Caffeine stimulates central nervous system on a short term. However, the long-term impact of caffeine on cognition remains unclear. We aimed to study the association between coffee and/or tea consumption at midlife and dementia/Alzheimer's disease (AD) risk in late-life. Participants of the Cardiovascular Risk Factors, Aging and Dementia (CAIDE) study were randomly selected from the survivors of a population-based cohorts previously surveyed within the North Karelia Project and the FINMONICA study in 1972, 1977, 1982 or 1987 (midlife visit). After an average follow-up of 21 years, 1409 individuals (71%) aged 65 to 79 completed the re-examination in 1998. A total of 61 cases were identified as demented (48 with AD). Coffee drinkers at midlife had lower risk of dementia and AD later in life compared with those drinking no or only little coffee adjusted for demographic, lifestyle and vascular factors, apolipoprotein E epsilon4 allele and depressive symptoms. The lowest risk (65% decreased) was found in people who drank 3-5 cups per day. Tea drinking was relatively uncommon and was not associated with dementia/AD. Coffee drinking at midlife is associated with a decreased risk of dementia/AD later in life. This finding might open possibilities for prevention of dementia/AD.

Epidemiologists are increasingly looking to large-scale sample surveys to provide data for studies of the associations between known or suspected risk factors and disease. More often than not, widely available statistical software packages have been used to analyze such data, particularly when multivariable modeling is involved. Such packages assume that the data have resulted from simple random samples. However, when the survey design incorporates such features as clustering and stratification, the results of statistical analyses based on this assumption can be incorrect. The authors utilized data from the PAQUID (Personnes Agees Quid) study, collected periodically from 1988 to 1996, to illustrate the ease of performing a "design-based" (vs. a "model-based") analysis of complex survey data, and they compared the results obtained using both approaches. The PAQUID study is a stratified cluster sample of elderly community residents in the southwestern departments of Gironde and Dordogne, France. In the illustration presented-in which 3,777 community residents aged 65 years or older were selected to permit identification of baseline and lifetime factors that might be related to cognitive loss, dementia, and Alzheimer's disease--measures of association (such as odds ratios and their associated standard errors) were comparable for both analytical strategies. However, this may not be the case for other examples. Descriptive measures (such as estimates of means and proportions) may be more seriously compromised by the decision to ignore the sampling design. The availability of modern statistical packages with survey analysis capabilities should encourage data analysts to perform design-based analyses whenever possible.