Biomarkers of Alzheimer's disease modification using adaptive cognitive assessments to improve responsiveness-a simulation study.

Introduction: Clinical studies assessing cognition in Alzheimer's and other neurodegenerative diseases require endpoints that are sensitive to treatment response across a broad range of cognitive abilities. However, responsiveness of conventional cognitive assessments typically varies with the performance level, especially due to non-linearities such as floor or ceiling effects. Here, we evaluate 6 newly developed smartphone-based and gamified Adaptive Cognitive Assessments (ACAs) entailing a system of dynamic difficulty adaptation to individual performance, which is expected to improve adherence but also measurement properties. Deployment of such ACAs to maximize their discriminative ability in comparative studies requires exploration of many free parameters and complex dynamics. Methods: In simulations of cohorts of patients with cognitive impairment, we compared two ACAs paradigms: after 14 daily tests allowing performance-based difficulty adaptation, the difficulty level was either (1) fixed or (2) kept adaptive for a period of 4 years with weekly testing. Sensitivity to between-group effects was assessed in cohorts characterized by cognitive decline observed in neurodegenerative diseases. Results: The discriminative ability of the two paradigms depends on features of the study design and subjects. At study end, the adaptive difficulty paradigm clearly outperformed the fixed-difficulty paradigm in terms of responsiveness for cognitive decline rates >2.5% per year. Discussion: ACA can increase biomarker responsiveness to treatment effects over fixed difficulty. ACA deployment should be guided by study and assessment features, including duration, expected cognitive decline rates and effect size. In the high-dimensional parameter space of ACA instruments, study simulations are indispensable to identify suitable deployment strategies.