A multi‐site study of Alzheimer’s disease neuroimaging and cognitive biomarkers in 2819 cognitively normal individuals

Abstract

AbstractBackgroundDeveloping effective Alzheimer’s disease (AD) biomarkers in cognitively normal (CN) individuals is crucial. Prior studies have identified neuroimaging and cognitive measures that are sensitive to preclinical AD. However, the generalizability of these biomarkers in large, multi‐site cohorts and their power to predict disease progression remain unclear. Structural MRI, tau PET and standardized cognitive tests of 2819 CN adults were pooled from three studies to investigate the power of these measures in discriminating β‐amyloid positivity (A+/‐) and predicting disease progression.MethodsT1‐weighted MRI and cognitive data of 2819 CN (A‐/A+: 1180/1540, Table 2) individuals from ADNI, HABS and A4 were included. Baseline medial temporal lobe structural measures were extracted from MRI (complete list in Table 1). For participants with prospective longitudinal MRI or cognitive measures (4.5 years), annualized change rate of each measurement was estimated. ANCOVA and receiver operating characteristic analyses were used to test biomarker differences between A+/A‐. In 563 CN with cross‐sectional tau PET available, stepwise linear mixed effect modeling was performed to identify the subset of baseline cross‐sectional biomarkers (tau, MRI and cognition) that yields the optimal model (smallest Akaike information criterion) in predicting longitudinal atrophy and cognitive decline. The area under the curve (AUC) of each model in discriminating the first (fast) and last (slow progressors) tertiles of each longitudinal measurement was reported.ResultsSignificant differences between A+/A‐ (Table 2) were observed in cross‐sectional tau (most significant), cross‐sectional and longitudinal cognitive, as well as longitudinal structural MRI measures. When predicting disease progression with baseline measures (Figure 1‐a), tau (selected in all the models) and cognitive biomarkers were included in the most predictive models while MRI biomarkers did not provide additional information. The selected biomarkers can identify fast versus slow progressors with AUCs ranging from 0.63‐0.78 (Figure 1‐b).ConclusionsThe results demonstrated that both cross‐sectional (tau PET and cognition) and longitudinal biomarkers (MRI and cognition) are sensitive to amyloid status in CN. In addition, baseline tau PET and cognitive measures provide complementary information in predicting disease progression regardless of amyloid status. This indicates that these biomarkers may have important utility in preclinical AD clinical trials and normal aging studies.