INTRODUCTION
By the end of 2022, the number of refugees reached 35.3 million worldwide [1]. Refugees face immense challenges not only in fleeing their homes and seeking safety but also in coping with traumatic experiences that significantly impact their mental health and well-being. Displacement involves danger, deprivation, and uncertainty, which can lead to substantial psychological distress. Additionally, separation from family, asylum procedures, detention, unemployment, inadequate housing, and acculturation issues can worsen existing mental health conditions or trigger new ones [2,3]. Refugees are at a heightened risk of developing mental disorders, including depression (2.3%-80%), post-traumatic stress disorder (PTSD; 4.4%-86%), and unspecified anxiety disorder (20.3%-88%), with prevalence estimates typically >20% [4]. These mental health challenges not only affect individuals but also have broader implications for the societies that receive them. Thus, addressing mental health issues among refugees is a pressing global concern that requires immediate action.
The division of the Korean Peninsula into North and South Korea originated at the end of World War II in 1945, when liberation from Japanese colonial rule was followed by a division along the 38th parallel. This separation was later entrenched by the 1950-1953 Korean War, after which the two Koreas remained divided and developed markedly different socio-political systems [5,6]. Because of the division, the Korean Peninsula has a unique context for refugees. As of March 2024, South Korea had 34,215 North Korean refugees (NKRs). Before the coronavirus disease 2019 pandemic, approximately 1,000 NKRs resettled in South Korea annually. However, this number declined due to border closures in North Korea, although it is expected to rise again [7]. NKRs often report severe traumatic experiences, such as starvation, deaths, lack of medical care, public executions, imprisonment, physical violence, and torture [8]. Additionally, during defection, they endure human trafficking, evasion from authorities, border encounters, shortages of food and water, and robbery [8], with female defectors facing higher risks of sexual violence [9]. In South Korea, the NKRs face significant adaptation challenges. Cultural adaptation stress impacts mental health more than the trauma experienced in North Korea or during escape [10]. Economically, they report high subsistence benefit rates (22.7% in 2023) and face limited job opportunities that are often restricted to low-paying jobs [11]. Numerous studies have indicated a high prevalence of psychiatric disorders such as PTSD, depression, and anxiety among NKRs [12-14], with NKRs being more likely to experience psychiatric disorders than South Koreans are [15].
Neuroimaging techniques are crucial for understanding the neurobiological underpinnings of PTSD and depression, revealing significant structural and functional changes in key brain regions, such as the amygdala, hippocampus, and prefrontal cortex [16]. Resting-state functional connectivity (RSFC) specifically offers insights into intrinsic brain interactions, without the influence of task-related confounders [17,18]. Previous RSFC studies on NKRs have shown that enhanced connectivity between frontal regions and the amygdala is linked to alexithymia [19], and that changes in the thalamus correlate with trauma-related features and compensatory mechanisms following traumatic events [20]. RSFC studies have been instrumental in identifying neuronal communication abnormalities in anxiety and mood disorders [21]. However, despite its potential, RSFC studies on PTSD, depression, and anxiety in NKRs are still rare. Our study addresses this gap by investigating RSFC patterns and their correlation with psychiatric symptoms in NKRs, with the aim of illuminating post-traumatic neurobiological alterations in this population.
To our knowledge, this is the first study to examine the relationship between RSFC and clinical symptoms of PTSD, depression, and anxiety in NKRs. This study aimed to investigate differences in RSFC between NKRs and South Korean healthy controls (SKCs). Additionally, we explored the correlations between these RSFC differences and psychiatric conditions such as PTSD, depression, and anxiety. We hypothesized that NKRs would exhibit distinct RSFC patterns compared with those of SKCs, reflecting altered neural circuitry associated with emotional, cognitive, reward, and somatosensory processing, and that these RSFC differences would correlate with the severity of psychiatric symptoms.
METHODS
Participants
In this study, 28 NKRs aged 19-60 years were recruited from the outpatient psychiatric clinic of the Korea University Anam Hospital in Seoul, Republic of Korea, between July 2020 and August 2021. A board-certified psychiatrist (K.-M. Han) assessed psychiatric disorders in NKRs according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) and conducted structured diagnostic interviews using the Structured Clinical Interview for DSM-IV-TR Axis I Disorders (SCID-I). Among the 28 NKRs, 22 (78.6%) had current psychiatric diagnoses such as major depressive disorder (MDD), PTSD, adjustment disorder, panic disorder, and primary insomnia, and 18 (64.3%) had been exposed to traumatic events during their escape, including life-threatening situations, detention and torture in North Korean camps, forced repatriation, and trafficking and sexual violence in China (while escaping through China). Exclusion criteria included presence of current psychotic symptoms, high suicide risk requiring immediate inpatient treatment, serious medical illnesses, neurological disorders, pregnancy or nursing, and contraindications to brain imaging (e.g., claustrophobia or metal implants). Additionally, 28 age-and sexmatched native SKCs were recruited from the community and verified as having no history of psychiatric disorders using the same SCID-I methodology and exclusion criteria as the NKR group. All the participants provided written informed consent and had the right to withdraw from the study at any time. Detailed participant information was obtained from a previous study on subcortical structure analysis [22]. The study protocol was approved by the Institutional Review Board of Korea University Anam Hospital (IRB No. 2008AN0054).
Psychometric measurements
Depressive symptom severity was assessed using the 17-item Hamilton Depression Rating Scale (HDRS-17) [23] and the Beck Depression Inventory-II (BDI-II) [24]. The HDRS-17 is a clinician-rated scale evaluating symptoms over the past week, while the BDI-II is a self-report measure covering the past 2 weeks, with responses rated on a 4-point Likert scale. Anxiety levels were evaluated using the State-Trait Anxiety Inventory (STAI) [25] for current anxiety symptoms and temperamental anxiety levels and the Hamilton Anxiety Rating Scale (HARS) [26] for clinician-rated anxiety symptoms experienced during the past week. In the NKR group, PTSD-related symptoms were assessed using the Clinician-Administered PTSD Scale (CAPS) [27], PTSD Checklist for DSM-5 (PCL-5) [28], and Impact of Event Scale-Revised (IES-R) [29]. The CAPS is a structured clinician-rated interview consisting of 30 items that assess the frequency and intensity of PTSD symptoms, and the “current” CAPS score was used for analysis in this study. The PCL-5 is a self-report measure of symptom severity based on 20 questions rated on a 5-point Likert scale, evaluating distress (according to the DSM-5 criteria) from the experiences of the past month. The IES-R is a self-report measure that assesses PTSD symptom severity over the past week, using 22 items rated on a 5-point Likert scale. To ensure consistency, all assessments were conducted on the same day as the magnetic resonance imaging (MRI) scans.
Functional MRI data acquisition
All participants underwent MRI on the day of clinical assessment. Functional and T1-weighted images of the participants were obtained using a 3.0-Tesla TrioTM whole-body imaging system (Siemens Healthcare GmbH) at the Korea University MRI Center. Anatomical images were acquired using the MPRAGE sequence (repetition time [TR], 1,900 ms; time to echo [TE], 2.52 ms; flip angle, 9°; matrix, 256×256; and resolution, 1×1×1 mm3), covering the entire head. Functional images were obtained using a single-shot echo planar imaging sequence (TR, 2,000 ms; TE, 30 ms; flip angle, 90°; number of slices, 42; matrix, 80×80; and resolution, 3×3×4 mm3). All images were checked by visual inspection for gross motion artifacts and potential artifacts due to metallic foreign bodies. No images were discarded after visual inspection.
Functional MRI data preprocessing
Resting state functional MRI (fMRI) was preprocessed using the CONN Functional Connectivity Toolbox (version 22a) [30] and SPM12 (http://www.fil.ion.ucl.ac.uk/spm). Preprocessing methods described in previous studies were employed to preprocess the functional and anatomical data using a flexible pipeline implemented in the CONN Toolbox [31]. In summary, this pipeline includes the following steps: slice-timing correction, outlier identification, direct segmentation and Montreal Neurological Institute (MNI)-space normalization, realignment with correction of susceptibility distortion interactions, and smoothing. After realigning functional data using the SPM realign and unwarp procedure [32], all scans were co-registered to the first scan of the first session using the least-squares method and a six-parameter (rigid body) transformation [33]. Potential outlier scans were detected using the Artifact Detection Tools scrubbing procedure as acquisitions with framewise displacement >0.9 mm or global blood-oxygen-level-dependent (BOLD) signal changes above five standard deviations [34]. Following a direct normalization procedure, functional and anatomical data were normalized to the standard MNI space, segmented into gray matter, white matter, and cerebrospinal fluid tissue classes, and resampled to 2 mm isotropic voxels using the SPM algorithm. Finally, spatial convolution was used to smooth the functional data using a Gaussian kernel with an 8 mm full width at half maximum. Furthermore, a standard denoising pipeline was employed to denoise the functional data using component-based noise correction (CompCor) [31], and the BOLD time series were bandpass frequency filtered between 0.008 Hz and 0.09 Hz to minimize the noise generated by scanner drift and physiological effects [35].
Functional connectivity analysis
RSFC between the NKRs and SKC groups was compared using the CONN functional connectivity toolbox (version 22a) [30]. For the seed-to-voxel FC analysis, brain regions such as pregenual ventromedial prefrontal cortex (vmPFC), dorsolateral prefrontal cortex (dlPFC), ventrolateral prefrontal cortex (vlPFC), pregenual anterior cingulate cortex (pgACC), dorsal anterior cingulate cortex (dACC), posterior cingulate cortex (PCC), anterior insula, amygdala, and ventral striatum of the bilateral hemispheres, which belong to the hubs of the default mode network, salience network (SN), cognitive control network, and reward network according to previous conceptual frameworks on neural circuitry of depression and anxiety by Williams [36] and Phillps et al. [37], were selected as seeds. The following regions of interest (ROIs) were selected as seed regions from recommendations of previous RSFC studies on MDD as follows: vmPFC (±6, 36, -12) [38], dlPFC (±46, 18, 44) [39], vlPFC (±30, 57, -6) [40], pgACC (±5, 38, 6) [41], dACC (±5, 27, 21) [41], PCC (±9, -56, 25) [42], anterior insula (±41, 3, 6) [43], amygdala (±25, -5, -15) [40], and ventral striatum (±8, 8, -8) [39]. The ROIs were defined as spheres with a radius of 5 mm and centered at MNI coordinates according to the Harvard-Oxford Subcortical Structural Probability Atlas in fMRI Imaging of the Brain Software Library.
Statistical analysis
The mean time series of each seed was used as a predictor in a multiple regression general linear model (GLM) at each voxel, and Pearson’s correlation analyses were performed for all other voxels in the cerebral region. For group-level analysis, Pearson’s correlation coefficients were converted to z-scores using Fisher’s r-to-z transformation. For the comparison analysis, GLM was used to measure group differences in functional connectivity (FC) maps between the NKR and SKC groups, including age and sex as nuisance covariates in all GLM analyses. A GLM was performed with an independent t-test to compare the mean z-scores of the two groups. For the statistical threshold, voxel-wise uncorrected p<0.001 was used to define the voxel size, and a cluster-wise correction was applied at a threshold of family wise error (FWE)-corrected p<0.05 to determine significant clusters.
Statistically significant seed-to-cluster connectivity was identified using the above-mentioned methods, and the mean z-scores (i.e., a measure of FC) were extracted for statistically significant seed-to-cluster connectivity using the CONN toolbox. For the correlation analyses between FC and severity of symptoms (i.e., PTSD, depression, and anxiety), Pearson’s correlation analyses between the mean z-scores of the significant seed-to-cluster connectivity and scores of the psychiatric scales were performed, including age and sex as nuisance covariates, with a statistical threshold of p<0.05. Correlation analyses were performed between NKR and SKC groups.
To investigate the differences in sociodemographic and clinical characteristics between the NKR and SKC groups, an independent t-test was used to analyze age and psychiatric scale scores, and a chi-square test was used to analyze the differences in sex distribution. Above-mentioned statistical and correlation analyses were performed using IBM SPSS Statistics for Windows (version 25.0, IBM Corp.).
RESULTS
Demographics and clinical characteristics of the participants
The sociodemographic characteristics and clinical variables of the NKR and SKC groups are presented in Table 1. There were no significant differences in age or sex between the two groups (p>0.1). Of the 28 NKRs, 22 had a psychiatric diagnosis (MDD, n=18; PTSD, n=11; adjustment disorder, n=3; panic disorder, n=1; primary insomnia, n=1), and 6 had no current psychiatric diagnosis. Of the 22 NKRs with psychiatric disorders, 12 had comorbid psychiatric disorders (MDD and PTSD, n=11; MDD and panic disorder, n=1). Psychometric results showed that compared with SKCs, NKRs had significantly higher scores on the HDRS-17, BDI-II, and HARS (all, p<0.001); however, there was no significant difference in state and trait anxiety (i.e., STAI-S and STAI-T) between the two groups. In the NKR group, the mean current scores of the CAPS, PCL-5, and IES-R were 21.54±13.86, 30.36±20.62, and 41.96±23.99, respectively.
Differences in RSFC between the NKRs and SKCs
Regarding FC between the two groups, NKRs had a significantly lower FC than that of SKCs in the following seeds and clusters: between the left dlPFC and left postcentral gyrus (FWE corrected p=6.08×10-4); between the right amygdala and right occipital pole (FWE corrected p=0.017); between the right ventral striatum and left lateral occipital cortex (FWE corrected p=0.007); and between the right ventral striatum and left cerebellum 6 (FWE corrected p=0.034). In contrast, NKRs showed significantly higher FC than that of the SKCs in the following seeds and clusters: between the left vmPFC and right putamen (FWE corrected p=3.20×10-4); between the left vmPFC and left putamen (FWE corrected p=0.026); and between the right dACC and right insula (FWE corrected p=0.043) (Table 2 and Figure 1).
Correlation between PTSD-related symptoms and FC in the NKR
In the NKR group, there was a significant negative correlation between the IES-R score and z-score of the FC between the right amygdala and right occipital pole (r=-0.427, p=0.030) (Table 3 and Figure 2A). A statistically non-significant trend of negative correlation was also observed between the PCL-5 score and FC between the right amygdala and right occipital pole (r=-0.377, p=0.058). Except for the above results, no significant correlation was found between FC and PTSD-related symptoms across the different seeds and brain regions (Table 3).
Correlation between depressive and anxiety symptoms and FC in the NKRs and SKCs
In the NKR group, there was a significant negative correlation between the STAI-T score and FC between the right ventral striatum and left cerebellum 6 (r=-0.416, p=0.035) (Supplementary Table 1 and Figure 2B). A statistically non-significant trend of negative correlation was also observed between the STAI-T score and FC between the right ventral striatum and clusters, including the lateral occipital cortex (r=-0.383, p=0.053) (Supplementary Table 1). In the SKC group, the FC between the left dlPFC and left postcentral gyrus was significantly negatively correlated with the HDRS-17 (r=-0.465, p=0.017) (Supplementary Table 2) and HARS scores (r=-0.426, p=0.030). In addition, there was a significant positive correlation of the FC between the right ventral striatum and left cerebellum 6 and the BDI-II score (r=0.425, p=0.031). A statistically significant positive correlation was also observed between the STAI-T score and FC between the right amygdala and right occipital cortex (r=0.431, p=0.028).
DISCUSSION
This study explored the RSFC differences between NKRs and SKCs and their association with psychiatric disorders, including PTSD, depression, and anxiety. NKRs showed a significantly lower RSFC between the left dlPFC and primary somatosensory cortex (i.e., postcentral gyrus), between the right amygdala and visual cortex (i.e., occipital pole), between the right ventral striatum and left lateral occipital cortex, and between the right ventral striatum and left cerebellum 6. In contrast, they exhibited higher RSFC between the left vmPFC and bilateral putamen and between the right dACC and right insula. The FC between the right amygdala and occipital pole was negatively correlated with IES-R scores, and the FC between the right ventral striatum and left cerebellum 6 was negatively correlated with STAI-T scores.
NKRs showed lower RSFC than that of SKCs between the right amygdala and primary visual cortex. The amygdala, which is pivotal in the SN, detects significant environmental changes [44], whereas the occipital cortex plays a crucial role in visual information processing [45]. Previous studies have linked the FC between the amygdala and visual cortex to emotional awareness and processing [46-48]. Individuals with PTSD often demonstrate heightened amygdala-visual cortex FC in response to trauma-related stimuli or negative emotional memories [49,50], which can lead to hyperreactive responses such as hypervigilance, visual intrusions, and flashbacks. Reduced FC in these regions at rest has been observed in generalized anxiety disorders [51] and social anxiety disorders [52], indicating disruptions in the processing of fear or threat stimuli. In MDD, reduced RSFC between the amygdala and visual cortices also indicates impaired bottom-up processing of emotionally charged visual stimuli [53]. Therefore, in PTSD, decreased FC between the amygdala and visual cortex in the absence of specific stimuli may be associated with hypoactive responses, hindering emotional recognition and processing. Furthermore, more severe PTSD symptoms correlated with weaker connectivity between the right amygdala and occipital pole in NKRs. Research has indicated that childhood trauma, particularly sexual abuse, may lead to structural changes in the occipital region that affect visual processing [54,55]. Given the high prevalence rate of sexual abuse and violence among North Korean women during escape, our findings suggest that psychological trauma may disrupt visual information-associated emotional recognition and processing and eventually contribute to the development of PTSD symptoms in NKRs.
Meanwhile, NKRs exhibited higher RSFC than SKCs between the left vmPFC and bilateral putamen. The striatum and vmPFC are integral components of reward processing in the affective circuitry [36,56]. Dysfunction in reward processing has been implicated in anhedonia [57,58], which is not only a primary symptom of MDD [59], but also a widely reported symptom of PTSD [60,61]. Striatal hypoactivation is a prominent feature of patients with depression, particularly those experiencing anhedonia [62-65]. Moreover, anhedonia has been associated with increased PFC activation, including the vmPFC, during the processing of positive-valence images [66-68] and anticipation of rewards [69,70], potentially reflecting compensation for striatal hypoactivation [36]. In other words, vmPFC hyperactivation may represent an adaptive compensatory mechanism in response to striatal dysfunction. Therefore, the increased FC observed between the vmPFC and putamen in NKRs in the present study suggests potential compensatory mechanisms following traumatic experiences to counteract reduced reward responsiveness (i.e., anhedonia).
NKRs also exhibited a higher RSFC than SKCs between the right dACC and right insular cortex. The dACC and anterior insula are essential components of the SN, which is crucial for detecting and responding to significant environmental changes [44,71]. Enhanced connectivity between these regions suggests an increased allocation and processing of attention to external stimuli or internal states. Individuals with heightened dACC connectivity within the SN often exhibit greater anticipatory anxiety in response to stressors [44]. Furthermore, the dACC and anterior insula are activated in response to pain, uncertainty, and threats to homeostasis [72-74], reflecting the hyperactive state of the SN associated with anxiety [75]. NKRs are likely to experience severe chronic stress and trauma, both physical and psychological, resulting in sustained high levels of anxiety and tension as they adapt to new environments. Therefore, our findings indicate chronic environmental stress and psychological anxiety in this group. Additionally, heightened FC within the SN in NKRs may serve as an adaptive mechanism, facilitating quick identification of and response to threatening stimuli, thereby supporting appropriate behavioral responses [44,76]. This enhanced FC could signify heightened sensitivity to environmental cues, potentially aiding survival in challenging circumstances.
In NKRs, the RSFC between the dlPFC and postcentral gyrus was lower than that in SKCs. The dlPFC serves as a critical neural substrate for cognitive control [77-79], whereas the postcentral gyrus is primarily involved in hierarchical somatosensory processing [80]. Therefore, the reduced FC between these regions may indicate diminished cognitive regulation during somatosensory information processing. This could potentially contribute to somatization symptoms, which are often characterized by a heightened awareness or misinterpretation of bodily sensations [81,82]. Previous functional neuroimaging studies have demonstrated abnormal activation patterns in the dlPFC and postcentral gyrus in individuals with conditions such as fibromyalgia, suggesting altered brain function in areas associated with somatosensory processing [83]. Similarly, reduced FC between the dlPFC and thalamus, another pivotal region in somatosensory processing [84,85], has been observed in patients with chronic pain and fibromyalgia [86,87]. These findings suggest that weakened connectivity between the dlPFC and somatosensory cortex may contribute to the manifestation of somatic symptoms in individuals with a history of psychological trauma.
NKRs also exhibited a lower RSFC than that of SKCs between the right ventral striatum and left cerebellum 6. The ventral striatum, which is part of the basal ganglia [88], and the cerebellum, which is traditionally linked to motor functions, are increasingly recognized for their roles in affective processes, including emotional recognition, subjective feeling elicitation, and reward valuation [89-95]. Disruption in this neural circuitry may introduce noise, impairing neural coordination and emotional recognition [95,96], potentially contributing to alexithymia, a difficulty in experiencing and expressing emotions [97], commonly observed as emotional numbing in PTSD [98]. These disruptions may also be associated with difficulties in emotional regulation such as emotion suppression, which is frequently observed in PTSD [99]. Therefore, the lower FC observed between the cerebellum and striatum in NKRs suggests challenges in emotional processing, including recognition and expression. Furthermore, weaker FC between the ventral striatum and cerebellum correlated with higher trait anxiety in NKRs. Indeed, both the striatum, which influences attention and motivation [100], and the cerebellum, which interacts with anxiety-regulating regions [101], are integral to the anxiety circuitry. Hence, the reduced FC observed between the ventral striatum and cerebellum may lead to difficulties in emotional recognition and processing, potentially exacerbating trait anxiety in NKRs.
The present study had several limitations. First, its cross-sectional design precludes the determination of the exact timing or causal relationship between traumatic experiences and FC in refugees. Future research employing a longitudinal design is essential to elucidate these relationships. Secondly, the sample size was small (28 participants per group). However, considering the challenges in recruiting a large sample of NKRs in a clinical setting, we believe that the sample remains meaningful despite its modest size. Third, the NKR group demonstrated considerable psychiatric heterogeneity, including varying levels of comorbidity. Because different psychiatric diagnoses can differentially affect RSFC, the observed RSFC differences may reflect trauma-related mechanisms, disorder-specific effects, or comorbidity, making interpretation challenging. Fourth, detailed information on psychotropic medication use was not systematically controlled for. Given that medications can alter RSFC, we cannot exclude the possibility that medication effects contributed to the findings. Finally, the current findings are specific to traumatized NKRs, thus limiting their generalizability to other refugee populations. Extending research to diverse refugee groups using data-driven approaches, such as independent component analysis, would enhance our understanding of the effects of trauma on functional brain connectivity and psychological health.
Our study revealed distinct patterns of RSFC in NKRs compared with SKCs, highlighting the potential neural correlates underlying psychiatric symptoms in NKRs. In the present study, significant alterations in RSFC in NKRs were observed across several neural circuits, including the visual cortex, somatosensory cortex, cerebellum, SN, and frontostriatal reward network. These findings implicate dysfunction in emotion recognition and processing, cognitive control over somatosensory information processing, heightened and maladaptive responses to environmental stressors, and reward processing. Alterations in FC in the NKRs highlight the necessity for precise and tailored psychological interventions to address these neurobiological and psychiatric challenges.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
