This study was a 2-arm, parallel-group randomized controlled trial. The trial was retrospectively registered with the Pan African Clinical Trials Registry under the identification number PACTR202601838702413 on January 30, 2026. The registration was conducted retrospectively due to an administrative oversight during the initial recruitment phase as the research team prioritized data collection commencement following funding approval. This study was conducted between October 2024 and February 2025 across 3 public middle schools in the Sfax region of Tunisia during routine health screenings.

A total of 960 adolescents (aged 14-16 years) were initially recruited during routine health screenings across 3 public middle schools in the Sfax region. A total of 151 students with PSU were identified. The inclusion criteria were (1) meeting the clinical cutoff for smartphone addiction and (2) reporting structured PA not exceeding 3000 metabolic equivalent of task minutes per week as assessed using the Arabic-language International Physical Activity Questionnaire [20]. The exclusion criteria were (1) BMI below the 3rd or above the 97th percentile [21] and (2) habitual sleep duration outside the range of 6 to 10 hours per night [22]. This resulted in the exclusion of 76 students (n=37, 48.7% due to excessive activity; n=39, 51.3% due to the BMI and sleep criteria).

Eligible participants (n=75) were randomly assigned to the experimental group (EG; n=36) or the CG (n=33) using stratified block randomization. Randomization was stratified by sex (male and female), with a combination of random block sizes of 4, 6, and 8 to ensure balanced allocation across groups throughout the recruitment period. The allocation sequence was obtained using computer-generated random numbers by an independent researcher who was not involved in participant recruitment, intervention delivery, or outcome assessment.

Allocation concealment was ensured using sequentially numbered opaque, sealed envelopes prepared by the independent researcher and opened only after completion of baseline assessments and confirmation of eligibility. Participants were enrolled by the lead author, and group assignment was implemented by a research assistant not involved in outcome assessment.

Due to the nature of the behavioral intervention, participants and intervention providers could not be blinded. However, outcome assessors and data analysts were blinded to group allocation throughout the trial.

During the trial, 8% (6/75) of the participants were excluded due to extended absences (>1 week), incomplete assessments, or newly emerging clinical conditions (EG: n=4; CG: n=2), resulting in a final analytic sample of 69 participants (EG: n=36, 52.2%; CG: n=33, 47.8%). The Little test was nonsignificant (χ²₁₈=12.5; P=.82), supporting the assumption that data were missing completely at random. Given the low rate of missingness and the missing completely at random assumption, we conducted complete-case analyses without multiple imputation. Baseline comparisons confirmed equivalence between groups on all demographic and outcome variables (P>.05 in all cases). A detailed flowchart of participant screening, randomization, allocation, and follow-up is presented in Figure 1.

An a priori power analysis was conducted using G*Power (version 3.1) [34] for a repeated-measure ANOVA with a within-between interaction. To detect a medium effect size (η_p²=0.06, equivalent to f=0.25) for the primary group × time interaction, with α of .05 and 1 − β of 0.80, a total of 64 participants were required; the final sample comprised 69 adolescents. The choice of a medium effect size is consistent with prior intervention research on adolescent digital addiction and related school-based programs, which reports moderate intervention effects on problematic digital technology use outcomes [35].

Data were analyzed using SPSS Statistics (version 28.0; IBM Corp). Baseline group differences were examined using independent 2-tailed t tests for continuous variables and chi-square tests for categorical variables. Primary intervention effects were evaluated using linear mixed-effects models, which included group, time, and the group × time interaction as fixed effects, with participants as a random effect. These models were adjusted for the covariates of age, sex, and BMI and tested for each outcome. Exploratory within-group pretest-posttest changes were assessed using paired t tests, and between-group differences at the posttest time point were examined using independent t tests. Analyses of covariance were conducted to compare adjusted postintervention outcomes between groups controlling for baseline scores as well as age, sex, and BMI. Effect sizes were calculated using η_p² and interpreted according to conventional benchmarks [36]. Mediation analyses were performed using bootstrapping procedures with 5000 resamples to estimate indirect effects and 95% CIs [37]. Statistical significance was set at an α value of .05 (2 tailed).

This study received ethics approval from the local Committee for the Protection of Persons at the Faculty of Medicine of University of Sfax, Tunisia (approval 0554/2023). The research was conducted in full compliance with the ethical standards outlined in the Declaration of Helsinki (2013). Written informed consent was obtained from participants’ legal guardians, and verbal assent was obtained from each adolescent prior to participation. All data, including smartphone use records, questionnaire responses, and behavioral observations, were anonymized. Any potentially identifiable information was stored separately on a password-protected server accessible only to the principal investigator. No financial or material compensation was provided to participants.

A detailed flowchart of participant screening, randomization, allocation, and follow-up is presented in Figure 1. At baseline, no statistically significant differences were observed between the EG (n=36) and CG (n=33) across demographic, anthropometric, or psychological variables (Table 1). The groups were comparable in age, sex distribution, BMI, smartphone addiction, nomophobia, psychological distress (global and subscale scores), and loneliness (P>.05 in all cases). Effect size estimates were negligible, indicating satisfactory baseline equivalence prior to the intervention.

Linear mixed model analyses revealed statistically significant group × time interaction effects across all primary outcomes (Table 2), consistently favoring the EG. Interaction effects were of moderate magnitude, including smartphone addiction (η_p²=0.12; P=.004), nomophobia (η_p²=0.10; P=.007), and global psychological distress (η_p²=0.10; P=.009). Similar moderate interaction effects were observed for depression (η_p²=0.09), anxiety (η_p²=0.08), stress (η_p²=0.08), and loneliness (η_p²=0.08), indicating a consistent intervention-related improvement across behavioral and psychosocial domains.

Figure 2 illustrates the distribution of pre- and postintervention scores across outcome measures for the EG and CG. Visual inspection indicates clear reductions from before to after the intervention in the EG across smartphone addiction, nomophobia, loneliness, and psychological distress (global score and subscales), whereas score distributions in the CG remained largely stable over time. Exploratory within-group comparisons showed statistically significant pretest-posttest reductions in the EG across all outcomes (P<.05 in all cases), whereas no significant changes were observed in the CG (P>.05 in all cases). Between-group comparisons at the posttest time point favored the EG across all measures (P≤.02).

After controlling for baseline scores, analyses of covariance confirmed significant postintervention group differences across all outcomes (Table 3). The EG demonstrated significantly lower adjusted posttest scores than the CG for smartphone addiction (η_p²=0.12; P=.004), nomophobia (η_p²=0.11; P=.006), global psychological distress (η_p²=0.12; P=.005), depression (η_p²=0.10; P=.008), anxiety (η_p²=0.09; P=.01), stress (η_p²=0.09; P=.02), and loneliness (η_p²=0.09; P=.01). All effects were in the moderate range, supporting the robustness of the intervention after adjustment for baseline variability.

Mediation analysis showed that reductions in loneliness partially mediated the effect of the intervention on postintervention smartphone addiction (Table 4 and Figure 3). The indirect effect was statistically significant (ab=−1.10, 95% CI −2.05 to −0.35), accounting for 34.4% of the total effect. The intervention significantly reduced loneliness (a=−1.80; P=.002), and higher loneliness was associated with greater smartphone addiction (b=0.61; P<.001). The direct effect of the intervention remained significant after accounting for loneliness, indicating partial mediation. No adverse events or unintended effects were reported in either group during the trial period.

This randomized controlled trial evaluated a 12-week, school-based intervention integrating traditional social sports games with mental exercises for Tunisian adolescents with elevated smartphone addiction. Linear mixed model analyses revealed significant group × time interaction effects of moderate magnitude (η_p²=0.08-0.12), consistently favoring the EG over the CG. Participants receiving the TSSG-ME intervention demonstrated statistically significantly greater reductions in smartphone addiction, nomophobia, psychological distress (global and subscale scores), and loneliness. After adjusting for baseline scores, between-group differences at the posttest time point remained significant with moderate effect sizes. Furthermore, mediation analysis indicated that reductions in loneliness accounted for approximately 34.4% of the intervention’s total effect on smartphone addiction, supporting a partial mediation model.

Reductions in smartphone addiction were among the most robust outcomes. These findings are consistent with those of prior studies showing that structured physical engagement and psychological support can reduce behavioral dependence on mobile technologies [38]. This aligns with meta-analytic findings that combined interventions integrating PA with psychological skill building produce more consistent reductions in smartphone addiction than single-component approaches [11,14]. This study contributes by demonstrating the efficacy of a culturally grounded program in a Tunisian context.

The term “traditional social sports games” denotes structured, rule-based PAs with inherent social organization and deep cultural resonance, distinguishing them from generic physical exercise [18,19]. The mental exercise component refers to a standardized protocol of evidence-based techniques adapted from sports psychology designed to build self-regulation and stress tolerance beyond simple relaxation [26,27]. The observed moderate effect sizes were clinically meaningful and consistent with those reported in other behavioral intervention trials targeting PSU [39]. They reflect a realistic and attainable impact, particularly given that our sample comprised adolescents with clinically elevated baseline symptoms who may have had greater room for improvement.

The reduction in nomophobia underscores the intervention’s potential to alleviate the anxiety associated with smartphone separation. By fostering rewarding, in-person social experiences and peer validation, the traditional games likely diminished the perceived indispensability of constant digital connectivity [6]. Concurrently, the mental exercise component equipped participants with concrete skills for managing emotional arousal, offering a healthier alternative to digital coping mechanisms [40].

Critically, our findings provide empirical support for the psychological mechanism proposed by SDT [15]. Reductions in loneliness mediated 34.4% of the intervention’s total effect on smartphone addiction, highlighting a key pathway. This suggests that the TSSG-ME program enhanced the basic psychological need for relatedness through prosocial, face-to-face interaction in traditional games, reducing perceived social isolation. Consequently, adolescents’ compensatory reliance on smartphones for social fulfillment likely decreased, alleviating addiction symptoms. This mechanism aligns with research identifying loneliness as a critical link between unmet social needs and PSU [41,42] and supports evidence that loneliness mediates the relationship between health-promoting behaviors and reduced maladaptive digital use [13,43]. Using culturally embedded games to foster relatedness also aligns with literature emphasizing their effectiveness in promoting sociability [44]. A key limitation is that mediation was tested with concurrent postintervention measures; temporal precedence cannot be fully established, so causal interpretation should be made cautiously.

The dual-component design posits complementary pathways: the traditional social sports games primarily targeted relatedness through social engagement and shared PA, whereas the mental exercise component targeted competence and autonomy through enhanced self-regulation [45]. Their integration likely addressed the multifaceted social-motivational and emotional-cognitive vulnerabilities underpinning smartphone addiction more comprehensively than either component alone [46]. Future research using factorial designs is necessary to disentangle their independent and synergistic contributions.

Several limitations should be acknowledged. First, the study relied on self-reported measures, which may be subject to response bias; future research should incorporate objective metrics such as device-based screen time tracking. Second, the culturally specific sample, while emphasizing the importance of contextual adaptation, may limit the generalizability of the findings to broader populations. Third, the absence of long-term follow-up assessments precludes conclusions regarding the sustainability of the intervention effects. Fourth, the use of a passive CG (standard physical education) rather than an active, time-matched comparator (eg, generic group sports or relaxation program) restricts the ability to attribute observed outcomes specifically to the traditional social games and mental training components as opposed to nonspecific factors such as increased social engagement or novelty. Finally, although effect sizes were statistically significant and clinically meaningful, they were moderate, highlighting the complex, multifactorial nature of PSU and suggesting the need for further optimization of the intervention.

This study provides evidence that a culturally adapted, school-based intervention combining traditional social sports games and mental training can yield significant, moderate improvements in smartphone addiction, nomophobia, psychological distress, and loneliness among adolescents. The mediation of the treatment effect through reduced loneliness highlights the importance of enhancing real-world social connectedness, supporting a key mechanism derived from SDT. By providing alternative, healthier avenues for satisfying psychological needs for relatedness and competence, programs such as TSSG-ME may reduce reliance on maladaptive smartphone use.

Beyond individual outcomes, these findings suggest that integrating culturally relevant social activities into school-based programs may serve as a scalable strategy to support adolescent mental health in increasingly digital contexts. Policymakers and educators could leverage such approaches to mitigate the negative consequences of excessive smartphone use. Future research should use longitudinal designs, active CGs, and diverse populations to confirm causal pathways and assess long-term efficacy. However, given that mediation was examined cross-sectionally, causal interpretations should be made cautiously.