This paper sets a research agenda for the prediction and prevention of future onset of Alzheimer's disease (AD). From a MEDLINE review of the literature, the authors found age to be a predictor of AD. The literature also indicates that memory and attentional impairments predict AD, although the relative risk is relatively low. Late-onset depression may also predict AD, but these data are limited by a lack of cohort studies. Studying cognitively impaired subjects with late-onset depression may identify a high-risk group, facilitating prevention trials. Characteristics of an "ideal" preventive agent are suggested. There is a biologic rationale, and preliminary evidence, that non-steroidal anti-inflammatory drugs (including ASA), estrogen and vitamin E may play a preventive role in AD. Other compounds (such as acetylcholinesterase inhibitors) are also promising, but costs, side effects, and lack of other health benefits may preclude their use in all but very high-risk groups.