Comparative effects and dose–response relationships of exercise for finger dexterity and upper-limb function in Parkinson’s disease: a Bayesian network and model-based dose–response meta-analysis

Background Upper-limb impairment limits independence in Parkinson’s disease (PD), but the comparative effects and dose–response profiles of exercise modalities remain uncertain. Methods PubMed, Embase, CENTRAL, Web of Science, Scopus, and SPORTDiscus were searched from inception to June 20, 2026. Studies that qualified were randomized controlled trials of structured exercise or rehabilitation in adults with PD. Bayesian random-effects network meta-analysis and model-based dose–response network meta-analysis were conducted separately for finger dexterity and upper-limb function. Weekly dose was expressed as metabolic equivalent of task minutes per week (MET·min/week), and evidence certainty was assessed using CINeMA. Results Thirty-six reports comprising 1,575 participants were included. Aerobic exercise training (Hedges’ g 1.46, 95% CrI 0.26–2.69) and task-oriented finger-dexterity training (0.96, 0.23–1.67) improved finger dexterity versus usual care. Low-certainty evidence suggested that mind–body exercise may improve upper-limb function, although the unusually large estimate (Hedges’ g = 6.90, 95% CrI 4.28–9.43) was based on only two small trials involving 53 participants in the relevant comparison arms. Given the sparse evidence and local inconsistency in the modality-specific dose–response analysis, this result should be regarded as exploratory and interpreted cautiously. Models fitted to mapped dose nodes suggested an approximately linear association for finger dexterity and diminishing incremental gains for upper-limb function. Predictions toward 1,800 MET·min/week partly extended beyond the highest representative dose nodes and remained uncertain. Conclusion Exercise effects were outcome specific. The findings can inform exercise-modality selection and future dose-ranging trials, but they do not establish an optimal dose or justify a uniform clinical prescription. Systematic review registration https://www.crd.york.ac.uk/PROSPERO/view/CRD420261463560 , identifier: CRD420261463560.