OBJECTIVE: To examine the relation between apolipoprotein E status and risk of Alzheimer disease (AD) in a defined population and estimate the fraction of incident AD attributable to the epsilon4 allele. DESIGN: Community-based cohort study. SETTING: East Boston, Mass. PARTICIPANTS: A random sample of 578 community residents aged 65 years and older free of AD. MAIN OUTCOME MEASURE: Clinical diagnosis of AD by uniform, structured evaluation. RESULTS: The increased risk of AD associated with the presence of the epsilon4 allele was less than that found in most family and case-control studies. Persons with the epsilon4/epsilon4 or epsilon3/epsilon4 genotypes had 2.27 (95% confidence interval, 1.06-4.89) times the risk of incident disease compared with those with the epsilon3/epsilon3 genotype. The epsilon4 allele accounted for a fairly small fraction of the incidence of AD; if the allele did not exist or had no effect on disease risk, the incidence would be reduced by only 13.7%. The effect of the epsilon4 allele on risk of AD did not appear to vary with age. CONCLUSIONS: The apolipoprotein E epsilon4 allele is an important genetic risk factor for AD but accounts for a fairly small fraction of disease occurrence in this population-based study. Continued efforts to identify other environmental and genetic risk factors are warranted.