Formula Performance According to Gender and Age
Gender and age did not modify the results of the MDRD formula, which remained superior to that of other formulas in each subgroup (P<0.001 vs. bias of CKD-EPI, CG, and NK formulas), except for patients older than 60 years where the CG formula yielded equivalent results to the MDRD formula. Results of the NK formula seemed to be influenced by gender (poorer performance in women).
Formula Performance According to BMI
Overall, the MDRD formula provided the best performance in the different subgroups (P<0.001 vs. bias of CKD-EPI, CG, and NK formulas) except in patients with BMI less than 18.5 kg/m2, whereas the results of the CG formula were equivalent to those of the MDRD formula.
Formula Performance According to GFR
The MDRD formula provided the best performance for patients with mGFR ranging from 15 to 60 mL/min/1.73 m2 (P<0.001 vs. bias of CKD-EPI, CG, and NK formulas). In patients with GFR between 60 and 90 mL/min/1.73 m2, CG and CKD-EPI had the least bias (P<0.001). However, the MDRD formula provided the best accuracy.
Formula Performance in Relation to Posttransplantation Period
Figure 2 shows mGFR and eGFR in relation to time elapsed since transplantation. Bias in estimating GFR by each formula was similar irrespective to time elapsed since transplantation.
The accurate assessment of renal function is necessary for monitoring kidney transplant recipients and is recommended by the National Kidney Foundation (2). The performance of GFR prediction equations must be assessed in this specific population.
In our population of kidney transplant recipients who had undergone inulin clearance procedure to measure GFR, the MDRD formula provided the best performance for estimating GFR. For the total population, the mean bias was low, measured at −0.5 mL/min/1.73 m2. Moreover, 85% of the estimates fell within 30% of the GFR measured by inulin clearance and were close to the 90% recommended by the Kidney Disease Outcomes Quality Initiative (15).
Although the NK formula is the only one developed for use in kidney transplant recipients, it provided the poorest performance. It was developed in 1995 in a small number of patients (n=146), who underwent several GFR measurements with 99mTc-DTPA (249 in all), in an Australian population (7). One other factor that could explain the poor performance of the NK formula was absence of standardization of serum creatinine assay methods.
The CG formula had intermediate performance with a mean bias at 5.3 mL/min/1.73 m2 and 30% accuracy at 77%. As known in patients with chronic kidney disease, the performance of the CG formula was inferior to that of the MDRD formula.
The main results of the 13 studies that tested the MDRD formula in kidney transplant recipients are given in Table 3. These results are heterogeneous with bias varying from −11.4 to +4.5 mL/min/1.73 m2 and 30% accuracy from 65% to 94%. This heterogeneity can be explained by several factors. The populations tested were different with regard to sample size and demographic characteristics. Few studies reported the ethnic origin of patients. The mean GFR (40–77 mL/min/1.73 m2) and its distribution also varied between studies. Different reference methods used for measuring GFR (plasma clearance of iohexol or urinary clearance of 99mTc-DTPA, 125I-iothalamate, cold iothalamate, or inulin) could also explain a part of the heterogeneity of the results. Another important source of variation was standardization of the creatinine assay method to that of the laboratory, which published the MDRD formula (Cleveland Clinic Laboratory, Cleveland, OH). Not all these studies used the standardized creatinine assay method. Moreover, the MDRD team recently reexpressed its formula, calibrating the creatinine assay method on an IDMS reference method (10) and showed improved performance (11). Only the latest studies tested the MDRD formula in kidney transplant recipients using the reexpressed formula. In this context, the strengths of our research lie in the facts that we had a larger patient cohort, we used inulin clearance as the gold standard for measuring GFR, and we used a calibrated liquid chromatography-mass spectrometry (LCMS) serum creatinine assay method to obtain the data used to calculate the reexpressed MDRD formula.
The recently published CKD-EPI formula (12) was developed to improve estimation of GFR in patients with normal GFR. Our results in kidney transplant recipients show that CKD-EPI performance is close to MDRD performance, but is not any better. This could be explained by the fact that few kidney transplant recipients have normal GFR. Three studies assessed the performance of the CKD-EPI formula in kidney transplant recipients. Kukla et al. (26) and Pöge et al. (28) also found that the CKD-EPI formula had lower performance than the MDRD formula. Conversely, White et al. (27) found better performance for the CKD-EPI formula. Studies reporting performances of the CKD-EPI formula demonstrated some differences (Table 3). All the studies used a calibrated creatinine assay with an IDMS method. However, the previously mentioned factors explaining heterogeneity in studies evaluating eGFR formula performances remain valid. Pöge et al. found poorer performance of the CKD-EPI formula (bias +8.1 mL/min/1.73 m2 and 30% accuracy of 64%) possibly because the mean mGFR of their population was low (39.6 mL/min/1.73 m2). Although White et al. reported CKD-EPI performance similar to our study, they found that CKD-EPI underestimated mGFR.
Bland-Altman plots (Fig. 1) show the variability in GFR estimates for individual values. If mean estimation of GFR by formulas is satisfactory for the whole population, the variability of individual estimation of GFR is the main limitation when using eGFR in clinical practice.
In our analysis, we assessed the performance of each formula taking into account main patients' characteristics: gender, age, BMI, and the stage of renal failure. Unlike the others, the MDRD formula showed equivalent bias for both genders. It also gave an equivalent estimate of GFR according to age, BMI, and the stage of renal failure. In our opinion, this robustness is important for reliably estimating GFR in clinical practice. For the MDRD and CG formulas, our data are consistent with those known for nontransplanted patients with renal failure (13). Two studies in kidney transplant recipients tested the same parameters (i.e., gender, age, BMI, and stage of renal failure). Results were discordant with regard to gender (in one study, the MDRD formula underestimated GFR in women  and in the other study, the MDRD formula underestimated GFR in men ). We also identified the limits of the CKD-EPI formula in some subgroups: patients younger than 40 years, patients with BMI less than 18.5 kg/m2, and patients with GFR less than 30 mL/min/1.73 m2. It is important to note that most of our patients were white (92%), which limits its application in black patients.
We also assessed the effect of posttransplantation period on formula performances. We found no effect on bias between 1 and 15 years posttransplantation. However, we checked all the GFR measurements available, because not all the patients had the same number of measurements or the same posttransplantation periods, and this could generate bias. In a study of 478 patients whose GFR was measured at 1, 2, and 5 years posttransplantation, Bosma et al. (17) also found that bias remained stable over time with the MDRD, CG, and NK formulas.
To conclude, in a population of largely white patients, the MDRD formula provided the best performance in estimating GFR in kidney transplant recipients. Its limits are the same as with patients presenting chronic renal failure, especially thin patients. Its performance was not modified by the time elapsed since transplantation. In our study, the new CKD-EPI formula was not more reliable for estimating GFR in kidney transplant recipients than the MDRD formula. Our results suggest that improved estimation of GFR will depend more on minimizing SD of bias than improving bias, which is low for the MDRD formula, irrespective of patient characteristics. We currently recommend using the MDRD formula to assess renal function in kidney transplant recipients, although remaining fully aware of its limitations.
MATERIALS AND METHODS
Population Studied and Definition of Subgroups
The performance of the four formulas used to estimate GFR was assessed on the first measurement of GFR in all included patients, in the total population, and by subgroups, using the criteria recommended by National Kidney Foundation (15). Between February 2003 and March 2009, 2520 inulin clearances were measured in 1297 adult kidney transplant recipients at the Edouard Herriot Hospital in Lyon (France). Inulin clearance values were measured at stable state, at 1 year, 3 years, 5 years, and then every 5 years after kidney transplantation. Trimethoprim for anti-pneumocystis prophylaxis was stopped 6 months after transplantation. Steroid dosage was tapered off to 5 mg per day during the first 3 months after transplantation, and then steroids were continued at this dosage. The patients were divided into subgroups according to gender, age (18–40, 41–60, and >60 years), BMI (<18.5, 18.5–24.9, 25–29.9, and ≥30 kg/m2), and GFR (15–29.9, 30–59.9, and 60–89.9 mL/min/1.73 m2). Patients with a GFR above 90 mL/min/1.73 m2 or below 15 mL/min/1.73 m2 per min were excluded from the analysis because the sample sizes were too small (n=8 and n=40, respectively). The performance of GFR prediction equations was also tested as a function of time elapsed since transplantation (1, 3, 5, 10, and 15 years). Among the 2520 inulin clearances performed, 138 GFR measurements made more than 17.5 years posttransplantation are not given because of the small sample size per 5-year sets.
Measurement of GFR
GFR was measured by the gold standard method, that is, urinary inulin clearance (29). Briefly, inulin clearance was measured in the morning after fasting. Two peripheral venous catheters were inserted, one in each arm, one for inulin infusion and one for blood sampling. The inulin infusion (Inutest, Frésénius Kabi, Graz, Austria) was started as a 0.03 g/kg loading dose over 12 min and was then continued at a constant rate of 0.33 mg/kg/min throughout the procedure. At the beginning, the patient was asked to empty the bladder and a first blood sample was taken for serum creatinine assay. At 45 min, the bladder was emptied again. Then followed three consecutive 30-min periods with blood sampling (for inulin assay) in the middle of each period and a urine collection (volume and inulin assay) at the end. Where necessary, periods were added and/or urinary catheterization performed, depending on urine output and voiding disorders. Inulin clearance was calculated separately for each period (urinary excretion rate over plasma concentration). The results were expressed as the mean of three to five clearances. Inulin was assayed using an enzymatic colorimetric technique with a Molecular Devices (Sunnyvale, CA) Versamax microplate reader (29).
Measurement of Serum Creatinine Concentrations
Sampling was performed before inulin infusion. The method used was a kinetic colorimetric compensated Jaffé assay. The results for serum creatinine were standardized by linear regression adjustment of the concentrations obtained by the compensated Jaffé assay and the concentrations obtained by LCMS. Briefly, the LCMS apparatus was calibrated with three European standards (Bureau Community Reference 573, 574, and 575) and two American standards (Standard Reference Material), in which creatinine concentrations range from 66.5 to 404 μmol/L. The parameters for the linear regression line were obtained for 54 patients with serum creatinine values ranging from 41 to 220 μmol/L. Ninety-four percent of the kidney transplant recipients in our population had serum creatinine values within this range. Calibration equation was as follows: standardized serum creatinine=0.9395×(Jaffé compensated serum creatinine in μmol/L)+4.6964. Intercept (4.6964; 95% CI: −2.4619–11.8656) and slope (0.9395; 95% CI: 0.8719–1.0072) were not significantly different from 0 and 1, respectively. The coefficient of correlation r was 0.97. Mean difference between LCMS and compensated Jaffé was 1.24±10.05 μmol/L. Stability of the serum creatinine assays was assessed during the study. Blinded ProBioQal controls were tested every 5 weeks, and a nationwide-blinded control was tested each year.
Estimation of GFR
GFR was estimated using the CG formula (6), the NK formula (7), the simplified standardized MDRD formula (10, 11), and the CKD-EPI formula (12). All the eGFRs were normalized by body surface area and expressed in mL/min/1.73 m2. Because eGFR performed better using the standardized creatinine values for CG and NK formulas (data not shown), standardized creatinine values were used for all the calculations.
A systematic literature search was performed using the MEDLINE database (1966 to May 2011). The medical subject heading term “kidney transplantation” was combined with “MDRD” and “CKD-EPI.” Eligible studies met the following criteria: (i) kidney transplant recipient patients, (ii) patients older than 17 years, (iii) MDRD and/or CKD-EPI equation validation as objective of the study, (iv) comparison to a reference measurement of GFR, and (v) report of mean bias and accuracy of the estimation equation. When several eligible studies were published by the same group, only the latest was included.
Bias expresses systematic deviation from the gold standard measure and was calculated as the difference between eGFR and mGFR. Biases were compared using analysis of variance. In the event of statistically significant differences, Student's paired t tests were used. Statistical significance was assessed with a Bonferroni-adjusted threshold (P<0.004) because multiple comparisons were performed. Precision, which expresses the variability of prediction equation estimates around the gold standard GFR measure, was estimated as the standard deviation of the mean difference between eGFR and mGFR (SD of bias). Thirty percent accuracy was calculated as the percentage of eGFR within 30% deviation of the mGFR (1). Bland-Altman plots illustrated the disagreement between eGFR values calculated by different formulas and the reference GFR. The regression trend of the difference and the mean bias of eGFRs toward the reference GFR illustrated the disagreement of different formulas. Statistical analysis was performed using Med-Calc Software version 11 5.1 (MedCalc Software, Mariakerke, Belgium).
The authors thank Mrs. Lynn Richardson for translation support and Mrs. Annie Varennes for creatinine assays.
1. National Kidney Foundation. K/DOQI clinical practice guidelines for chronic kidney disease: Evaluation, classification and stratification. Am J Kidney Dis
2002; 39: S1.
2. Levey AS, Eckardt KU, Tsukamoto Y, et al. Definition and classification of chronic kidney disease: A position statement from Kidney Disease: Improving Global Outcomes (KDIGO). Kidney Int
2005; 67: 2089.
3. First MR. Renal function as a predictor of long-term graft survival in renal transplant patients. Nephrol Dial Transplant
2003; 18(suppl 1): i3.
4. Fellström B, Jardine AG, Soveri I, et al. Renal dysfunction is a strong and independent risk factor for mortality and cardiovascular complications in renal transplantation. Am J Transplant
2005; 5: 1986.
5. Briggs JD. Causes of death after renal transplantation. Nephrol Dial Transplant
2001; 16: 1545.
6. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron
1976; 16: 31.
7. Nankivell BJ, Gruenewald SM, Allen RD, et al. Predicting glomerular filtration rate
after kidney transplantation
1995; 59: 1683.
8. Levey AS, Bosch JP, Lewis JB, et al. A more accurate method to estimate glomerular filtration rate
from serum creatinine: A new prediction formula. Modification of Diet in Renal Disease Study Group. Ann Intern Med
1999; 130: 461.
9. Levey AS, Greene T, Kusek J, et al. A simplified equation to predict glomerular filtration rate
from serum creatinine [abstract]. J Am Soc Nephrol
2000; 11: AO828.
10. Levey AS, Coresh J, Greene T, et al. Expressing the MDRD study equation for estimating GFR with IDMS traceable (Gold Standard) serum creatinine values. [abstract]. J Am Soc Nephrol
2005; 16: 69A.
11. Levey AS, Coresh J, Greene T, et al. Using standardized serum creatinine values in the modification of diet in renal disease study equation for estimating glomerular filtration rate
. Ann Intern Med
2006; 145: 247.
12. Levey AS, Stevens LA, Schmid CH, et al. CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration). A new equation to estimate glomerular filtration rate
. Ann Intern Med
2009; 150: 604.
13. Froissart M, Rossert J, Jacquot C, et al. Predictive performance of the modification of diet in renal disease and Cockcroft-Gault equations for estimating renal function. J Am Soc Nephrol
2005; 16: 763.
14. Poggio ED, Wang X, Greene T, et al. Performance of the modification of diet in renal disease and Cockcroft-Gault equations in the estimation of GFR in health and in chronic kidney disease. J Am Soc Nephrol
2005; 16: 459.
15. National Kidney Foundation. K/DOQI clinical practice guidelines for chronic kidney disease: Part 5. Evaluation of laboratory measurements for clinical assessment of kidney disease. Am J Kidney Dis
2002; 39: S76.
16. Gaspari F, Ferrari S, Stucchi N, et al. Performance of different prediction equations for estimating renal function in kidney transplantation
. Am J Transplant
2004; 4: 1826.
17. Bosma RJ, Doorenbos CR, Stegeman CA, et al. Predictive performance of renal function equations in renal transplant recipients: An analysis of patient factors in bias. Am J Transplant
2005; 5: 2193.
18. Mariat C, Alamartine E, Afiani A, et al. Predicting glomerular filtration rate
in kidney transplantation
: Are the K/DOQI guidelines applicable? Am J Transplant
2005; 5: 2698.
19. Raju DL, Grover VK, Shoker A. Limitations of glomerular filtration rate
equations in the renal transplant patient. Clin Transplant
2005; 19: 259.
20. Risch L, Huber AR. Assessing glomerular filtration rate
in renal transplant recipients by estimates derived fromserum measurements of creatinine and cystatin C. Clin Chim Acta
2005; 356: 204.
21. Poggio ED, Wang X, Weinstein DM, et al. Assessing glomerular filtration rate
by estimation equations in kidney transplant recipients. Am J Transplant
2006; 6: 100.
22. Gera M, Slezak JM, Rule AD, et al. Assessment of changes in kidney allograft function using creatinine-based estimates of glomerular filtration rate
. Am J Transplant
2007; 7: 880.
23. Harzallah K, Sellam A, Mhiri A, et al. Creatinine clearance estimation after kidney transplantation
: An analysis of three methods. Transplant Proc
2007; 39: 2571.
24. Zahran A, Qureshi M, Shoker A. Comparison between creatinine and cystatin C-based GFR equations in renal transplantation. Nephrol Dial Transplant
2007; 22: 2659.
25. Yeo Y, Han DJ, Moon DH, et al. Suitability of the IDMS-traceable MDRD equation method to estimate GFR in early postoperative renal transplant recipients. Nephron Clin Pract
2010; 114: c108.
26. Kukla A, El-Shahawi Y, Leister E, et al. GFR-estimating models in kidney transplant recipients on a steroid-free regimen. Nephrol Dial Transplant
2010; 25: 1653.
27. White CA, Akbari A, Doucette S, et al. Estimating glomerular filtration rate
in kidney transplantation
: Is the new chronic kidney disease epidemiology collaboration equation any better? Clin Chem
2010; 56: 474.
28. Pöge U, Gerhardt T, Stoffel-Wagner B, et al. Validation of the CKD-EPI formula in patients after renal transplantation. Nephrol Dial Transplant
[Epub ahead of print: May 5, 2011].
29. Dubourg L, Hadj-Aïssa A, Ferrier B. Adaptation of an enzymatic polyfructosan assay to clinical practice. Anal Biochem
2010; 405: 266.
Keywords:© 2011 Lippincott Williams & Wilkins, Inc.
Glomerular filtration rate; Kidney transplantation; Inulin clearance; Creatinine-based formulas