Objectives: Management of renal transplant recipients involves measuring glomerular filtration rate and albuminuria; however, data are conflicting on the use of estimating equations or creatinine clearance and albumin-creatinine ratio in early morning urine or albumin excretion in 24-hour urine. We aimed to determine the performance of creatinine clearance and 3 estimated creatinine-based formulas and compare the usefulness of albumin-creatinine ratio related to albumin excretion in kidney transplant patients.
Materials and Methods: This cross-sectional study examined 300 consecutive kidney transplant patients. Serum creatinine was measured with Cobas-8000 and albumin-creatinine ratio, and albumin excretion was measured with Cobas-C311 (Roche Diagnostics, Hitachi, Tokyo, Japan). We quantified bias and percent bias, Bland-Altman results, and concordances in the classification of chronic kidney disease between formulas and creatinine clearance. We also conducted linear regression analyses of all parameters and for cutoffs of 30 and 300 mg/24 hours and determined the ability of albumin-creatinine ratio to predict abnormal albumin excretion (receiver operator characteristic curve analysis).
Results: Bias (mL/min/1.73 m2), percent bias, and concordances between creatinine clearance and Cockcroft-Gault, Modification of Diet in Renal Disease, and Chronic Kidney Disease Epidemiology Colla-boration formulas in the classification of chronic kidney disease were as follows: 15.89, 20.91%, and 0.35; 20.52, 27.89%, and 0.21; and 18.24, 25.39%, and 0.27, respectively. Regression analyses showed a weak but significantly linear relationship for the cutoff values (P < .001). Receiver operator characteristic curve analyses showed areas under the curve of 0.957 and 0.997 at cutoffs of 30 and 300 mg/24 hours. In our patients, the cutoffs were 27 mg/g (88.38% sensitivity, 92.16% specificity) and 238 mg/g (80.00% sensitivity, 97.45% specificity).
Conclusions: We suggest using estimating equations and albumin-creatinine ratio with caution. In routine management of patients with successive stable revisions, we recommended using the Cockcroft-Gault or Chronic Kidney Disease Epidemiology Collaboration formulas and albumin-creatinine ratio.
Key words : Albuminuria, Chronic kidney disease, Renal transplant
Introduction
In renal transplant recipients, total urine albumin excretion is associated with progressive kidney disease, graft loss, and cardiovascular disease.1,2 In a recent study,3 estimated glomerular filtration rate (GFR) and albuminuria were already associated with biomarkers of cardiac injury at levels that do not fulfil the chronic kidney disease (CKD) criteria. Furthermore, estimated GFR and albuminuria are the main kidney measures used for detection and staging of kidney disease (GFR is a measure of kidney function, and albuminuria is a marker of kidney damage).4-6
The care of kidney transplant recipients requires an accurate estimate of GFR. Methods for measuring true GFR (urinary clearance of plasma inulin, decrease in plasma radioisotopes, or nonradioactive contrast media) are complex, invasive, lengthy, expensive, not easily available, and require the administration of substances not feasible in routine monitoring.7 Methods for obtaining estimated GFR can be made by measuring clearance of endogenous substances, often urinary clearance of creatinine (CrCl), computed from 24-hour urine collection. Creatinine clearance is a marker that overestimates measured GFR because it does not account for tubular secretion and extrarenal creatinine elimination. Over the past few decades, serum creatinine has been the most frequently used marker to estimate GFR; however, equations have recently been developed for estimating GFR using endogenous filtration markers such as serum creatinine.
There are several creatinine-based equations that are used to estimate GFR: the Cockcroft-Gault (CG)8 equation, the Modification of Diet in Renal Disease (MDRD)9 equation, and the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI)10 formula. However, these have not been developed in the transplant population.11 The CG equation is mainly for drug dosing. The CKD-EPI equation was developed to create a more accurate and precise formula than MDRD, especially when GFR is greater than 60 mL/min/1.73 m2. In patients with GFR greater than 60 mL/min/1.73 m2, bias was reduced and accuracy increased with the CKD-EPI equation compared with the MDRD equation.12
In our setting, 24-hour urine collection has been used to quantify albuminuria in renal transplant recipients. However, there are several downsides, including patient compliance, the time required, cost involved, inconvenience of handling urine for 24 hours, and inappropriate collection.2,13 Studies in nontransplant patients have shown good correlations between albumin-to-creatinine ratio (ACR) in early-morning urine (EMU) and albumin excretion in 24-hour urine collection.14,15 On the other hand, Akbari and associates2 explained that the ability of ACR to accurately predict 24-hour albumin excretion in kidney transplant recipients is modest. A 24-hour measurement of albumin excretion should, however, be considered when making major clinical decisions (eg, the need for biopsy). According to the Improving Global Outcomes (KDIGO 2012) Clinical Practice Guideline for the evaluation and management of CKD,16 24-hour urine collection is not necessary for the confirmation of proteinuria. However, the validity of spot albumin measurements in the renal transplant population remains unclear.17
The objectives of this study were to examine the performance of CrCl and 3 creatinine-based formulas for estimating GFR (CG, MDRD, and CKD-EPI) and to determine whether ACR in EMU is equivalent to 24-hour albumin excretion and the ability of ACR to predict 24-hour albumin excretion in kidney transplant patients.
Materials and Methods
Patients and sampling
This cross-sectional study was conducted in accor-dance with the ethical
principles of the Declaration of Helsinki and the principles of current Good
Clinical Practice guidelines. The study protocol was approved by the Ethical
Committee of the Puerta del Mar University Hospital and Bahía de Cádiz la Janda
District (no. 19/15). All patients provided written informed consent, and no
financial incentives were provided. From 9 February to 23 April 2015, 300
consecutive kidney transplant patients of the nephro-logy department at a
tertiary care teaching hospital in Cádiz (Spain) were enrolled in the study.
Patients were required to have 24-hour urine, EMU, and serum collection as part
of their clinical care. We used all samples for our study. Patients transplanted
within the previous 6 months, those with diuresis less than
500 mL, pregnant women, patients with unstable
renal function (outpatients with more than ± 15% variation in creatinine over
the preceding 2 weeks), patients under the age of 18 years, and patients who did
not sign the written informed consent were excluded.
Methods
Serum creatinine was measured with the Cobas 8000 system (Roche Diagnostics,
S.L; Hitachi, Tokyo, Japan), an automatic modification of the kinetic Jaffé
enzymatic method. The creatinine assay was adjusted for calibration with the
isotope dilution mass spectrometry method as reported elsewhere.18 The assay
range was from 0.17 to 24.90 mg/dL
(15-2200 μmol/L). The percent coefficient of variation was 3.5% at 1.14 mg/dL
(101 μmol/L) and 2.2% at 3.97 mg/dL (351 μmol/L). The GFR study used the
formulas described in Figure 1.
For 24-hour urine collection, participants were instructed to begin urine collection immediately after discarding the first morning urine and to collect all urine in the same container, including the final urine voided the following morning until completion of 24 hours. In the same way, patients were instructed to collect midstream urine for EMU. Urine was collected in containers without conservatives and stored at 4ºC before analysis, which was usually within 24 to 48 hours after reception. A collection was deemed improper if it fell outside the range of daily creatinine excretion (female patients = 740-1570 mg/L [65.49-138.94 mmol/L] and male patients = 1040-2350 mg/L [92.04-207.97 mmol/L]).13
The ACR in EMU and albumin in 24-hour urine samples were measured with the Roche Cobas C311 system (Roche Diagnostics) using reagent kits and calibrators supplied by the manufacturer. Urine creatinine levels were measured by a kinetic Jaffé method and urine albumin by immunoturbidimetry. The Cobas C311 system was validated for albumin and creatinine in urine. The results of reproducibility are presented in Table 1. A linearity study was made using calibrator dilutions (1:0, 3:1, 2:2, 1:3, and 0:1); the equation for albumin was Y = 1.07× – 1.51, and the equation for creatinine was Y = 1.02× + 2.62. The lower limits of detection for urine albumin and creatinine assays were 0.3 mg/dL (3.0 mg/L) and 4.2 mg/dL (0.375 mmol/L), respectively. Percent coefficient of variation was 1.7% at 3.12 mg/dL (31.2 mg/L) for urine albumin and 2% at 91.9 mg/dL (8.13 mmol/L) for urine creatinine concentration. At a concentration of 10.5 mg/dL (105 mg/L) and 176 mg/dL (15.53 mmol/L) of albumin and creatinine, coefficients of variation were 1.2% and 1.6%, respectively. Data were obtained with the Omega 3000 system (Roche Diagnostics, Barcelona, Spain).
Statistical analyses
Statistical analyses were performed using SPSS version 18.0 software
(Statistical Package for the Social Sciences: Chicago, IL, USA). We divided
24-hour albumin excretion into 3 groups (< 30,
30-300, and > 300 mg/24 h), and the mean differences were compared using
analysis of variance for normally distributed variables or Kruskal-Wallis
test for skewed variables. P < .05 was considered significant. Using the
Kolmogorov-Smirnov test,
we found recipient age and donor age to be normally distributed. Continuous
variables with and
without normal distribution are expressed as means and standard deviation (SD)
and median and interquartile range (IQR). Qualitative variables
are presented as absolute and relative frequencies.
We assessed the performance of the 3 equations against CrCl in the transplant patients by formally testing the differences in bias (CrCl – [CG/ MDRD/CKD-EPI]), percent bias (CrCl – [CG/ MDRD/CKD-EPI]/CrCl × 100), and Bland Altman plots for the comparison between formulas and GFR according to CrCl determinations. Cohen kappa results were used to measure agreements between classifications according to the KDIGO criteria.16
A simple linear regression test was calculated to study the linearity for patients and different cutoff points (< 30, 30-300, and > 300 mg/24 h). Linear regression, as an appropriate means of analysis, requires a constant variance of deviation of the observed points around the regression line.
A log/log transformation of the data corrected this nonconstant variability and allowed linear regression to be used for these comparisons. The ability of ACR to predict abnormal 24-hour albumin excretion at the cutoff of 30 and 300 mg/24 hours was determined from receiver operator characteristic curve analysis and by calculating sensitivities, specificities, likelihood ratios, and predictive values.
Results
Patient characteristics
We recruited 300 patients who consented to participate in the trial. We studied
their characteristics in the entire group and in separate groups depending on
24-hour albumin excretion (< 30, 30-300, >
300 mg/24 h). Of 300 patients, 180 (60%) were men and 120 (40%) were women; mean
age was 55 years (22-83 y) for men and 54 years (19-79 y) for women. The most
frequent reason for transplant was glomerulonephritis (103 patients, 34%).
Some patients (n = 135; 46%) took nephrotoxic drugs, including angiotensin-converting enzymes and/or angiotensin-receptor blockers. Most patients (n = 279; 93%) received triple immunosuppressive therapy, consisting of steroids, mycophenolate mofetil, and cyclosporine or tacrolimus. The remaining patients received different combinations of immuno-suppressive drugs. We also measured body mass index, mean arterial pressure, donor age, and whether the kidney was from a deceased donor. The mean time after transplant was 8 years (range, 6 mo to 28 y). Patient characteristics are listed in Table 2.
Creatinine clearance and glomerular filtration rate-estimating equations
The mean (SD) laboratory results are as follows: CrCl = 69.72 (33.19)
mL/min/1.73 m2; CG = 61.16 (25.07) mL/min/1.73 m2; MDRD = 52.17
(21.04) mL/min/1.73m2; and CKD-EPI = 55.88 (21.05) mL/min/1.73 m2.
Median serum creatinine (IQR) was 1.30 (0.60) mg/dL or 115.04 (53.10) μmol/L.
Mean biases (SD) of the equations were similar, but the CG equation had the
smallest bias at 15.89 (20.91) mL/min/1.73m2, with MDRD at 20.52
(21.85) mL/min/1.73m2 and CKD-EPI at 18.24 (20.75) mL/min/1.73m2.
Consequently, the relative bias found for the formulas were as follows: CG =
20.91%, MDRD = 27.89%, and CKD-EPI = 25.39%. Figure 2 shows the Bland Altman
plots for comparisons between GFRs calculated with CG, MDRD, and CKD-EPI and GFR
according to CrCl determinations. Results are represented as mean ± 1.96 SD.
Table 3 shows the results of cross tabulation between classifications according to the KDIGO criteria14 for CG versus CrCl, MDRD versus CrCl, and CKD-EPI versus CrCl. With Cohen kappa, concordances were 0.35, 0.21, and 0.27, respectively. We also measured Cohen kappa clustering for G1 and G2 (≥ 60 mL/min/1.73 m2) and G3a, G3b, and G4 (< 60 mL/min/1.73 m2), which allowed better results to be obtained with the formulas for CG, MDRD, and CKD-EPI by 0.60, 0.44, and 0.54, respectively. The G5 group was removed from the study because there were few patients, which could skew outcomes.
Albumin-creatinine ratio versus 24-hour albumin excretion
The medians and IQRs of ACR and 24-hour albumin excretion were 47.0 (165.0) mg/g
and 70.00 (198.75) mg/24 hours, respectively. There was a significant
correlation between ACR and 24-hour albumin excretion for all selected ranges,
as shown by the correlation coefficient (Table 4) being better when the study
was done for all patients and with increasing levels of albuminuria. The results
of the regression test are shown in Table 4.
The receiver operator characteristic analyses outcomes are shown in Table 5. In general, ACR demonstrated good discriminatory ability.
Discussion
Measurements of albuminuria and GFR are a central aspect of care and prognosis in the renal transplant recipient population. However, there is uncertainty regarding the best measure of urinary albumin excretion and the best estimated GFR equation, and this has clinically important implications from a practical and cost-effectiveness perspective.
The research presented here shows more 24-hour albumin excretion for older recipients and donors, in agreement with Kwiatkowska and associates.19 Our study compared different methods to estimate GFR that are routinely used to assess kidney function. In clinical practice, it is difficult to measure a true GFR. Creatinine clearance is a GFR marker that overes-timates measured GFR (by about 15%) because it does not account for tubular secretion and extrarenal creatinine elimination. Moreover, it requires accurate collection of 24-hour urine. Comparisons of differences, from Bland-Altman analyses, showed a low bias of estimated GFR when obtained with the formulas compared with CrCl, especially when GFR was < 60 mL/min/1.73 m2. Between 60 and 90 mL/min/1.73 m2, we found more bias, with the highest bias shown when GFR was more than 90 mL/min/1.73 m2. The plots showed that, at lower values of the estimated GFR equations, higher values of CrCl are shown. However, in general, Bland-Altman analysis revealed a good concordance between estimated GFR equations and CrCl.
The concordance in classification of CKD between the CrCl and equations was weak. We found better concordance when we made comparisons between groups (G1-G2, and G3-G4); therefore, we could conclude that formulas classified better if we only considered GFR < 60 mL/min/1.73 m2 or GFR > 60 mL/min/1.73 m2, but this could make it difficult to detect patients whose renal function is just beginning to deteriorate. In both cases, according to our results, the best formula to classify GFR categories is CG and then CKD-EPI. In the literature, we found a study comparing the CG formula versus MDRD in renal transplant patients, with the authors concluding that the predictive performance of the equations is inadequate, although CG formula did perform better.20 In contrast, others have considered that the MDRD formula provides the best estimate of GFR in kidney transplant participants.21 Kukla and associates22 and Pöge and associates23 found that the CKD-EPI formula had lower performance than the MDRD formula. Conversely, White and colleagues24 found better performance with the CKD-EPI formula.
We also aimed to determine the ability of ACR to predict albumin excretion in 24-hour urine. Similar to Wang and associates, we found that ACR underestimated 24-hour albumin excretion (47 mg/g vs 70 mg/24 h).11 Several studies have assessed the benefits of using albumin in EMU as a surrogate for 24-hour albumin excretion and demonstrated adequate sensitivity and specificity in predicting 24-hour albumin excretion.4,25 However, studies have rarely involved kidney transplant patients or have examined the ability of ACR to rule in or rule out abnormal albumin loss.1,26,27 This heterogeneity is due to differences among populations studied, the consideration of different time points for the control posttransplant, or the different laboratory method-ologies used. We believe that it is important to determine whether EMU can be used instead of 24-hour urine in transplant recipients. Our linear regression study showed that, when ACR increased or decreased, this was significantly (P < .001) accompanied by a similar increase or decrease in 24-hour albumin excretion. Although the directional changes were significantly correlated, there were more differences at lower levels of 24-hour albumin excretion. At ranges of 24-hour albumin excretion < 30 mg/24 hours, there was a 22.10% (R2) ACR variance from 24-hour albumin excretion. This could be a problem to begin to detect kidney damage in kidney transplant patients. When this occurs, ACR becomes a less accurate predictor of 24-hour albumin excretion at lower ranges of urinary albumin excretion. In these cases, we recommend caution because we did not obtain good regression coefficients. In contrast, we found a better correlation when we performed regression analysis for all ranges, with our results showing that ACR in EMU could be used as a screening method and, if kidney damage is detected, extended with 24-hour urine collection.
Similar to the correlation coefficient, sensitivity and specificity also provide useful information but do not inform the clinician about the quantitative accuracy of the test. The sensitivity is a statistical measurement of the performance of a binary classification test. It is defined as the proportion of true positives that are correctly identified by the test.28 In our case, if a condition (albuminuria) is defined as 24-hour albumin excretion > 30 mg/24 hours and if ACR at a cutoff value of 27.00 mg/g has a sensitivity of 88.38%, it means an ACR above 27.00 would detect 88.38% of patients with 24-hour albumin excretion greater than 30 mg/24 hours. Similarly, specificity is the proportion of true negatives that are correctly identified by the test. If the condition is defined again as 24-hour albumin excretion > 30 mg/24 hours, a specificity of 92.16% at an ACR cutoff of 27.00 mg/g would indicate that 92.16% of patients who have ACR less than 27.00 mg/g do not have an albuminuria of more than 30 mg/24 hours. Based on limited data in the transplant recipient population, the KDIGO guideline recommends using cutoffs to define albuminuria similar to those used in the general population.16 Receiver operator characteristic analyses confirmed that ACR is a good test for the diagnosis of albuminuria at cutoff levels of 30 and 300 mg/24 hours, with areas under the curve of 0.96 and 0.97. Erman and associates showed that a cutoff of 30 mg/g is not optimal in the renal transplant population27; in our trial, with conventional cutoff points of 30 and 300 mg/g for ACR, the sensitivity for detection of albuminuria would be lower (83.36%). For our transplant population, we could use 27.00 and 238.00 mg/g for ACR to better classify these patients.
Some limitations to our study should be noted. First, only a small number of patients were included. Second, we selected patients with stable renal function; thus, our results do not apply to patients with proteinuria and acute deterioration in kidney function.
In conclusion, our data showed acceptable correlation, sensitivity, and specificity for ACR, but we cannot always replace 24-hour albumin excretion. In routine management of patients with successive stable revisions, we recommend EMU for ACR (with a cutoff of 27.00 mg/dL) and CG or CKD-EPI equations to estimate GFR. We obtained a good regression coefficient for all patients and an acceptable concordance between estimated GFR equations and CrCl. In the event of suspected deteriorating renal function, the patient should be carefully instructed to collect a 24-hour urine sample to measure CrCl and albumin excretion.
References:
Volume : 17
Issue : 4
Pages : 450 - 456
DOI : 10.6002/ect.2017.0335
From the 1Department of Laboratory, the 2Department of Nephrology, and the
3Department of Preventive Medicine and Public Health, Puerta del Mar University
Hospital, Cádiz, Spain; and the 4Department of Nephrology, Costa del Sol
Hospital, Marbella, Málaga, Spain
Acknowledgements: The authors have no sources of funding for this study and have
no conflicts of interest to declare.
Corresponding author: Iratxe López Pelayo, Avda Cayetano del Toro 54, 5ºT. 11011
Cádiz, Spain
Phone: +34 630 52 10 15
E-mail: iratxelp@yahoo.es
Table 1. Reproducibility Study for Urine Albumin and Creatinine With the Cobas C311 Device
Table 2. Baseline Characteristics of Study Participants
Table 3. Patients Included in the Categories Specified by “Kidney Disease: Improving Global Outcomes,” According to Estimates Using Glomerular Filtration Rate Equations
Table 4. Simple Linear Regression Analyses
Table 5. Receiver Operator Characteristics Curve Analyses
Figure 1. Formulas to Estimate Glomerular Filtration Rate
Figure 2. Bland-Altman Analyses