One assumption of our model is that all subjects in each family are at risk of developing AD, because they belong to high-risk families that were selected for high prevalence of alcohol use disorders. Many people in these COGA families are aged between 12-25yrs. These individuals may be unaffected but are still at substantial risk of developing AD because they have not yet passed through the age of peak risk for developing AD. In contrast, there is little or no increase in power to be gained by using the age at onset of AD as a phenotype in population based cohorts or case-control datasets because the unaffected subjects in these datasets are generally older than the mean age at onset of AD, are not necessarily ascertained from high-risk alcoholic families and might not be exposed to predisposing genetic or environmental factors. We are not aware of any other dataset with a similar study design to the densely affected COGA families. This lack of a well-matched study design was the main reason we did not attempt to replicate the strongest signals.
