Sorry for the delay in responding to your post. You're correct that the percentage of 23& me customers with ApoE 4 is higher than would be expected based on population studies. For those who want the numbers, here's a slightly reformatted report:
23andMe® Personal Genome Service® (PGS)Genetic Health Risk Reports Package Insert
Published studies of people who don’t have Alzheimer’s disease estimate that 13-16% of people of European
descent, 18-23% of people of African American descent, 11-23% of people of Hispanic descent, and 7-14% of people of East Asian descent carry at least one copy of the ε4variant. Among people with Alzheimer’s disease, published studies estimate that 34-41% of people of European descent, 32-42% of people of African American descent, 19-32% of people of Hispanic descent, and 25-30% of people of East Asian descent carry at least one copy of the ε4 variant.
Frequency of the APOE ε4 variant in 23andMe customers
Variant name: ε4
African- American 34.10%
Ashkenazi Jewish 21.84%
East Asian 17.39%
Hispanic or Latino 22.44%
South Asian 17.16%
The insert data also claims that their tests have achieved about 99% accuracy, and they cite that the following "rare mutations" may affect the accuracy: may have rs11542041, rs573658040, or rs543363163. So I think there's a dissertation, or at least master's thesis in here somewhere for some aspiring sociologist or public health policy advocate to figure out what's driving this over-representation: family history of dementia? ancestral curiosity of people in the US whose ancestors came over from Ireland, the UK and northern Europe too long ago to know the family history? The latter was the case in my family; I was the only one who had heard of ApoE 4/4 and never expected to find out I had it, even with family history.