Using Standardized Serum Creatinine Values in the Modification of Diet in Renal Disease Study Equation for Estimating Glomerular Filtration Rate

Context Guidelines recommend that laboratories estimate glomerular filtration rate (GFR) with equations that use serum creatinine level, age, sex, and ethnicity. Standardizing creatinine measurements across clinical laboratories should reduce variability in estimated GFR. Contribution Using standardized creatinine assays, the authors calibrated serum creatinine levels in 1628 patients whose GFR had been measured by urinary clearance of 125I-iothalamate. They used these data to derive new equations for estimating GFR and to measure their accuracy. The equations were inaccurate only when kidney function was near-normal. Cautions There was no independent sample of patients for measuring accuracy. Implications By using this equation and a standardized creatinine assay, different laboratories can report estimated GFR more uniformly and accurately. The Editors Chronic kidney disease is a recently recognized public health problem. Current guidelines define chronic kidney disease as kidney damage or a glomerular filtration rate (GFR) less than 60 mL/min per 1.73 m2 for 3 months or more, regardless of cause (13). Kidney damage is usually ascertained from markers, such as albuminuria. The GFR can be estimated from serum creatinine concentration and demographic and clinical variables, such as age, sex, ethnicity, and body size. The normal mean value for GFR in healthy young men and women is approximately 130 mL/min per 1.73 m2 and 120 mL/min per 1.73 m2, respectively, and declines by approximately 1 mL/min per 1.73 m2 per year after 40 years of age (4). To facilitate detection of chronic kidney disease, guidelines recommend that clinical laboratories compute and report estimated GFR by using estimating equations, such as equations derived from the Modification of Diet in Renal Disease (MDRD) Study (13, 510). The original MDRD Study equation was developed by using 1628 patients with predominantly nondiabetic kidney disease. It was based on 6 variables: age; sex; ethnicity; and serum levels of creatinine, urea, and albumin (11). Subsequently, a 4-variable equation consisting of age, sex, ethnicity, and serum creatinine levels was proposed to simplify clinical use (3, 12). This equation is now widely accepted, and many clinical laboratories are using it to report GFR estimates. Extensive evaluation of the MDRD Study equation shows good performance in populations with lower levels of GFR but variable performance in those with higher levels (1332). Variability among clinical laboratories in calibration of serum creatinine assays (33, 34) introduces error in GFR estimates, especially at high levels of GFR (35), and may account in part for the poorer performance in this range (13, 14, 16, 1821, 27, 30). The National Kidney Disease Education Program (NKDEP) has initiated a creatinine standardization program to improve and normalize serum creatinine results used in estimating equations (36). The MDRD Study equation has now been reexpressed for use with a standardized serum creatinine assay (37), allowing GFR estimates to be reported in clinical practice by using standardized serum creatinine and overcoming this limitation to the current use of GFR estimating equations. The purpose of this report is to describe the performance of the reexpressed 4-variable MDRD Study equation and compare it with the performance of the reexpressed 6-variable MDRD equation and the CockcroftGault equation (38), with particular attention to the level of GFR. This information should facilitate implementation of reporting and interpreting estimated GFR in clinical practice. Methods Laboratory Methods Urinary clearances of 125I-iothalamate after subcutaneous infusion were determined at clinical centers participating in the MDRD Study. Serum and urine 125I-iothalamate were assayed in a central laboratory. All serum creatinine values reported in this study are traceable to primary reference material at the National Institute of Standards and Technology (NIST), with assigned values based on isotope-dilution mass spectrometry. The serum creatinine samples from the MDRD Study were originally assayed from 1988 to 1994 in a central laboratory with the Beckman Synchron CX3 (Global Medical Instrumentation, Inc., Ramsey, Minnesota) by using a kinetic alkaline picrate method. Samples were reassayed in 2004 with the same instrument. The Beckman assay was calibrated to the Roche/Hitachi P module Creatinase Plus enzymatic assay (Roche Diagnostics, Basel, Switzerland), traceable to an isotope-dilution mass spectrometry assay at NIST (37, 39). On the basis of these results, the 4-variable and 6-variable MDRD Study equations were reexpressed for use with standardized serum creatinine assay. The CockcroftGault equation was not reexpressed because the original serum creatinine samples were not available for calibration to standardized serum creatinine assay. Derivation and Validation of the MDRD Study Equation The MDRD Study was a multicenter, randomized clinical trial of the effects of reduced dietary protein intake and strict blood pressure control on the progression of chronic kidney disease (40). The derivation of the MDRD Study equation has been described previously (11). Briefly, the equation was developed from data from 1628 patients enrolled during the baseline period. The GFR was computed as urinary clearance of 125I-iothalamate. Creatinine clearance was computed from creatinine excretion in a 24-hour urine collection and a single measurement of serum creatinine. Glomerular filtration rate and creatinine clearance were expressed per 1.73 m2 of body surface area. Ethnicity was assigned by study personnel, without explicit criteria, probably by examination of skin color. The MDRD Study equation was developed by using multiple linear regression to determine a set of variables that jointly estimated GFR in a random sample of 1070 patients (development data set). The regressions were performed on log-transformed data to reduce variability in differences between estimated and measured GFR at higher levels. Several equations were developed, and the performance of these equations was compared in the remaining sample of 558 patients (validation data set). To improve the accuracy of the final equations, the regression coefficients derived from the development data set were updated on the basis of data from all 1628 patients (11). Estimation of GFR Glomerular filtration rate was estimated by using the following 4 equations: the reexpressed 4-variable MDRD Study equation (GFR= 175standardized Scr 1.154age0.2031.212 [if black]0.742 [if female]), the reexpressed 6-variable MDRD Study equation (GFR= 161.5standardized Scr 0.999age0.176SUN0.17albumin0.3181.18 [if black]0.762 [if female]), the CockcroftGault equation adjusted for body surface area (Ccr= [140age]weight0.85 [if female]1.73/72 standardized ScrBSA), and the CockcroftGault equation adjusted for body surface area and corrected for the bias in the MDRD Study sample (Ccr= 0.8[140age]weight0.85 [if female]1.73/72 standardized ScrBSA). In these equations, GFR and creatinine clearance (Ccr) are expressed as mL/min per 1.73 m2, serum creatinine and urea nitrogen (SUN) are expressed as mg/dL, albumin is expressed as g/dL, weight is expressed as kg, age is expressed as years, and body surface area (BSA) is expressed as m2. Correction for bias improves performance of the CockcroftGault equation because it adjusts for systematic differences between studies, such as differences in the measures of kidney function (GFR in the MDRD Study and creatinine clearance in the study by Cockcroft and Gault), the serum creatinine assays, and the study samples. Hence, the bias correction for the CockcroftGault equation provided here reexpresses that equation for the estimation of GFR for use with standardized creatinine in study samples similar to that in the MDRD Study. Measures of Performance Measures of performance include bias (median difference of measured minus estimated GFR and measured GFR) and percentage bias (percentage of bias divided by measured GFR), precision (interquartile range of the difference between estimated and measured GFR, and percentage of variance in log-measured GFR explained by the regression model [R2 values]), and accuracy (percentage of estimates within 30% of the measured values). In the overall data set, bias is expected to be close to 0 for equations derived in the MDRD Study database, including the 4-variable and 6-variable equations and the CockcroftGault equation adjusted for bias. The bootstrap method (based on percentiles, with 2000 bootstrap samples) was used to estimate 95% CIs for interquartile ranges and R2 values. Confidence intervals for the percentage of estimates within 30% of measured values were computed by using the normal approximation to the binomial or exact binomial probabilities, as appropriate. We also computed sensitivity, specificity, positive and negative predictive value of estimated GFR less than 60 mL/min per 1.73 m2, and receiver-operating characteristic (ROC) curves by using measured GFR less than 60 mL/min per 1.73 m2 as the criterion standard. Areas under the ROC curves were compared by using the method of DeLong and colleagues (41). R, version 2 (Free Software Foundation, Inc., Boston, Massachusetts), and SAS, version 9.1 (SAS Institute, Inc., Cary, North Carolina), were used for statistical analysis. We used the lowess function in R to plot smoothed functions in the figures. Role of the Funding Source The study was funded by grants from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) as part of a cooperative agreement that gives the NIDDK substantial involvement in the design of the study and in the collection, analysis, and interpretation of the data. The NIDDK was not required to approve publication of the finished manuscript. The institutional review boards of all participating institutions approved the study. Results Clinical characteristics of