Many but not all studies have indicated that smoking is inversely associated with Parkinson's disease (PD). Meta-analysis of epidemiological studies on smoking and PD was performed to summarize data from published studies. Fifty-four epidemiological studies (48 case-control and 6 cohort studies, 53 publications) were identified for potential inclusion in meta-analysis. The summary risk estimates for current smokers, former smokers, and ever (current and former) smokers were 0.31 (95% confidence interval (CI) = 0.25-0.38), 0.72 (95% CI = 0.63-0.83) and 0.55 (95% CI = 0.51-0.59), respectively. In stratified analysis by study design, smoking had a somewhat greater impact on PD risk in cohort studies than in case-control studies. However, meta-regression indicated that the study design did not significantly contribute to heterogeneity. Additional analyses were restricted to case-control studies because of the sufficient number of studies. Stratified analysis by ethnicity indicated that the summary OR for ever-smokers was nonsignificantly smaller in Asian populations than in Caucasian populations. In stratified analysis by source of controls, former smoking was significantly associated with a decreased risk of PD in hospital-based case-control studies but was marginally associated with a decreased risk in population-based case-control studies. The source of controls did not contribute significantly to heterogeneity. PD risk associated with ever-smoking was significantly lower for a hospital-based approach than a population-based approach. Among current smokers, the association held true to the same extent for both approaches. This meta-analysis indicated that smokers have a lower risk of PD. As PD is a multifactorial disease, further investigation of the smoking-gene interaction on PD risk may lead to a better understanding of the pathogenesis of PD.