Abstract
Significance Statement
New eGFR equations from Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) and European Kidney Function Consortium (EKFC) using creatinine (eGFRcr), cystatin C (eGFRcys), and both (eGFRcr-cys) have sufficient accuracy for use in clinical practice, leading to uncertainty in selecting equations for implementation. The authors evaluated performance of equations in an independent population of 4050 adults and evaluated other considerations important for implementation. They found that CKD-EPI and EKFC equations are approaching convergence, with better performance of eGFRcr-cys equations in the overall group and fewer differences among race, sex, and age subgroups than eGFRcr equations. Larger differences among eGFRcr equations reflect regional population differences in creatinine, forcing a trade-off between accuracy and uniformity in global implementation of eGFRcr equations. More widespread use of cystatin C could avoid this trade-off.
Background
New CKD-EPI and EKFC eGFR equations using eGFRcr, eGFRcys, and both (eGFRcr-cys) have sufficient accuracy for use in clinical practice. A better understanding of the equations, including their performance in race, sex and age subgroups, is important for selection of eGFR equations for global implementation.
Methods
We evaluated performance (bias and P30) of equations and methods used for equation development in an independent study population comprising 4050 adults pooled from 12 studies. The mean (SD) measured GFR was 76.4 (29.6) ml/min per 1.73 m2 and age 57.0 (17.4) years, with 1557 (38%) women and 579 (14%) Black participants.
Results
Coefficients for creatinine, cystatin C, age, and sex in the CKD-EPI and EKFC equations are similar. Performance of the eGFRcr-cys equations in the overall population (bias <±5 ml/min per 1.73 m2 and P30 >90%) was better than the eGFRcr or eGFRcys equations, with fewer differences among race, sex, and age subgroups. Differences in performance across subgroups reflected differences in diversity of source populations and use of variables for race and sex for equation development. Larger differences among eGFRcr equations reflected regional population differences in non-GFR determinants of creatinine.
Conclusion
CKD-EPI and EKFC equations are approaching convergence. It is not possible to maximize both accuracy and uniformity in selecting one of the currently available eGFRcr equations for implementation across regions. Decisions should consider methods for equation development in addition to performance. Wider use of cystatin C with creatinine could maximize both accuracy and uniformity of GFR estimation using currently available equations.
Keywords: CKD, clinical nephrology, creatinine
Introduction
GFR estimation is important for the detection, evaluation, and management of kidney disease. Previous guidelines prioritized performance relative to measured GFR (mGFR) in selecting eGFR equations for implementation and encouraged uniformity across regions. There are several new equations using creatinine (eGFRcr), cystatin C (eGFRcys), and both (eGFRcr-cys) that have sufficient accuracy for use in clinical practice in adults, leading to differing recommendations for implementation by kidney societies in the United States and Europe and uncertainty in global use of eGFR (Table 1).1–5 The equations differ in diversity of the source populations and how variables for race, sex, and age are included. This can lead to systematic differences in performance among subgroups defined by these variables and thus have substantial implications for clinical practice and public health. A better understanding of the equations, including their performance in race, sex, and age subgroups, is important for decisions for global implementation. Our goals were to evaluate performance in a population not included in the development of the equations (an independent study population) and other considerations important for implementation.
Table 1.
Equations, variables, and recommendations for use
| Equations | Demographic and Population Variables Included | Recommendations for GFR Evaluation | Recommendations for GFR Evaluation Equation for eGFR Reporting |
Performance Evaluation in This Report |
|---|---|---|---|---|
| Creatinine | ||||
| CKD-EPI 2009 (ASR)1 | Age, sex, race (B, NB) | Initial test | Recommended by KDIGO in 2012 for use in North America, Europe, and Australia9 | No |
| CKD-EPI 2021 (AS)3 | Age, sex | Initial test | Recommended by NKF, ASN in 2021 for use in the United States6 | Yes |
| CKD-EPI 2009 (ASR-NB)3 | Age, sex, race (NB) | Initial test | Recommended by ERA and EFLM in 2022 for use in Europe7,8 | Yes |
| EKFC 2021 (AS)4 | Age, Q for region (same Q for B and NB, separate Q for M and F, Q varies with age for M and F) | Initial test | Considered by ERA and EFLM in 2022 for use in Europe7,8 | Yes |
| Cystatin C | ||||
| CKD-EPI 2012 (AS)2 | Age, sex | Supportive test in specific circumstances | Recommended by KDIGO in 2012 for use in North America, Europe, and Australia9 | Yes |
| CKD-EPI 2023 (A) | Age | Supportive test in specific circumstances | New equation | Yes |
| EKFC 2023 (A)5 | Age, sex, Q for region (same Q for B and NB, same Q for M and F, Q varies with age) | Supportive test for GFR evaluation in specific circumstances | New equation | Yes |
| Creatinine–cystatin C | ||||
| CKD-EPI-2012 (ASR)2 | Age, sex, race (B, NB) | Supportive test Most accurate eGFR |
Recommended by KDIGO in 2012 for use in North America, Europe, and Australia9 | No |
| CKD-EPI 2021 (AS)3 | Age, sex | Supportive test Most accurate eGFR |
Recommended by NKF, ASN in 2021 for use in the United States6 | Yes |
| CKD-EPI 2012 (ASR-NB)3 | Age, sex, race (NB) | Supportive test Most accurate eGFR |
Recommended by ERA and EFLM in 2023 for use in Europe7,8 | Yes |
| EKFC 2023 (AS)4,5 | Age, Q for region (same Q for B and NB, separate Q for M and F) | Supportive test Most accurate eGFR |
New combination of equations | Yes |
CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; EKFC, European Kidney Function Consortium; A, age; S, sex; R, race; B, Black; NB, non-Black; Q, Q value factor derived by EKFC; M, male; F, female; KDIGO, Kidney Disease Improving Global Outcomes; NKF, National Kidney Foundation; ERA, European Renal Association; EFLM, European Federation of Laboratory Medicine.
Abbreviated Methods
Equations
Consistent with the recommendations not to include an individual's race in computing eGFR,6–8 we compared performance in adults for equations that do not explicitly include race developed by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) for use in adults and European Kidney Function Consortium (EKFC) for use in adults and children (Table 1).9 We included nine equations: five previously published CKD-EPI equations1–3 and a new eGFRcys equation without sex (Supplemental Table 1) and three previously published EKFC equations,4,5 using Q values for White Europeans, irrespective of race of the study participant. For additional information, see the Supplemental Material.
Study Population
We evaluated equation performance in a previously described independent study population of 4050 adults in 12 research studies in North America and Europe.3 Mean (SD) mGFR was 76.4 (29.6) ml/min per 1.73 m2 and age 57.0 (17.4) years, with 1557 (38%) women and 579 (14%) Black participants (Supplemental Table 2).
Comparisons among Equations
We evaluated pair-wise agreement in eGFR between equations using Deming regression (R2, P15, difference). We evaluated accuracy of eGFR versus mGFR (R2, P15, P30, bias) in the overall study population and in subgroups according to race (Black and non-Black), sex (male and female), and age (18–younger than 40 years, 40–65 years, and older than 65 years), as well as differential bias among subgroups. Confidence intervals for metrics were calculated using bootstrap methods. To avoid conclusions based on small differences that could reflect variation in measurement methods and may not be clinically relevant, metrics for accuracy were categorized according to small, medium, or large bias or differential bias and high, moderate, or low P30.6 In addition to performance, we evaluated equations according to methods used for their development, including diversity of the source populations and use of variables for race and sex.
Results
Review of Equation Development
The CKD-EPI and EKFC equations differ in the methods for equation development, including racial diversity of the source populations, assumptions of normal values for GFR and the filtration markers, inclusion of variables for race and sex, and methods for mGFR (Supplemental Table 3). Despite differences in form of the equations, the coefficients for creatinine, cystatin C, sex, and age are similar (Supplemental Table 4).
Performance of Equations in the Overall Population
Agreement between equations was high (Figure 1). In general, eGFR values for CKD-EPI equations were higher than for EKFC equations, with lesser agreement at higher eGFR and lesser agreement among the eGFRcr equations than among the eGFRcys and eGFRcr-cys equations. Accuracy relative to mGFR was adequate for all equations with small bias (<±5 ml/min per 1.73 m2) and moderate to high P30 (>80%). P30 was high (>90%) for eGFRcr-cys equations (Figure 2). In general, differences in accuracy among the eGFRcr equations were larger than among the eGFRcys and eGFRcr-cys equations.
Figure 1.
Agreement between CKD-EPI and EKFC eGFR creatinine, cystatin C, and creatinine–cystatin C equations. Panels show the pairwise agreement in eGFR between CKD-EPI and EKFC eGFRcr, eGFRcys, and eGFRcr-cys equations in the study population. Top row panels show agreement between eGFRcr equations. Middle row panels show agreement between eGFRcys equations. Bottom row panels show agreement between eGFRcr-cys equations. Each dot represents a participant. Black lines represent line of identity. Gray lines represent Deming regression of eGFR versus eGFR ±15%. Units of difference and P15 are ml/min per 1.73 m2 and percent, respectively. CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFRcr, creatinine; eGFRcys, cystatin C; EKFC, European Kidney Function Consortium.
Figure 2.
Accuracy of CKD-EPI and EKFC eGFR creatinine, cystatin C, and creatinine–cystatin C equations. Panels show accuracy of eGFR relative to mGFR for CKD-EPI and EKFC eGFRcr, eGFRcys, and eGFRcr-cys equations in the study population. Top row panels show accuracy of eGFRcr equations. Middle row panels show accuracy of eGFRcys equations. Bottom row panels show accuracy of eGFRcr-cys equations. Each dot represents a participant. Black lines represent line of identity and ±15% and 30%. Gray lines represent regression line of mGFR versus eGFR. Units of bias and P15/P30 are ml/min per 1.73 m2 and percent, respectively. mGFR, measured GFR.
Performance of Equations in Subgroups of the Population and Other Considerations
For all equations, bias was medium to small, and P30 was moderate to high in all subgroups (Tables 2 and 3) and by eGFR and age (Figures 3 and 4). Lack of racial diversity of source populations or exclusion of race and sex variables in equation development or eGFR reporting led to medium bias in subgroups and differential bias among subgroups (Table 4). All eGFRcr equations had medium differential bias by race, but the CKD-EPI (ASR-NB) and EKFC (AS) equations had medium bias only in the Black race subgroup, while the CKD-EPI (AS) equation had medium bias only in the older age and female sex subgroups. None of the eGFRcys equations had differences in bias by age or race, but omission of a sex variable led to medium differential bias between sex subgroups for both CKD-EPI (A) and EKFC (A) equations. Use of eGFRcr-cys equations attenuated differences in bias for all subgroups compared with eGFRcr and eGFRcys equations, with differences remaining only between race subgroups for the EKFC (AS) equation.
Table 2.
eGFR and bias of Chronic Kidney Disease Epidemiology Collaboration and European Kidney Function Consortium eGFR creatinine, cystatin C, and creatinine–cystatin C equations in the overall study population and in subgroups of race, sex, and age
| Equations | eGFR | Bias | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Overall Group | Overall Group | Race Subgroups | Sex Subgroups | Age Subgroups | |||||
| Black | Non-Black | Female | Male | <40 | 40–65 | >65 | |||
| N (%) | 4050 (100) | 4050 (100) | 579 (14.3) | 3471 (85.7) | 1557 (38.4) | 2493 (61.6) | 715 (17.7) | 1989 (49.1) | 1346 (33.2) |
| Creatinine | |||||||||
| CKD-EPI 2021 (AS) | 78.8 (27.3) | −3.1 (−3.5 to −2.6) | 3.6 (1.8 to 5.5)a | −3.9 (−4.4 to −3.5)a | −5.4 (−6.3 to −4.6)a | −1.7 (−2.2 to −1.1)a | −2.1 (−3.2 to −0.8)a | −1.3 (−1.9 to −0.5)a | −5.6 (−6.2 to −4.8)a |
| CKD-EPI 2009 (ASR-NB) | 75.2 (27.0) | 0.4 (0.0 to 0.8) | 7.1 (5.9 to 8.8)a | −0.5 (−0.9 to 0.0)a | −1.9 (−2.6 to −1.3) | 1.8 (1.2 to 2.2) | 0.2 (−1.3 to 1.2) | 2.0 (1.3 to 2.6) | −1.4 (−1.8 to −0.9) |
| EKFC 2021 (AS) | 71.6 (25.3) | 3.5 (3.1 to 3.9) | 9.2 (7.8 to 10.8)a | 2.8 (2.4 to 3.2)a | 2.1 (1.5 to 2.8) | 4.4 (3.8 to 5.0) | 3.6 (2.5 to 5.4) | 3.4 (2.8 to 4.0) | 3.6 (3.1 to 4.1) |
| Cystatin C | |||||||||
| CKD-EPI 2012 (AS) | 75.0 (29.4) | 0.6 (0.1 to 1.1) | −0.1 (−1.5 to 1.6) | 0.7 (0.2 to 1.2) | −1.1 (−1.8 to −0.5) | 1.8 (1.2 to 2.5) | −1.3 (−3.5 to 0.2) | 0.7 (−0.1 to 1.7) | 1.0 (0.5 to 1.6) |
| CKD-EPI 2023 (A) | 74.8 (29.2) | 0.9 (0.3 to 1.3) | 1.3 (−0.6 to 2.6) | 0.8 (0.2 to 1.3) | −3.7 (−4.5 to −3.0)a | 3.6 (3.0 to 4.3)a | −1.9 (−3.4 to 0.2) | 1.5 (0.8 to 2.2) | 0.7 (0.0 to 1.6) |
| EKFC 2023 (A) | 73.5 (25.0) | 1.2 (0.6 to 1.6) | 2.1 (0.7 to 3.8) | 1.0 (0.4 to 1.5) | −3.0 (−3.7 to −2.2)a | 4.0 (3.4 to to 4.5)a | −1.1 (−2.5 to 0.7) | 1.9 (1.0 to 2.6) | 1.1 (0.4 to 1.5) |
| Creatinine–cystatin C | |||||||||
| CKD-EPI 2021 (AS) | 78.4 (28.7) | −2.5 (−2.9 to −2.1) | 0.1 (−0.9 to 1.6) | −2.9 (−3.3 to −2.5) | −4.5 (−5.2 to −3.9) | −1.2 (−1.7 to −0.6) | −2.4 (−3.7 to −0.4) | −1.2 (−1.8 to −0.6) | −3.9 (−4.3 to −3.3) |
| CKD-EPI 2012 (ASR-NB) | 75.8 (28.3) | −0.1 (−0.5 to 0.2) | 3.4 (1.5 to 4.5) | −0.6 (−0.9 to −0.1) | −2.0 (−2.9 to −1.5) | 1.1 (0.7 to 1.6) | −0.9 (−2.4 to 0.7) | 1.0 (0.4 to 1.5) | −1.0 (−1.4 to −0.7) |
| EKFC 2023 (AS) | 72.5 (24.4) | 2.2 (1.8 to 2.6) | 5.8 (4.3 to 7.1)a | 1.6 (1.2 to 2.1)a | −0.7 (−1.4 to −0.1) | 4.0 (3.4 to 4.3) | 1.1 (−0.3 to 2.4) | 2.4 (1.7 to 3.1) | 2.2 (1.7 to 2.8) |
eGFR is mean (SD). Systematic error (bias) was expressed as the median difference (95% confidence intervals) between measured GFR and eGFR. Unit for eGFR and bias is ml/min per 1.73 m2. A positive value for bias indicates eGFR underestimates measured GFR, and a negative value indicates eGFR overestimates measured GFR. Bias closer to zero is optimal. For eGFRcr and eGFRcr-cys, we implemented the EKFC equations using the creatinine Q values for White Europeans and CKD-EPI ASR equations using the NB coefficient for race to avoid specification of race for individual participants. CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; A, age; S, sex; R, race; B, Black; NB, non-Black; EKFC, European Kidney Function Consortium; mGFR, measured GFR.
It indicates bias >±5 for one or more subgroups or differential bias within subgroups >±5.
Table 3.
P30 of Chronic Kidney Disease Epidemiology Collaboration and European Kidney Function Consortium eGFR creatinine, cystatin c, and creatinine–cystatin C equations in the overall study population and by subgroups of race, sex, and age
| Equations | Overall Group | Race Subgroups | Sex Subgroups | Age Subgroups | ||||
|---|---|---|---|---|---|---|---|---|
| Black | Non-Black | Female | Male | <40 | 40–65 | >65 | ||
| N (%) | 4050 (100) | 4050 (100) | 579 (14.3) | 3471 (85.7) | 1557 (38.4) | 2493 (61.6) | 715 (17.7) | 1989 (49.1) |
| Creatinine | ||||||||
| CKD-EPI 2021 (AS) | 86.6 (85.5 to 87.6) | 87.2 (84.5 to 90.0) | 86.5 (85.4 to 87.6) | 83.5 (81.7 to 85.2) | 88.5 (87.2 to 89.8) | 89.4 (87.1 to 91.7) | 87.1 (85.7 to 88.6) | 84.3 (82.3 to 86.3) |
| CKD-EPI 2009 (ASR-NB) | 89.0 (88.0 to 90.0) | 86.4 (83.4 to 89.1) | 89.5 (88.5 to 90.4) | 88.5 (86.9 to 90.0) | 89.3 (88.1 to 90.6) | 90.5 (88.3 to 92.6) | 87.7 (86.2 to 89.1) | 90.2 (88.6 to 91.8) |
| EKFC 2021 (AS) | 90.6 (89.7 to 91.5) | 85.7 (82.7 to 88.5) | 91.4 (90.5 to 92.3) | 91.7 (90.3 to 93.0) | 89.9 (88.7 to 91.1) | 91.6 (89.5 to 93.6) | 88.5 (87.1 to 90.0) | 93.1 (91.8 to 94.4) |
| Cystatin C | ||||||||
| CKD-EPI 2012 (AS) | 88.2 (87.2 to 89.2) | 84.6 (81.7 to 87.6) | 88.9 (87.9 to 89.9) | 88.7 (87.1 to 90.3) | 88.0 (86.7 to 89.2) | 88.3 (85.9 to 90.5) | 86.3 (84.9 to 87.8) | 91.1 (89.5 to 92.6) |
| CKD-EPI 2023 (A) | 86.3 (85.3 to 87.4) | 82.9 (79.8 to 85.8) | 86.9 (85.8 to 88.0) | 84.8 (83.0 to 86.6) | 87.3 (86.0 to 88.6) | 86.0 (83.4 to 88.7) | 84.0 (82.4 to 85.6) | 89.9 (88.2 to 91.5) |
| EKFC 2023 (A) | 88.2 (87.2 to 89.2) | 87.0 (84.3 to 89.6) | 88.4 (87.3 to 89.5) | 87.5 (85.9 to 89.1) | 88.6 (87.4 to 89.8) | 87.0 (84.3 to 89.5) | 86.2 (84.8 to 87.7) | 91.7 (90.1 to 93.1) |
| Creatinine–cystatin C | ||||||||
| CKD-EPI 2021 (AS) | 90.8 (89.9 to 91.6) | 90.5 (88.1 to 92.9) | 90.8 (89.9 to 91.8) | 88.3 (86.7 to 90.0) | 92.3 (91.3 to 93.3) | 93.0 (91.1 to 94.8) | 90.5 (89.2 to 91.9) | 89.9 (88.2 to 91.5) |
| CKD-EPI 2012 (ASR-NB) | 92.2 (91.4 to 93.0) | 90.8 (88.4 to 93.1) | 92.5 (91.6 to 93.3) | 91.5 (90.0 to 92.8) | 92.7 (91.7 to 93.7) | 93.0 (91.0 to 94.8) | 91.4 (90.1 to 92.6) | 93.1 (91.6 to 94.5) |
| EKFC 2023 (AS) | 93.1 (92.3 to 93.9) | 92.7 (90.5 to 94.8) | 93.1 (92.3 to 93.9) | 92.9 (91.7 to 94.2) | 93.1 (92.1 to 94.1) | 92.6 (90.6 to 94.4) | 91.5 (90.2 to 92.7) | 95.6 (94.5 to 96.7) |
P30 (95% CI) was defined as the percentage of individuals with eGFR within 30% of measured GFR. Unit is percent for P30. Higher P30 is optimal. For eGFRcr and eGFRcr-cys, we implemented the EKFC equations using the creatinine Q values for White Europeans and CKD-EPI ASR equations using the NB coefficient for race to avoid specification of race for individual participants. CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; A, age; S, sex; R, race; NB, non-Black; EKFC, European Kidney Function Consortium; mGFR, measured GFR.
Figure 3.
Bias and P30 of CKD-EPI and EKFC eGFR creatinine, cystatin C, and creatinine–cystatin C equations by eGFR. Panels show bias and P30 of CKD-EPI and EKFC eGFRcr, eGFRcys, and eGFRcr-cys equations in the study population across the eGFR range (ml/min per 1.73 m2). Systematic error (bias) was expressed as the median difference of mGFR minus eGFR. P30 was defined as the percentage of individuals with eGFR within 30% of mGFR. Units of bias and P30 are ml/min per 1.73 m2 and percent, respectively. Top row shows eGFRcr equations. Middle row shows eGFRcys equations. Bottom row shows eGFRcr-cys equations.
Figure 4.
Bias and P30 of CKD-EPI and EKFC eGFR creatinine, cystatin C, and creatinine–cystatin C equations by age. Panels show bias and P30 of CKD-EPI and EKFC eGFRcr, eGFRcys, and eGFRcr-cys equations in the study population across the age range (years). Systematic error (bias) was expressed as the median difference of mGFR minus eGFR. P30 was defined as the percentage of individuals with eGFR within 30% of mGFR. Units of bias and P30 are ml/min per 1.73 m2 and percent, respectively. Top row shows eGFRcr equations. Middle row shows eGFRcys equations. Bottom row shows eGFRcr-cys equations.
Table 4.
Considerations in selection of GFR estimating equations for implementation
| Approach | Diversity of Source Population for Equation Development | Inclusion of Race and Sex Variables in Equation and Reporting | Equation Performance in Overall Groupa | Equation Performance in Subgroupsa |
|---|---|---|---|---|
| Creatinine | ||||
| CKD-EPI 2021 (AS) | Black population included | Race not included in equation | Small bias and moderate P30 | Medium bias (underestimate) for older age and female sex; medium differential bias by race |
| CKD-EPI 2009 (ASR-NB) | Black population included | Race included in equation, non-Black race assumed for reporting | Small bias and moderate P30 | Medium bias (underestimate) for Black race; medium differential bias by sex and race |
| EKFC 2021 (AS) | Black population not included | Race not included in equation | Small bias and high P30 | Medium bias (underestimate) for Black race; medium differential bias by race |
| Cystatin C | ||||
| CKD-EPI 2012 (AS) | Black population included | Race not included in equation | Small bias and moderate P30 | Similar |
| CKD-EPI 2023 (A) | Black population included | Neither race nor sex included in equation | Small bias and moderate P30 | Medium differential bias by sex |
| EKFC 2023 (A) | Black population not included | Neither race nor sex included in equation | Small bias and moderate P30 | Medium differential bias by sex |
| Creatinine–cystatin C | ||||
| CKD-EPI 2021 (AS) | Black population included | Race not included in equation | Small bias and high P30 | Similar |
| CKD-EPI 2012 (ASR-NB) | Black population included | Race included equation, non-Black race assumed for reporting | Small bias and high P30 | Similar |
| EKFC 2023 (AS) | Black population not included | Race not included in equation | Small bias and high P30 | Medium bias (underestimate) for Black race |
CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; A, age; S, sex; R, race; NB, non-Black; EKFC, European Kidney Function Consortium.
Bias in one or more subgroups >±5 ml/min per 1.73 m2 or differential bias within subgroups >±5 ml/min per 1.73 m2.
Discussion
We reviewed the performance of CKD-EPI and EKFC equations relative to mGFR in a large, independent study population from North America and Europe, with adequate representation of Black and White race, sex, and age subgroups. We have considered how diversity of the study populations and inclusion of variables for race and sex affect equation performance. Our findings have implications for selection of eGFR equations for global implementation in adults.
First, despite differences in equation development, the CKD-EPI and EKFC equations are approaching convergence, with similar coefficients for creatinine, cystatin C, age, and sex and smaller differences in eGFR among equations compared with the magnitude of differences between eGFR and mGFR for all equations. In particular, performance of equations was similar in the range of eGFR from 30 to 60 that is most important for clinical decision making. These findings suggest that development of additional new equations using these variables may not substantially improve performance in North America and Europe. Second, as reported by others,3,4,10 eGFRcr-cys equations performed better than eGFRcr or eGFRcys equations, irrespective of the research group that developed the equation, supporting recommendations for more frequent use of cystatin C as a supportive test for GFR evaluation. Third, eGFRcr equations had larger variation in performance across most subgroups than eGFRcys or eGFRcr-cys equations, consistent with known greater heterogeneity in associations of race, sex, and age with non-GFR determinants of creatinine than cystatin C. Unlike other studies,4,5,10 the bias in eGFRcr in young adults was not larger for the CKD-EPI than EKFC equations, suggesting differences between development populations in this subgroup, which should be explored (e.g., lower mGFR in EKFC because of inclusion primarily of children and young adults with CKD and higher mGFR in CKD-EPI because of inclusion of young adults with type 1 diabetes and kidney donor candidates). Finally, differences in bias among all subgroups were related to the diversity of the study populations and use of variables for race and sex.
In the past, guidelines have prioritized performance and uniformity in selection of eGFR equations to improve outcomes and to facilitate implementation in clinical practice, research, and public health.9 Indeed inclusion of race, sex, and age variables in the CKD-EPI equations and use of race-specific, sex-specific, and age-specific Q values for application of the EKFC equations reflect strategies to maximize performance overall and in these subgroups. It is now accepted in the United States that race should not be used in eGFR equations, although omission of race reduces accuracy of eGFR equations relative to mGFR. US national kidney disease societies recommended, and clinical laboratory societies subsequently endorsed, the CKD-EPI 2021 (AS) equation, which is less accurate in both Black and non-Black race subgroups.8,11,12 The use of racial or ethnic factors is not permitted in some countries in Europe, limiting the use of formulas with such coefficients.13 The European Renal Association and European Federation of Laboratory Medicine recommended the CKD-EPI 2009 (ASR-NB) equation and favorably considered the EKFC 2021 (AS) equation using the Q value for White Europeans, which is less accurate only in the Black race subgroup.7,8
Our findings suggest that it will not be possible to optimize both accuracy and uniformity in the selection of any of the currently available eGFRcr equations across multiracial populations from North America and Europe. This trade-off between accuracy and uniformity will have important consequences around the world. Prioritizing accuracy over uniformity would lead to region-specific eGFRcr equations, which has already occurred to a limited extent in Asia and Africa,14,15 and would reduce systematic bias in eGFRcr for the overall population of the region and differential bias among subgroups within the population. This, in turn, would strengthen consistent use of GFR thresholds for disease definition, classification, and risk estimation across regions; consequently, an eGFR value of “60” would have the same meaning in all regions. However, the need for region-specific eGFR equations would be a barrier to implementation because it would be necessary to determine the most accurate eGFRcr equation in each region. Ideally, regions would be large to avoid inconsistency in eGFRcr among institutions in nearby locations and willing to accept small to medium differences in equation performance within the region and across subgroups. More frequent use of cystatin C may avoid this trade-off by allowing better performance and selection of uniform eGFRcys and eGFRcr-cys equations across regions. Cost is higher for assays of cystatin C than creatinine, although not higher than for many commonly used supportive tests, and further research is necessary to determine the cost-effectiveness of this strategy.
There is growing concern that use of a binary sex variable in eGFR equations may not be appropriate for all sex and gender subgroups within the population. The availability of eGFRcys equations that do not include a sex variable allows options for use in all gender subgroups. Further research is necessary to understand the accuracy of eGFR equations with and without a sex variable for subgroups according to gender identification and how to use nonbinary sex or gender identifiers in eGFR reporting.
Limitations of our analysis include few study participants of non-Black, non-White race, possible residual variation in assay methods for creatinine and cystatin C despite standardization, use of multiple mGFR methods in development and evaluation without calibration, absence of Q values appropriate for use in North America populations, and lack of knowledge of indications for mGFR in clinical populations within CKD-EPI and EKFC studies.
In conclusion, our analysis demonstrates that the CKD-EPI and EKFC equations for eGFRcr, eGFRcys, and eGFRcr-cys are approaching convergence, with better performance of eGFRcr-cys than eGFRcr or eGFRcys equations and fewer differences among race, sex, and age subgroups. Larger differences in performance among eGFRcr equations than eGFRcys and eGFRcr-cys equations reflect regional population differences in non-GFR determinants of creatinine, necessitating a trade-off between optimizing accuracy versus uniformity across regions in selecting among currently available eGFRcr equations for implementation. We support worldwide increased use of cystatin C with creatinine to maximize both accuracy and uniformity of GFR estimation. We support ongoing international discussions to promote consensus in the implementation of eGFR equations.
Supplementary Material
Acknowledgments
We would like to acknowledge collaborators of studies included in our analyses. Age, Gene/Environment Susceptibility Reykjavik study (AGES-RS): Margret B. Andresdottir, Hrefna Gudmundsdottir, Olafur S. Indridason, and Runolfur Palsson; Assessing Long Term Outcomes in Living Kidney Donors (ALTOLD): Bertram Kasiske, Matthew Weir, Todd Pesavento, and Roberto Kalil; Study of people living with HIV: Christina Wyatt, Zipporah Krishnasami, and James Hellinger; Multicenter AIDS Cohort Study (MACS), now the MACS/WIHS Combined Cohort Study (MWCCS): Alison Abraham; Multi Ethnic Study of Atherosclerosis (MESA): Tariq Shafi, Wendy Post, and Peter Rossing; NephroTest: Jerome Rossert and Benedicte Stengel; Prevent Kidney Function Loss (PERL): Andrzej Galecki, Catherine Spino, Michael Mauer, and Amy Karger; Renin Angiotensin System Study (RASS): Bernard Zinman and Ronald Klein; Steno Diabetes Center study: Hans-Henrik Parving.
Disclosures
J. Coresh reports grants from National Institute of Health and National Kidney Foundation and consulting fees from Healthy.io. J. Coresh also reports consultancy: Soma Logic; ownership interest: Healthy.io. A. Doria reports receiving grants from the JDRF and Novo Nordisk Foundation. A. Doria also reports speakers bureau: Novo Nordisk though Aristea “D-zOne Project: Global perspectives for diabetes and obesity management: getting out of the comfort zone?—November, 26th 2021.” M.M. Estrella reports grant to institution from Bayer Inc.; scientific meeting co-chair for National Kidney Foundation; and Editorial board member for American Journal of Kidney Diseases and CJASN. M.M. Estrella reports consultancy: Boehringer-Ingelheim, Inc.; spouse: Castle Biosciences: Isothrive, Medtronic, Phathom Pharmaceuticals, and Sanofi; research funding: Bayer, Inc.; and other interests or relationships: American Journal of Kidney Diseases, CJASN, and National Kidney Foundation. M. Froissart reports consultancy: CSL Vifor. V. Gudnason reports research funding: Novartis. L. Inker reports research grant from Chinook Therapeutics, National Kidney Foundation, NIH/NIDDK, Omeros, and Reata Pharmaceuticals and consulting agreements with Tricida Inc.; advisory role for Alport Foundation—Medical Advisory Council and NKF—Scientific Advisory Board; consultancy to Diamtrix; and other interests or relationships: American Society of Nephrology member, National Kidney Foundation member. R. Kalil reports research grant from Eurofins. R. Kalil reports research funding: Eurofins/Viracor. A.S. Levey reports research from the NIH and NKF to Tufts Medical Center, honoraria for lectures at academic medical centers, advisory board participation for clinical trials of dapagliflozin at AstraZeneca, and royalties from Up-To-Date. M. Mauer reports receiving grant (1R01DK129318) for a pilot study of fenofibrate to prevent kidney function loss in type 1 diabetes. M. Mauer also reports consultancy: Acelink, Acelink Theraputics, Amicus, Avrobio, Chiesi, Freeline Theraputics, Sangano, and Sanofi/Genzyme; research funding: Amicus, Freeline Theraputics, and Sanofi/Genzyme; honoraria: Acelink Theraputics, Amicus, Freeline Therapeutics, and Sanofi/Genzyme; and advisory or leadership role: North American Fabry Registry Board and International Fabry Registry Board. P. Rossing reports receiving honoraria to his institution from Abbott, AstraZeneca, Bayer, Boehringer Ingelheim, Eli Lilly, Gilead, Novo Nordisk, and Sanofi. P. Rossing also reports research funding: AstraZeneca, Bayer, and Novo Nordisk and advisory or leadership role: Astra Zeneca, Bayer, Gilead, Novo Nordisk, all honoraria to institution. J. Seegmiller reports subcontracts from CDC and NIH and participation on data and safety monitoring board for HIV in Nigeria (H3). M.G. Shlipak reports research grant from Bayer Pharmaceuticals, NIA, NIDDK, and NIH-NHLBI; consultancy fees from Cricket Health and Intercept Pharmaceuticals; and honorarium from AstraZeneca, Bayer Pharmaceuticals, and Boehringer Ingelheim. M.G. Shlipak reports research funding: Bayer Pharmaceuticals; honoraria: AstraZeneca, Bayer, and Boeringer Ingelheim; patents or royalties: Kidney Health Monitoring in Hypertension Patients; advisory or leadership role: American Journal of Kidney Disease, Circulation, JASN, and Board Member, Northern California Institute for Research and Education; and other interests or relationships: Committee Member—KDIGO Guidelines Committee. H. Tighiouart reports ownership interest: Bank of America, Cheniere Energy, Citi, Exxon Mobile, Ford, Merck, Oracle, Organon, and TJ Maxx. All remaining authors have nothing to disclose.
Funding
This work is supported by National Institute of Diabetes and Digestive and Kidney Diseases from 1R01DK116790 (L.A. Inker).
Author Contributions
Conceptualization: Lesley Inker, Andrew S. Levey.
Data curation: Alessandro Doria, Michelle M. Estrella, Marc Froissart, Anders Grubb, Vilmundur Gudnason, Roberto Kalil, Michael Mauer, Peter Rossing, Jesse Seegmiller, Michael G. Shlipak.
Formal analysis: Hocine Tighiouart.
Funding acquisition: Lesley Inker.
Investigation: Ogechi M. Adingwupu, Lesley Inker, Andrew S. Levey, Hocine Tighiouart.
Methodology: Ogechi M. Adingwupu, Josef Coresh, Lesley Inker, Andrew S. Levey, Hocine Tighiouart.
Project administration: Ogechi M. Adingwupu.
Supervision: Lesley Inker, Andrew S. Levey.
Writing – original draft: Ogechi M. Adingwupu, Lesley Inker, Andrew S. Levey, Hocine Tighiouart.
Writing – review & editing: Ogechi M. Adingwupu, Josef Coresh, Alessandro Doria, Michelle M. Estrella, Marc Froissart, Anders Grubb, Vilmundur Gudnason, Lesley Inker, Roberto Kalil, Andrew S. Levey, Michael Mauer, Peter Rossing, Jesse Seegmiller, Michael G. Shlipak, Hocine Tighiouart.
Data Sharing Statement
Previously published data were used for this study. Partial restrictions to the data and/or materials apply.
Supplemental Material
This article contains the following supplemental material online at http://links.lww.com/JSN/E540.
Supplemental Table 1. Performance of CKD-EPI 2012 AS and the new 2023 A eGFR Cystatin C Equations in the Development and Internal Validation Population (N=5352).
Supplemental Table 2. Demographic and Clinical Characteristics in the Overall Study Population and by Subgroups of Race, Sex, and Age.
Supplemental Table 3. Methods and Source Populations for Development of eGFR Equations.
Supplemental Table 4. Equations, Variables, Intercepts, and Coefficients.
References
- 1.Levey AS Stevens LA Schmid CH, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150(9):604–612. doi: 10.7326/0003-4819-150-9-200905050-00006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Inker LA Schmid CH Tighiouart H, et al. Estimating glomerular filtration rate from serum creatinine and cystatin C. N Engl J Med. 2012;367(1):20–29. doi: 10.1056/NEJMoa1114248 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Inker LA Eneanya ND Coresh J, et al. New creatinine- and cystatin C-based equations to estimate GFR without race. N Engl J Med. 2021;385(19):1737–1749. doi: 10.1056/NEJMoa2102953 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Pottel H Björk J Courbebaisse M, et al. Development and validation of a modified full age spectrum creatinine-based equation to estimate glomerular filtration rate: a cross-sectional analysis of pooled data: a cross-sectional analysis of pooled data. Ann Intern Med. 2021;174(2):183–191. doi: 10.7326/M20-4366 [DOI] [PubMed] [Google Scholar]
- 5.Pottel H Björk J Rule AD, et al. Cystatin C-based equation to estimate GFR without the inclusion of race and sex. N Engl J Med. 2023;388(4):333–343. doi: 10.1056/NEJMoa2203769 [DOI] [PubMed] [Google Scholar]
- 6.Delgado C Baweja M Crews DC, et al. A unifying approach for GFR estimation: recommendations of the NKF-ASN task force on reassessing the inclusion of race in diagnosing kidney disease. J Am Soc Nephrol. 2021;32(12):2994–3015. doi: 10.1681/ASN.2021070988 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Gansevoort RT Anders H-J Cozzolino M, et al. What should European nephrology do with the new CKD-EPI equation? Nephrol Dial Transplant. 2023;38(1):1–6. doi: 10.1093/ndt/gfac254 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Delanaye P Schaeffner E Cozzolino M, et al. The new, race-free, Chronic Kidney Disease Epidemiology Consortium (CKD-EPI) equation to estimate glomerular filtration rate: is it applicable in Europe? A position statement by the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM). Clin Chem Lab Med. 2023;61(1):44–47. doi: 10.1515/cclm-2022-0928 [DOI] [PubMed] [Google Scholar]
- 9.Kidney Disease: Improving Global Outcomes (KDIGO). KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int Suppl. 2013;3(1):1–150. https://www.kisupplements.org/issue/S2157-1716(13)X3100-4 [DOI] [PubMed] [Google Scholar]
- 10.Fu EL, Levey AS, Faucon A, Delanaye P, Inker LA, Carrero J. Accuracy of novel GFR estimating equations based on creatinine, cystatin c or both in routine care. J Am Soc Nephrol. 2023;34(7):1241–1251. doi: 10.1681/ASN.0000000000000128 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Delgado C, Powe NR. Cystatin C-based equation to estimate GFR without race and sex. New Engl J Med. 2023;388(18):1721–1722. doi: 10.1056/NEJMc2303329 [DOI] [PubMed] [Google Scholar]
- 12.Pottel H, Delanaye P, Rule AD. Cystatin C-based equation to estimate GFR without race and sex. Reply. N Engl J Med. 2023;388(18):1722. doi: 10.1056/NEJMc2303329 [DOI] [PubMed] [Google Scholar]
- 13.Directorate-General for Justice and Consumers (European Commission), Farkas L. Analysis and Comparative Review of Equality Data Collection Practices in the European Union: Data Collection in the Field of Ethnicity. Publications Office; 2020. [Google Scholar]
- 14.Teo BW Zhang L Guh J-Y, et al. Glomerular filtration rates in Asians. Adv Chronic Kidney Dis. 2018;25(1):41–48. doi: 10.1053/j.ackd.2017.10.005 [DOI] [PubMed] [Google Scholar]
- 15.Bukabau JB Yayo E Gnionsahé A, et al. Performance of creatinine- or cystatin C-based equations to estimate glomerular filtration rate in sub-Saharan African populations. Kidney Int. 2019;95(5):1181–1189. doi: 10.1016/j.kint.2018.11.045 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Previously published data were used for this study. Partial restrictions to the data and/or materials apply.