Comparison of Estimating Equations for the Prediction of Glomerular Filtration Rate in Kidney Donors before and after Kidney Donation

The aim of this study is to investigate the usefulness of the GFR-estimating equations to predict renal function in kidney donors before and after transplantation. We compared the performance of 24-hour-urine–based creatinine clearance (24 hr urine-CrCl), the Cockcroft-Gault formula (eGFRCG), the Modification of Diet in Renal Disease equation (eGFRMDRD), and the Chronic Kidney Disease Epidemiology Collaboration equation (eGFRCKD-EPI) with technetium-diethylenetriaminepentaacetic acid (99mTc-DTPA) clearance (mGFR) in 207 potential kidney donors and 108 uninephric donors. Before donation, eGFRCKD-EPI showed minimal bias and did not show a significant difference from mGFR (P = 0.65, respectively) while 24 hr urine-CrCl and eGFRMDRD significantly underestimated mGFR (P<0.001 for each). Precision and accuracy was highest in eGFRCKD-EPI and this better performance was more dominant when renal function is higher than 90 mL·min−1·1.73 m−2. After kidney donation, eGFRMDRD was superior to other equations in precision and accuracy in contrast to before donation. Within individual analysis, eGFRMDRD showed better performance at post-donation compared to pre-donation, but eGFRCKD-EPI and eGFRCG showed inferior performance at post-donation. In conclusion, eGFRCKD-EPI showed better performance compared to other equations before donation. In a uninephric donor, however, eGFRMDRD is more appropriate for the estimation of renal function than eGFRCKD-EPI.


Introduction
Assessment of renal function is a critical component of donor evaluation in kidney transplantation and regular monitoring of it is recommended for the long-term safety of kidney donors. [1] In many centers, measurement of GFR using the 125 I-iothalamate GFR or technetium-99 m diethylenetriaminepentaacetic acid ( 99m Tc DTPA) clearance is performed before kidney donation. [2,3] However, those studies are available only in a limited number of institutions. Moreover, they are not feasible for the post-donation monitoring of GFR in everyday clinical practice. Therefore, creatinine-based GFR estimations have been used as alternatives for the estimation of renal function before and after donation.
The 2 equations most commonly used are the Modification of Diet in Renal Disease (MDRD) Study equation and the Cockcroft-Gault (CG) formula. These formulas have some limitations for use in kidney donor workup, because they were developed based on data from patients with reduced renal function. [4,5,6] Recently, the Chronic Kidney Disease Epidemiology Collaboration developed a new equation (CKD-EPI). [7] Its aim was to eliminate the weak point of the MDRD formula and the underestimation of GFR; the data set of the CKD-EPI formula included many participants with normal GFR in the development process. During the validation process in several populations, it has shown greater precision and reliability compared with those of the MDRD formula, especially for subjects with GFR of .60 mL?min 21 ?1.73 m 22 . [7,8,9].
The aim of this study was thus to investigate the performance of each GFR-estimating equations in the prediction of renal function in kidney donors. Second, we intended to determine the usefulness of those equations for the post-donation monitoring of renal function in uninephric donors.

Patients and Methods
A total of 207 healthy Korean adults who underwent the kidney donor workup at our center between March 2009 and September 2011 were included in this study. Laboratory evaluation included blood urea nitrogen, serum creatinine (Scr), and 24 hour urine-based creatinine clearance . Scr values were measured in a single laboratory using a ''compensated'' IDMStraceable method (Hitachi Modular P-800; Roche Diagnostics, Germany). GFR was measured (mGFR) by 99m Tc DTPA clearance with a single injection technique with a 4-point sampling approach at 10, 30, 180, and 240 minutes after injection, according to the method described by Russel et al. [10] After kidney donation, donors visited an outpatient clinic every 3 months, where blood chemistry examination including Scr was performed for 1 year after kidney transplantation (KT). 99m Tc DTPA clearance was performed at around 6 months from KT, as part of the routine follow-up process. Out of 207 patient populations, 108 subjects completed studies after kidney donation, and they were included in the post-donation analysis.
Estimated glomerular filtration rate (eGFR) was calculated using the following equations.

Statistical Analysis
Data are presented as mean 6 SD or counts and percentages, depending on the data type. For continuous variables, mean values were compared using Oneway ANOVA and Dunnett's test. The mean difference between equation-based GFR and the measured GFR was used to determine the bias. Pair-wise comparison of the mean difference was performed using the paired t-test. The precision of the estimates was determined as SD of the mean difference between mGFR and eGFR. [12] Accuracy-integrating precision and bias was calculated as the percentage of GFR estimates within 10% and 30% of the measured GFR as suggested.
[13] McNemar's test was used to evaluate the degree of accuracy. [14] Comparison of the correlation coefficients was performed using Z-statistics. Moreover, a graphical approach to assess accuracy was carried out according to the Bland-Altman method. [15] Statistical analyses were performed using SPSS software (version 15.0; SPSS Inc., Chicago, IL, USA) and MedCalc 11.2.1.0 (Medcalc, Mariakerke, Belgium). All tests were 2-tailed, and the results were considered significant when the P value was below 0.05.

Baseline Characteristics of Patient Population
In the pre-donation analysis, the mean age of donors was 40.4611.3 years; 87 were male (42.0%). The mean height and weight were 164.068.9 cm and 64.0611.9 kg, respectively. The mean body surface area (BSA) was 1.6760.17 m 2 ; the mean body mass index (BMI) was 23.164.4 kg/m 2 . The mean Scr was 0.7860.16 mg/dL. In the post-donation analysis in 108 uninephric donors, the interval from KT to the measurement of GFR and Scr was 7.363.9 months. Mean patient age was 39.0611.5 years; 51 patients were male (47.2%). Scr at the measurement of mGFR was 1.0760.25 mg/dL. Table 1 provides overall results for the bias, precision, and accuracy of all equations for the estimation of mGFR in this predonation cohort. eGFR CG and eGFR CKD-EPI showed minimal bias (P = 0.99 and P = 0.92 vs. mGFR, respectively), while 24 hr urine-CrCl and eGFR MDRD significantly underestimated mGFR (P,0.001 vs. mGFR in each case). eGFR CKD-EPI showed highest precision (lowest SD of mean bias) among three equations. In addition, the accuracy of eGFR CKD-EPI within 30% of mGFR was 91.8%, which is significantly higher than that of 24 hr urine-CrCl (71.5%) and eGFR MDRD (84.1%) (P,0.001 in each case) and it showed a higher tendency compared with eGFR CG (86.0%) (P = 0.06).

Comparison of Each Equation's Performance According to mGFR Level before Kidney Donation
We analyzed the performance of equations according to renal function (Table 1). In 170 subjects with normal renal function (mGFR $90 mL?min 21 ?1.73 m 22 ), both 24 hr urine-CrCl and eGFR MDRD significantly underestimated mGFR (P,0.001 vs. mGFR in each case), but eGFR CKD-EPI showed little bias (P = 0.92). eGFR CKD-EPI showed higher precision and accuracy within 30% of mGFR (P,0.05 in each case) than the other 3 equations as well. In 37 subjects with decreased renal function (mGFR #90 mL?min 21 ?1.73 m 22 ), eGFR CKD-EPI significantly overestimated mGFR (P,0.001 vs. mGFR), and 24 hr-urine CrCl, eGFR CG and eGFR MDRD did not show significant bias to mGFR (P = 0.094, P = 0.211 and P = 0.123 vs. mGFR, respectively). In precision and accuracy, no significant differences were detected in any comparisons between equations. Table 2 provides overall results for the bias, precision, and accuracy of all three equations for the estimation of mGFR in this post-donation cohort. eGFR CKD-EPI showed the least bias as compared with mGFR (P = 1.0 vs. mGFR) as like in pre-donation analysis. In contrast, eGFR MDRD (SD of mean bias: 17.2) showed significantly higher precision as compared to eGFR CG (20.8) and eGFR CKD-EPI (22.9) (P,0.001 in each case). The accuracy within 10% and 30% of mGFR was significantly higher for eGFR MDRD as compared with eGFR CG and eGFR CKD-EPI as well (P,0.05 in each case) ( Table 2).

Comparison of Each Equation's Performance According to mGFR Level after Kidney Donation
We analyzed the performance of equations according to renal function in post-donation cohort ( Table 2). In 23 subjects who showed normal renal function (mGFR $90 mL?min 21 ?1.73 m 22 ), both eGFR MDRD (P,0.001 vs. mGFR) and eGFR CKD-EPI (P,0.05 vs. mGFR) significantly underestimated mGFR and bias was minimal in eGFR CG (P = 0.74 vs. mGFR). In precision, however, eGFR MDRD (SD of mean bias: 17.2) showed superior value as compared to eGFR CKD-EPI (SD of mean bias: 26.5) (P,0.05) and superior tendency compared to eGFR CG (SD of mean bias : 21.9) (P = 0.110). In 85 subjects who showed decreased renal function (mGFR ,90 mL?min 21 ?1.73 m 22 ), eGFR MDRD demonstrated the least bias (P = 0.95 vs. mGFR) and the highest precision (SD of mean bias: 12.6) among three equations but only eGFR CG significantly overestimated mGFR (P,0.01 vs. mGFR). No significant differences in accuracy were detected in any pairs of comparisons in all equations in both groups with normal and decreased renal function (P.0.05, respectively).

Comparison between Pre-donation and Post-donation Performance of Each Equation
For each individual equation, we compared the performance value between pre-donation and post-donation in 108 patients who took 99m Tc-DTPA clearance (mGFR) before and after kidney donation (Table 3). In this analysis, eGFR MDRD showed overall improved performance at post-donation. Precision significantly improved after donation (P,0.001) and the values of bias and accuracy were similar between pre-and post-donation. In contrast, eGFR CG and eGFR CKD-EPI showed overall inferior performance at post-donation compared to pre-donation. Bias from mGFR significantly increased in eGFR CG (P,0.001) and precision significantly decreased in eGFR CKD-EPI (P,0.05) and both equations showed inferior accuracy at post-donation as compared to pre-donation.

Bland and Altman Plots
The differences between each eGFR and mGFR were illustrated using a graphic technique developed by Bland and Altman. These figures display the span between +1.96 and 21.96 SD of the mean difference (limit of agreement), which represents 95% CI. Before kidney donation, a smaller limit of agreement was found for the eGFR CKD-EPI (37.5) in comparison with the eGFR MDRD (40.7), eGFR CG (44.9), and 24 hr-CrCl (57.7) ( Figure 1A-D). However, eGFR MDRD (31.0) showed a smaller limit of agreement than eGFR CG (40.9) and eGFR CKD-EPI (44.9) after kidney donation (Figure 2A-C).

Discussion
This study investigated the performance of formulas for estimating mGFR in both the pre-donation state (healthy adult) and the post-donation state (uninephric donors). In this study, as compared to other equations, eGFR CKD-EPI showed superior performance in healthy donors, the pre-donation state. In contrast, its performance at predicting mGFR was worse than that of eGFR MDRD in uninephric donors, and the inferiority was more significant in subjects with reduced renal function.
At the pre-donation state, 24 hr-urine CrCl and eGFR MDRD significantly underestimated mGFR, but eGFR CKD-EPI showed  only minimal bias. The SD of mean bias was lowest in eGFR CKD-EPI, which suggests the highest precision of this equation. The percentage within 30% of mGFR was significantly higher in eGFR CKD-EPI than in other equations , which suggests the superior accuracy of this equation compared to other equations. This result is fully consistent with the previous reports. [6,7,16,17]. The better performance of eGFR CKD-EPI was more significant when we only included subjects with normal renal function. As reported previously, trends of mGFR underestimation were found in 24 hr-urine CrCl, eGFR MDRD , and eGFR CG , but only eGFR CKD-EPI showed minimal bias in subjects with normal GFR in this study. [17,18] In precision and accuracy, eGFR CKD-EPI was superior to the other 3 equations as well for that patient group. However, in subjects with reduced renal function, this better performance was not dominant. This discrepancy of performance according to renal function level may result from differences in the process of equation development. eGFR MDRD and eGFR CG were developed based on CKD patients with reduced renal function, but eGFR CKD-EPI was not specifically developed for that patient population. [4,7] Indeed, it was previously reported that performance was similar between eGFR CKD-EPI and eGFR MDRD. in CKD patients. [19].
At post-donation state, we directly compared the performance between eGFR CKD-EPI and eGFR MDRD , and our results showed that eGFR CKD-EPI was inferior to eGFR MDRD in overall performance. eGFR CKD-EPI showed less bias compared to eGFR MDRD. /But as shown in high SD of mean difference between mGFR and eGFR CKD-EPI , which suggests low precision, the difference from mGFR was distributed widely in both the positive and negative directions. Negatively and positively biased values may offset each other during the calculation of mean value and may have resulted in the minimal bias of eGFR CKD-EPI . In another performance such as,precision and accuracy, eGFR CKD-EPI showed inferior performance compared to eGFR MDRD . In addition, eGFR MDRD showed better or similar performance at post-donation compared to pre-donation, but eGFR CKD-EPI showed inferior performance at post-donation compared to predonation within individual analysis.
The reason for the superior performance of eGFR MDRD at postdonation state is unclear. One possible reason is that the proportion of subjects with reduced renal function was greater in this group compared to the pre-donation group. Indeed, the performance of eGFR MDRD is not inferior to eGFR CKD-EPI in subjects with reduced renal function in pre-donation cohort. But it cannot explain the better performance of the eGFR MDRD than eGFR CKD-EPI in post-donation cohort with normal renal function. Hence, the more important reason may bethe specific situation of uninephric kidney donors, which is different not only from healthy populations but also from patients with chronic kidney disease. In these subjects, removal of 1 kidney leads to a subsequent reduction in GFR without disease-associated changes in body composition. [17] Renal tissue reduction is accompanied by compensatory hyperfiltration by the remaining nephrons with increases in singlenephron GFR. [20,21] Therefore, the renal function only showed a modest decrease compared to its level before KT because of the remaining kidney's hyperfiltration. In this specific condition, the performance of estimating equations in those patients may show different pattern compared to healthy populations or chronic kidney disease state.
Another possible reason is that the performance of equations for estimating GFR could be affected by the demographic and ethnic factors. Most estimating equations developed primarily based on western populations, hence they may show different performance when used in Asian because of the significant anthropometric difference. Of note, many studies about the performance of estimating equations conducted on Asian showed different outcomes compared to the result from Western populations. [22,23,24] For those reasons, it has been reported that modification is necessary in the use of eGFR CKD-EPI on multiethnic Asian populations. [19,25] But clear conclusion about this issue in Korean may need further investigation.
It is interesting that eGFR CG seems to be nearly unbiased in the potential kidney donors. Because the eGFR CG was derived to estimate creatinine clearance, which is known to overestimate mGFR by 10% to 20% as a result of creatinine secretion, this may be interpreted as a fortuitous cancellation of errors. [14,26] eGFR CG apparently underestimated creatinine clearance by 10% to 20%, thus producing a mean value close to the mean mGFR. In addition, some previous reports indicated that eGFR CG is more appropriate than eGFR MDRD in subjects without kidney disease. [23,27] Therefore, it is possible that eGFR CG showed superior accuracy and less bias compared to eGFR MDRD before kidney donation. After donation, however, when a significant portion of subjects showed reduced renal function, the overall performance of eGFR CG was inferior to eGFR MDRD, as expected.
Usually, CrCl using 24-hour urine collection is not recommended for the estimation of renal function because of the possibility of urine loss during collection, which can cause an inaccurate result. In addition, this method is so inconvenient for patients compared to other methods. [12] Indeed, 24 hour urine was adequately collected only in 31.9% of total donors according to normal range of creatinine excretion. [28] Hence the inaccuracy of 24-hour urine CrCl in the estimation of renal function, including the underestimation of mGFR, may not result from its own low performance but from the inadequate urine collection. Therefore, 24 hr-urine CrCl may not be appropriate for the estimation of GFR before or after donation considering the difficulty of adequate urine collection and patient's convenience.
This study does have some limitations. We could not use the inulin clearance, the gold standard method for measuring true GFR. However, 99m Tc-DTPA clearance is relatively less biased and has been accepted as the accurate method for the measurement of GFR in previous reports. [10,29,30,31] Second, this is a retrospective single-center study, which only included Korean adults. Therefore, the results of this study may not be definite in Western populations which have different anthropomorphic characteristics. To apply our results in those populations, further investigation may be required. Third, eGFR MDRD showed tendency to overestimate the prevalence of CKD, as shown in a previous report, which means that this estimating equation must be used with some caution in the follow-up of uninephric donors. [32].
Nevertheless, this study differs from previous studies in that we used unified and standard methods to measure Scr. A weak point detected in many previous reports was that the Scr assay was either not standardized or not unified, hence the need for a calibration process, which could induce some bias in the results. [17,32] It is possible that the divergences in Scr determination and calibration may have accounted for the heterogeneity of the results in previous studies. [14,33,34] To overcome it, we only included subjects who were tested with isotope dilution mass spectrometry (IDMS)-traceable creatinine, which helps to estimate GFR more accurately. [35].
In conclusion, in the potential kidney donor, eGFR CKD-EPI showed better performance than other GFR estimating equations including eGFR MDRD in the prediction of renal function. However, in the uninephric state after kidney donation, the overall performance of eGFR CKD-EPI was inferior to eGFR MDRD , which suggests that the eGFR MDRD is more appropriate for the estimation of renal function during follow-up of uninephric kidney donors.

Author Contributions
Conceived and designed the experiments: BHC. Performed the experiments: JHY HJC J-IK ISM. Analyzed the data: CWP CWY Y-SK. Wrote the paper: BHC BSC.