The review examined 11 outcomes, including perceived or general stress, anxiety, depressive symptoms, fatigue, resting systolic and diastolic blood pressure, resting heart rate, blood cortisol and three heart rate variability indices. The interventions included aerobic exercise, high-intensity interval training, resistance training, sport-based training, yoga, qigong and multicomponent school programs, lasting from three weeks to 10 months. Comparators included usual physical education, usual activity, health or attention education, no-intervention and wait-list conditions. The studies covered 18 countries or regions. Outcome-specific analyses contained 4–15 studies and 90–3,157 participants, so each pooled result rested on a different and often limited body of evidence.

For perceived or general stress, seven studies produced a pooled standardized mean difference of Hedges’ g = −0.28, with a 95% confidence interval of −0.48 to −0.08 and p = 0.006. Negative values represented lower post-intervention stress scores in intervention groups. Heterogeneity was moderate, with I² = 49.9% and τ² = 0.030. The estimate remained statistically significant under an adjusted Knapp–Hartung method and when non-randomized studies were excluded, but it became non-significant when Ranjani and colleagues’ study was removed. The analysis therefore supports a possible benefit rather than demonstrating a robust, generalizable effect. This estimate was one of several statistically significant psychological results, and the review did not establish a clear hierarchy among them.

Physical activity interventions were also associated with lower depressive symptom scores in a synthesis of 13 studies. The primary pooled estimate was g = −0.53, with a 95% confidence interval of −1.04 to −0.02 and p = 0.041. Most individual estimates favored intervention groups, but their magnitude varied widely: I² reached 94.8% and τ² was 0.809, indicating substantial variation in the studies’ estimated effects. The result was statistically fragile. Its confidence interval crossed zero under modified Knapp–Hartung inference, it became non-significant after several individual studies were omitted, and it also depended partly on the inclusion of non-randomized evidence. Depressive symptoms are not a direct measure of stress and can have other determinants, so the finding cannot by itself establish improved stress regulation.

No statistically significant pooled effect was found for anxiety or fatigue. For anxiety, 15 studies yielded g = −0.34, but the 95% confidence interval ranged from −0.79 to 0.11, with p = 0.137 and very high heterogeneity of I² = 93.7%. The point estimate favored lower anxiety, yet the data did not show a clear overall difference. Fatigue was assessed in only four studies; its pooled estimate was close to no effect at g = −0.11, with a 95% confidence interval of −0.32 to 0.10. An estimated I² of zero did not establish that the studies were genuinely alike, because the number of contributing studies was too small to characterize heterogeneity reliably. These non-significant results indicate that the available evidence is insufficient to confirm an effect, not that exercise has been shown to have no effect.

On the physiological side, one pooled signal was a small reduction in resting systolic blood pressure. Ten studies produced g = −0.15, with a 95% confidence interval of −0.26 to −0.04 and p = 0.008. The direction remained negative when individual studies were removed, although some confidence intervals then crossed zero. The estimate also survived the modified Hartung–Knapp inference used in sensitivity analysis and restriction to randomized studies. This outcome received the review’s highest certainty rating, but that rating was still low because of risk-of-bias concerns and imprecision. The confidence interval included effects smaller and larger than the review’s 0.20-standard-deviation benchmark. A lower resting blood pressure may reflect a general cardiovascular adaptation to training rather than a changed response to psychological stress.

Resting diastolic blood pressure showed a larger pooled association, with g = −0.33, a 95% confidence interval of −0.64 to −0.01 and p = 0.041 across 10 studies. However, heterogeneity was substantial at I² = 90.5%, and statistical significance disappeared under modified Hartung–Knapp inference, after exclusion of non-randomized studies, and in leave-one-out analyses that removed several individual studies. An Egger regression also indicated possible small-study effects for this outcome, making publication bias a plausible explanation for some of the observed asymmetry. No significant changes were found in resting heart rate, blood cortisol, or two heart rate variability measures, RMSSD and SDNN. Blood cortisol, for example, had g = −0.07 and a 95% confidence interval of −0.80 to 0.67.

Heart rate variability was assessed through three separate indices. Across four studies, the LF/HF ratio yielded g = −0.43, with a 95% confidence interval of −0.80 to −0.05 and p = 0.028, indicating a lower ratio in intervention groups. Yet the finding was fragile: statistical significance was lost when any one of the four studies was omitted, when the analysis was restricted to randomized studies, and under modified Hartung–Knapp inference. LF/HF is influenced by both sympathetic and parasympathetic activity, and its traditional use as a direct index of “sympathovagal balance” has been increasingly questioned. The direction of change therefore cannot confidently be translated into better or worse autonomic regulation. Estimates for RMSSD, g = 0.03, and SDNN, g = 0.01, were both close to no effect, with confidence intervals that crossed zero.

The evidence base also had important methodological weaknesses. None of the randomized analytical entries was judged to have an overall low risk of bias: after consolidating entries from the same report, seven of 32 randomized studies were classified as high risk and 25 raised concerns. Common problems included insufficient detail about allocation concealment, intervention deviations, analysis populations and prespecification of results. All three non-randomized studies had serious overall risk of bias, including serious confounding. Psychological outcomes were commonly self-reported and participants were generally unblinded, creating the possibility that expectations could inflate apparent benefits. Missing data, baseline imbalances, selective reporting and inadequate information about intervention dose further limited confidence in the estimates. The analyses used unadjusted post-intervention comparisons, so residual differences between groups could influence the pooled results.

These limitations matter because psychological symptoms and resting physiological measurements are related to, but not specific to, stress. Cortisol varies with sampling time, sleep, acute stress and circadian rhythms; resting heart rate changes with age; and blood pressure and heart rate variability describe cardiovascular or autonomic state rather than reactions to a defined stressor. To measure stress reactivity and recovery, measurements would need to be taken during and after a standardized challenge. The review therefore cannot establish that physical activity produces a unified improvement in stress regulation, that subjective benefits are accompanied by physiological changes, or that the reverse relationship holds. Differences in age, activity type, intensity, frequency, session duration, intervention length, psychological instruments and measurement conditions may also explain why pooled effects varied, especially for anxiety and depressive symptoms.

Overall, the analysis provides tentative evidence that structured movement may help some young people feel less stressed and may modestly lower resting blood pressure. It does not establish a reliable effect on anxiety, fatigue, cortisol, resting heart rate, RMSSD or SDNN, and the apparent reductions in depressive symptoms, diastolic blood pressure and LF/HF depend on analytical choices or individual studies. Larger, well-designed randomized trials with clearly specified activity doses, standardized psychological and physiological measures, adequate follow-up and assessment during and after stressors are needed to determine which programs benefit which populations. Until then, the evidence supports physical activity as a plausible support for youth stress management, but it does not yet demonstrate a broad or uniform effect across psychological and physiological outcomes.

Subject of Research: Effects of physical activity interventions on stress-related psychological and physiological outcomes in children and adolescents

Article Title: The effects of physical activity interventions on stress-related psychological and physiological outcomes in children and adolescents: a systematic review and meta-analysis

Article References: Chen, L., Zhang, J., Wang, B., & Gao, Z. (2026). The effects of physical activity interventions on stress-related psychological and physiological outcomes in children and adolescents: a systematic review and meta-analysis. Frontiers in Psychology, 17, Article 2000159. https://doi.org/10.3389/fpsyg.2026.2000159

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.2000159

Keywords: physical activity, children, adolescents, stress, exercise, depressive symptoms, blood pressure, heart rate variability, cortisol, systematic review, meta-analysis, effects