Alzheimer's disease (AD) is the most common cause of dementia in the elderly. One of the most successful therapeutic strategies for Alzheimer's disease has been the use of acetylcholinesterase inhibitors to enhance surviving cholinergic neurotransmission by inhibiting breakdown of released acetylcholine. The first generation acetylcholinesterase inhibitors, such as tacrine, revealed major limitations to use including hepatotoxicity. Several second generation acetylcholinesterase inhibitors have now been introduced, including rivastigmine, which are believed to have superior proprieties. The mode of action and metabolism of rivastigmine suggest that it is unlikely to interact significantly with other medications. This is of particular relevance in elderly AD patients, the majority of whom are likely to be receiving concomitant medication. Large multi-centre trials have been completed in the USA, Canada, Europe and South Africa. Rivastigmine has received EU approval for use in all member states. It has approval in 30 countries but not the US. It is currently under review by the Food and Drug Administration, who requested additional analyses in 1998. To determine the clinical efficacy and safety of rivastigmine for patients with dementia of the Alzheimer's type. The Cochrane Controlled Trials Register, the Dementia Group Register of Clinical Trials, other electronic databases and other sources of reports were searched using the terms ENA 713, EXELON, and rivastigmine in addition to the terms for controlled trials in dementia (see the Group's search strategy for full details). All unconfounded, double-blind, randomised trials in which treatment with rivastigmine was administered for more than one day and compared to placebo for patients with dementia of the Alzheimer's type. Data were extracted by the reviewer (JSB) and entered into an appropriate meta-analysis. The data extracted were cross-checked by the second reviewer (VI). For each outcome measure, data were sought on every patient randomised. To allow an intention-to-treat analysis, the data were sought irrespective of compliance, whether or not the patient was subsequently deemed ineligible, or otherwise excluded from treatment or follow-up. If these data were not available, an analysis of data on patients who completed treatment was conducted. There are seven included trials. There are no published reports for two large phase III trials, B304 and B351, although they were completed more than 3 years ago. These are part of the Novartis ADENA programme and comprise 1379 (49%) out of 2803 phase III patients. It is unclear how missing data are replaced in ITT analyses, as reports from the ADENA programme provide no description of the use of this method. This has a profound effect on the results: if the method is substantially the same as LOCF, the benefits of treatment inferred from the analyses described in the publications as ITT, may be exaggerated. The meta-analysis reveals benefits on cognitive function as measured by ADAS-Cog test scores for the higher dose of rivastigmine compared to placebo at 26 weeks and for the lower dose. An additional analysis of ADAS-Cog dichotomised into those showing less than 4 points improvement and those showing 4 or more points improvement at 26 weeks shows benefit for cognitive function for the higher dose of rivastigmine compared to placebo and not for the lower dose. Global clinical state, dichotomised, counting those showing no change or decline, against those showing improvement shows benefit due to lower dose rivastigmine compared to placebo at 26 weeks and not for the higher dose. One trial reported results at 18 weeks and there are no significant differences between higher dose rivastigmine and placebo. One trial reported results at 13 weeks, and there are no significant differences between the 4 or 6 mg/d rivastigmine group and p