Chronic kidney disease (CKD) affects 9.1% of the global population1 and 15% of the Spanish population.2 It is associated with a high risk of adverse events and is well-known for the poor outcomes in affected patients.3 In Spain, 1% of these patients require renal replacement therapy (RRT), which consumes up to 5% of the health care budget.4 Among the options for RRT, 55% of patients with CKD undergo functional kidney transplantation (KT), including 3688 KT procedures performed in 2023.5

Regardless of whether KT is performed, renal function is assessed using the glomerular filtration rate (GFR) and proteinuria. The Kidney Disease: Improving Global Outcomes (KDIGO) criteria are used for CKD staging on the basis of these parameter.1 Proteinuria and albuminuria are measured directly, while the GFR can be either measured directly (mGFR) or estimated (eGFR) using validated equations.

The mGFR, which is precise but complex to obtain (currently requiring inulin clearance or the use of isotopes),6 is reserved for specific cases in specialized centers. The eGFR, on the other hand, can be obtained both in hospital settings and in outpatient centers. The eGFR is key in KT for understanding and anticipating comorbidities associated with CKD; by staging CKD, we can better understand the severity of the disease that can be useful for different aspects of the medical community beyond nephrology, for example, for guiding medication adjustment, diagnostic testing, and the development of appropriate therapeutic strategies.

The CKD-EPI 2009 equation remains the most commonly used formula for estimating the GFR in clinical practice regardless of the use of KT; however, while it exhibits advantages over other formulas, it also demonstrates a number of disadvantages.6 This equation has undergone multiple updates, such as the elimination of the racial factor in 2021,7 and while its accuracy in children and elderly individuals has diminished, that for individuals aged 18–70 years remains validated.8 That said, to reduce the imprecision in estimating the GFR, more reliable formulas that include the entire age spectrum are needed.

In 2020, the European Kidney Function Consortium (EKFC) proposed a new equation based on the European population while using the same parameters as those of the CKD-EPI 2009: serum creatinine level, age and sex. The EKFC equation aims to overcome the limitations of the CKD-EPI with respect to age, its validity at advanced ages, and its general accuracy with respect to the mGFR.9 However, few studies have compared the two formulas in transplant patients, who are generally excluded from the development of these equations. Our objective was to evaluate the clinical impact of the two equations in terms of behavior, their relationship with comorbidities, and the classification of CKD in our KT cohort.

We conducted a retrospective observational study of patients who underwent KT at our center between January 1, 2006, and February 15, 2022, and whose graft survival was greater than one year. We excluded those under 18 years of age and those who were non-Caucasian. We collected analytical parameters one year post-transplantation and input serum creatinine level (measured using the standardized enzymatic method), sex and age into the CKD-EPI 2009 and EKFC equations to calculate the eGFR (Appendix B, Table S1).

We collected data on the patients’ serum hemoglobin, bicarbonate, phosphorus and potassium levels to define the following comorbidities associated with CKD: anemia (hemoglobin < 11 g/dL in women, <12 g/dL in males), hyperphosphatemia (phosphorus > 4.5 mg/dL), hyperkalemia (potassium > 5 mEq/L) and acidosis (bicarbonate < 22 mEq/L). We evaluated the discriminative capacity of both equations for these comorbidities with receiver operating characteristic (ROC) curve analysis and calculation of the area under the curve (AUC).

We studied the association between eGFR values calculated with each equation using Spearman’s correlation and their concordance and interchangeability using Lin’s intraclass correlation coefficient (ICC).

Bias between the equations was defined as the difference between the GFR values obtained by the CKD-EPI 2009 and those obtained by the EKFC (GFR-CKD-EPI 2009 – GFR-EKFC). We constructed a Bland‒Altman plot to detect systematic biases or dispersion patterns. We also constructed a scatter plot with LOESS adjustment to visualize the relationship between age and bias. We evaluated the statistical significance of the bias using the Wilcoxon test for paired data.

In subgroup analyses, we used the Mann‒Whitney U test for sex and diabetes mellitus (DM) groups and the Kruskal‒Wallis test for age group, followed by the post hoc Mann‒Whitney U test with Bonferroni correction for post hoc analyses between paired groups. We constructed a multiple linear regression model with bias as the dependent variable and age, sex and DM as independent variables to evaluate the simultaneous influence of these three variables.

We used contingency tables to analyze how the EKFC stages CKD with respect to the CKD-EPI 2009-based CKD staging. We used the chi-square test (χ2) to assess the associations and the weighted kappa coefficient (κ) as a measure of agreement between equations, with > 0.80 considered very good agreement.

SPSS and MedCalc were used for statistical analysis. We present quantitative variables as the mean ± standard deviation or the median and interquartile range (IQR) according to the distribution of the data. Qualitative variables are presented as absolute frequencies and percentages. This study was approved by the Ethics Committee of our center (2025.139).

Between January 2006 and February 2022, 628 patients underwent KT and had graft survival greater than 1 year. Twelve non-Caucasian patients were excluded. In total, the data of 616 patients were analyzed; their characteristics are shown in Table 1. A total of 31.2% (192) of the patients were women, and 41.2% (254) had diabetes. A total of 14.1% (87) were between 18 and 40 years old, and 37.3% (230) were > 60 years old. According to the CKD-EPI 2009, 54.3% of patients had stage 3 CKD.

The eGFRs obtained with both equations were strongly correlated, with a Spearman coefficient of 0.9956 (p < 0.001) and an ICC of 0.9945 (95% CI: 0.9936–0.9952), as represented in Appendix B, Fig. S1, and remained > 0.99 in all the subgroups analyzed (age, sex and DM).

According to the Wilcoxon signed-rank test, the eGFR values obtained with the two equations differed significantly (Appendix B, Table S2). The median bias was +0.2 mL/min [−1.0–1.6], with extreme values of −3.2–9.1 mL/min, indicating that compared with the EKFC, the CKD-EPI tended to estimate higher values.

In stratified analyses, the bias remained statistically significant among women (0.6 mL/min; p < 0.001) and diabetic patients (0.4 mL/min; p = 0.046) according to the Mann‒Whitney test. After adjustment by multiple regression, the differences were maintained for sex and age (p < 0.001) but not for DM (p = 0.624). With respect to age groups, we did not detect significant differences in the bias between patients aged 18–40 years and those aged > 40–60 years, but compared with patients aged > 60 years, both groups showed significant differences (Appendix B, Fig. S2), where the bias was greater (median: 1.4 mL/min). The relationship between bias and age is represented by a V curve (Fig. 1): the bias decreases from 20 to 40 years of age and increases thereafter.

Figure 1.

Scatter plot with LOESS adjustment for age and bias, defined as the difference between the outputs of the CKD-EPI 2009 and EKFC equations, of the estimated glomerular filtration rate.

Bland‒Altman analysis (Fig. 2) revealed a nonlinear relationship, with an inflection point at an average GFR 45 mL/min/1.73 m2, with bias in favor of the CKD-EPI 2009 for higher values and of the EKFC for lower values. The dispersion was greater for values indicative of stages 1–3a CKD: the median (IQR) bias was +2.3 mL/min (0.9–5.4) in stage 1, +1.2 mL/min (0.3–2.7) in stage 2, +0.5 mL/min (-0.6–1.7) in stage 3a, +0.2 mL/min (−0.9–1.1) in stage 3b and −1.0 mL/min (−1.5 to −0.3) in stage 4.

Figure 2.

Bland‒Altman plot of the estimated glomerular filtration rate (GFR); specifically the difference between the GFR estimated with the CKD-EPI 2009 and the GFR estimated with EKFC (bias) for each patient as the ordinate and the average of the two values for each patient as the abscissa. The LOESS fit line was obtained with the Tri-cube weight function (SPSS).

The predicted CKD stage showed very good agreement between the two equations (weighted κ = 0.946). In all the subgroups, κ exceeded 0.9, with the lowest being 0.921 in those > 60 years. The results of the χ2 test were significant (p < 0.001) both overall and in all the subgroups.

A total of 3.57% of patients were reclassified to a higher stage of CKD with the EKFC classification, whereas 2.92% were reclassified to a lower stage. As shown in Fig. 3, reclassification to a higher stage mainly occurred for those with higher GFRs, whereas reclassification to a lower stage mainly occurred for those with lower GFRs. This pattern was maintained when patients were stratified by age and was more pronounced in those > 60 years.

Figure 3.

Stacked bar graph of CKD stage reclassification. The global sample with the stages and the gross number of patients according to whether they remained in the same stage is shown in Figure A; reclassification was performed with the EKFC equation and is depicted as better or worse with respect to the CKD-EPI 2009 CKDEPI equation, with the percentage with respect to the total on the vertical axis; note that a greater percentage of patients are reclassified to a lower stage with better renal function, whereas the opposite occurs in those with worse renal function (the data for stage 5 are nonsignificant because of the small sample). The percentage of reclassification in each of the subgroups analyzed is shown in Figure B, alongside the kappa coefficient of agreement in each subgroup; notably, women, those with DM and those aged > 60 years were more likely to be reclassified to higher stages, in contrast to the lower reclassification in those aged < 60 years. DM: diabetes mellitus; CKD: chronic kidney disease.

The AUCs of the eGFRs for both equations were less than 0.8 for all events. We observed that the CKD-EPI 2009 eGFR performed best for predicting hyperphosphatemia (AUC = 0.770; 95% CI: 0.735–0.803) and the worst for predicting hyperkalemia (AUC = 0.619; 95% CI: 0.589–0.648).

Direct comparison with the EKFC eGFR revealed only statistically significant differences in favor of the CKD-EPI 2009 eGFR in predicting acidosis in males (AUC = 0.784 vs. 0.777; p = 0.0436) and in predicting hyperkalemia in individuals > 40−60 years (AUC = 0.601 vs. 0.594; p = 0.0127) (Appendix B, Table S3, Fig. S3).

Our study evaluated the clinical impact of the eGFR-EKFC vs. the eGFR-CKD-EPI 2009 in KT recipients. The values obtained with both equations were positively correlated and showed excellent agreement; however, at times the EKFC equation estimated significantly lower values than the CKD-EPI equation did, especially in women, individuals older than 60 years and those with stages 1−3a CKD. While the inflection point of the EKFC equation is 40 years,10 we observe the greatest difference after 60 years. This is not unexpected, since the EKFC equation is adjusted to adapt the eGFR to the physiological changes associated with age, and thus its estimates more accurately reflect the mGFR in the Caucasian population.9,11–13 In a study comparing the eGFRs obtained with the EKFC equation with those obtained with the CKD-EPI 2009, CKD-EPI 2021 and MDRD equations with respect to the inulin clearance-derived mGFR in a KT population, the MDRD equation showed the best results in all patients except without CKD; the EKFC equation obtained the best results in people over 65 years of age, and overall, its performance was non-inferior performance to MDRD in the overall cohort. The CKD-EPI 2021 equation had the greatest errors with respect to the mGFR.12

Given their design, it is unsurprising that the bias between the equations can be attributed to variations in sensitivity to age but not to comorbidities such as DM, whose significance in the bivariate model disappears when adjusted for age and sex. Both equations are less accurate in diabetic patients, although the reason is unclear, and the lack of accuracy does not favor one or the other in particular,14 suggesting that global renal assessments are needed, rather than guidance solely by the eGFR. Delanaye et al. reported that the EKFC equation is more accurate when cystatin C (CysC) is alone or in combination with creatinine.15 The inclusion of CysC has some limitations,7 however, and creatinine continues to be an essential parameter for using the equations in clinical practice, while the gold standard continues to be the mGFR.1,16

The CKD classification showed very good agreement between the equations. The percentage of reclassified patients was lower than that observed in the literature; Napier et al. reported a reclassification rate of 12.6%, with 97.43% of patients with stage 1−3a disease being reclassified to a higher stage17; this was not the case for patients with stage 3b–5 disease, however as also reported in a study by Pottel et al., where the differences were smaller and less in line with the mGFR.9 Lu et al. reported that the agreement between the equations was best for patients with stage 3b-5 disease; however, as in our cohort, no patient with stage 5 CKD was reclassified to stage 4.18 They performed the same analysis in a Korean population and observed that the EKFC equation was better than the CKD-EPI 2009 equation was for classifying patients into CKD stages except for those with stage 1 disease; specifically, the eGFR was underestimated in the population with better renal function.11 However, among those who were reclassified, a small magnitude of bias was observed in patients with renal failure (eGFR < 60 mL/min/1.73 m2), reducing the risk of possible complications associated with pharmacological adjustment.19

The predictive capacity of the two equations was significantly different (in favor of the CKD-EPI 2009) for acidosis in men and hyperkalemia in those > 40–60 years, although neither performance was clinically relevant; all the AUCs indicate a low level of accuracy. However, the eGFR continues to be the main indicator of renal function 1 as well as the main factor that indicates whether a certain comorbidity may be associated with CKD. To our knowledge, no studies have compared the accuracy of both these equations in predicting these events; we have observed similar results in comparisons with other equations such as the MDRD equation, which had better performance than the CKD-EPI in predicting metabolic syndrome,20 although overall, the MDRD equation provides lower eGFR values, as does the EKFC equation.

Our study has important limitations, including its retrospective design and the lack of comparisons with the mGFR. Even though the goal of the EKFC equation was to provide a convenient solution to the need for two different equations, we did not explore how the eGFR changes from the pediatric age to the adult age, since we do not perform pediatric transplantation in our center.21 The data in Fig. 1 demonstrate the importance of expanding this study to the pediatric population.

In conclusion, our study on KT revealed that the results for both equations are comparable to those reported in studies on nontransplant populations, reinforcing the need to obtain suitable information in this underrepresented population. We detected differences among women, patients aged > 60 years and those with CKD stages 1–3a, while the EKFC result was highly concordant and interchangeable with that of the CKD-EPI 2009 equation in stages 3b–5. The magnitude of the difference is small, particularly in patients with stages 1−3a CKD, but in the population > 60 years, we observed substantial variability; an interdisciplinary approach may be necessary for managing patients in this age range, but interventional studies are needed to confirm this statement.

This research has not received specific support from public sector agencies, the commercial sector or nonprofit entities.