2.1 Patients

This retrospective quality assurance project included data from 23rd January 2018 to 23rd January 2023 on pediatric patients at Odense University Hospital, a renal transplantcenter for pediatric patients in the western part of Denmark. All patients were identified from the outpatient clinic. The inclusion criteria were age below 19 years, CKD, and/or kidney transplant requiring monitoring of kidney function. The collected data were retrieved from the patients' health journals and included only already accessible data. In this period, all patients had an mGFR in terms of plasma clearance of the radiotracer, available information on contemporary P-cystatin C, gender, height, weight, P-creatinine, P-carbamide, and underlying disease to create a full dataset.

The study conformed to the Declaration of Helsinki. It was approved by The Danish Data Protection Agency. As a quality assurance project, no approval by the ethics committee was required, and due to the retrospective nature, informed consent from the patients was not needed.

2.2 Measured GFR

The mGFR was determined by a nuclear medicine multi-sample examination at Odense University Hospital. The test included an intravenous injection of the radiotracer (3.7 MBq ⁵¹Cr-EDTA, GE Healthcare, or 8 MBq ^99mTc-DTPA, Mallinckrodt Pharmaceuticals) followed by blood sampling at minutes 120, 150, and 240.¹² Blood samples and spill samples were analyzed in duplicate on a gamma counter (2470 Wallac Wizard 2TM, Perkin Elmer) with an energy window 121–159 keV. Standard GFR (mL/min/1.73m²) was calculated according to the formula by Brøchner-Mortensen & Jødal,^{13, 14} body surface area (m²) was calculated from the Haycock formula: BSA(m²) = 0.024265 · W^0.5378(kg) · H^0.3964(cm).¹⁵

2.3 Estimated GFR

CKiD_Crea-CysC, CKiD_U25, and CKiD_Crea were calculated according to the formulas shown in Table 1.

Cystatin C values were analyzed on a Cobas 8000 module c702 (Roche). The principle of the method was particle-enhanced immunoturbidimetry, where human cystatin C agglutinates with latex particles coated with anti-cystatin C-antibodies. The aggregate was determined turbidimetric at 546 nm. The analysis method was standardized to reference material ERM-DA471/IFCC. The total imprecision was determined at two levels: 0.87 mg/L, CV% = 2.28% and 2.82 mg/L, CV% = 1.39% and the lower limit of quantification was 0.40 mg/L. All P-creatinine was measured enzymatically at Odense University Hospital and were all IDMS traceable. The total imprecision was CV% <2. The lower limit of detection was 5 μmol/L. P-creatinine and P-carbamide values obtained within 1 week from the mGFR were accepted. The height of the patient was accepted up to 3 months apart from the mGFR if there was a following height measurement that showed that the height was stable. Regarding the P-cystatin C values, a time gap of 3 months was also accepted as long as several stable P-creatinine values were available to indicate a steady state in P-creatinine as well as in P-cystatin C. We also monitored if the patients' height and weight were stable during the gap time. In the majority of the datasets (68 out of 73), P-cystatin was analyzed within less than 14 days compared to mGFR.

2.4 Statistical analysis

Descriptive statistics comprise mean ± standard deviation (SD) for continuous and roughly normally distributed variables (as judged by histograms with approximating normal distributions), median (minimum-maximum) for continuous, non-normally distributed variables, and frequencies with respective percentages for categorical variables. Bland–Altman analyses were used to assess the performance of both eGFR equations compared to mGFR.¹⁶ We regressed the differences on the mean values and detected a negative slope over the measurement range, suggesting a non-constant bias; therefore, we applied a regression approach for non-uniform differences to derive the Bland–Altman Limits of Agreement.^17-19 Relative differences between the eGFR and mGFR were described with P10 and P30, which show the percentage of eGFR values that differ by <10% or <30% from the corresponding mGFR. All analyses were done with STATA/MP 17.0 (StataCorp, College Station, Texas 77 845 USA).

3.1 Description of patients

A total of 35 patients were assessed for eligibility. Twenty-eight patients were included in a total of 73 datasets because several of the patients had more than one dataset available in the course of our 5-year period. A dataset consisted of an mGFR and two corresponding eGFR values calculated from the retrieved patient data. Nine patients contributed with one dataset each. Six patients had two datasets. Three patients had three datasets. Seven patients had four datasets and three patients had five comparable datasets available. Seven patients (#6, 11, 15, 16, 18, 22, 29) were not included, as the needed data were not available. The included patients were children and adolescents 3.5–18.5 years of age. All patients had CKD, and the majority had a renal transplant. The CKD was due to several known diseases, e.g., congenital renal hypo-dysplasia, posterior urethral valves, and ANCA-associated vasculitis. The datasets were subdivided into GFR categories based on mGFR.² Two measurements in one patient showed normal kidney function. Approximately one-third showed mildly decreased kidney function. Just over 50% were in the category of moderately decreased kidney function. Five patient datasets showed severely decreased kidney function, and two patient datasets revealed definitive kidney failure. Patient characteristics are presented in Table 2.

3.2 Agreement analysis

The mGFR mean ± SD was 56.6 ± 17.9 mL/min/1.73m². The CKiD_Crea-CysC estimate was 61.1 ± 17.4 mL/min/1.73m². Of all the paired observations, 58.9% of the estimates were within P10 and 87.7% within P30 of the mGFR results. The mean difference ± SD was 4.8 ± 8.5 mL/min/1.73m² with the mean value being positive over the whole GFR spectrum (Figure 1). Figure S1 shows a scatterplot of CKiD_Crea-CysC and mGFR values.

The CKiD_U25 estimate resulted in an eGFR of 53.7 ± 15.0 mL/min/1.73m²; 53.4% of the observations were within P10 and 93.2% within P30 of the mGFR values. The mean difference ± SD was −2.9 ± 8.4 mL/min/1.73m² and varied with the GFR so that GFR was overestimated when mGFR was lower than approximately 40 mL/min/1.73m², and underestimated when mGFR was higher than that (Figure 2). Figure S1 shows a scatterplot of CKiD_Crea-CysC and mGFR values.

The original CKiD_Crea equation resulted in an eGFR of 64.0 ± 18.8 mL/min/1.73m². CKiD_Crea overestimated GFR significantly, with a mean difference ± SD of 11.1 ± 12.3 mL/min/1.73m². P10 and P30 were assessed as 24.7% and 54.8%, respectively (Table 3, Figures 3 and 4).

In Figures 1 and 2, datasets belonging to the same patient can be identified, allowing for evaluation of the difference between eGFR and mGFR at different time points. It shows that for some of the datasets of patient #1, the difference between eGFR based on CKiD_U25 and mGFR exceeded the 95% limit of agreement (Figure 2). With CKiD_Crea-CysC this was the case for all of the datasets of patient #1 and some of the datasets of patient #7 (Figure 1). There seemed to be a tendency towards intraindividual differences being at the same level in the majority of the patients.

4.1 Summary of main findings

This study examined how eGFR based on two cystatin C equations performed in comparison to mGFR. In the Region of Southern Denmark, the CKiD_Crea-CysC equation^{20, 21} is the standard used to estimate GFR in pediatric patients. To consider age, we also estimated GFR using CKiD_U25. Furthermore, we compared mGFR with the original CKiD_Crea equation.

Regarding P10 and P30 values, the two GFR estimates were comparable. The CKiD_Crea-CysC tended to overestimate GFR and therefore overrate the kidney function within all the GFR ranges. The CKiD_U25 tended to overestimate GFR in patients with very low kidney function but underestimated GFR in patients with higher GFR.

4.2 Comparison

The findings in this study are similar to others. The P10 and P30 values found were marginally more accurate than previously proven.^{22, 23} For instance, Edward J. Nehus et al. found a P30 of 86.2% and a P10 of 43.7% using the CKiD_Crea-CysC equation.²⁴ George J. Schwartz found similar results using the CKiD_U25 with a P30 of 89.4% and a P10 of 47.1.²⁵

Studies show an association between higher values of P-cystatin C and obesity-related conditions such as type-2 diabetes mellitus (DM2) and metabolic syndrome. The thesis is that adipocytes might increase the secretion of cystatin C as a response to the expansion of the adipose tissue. In this study, none of the patients were diagnosed with DM2 or metabolic syndrome, which eliminates the risk of this association causing an unbalanced bias. Other non-GFR factors that might affect a cystatin C-based eGFR are inflammation, smoking, hypo- and hyperthyroidism, overhydration, and acute kidney injury (AKI). Therefore, the clinical use of a cystatin C-based eGFR should be reconsidered or at least not stand alone in these patient groups.^{4, 26}

Previously, it has been proven in several studies, that the Schwartz CKiD_Crea equation tends to overestimate GFR and therefore kidney function.^{27, 28} In the Foundation for the National Institutes of Health-sponsored CKiD study, the CKiD_Crea equation overestimated GFR by 12 mL/min/1.73m².¹ Our study showed a mean difference of 11.1 mL/min/1.73m² from the mGFR which correlated very well with previous results.

4.3 Strengths and limitations

Our gold standard is not precise in itself. The mGFR by radiotracer has an uncertainty of up to 13% although less with the multi-sample versus the single-sample method.^{14, 19} It is, however, used in daily practice. We only processed two different contemporary GFR estimates, since we chose to focus on the one most used in clinical practice in Denmark, which is CKiD_Crea-CysC. Additionally, we examined the CKiD_U25 to investigate whether it would be a more accurate estimate because it took age into account.^{21, 29}

The rarity of CKD in children is the reason for the small number of patients included. Apart from one dataset, we only included patients with reduced kidney function, and we cannot tell from our results, how the equations perform in patients with normal kidney function. One further limitation was that this study primarily included patients of Caucasian ethnicity, however, a previous study showed that a CKiD estimate that does not consider race had the best overall performance, but also highlighted that further research was needed to conclude the rejection of race-corrected equations.³⁰

Because several patients had several datasets, it allowed the monitoring of the consistency of eGFR versus the mGFR within an individual, in contrast to other similar studies, which often only included one dataset per patient.²⁴

Filler G addresses an important issue regarding mGFR indexed to BSA. The standard is to use the actual weight, but if the ideal weight is better, it needs to be clarified. The concern is that overweight overestimate mGFR and underweight overestimate mGFR. In our study, the majority of patients had an ideal weight within ± 2 SD.³¹

During the time of the study, the radiotracer was changed from ⁵¹Cr-EDTA to ^99mTc-DTPA. However, several studies found that the two clearance methods correlated well (r = .996).³² A recent study concluded that no clinically relevant differences were found between the ^99mTc-DTPA and ⁵¹Cr-EDTA plasma clearances, and that ^99mTc-DTPA can replace ⁵¹CrEDTA.³³

4.4 Perspectives

It was clear that the eGFR was not a perfect substitute for mGFR. However, given that intraindividual differences tended to be in the same order of magnitude over time, estimation of GFR might be a possible means of monitoring the kidney function if there is a combined mGFR and eGFR at the baseline. The results clearly showed that GFR estimates were considerably more inaccurate for patients #1 and #7. Patient #1 suffered from syndromes that cause low muscle mass with a possible impact on P-creatinine, thereby affecting the eGFR. Patient #7, on the contrary, had no obvious reason for the eGFR to be so inaccurate. This example causes doubt about the use of eGFR. Therefore, a baseline mGFR should at least be available, and if the eGFR differentiates significantly, eGFR would supposedly be discouraged as a valid estimate of GFR and kidney function.