Fig. 3
Proportion of variation in plasma NTA levels explained by Aβ, tau and neurodegeneration. Performance of different models for predicting plasma NTA-tau levels are shown in A. Each barplot represents one independent model, including CSF Aβ42/40 (A), tau-PET SUVR (T) and/or cortical thickness (N) as predictors in multivariable models. All models are adjusted for age and sex. Basic model includes only covariates. R2 values are shown inside the barplots and AICc of each model is shown on top. The optimal model predicting plasma NTA-tau was A&T because it had the highest R2 and the lowest AICc. Partial R2 of Aβ (CSF A β42/40) and tau (tau-PET) for predicting NTA-tau levels are shown in B (all participants, n = 1,294) and C (subsample with t-tau, n = 715). Light green bars represent the variance explained by tau and dark green bars represent that explained by Aβ. Other plasma biomarkers available are shown for comparison. Partial R2 values are shown inside the bars and percentage of the total R2 of the model are shown on top. In all cases, age and sex were used as covariates. NonAD participants were excluded from these analyses to avoid bias from neurodegeneration markers. Abbreviations: Aβ, amyloid-β; AICc, corrected Akaike criterion; CSF, cerebrospinal fluid; nonAD, non-Alzheimer’s type dementia; SUVR, standardized uptake value ratio