If you read it, one of the first things they discuss is methodology for adjusting for age at death as it relates to Alzheimer's mortality, citing exactly this objection. I'm not a statistician and I don't know if their methodology is solid or not, but it's been addressed.