Demographics & eGFR Accuracy
Estimated glomerular filtration rate (eGFR) equations incorporate demographic variables to adjust for variation in concentrations of filtration markers across population subgroups. Adjusting for these variations helps to minimize error in patient subgroups.
The recommended eGFR equations for adults adjust for age and sex.
Previous eGFR equations also adjusted for race. However, these equations are no longer recommended and are provided for reference.
Age
To improve eGFR accuracy, adult equations include age to adjust for the observed decline in serum creatinine with increasing age.
Kidney function declines with age in adults (see Table 1).
Table 1. Population mean eGFRs from the National Health and Nutrition Examination Survey (NHANES) III1
| Age (Years) | Mean eGFR (mL/min/1.73 m2) |
|---|---|
| 20–29 | 116 |
| 30–39 | 107 |
| 40–49 | 99 |
| 50–59 | 93 |
| 60–69 | 85 |
| 70+ | 75 |
Sex
Current eGFR equations are based on sex assigned at birth and include different coefficients for female and male.
Race Coefficient
Research has demonstrated that race does not reliably account for genetic diversity.2 As such, use of race in estimating equations is no longer recommended.
Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) developed a new equation that does not include race [2021 CKD-EPI eGFRcr (age+sex)]. Elimination of race introduced a small bias but at levels considered acceptable for routine clinical decision making. Consistent with this, results from the Chronic Renal Insufficiency Cohort (CRIC) and other studies suggested that the use of serum creatinine to estimate GFR without race resulted in systematic misclassification. Data from both CKD-EPI and CRIC demonstrate that this error is mitigated with the use of cystatin C.3 Thus, NIDDK supports the National Kidney Foundation–American Society of Nephrology and Kidney Disease Improving Global Outcomes (KDIGO) recommendation to assess eGFR using both cystatin C and serum creatinine, when possible, especially when eGFR results are within range of key clinical decision cut points.4,5 However, a single parameter equation may be preferred when important non-GFR determinants are present for either creatinine or cystatin C.
Tables 2 and 3 summarize the systematic error (bias) and accuracy, respectively, of the previous and updated CKD-EPI equations.
Table 2. Systematic error (bias) in previous vs updated CKD equations for creatinine, cystatin C, and combined in the CKD-EPI 2021 validation dataset
| Filtration marker | Equation | Black individuals | non-Black individuals | Difference between Black and non-Black individuals (95% CI) |
|---|---|---|---|---|
| creatinine | 2009 eGFRcr (age+sex+race) | -3.7 (-5.4, -1.8) | -0.5 (-0.9, 0.0) | -3.2 (-5.0, -1.3) |
| creatinine | 2021 eGFRcr (age+sex) | 3.6 (1.8, 5.5) | -3.9 (-4.4, -3.4) | 7.6 (5.6, 9.5) |
| cystatin C | 2012 eGFRcys (age+sex) | -0.1 (-1.5, 1.6) | 0.7 (0.2, 1.2) | -0.8 (-2.5, 0.8) |
| creatinine-cystatin C | 2012 eGFRcr-cys (age+sex+race) | -2.5 (-3.7, -1.2) | -0.6 (-0.9, -0.2) | -1.9 (-3.2, -0.6) |
| creatinine-cystatin C | 2021 eGFRcr-cys (age+sex) | 0.1 (-0.9, 1.6) | -2.9 (-3.3, -2.5) | 3.0 (1.6, 4.4) |
Abbreviations: eGFR, estimated glomerular filtration rate; mGFR, measured GFR; eGFRcr, estimating GFR using creatinine; eGFRcys, estimating GFR using cystatin C; eGFRcr-cys, estimating GFR using creatinine and cystatin C; CI, confidence interval
Bias is computed as the median of the difference between measured and estimated GFR. Units for bias are ml/min per 1.73 m2.
Table 3. Accuracy (P-30 in percent) of previous vs updated CKD equations for creatinine, cystatin C, and combined in the CKD-EPI 2021 validation dataset
| Filtration marker | Equation | Black individuals | non-Black individuals | Difference between Black and non-Black individuals (95% CI) |
|---|---|---|---|---|
| creatinine | 2009 eGFRcr (age+sex+race) | 85.1 (82.2, 87.9) | 89.5 (88.5, 90.4) | -4.4 (-7.6, -1.2) |
| creatinine | 2021 eGFRcr (age+sex) | 87.2 (84.5, 90.0) | 86.5 (85.4, 87.6) | 0.7 (-2.4, 3.8) |
| cystatin C | 2012 eGFRcys (age+sex) | 84.6 (81.7, 87.6) | 88.9 (87.9, 89.9) | -4.3 (-7.5, -1.1) |
| creatinine-cystatin C | 2012 eGFRcr-cys (age+sex+race) | 88.6 (85.8, 91.2) | 92.4 (91.5, 93.2) | -3.8 (-6.7, -0.9) |
| creatinine-cystatin C | 2021 eGFRcr-cys (age+sex) | 90.5 (88.1, 92.9) | 90.8 (89.9, 91.8) | -0.3 (-3.0, 2.4) |
Abbreviations: eGFR, estimated glomerular filtration rate; mGFR, measured GFR; eGFRcr, estimating GFR using creatinine; eGFRcys, estimating GFR using cystatin C; eGFRcr-cys, estimating GFR using creatinine and cystatin C; CI, confidence interval
P30 is a measure of accuracy and is computed as percent (%) of estimates within 30% of the mGFR.
Comparing the Race-based 2009 and Race-free 2021 CKD-EPI Equations
During the transition from the race-based GFR estimating equations to the race-free equations, it is important that laboratories clearly identify which equation was used to report eGFR values. Clinicians may need to continue to compare a patient’s eGFR values to results from older equations so as not to miss changes in kidney function that could be obscured by the use of a new equation with a different estimation output. The following figures show the differences in eGFR values using the previous CKD-EPI equations with age, sex, and race compared to the new CKD-EPI equations with only age and sex coefficients.
As demonstrated in Figure 1, the 2009 CKD-EPI creatinine equation that uses age, sex, and race may overestimate GFR in Black individuals and to a lesser degree in non-Black individuals. The 2021 CKD-EPI creatinine equation that uses age and sex, and omits race, may underestimate measured GFR (mGFR) in Black individuals and overestimate mGFR in non-Black individuals. Therefore, using the 2021 CKD-EPI creatinine equation without the race coefficient may increase differential bias between Black and non-Black participants compared to the 2009 race-based equation. The 2021 CKD-EPI creatinine-cystatin C equation was more accurate than the 2021 CKD-EPI creatinine equation, with smaller differences between race groups.
Figure 1. Comparison of measured vs estimated GFR by race groups across previous and updated CKD-EPI eGFR equations
View full-sized image Courtesy of: Lesley A. Inker, M.D., Tufts University School of Medicine and Tufts Medical Center Figure 1: Comparison of measured vs estimated GFR by race groups across previous and updated CKD-EPI eGFR equations. Shows five graphs comparing mGFR and eGFR for Black individuals and non-Black individuals using previous and updated CKD-EPI eGFR equations in the validation data set for the 2021 CKD-EPI equations.2 The first graph uses the previous 2009 eGFR creatinine equation. The second graph uses the updated 2021 eGFR creatinine equation. The third graph uses the previous 2012 eGFR creatinine-cystatin C equation. The fourth graph uses the updated 2021 creatinine-cystatin C equation. The final fifth graph uses the 2012 eGFR cystatin C equation.
Each graph contains bias, P30, and correct classification data. The equations are referred to by the filtration marker or markers (creatinine [eGFRcr], cystatin C [eGFRcys], or creatinine–cystatin C [eGFRcr-cys]), the demographic factors (age, sex, and race or age and sex) that were used in their development, and the year in which the equations were developed (2009, 2012, or 2021). Bias is defined as the median difference between mGFR and eGFR. A positive sign indicates underestimation of mGFR, and a negative sign overestimation of mGFR. P30 is the proportion of eGFR within 30% of mGFR. Correct classification refers to agreement between mGFR and eGFR categories of more than 90, 60-89, 45-59, 30-44, 15-29, and less than 15 ml per minute per 1.73 m2.
Figures 2 and 3 demonstrate the differences in mean bias and P30 using the old and new equations.
Figure 2. Systematic error (bias) and precision of the 2009 CKD-EPI creatinine equation vs 2021 CKD-EPI creatinine equation across eGFR subgroups
View full-sized image Courtesy of: Lesley A. Inker, M.D., Tufts University School of Medicine and Tufts Medical Center Figure 2: Systematic error (bias) and precision of the 2009 CKD-EPI creatinine equation vs 2021 CKD-EPI creatinine equation across eGFR subgroups. Shows two graphs comparing bias and precision of the previous and updated CKD-EPI eGFR creatinine equations. The upper graph shows bias, measured as the difference between measured and estimated GFR, for the previous 2009 CKD-EPI creatinine equation with race coefficient versus the updated 2021 CKD-EPI creatine equation without race coefficient. The bottom graphs show accuracy, as measured by 1 - P30 or the percentage of estimates within 30% of mGFR, of the previous 2009 CKD-EPI creatinine equation with race coefficient versus the updated 2021 CKD-EPI creatine equation without race coefficient.
The vertical bars indicate 95% confidence intervals. The dotted black line represents the difference in the GFR equation performance by race. eGFR stages include < 30, 30-59, 60-89, and > 90 ml/min/1.73m2. eGFR was defined separately for each equation. The equations are referred to by the demographic factors (age, sex, and race or age and sex) that were used in their development and the year in which the equations were developed (2009 or 2021). The units for GFR are ml/min per 1.73 m2.
Figure 3. Systematic error (bias) and precision of the 2012 CKD-EPI creatinine-cystatin C equation vs the 2021 CKD-EPI creatinine-cystatin C equation across eGFR subgroups
View full-sized image Courtesy of: Lesley A. Inker, M.D., Tufts University School of Medicine and Tufts Medical Center Figure 3: Systematic error (bias) and precision of the 2012 CKD-EPI creatinine-cystatin C equation vs the 2021 CKD-EPI creatinine-cystatin C equation across eGFR subgroups. Shows two graphs comparing bias and precision of the previous and updated CKD-EPI eGFR creatinine-cystatin C equations. The upper graph shows bias, as measured as the difference between measured and estimated GFR, of the previous 2012 CKD-EPI creatinine-cystatin C equation with race coefficient versus the updated 2021 CKD-EPI creatine-cystatin C equation without race coefficient. The bottom graphs show accuracy, as measured by 1 - P30 or the percentage of estimates within 30% of mGFR, of the previous 2012 CKD-EPI creatinine-cystatin C equation with race coefficient versus the updated 2021 CKD-EPI creatine-cystatin C equation without race coefficient.
The vertical bars indicate 95% confidence intervals. The dotted black line represents the difference in the GFR equation performance by race. eGFR stages include < 30, 30-59, 60-89, and > 90 ml/min/1.73m2. eGFR was defined separately for each equation. The equations are referred to by the demographic factors (age, sex, and race or age and sex) that were used in their development and the year in which the equations were developed (2012 or 2021). The units for GFR are ml/min per 1.73 m2.