Biomarker Variability Limits Individualized Amyloid Time Estimation in Alzheimer Disease

Area of Science:

• Neuroscience
• Biomarker Development
• Computational Biology

Background:

• Disease Progression Modeling (DPM) is crucial for staging Alzheimer's disease (AD).
• DPM accuracy relies on biomarker accumulation rates and reliability.
• Systematic assessment of variability's impact on DPM is needed for clinical applications.

Purpose of the Study:

• To assess the contributions of inter- and intra-individual variability to DPM performance.
• To evaluate DPM robustness across different amyloid biomarkers (PET, CSF, plasma).
• To inform the use of DPM for individual-level predictions in AD research and trials.

Main Methods:

• Simulation studies using empirical amyloid biomarker noise properties.
• Analysis of autosomal dominant and sporadic AD cohorts with PET, CSF, and plasma biomarkers.
• Group-level DPM performance assessed by Mean Average Error (MAE) and Root Mean Squared Error (RMSE).

Main Results:

• Inter-individual variability was the primary driver of temporal estimation errors.
• Intra-individual variability decreased the stability of DPM estimates.
• Amyloid PET demonstrated superior performance compared to CSF and plasma biomarkers in research data.

Conclusions:

• DPM performance is fundamentally limited by biomarker dynamic range and reliability.
• Current DPM methods are more robust at the group level.
• Biomarkers with higher variability (e.g., fluid biomarkers >10-15%) are more suitable for group-level DPM.