Tau-Clinical Mismatch Identifies Individuals with Co-Pathology and Predicts Clinical Trajectory
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Importance
The heterogeneous course of Alzheimer’s disease makes it difficult to predict individuals’ cognitive trajectories, which is particularly important in the era of disease modifying therapy. Identifying individuals more likely to have co-pathology and differing disease courses using clinically practical tools remains a critical gap.
Objective
To evaluate tau-clinical mismatch for identifying resilient and vulnerable individuals and compare levels of co-pathology and clinical trajectories between groups.
Design, Setting, and Participants
Participants were selected from the Alzheimer’s Disease Neuroimaging Initiative (ADNI, inclusion from 2005-2024), Penn Alzheimer’s Disease Research Center Cohort (Penn-ADRC, inclusion from 2002-2025), and Penn Anti-amyloid Therapy Monitoring (Penn-ATM) cohort (inclusion from 2024-2025). All participants were amyloid-β positive, had clinical assessment, and measures of Tau-PET or plasma p-tau217 available.
Exposures
Clinical assessment (CDR-SB) and tau burden (tau-PET or p-tau 217 ) for mismatch group classification.
Main Outcomes and Measures
Cross-sectional measures of neurodegeneration (medial temporal lobe volume and thickness, cortical thickness, TAR DNA-binding protein 43 [TDP-43] imaging signature), α-synuclein cerebrospinal fluid seed-amplification assay, longitudinal CDR-SB
Results
365 ADNI Tau-PET participants (ages 55-93, 52.6% women) and 524 ADNI p-tau 217 participants (ages 56-95, 49.0% women) were used to generate tau-clinical mismatch models with 55.6-57.1% classified as canonical (CDR-SB ∼ Tau), 23.7-24.7% as resilient (CDR-SB < Tau), and 19.3-19.7% as vulnerable (CDR-SB > Tau). Groups showed diverging clinical courses with earlier cognitive impairment seen in vulnerable groups and later impairment in resilient groups. Vulnerable groups showed higher frequencies of co-pathology, with TDP-43 neurodegeneration patterns and α-synuclein positivity. Similar findings were seen when applying these models to an independent dataset of 244 individuals (54-92 age, 57.0% women) in Penn-ADRC. Finally, these models were applied to a cohort receiving anti-amyloid therapy to show the utility of this method for predicting individual cognitive trajectories during therapy.
Conclusion and Relevance
Tau-clinical mismatch identifies individuals more likely to harbor co-pathology and have diverging clinical trajectories. Plasma-based models produced similar results to Tau-PET models and could be replicated in independent datasets. These models provide an important tool that can be implemented in clinical practice to provide improved individualized prognosis and, potentially, monitoring of response to disease-modifying therapy.
