Extended Data Table 4.

Survival analyses for the full A^-T^- sample versus an age-matched A^-T^- sample. In this analysis, the A^-T^- sample was restricted to individuals older than 65 years (n = 485), so that there was no longer any significant age difference between the AT groups. The hazard ratios are derived from a Cox proportional-hazards model with clinical progression (progression to MCI or to all-cause dementia) as the outcome and age, sex, education and cohort as covariates; A⁻T⁻ served as the reference group. P values from two-sided statistical tests without adjustment for multiple comparisons are reported