2023 Impact Factor
2023 Impact Factor
Nonsuicidal self-injury (NSSI) refers to intentional and self-inflicted harm to body tissues without the intention to commit suicide and is not socially sanctioned [1]. It encompasses various behaviors including cutting, burning, biting, and scratching the skin [1]. While the prevalence of NSSI in nonclinical samples is estimated to be approximately 17.2% among adolescents and 5.5% among adults [2], the prevalence in clinical samples consisting of individuals receiving clinical care or treatment is estimated to be between 13% and 37% among adults [3-5]. The prevalence of NSSI has been rapidly increasing [6], raising significant clinical concerns. NSSI is particularly common among adolescents [7] and poses a significant social issue, with methods being shared through social networking sites and other means. Engaging in NSSI can lead to a range of negative outcomes such as bodily harm, strained relationships, and reduced academic or occupational performance. While NSSI is defined as intentional self-injury without suicidal intent, patients engaged in NSSI may show an increased risk of suicidal behavior [8].
Functional connectivity (FC) is defined as the statistical dependency or association between two signals recorded simultaneously [9]. FC, which reflects the pattern of correlations observed between various brain regions, has been found to be associated with a range of psychiatric, behavioral, and cognitive phenotypes [10-12]. FC is generally evaluated using functional magnetic resonance imaging (fMRI), and fMRI has been used to analyze FC in patients with NSSI. One such study investigated regional brain activity and remote FC in the right fusiform gyrus as well as the right median cingulate and paracingulate gyri in adolescents with major depressive disorder (MDD) alone and adolescents with MDD who engaged in NSSI. The findings indicated significantly reduced FC in patients with NSSI [13]. Another study that compared the FC of adolescents with NSSI and healthy controls using fMRI found a decrease in amygdala-frontal connectivity. Conversely, amygdala FC appears to be increased in certain regions, including the supplementary motor area and bilateral dorsal anterior cingulate [14].
However, fMRI is difficult to use in clinical settings and is primarily used for research purposes. FC can also be measured using electroencephalography (EEG) [15], which is more practical for clinical applications owing to its cost-effectiveness and noninvasiveness. Therefore, EEG may be a good alternative to fMRI for FC analysis. The use of EEG to investigate changes in FC in patients with depression has received much research attention recently. In most EEG studies focusing on FC in patients with depression, it has been observed that alpha FC appears to be higher in individuals with depression [16]. However, to the best of our knowledge, no studies have used EEG to investigate the differences in FC in individuals with NSSI.
We aim to investigate whether differences exist in the FC of the frontal lobe or limbic system in patients with depression, specifically comparing those with and without NSSI, based on the assumption that there may be differences in areas such as impulse control and addiction. Our goal was to elucidate the biological mechanisms underlying NSSI, with the expectation that the findings would provide insights for the prevention and treatment of NSSI.
We conducted a retrospective study of inpatient patients with mood disorders (major depressive disorder, bipolar I disorder, and bipolar II disorder) who experienced major depressive episodes and underwent EEG between 2015 and 2023. Data regarding suicide attempts and NSSI were collected, and NSSI was defined according to the Diagnostic and Statistical Manual of Mental Disorders 5th edition (DSM-5) diagnostic criteria as ‘individuals who engaged in intentional self-inflicted harm with the expectation that the injury will lead to only minor or moderate physical harm on 5 or more days in the last year.’ Consequently, we investigated whether there was a history of NSSI and suicide attempts in the year prior to admission. Patients with neurological disorders, such as epilepsy, or a medical history of head trauma, brain surgery, cerebrovascular accidents, or hemorrhage were excluded from the study. Since NSSI is common during adolescence and early adulthood, the age range of participants was limited to 15−30 years. Ultimately, we analyzed the medical records of 77 patients who were grouped as follows on the basis of the presence of a history of NSSI: Group A (NSSI group; n = 31) and Group B (non-NSSI group; n = 46). We compared EEG findings and scores on the Beck Depression Inventory (BDI), Hamilton Rating Scale for Depression (HAM-D), and Suicidal Ideation Questionnaire (SIQ) between the two groups. All scales and EEG tests were conducted within one week of admission. This study was approved by the Institutional Review Board of the Inje University Ilsan Paik Hospital (2023-07-012).
The EEG data recorded before March 2021 were collected using a Neuroscan 64-channel QuickCap (sampling rate: 1,000 Hz; bandpass: 1−55 Hz; references: both mastoids). A total of 64 Ag-AgCl electrodes were positioned in QuickCap using an extended 10-20 international placement system. The data recorded after March 30, 2021, were collected using the EGI 64-channel cap and the 10-10 system (sampling rate: 1,000 Hz, bandpass: 1−55 Hz, reference: Cz). The participants were instructed to close their eyes and undergo the examination during a 5-minute resting state.
Independent component analysis was applied to both EEG systems to remove artifacts. Only 48 channels based on electrode names were used for FC analysis (Fig. 1). All recorded segments were bandpass-filtered into the following frequency bands: delta (1−4 Hz), theta (4−8 Hz), alpha (8−12 Hz), low-beta (12−20 Hz), high-beta (20−30 Hz), and gamma (30−55 Hz).
Functional connectivity analyses generally evaluate functional communication between different regions of the brain by measuring the degree of synchronization among EEG signals. The following three well-known types of FC measures were applied: coherence (Coh) and the imaginary part of coherence (iCoh) as frequency-domain measures, and the phase-locking value (PLV) as a phase-domain measure.
Coh is used to quantify the linear relationship between two signals [17] and is quantified on a scale of zero to one, where a value of zero indicates complete absence of correlation between the signals. High coherence indicates a strong level of connectivity, whereas low coherence indicates a weaker level of connectivity. The concept of coherence involves a complex value consisting of both real and imaginary components. By focusing on the imaginary axis (y-axis), the iCoh can be extracted, as discussed by Nolte et al. [17]. This measure can correct for the effects of volume conduction and signal leakage.
The PLV captures the level of phase synchronization between two signals by evaluating their instantaneous phase differences [18]. Connectivity is assessed on the basis of the stability of the phase differences observed between the two signals. A constant phase difference between two signals indicates a high likelihood of coupling, resulting in a PLV of 1. Conversely, if the two signals display a random phase difference, this suggests independence between them, leading to a PLV value of 0. For the other scenarios, the PLV values fell within the range of 0 to 1, reflecting different levels of connectivity.
Graph theory analysis (GTA) is conducted through a mathematical approach, which involves transforming the connectivity matrix into a binarized matrix [19]. Within the framework of GTA, brain networks can be conceptualized as graphs, where individual anatomical and/or functional brain regions are depicted as nodes, and any form of interaction between pairs of brain regions is represented by edges. In this study, the graphical analysis measures were derived from the concepts of node strength, clustering coefficient, and characteristic path length. These network indices were calculated on the basis of each FC measure.
Nodal strength refers to the cumulative sum of the edge weights connecting a node to other nodes within a weighted network [20]. It serves as a measure for identifying the engagement and information transmission of a specific region within a functional brain network.
The clustering coefficient measures the ratio of the existing edges between neighboring nodes to all possible connected edges [20,21]. It quantifies the level of functional segregation within the brain regions and provides information on local efficiency. A higher clustering coefficient signifies stronger and more efficient local interactions, indicating a network with increased segregation.
Characteristic path length refers to the average shortest path length between all possible pairs of nodes in a network [20,21]. This metric is used to assess the degree of functional integration among the different brain regions. A lower path length indicates enhanced functional integration, suggesting smoother information flow between brain regions.
To investigate the differences in the demographic and clinical characteristics of the two groups, the independent t test and chi-square test were used to analyze continuous and categorical variables. We used binary logistic regression models to test the associations of NSSI with age; sex; diagnosis; BDI, HAM-D, and SIQ scores; and history of suicide attempts. IBM SPSS for Windows version 29 (IBM Corp.) was used for all the analyses.
An independent t test with false discovery rate correction was used to investigate the differences in FC between groups A and B, while an independent t test was used to compare the network indices. The analysis was conducted using MATLAB software (MathWorks Inc.). A significance level of p < 0.05 was set for all tests.
The mean ages of the patients in groups A and B were 20.0 and 20.8 years, respectively (p = 0.386; Table 1). Of the 77 patients, 26 were male and 51 were female. The male-to-female sex ratio in groups A and B was 5:26 and 21:25, respectively (p = 0.007; Table 1). The ratio of individuals with MDD to BD was 18:13 in Group A and 28:18 in Group B (p = 0.806; Table 1). The BDI score differed significantly between the two groups (p = 0.019), whereas the HAM-D score did not show a significant difference (p = 0.465). In comparison with Group B, Group A showed significantly higher SIQ scores (p < 0.001; Table 1). Of the 31 individuals in Group A, 14 had previously attempted suicide, whereas in Group B, 11 of 46 individuals had a history of suicide. However, this difference was not statistically significant (p = 0.051; Table 1). Binary logistic regression analysis revealed that the risk of NSSI is higher in females, and that a higher SIQ score is associated with a greater risk of NSSI (odds ratio = 5.099 and 1.019, p = 0.020 and 0.033, respectively; Table 2).
We examined the connectivity of 1,326 pairs of signals across 6 frequency bands, yielding 7,956 signal pairs. After calculating connectivity measures for each participant in each group, group comparisons were performed using these values.
The Coh was calculated for each pair, and it showed no significant differences between the two groups (p > 0.05). Similarly, the iCoh showed no significant intergroup differences (p > 0.05). For the PLV, an independent t test with FDR correction showed no statistically significant differences between the two groups (p > 0.05). However, an uncorrected t test showed significant intergroup differences in all the frequency bands (p < 0.05; Fig. 2).
Network indices were calculated using the previously computed connectivity measures as inputs. The network indices showed no significant differences between the two groups when the connectivity values were estimated using iCoh (Table 3). Similarly, no significant intergroup differences were observed when connectivity values were estimated using Coh (Table 3). Additionally, when PLV was used as the connectivity value, no statistically significant differences were observed in the network indices between the two groups (Table 3).
In the present study, we investigated 77 patients with depression, of whom 31 had a history of NSSI. To explore connectivity using a clinically accessible approach, we employed FC measures such as Coh, iCoh, and PLV, along with GTA based on EEG. The results indicated no difference when Coh, iCoh, and PLV were used with correction. However, a lower uncorrected PLV was observed across all frequency bands in patients with NSSI.
Decreased FC implies that the brain regions are less frequently coactivated and are, therefore, less likely to effectively communicate with each other [22]. In this study, however, there were no significant differences in FC between the two groups. In our study, we classified individuals who had engaged in self-injury at least five times in the past year as NSSI patients, which means that not all NSSI group patients had engaged in self-injury immediately before admission. Since EEG is a high-temporal resolution device, if we did not measure brainwaves immediately after self-injury, it is possible that there might not be significant differences in functional connectivity between the two groups. Therefore, conducting a study comparing EEG data taken immediately after self-injury and shortly thereafter in NSSI patients in the future could be beneficial.
On the other hand, NSSI can be viewed as a form of addiction and is theoretically associated with the reward system. Building on the theory that addiction is closely connected to the reward system, one study investigated the functional connectivity of NSSI-related reward circuits in adolescents with depression and NSSI. The results revealed abnormalities in the functional connectivity of patients with depression and NSSI within the reward circuitry, with both increased and decreased functional connectivity observed in various regions [23]. In this way, there may be biological findings that differentiate NSSI patients from those who do not engage in self-injury, even if the self-injury is not happening at the moment. It will be beneficial to conduct more controlled studies than the current research.
Our findings showed no statistically significant difference in the HAM-D scores between the two groups, but a statistically significant difference was observed in the BDI scores. Since the HAM-D involves objective questioning and the BDI reflects patients’ subjective emotional states, the absence of differences in HAM-D scores and the presence of differences in BDI scores could indicate the potential influence of patients’ personalities. Considering the high correlation between personality and NSSI, such results could be feasible [24]. To determine whether the observed FC in our study was influenced by the severity of depression, we conducted binary logistic regression analyses to examine the potential influence of the BDI and HAM-D scores on the presence of NSSI. The results indicated that the BDI and HAM-D scores did not have a significant impact.
Notably, SIQ scores differed significantly between the two groups, whereas the two groups showed no significant difference in suicide attempt history. These findings contrast somewhat with research results indicating a higher incidence of suicide attempts among patients with NSSI [25]. A previous study investigated FC changes in the default mode network (DMN) using fMRI, comparing suicidal and nonsuicidal patients with depression and healthy controls. In that study, patients with depression exhibited greater FC within the DMN in comparison with healthy controls. Furthermore, in comparison with nonsuicidal patients, patients with a history of suicide attempts showed increased FC in the left cerebellum and left lingual gyrus and decreased FC in the right precuneus [26]. Considering the findings of these previous studies, the absence of significant differences between the two groups in other FC measures and network indices, excluding uncorrected PLV, in our study could be interpreted as indicating that the absence of differences in the history of suicide attempts and depression severity between the groups resulted in the absence of differences in FC.
This study had some limitations. First, we did not control for demographic characteristics such as age, sex, diagnosis, severity of depression, and medication status. Second, comorbid psychiatric disorders, such as personality disorders, were not considered. Third, the sex ratio differed between the two groups, with one group containing significantly more female participants. The results of binary logistic regression analyses showed that sex had an impact on NSSI. Fourth, we compared the presence of NSSI among patients with depression and did not compare them to a healthy control group.
To the best of our knowledge, no previous study has analyzed the FC of patients with NSSI using EEG. Therefore, this study is significant in that it compared FC between patients with and without NSSI by using EEG.
In this study, we did not observe differences in FC between patients with depression with and without NSSI. However, we anticipate that conducting future research with controlled factors such as the timing of EEG assessments or patient characteristics may reveal distinctive FC differences in NSSI patients. Studies with larger sample sizes can allow further research while controlling for variables to exclusively investigate the presence or absence of NSSI.
The authors express their gratitude to Kim B.E. for her support in collecting the data.
No potential conflict of interest relevant to this article was reported.
Conceptualization: Young-Min Park, Miseon Shim. Formal analysis: Hye-Jin Lee. Methodology: Young-Min Park, Miseon Shim. Software: Hye-Jin Lee. Writing—original draft: Hye-Jin Lee. Writing—review and editing: Young-Min Park.
CrossRef (0) Scopus (0) Download (336)
Author ORCID Information Hye-Jin Lee Young-Min Park Miseon Shim |
 |
