Digital completion time relates more strongly to Alzheimer's disease‐type atrophy than to vascular burden in a community cohort
Kenichi Iwasa, Naoki Omori, Manabu Ishida, Tomotaka Tanaka, Atsushi NagaiAbstract
INTRODUCTION
Slowing on digital cognitive testing may reflect Alzheimer's disease (AD)‐type atrophy, small‐vessel disease, or both. Single‐outcome studies cannot determine whether slowing preferentially reflects one type of brain change or general brain health. We tested whether response‐time features from the tablet‐based Cognitive Assessment for Dementia, iPad version 2 (CADi2) were more strongly associated with AD‐type atrophy than with vascular burden.
METHODS
In 1163 community‐dwelling adults undergoing brain‐health screening (Izumo cohort, 2019–2025), CADi2 response‐time features predicted two primary MRI markers from one framework under repeated nested cross‐validation: Alzheimer's disease likelihood score (ADLS; similarity to an AD‐type atrophy pattern) and normalized white matter lesion (WML) volume. The primary metric was the incremental observed–predicted correlation (Δ R ) over age and sex. Silent brain infarcts, microbleeds, and a composite small vessel disease score were additional vascular outcomes.
RESULTS
CADi2 features improved prediction of ADLS (Δ R = +0.054, 95% confidence interval [CI] +0.010 to +0.095) but not WML (Δ R = −0.006). The improvement reflected completion time, not accuracy. Because predictive headroom differed, we repeated the comparison in residual space. There, the increment correlated about twice as strongly with ADLS as with WML ( r = 0.160 vs. 0.080; p = 0.044), although both remained significant. A parallel contrast against the composite vascular score gave Δ r = 0.100 ( p = 0.008).
DISCUSSION
Digital completion time was preferentially, but not exclusively, associated with AD‐type atrophy (neurodegeneration axis) relative to vascular burden. Because ADLS indexes an atrophy pattern rather than amyloid or tau pathology, these cross‐sectional findings characterize a structural phenotype.