Effects of physical activity on executive function in children and adolescents: A Bayesian dose-response network meta-analysis

RCTs identified across PubMed, Web of Science, Embase, Scopus and Cochrane to April 2025 (study count not stated in abstract)

Children and adolescents

Dose-response model-based network meta-analysis with Bayesian estimation, comparing exercise modalities (aerobic, skill/coordination, HIIT, cognitively engaging PA) and estimating dose gradients rather than binary exposure. This is the best available attempt at an actual dose-response function in this literature. Graded C: it pools mixed-quality RCTs, reports no selection-model publication-bias adjustment, and its dose curves are fitted across studies (an ecological contrast) rather than within randomized dose arms.

The dose question has an answer and the answer is small. Overall effect on executive function is g=0.14 (95% CrI 0.08 to 0.19). By modality: aerobic exercise g=0.22 (0.13 to 0.32), skill/coordination g=0.17 (0.07 to 0.26); HIIT and cognitively engaging PA "required further data". The dose-response shape is an INVERTED U at moderate intensity — meaning more is not better past a modest point, so the common "add more minutes" policy inference does not follow even from the optimistic reading. These estimates are uncorrected for publication bias; the same outcome class in the same age group falls to SMD 0.012 with moderate-to-strong evidence against any effect once study-level bias adjustment is applied (Bartos et al. 2025). Note also the outcome class: executive function on lab tasks, not achievement.

Randomized corpus; the dose contrasts are between-study, which reintroduces study-level confounding but not participant-level genetic confounding.