Group-level versus individualized approaches for capturing tau PET changes and clinical progression in Alzheimer's disease.

INTRODUCTION

Quantification approaches based on group-level regions of interest (ROIs) are widely used for evaluating tau positron emission tomography (PET) and clinical progression in Alzheimer's disease (AD), but we hypothesize that personalized approaches better capture inter-individual heterogeneity.

METHODS

We included 556 amyloid beta-positive individuals from seven cohorts (335 cognitively unimpaired [CU]; 221 cognitively impaired [CI]) with ≥ 2 [18F]flortaucipir (tau) PET scans. We compared three group-level and seven individualized tau PET quantification approaches (and a voxel-wise spatial extent metric) by (1) percentage change, (2) sensitivity to change, and (3) clinical associations.

RESULTS

Percentage change was highest in a composite temporal ROI (temporal-meta) in CU and CI. Individualized Braak stage-based (CU) and connectivity-informed (CU/CI) approaches, and tau spatial extent (CI), performed well for sensitivity to change and clinical associations, but temporal-meta was consistently a top performer.

DISCUSSION: Contrary to our hypothesis, temporal-meta quantification was most robustly associated with biological and clinical progression across disease stages, making it well suited for monitoring AD progression with tau PET.