Chronic kidney disease classification according to different formulas and impact on adverse outcomes in patients with atrial fibrillation: A report from a prospective observational European registry

Abstract

Background: Chronic kidney disease (CKD) and atrial fibrillation (AF) often coexist, making accurate renal function estimation crucial, typically through equations calculating estimated glomerular filtration rate (eGFR) or creatinine clearance (CrCl). Objective: To compare the concordance and predictive performance of different renal function estimation equations in a European cohort of AF patients. Methods: We analyzed data from AF patients enrolled in a prospective observational European registry. Renal function was estimated using eight formulas: BIS-1, CG, CG-BSA, CKD-EPI, EKFC, FAS, LMR and MDRD. Concordance between formulas was assessed using weighted Cohen's Kappa, while Cox regression and receiver operating characteristic (ROC) curves evaluated their association with outcomes (composite of all-cause death, any coronary revascularization and any thromboembolism). Results: We included 8,506 patients. CKD-EPI demonstrated good to excellent concordance with other formulas, with the lowest concordance with CG (K = 0.607; 95% CI, 0.595-0.618) and the highest with MDRD (K = 0.880; 95% CI, 0.873-0.887). The risk of adverse outcomes increased sharply when renal function dropped below 60 ml/min across all formulas. CG-BSA and CG formulas showed the best discriminative ability for predicting composite outcomes (AUC 0.660, 95% CI 0.644-0.677, and 0.661, 95% CI 0.644-0.678, respectively). Based on integrated discrimination improvement (IDI) analysis, compared to the CKD-EPI equation, the CG and CG-BSA formulas showed significant improvements in sensitivity of 0.9% and 1.1%, respectively Conclusion: Equations for estimating renal function vary in concordance, with potential implications for drug prescription and predicting adverse events. CG and CG-BSA formulas showed superior performance in identifying patients at risk for adverse outcomes. © 2025 The Authors