Predicting CDR status over 36 months with a recall-based digital cognitive biomarker

Area of Science:

• Neuroscience
• Cognitive Science
• Biomarker Discovery

Background:

• Word-list recall tests are standard for cognitive assessment.
• Process scoring may enhance the accuracy of these cognitive tests.
• Alzheimer's Disease Assessment Scale-Cognitive subscale (ADAS-Cog) is a common cognitive assessment tool.

Purpose of the Study:

• To evaluate if process-based digital cognitive biomarkers (DCBs) derived from ADAS-Cog predict longitudinal Clinical Dementia Rating (CDR) progression.
• To compare the predictive accuracy of DCBs against standard ADAS-Cog scores.

Main Methods:

• Utilized Alzheimer's Disease Neuroimaging Initiative (ADNI) data from 330 participants.
• Performed regression analyses predicting CDR at 36 months.
• Controlled for demographics and genetic risk, using ADAS-Cog scores and DCBs as predictors.

Main Results:

• Latent recall ability (M), a DCB, was the strongest predictor of CDR at 36 months (AUC = 0.84).
• M significantly outperformed traditional ADAS-Cog scores in predicting dementia decline.
• The top DCB model demonstrated substantially higher predictive power (odds ≈ 90x) than the top ADAS-Cog model.

Conclusions:

• Process scoring and latent modeling offer superior predictive accuracy compared to traditional scoring for cognitive tests.
• The DCB 'M' shows potential for identifying individuals unlikely to experience cognitive decline.
• Further research is warranted to explore the applicability of these DCBs across different tests and populations.