Age-dependent diagnostic and correlational architecture of multiplex plasma biomarkers in Alzheimer’s disease: a cross-ethnic, cross-platform validation study

BACKGROUND: Plasma phosphorylated tau-217 (p-tau217) is one of the most accurate blood-based biomarkers for Alzheimer's disease (AD), yet whether its diagnostic and correlational properties vary with age at onset has not been systematically examined. METHODS: We measured eight plasma biomarkers (tau, amyloid, neurodegeneration, neuroinflammation) via automated chemiluminescence immunoassay in a Chinese memory clinic cohort (n = 604; amyloid positron emission tomography (PET) reference), with validation in the Alzheimer's Disease Neuroimaging Initiative (ADNI) (n = 1,615; amyloid and tau PET). Diagnostic accuracy and biomarker-clinical correlations were compared between early-onset (EO < 65 years) and late-onset (LO ≥ 65 years) subgroups. RESULTS: Overall accuracy was high (primary area under the curve (AUC) 0.923; ADNI 0.904). In ADNI, p-tau217 showed higher accuracy in EO than LO (AUC 0.940 vs 0.892, P = 0.012), most pronounced at age 60 and absent by 70. The optimal biomarker shifted from tau-centric markers in LO to Aβ42/Aβ40 in EO. Tau biomarker-clinical correlations diverged sharply by onset age and were stronger in EO for cognitive and neuropsychiatric outcomes, whereas glial fibrillary acidic protein (GFAP) and neurofilament light chain (NfL) showed no age-dependent heterogeneity. Opposing-direction associations were observed for p-tau217%, which correlated negatively with onset age in AD but positively in Non-AD cognitive impairment. In an exploratory analysis, baseline p-tau217 was associated with longitudinal cognitive decline, numerically greater in EO though based on a small subgroup. Cross-ethnic generalizability was supported by the ADNI-Asian subgroup. CONCLUSIONS: This cross-ethnic, cross-platform study demonstrates that the diagnostic accuracy, biomarker-clinical correlation architecture, and optimal analyte selection of plasma p-tau217 vary systematically with age at onset, most markedly for tau-related measures. These findings challenge universal threshold paradigms and support age-stratified biomarker interpretation in clinical practice and therapeutic trial design.