Frailty, social disconnection, metabolomic signatures, and incident peripheral artery disease in type 2 diabetes: a prospective cohort study

Abstract

Background Frailty and social disconnection may increase vascular risk in type 2 diabetes mellitus (T2DM), but their associations with peripheral artery disease (PAD) and related metabolic pathways remain unclear. Methods This secondary analysis of prospectively collected data from the UK Biobank included 22,582 participants with T2DM and without PAD at baseline. Frailty was defined using an adapted Fried phenotype, and social disconnection was assessed using measures of loneliness and social isolation. Cox proportional hazards models examined their separate and joint associations with incident PAD. Among 21,986 participants with nuclear magnetic resonance (NMR) metabolomic data, exposure-specific metabolomic risk scores (MRSs) were developed using least absolute shrinkage and selection operator Cox (LASSO-Cox) regression in a 70% training set and evaluated in a 30% validation set. Exploratory mediation analyses were performed in the validation set. Extreme gradient boosting (XGBoost) models with Shapley additive explanations (SHAP) were used to explore predictive patterns. Results During a mean follow-up of 14.43 years, 1,873 incident PAD cases occurred. In the mutually adjusted model, loneliness [hazard ratio (HR), 1.19; 95% confidence interval (CI), 1.03–1.38], social isolation (HR, 1.24; 95% CI, 1.10–1.41), pre-frailty (HR, 1.22; 95% CI, 1.10–1.36), and frailty (HR, 2.02; 95% CI, 1.75–2.34) remained associated with incident PAD. Participants with frailty and either loneliness or social isolation had the highest PAD risk. In the validation set, the frailty-, loneliness-, and social isolation-related metabolomic risk scores were associated with incident PAD, with HRs per standard deviation increase of 1.22 (95% CI, 1.11–1.34), 1.23 (95% CI, 1.12–1.34), and 1.22 (95% CI, 1.11–1.33), respectively. Exploratory mediation analyses indicated that these scores statistically accounted for 16.6, 20.4, and 22.6% of the corresponding associations with PAD, respectively. Conclusion Frailty, loneliness, and social isolation remained associated with incident PAD after mutual adjustment among individuals with T2DM. Training-derived metabolomic scores were associated with PAD in the validation set and statistically accounted for modest proportions of these associations. These findings support the potential relevance of physical vulnerability and social disconnection to PAD risk, while the metabolomic findings require external validation.