No patient or public involvement was incorporated into the design, conduct, reporting, or dissemination of this research. This was a prospective, randomized, controlled, single-blind trial with two parallel groups. The research was conducted at the Department of Psychology of Yunnan Normal University in Kunming, China, from November 1, 2024, to December 1, 2024. The study design and reporting adhered to the CONSORT 2025 guidelines, specifically following the CONSORT checklist for reporting information in randomized trials (Checklist 1) [43]. The trial proceeded to its planned completion without early termination.

The study protocol was reviewed and approved by the Human Investigation Committee (Institutional Review Board) at Yunnan Normal University (Approval No: yunnuethic2024-044; registered prospectively on 2024-10-14). No changes were made to the trial protocol after its commencement. Written informed consent was obtained from all participants prior to their enrollment in the study. To ensure participant privacy and confidentiality, all collected data were immediately deidentified following collection. Participants were assigned anonymous codes, and all analyses were performed using these coded identifiers. No personally identifiable information is included in this manuscript or any supplementary materials. All participants who completed the entire study protocol, including pretest and posttest assessments and all training sessions, received financial compensation of 400 RMB (approximately US $56.29).

Figure 1 shows the process and sequence of the study. This study initially recruited 72 participants with IGD. Eight participants dropped out during the intervention training, resulting in a final cohort of 64 participants (PAT: n=33; FT: n=31). No missing data were encountered (no additional prespecified or post-hoc analyses were performed).

We recruit patients with IGD from schools through advertisements to obtain treatment opportunities for improving mental health. Researchers require interested participants to participate in the internet addiction test (IAT) online [44], which has been proven to effectively diagnose internet addiction. Participants with scores greater than 50 are required to undergo an offline professional interview with a psychiatrist. Select eligible research subjects using DSM-5 criteria [45], and Mini International Neuropsychiatric Interview (MINI) [46]. 

The criteria for inclusion in participants are as follows: (1) IAT score >50; (2) DSM-5 score >5; (3) active engagement in online gaming exceeding 1 year; (4) BMI within normal to overweight range (15≤BMI≤28); (5) no cognitive impairment and no other psychiatric disorders confirmed by MINI; (6) no previous treatment for IGD; (7) no experience of regular use of any psychotropic drugs. The exclusion criteria for participants were as follows: (1) majors related to sports; (2) DSM-5 score <5; (3) the body contains metals. All participants who met the inclusion criteria were randomly assigned to either the PAT group or the FT group. Detailed demographic data for the two groups are shown in Table 1.

In this study, the American College of Sports Medicine guidelines were used to develop an aerobic training program with an aerobic intensity of 46% in the first phase and 55% in the second phase, as well as a training duration of 40 minutes per session for a total of 20 sessions in 1 month [47-49]. The PAT group trained only on the treadmill at the speed calculated from the step test [50]. The specific training process and related contents are shown in Multimedia Appendix 1. The interventions were delivered by research assistants who were trained in the specific protocols for both the PAT and FT groups to ensure consistency and safety.

As adverse events were not systematically assessed, none were reported during the trial.

A comparison between study completers and dropouts showed no significant differences in key demographic or clinical measures at baseline (Multimedia Appendix 1), indicating that attrition bias is unlikely to have substantially influenced the primary findings. This trial involved no interim analyses or stopping guidelines.

Using G-power, setting the effect size at f=0.25 and α error probability at 0.05, we estimated the required sample size at 66. The current sample is a bit lower than the required 66.

Participants were randomly allocated to either the PAT group or the FT group using simple randomisation with a 1:1 allocation ratio. The random allocation sequence was generated by an independent researcher not involved in participant recruitment or intervention. The sequence was concealed using sequentially numbered, opaque, sealed envelopes. The principal investigator enrolled participants, and a research assistant then assigned them to groups by opening the next envelope in the sequence.

The participants were indeed blinded to the existence of the other experimental group. Both groups were in the same training room but did not know the existence of the other group (the PAT group trained from 14:00-19:00; the FT group trained from 20:00-22:00). Research assistants conducting assessments were aware of group allocations due to the practical constraints of implementing the intervention protocols.

The current analysis is a per-protocol analysis; only subjects who finished the pretest, 20-time training, and posttest were included. Before the first training and after the entire training period, all participants’ addiction levels were measured based on the IAT, DSM-5 criteria, and craving scores. Craving was measured with the Questionnaire for Gaming Urges, adapted from the Tiffany Questionnaire for Smoking Urges [51]. Meanwhile, before the first training session, after 10 interventions and all 20 interventions, we conducted a 3-minute step test [50] on all participants to obtain the maximum oxygen uptake for dynamically adjusting the intensity of PAT.

The magnetic resonance imaging data were acquired using a Siemens Trio 3T scanner. The specific parameters used were as follows: repetition time=2000 ms, 54 interleaved slices, echo time=30 ms, thickness=3.0 mm, flip angle=90°, field of view=220 mm × 220 mm, and matrix=64×64. Head motions were minimized by filling the empty space around the subjects’ heads with sponge and fixing their lower jaws with tape.

The preprocessing of the rs-fMRI data was completed according to Statistical Parametric Mapping principles [52] using SPM12 [53], and data processing and analysis were performed using Data Processing & Analysis of Brain Imaging (DPABI) [54,55]. The preprocessing of the functional rs-fMRI data included the following steps: (1) the quality of the raw rs-fMRI data for all the participants was examined; (2) the raw data were transformed from the Digital Imaging and Communication in Medicine (DICOM) format to the Neuroimaging Informatics Technology Initiative (NIFTI) format; (3) the initial 10 volumes were discarded to maintain magnetic field stabilization; (4) slice timing was performed; (5) realignment was performed (excluding subjects with maximum head motion >2.5 mm or rotation >2.5°); (6) regressing nuisance covariates were included, including the signal noise of white matter, cerebrospinal fluid, and Friston-24 head motion parameters [56]; (7) the data were normalized using the standard procedure within the DPABI pipeline, which registers the images to the MNI-152 template; (8) the data were filtered by a bandpass filter with a frequency range of 0.01 to 0.1 Hz before the calculation of ReHo and DC; and (9) the data were smoothed using a Gaussian kernel with a full width at half maximum of 6 mm.

We first computed ReHo and DC values for both the PAT and FT groups using pretest and posttest measurements. Repeated-measure ANOVAs were then conducted for two groups using the DPABI Viewer toolbox, with a statistical threshold set at P<.01 and a cluster size threshold of >100 voxels (corrected for the false discovery rate).

The prespecified primary outcome was the change in the Questionnaire for Gaming Urges score from baseline to post-intervention. Secondary outcomes included other behavioral measures and functional connectivity measures (including their correlation with normative neurotransmitter maps).

We conducted whole-brain FC analysis using the ROI obtained by overlapping DC and ReHo. Then, significant brain region data were extracted with the DPABI toolbox for repeated measures ANOVA analysis in JASP (Version 0.19.3) [57,58], applying a Bonferroni correction (α=.05, corrected for multiple comparisons). The correlation analysis of behavioral data was also completed using JASP. Subsequently, we performed a correlation analysis between FC results and neurotransmitter levels using JuSpace 1.5 (10,000 permutations, P<.05, with correction for the false discovery rate) [59]. The analysis demonstrates a spatial correspondence between our FC findings and the normative neurotransmitter maps, rather than a correlation with measured individual neurochemical levels.

The following analysis was conducted with 64 participants who completed the interventions without receiving other therapies, as shown in Figure 2. Consistent with the methods, no intervention-related adverse events were observed, and no additional prespecified or post-hoc analyses were performed.

We performed repeated measures ANOVA on game craving in the PAT and FT groups and found that the interaction was not significant. Then, paired-samples t tests were conducted to compare pretest and posttest gaming cravings in the PAT group and the FT group. The results revealed a significant reduction in gaming cravings in the PAT group (t₃₂=2.278, P_bonf=0.03, Cohen d=0.397, 95% CI 0.045-0.851), whereas no significant change was observed in the FT group (t₃₀=0.481, P_bonf=1, Cohen d=0.086, 95% CI −0.306 to 0.502). Compared to the FT group (t₃₀=2.277, P_bonf=0.03, Cohen d=0.409, 95% CI 0.039-0.773), the PAT group showed a significantly greater reduction in the IAT score (t₃₂=4.333, P_bonf<0.001, Cohen d=0.754, 95% CI 0.362-1.137). This finding was consistent with DSM-5 scores (Figure 3).

In this study, we evaluated the efficacy and neural mechanisms of the PAT intervention for IGD. The results indicate that PAT was associated with reduced severity of addiction and craving levels, alongside alterations in the FC within the ECN and between the ECN and the RN.

Following the PAT intervention, participants in the PAT group exhibited markedly lower gaming cravings and decreased addiction severity relative to the AT group. Although PAT has not been used in previous literature for the treatment of IGD, previous studies on alcohol and nicotine dependence have shown that PAT training can reduce cravings for addictive substances and the severity of addiction [60,61]. Therefore, this study found decreased game craving and addiction severity after PAT, indicating that PAT is an effective intervention for patients with IGD, significantly mitigating the symptoms of IGD.

This study found enhanced FC between the prefrontal cortex (mSFG and IFG), SPG, middle cingulate gyrus, and postcentral and precentral gyrus in IGD after PAT. The prefrontal cortex and postparietal cortex are implicated in a wide range of cognitive functions, particularly cognitive control [62,63]. The SPG has a crucial role in visual perception, spatial cognition, and attention [64,65]. Resting-state fMRI has found that compared to healthy controls, patients with IGD showed decreased FC in the ECN (including several parts of the frontoparietal cortex) [66,67], indicating impaired executive functioning. The findings of this study demonstrate that PAT contributed to enhancing the functional integrity of ECN implicated in IGD. The precentral gyrus and postcentral gyrus are the key regions of sensorimotor networks associated with integrating sensorimotor information and coordinating physical movement [68,69]. Although the precentral gyrus is not part of the ECN, executive control processes frequently require integrating sensory information transmitted from the precentral gyrus to achieve efficient behavioral regulation [70]. Therefore, this study found that the FC within the ECN was significantly enhanced following PAT, indicating that PAT may help normalize sensorimotor dysfunction of IGD, thereby contributing to the restoration of executive control and emotional regulation capabilities.

The present study found enhanced FC between the mSFG, bilateral insula, and substantia nigra. The insula plays a pivotal role in the RN, engaging in reward processing and the regulation of motivated behaviors by integrating sensory and emotional information. It influences decision-making and is closely associated with mental health [71,72]. In the context of addiction, insula dysfunction may contribute to heightened cravings and impulsive behaviors, disrupting the balance between impulse control and reward processing [73]. The substantia nigra, a core component of the basal ganglia, plays a pivotal role in reward-related processing [74]. Meanwhile, the superior frontal gyrus, a key region within the ECN and a default mode network seed region, is critical for top-down cognitive control, including decision-making, selective attention, inhibitory control, and emotion regulation [75,76]. Consistent with prior research, individuals with IGD exhibit reduced FC between the ECN and RN compared to healthy controls [77-79]. Therefore, this study revealed that the FC between the ECN and RN was significantly enhanced following PAT, indicating that PAT may weaken gaming cue reactivity and strengthen top-down regulation of reward processing in patients with IGD.

Few studies have correlated the neuroimaging changes of IGD with the activity of microscopic neurotransmitters [80,81]. This study found an association between the FC and activity maps of CB1 and mGluR5 in the PAT group. CB1 cannabinoid receptors are located presynaptically in multiple brain regions, including reward-related structures like the caudate-putamen, substantia nigra, and globus pallidus [82,83]. It is well-established that aerobic exercise is a potent trigger for the release of endogenous cannabinoids, such as anandamide, in the brain [84]. The primary central receptor for these endogenous cannabinoids is the CB1 receptor. Therefore, our finding that functional connectivity changes correlate with the spatial distribution of CB1 receptors provides a direct neurobiological substrate for the effects of PAT. We posit that the PAT in our study enhanced the signaling of endogenous cannabinoids, which in turn modulated neural circuitry in CB1-rich brain regions that are critically involved in IGD pathology.

Furthermore, the significant correlation with the mGluR5 map is highly informative. The mGluR5 is highly expressed in key addiction-related regions, including the nucleus accumbens, cerebral cortex, hippocampal formation, and amygdala [28]. Both CB1 and mGluR5 are considered to exhibit functional interactions and are known to form heteromeric complexes that regulate synaptic plasticity [85,86]. This provides a compelling rationale for our concurrent findings regarding both receptors. We thus hypothesize that the therapeutic action of PAT is mediated not by a single system, but by the coordinated modulation of CB1 and mGluR5 signaling, which promotes adaptive synaptic plasticity to normalize the dysregulated neural circuitry underlying IGD.

This study has several limitations. First, while our FC analysis revealed altered connectivity between key brain regions, it did not elucidate the directionality of these changes. Future studies could employ effective connectivity analysis to determine the causal interactions underlying these FC modifications. Second, the FT group exhibited a slight therapeutic effect on IGD symptoms, suggesting a potential placebo effect. To address this limitation, future research could incorporate multiple control groups to better isolate the specific effects of PAT. Third, our sample was recruited solely from schools, which may limit the generalizability of the findings to older individuals with IGD. Fourth, it should be noted that potential confounding factors, such as general fitness improvement, social interaction during training, and participant expectancy effects, may have contributed to the observed outcomes. Finally, the modest sample size in this study may limit the generalizability of the findings. Future research with a larger, multicenter cohort, incorporating long-term follow-ups, would enhance the statistical power and robustness of the findings.

This study provides evidence for the beneficial effects of PAT on IGD symptoms. A key innovation of this trial is its implementation of a progressive aerobic intervention, which dynamically adjusted intensity based on the real-time performance of individuals with IGD. This approach moves beyond conventional fixed-intensity protocols to offer a more personalized and potentially more effective therapeutic approach. At the cognitive-neural level, our findings suggest that PAT may enhance executive control function in patients with IGD concurrently with alterations in its regulatory influence over the reward system. These neural improvements were substantiated by significant correlations between neurofunctional markers and behavioral measures. The results offer insights into the neural mechanisms associated with PAT’s positive outcomes in IGD and provide a foundation for future clinical intervention research.