This cross-sectional study evaluated four dementia clinical risk scores (CRSs) — mCAIDE, WHICAP, LIBRA, and CogDRisk — against AD endophenotypes (plasma biomarkers, neuroimaging, cognition) and pTau217/Aβ42-defined amyloid positivity in 2,627 adults (≥55 years) from the multiethnic HABS-HD cohort (23.5% Black/AA, 36.1% Hispanic/Latinx, 40.4% Non-Hispanic White).
CogDRisk showed the strongest, most consistent associations across all endophenotypes and pTau217/Aβ42 positivity (OR ~1.63 overall; 63% NHW, 60% LA, 76% AA increased odds), and the highest AUC for dementia (0.65) among CRSs. Each SD increase in CRS corresponded to a 35–73% increase in odds of dementia. No CRS outperformed the demographics-only model (AUC 0.70) or demographics+APOE model. mCAIDE performed worst and lacked plasma biomarker associations.
- Cross-sectional design prevents causal inference and cannot exclude reverse causality between CRS components and early AD pathology. - Some CRS components (diet, cognitive activity, atrial fibrillation, insomnia) were unavailable in HABS-HD and had to be omitted, potentially underestimating CRS performance. - Native American, Asian, and Pacific Islander groups were not included, limiting generalizability beyond Black, Hispanic/Latinx, and White populations.
CogDRisk is the best-performing off-the-shelf risk score for flagging AD-related biological risk across Black, Hispanic/Latinx, and White patients in community settings, but should be paired with plasma biomarkers (e.g., pTau217/Aβ42) and APOE status rather than used as a standalone predictor. CRSs are most useful as low-cost, first-line screening tools to identify who warrants further biomarker or genetic workup.
Explore related topics
