Memantine, a low affinity antagonist to glutamate NMDA receptors, may prevent excitatory neurotoxicity in dementia. To determine efficacy and safety of memantine for people with Alzheimer's disease (AD), vascular (VD) and mixed dementia. The Specialized Register of the Cochrane Dementia and Cognitive Improvement Group was searched on 28 October 2004. This register contains references from all major healthcare databases and many ongoing trial databases and is updated regularly. In addition the search engines Copernic and Google were used to identify unpublished trials through inspection of the websites of licensing bodies like the FDA , EMEA an Nice and of companies' websites (Lundbeck, Merz, Forest, Suntori etc). Double-blind, parallel group, placebo-controlled, randomized trials of memantine in people with dementia. Data were pooled where possible. Intention-to-treat (ITT) and observed case (OC) analyses are reported. 1. Moderate to severe AD. A major study (MD-01) is unpublished. Published data from two six month studies show a small beneficial effect of memantine at six months on cognition (4.12 SIB points, 95% Confidence Interval (CI) 2.14 to 5.74, P < 0.00001), activities of daily living (1.70 ADCS-ADLsev19 points, 95% CI 0.63 to 2.76, p = 0.002) and behaviour (3.64 NPI points, 95% CI 1.38 to 5.90, p = 0.002), supported by clinical impression of change (0.27 CIBIC+ points, 95% CI 0.10 to 0.43, p = 0.002). The unpublished study would need to have found a detrimental effect of memantine to overturn the statistical significance of the benefits apparent in the two published studies. 2. Mild to moderate AD. In a single six month trial, memantine had a beneficial effect on ITT analysis of cognition, (1.9 ADAS-Cog points, 95% CI 0.35 to 3.43, p = 0.02) and behaviour (3.50 NPI points 95% CI 0.15 to 6.85, p = 0.04) supported by clinical global impression of change (0.30 CIBIC+ points, 95% CI 0.09 to 0.51, p = 0.005), but no effect on activities of daily living or OC analysis of cognition. The statistical significance of these benefits could be overturned by data from two unpublished studies which are known to show no significant effect. 3. Mild to moderate vascular dementia. In two six month studies, memantine improved cognition (1.85 ADAS-Cog points, 95% CI 0.88 to 2.83, p = 0.0002), and behaviour (Weighted Mean Difference (WMD) 0.84 95% CI 0.06 to 0.91, p = 0.03) but this was not supported by clinical global measures.4. Patients taking memantine were less likely to develop agitation (Odds Ratio (OR): 0.65, 95% CI 0.48 to 0.89, p = 0.007). The effect on agitation which is already present is unknown.5. Memantine is well tolerated. Published data suggest a small beneficial effect of memantine at six months in moderate to severe AD. The beneficial effect on cognition in patients with mild to moderate vascular dementia was not clinically discernible at six months. Whether memantine has any effect in mild to moderate AD is unknown.