The epidemiology of Parkinson's disease: risk factors and prevention

Summary

Since 2006, several longitudinal studies have assessed environmental or behavioural factors that seem to modify the risk of developing Parkinson's disease. Increased risk of Parkinson's disease has been associated with exposure to pesticides, consumption of dairy products, history of melanoma, and traumatic brain injury, whereas a reduced risk has been reported in association with smoking, caffeine consumption, higher serum urate concentrations, physical activity, and use of ibuprofen and other common medications. Randomised trials are investigating the possibility that some of the negative risk factors might be neuroprotective and thus beneficial in individuals with early Parkinson's disease, particularly with respect to smoking (nicotine), caffeine, and urate. In the future, it might be possible to identify Parkinson's disease in its prodromal phase and to promote neuroprotective interventions before the onset of motor symptoms. At this time, however, the only intervention that seems justifiable for the primary prevention of Parkinson's disease is the promotion of physical activity, which is likely to be beneficial for the prevention of several chronic diseases.

Introduction

Major discoveries have profoundly changed our understanding of Parkinson's disease and its determinants. Whereas genetic studies¹ have revealed the heterogeneity of Parkinson's disease and provided insights into its pathogenesis and aetiology,² epidemiological investigations have provided robust evidence that behavioural and environmental factors have a key role in disease pathogenesis and progression. This evidence is strengthened and complemented by observations that 90% of Parkinson's disease cases have no identifiable genetic cause,³ and that many factors associated with an altered risk of Parkinson's disease have neuroprotective or neurotoxic properties in animal models of the disease.

In this Review, we provide an update on the descriptive epidemiology of Parkinson's disease, and then focus on the epidemiological advances of the last 10 years and their implications for Parkinson's disease prevention and treatment. Studies on genetic forms of Parkinson's disease or parkinsonism other than idiopathic Parkinson's disease, and the substantial advances in the identification and characterisation of prodromal Parkinson's disease are considered beyond the scope of this Review. Where evidence exists, we mention briefly the potential underlying biological mechanism of the epidemiological findings. We describe comprehensively the longitudinal investigations of nongenetic risk factors for Parkinson's disease, and provide a critical summary of current knowledge, knowledge gaps, and implications. Because most epidemiological studies do not distinguish idiopathic Parkinson's disease from Parkinson's disease due to genetic mutations, and rely on clinical rather than pathological diagnostic criteria, from here on we refer to Parkinson's disease without further specifications, with the understanding that the conclusions, being driven by the more common clinically defined sporadic Parkinson's disease, might not apply to monogenetic forms, and might be affected by the accuracy of the clinical diagnoses, which is typically only 80–90% when compared with pathological findings.⁴

Progress over the past 10 years in understanding the risk factors for Parkinson's disease can largely be attributed to prospective longitudinal studies, which have well known advantages over case-control studies that rely on the participants' recall of past events. Recall bias is particularly important in Parkinson's disease, which has a long prodromal phase characterised by symptoms such as hyposmia, constipation, and sleep disorders that might be present up to 20 years before manifestation of the characteristic motor symptoms,⁵ and are likely to affect several aspects of lifestyle, such as diet, physical activity, and medication. Further, for most case-control studies, the representativeness of the control group is uncertain. This Review, therefore, mostly relies on studies conducted within well defined cohorts of individuals without Parkinson's disease who have provided biological samples or information on the exposures of interest at time of recruitment, and were then followed prospectively for the occurrence of newly diagnosed Parkinson's disease; this category includes case-control studies nested within these cohorts (table). Validity of these investigations requires accurate information on the exposures and potential confounders and their changes over time, the duration and completeness of the follow-up and Parkinson's disease ascertainment, and the correctness of the Parkinson's disease diagnosis. Weaknesses in one or more of these aspects are common, and therefore understanding of risk factors for Parkinson's disease requires an assessment of each investigation and the exploration of potential alternative explanations of the reported findings.

The increased availability of large electronic databases has provided an additional source of epidemiological data, which are particularly useful to investigate the relation between prescription drugs and other medical events (eg, head trauma) and Parkinson's disease risk, but lack accurate information on confounders and dates of disease onset. Date of disease onset is often equated to the date of diagnosis or first treatment for Parkinson's disease, which in some individuals can be years after symptom onset.¹⁹ Keeping these limitations in mind, we have referred to these studies to complement the inference that could be made from prospective cohorts.