BACKGROUND: Dementia with Lewy bodies (DLB) and Parkinson's disease dementia (PDD) are common forms of dementia that substantially affect quality of life. Currently, the only treatment licensed for PDD is rivastigmine, and there are no licensed treatments for DLB. We aimed to test the safety and efficacy of the N-methyl D-aspartate (NMDA) receptor antagonist memantine in patients with PDD or DLB. METHODS: We did a parallel-group, 24-week, randomised controlled study of memantine (20 mg per day) versus placebo at four psychiatric and neurological outpatient clinics in Norway, Sweden, and the UK during 2005-08. Patients were included if they fulfilled the UK Parkinson's Disease Society Brain Bank clinical diagnostic criteria for Parkinson's disease (PD) and developed dementia according to the Diagnostic and Statistical Manual of Mental Disorders 4th edition (DSM IV) criteria at least 1 year after the onset of motor symptoms (PDD) or met the revised consensus operationalised criteria for DLB. Patients were assigned to a computer-generated randomisation list. All physicians who had contact with patients were masked to treatment allocation. The primary outcome measure was clinical global impression of change (CGIC), which ranged from 1 to 7 points, and a low score means a better outcome. Analysis was by intention to treat based on the last observation carried forward. This trial is registered, number ISRCTN89624516. FINDINGS: 72 patients with PDD or DLB were randomly assigned and started treatment: 34 with memantine and 38 with placebo. 56 (78%) completed the study. All withdrawals were owing to adverse events, but the proportion of withdrawals was similar in both groups. At week 24 the patients in the memantine group had better CGIC scores than those taking placebo (mean difference 0.7, 95% CI 0.04-1.39; p=0.03). With the exception of improved speed on attentional tasks in the memantine group (a quick test of cognition [AQT] form: difference 12.4, 95% CI 6.0-30.9; p=0.004), there were no significant differences between the groups in secondary outcome measures. INTERPRETATION: Patients with DLB or PDD might benefit from treatment with memantine, which was well tolerated. Large-scale studies are now required to confirm our preliminary findings. FUNDING: The Western Norway Regional Health Authority; H Lundbeck A/S.