Optimal exercise modalities for post-intervention motor function changes in stroke survivors: a pairwise and network meta-analysis.

BACKGROUND: Stroke is a leading cause of long-term disability worldwide, particularly among older adults, and severely compromises functional independence and quality of life. Exercise interventions are well established as effective strategies for promoting motor recovery through improvements in neuroplasticity, fitness, strength, and psychosocial well-being, yet most evidence to date has focused on immediate post-intervention outcomes. The durability of these benefits during follow-up periods-defined as the post-intervention phase when patients are observed after completing structured rehabilitation and begin transitioning to self-directed activities-remains poorly understood. Clarifying functional trajectories during this phase is critical for assessing the sustainability and real-world value of rehabilitation strategies. A systematic evaluation of motor function outcomes beyond formal intervention is therefore urgently needed to inform the design of programs that prioritize long-term recovery. METHODS: This systematic review and network meta-analysis followed PRISMA 2020 guidelines and was registered in PROSPERO (CRD420251117014). Five databases (PubMed, Embase, Web of Science, Scopus, Cochrane Library) were searched up to March 13, 2025. We included randomized controlled trials comparing structured exercise interventions in post-stroke adults during follow-up periods, reporting motor function outcomes. Risk of bias was assessed using RoB 2.0, and evidence quality was graded via CINeMA. Pairwise and Bayesian network meta-analyses were conducted using netmeta and gemtc packages. Heterogeneity, inconsistency, regression, sensitivity, and publication bias analyses were also performed. RESULTS: Sixty-four trials (n = 3074) were included, covering 23 exercise modalities. For the 6-min walk test, resistance training demonstrated the greatest effect [mean difference (MD) = 21.79, 95% CI 8.19-35.40], followed by robot-assisted therapy (MD = 19.78, 95% CI 3.30-36.27). In the 10-m walk test, robot-assisted therapy ranked highest (MD = 0.06, 95% CI 0.01-0.12). Task-oriented training significantly improved balance [standardized mean difference (SMD) = 0.35, 95% CI 0.07-0.64]. No significant effects were observed in lower or upper limb Fugl-Meyer outcomes. Consistency and heterogeneity were generally acceptable, and no publication bias was detected. CONCLUSIONS: Resistance training and robot-assisted therapy appear most effective for enhancing walking capacity after stroke. Task-oriented training benefits postural balance. Evidence for upper and lower limb recovery remains inconclusive due to the small number of trials and wide confidence intervals. These findings can inform individualized, modality-specific rehabilitation strategies for older adults recovering from stroke. TRIAL REGISTRATION: Registered on PROSPERO (CRD420251117014).