Frailty-related plasma proteomic signatures and key biomarkers for new-onset dementia and its subtypes: A prospective cohort study.

Frailty is a well-established clinical risk factor for dementia, but its underlying mechanisms remain poorly defined. We aimed to investigate whether proteomic signatures and individual proteins linked to frailty could predict and characterize the association between frailty and dementia. We analyzed over 52,000 UK Biobank participants free of dementia at baseline, with plasma profiles of 2,915 proteins. Physical frailty (PF) and a 49-item frailty index (FI) were assessed. Proteomic signatures were derived using multivariable linear regression and 100 repeated LASSO selections. Incident dementia was ascertained via linkage to hospital and mortality records. Associations with incident dementia were assessed using multivariable Cox proportional hazards models, dose-response, and mediation analyses. During 13.6 years of follow-up, 1,437 participants developed dementia. Proteomic signatures for PF and FI were independently associated with dementia risk (hazard ratios [HRs] up to 2.80). Stratification by signature quintiles showed clear gradients in both absolute and relative dementia risk. For all-cause dementia, cumulative incidence ranged from 1.19% to 5.27% (HR up to 3.29), and for vascular dementia, from 0.16% to 1.19% (HR up to 4.53). Similar trends were observed for Alzheimer's dementia. Mediation analysis indicated that proteomic signatures statistically accounted for 46%-53% of the association between frailty and dementia risk. Key proteins, including GDF15, HPGDS, ITGAV, SPP1, CHGA, and LGALS4, were identified as top contributors and mapped to immune-inflammatory, neuroimmune, and extracellular matrix pathways. Frailty-related proteomic signatures and key proteins predict dementia risk and capture biological features beyond clinical frailty. These molecular markers may facilitate early risk stratification and support future research into prevention strategies.