To control for diagnostic confounding by design, Model 1 was also re-estimated in the subset of cognitively normal (CN) participants from the NACC cohort. In a separate sensitivity analysis, we ...