114 Chapter 5 complete the TMT-A with the mean TMT-B/A ratio from the respective diagnostic group in the respective cohort. To be included in the cognition analyses, participants were required to have at least 6 out of 9 cognitive tests available and 72 (91%) participants met this criterion (30 early- and 42 late-onset AD). Among the excluded cases there was one PCA case (early-onset AD group). Supplementary Figure 1 shows the missing values in the cognition-subsample for each neuropsychological test. There were on average 5% missing data in this subsample, for which we performed single imputation (with 5 iterations) for missing data, using all demographic (except for age and APOE4 status), neuropsychological and neuroimaging variables as predictors[42, 43]. Statistical analysis Differences in demographics and clinical characteristics between groups were assessed using independent sample t-tests for continuous variables and χ2 for dichotomous data. Differences in neuropsychological test-scores were assessed using ANOVA, adjusting for sex and education. To assess differences in [18F]flortaucipir BP ND and R1 between early- and late-onset AD groups, we used ANOVA with each of the lobar regions-of-interest (separately) as dependent variable. We repeated these analyses, but now excluding the atypical AD cases (n=8), to investigate whether potential age-dependent results were not driven solely by these extreme phenotypes. To assess more fine-grained differences, we additionally performed voxel-wise contrasts in SPM12 for BPND or R1 between early- and late-onset AD. All analyses were adjusted for sex. Results from voxel-wise analyses are displayed at both more liberal (i.e., p<0.001, uncorrected) and more stringent (p<0.05, Family Wise Error (FWE) corrected) thresholds. Next, we investigated whether associations between BPND or R1 and cognition differed between early- and late-onset AD. We performed linear regression analyses between BPND or R1 and neuropsychological tests (separate models), adjusted for sex and education, including the interaction term “age-at-onset (dichotomous)*BPND or R1”. Results were considered significant if p-values were ≤0.05 for regional analyses. We considered a p-value ≤0.10 significant for interaction terms[44]. Additionally, adjustment for multiple comparisons was performed using the Benjamini-Hochberg False Discovery Rate (FDR) method (indicated by pFDR)[45]. Statistical analyses were performed using R software, version 4.0.2, and the ‘mice’ package was used for imputation.
RkJQdWJsaXNoZXIy MjY0ODMw