Abstract / Summary
Exercise modality and planned weekly structured-exercise exposure may provide complementary information for Parkinson's disease motor outcomes, but their outcome-specific effects and evidential confidence remain uncertain.
We synthesized 173 randomized trials using covariance-aware Bayesian random-effects network meta-analysis for motor symptoms, balance, walking speed, and Timed Up and Go (TUG). A key secondary analysis estimated exercise-class-adjusted associations per 250 metabolic equivalent task minutes/week, with pre-specified sensitivity analyses.
The category NMA networks included 110, 106, 89, and 96 trials for Unified Parkinson's Disease Rating Scale part III (UPDRS III), balance, walking speed, and TUG, respectively. Mind-body exercise had the largest posterior median versus non-exercise control for UPDRS III (g = 0.65, 95% credible interval [CrI] 0.42-0.87). In the reconstructed, exercise-category-adjusted dose models, planned weekly structured-exercise exposure was positively associated at the trial level with UPDRS III (0.121 SMD per + 250 MET-min/week, 95% CrI 0.055-0.194), balance (0.111, 0.047-0.184), and TUG (0.124, 0.058-0.190), but not walking speed (0.008, - 0.063 to 0.076). The TUG estimate attenuated after P95 dose truncation. All active-versus-control prediction intervals crossed zero; CINeMA confidence was low or very low. The evidence did not establish a universally best modality, an optimal dose, or an individual causal target. Higher planned weekly exposure was associated at the trial level with greater average improvement in UPDRS III, balance, and TUG, whereas walking speed showed no clear association. However, within-class randomized dose-contrast estimates were imprecise for all four outcomes, and the TUG association was sensitive to the high-dose tail.