Effects of different exercise interventions on motor function in patients at different post-stroke recovery phases: a systematic review and Bayesian network meta-analysis.

Background: Post-stroke motor dysfunction is common, with diverse intervention strategies; however, their efficacy across recovery phases remains unclear. Objective: To systematically review and compare the effects of different exercise interventions on motor function in patients at different post-stroke recovery phases. Methods: Web of Science, PubMed, Embase, and the Cochrane Library were searched from inception to October 2025. Randomized controlled trials (RCTs) in adults with stroke assessing exercise effects on motor function were included. A Bayesian Network Meta-analysis (NMA) was applied for measures with enough studies and comparisons. Risk of bias was assessed using the Cochrane tool. Results: A total of 35 RCTs involving 1,540 stroke patients. In the acute phase, aerobic exercise (AE) was potentially superior to conventional therapy (CT) in improving 6-min walk distance (6MWD), Barthel Index (BI), and Fugl-Meyer Assessment (FMA). In the subacute phase, AE also showed a certain advantage in enhancing 6MWD. According to the surface under the cumulative ranking curve (SUCRA), core stability exercise combined with resistance exercise (CSE + RE, 99.99%) ranked best for improving the 10-meter walk test (10MWT); CSE (80.84%) was most effective for BI; and AE (78.73%) was optimal for improving FMA. In the chronic phase, AE + RE ranked first for improving 6MWD (83.33%) and FMA (99.90%); CSE (98.62%) was the most effective for 10MWT; and AE (87.85%) remained the best intervention for enhancing BI. Conclusion: Exercise interventions exert phase-specific effects on motor recovery post-stroke. These findings underscore the importance of tailored, phase-specific rehabilitation programs. Long-term efficacy and individualized optimization require further investigation. Systematic review registration: PROSPERO CRD42024607395.