Fig. 2: Associations with health-related phenotypes in LBC1936 for lifetime brain atrophy (LBA) and observed atrophic changes (N = 286).
No additional covariates were included. Atrophy estimated with the ratio and residual method were flipped to match the difference score whereby larger values represent more brain atrophy. Associations with continuous traits in panels (A, B) were calculated with linear regressions, where the outcome was always one of the traits listed on the y-axes and the predictor was either LBA or longitudinal atrophic changes. It may be counterintuitive to include APOEε4 status as an outcome, but we make no claims of causality or directionality and prioritised treating all outcome traits equally. Beta effect sizes indicate change per SD in the health trait (e.g., iCog). Percentages indicate variance explained (R2) in the health trait, and is only printed if the association is statistically significant (p < 0.05/ 15 traits; two-sided tests). Associations with binary traits in panel (C, D) were calculated with logistic regressions; R2 estimates were obtained with Nagelkerke’s R2. Odds ratios indicate the increased chances of having one of the diseases listed on the y-axis associated with one SD deviation in LBA (or observed atrophic changes). Only packyears was analysed with hurdle regression where R2 is inferred with maximum likelihood pseudo R2. Confidence intervals are at 95% and were calculated as beta ± 1.96*SE for continuous traits, and exp(beta ± 1.96*SE) for binary traits. Note when interpreting the results that the x-axes on panel C-D are scaled differently. Variables for which we extracted intercepts and slopes (Cog, Frailty, BMI) were relative to the same baseline at age 73 as observed atrophic changes, but LBA represents loss since maximum brain size many years earlier. Visually rated atrophy was assessed at age 76 (wave 3) which was the last available time point. Ideally, we would have preferred to include visually rated atrophy at age 82 (wave 5), so that LBA, the final time point of longitudinal atrophic changes, and visually rated atrophy would have been recorded at the same time point.
