If feasible and meaningful, data were pooled by means of meta-analysis. Effect measures presented in this publication were reported as Mantel Haenszel risk ratios (RRs) including 95% confidence intervals (CIs) for binary data. Effects on rating scales were expressed as standardized mean differences (SMDs) with the 95% CIs, N and p values. A random-effects model was used to calculate a pooled effect estimate, because we expected considerable heterogeneity. When different dose levels were used in different study arms, the study arm with the higher dosage was chosen, and a sensitivity analysis was performed. Further sensitivity analyses were performed for type of dementia (Alzheimer's dementia versus vascular or mixed dementia) and exclusion of studies with a high number of participants without dementia.