Aerobic fitness mediates the intervention effects of a school-based physical activity intervention on academic performance. The School in Motion study - A cluster randomized controlled trial

2,084 students, 30 Norwegian lower-secondary schools (mediation analysis within the School in Motion cluster RCT)

14-year-old Norwegian lower-secondary students

Causal-steps mediation analysis on the School in Motion trial, testing whether aerobic fitness (Andersen test) carried the intervention effect on standardized national tests. Graded C, not B: the exposure is randomized but the mediator is not, so the mediator-outcome path is an observational association within a trial and inherits the full confounding structure of the observational fitness-achievement literature. Causal-steps mediation is also the weakest of the available mediation estimators.

This is the best available direct test of the "fitness is the input that raises achievement" hypothesis inside a randomized trial, and it comes out inconsistent. In the Physically Active Learning arm, fitness partially mediated numeracy (28% of a 1.73-point total effect) and fully mediated reading (0.89 points total, 0.40 direct, CI now crossing zero). In the Don't Worry - Be Happy arm — which produced LARGER academic effects — fitness mediated nothing. A mediator that explains the smaller effect and none of the larger one is not behaving like a mechanism. Read alongside Tuvey et al. 2019, which found across 17 trials that intervention-induced fitness change does not predict academic change (beta = -0.04), the fitness-as-mediator story has no consistent support.

The mediator is the problem. Cardiorespiratory fitness is roughly 55-72% heritable (Schutte et al. 2016), so the fitness-achievement path estimated here is exactly the confounded association the trial design was supposed to escape: children who respond to training with larger fitness gains differ genetically from those who do not.