(REMISSION) Functional neural changes and altered cortical-subcortical connectivity associated with recovery from Internet gaming disorder (2019)

From 2016 to 2017, we recruited 154 IGD subjects for fMRI during a cue-craving task (described below). We contacted participants after approximately 1 year and reevaluated them for IGD. Twenty-nine IGD subjects (five females) who did not satisfy the criteria of IGD anymore agreed to participate during scanning when performing the cue-craving task. We then compare their more recent data (recovered IGD) to the baseline data (active IGD) to identify differences over time (Figure 1A).

At study onset, participants were classified as having IGD if they scored 50 or higher on Young’s Internet Addiction Test (a self-report questionnaire) and met at least five DSM-5 criteria for IGD (clinical interview; see “Supplemental Material” for additional details; Petry et al., 2014; Young, 2009). All participants underwent structured psychiatric interviews (MINI) conducted by an experienced psychiatrist (Lecrubier et al., 1997), and individuals with psychiatric disorders or behaviors were excluded (see “Supplemental Material”). In addition, no subjects reported previous experience with gambling or illicit drugs (e.g., cannabis and heroin). All subjects played League of Legends (LOL and Riot Games) for more than 1 year. This criterion was based on our use of gaming cues as stimuli in this study and LOL being the most popular online game during study onset. Individuals who recovered from IGD were needed to score less than 50 on Young’s Internet Addiction Test and meet less than five DSM-5 criteria for IGD at the 1-year time (Petry et al., 2014; Young, 2009; see Table 1 for details).

Table 1. Demographic features of IGD participants when IGD was active and recovered

An event-related cue-reactivity task was used in this study, as has been described previously (Dong et al., 2017; Dong, Wang, et al., 2018). The task contains two types of cue pictures: 30 gaming-related pictures and 30 typing-related pictures (neutral baseline). Within each type, half of the 30 pictures contained a face and hands and half contained only hands. Gaming-related pictures show a person who is playing the online game (LOL) on a computer. In typing-related pictures, the same person is typing an article on a keyboard in front of a computer. Participants were instructed to indicate whether or not there was a face in the picture by pressing the button “1” on the keyboard when a face was present and pressing “2” when there was no face present.

Figure 1B shows the timeline of a sample trial in the task. First, a fixed 500 ms of cross was presented, followed by a cue picture as described above. Pictures were presented in a randomized order to avoid order effects. Each picture was presented for up to 3,000 ms, during which time participants needed to respond. The screen turned to black after button-pressing and lasted for 3,000 (response time) ms. Then, in the craving evaluation stage, participants were asked to evaluate the level of their craving for the corresponding stimuli on a 5-point scale, ranging 1 (no craving) to 5 (extremely high craving). This stage lasted for up to 3,000 ms and was terminated by a button press. Finally, a 1,500–3,500 ms blank screen was presented between each trial. The whole task contained 60 trials and lasted approximately 9 min. The task was presented and behavioral data were collected using E-prime software (Psychology Software Tools, Inc., Sharpsburg, PA, USA). All participants were asked to complete a 10-item gaming urge questionnaire, with scores ranging from 1 to 10, to assess gaming-related craving prior to fMRI (Cox, Tiffany, & Christen, 2001).

Preprocessing of the fMRI data was conducted using SPM12 (http://www.fil.ion.ucl.ac.uk/spm) and Neuroelf (http://neuroelf.net), as described previously (Dong et al., 2017; Dong, Wang, et al., 2018). Images were slice-timed, reoriented, and realigned to the first volume, with T1-coregistered volumes used to correct for head movements. Images were then normalized to MNI space and spatially smoothed using a 6-mm full width at half maximum Gaussian kernel. No subjects were removed from analysis because of head motion (the exclusion criteria were 2 mm in directional movement or 2° in rotational movement). A general linear model (GLM) was applied to identify BOLD activation in relation to brain activities. Different types of trials (gaming-related, typing-related, incorrect, or missed) were separately convolved with a canonical hemodynamic response function to form task regressors. The duration of each trial was 4,000 ms. The GLMs included a constant term per run. Six head-movement parameters derived from the realignment stage and gaming history (self-reported years of gaming) were included to address these potential confounds. A GLM approach was used to identify voxels that were significantly activated for each event during the “response” stage.

The second-level analyses were conducted as follows. First, a voxel-wise repeated-measures analysis throughout the whole brain was conducted to investigate activity related to [(recovered_{Gaming-related stimuli} − recovered_{Typing-related stimuli}) −  (active_{gaming-related stimuli} − active_{Typing-related stimuli})]. Family-wise error thresholds (p < .001) were determined using 3dClustSim (an updated version of Alphasim), and all comparisons were corrected using 3dClustSim (https://afni.nimh.nih.gov/pub/dist/doc/program_help/3dClustSim.html), p < .001, two-tailed, with an extent of at least 40 voxels.

This experiment was approved by the Human Investigations Committee of Zhejiang Normal University and conformed to The Code of Ethics of the World Medical Association (Declaration of Helsinki). All participants provided written informed consent before scanning.

We analyzed the correlations between the brain responses in the left ACC and lentiform and self-reported craving to cues. Significant correlations between self-reported cravings and lentiform activations were found, regardless of IGD status (Figure 3). No significant correlations were observed between ACC activations and cravings.

We further analyzed the effective connectivity between the left ACC and left lentiform using dynamic causal modeling (DCM) at the two timepoints. The nodes used were defined through the interaction results presented above. Of the several brain regions identified in the whole-brain analyses, the ACC is located in the executive-control network and the lentiform nucleus in the reward network. Given our hypothesis that executive control over craving should be changed in recovery from IGD, we selected these two brain regions as regions of interests in this study for connectivity analyses. In other words, we selected these two regions as components of executive-control and reward networks to investigate interactions between these two systems in recovery from IGD.

We took the coordinate of the peak of the clusters (local maxima in the statistical map) as the central point to create spheres with 6-mm radii [left lentiform (−21, 3, 21); ACC (−3, 39, 6)]. About 33 voxels were included in each sphere. These regions identified for each group were included in a dynamic network, and DCM was used to determine the network’s most likely structure, given the data.

In fixed connectivity, DCM estimates demonstrated significantly increased ACC–lentiform connectivity when IGD subjects recovered (t = 3.167, p = .003). Similarly, lentiform–ACC connectivity was also significantly increased when IGD subjects recovered (t = 4.399, p < .001).

Similar features were also observed when subjects were exposed to gaming cues. In modulating effects, DCM estimates demonstrated significantly increased ACC–lentiform connectivity when IGD subjects recovered (t = 2.769, p = .009). However, lentiform–ACC connectivity was only marginally increased when IGD subjects recovered (t = 1.798, p = .09; Figure 4).

Decreased sensitivity to gaming cues

Consistent with our hypothesis, decreased gaming cue-related activations in reward circuit-related brain regions [lentiform, ventromedial prefrontal cortex (vmPFC, including orbitofrontal cortex (OFC)] were found when IGD subjects recovered from gaming. Reward circuitry may influence motivated or goal-directed behaviors and reward processing (Ikemoto, Yang, & Tan, 2015; Sayette, 2016), including in addictions (Balodis & Potenza, 2015; Cheng et al., 2016; Tobler et al., 2016; Yang et al., 2017). The reward system could be activated when individuals are exposed to relevant stimuli in substance use or gambling disorders (Balodis et al., 2012; Worhunsky, Malison, Rogers, & Potenza, 2014) as well as in IGD (Ko et al., 2009; Liu et al., 2017; Sun et al., 2012). Individuals with IGD as compared to those with regular game use have shown higher lentiform activation to gaming cues, consistent with cue reactivity and craving findings in substance use disorders (Dong et al., 2017; Dong, Wang, et al., 2018).

In this study, decreased activations were found in the lentiform nucleus and other reward-related brain regions after recovery. The findings suggest that neural response to gaming cues decreases following recovery, which is consistent with previous studies comparing IGD with controls (Kim et al., 2018; Ko et al., 2013; Ma et al., 2019). The correlations between decreases in lentiform activation and self-reported cue-elicited craving provide support to the notion that decreased neural reactivity in the lentiform may underlie decreased cue-elicited craving responses in recovery in IGD and may relate importantly to diminished motivations to engage excessively in gaming behaviors. Our previous study showed that gaming behaviors could increase IGD subjects’ craving (Dong, Wang, et al., 2018). Furthermore, we previously reported that greater lentiform activation to gaming cues was linked to emergence of IGD in individuals with regular game use (Dong, Wang, Liu, et al., 2019). This study suggests that during recovery a decrease in problematic gaming is linked to decreased cravings in IGD, with the lentiform nucleus implicated in this relationship. Taken together, the findings suggest an important role for the lentiform nucleus and cue-elicited craving in transitions between IGD and regular game use and vice versa. The precise relationships (e.g., whether decreased gaming leads to decreased lentiform responsivity and decreased craving or whether decreased lentiform responsivity leads to decreased craving and decreased gaming) require further investigation.

Control of craving after recovery

Another brain region showing group differences was the ACC, which has been implicated in executive control and other processes. In contrast to our hypothesis, activation was decreased in the ACC (as well as in the MFG) after recovery. The cluster identified included the ACC and MFG and extended ventrally to include the vmPFC and OFC. Notably, the medial prefrontal cortex has been implicated in cue-elicited craving in substance addictions like cocaine-use disorder (Kober et al., 2016; Wexler et al., 2001), processing of rewards, especially during notification or outcome phases (Knutson, Fong, Adams, Varner, & Hommer, 2001; Knutson & Greer, 2008), decision-making (Tanabe et al., 2007), default-mode processing (Harrison et al., 2017), and other processes (Li, Mai, & Liu, 2014). Given that the task employed in this study focused on cue-elicited craving, it is tempting to speculate that the relatively decreased activation observed in the cluster involving the OFC/vmPFC/ACC/MFG may relate to diminished cue reactivity, although this interpretation is less supported by data than the lentiform findings given the absence of a correlation with self-reported cravings.

Given that the ACC and other cortical brain regions have been implicated in executive or cognitive control (Rolls, 2000), including in people with addictive disorders (Filbey et al., 2008; Franklin et al., 2007; Kosten et al., 2005; Myrick et al., 2004; Wrase et al., 2002), it is possible that individuals with IGD who have recovered are demonstrating more efficient processing of control regions relative to when they were gaming problematically. To examine relationships between ACC and lentiform activities, we applied DCM and found that the connectivities were increased following recovery. According to psychophysiological interpretations of functional connectivities among these brain regions (Havlicek et al., 2015; Stephan et al., 2010), higher values in ACC–lentiform and lentiform–ACC connectivities during recovery relative to times when gaming problematically suggest that the interactions among these two brain regions are more efficient in subjects following recovery. As such, future research should examine the extent to which this reflects a mechanism for controlling cravings more efficiently, concurrent coupling of regions involved in reward processing, or craving-related motivations or other possibilities.

Importance and clinical implications

Theoretical models have proposed important roles for cortical and subcortical brain regions in Internet-use behaviors and disorders. A recent update of the I-PACE model (Brand et al., 2019) proposed behavioral and neural mechanisms related to transitions in Internet-use disorders like IGD. In this model, cue reactivity and changes in cortical-to-basal-ganglia circuitries were important components, consistent with the findings in this study. Of note, the updated I-PACE model also proposes a role for the insula (Brand et al., 2019), consistent with changes in cue-reactivity and craving findings and insular activation and connectivity in individuals with IGD receiving a craving-behavioral intervention (Zhang et al., 2016b). Furthermore, resting-state data from the same cohort suggested decreased connectivity (e.g., between the OFC and hippocampus and between the posterior cingulate and motor-related regions; Zhang et al., 2016a). As such, this study and other recent ones suggest potential neural targets for interventions (e.g., using brain modulation methods like rapid transcranial magnetic stimulation or transcranial direct current stimulation) to reduce cravings and promote recovery in IGD. Behavioral approaches that target craving and might operate through shared or distinct neural mechanisms (e.g., cognitive-behavioral and mindfulness-based therapies) should also be considered in light of the current findings, especially given the important role for behavioral therapies in the treatment of addictions and the value of understanding how specific therapies may operate at neurobiological levels.

Several limitations should be mentioned. First, we did not include healthy control subjects in this study. Although we have found that gaming history was not related to IGD severity (r = .088, p = .494) and also included gaming history as a factor in the GLM, a control group may have been helpful in understanding the data (e.g., with respect to possible test–retest effects). Second, most study subjects were male (only five females). As such, future studies should examine the extent to which the findings may apply to female populations, especially as gender-related differences have been observed in neural correlates in IGD populations (Dong, Wang, et al., 2018; Dong, Wang, Wang, et al., 2019; Dong, Zheng, et al., 2018). Third, although we performed a DCM analysis that suggests that executive control over lentiform activation may improve with recovery, we cannot exclude other possible explanations that should be investigated directly in future studies.

Conclusions