Regional tau PET patterns predict prospective domain-specific cognitive decline in early symptomatic Alzheimer's disease.

Background: Tau-PET binding patterns are heterogenous, with regional binding showing strong cross-sectional correlations with domain-specific cognitive performance and longitudinal correlations with prospective neurodegeneration. Here we evaluated whether regional patterns of baseline tau-PET predict prospective longitudinal decline in specific cognitive domains in early symptomatic Alzheimer's disease (AD), including participants from clinical trial cohorts.

Methods: 731 amyloid-positive participants with a clinical diagnosis of mild cognitive impairment (MCI) or mild AD dementia underwent a flortaucipir F 18 PET (FTP-PET), structural MRI, and neuropsychological testing with the AD Assessment Scale-Cognitive Subscale (ADAS-Cog). Cognitive assessment was repeated after 9-18 or 12-24 months. Sub-scale annualized w-scores at each time point were combined as composite scores according to domains, including episodic memory, semantic memory, executive function, language and praxis. Latent growth curve models were applied to longitudinal composite scores to estimate the rate of annual decline (slope) in each participant. Standardized uptake value ratio (SUVR) images were created using cerebellar crus as the reference region. Regional and voxel-wise correlation analyses were implemented to identify baseline FTP-PET patterns and MRI grey matter volumes associated with longitudinal changes in each cognitive domain, and to evaluate whether MRI mediates the association between FTP-PET and cognitive decline.

Results: Differential FTP-PET signal patterns showed significant negative associations with domain-specific annual rates of decline. Higher temporo-parietal FTP-PET SUVR was associated with faster decline in episodic memory, while higher left anterior temporal SUVR was associated with faster decline in semantic memory. FTP-PET signal in a left-dominant fronto-temporal pattern was associated with faster decline in language, while FTP-PET signal in a right-dominant fronto-parietal pattern was associated with faster decline in praxis. Executive decline showed limited spatial associations with FTP-PET. Overall, regional and voxel-wise analyses identified similar pairwise associations between FTP-PET signal and domain-specific longitudinal decline. Baseline MRI showed weaker associations with domain-specific cognitive decline than FTP-PET, and did not mediate the predictive effect of the latter.

Conclusion: Differential regional tau-PET signal patterns were associated with domain-specific cognitive decline in MCI and early AD dementia. Tau-PET may be a useful precision medicine tool to support individualized predictions of cognitive decline trajectories in early symptomatic AD.