OBJECTIVE: To explore the association between high homocysteine levels and risk of Alzheimer disease (AD) in the Washington Heights-Inwood Columbia Aging Project (WHICAP). METHODS: The authors obtained fasting plasma samples in 909 elderly subjects chosen at random from a cohort of Medicare recipients; there was longitudinal data in 679 subjects without dementia at baseline who were followed for 3,206 person-years. Prevalent and incident dementia and its subtypes were diagnosed using standard methods. RESULTS: There were 128 persons with prevalent AD and 109 with incident AD in 3,206 person-years of follow-up. The adjusted OR of prevalent AD for the highest quartile of homocysteine compared to the lowest was 1.3 (95% CI = 0.7, 2.3; p for trend = 0.25). In longitudinal analyses, the authors found that the adjusted hazard ratio of AD for the highest quartile of homocysteine was 1.4 (95% CI = 0.8, 2.4; p for trend = 0.31). The authors also found that high homocysteine levels were not related to a decline in memory scores over time. Age was a significant confounder in all the analyses. The study had 80% power to detect a hazard ratio of 1.3 in the longitudinal analyses. CONCLUSION: High homocysteine levels were not associated with AD and were not related to a decrease in memory scores over time.