Digital cognitive phenotyping enhances risk stratification in preclinical Alzheimer's disease.

INTRODUCTION

Amyloid positivity does not reliably predict near-term cognitive decline in preclinical Alzheimer's disease (AD). We tested whether digital cognitive phenotyping improves prognostic precision.

METHODS

Baseline performance on a brief self-administered digital battery was analyzed in 1146 amyloid beta (Aβ)+ and 538 Aβ- cognitively unimpaired older adults from the Anti-Amyloid Treatment in Asymptomatic Alzheimer's/Longitudinal Evaluation of Amyloid Risk and Neurodegeneration studies. Unsupervised k-means clustering identified cognitive phenotypes. Survival models examined time to Preclinical Alzheimer Cognitive Composite decline and Clinical Dementia Rating (CDR) progression. Associations with tau positron emission tomography and plasma phosphorylated tau (p-tau)217 were evaluated.

RESULTS

Four cognitive phenotypes emerged. One phenotype (18% of Aβ+ participants) showed the steepest cognitive decline, highest progression on the CDR, greater baseline medial temporal tau, and fastest neocortical tau accumulation over 4.6 years (hazard ratio [HR] for decline ≈ 2.9). Combining this phenotype with elevated plasma p-tau217 identified a subgroup at markedly amplified risk (HR ≈ 4.5).

DISCUSSION: Digital cognitive phenotyping, especially when integrated with plasma biomarkers, identifies actionable risk heterogeneity and enables scalable precision prognostication in preclinical AD.