Studies were combined using fixed-effect meta-analyses to calculate pooled relative risks with 95% CI for the comparisons defined above. Statistical heterogeneity was assessed using the I2 statistic; where this exceeded 50%, a random-effects model was adopted for the analysis. Indirect comparisons were used for some of the analyses, but the methodology used was not described.
Sensitivity analyses excluded individual studies in turn and analysed studies in which levodopa was or was not used. Analyses were conducted for adverse events of nausea, dyskinesia, dizziness, orthostatic hypotension, somnolence, insomnia, headaches, constipation, hallucinations, abdominal pain, vomiting and confusion.