To analyse differences between sleep groups in demographic characteristics and concurrent health measures (n = 619), we used chi-square or Fisher’s exact test for categorical variables, analysis of variance (ANOVA) for continuous variables [mean (SD) reported] and Kruskal–Wallis tests for Likert-scale variables [median (Q1–Q3) reported]. Post hoc pairwise group differences at unadjusted P < 0.05 were reported. We excluded people who took insulin medication (n = 9) when comparing the HOMA-IR difference across the sleep groups (see Wallace et al.68 (link)). Linear regression was used to assess the relationship between sleep group and concurrent cognitive composite scores after adjusting for covariates [age, sex, education, WRAT3 reading score and the number of prior exposures to the cognitive tests (the practice effect)].
Similarly, to analyse differences between sleep groups and concurrent amyloid burden, we examined data from the subset that had completed at least one PiB PET study [n(%) = 108 (17.4%)]. Kruskal–Wallis tests were used to assess the difference between sleep groups in estimated concurrent global PiB DVR and amyloid chronicity, and Fisher’s exact test was used to analyse the concurrent amyloid PET status difference between sleep groups. In sensitivity analyses, we tested whether there was significant difference of amyloid burden at the most recent PET scan across the alternative sleep group assignments. In the imputed data set, 285 (23.0%) had at least one PiB PET scan, and we tested the difference in estimated concurrent and most recent global PiB DVR and amyloid chronicity among sleep groups.
We compared corrected Akaike information criteria (AICc) model fit statistics across otherwise identical models and considered |ΔAICc| values <2 to represent comparable models. Linear regression was performed for the association between sleep groups and concurrent cognitive composite scores after we removed stroke (n = 10), epilepsy/seizures (n = 13), multiple sclerosis (n = 5) and Parkinson’s disease (n = 2). Since APOE genotype associates with cognition, additional linear regression was performed including APOE e4 carriers in the model, and we compared model fits with the fits of the model in Aim 2 with the participants who have APOE data (n = 538). ΔAICc values were reported.
Similarly, to analyse differences between sleep groups and concurrent amyloid burden, we examined data from the subset that had completed at least one PiB PET study [n(%) = 108 (17.4%)]. Kruskal–Wallis tests were used to assess the difference between sleep groups in estimated concurrent global PiB DVR and amyloid chronicity, and Fisher’s exact test was used to analyse the concurrent amyloid PET status difference between sleep groups. In sensitivity analyses, we tested whether there was significant difference of amyloid burden at the most recent PET scan across the alternative sleep group assignments. In the imputed data set, 285 (23.0%) had at least one PiB PET scan, and we tested the difference in estimated concurrent and most recent global PiB DVR and amyloid chronicity among sleep groups.
We compared corrected Akaike information criteria (AICc) model fit statistics across otherwise identical models and considered |ΔAICc| values <2 to represent comparable models. Linear regression was performed for the association between sleep groups and concurrent cognitive composite scores after we removed stroke (n = 10), epilepsy/seizures (n = 13), multiple sclerosis (n = 5) and Parkinson’s disease (n = 2). Since APOE genotype associates with cognition, additional linear regression was performed including APOE e4 carriers in the model, and we compared model fits with the fits of the model in Aim 2 with the participants who have APOE data (n = 538). ΔAICc values were reported.
Free full text: Click here
