The association between the sparteine/debrisoquine (CYP2D6) oxidation polymorphism and the risk of Parkinson's disease was examined in a meta-analysis of case-control studies. The odds ratio was calculated for the risk of Parkinson's disease among poor metabolisers compared with extensive metabolisers. Twenty-one studies were identified of which six were excluded because they were not reported as full papers (n = 3), used incomplete genotype analysis (n = 2) or used Parkinson patients as both control individuals and cases (n = 1). The overall odds ratio was 1.48 (95% confidence interval 1.10-1.99). The odds ratio was 1.05 (95% confidence interval 0.63-1.77) in studies discriminating extensive and poor metabolisers by phenotyping (n = 8) and 1.67 (95% confidence interval 1.11-2.50) in studies using genotyping (n = 7). This difference was caused by a single large study using genotyping. We conclude that there is no convincing evidence of an association between the debrisoquine/sparteine polymorphism and Parkinson's disease. However, it could prove worthwhile to perform another large study using genotyping.