Digital maze test reveals cognitive performance patterns associated with amyloid-β and tau.

OBJECTIVE: Digital cognitive tools provide novel ways to examine subtle cognitive processes not captured with traditional measures. This is particularly useful in examining subtle cognitive changes in early Alzheimer's disease (AD) stages. Here, we evaluate a digital maze test and its association to AD biomarkers in a largely cognitively unimpaired sample. METHOD: A total of 221 participants in the Harvard Aging Brain Study and affiliated studies completed a digital maze test, and neuropsychological battery including the trail making test. Cortical Amyloid-β and tau burden were visualized with C¹¹Pittsburgh compound B (PiB), and [¹⁸ F]Flortaucipir. Latent patterns among maze features were explored with principal component analysis and least absolute shrinkage and selection operator regression was applied to evaluate their associations with PiB and [¹⁸ F]Flortaucipir. Multiple regression models explored both maze features and trail making test variables in models predicting AD biomarkers. RESULTS: Principal component analysis results revealed maze performance is characterized by separate speed, path following, and impulsivity/planning components. Least absolute shrinkage and selection operator regression models revealed longer time duration (95% CI [5.31, 25.31]), greater pen-off-page time (CI [5.42, 16.81]), and pen stroke count (CI [.004, 0.021]) were associated with higher PiB. Higher pen stroke count (CI [0.003, 0.010]), and time deviating from the correct path (CI [0.0003, 0.002]) were associated with [¹⁸ F]Flortaucipir. In multiple regression models, specific maze features remained significant predictors of PiB beyond trail making test performance. CONCLUSION: Unique, digitally captured maze features are associated with AD pathology beyond what is seen by a traditional executive measure. This easy to administer, time-efficient digital tool provides unique information about amyloid and tau-related cognitive change. (PsycInfo Database Record (c) 2026 APA, all rights reserved).