SuO007THE CKD-EPI CREATININE FORMULA IS A SUPERIOR PREDICTOR OF END STAGE KIDNEY DISEASE THAN IOHEXOL PLASMA CLEARANCE: RESULTS FROM THE SWEDISH RENAL REGISTRY

Abstract

Introduction and Aims: It has been suggested that estimated GFR becomes less reliable in patients with advanced kidney disease due to loss of muscle mass and reduced creatinine generation. Iohexol plasma clearance is a reliable exogenous measure of GFR. The aim of this study was to compare the ability of estimated GFR (eGFR) versus measured GFR (mGFR) to predict end stage kidney disease (ESKD) in patients with CKD, both as an aetiological risk factor and for patient prognostication. Methods: From the Swedish Renal Registry (SRR-CKD) we identified patients >18 years with CKD stage 4 or 5 who, between 1st January 2005 and 31st December 2011 had a mGFR performed using iohexol plasma clearance during a nephrology clinic visit, along with a measure of serum creatinine. We performed data linkage with the national inpatient register to obtain comorbidity data, and the dialysis and death registers to obtain information on start of dialysis or vital status until 30th September 2013. Multiple imputation was performed for missing data (<10%). eGFR and mGFR were log-transformed and then standardised to allow fair comparison. A Cox proportional hazards model was constructed for the composite endpoint of end stage kidney disease (commencement of renal replacement therapy or death from kidney failure) Age, sex, primary renal diagnosis, mean arterial pressure, haemoglobin, albumin and Charlson Score were covariates. Measures of discrimination, calibration and re-classification were performed. Results: 1517 patients were included with a mean age of 68±14 years, 65% male and 37% had a diagnosis of diabetes. Median eGFR using CKD-EPI creatinine was 20ml/min/ 1.73m2 (interquartile range [IQR] 14 - 27) andmGFR was 18 (IQR 13 - 23). Median follow-up was 45 months (26-59), registering 471 deaths (31%). In Cox models, adjusted hazard ratio for eGFR CKD-EPI was 0.36 (95% Confidence intervals [CI] 0.33 - 0.40) and mGFR was 0.39 (95%CI 0.36-0.43). This means that for a rise of one standard deviation in (log)eGFR and (log)mGFR there was a 64% versus 61% fall in ESKD respectively. The C statistic (measure of discrimination) for 2-yr risk of ESKD was statistically significantly higher for eGFR (0.81, 95% CI 0.79-0.83) vs mGFR (0.76, 95%CI 0.73-0.78) (p value for difference <0.0001 using Hanley and McNeillmethod). Calibration was measured using the Hosmer-Lemeshow test (non-significant p value shows evidence of calibration); eGFR p=0.764, mGFR p=0.155. Net reclassification index showed a significant improvement of 20.7 using eGFR vsmGFR to predict ESKD (p<0.0001). Conclusions: In the aetiological Cox model, changes in eGFR and mGFR were associated with a similar risk of ESKD. However in the prognostic models, the performance of eGFR was superior to mGFR in predicting ESKD in adult patients with CKD stage 4 and 5. Estimation of GFR using creatinine-based formulae is considerably more convenient for patients and cheaper than formal measurement of GFR and these results support the continued use of this strategy for prognostication in patients with CKD.