We evaluated the sensitivity, specificity, and predictive value of the apolipoprotein E (ApoE) epsilon4 allele for the neuropathological diagnosis of Alzheimer disease (AD) in a clinical series of well-characterized AD patients and controls followed longitudinally in the multicenter study of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD). In the 162 patients for whom autopsy and ApoE genotype data were available, the clinical diagnosis of AD was verified as the primary cause of dementia in 139 cases (139 of 162, or 86%). The sensitivity and specificity of the epsilon4 allele for AD were both 83%. The positive predictive value of the epsilon4 allele was very high at 97% (115 of 119); whereas, the negative predictive value was 44% (19 of 43), providing no useful information for diagnosis of AD when the epsilon4 allele is not present. Of the cases where AD was not considered the primary or sole cause of dementia (n = 23), 6 cases exhibited concomitant neuropathological findings of definite or probable AD and 2 of these 6 cases had at least one epsilon4 allele. These multicenter data extend previous observations reported from smaller case series of single laboratories and demonstrate that once the clinical diagnosis of AD is established, the presence of an epsilon4 allele reliably predicts the ultimate CERAD neuropathological diagnosis of AD. The findings also suggest that ApoE genotype information is useful clinically in bolstering diagnostic confidence when an epsilon4 allele is present and in identifying a group of patients for whom additional diagnostic evaluations may be warranted when the epsilon4 allele is absent.