An umbrella review of randomized control trials on the effects of physical exercise on cognition

Healthy humans across lifespan (children, adults, older adults); 24 meta-analyses of RCTs

Umbrella review of 24 meta-analyses restricted to RCTs in healthy humans across the lifespan, plus re-meta-analysis of pooled primary RCTs with moderator adjustment (active vs passive control, baseline differences) and publication-bias correction. Assessed power, selective inclusion and analytic flexibility of source meta-analyses.

The strongest deflation of the exercise->cognition claim. Most source meta-analyses report positive effects, but primary RCTs are underpowered, selectively included and publication-biased; the effect shrinks ~4x with basic moderator control and vanishes after bias correction. Authors urge caution in claiming cognitive benefits of exercise in healthy people. Contested: Singh/Erickson et al. rebuttal ('Do not underestimate the cognitive benefits of exercise', Nat Hum Behav 2024) and a Ciria reply.

Not addressed; RCT design controls genetic confounding by randomization, which is why the near-null is informative — larger observational correlations likely reflect rGE/selection.

Directly contested in a 2024 Nature Human Behaviour exchange (comment + reply); the near-null is consistent with large preregistered field RCTs (e.g. Fit to Study).