For example, it is thought that the pattern of alpha asymmetry in patients with anxiety is the opposite of that of patients with MDD, in that the relative alpha power is greater in the right hemisphere than in the left.73) Another limitation is that although we believe that the alpha asymmetry indices might reflect the cognitive rigidity of suicidal individuals, the direct relationship was not examined in our study. Future studies ought to focus on identifying the association between source level alpha asymmetry indices and cognitive abilities in MDD patients with suicidal tendencies using more sophisticated designs.

Nevertheless, our study is one of only a few that has directly examined the relationship between frontal alpha asymmetry and suicidal behavior. Moreover, to our knowledge, our study is the first to explore the relationship between source level alpha asymmetry in various cortical areas and suicidal behavior. Our results indicate that individuals with depression and suicidal behaviors exhibit an increased relative alpha power in the left hemisphere of the DLPFC, vmPFC, rACC, and OFC, which may be reflective of ruminative tendencies and cognitive rigidity. Taken together, our results suggest that alpha asymmetry, especially that of the low alpha band, may be a plausible biomarker of suicide risk.