Effects of different types of exercise in patients with stroke: a network meta-analysis

Exercise is a cornerstone of post-stroke rehabilitation, but comparative evidence across exercise modalities and functional domains remains fragmented. This study compared the relative effects of different exercise modalities across major domains of physical recovery in adults with subacute or chronic stroke. PubMed, Embase, Web of Science, and the Cochrane Library were searched from inception to January 22, 2026 for randomized controlled trials of exercise-based rehabilitation after stroke. Bayesian random-effects network meta-analyses were conducted for the Berg Balance Scale (BBS), 6-minute walk test (6MWT), 10-meter walk test (10MWT), and Timed Up and Go test (TUG). The Modified Barthel Index (MBI) and Fugl-Meyer Assessment-Lower Extremity (FMA-LE) were synthesized narratively because the available evidence did not support reliable network meta-analysis. Treatment effects were expressed as mean differences (MDs) or standardized mean differences (SMDs) with 95% credible intervals (CrIs), and certainty of evidence was assessed using GRADE. Fifty-two randomized controlled trials involving 2,329 participants were included. Whole-body vibration showed the clearest benefit for balance compared with control (BBS: MD 9.86, 95% CrI 4.78 to 14.92) and was the only principal comparison supported by moderate-certainty evidence. Aerobic exercise, aerobic plus strength training, coordinative training, and strength training improved walking endurance compared with control, with no conclusive differences among active modalities. For gait performance, aerobic exercise, coordinative training, and strength training showed favorable effects on the 10MWT, whereas strength training improved functional mobility on the TUG (MD − 3.41 s, 95% CrI − 5.06 to − 1.66). Evidence for the 10MWT was particularly uncertain because of very-low-certainty evidence, local inconsistency, potential small-study effects, and sensitivity to individual studies. Evidence for the MBI and FMA-LE was insufficient for reliable comparative inference. Of the 40 principal comparisons assessed, one was rated as moderate certainty, 25 as low certainty, and 14 as very low certainty. The comparative effects of exercise after stroke differed across rehabilitation domains. Whole-body vibration showed the most consistent evidence for balance, whereas several exercise modalities improved walking endurance. Comparative evidence for gait performance and functional mobility was less robust, and evidence for activities of daily living and lower-extremity motor recovery remained limited. Exercise selection should therefore be aligned with the specific rehabilitation target and clinical context rather than guided by treatment rankings alone.