Comparison of Predictability of Future Cardiovascular Events Between Chronic Kidney Disease (CKD) Stage Based on CKD Epidemiology Collaboration Equation and That Based on Modification of Diet in Renal Disease Equation in the Japanese General Population

Abstract

Whether estimated glomerular filtration rate (eGFR) calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) Study equation (eGFRCKDEPI) improves risk prediction compared to that calculated using the Modification of Diet in Renal Disease (MDRD) study equation (eGFRMDRD) has not been examined in a prospective study in Japanese people. Participants (n=24,560) were divided into 4 stages (1, ≥90; 2, 60-89 (reference); 3a, 45-59; 3b+ <45 ml·min(-1)·1.73 m(-2)) according to eGFRCKDEPI or eGFRMDRD. Endpoints were all-cause death, myocardial infarction (MI) and stroke. Area under the receiver operating characteristic curves (95% confidence intervals) for predicting all-cause death, MI and stroke by eGFRCKDEPI vs. eGFRMDRD were 0.680 (0.662-0.697) vs. 0.582 (0.562-0.602); 0.718 (0.665-0.771) vs. 0.642 (0.581-0.703); and 0.656 (0.636-0.676) vs. 0.576 (0.553-0.599), respectively. Multivariate-adjusted Cox regression and Poisson regression analysis results were similar for adjusted incidence rates and adjusted hazard ratios in each corresponding stage between the 2 models and no differences were found in model assessment parameters. Net reclassification improvement (NRI) for predicting all-cause death, MI and stroke were estimated to be 6.7% (P<0.001), -1.89% (P=0.029) and -0.20% (P=0.421), respectively. Better discrimination was achieved using eGFRCKDEPI than eGFRMDRD on univariate analysis. NRI analysis indicated that the use of eGFRCKDEPI instead of eGFRMDRD offered a significant improvement in reclassification of death risk.