Alzheimer's disease (AD) lesions are characterized by the presence of numerous inflammatory proteins. This has led to the hypothesis that brain inflammation is a cause of neuronal injury in AD and that anti-inflammatory drugs may act as protective agents. Seventeen epidemiologic studies from nine different countries have now been published in which arthritis, a major indication for the use of anti-inflammatory drugs, or anti-inflammatory drugs themselves have been considered as risk factors for AD. Both factors appear to be associated with a reduced prevalence of AD. The small size of most studies has limited their individual statistical significance, but similarities in design have made it possible to evaluate combined results. We have used established methods of statistical meta-analysis to estimate the overall chance of individuals exposed to arthritis or anti-inflammatory drugs developing AD as compared with the general population. Seven case-control studies with arthritis as the risk factor yielded an overall odds ratio of 0.556 (p < 0.0001), while four case-control studies with steroids yielded odds ratios of 0.656 (p = 0.049) and three case-control studies with nonsteroidal anti-inflammatory drugs (NSAIDs) yielded an odds ratio of 0.496 (p = 0.0002). When NSAIDs and steroids were combined into a single category of anti-inflammatory drugs, the odds ratio was 0.556 (p < 0.0001). Population-based studies were less similar in design than case-control studies, complicating the process of applying statistical meta-analytical techniques. Nevertheless, population-based studies with rheumatoid arthritis and NSAID use as risk factors strongly supported the results of case-control studies. These data suggest anti-inflammatory drugs may have a protective effect against AD. Controlled clinical trials will be necessary to test this possibility.