Plasma proteomics predicts risk of incident chronic kidney disease in atrial fibrillation

Patients with atrial fibrillation (AF) have an increased risk of chronic kidney disease (CKD); however, plasma protein biomarkers for CKD prediction in this population remain poorly characterized. We aimed to identify proteins associated with AF and incident CKD, examine relevant pathways, and assess their value for CKD prediction in AF. This prospective analysis included 457,682 UK Biobank participants without CKD at baseline. Time-varying Cox proportional hazards models assessed the association between AF and incident CKD. In the proteomic subcohort, multivariable linear and Cox regression identified proteins associated with baseline AF and incident CKD, respectively. Weighted gene co-expression network analysis, pathway enrichment, and mediation analysis characterized protein modules and relevant pathways. Clinical, protein-only, and combined XGBoost Cox models were internally evaluated using repeated nested cross-validation among 868 participants with baseline AF. Performance at 5 and 10 years was assessed using the area under the receiver operating characteristic curve (AUC), calibration, prediction error, reclassification, and decision curve analysis. During a median follow-up of 13.6 years, 17,645 participants developed CKD. AF was associated with incident CKD after multivariable adjustment (hazard ratio [HR], 2.45; 95% confidence interval [CI], 2.35–2.56). Among 2,911 plasma proteins, 197 were associated with both baseline AF and incident CKD. Mediation analysis estimated that modules related to elastic fiber-vascular remodeling and to cytokine-vascular remodeling accounted for 22.9% and 10.5% of this association, respectively. Sequential feature selection yielded a nine-protein panel. The combined model achieved AUCs of 0.870 at 5 years and 0.832 at 10 years, exceeding the clinical model by 0.098 and 0.054, respectively. Reclassification metrics and estimated net benefits also favored the combined model. AF was associated with more than double the risk of incident CKD. Proteomic analyses highlighted extracellular matrix, vascular, cytokine, and cell matrix processes. Adding nine proteins to clinical factors improved CKD risk prediction at 5 and 10 years, although external validation is required.