Effects of school-based physical activity on academic achievement in children and adolescents: a systematic review and meta-analysis

17 randomized controlled trials, search to 8 December 2024; no total participant N is reported anywhere in the paper

Children and adolescents aged 6-18 in school-based physical-activity programmes (PE classes, active classroom breaks, physically integrated academic lessons; extracurricular and after-school activity explicitly excluded)

PROSPERO-registered (CRD42024623670) random-effects (DerSimonian-Laird) meta-analysis in RevMan 5.4, with Cochrane risk-of-bias assessment and subgroup analyses by intensity and duration. Corpus is ALL RCTs — inclusion criterion 5 is "the study design is a randomized controlled trial" and section 3.1 states "All investigations were RCTs"; an earlier version of this record wrongly described the corpus as mixing randomized with quasi-experimental designs. What actually earns the D grade is the outcome measures: inclusion criterion 4 states that "academic achievement is assessed through NON-STANDARDIZED tests", with overall achievement being "the average score across subjects or evaluated by teachers". A few included trials nonetheless used standardized batteries (Donnelly/A+PAAC used WIAT-III), so the pooled measure type is mixed in fact while the stated eligibility rule is non-standardized. All 17 studies were rated "moderate quality". Publication bias was inspected VISUALLY via funnel plots only (no Egger test, no trim-and-fill, no PET-PEESE); maths, reading and language funnels were called symmetrical, and for the headline overall-achievement estimate the authors state the small study count "hindered the assessment of the funnel plot's symmetry and precluded the removal of publication bias". Internal bookkeeping is loose: section 3.1 counts 12/5/6/5 studies per outcome while the forest plots pool 13/6/7/6; the overall-achievement forest plot cites reference (41) although the review states it included studies (24-40); and Table 2's "intervention intensity" and "intervention duration" row labels are transposed.

Two of the four pooled outcomes are positive and two are null, and the record previously carried only the positives. Positive: mathematics SMD 0.11 (95% CI 0.04 to 0.18, p=0.001, I2=55%, 13 studies) and overall achievement SMD 0.22 (95% CI 0.01 to 0.44, p=0.040, I2=74%, 6 studies). NULL: reading comprehension SMD 0.07 (95% CI -0.03 to 0.18, p=0.170, I2=74%, 6 studies) and language expression SMD -0.06 (95% CI -0.18 to 0.06, p=0.370, I2=68%, 7 studies) — negative point estimate, and negative in every one of its five subgroups. The authors themselves conclude "the evidence for improvements in reading comprehension and language expression remains inconclusive". The largest subgroup estimates are high-intensity activity on mathematics (0.41, 3 studies) and interventions of 24 weeks or longer (0.47 overall, 3 studies), both resting on two or three trials. Heterogeneity is severe (I2 up to 74%), the only bias check is visual funnel-plot inspection and it could not even be run on the headline overall estimate, the stated eligibility rule admits non-standardized tests and teacher grades, and the 17 studies substantially overlap the 26 in Alvarez-Bueno et al. 2017 and the 92 in Vasilopoulos et al. 2023 (Donnelly, Ahamed, Egger, Gall and Mullender-Wijnsma appear in both).

All-RCT corpus, so randomization removes genetic confounding of the treatment contrast. The residual risk is measurement, not heredity - teacher-graded overall achievement is the outcome carrying the largest pooled effect.