The association between apolipoprotein E (APOE)-epsilon 4 and Alzheimer's disease has been confirmed worldwide. We and others have observed a diminished association in the very old and among African-Americans compared to Caucasians and Hispanics in New York. In this review we describe a new method we developed to compare relative risks by APOE genotypes in an expanded cohort of cases and controls from three ethnic groups in a New York City community and discuss the association as between APOE epsilon 4 and Alzheimer's disease as modified by head injury. Compared to persons with APOE epsilon 3/epsilon 3 genotypes, relative risk (RR) for Alzheimer's disease associated with APOE epsilon 4 homozygosity was similar across ethnic groups (African-American RR = 3.3; 95% c.i. 1.6-6.8; Caucasian RR = 5.3; 1.6-16.0; Hispanic RR = 2.5; 1.1-5.8). The risk was also increased for APOE epsilon 4 heterozygous Caucasians (RR = 3.2; 1.8-5.8) and Hispanics (RR 1.5; 1.0-2.2) but not African-Americans (RR = 0.6; 0.4-0.9). Risk of AD was not significantly diminished for individuals in any group with APOE epsilon 2/epsilon 2 or -epsilon 2/epsilon 3 genotypes. A 10-fold increase in the risk of Alzheimer's disease was associated with both APOE epsilon 4 and a history of traumatic head injury, compared to a twofold increase in risk with APOE epsilon 4 alone. Head injury in the absence of an APOE epsilon 4 allele did not increase risk. These results imply that in Alzheimer's disease genotypic risk associated with APOE may be influenced by age, ethnicity, and certain environmental factors.