Biomarkers.

Area of Science:

• Neuroscience and Neurology
• Biomarker Discovery
• Epidemiology

Background:

• Urgent need for dementia risk models applicable to earlier disease stages.
• Blood-based biomarkers (BBMs) show promise but require validation in general populations.
• Potential for BBMs in early intervention, clinical trial access, and primary care diagnostics.

Purpose of the Study:

• To systematically review and meta-analyze the predictive performance of BBMs in longitudinal models.
• To assess the utility of BBMs for dementia prediction in primary care contexts.
• To evaluate BBM performance across different settings and population types.

Main Methods:

• Systematic review and meta-analysis of 32 studies evaluating longitudinal predictive performance of BBMs (p-tau181, p-tau217, p-tau231, Aβ, GFAP, NfL).
• Inclusion of pre-dementia populations reporting predictive performance using area under the receiver operating characteristic curve (AUROC).
• Sensitivity analyses on study setting, BBM combinations, baseline characteristics, and outcomes (disease conversion vs. cognitive decline).

Main Results:

• Pooled AUROC for dementia conversion was 0.89 in memory clinics and 0.85 in community settings.
• Models showed higher predictive performance for symptomatic versus asymptomatic individuals (AUROC 0.89 vs. 0.79).
• p-tau181 demonstrated high predictive accuracy, especially in single-marker models (AUROC 0.97) and for MCI to Alzheimer's dementia conversion (AUROC 0.91).

Conclusions:

• Longitudinal BBM models offer robust prediction for incident dementia and cognitive decline across heterogeneous settings.
• Early disease phases present opportunities for stratified prediction using BBMs.
• Further large-scale observational trials in primary care are essential for validating and optimizing BBM models for real-world use.