Suitability of the IDMS-Traceable MDRD Equation Method to Estimate GFR in Early Postoperative Renal Transplant RecipientsYeo Y.a · Han D.-J.b · Moon D.H.c · Park J.S.a · Yang W.S.a · Chang J.W.a · Byun S.W.a · Park S.-K.a
aDivision of Nephrology, Department of Internal Medicine, bDepartment of Surgery and cDepartment of Nuclear Medicine, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea Corresponding Author
Background/Aims: Accurate measurement of glomerular filtration rate (GFR) is critical for the management of kidney transplant recipients. Comparison of creatinine and cystatin C in renal transplant recipients gave conflicting results. We aimed to compare the performance of creatinine- and cystatin C-based equations and creatinine clearance in 102 early postoperative Korean renal transplant patients. Methods: We measured 51Cr-EDTA clearance using a 2-compartment model and considered this the reference GFR. Then, we estimated GFR using 13 creatinine- and 7 cystatin C-based equations. Serum creatinine value was calibrated to isotope-dilution mass spectrometry (IDMS). Results: The mean reference GFR was 76.77 ± 17.01 ml/min/1.73 m2. The IDMS-traceable MDRD (IDMS-MDRD) equation had the highest accuracy (94.12 within 30% of the reference; 99.02 within 50% of the reference) with a bias of 0.33 ml/min/1.73 m2 and a precision of 12.57 ml/min/1.73 m2. The Mayo Clinic equation also performed well (92.16% within 30% of the reference; 99.02% within 50% of the reference; bias: –0.19 ml/min/1.73 m2). As for cystatin C-based equations, the Filler equation had the least bias (0.03 ml/min/1.73 m2) but low accuracy (78.43% within 30% of the reference). Conclusions: We conclude that the IDMS-MDRD equation provided the best estimate of GFR in our early postoperative Korean renal transplant patients.
Copyright © 2009 S. Karger AG, Basel
An accurate measurement of glomerular filtration rate (GFR) is essential for the management of kidney transplantation recipients. The best measure to determine GFR is renal clearance of inulin . However, inulin infusions are not often used in clinical practice because inulin is expensive, commercial sources are limited, continuous intravenous infusions are required and timed urine collection by bladder catheterization is necessary.
Other exogenous substances, such as 51chromium-ethylenediaminetetraacetic acid (51Cr-EDTA), 125I-iothalamate or iohexol have been successfully used as markers for measurement of GFR [2,3]. However, these markers are not easily used in daily clinical practice because they are expensive, administration is invasive and measurements are time-consuming.
There are well-known limitations of using serum creatinine alone and creatinine clearance for the estimation of GFR . Serum creatinine is affected by several factors that are independent of GFR, such as age, race, muscle mass, gender, medication use and catabolic state. The measurement of creatinine clearance is inaccurate because of collection errors and changes in creatinine excretion [4,5].
In order to estimate GFR accurately and simply, numerous creatinine-based equations have been developed, mainly for the non-transplant population. The commonly recommended Modification of Diet in Renal Disease (MDRD) formula has been validated in patients with different renal diseases , but its diagnostic performance has not been shown to be reliable in renal transplant recipients [7,8,9].
Cystatin C can also be used as a biomarker for GFR measurement [10,11,12]. Cystatin C is a small, nonglycosy-lated basic protein (13 kDa) that is produced at a constant rate by all nucleated cells, is freely filtered by the glo- merulus, and is almost completely reabsorbed. Recently, several formulae for estimating GFR based on cystatin C also have been suggested [13,14,15,16,17,18,19]. However, the comparison of creatinine and cystatin C in renal transplant patients gave conflicting results [9,20,21,22,23].
Race is an important determinant of GFR estimation. The MDRD equations originally were developed from measurements in Caucasian and African-American patients. It has been suggested that the MDRD equations be modified for Asian population [24,25]. Almost all studies that have compared GFR estimations using serum creatinine and cystatin C in renal transplant recipients enrolled only Caucasian subjects.
Based on an improved knowledge of the impact of creatinine calibration on GFR estimation with a creatinine-based equation, Levey et al. [26 ] re-expressed the MDRD equation based on isotope-dilution mass spectrometry (IDMS), the gold standard of creatinine determina- tion. The National Kidney Disease Education Program (NKDEP) recommends that clinical laboratories using creatinine methods that have been calibrated to IDMS should use the IDMS-traceable MDRD (IDMS-MDRD) equation .
In this study, we compared the performance of creatinine-based equations, including the IDMS-MDRD equation, cystatin C-based equations, and creatinine clearance for measurement of GFR in early postoperative Korean renal transplant recipients.
Materials and Methods
From October 2007 to September 2008, a total of 193 Korean adult patients underwent renal transplantation at the Asan Medical Center. Among them, we enrolled 102 consecutive patients, all of whom had stable renal function. After creatinine levels reached to nadir, and the creatinine variability was less than 10% within 3 to 5 days, we measured the GFR using 51Cr-EDTA before discharge. We excluded patients who were unable or unwilling to provided informed consent or who had unstable renal function due to acute rejection, infection or calcineurin inhibitor toxicity. The protocol of this prospective study was approved by Institutional Review Board of the Asan Medical Center and written informed consent was obtained from all patients.
We determined the GFR by measuring the plasma clearance of 51Cr-EDTA using the 2-compartment model, which requires multiple blood samples. Although a simplified method, such as a single compartment model that requires only 2–3 blood samples, has been proposed, we used the 2-compartment model because it provides more accurate measurements [3,28].
After establishing intravenous access, a dose of 50 µCi 51Cr-EDTA was injected. Six plasma samples were taken at 10, 20, 30, 60, 180 and 300 min. Serum 51Cr-EDTA concentration was measured by a radio-immunoassay kit (Cobe Laboratories, Lakewood, Colo., USA).
GFR was corrected for the standard body surface area by multiplying the measured value by 1.73 and dividing by the patient’s body surface area, as estimated by the Du Bois formula .
On the day of 51Cr–EDTA GFR measurement, patients were weighed and a blood sample was taken for determination of serum creatinine, blood urea nitrogen, and cystatin C. The 24-hour urine collection for the measurement of creatinine clearance was completed by the same day. Creatinine clearance was corrected for standard body surface area. Serum creatinine was determined with a Toshiba 200-FR Neo (Toshiba Medical Systems Co., Ltd., Tokyo, Japan), using an assay based on the rate-blanked compensated kinetic Jaffe method. We used a Roche calibrator (Roche Diagnostics, Indianapolis, Ind., USA) that was traceable to the IDMS reference method. The intra-assay coefficient of variation (CV) and the interassay CV were 0.9–1.7% and 1.4–2.1%, respectively.
Cystatin C was measured by particle-enhanced immuno-nephelometric immunoassay (Behring Nephelometry II, Dade Behring Diagnostics, Marburg, Germany).
|Table 1. Formulas used to estimate GFR based on measurements of serum creatinine and cystatin C|
We validated versions of the MDRD equations that were modified for Japanese and Chinese patients [24,25]. The Mayo Clinic equation and the Rule’s refitted MDRD equation are based on creatinine value determined at the Mayo Clinic laboratory. Therefore, creatinine values were converted using the following equation .
CreatinineMayo Clinic = (creatinine IDMS/0.95 mg/dl) + 0.23 mg/dl
All cystatin C formulae are based on the Dade-Behring assay. All GFR estimates were expressed as ml/min/1.73 m2.
We used SPSS version 14.0 (SPSS Inc., Chicago, Ill., USA) and Medcalc software (Mariakerke, Belgium) for all statistical analyses. Correlations between the reference and estimated GFR were assessed using Pearson’s correlation coefficient.
Evaluation of the prediction equations was performed by calculating bias, precision, relative difference, accuracy, and agreement, as recommended in the National Kidney Foundation guidelines on chronic kidney disease . Bias, a measure of systematic error, was defined as the mean difference between the reference GFR (51Cr-EDTA clearance) and the estimated GFR. Pair-wise comparison of the mean difference was performed using paired t test. Relative difference was calculated as the percentage difference from the reference GFR. Precision was defined as the standard deviation (SD) of the difference between the reference and the estimated GFR. An F test was performed to compare the SD of the mean difference. Accuracy was measured as the proportion of GFR estimates that were within 30% and 50% of true GFR. Limits of agreement between the estimated and the reference GFR were calculated from the mean difference ± 1.96 SD of the differences and presented as Bland-Altman plots, performed with Medcalc software . A p < 0.05 was considered statistically significant.
Table 2 shows baseline characteristics of the 102 patients. The mean creatinine concentration was 1.01 ± 0.26 mg/dl (range: 0.5–1.8 mg/dl) and the mean cystatin C concentration was 1.26 ± 0.37 mg/l (range: 0.64–2.51 mg/l). The mean GFR measured by 51Cr-EDTA was 76.77 ± 17.01 ml/min/1.73 m2 (range: 41.4–124.5 ml/min/ 1.73 m2). The immunosuppression regimen consisted of cyclosporine A in 40 patients and tacrolimus in 62 patients. Calcineurin inhibitor treatment was combined with mycophenolate mofetil in 46 patients and with azathioprine in 45 patients. All patients received corticosteroids except 8 patients who received thymoglobulin as an immunosuppressive agent for a separate trial. Mean corticosteroid dosage was 23.5 ± 12.6 mg per day. All patients were given prophylactic doses of trimethoprim/sulfamethoxazole.
|Table 2. Baseline patient characteristics|
All tests of GFR showed good correlation with measurements of 51Cr-EDTA clearance (p < 0.001). The correlation coefficients (r) ranged from 0.592 to 0.777.
Table 3 shows the bias, precision, relative difference and accuracy of the 20 prediction equations. The paired t test indicated that all GFR estimates differed significantly (p < 0.05) from 51Cr-EDTA clearance, except the IDMS-MDRD estimate, the Mayo Clinic estimate, the Walser estimate, and the Filler estimate. The abbreviated MDRD equation significantly overestimated the measured GFR by 5.17 ml/min/1.73 m2 (p< 0.001). The IDMS-MDRD estimate, the Mayo Clinic estimate and the Filler estimate had the least bias of 0.33, –0.19 and 0.03 ml/min/1.73 m2, respectively.
|Table 3. Bias, precision, accuracy, and relative difference of creatinine and cystatin C estimates|
The Japanese Society of Nephrology-Chronic Kidney Disease Initiatives equation (JSN-CKDI) had the best precision (10.89 ml/min/1.73 m2). However, the precision of the IDMS-MDRD estimate, the Mayo Clinic estimate and JSN-CKDI estimate did not differ significantly from one another (p = 0.151, 0.421, F statistic: 1.3320, 1.1743).
The accuracy was best for the IDMS-MDRD estimate, the Mayo Clinic estimate, and Rule’s refitted MDRD estimate (within 30% of the reference: 94.12, 92.16 and 94.12%, respectively; within 50% of the reference: 99.02, 99.02 and 100%, respectively). Among these 3 estimates, Rule’s refitted MDRD estimate had the lowest accuracy within 10% (35.29%).
The Japanese-modified MDRD equation and the JSN-CKDI equation significantly underestimated GFR, with biases of –17.95 and –20.86 ml/min/1.73 m2, respectively. The Chinese-modified MDRD equation significantly overestimated the GFR, with a bias of 24.27 ml/min/ 1.73 m2.
All of the cystatin C-based equations underestimated GFR, except the Filler equation. The Filler equation had the least bias of 0.03 ml/min/1.73 m2, but it had a significantly lower precision (17.10 ml/min/1.73 m2) compared with Rule’s refitted MDRD equation (p <0.001, F statistic: 2.4657). The accuracy within 30% of the Filler equation was 78.43%.
The creatinine clearance had a bias of 7.69 ml/min/ 1.73 m2, precision of 17.12 ml/min/1.73 m2 and an accuracy of 84.31% within 30% of the reference, and 97.06% within 50% of the reference.
Figure 1 shows a Bland-Altman plot of our data. This displays the agreement between the estimated and measured GFR values by calculating the span (±1.96 SD) of the mean difference. Table 4 shows the limits of agreement of all equations. The spans of both limits for the IDMS-MDRD, Mayo Clinic and Japanese-modified MDRD equation were 49.3, 46.2 and 43.3 ml/min/1.73 m2 respectively. The span of the Le Bricon equation, which was the smallest span of all cystatin C-based equations, was considerably higher (54.8 ml/min/1.73 m2).
|Table 4. Limits of agreement of GFR estimates|
|Fig. 1. Bland-Altman analysis, showing the differences between the estimated and reference GFR vs. the average of the reference and estimated GFR for each patient. The solid middle line indicates the mean differences, and the upper and lower dotted lines (± 1.96 SD) indicate the limits of agreement. a IDMS-MDRD equation and reference GFR. b Japanese-modified MDRD equation and reference GFR. c Creatinine clearance and reference GFR. d Filler equation and reference GFR.|
Table 5 shows the performance of the GFR equations for the 16 patients who had GFR below 60 ml/min/ 1.73 m2. In these patients, the mean GFR was 49.90 ± 6.32 ml/min/1.73 m2. Almost all of the equations showed a similar accuracy to the measured GFR for all patients. The IDMS-MDRD equation, the Walser equation, and Rule’s refitted MDRD equation provided the most accurate GFR estimates in this population.
|Table 5. Bias, precision, and accuracy of creatinine and cystatin C estimates for patients with GFR <60 ml/min/1.73 m2|
In this study, we compared numerous equations for estimating the GFR with the gold standard for measuring GFR (51Cr-EDTA clearance) in early postoperative Korean renal transplant recipients. Our results indicate that the IDMS-MDRD equation and the Mayo Clinic equation provided the best estimates. The IDMS-MDRD equation had an accuracy of 94.12% within 30% of the reference, 99.02% within 50% of the reference, and a percent bias of 1.17%. A previous study showed that, within the validation sample of the MDRD study, the abbreviated MDRD equation had an accuracy of 91% within 30% of the reference, 98% within 50% of the reference, and percent bias of 3% . Our results are comparable.
In addition, for our 16 patients who had GFR <60 ml/min/1.73 m2, the IDMS-MDRD equation provided the most accurate estimate of GFR. Rule’s refitted MDRD equation and the Walser equation also performed well for these patients.
Differences in methods and equipment affect measurement of serum creatinine and cause inaccuracies in estimation of GFR [32,33]. The MDRD equation was developed in 1999 based on data from 1,628 chronic kidney disease patients . Evaluation of the MDRD equation requires calibration of serum creatinine to the laboratory used in the MDRD study (Beckman CX3, Kinetic Jaffe method). The NKDEP has initiated a creatinine standardization program to improve and normalize serum creatinine results that are used in the various equations for estimating GFR . The MDRD equation was re-expressed for use with a standardized serum creatinine assay (IDMS) which yielded serum creatinine values that were 5% lower . The NKDEP recommends that clinical laboratories using creatinine methods that have been calibrated to IDMS should use the IDMS-MDRD equation . Several studies have demonstrated that the IDMS-MDRD equation provides better estimates than the MDRD equation [24,30,34].
Race can affect GFR estimation. The MDRD equations originally were developed from measurements of Caucasian and African-American patients. A modification of MDRD equations were suggested for the diagnostic inaccuracy of the abbreviated MDRD equation in Asian patients [24,25]. In our study, the Japanese-modified MDRD equation and the JSN-CKDI equation significantly underestimated GFR by –17.95 ml/min/1.73 m2 and –20.96 ml/min/1.73 m2, respectively. On the contrary, the Chinese-modified MDRD equation significantly overestimated GFR by 24.27 ml/min/1.73 m2. Thus, the modification of the MDRD equation for Asian patients did not improve the GFR estimates for our Korean renal transplant recipients. As far as we know, there is no study regarding the accuracy of the MDRD equation, in non-transplantation Korean patients. Therefore, it is difficult to know whether the GFR in non-transplantation Koreans is similar to that of the Japanese or Chinese, which are conflicting with each other. As far as the results of our research is concerned, the modification of the MDRD equation which is needed for Asian patients is not required for Korean transplantation patients.
Several studies have reported a low accuracy of the MDRD equation for kidney transplant [7,8,35,36]. Serum creatinine is affected by several factors that are independent of GFR, such as age, race, muscle mass, gender, medication use and catabolic state. Cystatin C is another marker that can be used to estimate GFR. Some studies have suggested that cystatin C was superior to creatinine for measurement of GFR in renal transplant recipients [9,20,21], but other studies have questioned this conclusion [22,37].
White et al.  included 117 transplant patients (mean post-transplantation, 7.5 ± 10.8 years) using 99mTc-DTPA as a reference. They showed that the abbreviated MDRD equation had low accuracy (74% within 30% of the reference) and high bias (–10.0 ml/min/1.73 m2) and that the Filler equation (87% within 30% of the reference; bias of –1.7 ml/min/1.73 m2) and the Le Bricon equation (89% within 30% of the reference; bias of –3.8 ml/min/1.73 m2) performed best.
Poge et al.  studied 108 renal transplant patients (mean post-transplantation, 75.4 months, 95% confidence interval 61.2–87.6) using 99mTc-DTPA as the reference. They showed that the Hoek equation had the best accuracy (77.1% within 30% of the reference; 97.2% within 50% of the reference; bias of –4.4 ml/min/1.73 m2) and that the abbreviated MDRD equation had poor accuracy (67 within 30% of the reference; 85 within 50% of the reference). However, calibration of the serum creatinine to the laboratory used in the MDRD study  was not done in this study.
Zahran et al.  studied 103 transplant patients (mean post-transplantation, 147.8 ± 60.0 months) using inulin clearance as a reference. They showed that the Gates equation demonstrated the highest accuracy within 30% of 70% in patients with GFR <60 ml/min/1.73 m2 and the Nankivell equation achieved the highest accuracy within 30% of 73.91% in patients with GFR ≧60 ml/min/1.73 m2. They concluded that creatinine-based equations performed as well as cystatin C-based equations.
For our early postoperative Korean renal transplant recipients, the performance of cystatin C-based equations was disappointing. The generation of cystatin C appears to be less variable from person to person than that of creatinine. However, there is preliminary evidence that serum levels of cystatin C are influenced by corticosteroid use [39,40], age, sex, weight, height, smoking status, thyroid function, and the level of C-reactive protein [41,42,43]. Almost all of our patients were on the tapering period from high-dose prednisone (mean dose, 23.4 ± 12.6 mg/day). A high steroid dose can increase the level of serum cystatin C [39,40] and consequently lead to an underestimate of GFR in this early post-transplantation period.
There are several limitations to our study. First, the mean time from kidney transplantation to enrollment was 11.6 days. Although we only included patients with stable renal function, which was defined as the creatinine value reached baseline and its variation was less than 10% within 3 to 5 days, renal function could have changed in this early postoperative period.
Second, we only included patients within a relatively small range of kidney function and with a high level of GFR (mean GFR, 76.77 ± 17.01 ml/min/1.73 m2). Only 16 patients had GFR <60 ml/min/1.73 m2. Although the IDMS-MDRD equation also provided the best estimate of GFR in patients with low GFR, we cannot make definite conclusion about the accuracy of the prediction equation at this level of GFR from our limited data.
In addition, all patients were given trimethoprim/sulfamethoxazole prophylaxis. Trimethoprim blocks tubular secretion of creatinine and reportedly does not affect to GFR . In truth, trimethoprim can raise the serum creatinine by ∼0.2 mg/dl . That might cause a less positive bias in the creatinine-based estimations of GFR.
In conclusion, our study shows that the IDMS-MDRD equation provides the best estimate of GFR in early postoperative Korean renal transplant recipients after calibration of serum creatinine by IDMS. Further prospective study is needed to determine whether this equation will be sufficiently accurate to monitor long-term allograft function.
Asan Medical Center, University of Ulsan College of Medicine
388-1 Pungnap-dong, Songpa-gu
Seoul 138-736 (Republic of Korea)
Tel. +82 2 3010 3263, Fax +82 2 3010 6963, E-Mail email@example.com
Received: February 12, 2009
Accepted: June 2, 2009
Published online: November 3, 2009
Number of Print Pages : 10
Number of Figures : 1, Number of Tables : 5, Number of References : 52
Nephron Clinical Practice
Vol. 114, No. 2, Year 2010 (Cover Date: February 2010)
Journal Editor: El Nahas M. (Sheffield)
ISSN: 1660-2110 (Print), eISSN: 1660-2110 (Online)
For additional information: http://www.karger.com/NEC