Performance of the CKD-EPI and MDRD equations for estimating glomerular filtration rate: a systematic review of Latin American studies

ABSTRACT BACKGROUND: The most-used equations for estimating the glomerular filtration rate (GFR) are the CKD Epidemiology Collaboration (CKD-EPI) and Modification of Diet in Renal Disease (MDRD) equations. However, it is unclear which of these shows better performance in Latin America. OBJECTIVE: To assess the performance of two equations for estimated GFR (eGFR) in Latin American countries. DESIGN AND SETTING: Systematic review and meta-analysis in Latin American countries. METHODS: We searched in three databases to identify studies that reported eGFR using both equations and compared them with measured GFR (mGFR) using exogenous filtration markers, among adults in Latin American countries. We performed meta-analyses on P30, bias (using mean difference [MD] and 95% confidence intervals [95% CI]), sensitivity and specificity; and evaluated the certainty of evidence using the GRADE methodology. RESULTS: We included 12 papers, and meta-analyzed six (five from Brazil and one from Mexico). Meta-analyses that compared CKD-EPI using creatinine measured with calibration traceable to isotope dilution mass spectrometry (CKD-EPI-Cr IDMS) and using MDRD-4 IDMS did not show differences in bias (MD: 0.55 ml/min/1.73m2; 95% CI: -3.34 to 4.43), P30 (MD: 4%; 95% CI: -2% to 11%), sensitivity (76% and 75%) and specificity (91% and 89%), with very low certainty of evidence for bias and P30, and low certainty of evidence for sensitivity and specificity. CONCLUSION: We found that the performances of CKD-EPI-Cr IDMS and MDRD-4 IDMS did not differ significantly. However, since most of the meta-analyzed studies were from Brazil, the results cannot be extrapolated to other Latin American countries. REGISTRATION: PROSPERO (CRD42019123434) - https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42019123434.


INTRODUCTION
Chronic kidney disease (CKD) is a public health problem: in 2014, 10.6% of adults aged over 30 years had stage 3-5 CKD. 1 In 2017, CKD caused 35,800,000 disability-adjusted life-years (1.4% of all disability-adjusted life-years) worldwide, 2 and 1,230,200 deaths (2.2% of all deaths). 3 Assessing the glomerular filtration rate (GFR) is the cornerstone for performing adequate screening, diagnosis and classification of CKD. 4 However, the methods used for directly measuring GFR (measured GFR, mGFR) require use of exogenous filtration markers and are laborious and costly. Thus, some equations are routinely used to obtain estimated GFR (eGFR) from endogenous markers such as creatinine 5 or serum cystatin C. 6 The most commonly used equations are the CKD Epidemiology Collaboration (CKD-EPI) and the Modification of Diet in Renal Disease (MDRD) equations. 7 The MDRD equation originally used six variables (MDRD-6): serum creatinine, urea, albumin, age, sex and ethnicity. 8 A later version used only four variables (MDRD-4), excluding serum urea and albumin. 9 Most recently, the MDRD-4 was re-edited to use creatinine measured with calibration traceable to isotope dilution mass spectrometry (IDMS). 10,11 The CKD-EPI originally used the same four variables of the MDRD-4. 12 Later, other CKD-EPI equations were developed, which used serum cystatin C instead of creatinine, 13  Differences in the performance of these equations across certain ethnic groups have been reported, [15][16][17][18] and attributed to differences in the production and excretion of creatinine. 19 This, in turn, is related to diet (protein intake) and muscle mass (endogenous production of creatinine), which vary according to ethnicity. 19

OBJECTIVE
Latin American stakeholders and practitioners need to know which equation has the best diagnostic performance in their specific context, in order to better inform their decisions. Therefore, we conducted a systematic review with the aim of comparing the performance of the CKD-EPI and MDRD equations for estimating the GFR in Latin American countries, and we evaluated the certainty of the evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.

METHODS
The study protocol was registered in PROSPERO (CRD42019123434). We performed a systematic review following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. 23

Literature search and study selection
In this systematic review, we included original observational studies that were performed in Latin American countries and compared both the CKD-EPI and the MDRD equation with mGFR (the gold standard, measured using any exogenous filtration markers such as inulin, iohexol, iothalamate, 51Cr-EDTA or DTPA, among others) in adult populations (≥ 18 years). We did not exclude any study on the basis of language or any other criteria.
We performed a two-step sensitive search. First, we carried out a literature search in PubMed and Scopus in January 2019, and in "Biblioteca Regional de Medicina" (BIREME) in February 2019.

Data extraction
Two researchers (ABC and NBC) independently extracted data from each article that met the inclusion criteria, using a standardized Microsoft Excel sheet. Any differences were resolved by a third researcher (JHZT).
The following variables were extracted from each study: first author, year of publication, country, design (prospective or retrospective), population characteristics (inclusion and exclusion criteria, number of participants, sex, age, ethnic group, CKD diagnosis and CKD etiology), intervention (type of MDRD and CKD-EPI equations), gold standard (exogenous filtration marker), mGFR, eGFR and numerical results from diagnostic measurements.
The main diagnostic measurement comprised bias (defined as the mean of the difference between eGFR and mGFR), P30 (percentage of results of eGFR that did not deviate more than 30% from mGFR) and accuracy measurements (sensitivity, specificity and area under the curve).
Other measurements made included the following: precision (defined as one standard deviation of bias, or as the interquartile range), bias% (mean of the difference between eGFR and mGFR, as a function of mGFR), P15, P10, combined root mean square error (CRMSE), Pearson coefficient, intraclass correlation coefficient, kappa coefficient and limits of agreement (defined as bias ± 2 standard deviations).
When there were doubts about some information reported in the studies, we sent an email to the authors in order to clarify the information. We used the GRADE methodology 25 to report our certainty regarding the evidence of accuracy of the diagnostic test results.

Risk of bias and certainty of evidence
To show this certainty, we created tables of summary of findings (SoF), in accordance with the GRADE specifications. 26,27

Statistical analyses
When possible, we performed meta-analyses on P30, bias, sensitivity and specificity. This was done when studies compared similar equations, showed their confidence intervals or standard deviations, or enabled calculation of these values.
For P30 and bias, we calculated mean differences (MD) and their 95% confidence intervals (95% CI). For sensitivity and specificity, we built a 2 x 2 table when possible. As there were fewer than four studies to meta-analyze, we could not perform a meta-analytical hierarchical regression for diagnostic accuracy. Instead, we performed a meta-analysis of proportions using the exact binomial distribution. We assessed heterogeneity using an I² statistic and used random-effects models when I² was higher than 40%.
For bias and P30, we performed a subgroup analysis according to the presence of CKD (using the cutoff of 60 ml/min/1.73 m²), since a previous systematic review showed that the eGFR equation performance varies across these subgroups 28

Ethics committee approval
This was not applicable since this review did not directly involve human participants.

Studies characteristics
In total, we identified 379 records after removing duplicates.
The characteristics of the 12 studies included are summarized in Table 1 and detailed in Supplementary Material 3 (https://doi. org/10.6084/m9.figshare.14614788.v1). The numbers of participants ranged from 14 to 354 in these studies. Two studies reported results from the same cohort. 30,40 One study 38 added data from two cohorts, one of which 36 was also included in our review and the other had not been published as a separate original paper.
Regarding use of a correction factor for black race, these six studies included this in the MDRD-4 IDMS equation. Five studies (four from Brazil and one from Mexico) used a CKD-EPI-Cr equation that included the correction factor. One study from Brazil 32 did not included the correction factor in the CKD-EPI-Cr equation: the population of this study (n = 70) was mostly Caucasian (only 12 people aged ≥ 60 years were of other races and the study did not detail which races these were).

Risk of bias
Using the QUADAS-2 tool, we found that the risk of bias was uncertain for most studies, regarding patient enrolling, interpretation of index test results without knowledge of the reference standard, interpretation of the reference standard without knowledge of the index test results and the interval between the index and reference standard tests (Figure 2). [29][30][31][32][33][34][35][36][37][38][39][40]
Meta-analyses could only be performed for the comparison between CKD-EPI-Cr IDMS and MDRD-4 IDMS, since other versions of the equations were not evaluated or were evaluated only in one study for the outcomes of interest.
Meta-analyses on bias and P30 are shown in Figure 3.

Certainty of evidence
We used GRADE SoF tables to report the certainty of evidence.
Regarding bias and P30, the certainty of evidence was very low for both CKD-EPI-Cr IDMS and MDRD-4 IDMS ( Table 2).
Regarding differences in true positives, true negatives, false positives and false negatives between equations (obtained through sensitivity and specificity), the certainty of evidence was low ( Table 3).

Comparison with other studies
We performed meta-analyses on six studies conducted in Latin American countries (five from Brazil, one from Mexico) that was found in the previous systematic review. 28 This could be due  from Brazil (elderly); Veronese et al. 38 from Brazil (a: type 2 diabetics, CKD; b: healthy, type 2 diabetics); Arreola-Guerra et al. 29 from Mexico (healthy), and David-Neto et al. 32 from Brazil (elderly renal-transplanted).

Question: How good are the performances of the CKD-EPI-Cr IDMS and MDRD-4 IDMS equations for diagnosing CKD in adult populations (≥ 18 years) in Latin America? Patient or population: Adults in Latin American countries Settings:
The studies included involved community-dwelling adults and hospital-based patients (mean prevalence of CKD across studies included: 41%) New test: CKD-EPI-Cr IDMS Comparison test: MDRD-4 IDMS Reference test: The measured glomerular filtration rate (mGFR) was taken to be the gold standard and was obtained using the Cr-EDTA single-injection method in four studies, Iohexol clearance in one study, and 99m Tc DTPA in one study.  Bias: Defined as the mean of the difference between eGFR (from equations) and mGFR; P30: Defined as the percentage of results for eGFR that did not deviate more than 30% from mGFR. eGFR = estimated glomerular filtration rate; mGFR = measured glomerular filtration rate; CI = confidence interval; CKD = chronic kidney disease; CKD-EPI-Cr IDMS = CKD epidemiology collaboration equation using creatinine with isotope dilution mass spectrometry method to determine creatinine levels; MDRD-4 IDMS = modification of diet in renal disease (with four variables) equation with isotope dilution mass spectrometry method to determine creatinine levels. 1 It was decided to downgrade the level of evidence due to risk of bias because, in more than 50% of the studies, it was uncertain whether the gold standard and reference results were collected at the same time; 2 It was decided to downgrade the level of evidence due to high heterogeneity between the studies (I 2 higher than 90%); 3 It was decided to downgrade the level of evidence due to risk of bias (the gold standard was not the same in all the studies); 4 It was decided to downgrade the level of evidence due to imprecision (both equations could overestimate or underestimate the real value of the GFR); 5 It was decided to downgrade by one level due to risk of bias (it was uncertain whether the results for the gold standard and the reference were collected at the same time, and in one of the studies, no analysis was done on the results from some of the participants).

Outcome Number of studies (number of participants) Test result (95% CI) Quality of the evidence (GRADE) Bias
to the fact that the CKD-EPI-Cr equation was developed in a study in which the mean GFR was higher than the GFR of the study in which the MDRD-4 equation was created (94.5 ml/minute versus 39.8 ml/minute respectively). 12

How to better evaluate eGFR in Latin American populations
These equations may not be accurate for all racial groups due to differences in muscle mass and, consequently, differences in creatinine excretion. 21 Thus, attempts to correct the estimates according to race have been made in these equations using different coefficients for white or black people, but other races have not been taken into account.  [45][46][47] Given this ethnic heterogeneity, it is possible that equation performance may differ from one country to another.
However, among the six studies that could be meta-analyzed in our study, five were performed in Brazil, where the ethnic composition differs from that of other countries in the region.
As an example, while around 60% of the Brazilian population is Caucasian and less than 0.5% is Amerindian, 48 Table 3. Summary of sensitivity and specificity findings for the 60 ml/min/1.73 m 2 cutoff point CI = confidence interval; CKD = chronic kidney disease; CKD-EPI-Cr IDMS = CKD Epidemiology Collaboration equation using creatinine with isotope dilution mass spectrometry method to determine creatinine levels; MDRD4 IDMS = Modification of Diet in Renal Disease (with four variables) equation with isotope dilution mass spectrometry method to determine creatinine levels. 1 It was decided to downgrade the level of evidence due to risk of bias (in both studies, it was uncertain whether a consecutive or random sample of patients was enrolled and whether the results from the index test were interpreted without knowledge of the results of the gold standard); 2 It was decided to downgrade the level of evidence due to risk of bias (the gold standard was not the same in all the studies).

Limitations and strengths
Some limitations of this review should be considered: 1) not all studies had enough information to perform a meta-analysis on the outcomes of interest, even after the authors were consulted; and 2) we found differences in the characteristics of the populations included, but we were not able to perform any subgroup analysis to understand how these differences affected the accuracy of the formulas. 21 The influence of other factors, such as the different causes of CKD or the medicines taken, was not studied either. 50 In spite of these limitations, we believe that our study is important because this is the first systematic review that has compared the GFR equations in Latin American countries (mostly from Brazil), through a two-step sensitive search (the first in two international databases and one local database, and the second in the references and articles that cited each of the articles included in the first step).
In addition, we performed a comprehensive search that including papers in Spanish and Portuguese, and the selection and extraction of data were performed in duplicate.

CONCLUSION
We performed a systematic review to assess the performance of the CKD-EPI and the MDRD equations for estimating the GFR in Latin American countries. We found 12 studies and were able to meta-analyze six of them (five were conducted in Brazil). We found that the performances of CKD-EPI-Cr IDMS and MDRD-4 IDMS did not differ significantly, although CKD-EPI-Cr IDMS tended to have a non-significantly better performance in terms of P30 and among people with GFR ≥ 60 ml/min/1.73m 2 . However, since most of the meta-analyzed studies were from Brazil, the results cannot be extrapolated to other Latin American countries.