Statistical Analysis

Data were analyzed using SPSS statistical software version 24.0 (IBM). Groupwise differences for categorical variables were determined using the Fisher exact or Pearson χ² tests. Continuous measures were compared using the t test unless otherwise stated. The associations of the intergenerational difference in AAO of AD (dependent variable) with acquired (ie, years of education; body mass index; history of cardiovascular disease, hypertension, hypercholesterolemia, diabetes, active depression within 2 years, traumatic brain injury, tobacco abuse, or unhealthy alcohol use; or retrospective determination of AAO) and heritable factors (ie, ethnicity/race, paternal or maternal history of dementia, parental history of early-onset dementia, APOE ε4 allele status, or AD PRS) were explored using stepwise multivariable linear regression (α = .05 for entry; α = .10 for removal), controlling for age and sex (forced entry). To assess for collinearity, variables of interest were included in a linear regression (forced entry) model and variance inflation factor quotients were computed. Variance inflation factors less than 5 were considered to indicate low probability of collinearity. Model explanatory power and fit were assessed using the adjusted R² and analysis of variance. Residual values were determined on a participant-by-participant basis by subtracting the expected intergenerational difference in AAO of AD (ie, that predicted from the model) from the observed intergenerational difference in AAO so that negative residuals identified participants whose observed intergenerational difference in AAO occurred earlier than expected and positive residuals participants whose intergenerational difference in AAO of AD occurred later than predicted. Z scores of residuals were determined, and participants were rank ordered. Participants with greater-than-expected residuals (ie, highest variability in intergenerational difference in AAO of AD not associated with measured variables) were defined as those with residuals more than 1 SD below (risk) or above (resilience) the mean. P values were 2-tailed, and statistical significance was defined as P less than .05 and Bonferoni corrected where appropriate.