Apolipoprotein E type 4 allele (apoE epsilon4) is associated with Alzheimer's disease (AD) in the late-onset familial form and in sporadic cases, but the age-associated risk in a randomly sampled elderly population is not established. We examined the association of apoE epsilon4 with AD and other dementias (mainly multi-infarct or dementia following stroke) in 1,030 persons aged 71 to 100 years in the population-based Framingham Study cohort. Kaplan-Meier survival analysis revealed that 55% of the apoE epsilon4/epsilon4 homozygotes developed AD by age 80, whereas 27% of apoE epsilon3/epsilon4 heterozygotes developed AD by age 85, and 9% of those without a 4 allele developed AD by age 85 years. In comparison with persons without a 4 allele, the risk ration for AD was 3.7 (95% CI = 1.9 to 7.5) for apoE epsilon3/epsilon4 heterozygotes and 30.1 (95% CI = 10.7 to 84.4) for apoE epsilon4 homozygotes. ApoE epsilon2 (2/2, 2/3, or 2/4 genotypes) was associated with an absence of AD. One-half (n=21) of the 43 AD patients were either homozygous or heterozygous for apoE epsilon4. We found evidence for an association of apoE epsilon4 with other dementia, primarily multi-infarct dementia and stroke. The risk ratio was 2.3 (95% CI = 0.9 to 6.1) for non-AD dementias among persons with apoE epsilon3/epsilon4. Although the apoE epsilon4 allele is a potent risk factor for AD and may be associated with other forms of dementia, most apoE epsilon4 carriers do not develop dementia, and about one-half of AD is not apoE epsilon4 associated. The low positive predictive value of this marker (0.10) suggest that use of apoE genotyping as a screening test for AD is not supported.