Background: Although subthalamic stimulation is a recognised treatment for motor complications in Parkinson's disease, reports on behavioural outcomes are controversial, which represents a major challenge when counselling candidates for subthalamic stimulation. We aimed to assess changes in behaviour in patients with Parkinson's disease receiving combined treatment with subthalamic stimulation and medical therapy over a 2-year follow-up period as compared with the behavioural evolution under medical therapy alone. Methods: We did a parallel, open-label study (EARLYSTIM) at 17 surgical centres in France (n=8) and Germany (n=9). We recruited patients with Parkinson's disease who were disabled by early motor complications. Participants were randomly allocated (1:1) to either medical therapy alone or bilateral subthalamic stimulation plus medical therapy. The primary outcome was mean change in quality of life from baseline to 2 years. A secondary analysis was also done to assess behavioural outcomes. We used the Ardouin Scale of Behavior in Parkinson's Disease to assess changes in behaviour between baseline and 2-year follow-up. Apathy was also measured using the Starkstein Apathy Scale, and depression was assessed with the Beck Depression Inventory. The secondary analysis was done in all patients recruited. We used a generalised estimating equations (GEE) regression model for individual items and mixed model regression for subscores of the Ardouin scale and the apathy and depression scales. This trial is registered with ClinicalTrials.gov, number NCT00354133. The primary analysis has been reported elsewhere; this report presents the secondary analysis only. Findings: Between July, 2006, and November, 2009, 251 participants were recruited, of whom 127 were allocated medical therapy alone and 124 were assigned bilateral subthalamic stimulation plus medical therapy. At 2-year follow-up, the levodopa-equivalent dose was reduced by 39% (–363·3 mg/day [SE 41·8]) in individuals allocated bilateral subthalamic stimulation plus medical therapy and was increased by 21% (245·8 mg/day [40·4]) in those assigned medical therapy alone (p<0·0001). Neuropsychiatric fluctuations decreased with bilateral subthalamic stimulation plus medical therapy during 2-year follow-up (mean change −0·65 points [SE 0·15]) and did not change with medical therapy alone (–0·02 points [0·15]); the between-group difference in change from baseline was significant (p=0·0028). At 2 years, the Ardouin scale subscore for hyperdopaminergic behavioural disorders had decreased with bilateral subthalamic stimulation plus medical therapy (mean change −1·26 points [SE 0·35]) and had increased with medical therapy alone (1·12 points [0·35]); the between-group difference was significant (p<0·0001). Mean change from baseline at 2 years in the Ardouin scale subscore for hypodopaminergic behavioural disorders, the Starkstein Apathy Scale score, and the Beck Depression Inventory score did not differ between treatment groups. Antidepressants were stopped in 12 patients assigned bilateral subthalamic stimulation plus medical therapy versus four patients allocated medical therapy alone. Neuroleptics were started in nine patients assigned medical therapy alone versus one patient allocated bilateral subthalamic stimulation plus medical therapy. During the 2-year follow-up, two individuals assigned bilateral subthalamic stimulation plus medical therapy and one patient allocated medical therapy alone died by suicide. Interpretation: In a large cohort with Parkinson's disease and early motor complications, better overall behavioural outcomes were noted with bilateral subthalamic stimulation plus medical therapy compared with medical therapy alone. The presence of hyperdopaminergic behaviours and neuropsychiatric fluctuations can be judged additional arguments in favour of subthalamic stimulation if surgery is considered for disabling motor complications. Funding: German Federal Ministry of Education and Research, French Programme Hospitalier de Recherche Clinique National, and Medtronic.