OBJECTIVE: To investigate whether white matter microstructural changes can be used as a predictor of worsening of motor features or cognitive decline in patients with Parkinson's disease and verify whether white matter microstructural longitudinal changes differ between patients with Parkinson's disease with normal cognition and those with mild cognitive impairment. METHODS: We enrolled 120 newly diagnosed patients with early stage Parkinson's disease (27 with mild cognitive impairment and 93 with normal cognition) along with 48 controls. Participants were part of the incidence of cognitive impairment in cohorts with longitudinal evaluation in Parkinson's disease study and were assessed at baseline and 18 months later with cognitive, motor tests and diffusion tensor imaging. The relationships between fractional anisotropy and mean diffusivity with disease status, cognitive and motor function were investigated. RESULTS: At baseline, patients with early stage Parkinson's disease had significantly higher widespread mean diffusivity relative to controls, regardless of cognitive status. In patients with Parkinson's disease/mild cognitive impairment, higher mean diffusivity was significantly correlated with lower attention and executive function scores. At follow-up frontal mean diffusivity increased significantly when comparing patients with Parkinson's disease/mild cognitive impairment with those with normal cognition. Baseline mean diffusivity was a significant predictor of worsening of motor features in Parkinson's disease. CONCLUSIONS: Mean diffusivity represents an important correlate of cognitive function and predictor of motor impairment in Parkinson's disease: DTI is potentially a useful tool in stratification of patients into clinical trials and to monitor the impact of treatment on motor function.