The Evidence on Teaching

Effects of physical activity on schoolchildren's academic performance: The Active Smarter Kids (ASK) cluster-randomized controlled trial

Resaland GK, Aadland E, Moe VF, Aadland KN, Skrede T, Stavnsbo M, Suominen L, Steene-Johannessen J, Glosvik O, Andersen JR, Kvalheim OM, Engelsrud G, Andersen LB, Holme IM, Ommundsen Y, Kriemler S, van Mechelen W, McKay HA, Ekelund U, Anderssen SA · 2016

grade Brctdeveloper-ledunreplicated
Sample
1,129 fifth-graders from 57 elementary schools (28 intervention / 29 control), Sogn og Fjordane, Norway
Population
10-year-old Norwegian schoolchildren
Design
Seven-month cluster RCT, school-level randomization, registered NCT02132494. Three-component intervention: 90 min/week physically active educational lessons in the playground, 5 min/day classroom activity breaks, 10 min/day physical activity homework. Physical activity objectively measured by accelerometry. Outcomes were the Norwegian standardized national tests in numeracy, reading and English — independent, externally scored, high-stakes instruments, which is unusually strong for this literature. 57 clusters is well above the norm for the field.
Key findings
Primary analysis is null: standardized differences 0.01-0.06, all p>0.358, across numeracy, reading and English. The authors themselves conclude the trial "supports the notion that there is still inadequate evidence to conclude that increased physical activity in school enhances academic achievement in all children." A post-hoc subgroup — the lowest baseline-performance tertile — showed a large numeracy effect (standardized difference 0.62, 95% CI 0.19-1.07, interaction p=0.005). This subgroup result is the single most-cited positive finding in the school-PA literature and should be read as hypothesis-generating: it is one interaction out of many tested, in the tertile most exposed to regression to the mean, and it has not been replicated in a trial designed to test it.
Genetic confound
Randomization removes genetic confounding of the primary contrast. The subgroup finding is more exposed — baseline-defined tertiles are strongly heritability-loaded (low baseline achievers differ systematically from high achievers), so subgroup x treatment interactions in such strata are the estimates most likely to reflect measurement artifacts rather than differential causal response.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Numeracy, reading, English (primary, full sample)standardized difference0.01 to 0.06, all p>0.358standardizedend of 7-month interventionbusiness-as-usualend-of-treatmentdomain-skill
Numeracy in the lowest baseline-performance tertile (post-hoc subgroup)standardized difference0.62 (95% CI 0.19 to 1.07); subgroup x group interaction p=0.005standardizedend of 7-month interventionbusiness-as-usualend-of-treatmentdomain-skill

Cited by