Screen-based digital media use is widespread in childhood and adolescence, but its association with anxiety and dose–response relationship remain uncertain. We systematically reviewed observational studies of participants aged 6–19 years reporting time- or frequency-based screen exposure and anxiety-specific outcomes. PubMed/MEDLINE, Web of Science Core Collection, Scopus, and CNKI were searched through September 2026. Study quality was assessed using the AHRQ checklist or Newcastle–Ottawa Scale. Eligible estimates were synthesized using three-level random-effects meta-analysis and restricted cubic splines. Fifty-seven reports were retained. The primary adjusted dataset included 79 odds-ratio effects from 17 independent studies (pooled OR 1.40, 95% CI 1.26–1.54; I² = 51.7%). Screen type significantly moderated the association after Holm correction. The dose–response analysis included seven independent studies; the linear component was significant (p = 0.039), whereas nonlinearity was not supported (p = 0.132; wild-cluster p = 0.163). Evidence was predominantly observational and cross-sectional, and dose–response data were limited. Greater screen exposure was associated with higher odds of anxiety, supporting a graded pattern without a discrete threshold. Because the evidence was predominantly observational and cross-sectional, this exposure–response gradient should be interpreted as an association rather than evidence that greater screen use causes anxiety; reverse and reciprocal pathways remain possible. This review was not registered and was funded by two Jiangsu Provincial Education Science Planning projects.

Gameful digital mental health interventions are increasingly used to support university students, but evidence for the effectiveness of complete intervention packages should be distinguished from evidence that specific game mechanics independently improve mental health outcomes. This systematic review and meta-analysis evaluated the short-term package effectiveness of gameful digital mental health interventions for depression, anxiety, and stress/distress among university students, while separately examining active-comparator, component-level, and mechanism-oriented evidence to distinguish package effects from the incremental effects of gameful features and other intervention processes. We searched PubMed, Web of Science Core Collection, Scopus, CINAHL, CNKI, Embase, and APA PsycINFO (EBSCOhost) from database inception through 3 September 2026. Two reviewers independently screened titles/abstracts and full texts, with disagreements adjudicated by a third reviewer. Risk of bias was assessed with RoB 2, and outcome certainty was evaluated with GRADE. Randomized package-effect trials were synthesized using outcome-specific random-effects models, and an integrative three-level model retained multiple symptom effects within studies. Dependence was examined using study-cluster robust inference and sensitivity analyses across plausible within-study sampling correlations. Component- and mechanism-oriented studies were synthesized separately from package-effect trials. Database searches identified 4267 records. Twenty-three reports were retained through the database pathway, and backward citation chasing identified 1 additional report, yielding 24 retained reports overall; 3 reports were not retrieved. Compared with inactive controls, gameful interventions showed moderate short-term effects for depression (4 studies; Hedges’ g = 0.52, 95% CI 0.32–0.72) and anxiety (3 studies; g = 0.51, 95% CI 0.22–0.80), whereas stress/distress was highly imprecise (2 studies; g = 0.19, 95% CI −1.68 to 2.06). The integrative synthesis included nine symptom effects from five independent studies and yielded g = 0.410 (95% CI 0.245–0.575; study-cluster robust 95% CI 0.182–0.638), with a 95% prediction interval of −0.013 to 0.834, which includes no benefit and a small harmful effect in a future setting. Active-control effects were smaller in the full moderator model but were not leave-one-study-out stable. Expanded component and mechanism-oriented evidence did not show a consistent independent symptom benefit attributable to gameful features, although some studies reported differences in engagement, adherence, enjoyment, or related intervention processes. Gameful digital mental health intervention packages may reduce short-term depression and anxiety symptoms among university students, but certainty is low for depression and anxiety and very low for stress/distress; only five independent studies informed the integrative model. Generalizability across interventions and settings remains uncertain. Evidence against active comparators is sparse and unstable. Component and mechanism-oriented evidence indicates that gameful or contextual design choices may influence engagement, adherence, enjoyment, or training processes, but does not support a consistent independent symptom benefit from adding game mechanics. Future research should use meaningful active comparators and factorial or dismantling designs to identify which gameful features support therapeutic mechanisms, engagement, or potentially interfere with treatment processes.