Abstract
Early identification of Alzheimer's disease (AD) pathology is essential for timely intervention, particularly in primary care. We evaluated the diagnostic performance of a scalable, multimodal framework in a real-world, population-based cohort. A total of 277 community-dwelling individuals aged ≥ 60 years from the STOP-ALZHEIMER DEBA study (Basque Country, Spain) underwent brief cognitive screening (MMSE, M@T, Fototest, AD8) with optimized cut-offs, along with clinical risk assessment. Among them, 181 participants also completed structural MRI, plasma biomarker profiling (p-tau181, Aβ42/40, GFAP, NfL), and cerebrospinal fluid (CSF) analysis. We assessed performance for detecting cognitive impairment, CSF amyloid positivity (A+), and combined amyloid-tau positivity (A + T+). Optimized cognitive tests showed moderate accuracy (AUC 0.66-0.77), with the Fototest performing best. For biological outcomes, GFAP and p-tau181 had the highest predictive value (AUCs: 0.813 and 0.755 for A+; 0.852 and 0.710 for A + T+), and their combination further improved accuracy (AUC = 0.842). Fully adjusted models incorporating optimized cognitive scores, plasma biomarkers, APOE genotype, MRI, and demographics achieved high diagnostic performance (AUC = 0.886 for A+; 0.893 for A + T+). Results were consistent across sex and age strata. These findings support a stepwise diagnostic strategy combining brief, minimally invasive tools to enhance early AD detection in community settings.