Abnormal effective connectivity of default mode network subsystems affects depressive symptoms and sleep disturbances in adolescents with internet gaming disorder

Abstract

Background

Studies have demonstrated that Internet gaming disorder (IGD) was associated with sleep and psychological disorders. The default mode network (DMN) is not a unitary system, but rather is composed of smaller and distinct functional subsystems that interact with each other. We aimed to assess the functional relevance of these subsystems and the neural underpinnings of clinical symptoms in IGD.

Method

We recruited 47 IGD adolescents and 56 healthy controls, and examined their effective connectivity (EC) from three DMN subsystems by using dynamic causal modelling (DCM), including midline core network (PCC-aMPFC core DMN), medial temporal subsystem (MTL-DMN), and dorsal medial prefrontal subsystem (dMPFC-DMN). The relationships between neuroimaging findings, anxiety/depression and sleep disturbances were assessed through a mediation model. Finally, Leave-one-out cross-validation was used to investigate whether neural response could predict accompanying symptoms of IGD.

Results

The IGD group showed increased EC from PCC-aMPFC core DMN to MTL-DMN and increased self-connection of lateral temporal cortex (LTC) compared to the control group. The EC from anterior medial prefrontal cortex (aMPFC) to posterior inferior parietal lobule (pIPL) mediated the relationship between depression and sleep disturbances in IGD. Moreover, increased EC from aMPFC to pIPL could predict sleep disturbances of IGD adolescents.

Conclusions

This study emphasized the aberrant effective connectivity of different functional subsystems of DMN and revealed the neural correlates of accompanying symptoms in IGD adolescents. These findings suggested DMN subsystems-derived indicators could help to identify accompanying symptoms of IGD and could be explored as a target for neuromodulation interventions in future studies.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-026-08266-9.

Introduction

Internet gaming disorder (IGD) has been considered as a behavioral addiction and it is characterized by difficulties controlling internet gaming to the point of serious impairments in psychological and social functioning, showing similar neurobiological features to substance use disorders [1]. Evidence suggests that the risk of IGD may peak during adolescence, with an average prevalence of 5.5%, which has been increasingly attracting attention globally from public and scientific communities [2]. Studies have demonstrated that multiple negative elements such as anxiety, depression and family dysfunction are risk factors for IGD in adolescents [3], and comorbid conditions of these factors lead to a greater disease burden and a poorer treatment outcome [4]. Sleep disturbances are linked to IGD and are considered as a negative outcome or complication of IGD and further exacerbate psychological disorders [5, 6]. Previous studies have observed aberrant patterns of brain functional connectivity in the default mode network (DMN) across different classes of addiction (for example in Cannabis use and IGD), which are associated with behavioral impulsivity and negative emotions [7–9]. However, the neurological mechanisms behind the associations of accompanying symptoms in IGD is still unclear.

To further explore the psychological and neurological implications of IGD, recent studies have emphasized its strong links with a range of comorbid conditions, including sleep disturbances, depressive symptoms, and social functional impairments [5, 10]. In an investigation of American college students, IGD students had a greater proportion of suicidal thoughts (16.9% vs. 6.6%), depression (9.7% vs. 3.0%), social anxiety disorder (24.8% vs. 8.5%) and difficulties initiating sleep (13.0% vs. 5.7%) than the non-IGD group [11]. The relationships found in IGD, depression, anxiety, and sleep disturbances could be explained by mediated effects. Bhandari’s study demonstrated that IGD statistically mediated 16.5% of the indirect effect of sleep quality on depressive symptoms and sleep quality statistically mediated 30.9% of the indirect effect of internet addiction on depressive symptoms [12]. As a large-scale functional brain network that generally exhibits higher activity during resting state, the functional aberrance of DMN are believed to contribute to impaired self-awareness, negative emotions and ruminations in IGD [7]. Previous study found that the negative mood states of depression and anger in IGD individuals might reflect impaired functional connectivity (FC) among DMN regions [13]. Moreover, a growing body of evidence suggested that patients with sleep disturbances showed greater activation in core regions of DMN during self-reference-related tasks compared with controls [14]. Therefore, we hypothesized that a potential connection between IGD and sleep disturbances might be the dysfunction of DMN, which is involved in negative emotions and cyclically regulating sleep/wake states. Although research suggests separate robust associations among psychological disorders, theoretical models are still needed to explain the interaction between IGD, negative mood states and sleep disturbances.

IGD is a complex addictive disease and it is characterized by functional and structural aberrance in certain brain regions. Regarding IGD, previous studies have demonstrated abnormal resting-state functional connectivity (RSFC) in DMN-related regions and the DMN contributed most to distinguish IGD from healthy controls and could predict the severity of IGD [9, 15]. Therefore, the DMN is a potential biomarker for predicting clinical outcomes in IGD. Kim et al. found that IGD subjects showed increased local functional connectivity in the posterior cingulate cortex (PCC) and decreased local functional connectivity in the superior temporal gyrus compared to controls [16]. Previous study found that individuals with IGD displayed expanded RSFC between DMN-related regions (PCC-thalamus and PCC-medial prefrontal cortex [MPFC]) compared with controls [17]. In addition to these local or single connections, there are significant interactions between different components of DMN. Buckner’s study clarified the subsystems of DMN, and focused on how the subsystems activated in different emotion and cognition processes, including a midline core network (PCC-aMPFC core DMN), a medial temporal subsystem (MTL-DMN), and a dorsal medial prefrontal subsystem (dMPFC-DMN) [18–22]. Anterior medial prefrontal cortex (aMPFC) and PCC are involved in the midline core network, which integrates information from different cortical regions to facilitate self-related decision-making [23]. The MTL-DMN comprises medial temporal lobe (parahippocampal cortex), medial parietal cortex (retrosplenial cortex), the inferior parietal lobe (IPL) and ventral medial prefrontal cortex (vMPFC) and becomes active when decisions involve constructing a mental scene based on memory [18]. Dorsal medial prefrontal cortex (dMPFC) and lateral temporal cortex (LTC) are the critical components of dMPFC-DMN and involved in social cognition (e.g. understanding the mental states of others), self-reflection and conceptual processing [18, 23]. Both MTL-DMN and dMPFC-DMN subsystems are strongly connected with the midline core DMN [7]. The functional relevance of these subsystems was proved to be aberrant in IGD and associated with negative emotions. The negative mood states of depression and anger in individuals with IGD reflected impaired FC among DMN regions, for example, depression was negatively correlated with FC among ventral DMN regions and anger was negative correlated with FC among dorsal DMN regions [13]. Previous study has found the IGD group showed dysregulation in IPL-MPFC-PCC circuit of DMN and reduced self-connection in PCC and left IPL compared to the control group [24]. However, the DMN is not a unitary system, but rather is composed of distinct functional subsystems that interact with each other. The functional relevance between/within DMN subsystems is still unclear in IGD.

Previous studies investigated abnormal brain functional connections of IGD by using FC method. Seed-based RSFC analyses revealed that the left angular gyrus in the IGD group exhibited decreased connectivity with areas involved in DMN, and improved connectivity between the left angular gyrus and vMPFC might be the neural mechanism underlying the craving behavioral intervention effects [25]. However, FC could only measure statistical dependencies between neurophysiological signals, it could not uncover the directed effective/causal effects that underlie these dependencies [26]. This directed information flow is critical for understanding IGD pathophysiology, as different patterns of directed connectivity may underlie distinct cognitive and emotional dysregulations [2]. Previous study has demonstrated that adolescents with IGD exhibited reduced connectivity from the anterior cingulate cortex to the dorsolateral prefrontal cortex, a directional abnormality correlated with impaired cognitive control [27, 28]. Moreover, identifying causal connection could provide more precise targets for neuromodulation interventions, where the direction of abnormality informed the choice between excitatory and inhibitory interventions in the mood processing [29]. To overcome these limitations and better reveal information flow in the DMN, spectral DCM has been applied to resting-state fMRI [30]. Spectral DCM could not only estimate DCM parameters more efficiently, but also detects group differences sensitively by using parametric empirical bayes (PEB) analysis [31, 32]. Spectral DCM was shown to be effective with a great hypothesis testing capability, and has been widely used in neuropsychiatric diseases to assess aberrant communications within or between networks, such as tobacco use disorder and obsessive-compulsive disorder [33–35]. In IGD, Wang’s study conducted a preliminary study of EC in DMN using Spectral DCM, the main limitation of this study was that the authors only selected four key nodes of DMN to conduct the full-connected DCM modeling, while ignoring the subsystems of DMN [24]. The integration of functional networks and directed information flows among DMN subsystems of IGD need to be further understood.

This study aimed to assess effective connectivity alterations within three subsystems of DMN among adolescents with IGD by using spectral DCM. Then, correlation and mediation analyses were conducted to explore the neurological mechanisms behind the associations of accompanying symptoms in IGD. Finally, we investigated whether the altered neural response could predict the accompanying symptoms of IGD in each subject. We hypothesized that (i) adolescents with IGD would show aberrant ECs within/between DMN subsystems compared to healthy controls, especially in PCC-aMPFC core DMN and MTL-DMN subsystem, and such abnormalities could mediate the relationship between depressive symptoms and sleep disturbances in IGD; (ii) the specific EC abnormalities of DMN subsystems would be the mediator to predict accompanying symptoms (e.g. sleep disturbances) of IGD.

Methods and materials

Subjects

One hundred and three participants were recruited in Henan Province, including 47 adolescents with IGD and 56 healthy controls. All the participants were right-handed, aged from 12 to 18. IGD was diagnosed through clinical assessment by two experienced physicians and defined as meeting the proposed DSM-5 criteria for IGD, which has been recognized as a “condition for further study” in the DSM-5, and having Young’s Internet Addiction Test (IAT) score more than 40 (evaluating the severity of IGD) [36–38]. Healthy controls were recruited individuals who did not meet the proposed DSM-5 criteria for IGD and other mental disorders, and matched for age, gender and years of education with the IGD group. Exclusion criteria for all participants included: (i) any prior clinical diagnosis of neuropsychiatric disorders (e.g., schizophrenia, depressive disorder, or epilepsy); (ii) current use of psychotropic medications or concomitant substance use disorders; (iii) the presence of organic brain lesions; or (iv) contraindications for MRI scanning, including claustrophobia or ferromagnetic implants. This study was carried out in accordance with the Declaration of Helsinki and approved by the Medical Ethics Committee of First Affiliated Hospital of Zhengzhou University (2022-KY-0438–002), and informed consent was obtained from each participant. For participants under the age of 16, informed consent was obtained from their parents or legal guardians.

Clinical scales

Demographic data were collected by two experienced physicians. We collected clinical scales related to IGD including IAT [39], Pittsburgh Sleep Quality Index (PSQI) [40], Hamilton Depression Scale-24 (HAMD-24) [41], Hamilton Anxiety Rating Scale-14 (HAMA-14) [42] and Chinese version of the Family Environment Scale (FES-CV) [43]. The HAMD-24 was used to assess depressive symptom severity rather than to diagnose depressive disorder. More details about the above-mentioned scales can be found in supplementary.

Image acquisition

MRI data were obtained using a 3.0T German Siemens Magnetom Prisma magnetic resonance imaging equipment with a 64-channel head coil at First Affiliated Hospital of Zhengzhou University, Henan Province, China. Participants were instructed to rest with their eyes closed, keeping awake, not to think of anything, and to keep their head motionless during scanning. Earplugs were used to protect the hearing of subjects and spongy pads were used to fix their heads to minimize head movement. No external stimuli were exerted during image acquisition. The parameters were repetition time (TR)/echo time (TE)=1000/30 ms, flip angle = 70°, field of view = 220 mm × 220 mm, voxel size = 2 mm × 2 mm × 2.2 mm, slices = 52 and slice thickness = 2.2 mm, slice gap = 0.4 mm, a total of 400 volumes.

Image preprocessing

Data processing was conducted by the DPARSF toolbox based on MATLAB platform. The details of image preprocessing could be found in supplementary.

Dynamic causal modelling

Spectral DCM, as implemented in SPM12, was used for effective connectivity analysis. First, a general linear model was established in SPM with cosine basis function from 1/128 Hz to 0.1 Hz as the interest effects and the movement parameters, cerebrospinal fluid and white matter signals as nuisance regressors [44]. The regions of interest (ROIs) were selected based on previous researches [18]. A total of eight ROIs of DMN were selected, including PCC-aMPFC core DMN (anterior medial prefrontal cortex, and posterior cingulate cortex), dMPFC-DMN subsystem (dorsal medial prefrontal cortex, and lateral temporal cortex), and MTL-DMN subsystem (ventral medial prefrontal cortex, posterior inferior parietal lobule, retrosplenial cortex, and parahippocampal cortex). Previous studies have demonstrated that the components of DMN are predominantly left-lateralized, particularly the mid-orbital gyrus and middle temporal gyrus [45]. This left-lateralized organization is especially relevant in the context of IGD. Chen et al. recently reported that IGD individuals exhibited significant leftward gray matter asymmetry in core DMN regions [46]. These structural abnormalities suggest that left-hemispheric DMN abnormalities may be a critical neurobiological feature of IGD. Therefore, following Andrews-Hanna et al., we selected left-hemisphere ROIs to simplify the analysis, to avoid potential biases from strong interhemispheric correlations between homotopic regions, and to circumvent the strong laterality effects observed in lateral parietal regions [18, 47]. This approach ensured that the effective connectivity estimates could reflect intra-hemispheric dynamics within the left DMN, which has been shown to be predominantly lateralized in key nodes involved in self-referential and emotional processing [47]. ROIs were defined as 6-mm radius spheres centered at Montreal Neurological Institute (MNI) coordinates derived from previous studies (see in Supplementary Table S1). The time series from each spherical ROI was extracted with its center held fixed. Signal characteristics and spatial locations of all ROIs are displayed in Fig. 1. For each ROI, the first principal component of the time series across all voxels within the sphere was computed after regressing out confounds, including movement parameters, cerebrospinal fluid and white matter signals. Under resting-state conditions, a fully connected 8-node DCM model was built for each subject (including self-connection of each node, total 64 connections). Each full DCM was then inverted to estimate its free energy and posterior probability [48]. Finally, diagnostics of the model inversion was performed to assess the estimated parameters and the percentage variance explained by the model.

Fig. 1.

Eight ROIs are selected for DCM analysis. (A) spatial location of the ROIs. Yellow represents dMPFC-DMN subsystem (dMPFC and LTC). Red represents PCC-aMPFC core DMN (aMPFC and PCC). Green represents MTL-DMN subsystem (vMPFC, pIPL, Rsp and PHC). (B) sample blood oxygen level–dependent time series data of ROIs. These time series were used to invert the spectral DCM with the (fully connected) architecture. Note: DCM, dynamic causal modelling; ROI, region of interest; aMPFC, anterior medial prefrontal cortex; PCC, posterior cingulate cortex; dMPFC, dorsal medial prefrontal cortex; LTC, lateral temporal cortex; vMPFC, ventral medial prefrontal cortex; pIPL, posterior inferior parietal lobule; Rsp, retrosplenial cortex; PHC, parahippocampal cortex

Parametric empirical bayes (PEB) analyses

For the group-level analysis of all 64 tested effective connections (ECs), we employed the Parametric Empirical Bayes (PEB) approach [49]. PEB is a hierarchical empirical Bayesian framework that models inter-subject variability in connection strengths by estimating how individual connectivity parameters relate to group-level or condition-level means. This method utilizes the full posterior density of parameters derived from each subject’s DCM to inform inferences at the group level [50]. To identify differences in connectivity between IGD adolescents and healthy controls, we applied Bayesian Model Reduction (BMR) to systematically evaluate PEB models featuring various combinations of connections and group effects. BMR was implemented to iteratively remove redundant connection parameters from the full model until further pruning reduced model evidence. The parameters of the best 256 pruned models were then averaged, weighted by their evidence. In the PEB framework, group-level inference is conducted using Bayesian posterior probability, which inherently avoids the multiple comparisons problem due to the absence of false positives [51]. Bayesian posterior probability (Bayesian-PP) served as a measure of confidence, with higher values indicating stronger evidence. Here, the results were considered reliable if Bayesian-PP > 0.95 [52].

Leave-one-out cross validation

Assessing predictive validity is particularly important for studies determining the clinical significance of model parameters. In order to investigate whether the neural response could predict IGD and its accompanying symptoms (e.g. sleep disturbances, depressive and anxiety) in each subject, we performed leave-one-out cross validation analysis to explore relationship between IAT, PSQI, HAMD-24, HAMA-14 scores and ECs with significant group difference. A PEB model was fitted to all but one subject and IAT, PSQI, HAMD-24, HAMA-14 scores for the left-out subject were predicted, with age, gender and years of education as confounds [53]. This was repeated with each subject left out and the accuracy of the prediction was recorded. The Pearson correlation coefficient was calculated between the actual and predicted scores for overall predictive performance [53].

Statistical analyses

Data analysis was conducted using SPSS 22 software. Descriptive data for continuous variables are expressed as mean ± standard deviation. Group differences in continuous (age, years of education, fame-wise displacement [FD]) and categorical (gender) demographics were examined using two-sample t-tests and the chi-square test, respectively. Correlation analyses were performed to explore the relationships between ECs with significant group difference and clinical scales, including IAT, PSQI, HAMD-24, HAMA-14 and FES-CV.

Based on our hypothesis, mediation analyses were conducted to explore whether altered ECs of DMN mediated the relationship between depressive or anxiety symptoms and sleep disturbances in individuals with IGD by using SPSS PROCESS. The significance of the indirect (mediated) effect was assessed using 10,000 bootstrapped samples to obtain estimates and CIs. The indirect effect was considered significant if the 95% CI did not include zero [54].

Results

Demographic and clinical data

Table 1 summarize the demographics and clinical information related to IGD. No significant differences were found in age, years of education, gender and FD between the IGD group and the control group.

Table 1.

Demographic and clinical features

	IGD（n = 47）	HC（n = 56）	Statistical values	p
Age(year) a	14.6 ± 2.0	15.5 ± 5.5	−1.123	0.264
Education(year) a	8.8 ± 1.9	9.0 ± 4.1	−0.320	0.750
Gender(male/female) b	38/9	38/18	2.231	0.135
FD (mm)	0.16 ± 0.06	0.14 ± 0.07	1.059	0.294
IAT	62.1 ± 10.7	–	–	–
PSQI	8.3 ± 3.6	–	–	–
HAMD-24	23.3 ± 13.2	–	–	–
HAMA-14	14.5 ± 9.1	–	–	–
FES-CV
Cohesion	4.3 ± 2.7	–	–	–
Expressiveness	4.1 ± 1.5	–	–	–
Conflict	5.3 ± 2.3	–	–	–
Independence	5.4 ± 1.4	–	–	–
Achievement orientation	4.8 ± 1.8	–	–	–
Intellectual-cultural orientation	3.4 ± 1.9	–	–	–
Active-recreational orientation	3.3 ± 2.2	–	–	–
Moral-religious emphasis	4.0 ± 1.7	–	–	–
Organization	4.3 ± 1.8	–	–	–
Control	3.7 ± 1.5	–	–	–

Note: Data represent mean ± standard deviation; ^a two sample t test; ^b Chi-square test; IGD, Internet gaming disorder; HC, healthy control; FD, frame-wise displacement; IAT, Young’s Internet addiction test; PSQI, Pittsburgh Sleep Quality Index; HAMD-24, Hamilton Depression Scale-24; HAMA-14, Hamilton Anxiety Rating Scale-14; FES-CV, Chinese version of the Family Environment Scale

Group comparison

The average explained variance of the model was close to 88.47%, suggesting that the model had a good fitting effect. Compared to healthy controls, individuals with IGD showed significant increased EC from anterior medial prefrontal cortex (aMPFC) to posterior inferior parietal lobule (pIPL), EC = 0.072 HZ, Free energy Bayesian-PP = 1.00. Moreover, lateral temporal cortex (LTC) showed enhanced self-connection in the IGD group compared to the control group, EC = 0.125 HZ, Free energy Bayesian-PP = 1.00 (Fig. 2).

Fig. 2.

Results of group comparison. (A) EC difference between the IGD group and the control group, whereby positive values indicate increased EC in the IGD group. (B) compared to healthy controls, individuals with IGD showed significant increased EC from aMPFC to pIPL (EC = 0.072 HZ, Free energy Bayesian-PP = 1.00, yellow arrow) and increased self-connection of LTC (EC = 0.125 HZ, Free energy Bayesian-PP = 1.00, yellow ball represents enhanced self-connection in this brain region). Note: IGD, internet gaming disorder; EC, effective connectivity; aMPFC, anterior medial prefrontal cortex; PCC, posterior cingulate cortex; dMPFC, dorsal medial prefrontal cortex; LTC, lateral temporal cortex; vMPFC, ventral medial prefrontal cortex; pIPL, posterior inferior parietal lobule; Rsp, retrosplenial cortex; PHC, parahippocampal cortex

Leave-one-out cross validation

PSQI scores predicted by EC from aMPFC to pIPL are significantly correlated with the actual scores in individuals with IGD (r = 0.51, p = 0.003, FDR corrected). Figure 3C exhibited the results of this analysis. The left plot showed the actual and (out-of-sample) predicted PSQI scores (after mean correction and standardization). The shaded area represents the 90% Bayesian confidence intervals. The right plot showed the correlation between actual values and predicted values. Therefore, the effect size estimated by DCM could predict sleep disturbances in IGD.

Fig. 3.

(A) results of the correlation analysis. Self-connection of LTC was negatively correlated with Independence subscale of FES-CV. The EC from aMPFC to pIPL was positively correlated with PSQI scores and HAMD-24 scores. (B) the mediation effect of EC from aMPFC to pIPL on the influences of depressive symptoms and sleep disturbances in IGD (indirect effect = −0.1603, bootstrapped 95% CI = [0.0144, 0.4703], p < 0.05) with the approximate proportion of mediation of 27.1%. (C) Leave-one-out cross validation. Left: the actual and (out-of-sample) predicted PSQI scores based on EC from aMPFC to pIPL (after mean correction and standardization). The yellow dashed line was the actual values. The red line was the predicted values. The shaded area represents 90% confidence interval. Right: the out-of-samples correlation of the actual PSQI scores against the predicted scores based on EC from aMPFC to pIPL for each left-out subject (r = 0.51, p = 0.003, FDR corrected). Note: IGD, internet gaming disorder; LTC, lateral temporal cortex; aMPFC, anterior medial prefrontal cortex; pIPL, posterior inferior parietal lobule; PSQI, Pittsburgh sleep quality Index; HAMD-24, Hamilton depression Scale-24; FES-CV, Chinese version of the family Environment Scale

Correlation and mediation analyses

The self-connection of LTC was negatively correlated with Independence subscale of FES-CV (r = −0.460, p = 0.024) in IGD group. The EC from aMPFC to pIPL was positively correlated with PSQI scores (r = 0.559, p = 0.002) and HAMD-24 scores (r = 0.437, p = 0.005) in IGD group. See in Fig. 3A.

Our results showed that EC from aMPFC to pIPL partially mediated the relationship between depressive symptoms measured by HAMD-24 scores and sleep disturbances measured by PSQI scores in IGD (indirect effect = −0.1603, bootstrapped 95% CI = [0.0144, 0.4703], p < 0.05) with the approximate proportion of mediation of 27.1% (Fig. 3B).

Discussion

This study focused on the aberrance of effective connectivity in DMN subsystems among adolescents with IGD and the neural correlates of accompanying symptoms in IGD. Among the eight key nodes of DMN, the IGD group showed increased EC from PCC-aMPFC core DMN to MTL-DMN subsystem (especially the aMPFC-pIPL circuit) and increased self-connection of LTC compared to the control group. The EC from aMPFC to pIPL mediated the relationship between depressive symptoms and sleep disturbances in IGD. Moreover, increased EC from aMPFC to pIPL could predict the sleep disturbances of IGD adolescents. These findings of the DMN subsystems provide novel insights into the neural mechanism of IGD and its accompanying symptoms.

Self-connection changes of LTC

Firstly, we found that the IGD group exhibited increased self-connection of LTC compared to the control group. Since the self-connection of DCM is always inhibitory, the results of increased self-connection in LTC suggest that this brain area is more self-inhibited [53, 55]. Consistent with our results, Ma’s study found that individuals with IGD exhibited decreased local functional connectivity in temporal lobe compared to healthy controls [56]. During the cue-induced internet game video task (watching a game video from World of Warcraft during MRI data acquisition), individuals with IGD showed increased activation in superior temporal gyrus compared to healthy controls [57]. Together, these findings confirmed the critical role of LTC dysfunction in IGD. Notably, our study extended these findings by demonstrating that self-connection in LTC were correlated with family environment factors, especially it could reflect the degree of self-esteem, self-confidence and autonomy of adolescents in the family. Previous study found that stress, being bullied, low self-esteem and family dysfunction were the possible risk factors of IGD adolescents [3]. Therefore, a reasonable hypothesis is that a discordant family environment may be linked to the dysfunction of temporal cortex, potentially representing a risk factor for IGD. The current findings established a specific link between LTC intrinsic inhibitory dynamics and family environment and provided a novel direction for future research on psychosocial influences in IGD adolescents.

EC alterations between DMN subsystems

As mentioned before, the DMN is now recognized as being composed of distinct functional subsystems rather than a unitary neural system [18, 58]. Recent studies have predominantly focused on the midline core of the DMN, with widespread evidence demonstrating disrupted functional connectivity between the MPFC and PCC in IGD [24]. Our study extended this circuits by further investigating its interactions with two closely connected subsystems, the MTL-DMN and dMPFC-DMN. Previous studies have proposed that aberrant connectivity between midline core DMN and MTL-DMN contribute to the conditioning of internal affective states with experiences of drug intake in substance addictions [7, 59]. Song’ study demonstrated that the positive functional connection between midline core DMN and MTL-DMN could predict IGD, which might reflect interference of IGD in recollecting personal information and further affect the conditioning of emotional states in IGD [15]. On this basis, our study complemented the existing findings that the aberrant connection between midline core DMN and MTL-DMN in IGD was directional, and such directional connectivity changes was associated with emotional dysregulation. Moreover, by focusing specifically on adolescents with IGD, our study emphasized a critical developmental period characterized by functional network reorganization, shifting from local interactions to more distributed connectivity to support executive functions and cognitive control [60]. Within this developmental context, we proposed that increased EC from aMPFC to pIPL in IGD adolescents may reflect abnormal organization of functional network shifts, a process that has been related to emotional dysregulation and impulsive behaviors in previous studies [60, 61]. This perspective may offer a partial explanation for the vulnerability to addiction during adolescence period [61].

Nevertheless, these interpretations should be considered in light of a methodological feature of our study, namely that the DCM analysis was restricted to left-hemisphere ROIs. This decision was based on prior evidence of left‑lateralized DMN abnormalities in IGD, and on the goal of avoiding strong interhemispheric correlations that could confound effective connectivity estimates [18, 45, 46]. However, both emotional processing and sleep regulation, two domains closely linked to our findings, are known to involve bilateral DMN engagement. Previous studies have demonstrated that increased network homogeneity in the right precuneus has been linked to sleep disturbances in major depressive disorder, and functional connectivity of right DMN regions (e.g., right precuneus and right temporal pole) has been associated with both sleep quality and emotion regulation strategies [62–64]. Therefore, the aberrant effective connectivity we observed from aMPFC to pIPL should be understood primarily as reflecting left‑hemisphere dynamics. We cannot exclude the possibility that right‑hemisphere or cross‑hemispheric pathways also contribute to the association between DMN subsystem dysfunction and clinical symptoms in IGD. Future studies incorporating bilateral ROIs are needed to determine whether complementary or lateralized effects exist in the right hemisphere.

Relationships between IGD and accompanying symptoms

Our findings confirmed the above speculative conclusion that aberrant EC from midline core DMN to MTL-DMN in IGD adolescents could be associated with mood states. Correlation analyses showed that the EC from aMPFC to pIPL was positively correlated with PSQI scores and HAMD-24 scores. Previous studies have reported associations between gaming disorder and psychological distress (depression, anxiety and stress), as well as sleep problems in adolescents (insomnia, sleep duration and sleep latency) in adolescents [65, 66], with Bhandari et al. demonstrating statistical mediation effects among these variables [12]. In this current study, we revealed the neural correlates behind such mediated effects from the perspective of brain network. Our findings demonstrated that the EC from aMPFC to pIPL partially mediated the relationship between depressive symptoms and sleep disturbances in IGD. This finding aligns with previous observations that pharmacological intervention with escitalopram improved both depressive and IGD symptoms while reducing functional connectivity within DMN regions, including the PCC-aMPFC core DMN and left MPFC to left parietal cortex [67]. Moreover, previous study also found that individuals with sleep disturbances exhibited decreased FC among MPFC, inferior parietal cortex and medial temporal lobe areas compared to controls [68]. Combined, we infer that the EC from aMPFC to pIPL may represent a shared neurofunctional correlate of depressive symptoms and sleep disturbances in adolescents with IGD. In addition, internet gaming may begin as a strategy for adolescents to cope with emotional distress. However, adolescents may increasingly rely on games as a method of dealing with negative emotions, a pattern that may be accompanied by addiction to online activities, escape from the real world, and insufficient sleep duration and delayed bedtimes [69, 70]. The mediating role of the aMPFC-pIPL circuit provides a neurofunctional perspective for understanding the relationship between depressive symptoms and sleep disturbances in IGD adolescents. However, given the cross-sectional design, causal inference cannot be established, and these findings should be interpreted cautiously. The aMPFC-pIPL circuit may represent a potentially relevant neural biomarker for future longitudinal and intervention studies.

EC from PCC-aMPFC core DMN to MTL-DMN predicted sleep disturbances

Of note, this current study found that EC from PCC-aMPFC core DMN to MTL-DMN could predict PSQI scores in adolescents with IGD, which indicated that adaptive changes of this circuit might be a potential neural mechanism of sleep disturbances in IGD. The potential mechanisms in disrupting sleep may include psychological stimulation from gaming [71], bright screens [72] and alterations in sleep cycle [73]. The interrelationships of psychological stimulation and sleep disturbances in adolescents with IGD has been fully discussed in the previous section and will not be repeated here. For rapid eye movement (REM) sleep phase, it has been implicated in coordinating various brain functions. Previous study has demonstrated that greater functional connections between retrosplenial cortex/hippocampus (key nodes of MTL-DMN) and PCC, MPFC and inferior parietal cortex were associated with lower sleep efficiency, lower amounts of REM sleep and greater sleep-onset latency in individuals with primary insomnia [74]. Building on this evidence, our study extended prior findings by demonstrating that directional connectivity between the PCC-aMPFC core DMN and MTL-DMN (specifically the aMPFC-pIPL pathway) served as a significant predictor of subjective sleep disturbances in IGD adolescents. However, given that the modest sample size and model complexity may introduce a risk of overfitting, this finding should be interpreted cautiously. Independent validation in larger cohorts will be needed to confirm the robustness and generalizability of this association. It is also important to note that our current design cannot directly establish the links between sleep physiology and neural dynamics, and the underlying physiological sleep disruptions need further investigation using objective measurements in future studies.

Limitations and future work

Our findings suggest several avenues and challenges for future research. First, this study had a cross-sectional design and a relatively small sample size. Although the inversion model diagnostics and cross-validation procedures provide internal robustness evidence, future studies with larger samples and longitudinal data are still needed to further validate the dynamic changes in EC and their relationship with accompanying symptoms. Second, the eight ROIs of DMN are from the left hemisphere, which limit insight into the potential role of right-hemispheric or interhemispheric dynamics, particularly in emotional and sleep regulation. Previous study has demonstrated that the populations with comorbid depression and sleep disturbances showed interhemispheric functional connectivity aberrance [75]. Future studies incorporating bilateral and cross-hemispheric analyses will be important to validate and extend these findings. Third, sleep quality was assessed by subjective self-reports (PSQI questionnaire) in this study, it may be better to combine objective measures such as polysomnography. Therefore, the clinical implications of our findings, particularly the role of the aMPFC-pIPL circuit in sleep disturbances, should be interpreted with caution. Finally, our current study did not stratify the participants based on the severity of disease. Consequently, the subtle characteristics of DMN subsystems in IGD adolescents scoring near the cut-off threshold of IAT scale were overlooked, which may represent a clinically meaningful “at-risk or problematic” population. We consider that future work should prioritize longitudinal studies with well-documented sleep quality metrics to examine the long-term evolution of sleep disturbances and negative emotions across the development cycle of IGD. Particularly in adolescent populations, it is crucial to investigate stratified risk patterns or dynamic shifts in IGD severity levels, while considering the neurodevelopmental context. The aMPFC-pIPL circuit suggested by our findings could be considered as a potential target or ROI for future studies.

Conclusion

Overall, this study emphasized the aberrant effective connectivity in different functional subsystems of DMN in adolescents with IGD and revealed the neural correlates of accompanying symptoms in IGD. Specifically, our findings indicated dysfunction from PCC-aMPFC core DMN to MTL-DMN subsystem in IGD and it could mediate the relationship between depressive symptoms and sleep disturbances in adolescents with IGD. Moreover, the abnormal communication of aMPFC-pIPL could predict the sleep disturbances of IGD adolescents. From a clinical perspective, the predictive value of this circuit raises the possibility that DMN subsystem-derived indicators may contribute to identifying accompanying symptoms of IGD and could be explored as a target for neuromodulation interventions in future studies.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

All authors take responsibility for the content, gave approval of the submission and made substantive intellectual contributions to the submitted work. Z.M. and M.L. conceived and designed the study. D.J., T.Q. and S.J. analyzed the data. Z.M. drafted the manuscript. H.S. and W.W. revised the manuscript. Z.Y. supervised. Z.Y. provided financial support.

Funding

National Natural Science Foundation of China (National Natural Science Foundation of China) (82471962); Funding for Scientific Research and Innovation Team of The First Affiliated Hospital of Zhengzhou University (Funding for Scientific Research and Innovation Team of The First Affiliated Hospital of Zhengzhou University) (Z.Y. Grant/Award Number QNCXTD2023007).

Declarations

Ethics approval and consent to participate

This study was carried out in accordance with the Declaration of Helsinki and approved by the Medical Ethics Committee of First Affiliated Hospital of Zhengzhou University (2022-KY-0438–002), and informed consent was obtained from each participant. For participants under the age of 16, informed consent was obtained from their parents or legal guardians.

Conflict of interest

The authors declare no competing interests.

Data availability

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.