Practical algorithms for amyloid β probability in subjective or mild cognitive impairment

Area of Science:

• Neurology
• Biomarker Discovery
• Alzheimer's Disease Research

Background:

• Alzheimer's disease diagnosis relies on identifying amyloid pathology.
• Patients with subjective cognitive decline or mild cognitive impairment require accurate tools for predicting amyloid presence.
• Clinical decisions for confirmatory biomarker testing can be improved with predictive algorithms.

Purpose of the Study:

• To develop and validate practical algorithms for predicting the probability of amyloid pathology.
• To aid clinical decision-making in patients with cognitive decline.

Main Methods:

• Algorithm development utilized data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the Australian Imaging, Biomarkers and Lifestyle (AIBL) study.
• Nested cross-validation with subsampling generated 1000 decision trees for internal validation.
• Algorithms were validated across ADNI, AIBL, and the Mayo Clinic Study of Aging.

Main Results:

• Two algorithms were created using age and normalized immediate recall z-scores, with optional inclusion of apolipoprotein E ε4 carrier status.
• Both algorithms demonstrated robust performance across diverse datasets.
• Performance remained strong even when alternative recall memory tests were used.

Conclusions:

• The developed statistical framework provides reliable probability estimation for amyloid pathology.
• These algorithms can assist clinicians in deciding on further diagnostic testing for Alzheimer's disease.
• Improved prediction of amyloid pathology supports timely and accurate diagnosis.