Comparison of the Chronic Kidney Disease Epidemiology Collaboration, the Modification of Diet in Renal Disease study and the Cockcroft-Gault equation in patients with heart failure

Karolina Szummer, Marie Evans, Juan Jesus Carrero, Urban Alehagen, Ulf Dahlström, Lina Benson, Lars H Lund

It is unknown how the creatinine-based renal function estimations differ for dose adjustment cut-offs and risk prediction in patients with heart failure. The renal function was similar with the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) (median 59 mL/min/1.73 m(2), IQR 42 to 77) and Modification of Diet in Renal Disease Study (MDRD) (59 mL/min/1.73 m(2), IQR 43 to 75) and slightly lower with the Cockcroft-Gault (CG) equation (57 mL/min, IQR 39 to 82). Across the commonly used renal function stages, the CKD-EPI and the MDRD classified patients into the same stage in 87.2% (kappa coefficient 0.83, p<0.001); the CKD-EPI and the CG equation agreed in 52.3% (kappa coefficient 0.39, p<0.001). Hence, a differing number of patients will receive dose adjustment depending on which formula is used as cut-off. The CG equation predicted worse prognosis better (c-statistics 0.740, 95% CI 0.734 to 0.746) than CKD-EPI (0.697, 95% CI 0.690 to 0.703, p<0.001) and MDRD (0.680, 95% CI 0.734 to 0.746). Using net reclassification improvement (NRI), the CG identified 12.8% more patients at higher risk of death as compared with the CKD-EPI equation. Patients registered in the Swedish Heart Failure Registry (n= 40 736) with standardised creatinine values between 2000 and 2012 had their renal function estimated with the CKD-EPI, the MDRD and the CG. Agreement between the formulas was compared for categories. Prediction of death was assessed with c-statistics and with NRI. The choice of renal function estimation formula has clinical implications and differing results at various cut-off levels. For prognosis, the CG predicts mortality better than the CKD-EPI and MDRD.