Objective:

Background:

Preclinical AD is characterized by initial amyloid accumulation without cognitive symptoms. While functional connectivity changes have been observed, the relationship between AD pathology and the functional connectome during this stage remains unclear.

Design/Methods:

We applied Connectome-based Predictive Modeling (CPM), an approach that predicts outcome variables from the functional connectome, to baseline tau PET (18F-flortaucipir), amyloid PET (18F-florbetapir) and resting-state fMRI from the Anti-Amyloid in Asymptomatic Alzheimer's disease study (n=394, aged 65-85). Model performances were assessed using Spearman's correlations (⍴) of predicted vs. observed values. Model significance was assessed against permuted models (n=1000 iterations), corrected for the false-discovery rate. We characterized models and assessed generalizability using an external symptomatic AD cohort (ADNI, n=469, aged 55-90).

Results:

Whole-brain functional connectivity robustly predicted regional tau PET, outperforming amyloid PET models. The best-performing models were for regions associated with Braak stages IV/V (posterior cingulate ⍴=0.30, precuneus ⍴ = 0.22; p<0.05), while models for regions first impacted by tau accumulation performed poorly (parahippocampal ⍴=0.05, entorhinal ⍴=0.06, fusiform ⍴=0.09; p>0.05). The association between connectivity and regional tau is predominantly linear and prediction accuracies were robust to temporal and default mode network node lesions. Tau model accuracies correlated with global connectivity of brain regions (⍴=0.54, p<0.05) rather than underlying tau burden (⍴=0.17, p>0.05). Models generalized to ADNI, particularly in individuals with elevated tau (SUVR >1.19).

Conclusions:

Whole-brain functional connectivity predicts tau PET in preclinical AD and generalizes to a clinical dataset with abnormal tau PET, highlighting the functional connectome's potential as a biomarker for early AD detection and monitoring.