Impact of sample selection on APOE epsilon 4 allele frequency: a comparison of two Alzheimer's disease samples

In a highly selected sample of unrelated Alzheimer's disease (AD) patients, we found that the APOE epsilon 4 allele frequency was higher than previously reported. Differing selection and ascertainment criteria may lead to these differences. To address this possibility, we compared the epsilon 4 allele frequency in two samples of AD patients selected from the same geographical area.

Cases (n = 55) and controls (n = 99) from a research clinic-based sample were compared with subjects (n = 537) from a community-based AD patient sample. The samples consisted of unrelated cases who met NINCDS/ADRDA criteria for probable AD.

Clinical characteristics and APOE genotype data were obtained from AD cases and controls from both samples.

Frequency of APOE epsilon 4 allele in the research cases compared with the community cases (0.45 vs 0.36) was nearly significant. We compared demographic and clinical characteristics that might account for this difference and found that the research cases were younger, had an earlier age of onset, and had more advanced disease than the community cases. After onset age was controlled, there was no overall difference between epsilon 4 allele frequency of the two samples.

We found that the epsilon 4 allele frequency tended to be higher in the research AD sample compared the community-based sample. The two samples differed in several demographic and clinical characteristics. We conclude that research-based samples may lead to enrollment of younger patients with more severe disease who have higher APOE epsilon 4 allele load. This potential selection bias must be considered in the interpretation of studies of APOE allele frequency.