Among our pediatric kidney transplant cohort with stable kidney function and with predominantly steroid-based immunosuppression, CysC-based formulas (CysC alone or in combination with Cr) may estimate GFR better than equations that are solely Cr-based. We also assessed the performance of these equations among those with histologic changes to determine if any of the equations can better identify those who have acute changes in kidney function and may need an allograft biopsy to determine the cause. Our findings show the Gentian CysC formula had a smaller mean bias among the whole cohort and in the subgroup with histologic changes in biopsy. CysC-based U25-CysC, CKiD Cr-CysC, and U25 Cr-CysC equations had higher precision in the whole group and among those with histological changes on biopsy. Therefore, CysC-based formulas (alone or in combination with Cr) appeared to perform as well or better in estimating GFR among those with and without allograft injury compared to equations solely utilizing Cr. The ability to precisely detect acute changes in kidney function in the transplant population will have implications on allograft longevity.

Several studies in adult kidney transplant recipients also found CysC-based equations performed better compared to Cr-based formulas [28,29,30]. Nonetheless, there were variable GFR approximations even between different CysC formulas within each study, which can be attributed to the lack of standardized CysC assays and measurement techniques [27, 29]. In our study, we attempted to provide more accurate estimation of GFR by using a CysC assay that was standardized against IFCC reference material and accounted for variations when using eGFR equations developed before IFCC calibrated CysC values.

In contrast to our findings, previous studies showed CysC-based equations were inferior to Cr-based or combined Cr- and CysC-based formulas. Among pediatric kidney transplant patients on low-dose steroids < 2.5 mg/m2 per day, de Souza et al. concluded CysC-based formulas did not perform better than Cr-based formulas, but rather the CKiD combined Cr and CysC formula performed the best for patients with a GFR < 90 ml/min/1.73 m2 [31]. In a cross-sectional study with 1139 adult kidney transplant recipients > 1 year post-transplant and the majority (86.9%) on low-dose corticosteroid (5 mg or less per day), Cr-based equations alone or in combination with CysC were preferred among adult kidney transplant recipients due to low bias and better accuracy compared to CysC only formulas, which was similar to the findings of Selistre et al. in pediatric kidney transplantation [27, 32]. It is unclear, however, if Selistre’s study subjects were on high-dose steroid therapy which could have affected CysC results. In our study, Bland–Altman plots showed that most Cr- and CysC-based equations have good precision for GFR between 60 and 100 ml/min/1.73 m2, with a tendency in some to overestimate among GFRs > 100 ml/min/1.73 m2. While other studies have demonstrated the Cr-based Bedside Schwartz equation to overestimate GFR in general, our current study showed that this more likely occurs at GFR > 100 ml/min/1.73 m2, where it is somewhat less relevant [31, 33].

Although the CysC-based CAPA equation misclassified CKD stage less in our small cohort, misclassification remains a common occurrence for all equations, generally occurring at least 20% of the time in other studies, compared to approximately 40–50% of the time in this study [33,34,35]. Furthermore, our study highlights the trend of underestimating CKD G1 and overestimating CKD G2. In a pediatric liver transplant cohort, equations utilizing both Cr and CysC misclassified CKD stage the least compared to highest misclassification with Cr-based formulas [34]. Among 198 adult kidney transplant recipients with stable kidney function, the CysC-based Filler estimating equation classified more patients into the correct CKD stage compared to Cr-based equations, with the Filler equation accurately classifying 76% of patients vs. 65% and 69% with Cr-based MDRD and Cockcroft-Gault formulas [35]. Westland et al. evaluated eGFR equations among 77 children with solitary functioning kidney, noting the least misclassification in CysC-based Zapitelli equation at 22%, while urine Cr-clearance had the highest misclassification at 44% [33]. Therefore, CysC-based equations may perform better in classifying CKD stage compared to Cr-based formulas not only in the transplant population, but also in those with CKD. This finding is not surprising, as CysC values are not affected by medications, diet, and muscle mass, which varies widely in the pre-transplant and post-transplant population.

It is important to note that our cohort had relatively well-preserved kidney function with a median GFR of 93.3 ml/min/1.73 m2, compared to many other studies with a mean or median GFR of around 60 ml/min/1.73 m2 [27, 28, 30,31,32]. Although the CAPA and CKiD Cr-CysC equations were validated in subjects with a wide range of measured GFR (< 30 to > 90 ml/min/1.73 m2), the U25 and Bedside Schwartz’s populations had mild–moderate chronic kidney disease, with median measured GFR of approximately 48 ml/min/1.73 m2 (IQR 34–64 ml/min/1.73 m2) and 41 ml/min/1.73 m2 (IQR 32–52 ml/min/1.73 m2), respectively [6, 8, 25, 26]. Based on these studies’ validation groups and our cohort’s collective range of GFR, CysC-based equations may perform better than those that are solely Cr based. Although our study population overall had higher measured GFR, the most current eGFR equation (U25) likely performed well in our cohort in terms of bias, precision, and accuracy because U25 estimates account for changes between sex, age, and height/serum Cr, or 1/CysC – factors believed to strengthen limitations in earlier formulas [8].

Like most of the transplant studies discussed, our patients were largely maintained on steroid-based immunosuppression. One reason for CysC-based equations (with or without Cr) potentially performing better in our cohort compared to solely Cr-based equations could be that CysC is not influenced by the use of low-dose steroids, which was also noted in other studies [16, 28, 30]. While very large doses of glucocorticoids have been described to increase the production of CysC, low and medium doses do not seem to alter the production [11, 12, 15, 16, 34, 36]. Risch et al. showed that 5–10 mg/day of steroid exposure led to higher CysC concentrations compared to those not on steroids among adult kidney transplant recipients and that the rise in CysC was dose dependent. Nonetheless, Risch’s study found that CysC was more accurate than Cr in identifying GFR < 60 ml/min/1.73 m2 [12]. In comparison, Cr concentrations are dependent on a multitude of factors including sex, age, race, nutritional state, and muscle mass [37]. Moreover, medications commonly used in transplantation can interfere with Cr levels. Steroids have a direct catabolic effect leading to lower muscle mass, and tubular secretion of Cr can be blocked by trimethoprim [37]. In this study, the use of low-dose trimethoprim-sulfamethoxazole did not appear to impact the Cr concentrations and thus, the GFR estimates. In contrast, in a retrospective study of 76 adult kidney transplant recipients, Yamanaga et al. concluded very low-dose trimethoprim-sulfamethoxazole for Pneumocystis jirovecii pneumonia prophylaxis reversibly increased Cr by 6% [38]. The cumulative impact of factors that determine Cr levels can certainly affect the utility of Cr in accurately assessing GFR in the pediatric kidney transplant population and thus supports the use of both Cr and CysC post-transplantation.

There were several limitations in our study, with one attributed to not assessing thyroid function [14, 39, 40]. A meta-analysis evaluating serum CysC levels in 1265 patients with thyroid disease and 894 controls revealed higher CysC levels among hyperthyroid subjects compared to lower CysC values in those with hypothyroidism. Furthermore, with treatment of the thyroid disease, CysC levels were notably affected, and therefore, the study concluded that serum CysC could be a marker for monitoring thyroid disease [40]. Additionally, our study had a small sample of 45 patients that only included those who underwent a protocol biopsy in the first 2 years post-transplant, or a for-cause biopsy, which may have introduced selection bias. Furthermore, with over half of our cohort being Hispanic and over 65% male, applicability to other demographics may be limited. The inherent nature of graft attenuation over time could mean that for some subjects in our study, kidney function may not have been in steady state even though their GFR remained within the allotted 20% deviation from prior creatinine measurements. A majority of our subjects had relatively well-preserved graft function, with 91.1% of the group with mGFR > 60 ml/min/1.73 m2 (57.8% of the population with mGFR > 90 ml/min/1.73 m2), therefore, reducing the ability to generalize this study’s findings to those with more advanced stages of CKD. Moreover, our cohort only included subjects 6 months and more post-transplant. As a result, we are unable to extrapolate these results to those at earlier stages post-transplant. Lastly, there was also variability between equations assessed and CysC assays used in our study in contrast to others, resulting in substantial heterogeneity in the performance of equations and difficulty conducting head-to-head comparisons. With such variation in the performance of eGFR equations, one could consider utilizing the same equation to longitudinally follow a patient’s allograft function over time. Pottel et al. found that over a follow-up time of 20 years among 417 adult kidney transplant patients, Cr-based estimating equations correctly predicted the trajectory of measured GFR (urinary clearance of inulin) in kidney transplant recipients; however, they lacked precision and accuracy [41]. Among our cohort, CysC-based equations (CysC alone or in combination with Cr) may better estimate GFR in pediatric kidney transplant recipients, including those with perceived stable allograft function exhibiting changes on biopsy. However, it is important to note that while CysC-based equations appear to perform better, the degree of difference may not be clinically significant. Our data, therefore, does not demonstratively show superiority of one biomarker over the other. Thus, the utilization of CysC could be tailored to the individual patient. For example, a person with reduced or high muscle mass, or on medications that could impact Cr may benefit from CysC assessment as an adjunct measurement of kidney function. In conclusion, our study supports the monitoring of both CysC and Cr post-transplant, which should be validated in future prospective, multicenter clinical trials.