Life table methods were used to determine the relative risk of Alzheimer's disease (AD) in relatives of index cases with AD. Risk of AD was assessed in 967 first-degree relatives of 128 probands with clinically diagnosed AD and 572 first-degree relatives of a control group consisting of 84 subjects with Parkinson's disease (PD). Using a method that weights likelihood of correct diagnosis of AD, cumulative risk among AD relatives by age 93 was 24%, whereas relatives of the PD probands had a weighted risk of 16% by age 90. Overall, the total lifetime risk of developing dementia was similar among first-degree relatives of patients with AD and those of patients with PD. The age-specific risks were much different, however; from the age of 65 years to the age of 80 years, relatives of patients with AD had a twofold to fourfold increased risk of dementia. Equal risks were found for parents and siblings and for male and female relatives after adjustment for sex-specific patterns of survivorship. Relatives of probanwith early-onset (> 67 years) and late-onset (> 67 years) AD had equivalent risks of developing illness. Because onset age was found to cluster in families, the apparent increase in cumulative incidence for dementia in relatives of early-onset cases was likely due to relatives of the late-onset cases dying before developing illness. Our interpretation of these data is that a proportion of AD cases is caused by a fully penetrant gene transmitted as an autosomal dominant with age-dependent penetrance, whereas other cases, including some with multiple affected members, are clinically indistinguishable and of nongenetic cause.