INTRODUCTION
Internet gaming disorder (IGD) is a behavioral addiction characterized by persistent and excessive engagement in online gaming, which leads to significant distress, functional impairment, and withdrawal symptoms [1]. The World Health Organization has acknowledged gaming disorder as a public health concern, classifying it under ‘disorders due to addictive behaviors’ in the 11th edition of the International Classification of Diseases, highlighting the importance of continued research on this emerging psychiatric condition [2].
As concerns about IGD increase, therapeutic interventions are being investigated, including transcranial direct current stimulation (tDCS), a noninvasive neuromodulation method that delivers a low-intensity direct current across the scalp to alter neuronal resting membrane potentials [3]. Lee et al. [4] demonstrated that tDCS decreased the weekly hours spent on gaming and increased self-control scores in online gamers. Another recent study showed that tDCS targeting the dorsolateral prefrontal cortex (DLPFC) improves inhibitory control in patients with IGD [5]. Functional magnetic resonance imaging (fMRI) revealed that tDCS increases resting state functional connectivity between the right DLPFC and the anterior cingulate cortex (ACC). However, this connectivity change did not correlate with symptom improvement, and the neurophysiological changes induced by tDCS that lead to clinical improvement remain unclear.
Resting-state electroencephalography (EEG), which records spontaneous electrical activity in the brain, is commonly used to examine neurophysiological features associated with psychiatric disorders [6]. EEG measures neural activity directly with high temporal resolution and can identify dynamic changes in functional brain networks. Recently, source-level EEG analysis using algorithms such as low-resolution brain electromagnetic tomography (LORETA) has been introduced to estimate the activity and connectivity of specific cortical regions, offering improved spatial resolution compared to that with sensor-level analysis [7].
Advances in EEG source localization techniques enabled the estimation of brain networks previously defined in fMRI studies, such as the default mode network (DMN) and the reward/salience network (RSN) [8]. Previous fMRI studies in patients with IGD suggested that dysfunctional connectivity in the DMN and the RSN is associated with impaired cognitive and reward processing [9-11]. While fMRI remains the gold standard for brain network identification, Lee et al. [8] utilized EEG source localization to identify the DMN and RSN, and demonstrated that resting-state hyperconnectivity within the brain networks is a potential state marker in IGD.
This study is a reanalysis of our previously published data set, in which we used sensor-level analysis of resting-state EEG features to investigate the neurophysiological effect of tDCS on patients with IGD [12]. A previous study revealed that tDCS decreased absolute gamma power in the left parietal region, while suppressing the increase in intra-hemispheric beta coherence. However, there was no significant association between the sensor-level EEG changes and clinical improvement. Therefore, in the present study, we performed a source-level analysis of resting-state EEG data obtained before and after tDCS. We chose the brain networks known to play important roles in addiction neurophysiology, the DMN and RSN, as our target of analysis. We hypothesized that tDCS has a significant neuromodulating effect on both neural networks and that functional connectivity changes would be correlated with clinical improvements. To the best of our knowledge, this is the first source-level EEG analysis to investigate the neurophysiological effect of tDCS on functional neural networks in patients with IGD.
METHODS
Participants
To diagnose IGD, a clinical psychiatrist interviewed 31 participants who engaged in excessive internet gaming. All participants were male, representing a male predominance in patients with IGD, with a reported male-to-female ratio of 2.5:1 [13]. The Mini-International Neuropsychiatric Interview (MINI) was used to identify and exclude participants with comorbid psychiatric disorders [14]. Patients with head injuries were also excluded. All participants were medication-naive throughout the initial assessment and intervention. The research was fully explained, and all participants provided informed consent before participating in the study. The participants were given approximately $50 to participate in the study. Participants completed the Korean version of the Wechsler Adult Intelligence Scale-IV (WAIS-IV), and participants with WAIS-IV scores < 80 were excluded. Three participants dropped out after the baseline assessment, and two dropped out after the intervention, resulting in a total of 26 participants.
This clinical trial was conducted following the Consolidated Standards of Reporting Trials (CONSORT) guidelines and registered at ClinicalTrials.gov (Identifier: NCT03347643). The study adhered to the principles outlined in the Declaration of Helsinki and obtained ethical approval from the Institutional Review Board of SMG-SNU Boramae Medical Center, Seoul, Republic of Korea (IRB No. 30-2017-10).
Measurements
Resting-state EEG was measured before and 1 month after the intervention, and source-level analysis was used to calculate the coherence values of the neural connections within the DMN and RSN. For secondary outcomes, clinical features including IGD severity and craving were assessed at baseline and the 1-month follow-up. Depression and anxiety were included in the baseline assessment to control for confounding variables.
EEG measurements
The participants were positioned in a resting state in an isolated sound-shielded room that was separated from the recording area using a one-way window. EEG data were collected over a 10-minute period: 4 minutes with eyes closed, 2 minutes with eyes open, and 4 minutes with eyes closed. Recordings were obtained using a 64-channel Quik-Cap (Compumedics Neuroscan), according to the modified International 10 to 20 system. All EEG signals were acquired through Syn-Amps 2 (Compumedics) and Neuroscan software (Scan 4.5; Compumedics). The signals were amplified at a sampling frequency of 1,000 Hz with an online bandpass filter of 0.1 to 100 Hz and an offline bandpass filter of 0.1 to 50 Hz. Electrode impedances were maintained below 5 kΩ.
EEG data were preprocessed using NeuroGuide software (NG; ver. 3.0.5, Applied Neuroscience). Nineteen of the 64 channels were selected based on a linked ear reference montage in NG. Artifacts were removed using NG artifact rejection toolbox and visual inspection, and artifact-free epochs during eyes-closed conditions were used for the analysis. Source-level coherence was assessed using NeuroNavigator (NG Deluxe 3.0.5, Applied Neuroscience), which applies swLORETA, a source localization method developed by Palmero-Soler et al. [7]. The analysis used real MRI images of the standard Collin’s brain from the Montreal Neurological Institute, comprising 12,270 voxels with 5-mm spatial resolution [15]. Weighted, undirected networks were constructed from the source-level coherence matrix, with Brodmann areas (BA) as nodes and coherence values as edges. The coherence between BA regions was computed in the delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–40 Hz) bands using fast Fourier transforms.
The NeuroNavigator software includes a built-in toolbox that groups specific BA into predefined functional brain networks and provides them through its ‘Symptom Check List’ panel [16]. The DMN was defined as a group of 20 regions of interest (ROI) on the left and right hemispheres: the postcentral gyrus (PoG, BA2), superior parietal lobule (SPL, BA7), anterior prefrontal cortex (aPFC, BA10), orbitofrontal cortex (OFC, BA11), associate visual cortex (BA19), posterior cingulate cortex (PCC, BA 29, 30), perirhinal cortex (PRC, BA 35), angular gyrus (AG, BA 39), and supramarginal gyrus (SMG, BA40). The RSN was defined as a combination of RSN. The reward network consisted of 18 ROIs in the left and right hemispheres: insular cortex (IC, BA13), ACC (BA24, 32), subgenual anterior cingulate cortex (sgACC, BA25), dorsal entorhinal cortex (EC, BA34), DLPFC (BA46), and inferior frontal gyrus (IFG, BA44, 45, 47). The salience network consisted of 22 ROIs on the left and right hemispheres: frontal eye field (FEF, BA8), DLPFC (BA9), aPFC (BA10), IC (BA13), superior temporal gyrus (STG, BA22), PCC (BA23, 29, 30), ACC (BA24, 32), and sgACC (BA25). The overlapping edges between the two networks (BA13, 24, 25, and 32) were counted once. Our method of using swLORETA to define the DMN and RSN was derived from a previous study by Lee et al. [8].
Clinical measurements
The Korean version of Young’s internet addiction test (IAT) uses 20 items rated on a 5-point scale to assess the severity of internet addiction, with a higher score indicating stronger addiction [17]. A visual analog scale (VAS) was used to measure craving for gaming, ranging from 0 (none) to 10 (maximum). The Beck depression inventory-II uses 21 items to measure the severity of depressive symptoms during the previous 2 weeks [18]. The Beck anxiety inventory includes 21 items measuring anxiety levels during the previous week [19].
Randomization & intervention
This randomized, double-blind, sham-controlled study used a SPSS version 20 (IBM Corp.) block randomization list (block size 4) to assign participants in a 1:1 ratio to active or sham tDCS. The devices were preprogrammed for each group, and the investigator was blinded to the assignment.
The tDCS protocol followed that of a previous study [20], with the anode over the left DLPFC (F3) and the cathode over the right DLPFC (F4) per the 10–20 system. Active stimulation was delivered at 2.0 mA for 20 minutes per session (Ybrain). Sham stimulation used the same electrode placement, but the device was turned off after the current was ramped up and down. Both groups underwent two daily sessions for 5 days (10 sessions total), with 20-minute intervals between sessions.
Statistical analysis
All statistical analyses were performed using R version 4.3.3 (R Development Core Team) Student’s t-test was used to compare demographic characteristics between the active and sham groups. The effect of stimulation type (active vs. sham) on the primary outcome (neural connectivity within brain networks) and secondary outcome (IAT and craving score) was analyzed using repeated measures analysis of variance. The group×time effect on coherence was calculated for each neural connection within the DMN and the RSN. For each edge that showed a significant group×time interaction (P< 0.05), four post hoc pairwise comparisons were performed (pre–post within the active group, pre–post within the sham group, between-group comparison at baseline, and between-group comparison at 1-month follow-up). Bonferroni correction was applied using these four comparisons as denominators (n=4). To examine the clinical relevance of the tDCS-induced neural connectivity changes, Pearson’s correlation analyses were performed for neural connections that showed statistical significance in the primary analysis. Correlations between clinical symptom improvement and neural connectivity changes were assessed in both active and sham groups, and baseline depression and anxiety levels were used as covariates.
RESULTS
Demographic and clinical data
The demographic and clinical characteristics of the active and sham groups are show in Tables 1, 2. There were no significant baseline group differences in demographics, depression, and anxiety scores. The mean IAT and craving scores decreased in both groups after the intervention; however, the group× time effect was not significant (IAT: F=0.338, P=0.567, η²p=0.014; craving: F=0.206, P=0.654, η²p=0.009). The distribution for IAT and craving scores are depicted in Supplementary Fig. 1.
EEG activity
The neural connections that showed significant group× time effects are listed in Table 3 and depicted in Fig. 1. In the DMN, tDCS affected 10 neural connections and had a heterogeneous effect. Post hoc analysis revealed that tDCS decreased delta band connectivity within the active group. In the theta band, tDCS affected the anterior prefrontal cortex connectivity. Post hoc analysis revealed that theta band connectivity was significantly weaker in the active group than in the sham group 1 month after the intervention. Neural connectivity in the beta band increased in the sham group, whereas no change was observed in the active group.
In the RSN, tDCS affected neural connections involving the right DLPFC and cingulate cortex in the same direction; tDCS increased neural connectivity compared to that with sham stimulation. Detailed descriptive statistics are presented in Supplementary Table 1.
Correlation analysis
A significant negative correlation was observed between neural connectivity changes and craving (VAS) changes at the two edges of the RSN: theta coherence between the right DLPFC and left PCC (r = –0.549, P = 0.042), and beta coherence between the right DLPFC and right sgACC (r= –0.663, P< 0.001). In both neural connections, a greater increase in neural connectivity after tDCS was correlated with a larger decrease in the craving score. The same analysis was conducted for the sham group to assess whether this relationship was specific to the active group. In the sham group, no significant correlation was found between craving and connectivity changes (P = 0.153 for theta coherence between the right DLPFC and left PCC; P= 0.146 for beta coherence between the right DLPFC and right sgACC). The correlation plots are shown in Fig. 2.
DISCUSSION
This study examined tDCS effects on neural connectivity in patient with IGD, targeting the DMN and RSN using EEG source localization. In the DMN, tDCS suppressed beta-band hyperconnectivity as observed in the control group. In contrast, tDCS strengthened neural connectivity within the RSN. Furthermore, increased right DLPFC connectivity in the theta and beta bands correlated with a reduction in craving. These findings suggest that changes in neural circuits may be associated with improved symptomatology, implying that a targeted, personalized approach when applying neuromodulation treatments might have superior outcomes for individuals with IGD.
Our previous sensor-level analysis of the same cohort found that tDCS decreased the absolute gamma power in the left parietal region and suppressed the increase in intra-hemispheric beta coherence [12]. Thus, we postulate that tDCS has a stabilizing effect on frontoparietal neural connections in the beta band, which are vulnerable to hyperconnectivity as IGD progresses. This finding is in accordance with a previous resting-state EEG source-level study, which revealed that patients with IGD have increased beta functional connectivity between the left OFC and left/right SMG compared to that in healthy controls [8]. Furthermore, a longitudinal study on neural connectivity changes in patients with IGD by Park et al. [21] demonstrated that intra-hemispheric beta and gamma coherence was higher in patients with IGD than in healthy controls, both at baseline and after 6 months of outpatient management with selective serotonin reuptake inhibitors (SSRI). Interestingly, the same study concluded that even after patients with IGD exhibited improvement in IGD symptoms, increased beta and gamma coherence persisted. This could explain why the control group exhibited increased beta connectivity within the DMN 1 month after sham stimulation. There is a possibility that increased beta connectivity within the DMN is a neurophysiological marker of IGD, which worsens over the duration of illness when left untreated, and persists even after pharmacotherapy. Our finding that the progression of frontoparietal beta hyperconnectivity observed in patients with IGD may be suppressed by tDCS suggests that tDCS has the potential to treat the core neuropathology of behavioral addiction.
In the RSN, tDCS increased theta- and beta-wave connectivity, with the right DLPFC and cingulate cortex being the most affected. The DLPFC is an important neural substrate in addiction, as it plays a critical role in executive functioning. Impaired executive control over cravings and negative emotions is a core neuropathological concept in addictive disorders ranging from substance use disorders to behavioral addictions [22]. Neuromodulation techniques that stimulate the DLPFC have been investigated in IGD, resulting in decreased craving and improved emotional control [23]. ACC is also associated with various cognitive functions that are impaired during addiction, such as decision making, emotional inhibition, and motivation [24]. Therefore, tDCS may stimulate the DLPFC and cingulate cortex in patients with IGD, thereby increasing functional connectivity and improving cognitive control. This hypothesis is consistent with the results of a previous fMRI study of patients with IGD, which showed that tDCS increased connectivity between the right DLPFC and ACC [5].
The neurophysiological effects of tDCS depend on multiple factors, including baseline neural activity, pathological connectivity, and long-term, network-level plasticity [25]. Given this complexity, the preferential increase in right DLPFC connectivity over left DLPFC connectivity is difficult to attribute to a single mechanism. One possible explanation is that the participants may have exhibited baseline hemispheric asymmetry within the DLPFC networks, which may have led to lateralized responsiveness to neuromodulation. Further research is warranted to clarify whether such pre-existing network imbalances exist and whether they contribute to hemispheric differences in tDCS-induced plasticity.
Although our study demonstrated that tDCS affects various neural connections in the DMN and RSN, the clinical effects of tDCS, as measured by IAT and craving scores, fell short of statistical significance. This may be because of several reasons. First, the 1-month follow-up period may not have been long enough to show symptom improvements induced by neural connectivity changes. Another possible explanation is that the intensity of stimulation (10 sessions over 5 days) was sufficient to produce subtle changes in neural networks but not enough to improve clinical symptoms. Therefore, additional research with higher stimulation intensities and longer follow-up periods are warranted. Furthermore, since DLPFC connectivity showed clinical relevance, follow-up studies including patients with IGD with aberrant DLPFC connectivity at baseline assessment for targeted neuromodulation may also be helpful.
This study has some limitations. First, the sample size was relatively small and the low statistical power of this study limits its use as a pilot study. Therefore, further studies with larger sample sizes are warranted. Second, only males were included, which hinders the generalizability of the study. Third, this study lacks the validity of using EEG source localization to interpret brain networks. Liu et al. [26] concluded that although EEG source localization can accurately mimic brain networks defined in fMRI, the density of EEG, influence of head modeling, and the source localization algorithm employed influence the accuracy. Although our study used only 64 electrodes in terms of EEG density, swLORETA, which uses the boundary element method to derive a more realistic head model, ensured better accuracy [18].
To the best of our knowledge, this is the first study to reveal the neurophysiological effects of tDCS on the resting-state EEG activity of patients with IGD using source-level analysis to focus on the DMN and RSN. These results are consistent with previous findings and provide additional insights into the mechanisms of neuromodulation in addictive disorders.
In conclusion, tDCS affects neural connectivity in the EEG-derived DMN and RSN in patients with IGD. Our results suggest that abnormal functional connectivity within the DMN and RSN is a neurophysiological marker of IGD. The correlation observed between changes in DLPFC connectivity and symptom improvement implies that tDCS may be targeted in patients with IGD who have decreased DLPFC connectivity.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
