Performance of chronic kidney disease epidemiology collaboration creatinine-cystatin C equation for estimating kidney function in cirrhosis

Abstract

Conventional creatinine-based glomerular filtration rate (GFR) equations are insufficiently accurate for estimating GFR in cirrhosis. The Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) recently proposed an equation to estimate GFR in subjects without cirrhosis using both serum creatinine and cystatin C levels. Performance of the new CKD-EPI creatinine-cystatin C equation (2012) was superior to previous creatinine- or cystatin C-based GFR equations. To evaluate the performance of the CKD-EPI creatinine-cystatin C equation in subjects with cirrhosis, we compared it to GFR measured by nonradiolabeled iothalamate plasma clearance (mGFR) in 72 subjects with cirrhosis. We compared the "bias," "precision," and "accuracy" of the new CKD-EPI creatinine-cystatin C equation to that of 24-hour urinary creatinine clearance (CrCl), Cockcroft-Gault (CG), and previously reported creatinine- and/or cystatin C-based GFR-estimating equations. Accuracy of CKD-EPI creatinine-cystatin C equation as quantified by root mean squared error of difference scores (differences between mGFR and estimated GFR [eGFR] or between mGFR and CrCl, or between mGFR and CG equation for each subject) (RMSE = 23.56) was significantly better than that of CrCl (37.69, P = 0.001), CG (RMSE = 36.12, P = 0.002), and GFR-estimating equations based on cystatin C only. Its accuracy as quantified by percentage of eGFRs that differed by greater than 30% with respect to mGFR was significantly better compared to CrCl (P = 0.024), CG (P = 0.0001), 4-variable MDRD (P = 0.027), and CKD-EPI creatinine 2009 (P = 0.012) equations. However, for 23.61% of the subjects, GFR estimated by CKD-EPI creatinine-cystatin C equation differed from the mGFR by more than 30%.

Conclusion: The diagnostic performance of CKD-EPI creatinine-cystatin C equation (2012) in patients with cirrhosis was superior to conventional equations in clinical practice for estimating GFR. However, its diagnostic performance was substantially worse than reported in subjects without cirrhosis.

Figure 1

Side-by-side boxplots of the difference scores [(DS)=mGFR-eGFR (or the differences between mGFR and CrCl, or between mGFR and CrCl estimated by CG equation) for each subject] for CrCl, CG and GFR-estimating equations; circles in each box show mean difference score (bias); black bars in each box show median difference score.