Walking and subjective sleep quality in adults: a Bayesian three-level meta-analysis with probabilistic clinical...

To evaluate the effect of walking on subjective sleep quality in adults using a Bayesian three-level meta-analysis and to identify moderators of treatment response.

PubMed, Embase, the Cochrane Library, and Web of Science were searched from inception through March 2026 for randomized controlled trials (RCTs) comparing walking interventions with inactive or non-aerobic controls. Effect sizes were calculated as Hedges' g. A Bayesian three-level random-effects model was fitted to handle dependent effect sizes. Clinical relevance was assessed using posterior probabilities of exceeding the minimal clinically important difference (MCID, defined as g ≤ -0.39). Moderators were examined through subgroup analyses and meta-regressions.

Twenty-one RCTs (1,707 participants, 66 effect sizes) were included. Walking produced a medium-to-large improvement in subjective sleep quality (g = -0.76, 95% credible interval: -0.99 to -0.55), with extreme Bayesian evidence (Bayes factor > 10⁵) and a 99.97% posterior probability of exceeding the MCID. The 95% prediction interval fell entirely below zero. Baseline Pittsburgh Sleep Quality Index score was the strongest moderator (β = -0.302), with worse baseline sleep associated with larger effects. Supervision mode and exercise setting did not meaningfully influence outcomes. GRADE certainty of evidence was very low.

Current evidence suggests that walking may improve subjective sleep quality in adults, with a 99.97% posterior probability of reaching the prespecified minimal clinically important difference, although evidence certainty is very low. Moderate-intensity walking three to five times weekly, 30 to 60 min per session, for 12 to 24 weeks may serve as a feasible prescription window, particularly for those with poorer baseline sleep. Future RCTs incorporating objective sleep measures are needed.

PROSPERO, CRD420261351897.