A systematic review and Bayesian meta-analysis provide evidence for an effect of acute physical activity on cognition in young adults

Abstract

Physical exercise is a potential intervention for enhancing cognitive function across the lifespan. However, while studies employing long-term exercise interventions consistently show positive effects on cognition, studies using single acute bouts have produced mixed results. Here, a systematic review and meta-analysis was conducted to determine the impact of acute exercise on cognitive task performance in healthy young adults. A Bayesian hierarchical model quantified probabilistic evidence for a modulatory relationship by synthesizing 651 effect sizes from 113 studies from PsychInfo and Google Scholar representing 4,390 participants. Publication bias was mitigated using the trim-and-fill method. Acute exercise was found to have a small beneficial effect on cognition (g = 0.13 ± 0.04; BF = 3.67) and decrease reaction time. A meta-analysis restricted to executive function tasks revealed improvements in working memory and inhibition. Meta-analytic estimates were consistent across multiple priors and likelihood functions. Physical activities were categorized based on exercise type (e.g., cycling) because many activities have aerobic and anaerobic components, but this approach may limit comparison to studies that categorize activities based on metabolic demands. The current study provides an updated synthesis of the existing literature and insights into the robustness of acute exercise-induced effects on cognition. Funding provided by the United States Army Research Office.

Fig. 1. PRISMA flow diagram of literature search results.

A total of 113 studies were deemed eligible for meta-analytic modeling.

Fig. 2. Meta-analysis of the effect of acute exercise on general cognitive task performance.

a Funnel plot of 642 study effect sizes (black circles). Imputed effect sizes after using the trim and fill method are represented by the unfilled circles ($n=9$). Vertical blue line indicates the estimated pooled effect sizes, while dashed black lines represent a pseudo 95% confidence limits. b Posterior distribution of estimated pooled effect. Horizontal black line indicates bounds of 89% HDI derived using $n=651$ effect sizes. c Empirical cumulative density function of distribution in b, where the dashed black line indicates the pooled effect. d Representation of using the Savage-Dickey ratio to calculate BFs. The density of the null value in the prior distribution (red) is divided by its density in the posterior distribution (blue) to yield probabilistic evidence in favor of the alternative hypothesis. e Posterior distributions of between and within study heterogeneity.

Fig. 3. Subgroup analyses.

Posterior distributions of a cognitive and b exercise moderators. Horizontal black line indicates the 89% HDI interval, while the black dot represents the mode of the posterior distribution. Intervals derived using $\mathrm{n}=651$ effect sizes.

Fig. 4. Interactions between cognitive and exercise moderators.

Posterior mode estimates of models including interactions between cognitive domain and a exercise type and b task outcome measure. Width of line represents 89% HDI derived using $\mathrm{n}=651$ effect sizes.

Fig. 5. Sensitivity analyses.

Estimates for the a overall pooled effect size, b between- and c within-study heterogeneity parameters across the t-likelihood function (TL), weakly informed, null effect (NE), and positive effect (PE) priors. Color dots represent mode of posterior distributions, while color horizontal line depicts the 89% HDI derived using $n=651$ effect sizes.

Fig. 6. Subgroup analyses of effects from tasks testing executive function.

Posterior distributions for executive function sub-domain. Horizontal black line indicates the 89% HDI interval, while the black dot represents the mode of the posterior distribution, which was derived using $n=433$ effect sizes.