Many agents developed for neuroprotective treatment of Parkinson disease (PD) have shown great promise in the laboratory, but none have translated to positive results in patients with PD. Potential neuroprotective drugs, such as ubiquinone, creatine and PYM50028, have failed to show any clinical benefits in recent high-profile clinical trials. This 'failure to translate' is likely to be related primarily to our incomplete understanding of the pathogenic mechanisms underlying PD, and excessive reliance on data from toxin-based animal models to judge which agents should be selected for clinical trials. Restricted resources inevitably mean that difficult compromises must be made in terms of trial design, and reliable estimation of efficacy is further hampered by the absence of validated biomarkers of disease progression. Drug development in PD dementia has been mostly unsuccessful; however, emerging biochemical, genetic and pathological evidence suggests a link between tau and amyloid-β deposition and cognitive decline in PD, potentially opening up new possibilities for therapeutic intervention. This Review discusses the most important 'druggable' disease mechanisms in PD, as well as the most-promising drugs that are being evaluated for their potential efficiency in treatment of motor and cognitive impairments in PD.