- Design
- Systematic review, Bayesian network meta-analysis and meta-regression of randomised trials
- Population
- 173 randomised trials of exercise in Parkinson's disease
- Primary outcome
- UPDRS III, balance, walking speed and Timed Up and Go
- Effect
- Mind-body exercise vs control on UPDRS III: g 0.65 (95% CrI 0.42 to 0.87); confidence low or very low
This Bayesian network meta-analysis pooled 173 randomised trials of exercise in Parkinson's disease, with networks of 89 to 110 trials for each of four outcomes: motor symptoms (UPDRS III), balance, walking speed and Timed Up and Go.
Mind-body exercise had the largest effect against non-exercise control on UPDRS III (standardised mean difference 0.65, 95% credible interval 0.42 to 0.87). In dose models, more planned weekly exercise was associated at trial level with better UPDRS III, balance and Timed Up and Go, but not walking speed. The Timed Up and Go association weakened when the highest doses were trimmed. All prediction intervals for active exercise versus control crossed zero, and confidence in the evidence was low or very low.
The authors state that the evidence did not establish a best modality, an optimal dose or an individual causal target. Dose findings are across-trial associations and not randomised comparisons of dose. The practical conclusion is modest: exercise is worth recommending, and the type matters less than whether the patient will keep doing it.
- Recommend regular structured exercise to every person with Parkinson's disease who can do it.
- Mind-body exercise had the largest average effect on motor scores.
- Choose a type the patient will keep doing; no modality was shown to be best.
- Do not promise a specific benefit; confidence in the evidence was low or very low.
Why it matters
It asks which exercise to prescribe, and the honest answer from 173 trials is that nobody knows yet.
Don't overread it
Dose associations are trial-level and do not show that doing more exercise causes better outcomes in an individual.
The statistics, in plain English
A prediction interval describes where the effect in a new trial might fall. When it crosses zero, as here for every active-versus-control comparison, a future trial could find no benefit, even though the average effect is positive.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for neurology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free