Estimated versus measured glomerular filtration rate in children prior to hematopoietic cell transplantation

Abstract

An accurate assessment of kidney function prior to hematopoietic cell transplantation (HCT) can help to properly dose conditioning chemotherapy and follow patients for the development of chronic kidney disease. We cross-sectionally examined 94 children and young adults prior to HCT to compare formal nuclear GFR testing with estimated GFR using creatinine and cystatin C-based equations including the original Schwartz formula and the more recent formulas developed in the Chronic Kidney Disease in Children (CKiD) cohort. The median age of the cohort was 5.9 years (range 0.26–30.5 years). The mean cohort nuclear GFR was 107.4 ± 24.7 ml/min/1.73m², with 18/94 (19.1%) subjects having abnormal kidney function (GFR <90 ml/min/1.73m²) prior to HCT. The creatinine-based original Schwartz and bedside CKiD formulas showed significant bias (95% confidence interval), overestimating the nuclear GFR by 57.4 (49.0–65.8) and 14.1 (7.1–21.1) ml/min/1.73m², respectively. Cystatin C formulas had less mean bias and improved accuracy, but also had decreased sensitivity to detect abnormal kidney function prior to HCT. The Full CKiD equation showed the best performance, with a mean bias of −3.6 (−8.4–1.2) ml/min/1.73m² that was not significantly different from the measured value and 87.7% of estimates within ±30% of the nuclear GFR. While the more recent bedside CKiD formula performed better than the original Schwartz formula, both formulas had poor sensitivity for detecting a low GFR. An abnormal pre-transplant nuclear GFR was not associated with post-HCT acute kidney injury, the need for dialysis, or death in the first 100 days. In conclusion, we observed cystatin C-based equations outperformed creatinine-based equations in estimating GFR in children prior to HCT. However, all formulas had decreased sensitivity to detect impaired GFR. Formal measurement of kidney function should be considered in children and young adults who need an accurate assessment of kidney function prior to HCT.

INTRODUCTION

Chronic kidney disease (CKD) occurs in at least 15% of patients after hematopoietic cell transplantation HCT [1]. Certain chemotherapeutic agents used for conditioning prior to HCT need to be dose-adjusted depending on the glomerular filtration rate (GFR) [2]. Following transplant, an accurate assessment of GFR is needed to dose other medications including antibiotics and calcineurin inhibitor therapy for graft versus host disease (GVHD) prophylaxis, as well as to monitor patients over time for the development of CKD [3].

Outside of research protocols, there are no established guidelines for how to assess kidney function (GFR) before HCT. Available options include serum creatinine, 24 hour urine collections for creatinine clearance, and formal measurements of GFR using an injected nuclear isotope or contrast agent such as iohexol. Some have suggested that a pre-HCT serum creatinine ≤1.5 mg/dL and a creatinine clearance >60 ml/min are preferred prior to starting transplant [4]. While creatinine may have a limited ability to estimate GFR in patients with low muscle mass, formal GFR testing is more costly, invasive, and time consuming [3, 5].

We reported that cystatin C was more accurate than creatinine in estimating GFR in 16 children receiving autologous HCT [6]. Unlike serum creatinine, cystatin C may be independent of muscle mass, but possibly affected by other non-renal factors such as corticosteroid treatment or body weight [5, 7]. Cystatin C has been less studied in patients undergoing HCT [8–12]. Our objective was to expand on our prior work by examining GFR in a larger cohort of children and young adults before HCT which included allogeneic recipients. We used the most recent estimating equations recommended by the 2012 Kidney Disease Improving Global Outcomes (KDIGO) consensus guidelines [13] and focused on comparing creatinine-estimated GFR to a measured nuclear GFR and secondarily included cystatin C-estimated GFR.

MATERIALS and METHODS

Study population

We conducted a cross-sectional analysis of children and young adults who were enrolled in a prospective cohort originally designed to study risk factors for thrombotic microangiopathy after HCT. The cohort included 100 consecutive children and young adults receiving a HCT at Cincinnati Children’s Hospital Medical Center (CCHMC) from September 2010 to December 2011. Of these 100 subjects, 95 had a nuclear GFR performed for clinical indications prior to transplant. Two autologous recipients have been reported in our prior study [6]. Clinical data were recorded from the medical record and included age, gender, primary diagnosis, race, height, weight, exposure to prior chemotherapy, number of prior HCT (if any), corticosteroid therapy prior to transplant, and creatinine and blood urea nitrogen values. We also captured outcome data including the development of acute kidney injury (AKI, defined as a doubling of each subject’s baseline serum creatinine), the need for dialysis, or death within the first 100 days after HCT. Cystatin C was measured on prospectively collected and stored plasma samples, as described below.

We included subjects with a creatinine and/or cystatin C measurement within 3 days of nuclear GFR testing. For subjects with a >3 day lag between the measured and the creatinine and/or cystatin C-estimated GFRs, we only included subjects who were not admitted to the hospital and had not received nephrotoxic antibiotics between testing. In addition, for analyses incorporating cystatin C, we excluded subjects with cystatin C testing performed after the start of conditioning and also those with unstable kidney function (defined as an intra-subject creatinine standard deviation of ≥0.1, as reported previously [6]) between cystatin C and nuclear GFR testing. The research was approved by the Institutional Review Board at CCHMC.

Measurement of GFR

GFR was measured using ^99mTc-labeled diethylene triamine pentaacetic acid (DTPA). The procedure was performed and clearance was calculated as previously described [6, 14]. In brief, after a single injection of DTPA, plasma samples were obtained at approximately 120, 150, 180, and 210 minutes. GFR was calculated using a slope-intercept, or one compartment, method based on the terminal portion of the plasma clearance curve. A quadratic correction factor was then used to adjust the slope-intercept GFR to a two-compartment model using the methods of Brochner-Mortensen with coefficients that were recently validated in children by Schwartz et al (GFR corrected = 0.995*GFRmeasured − 0.001159(GFRmeasured²) [3].

Estimation of GFR

Primary Analysis: Creatinine-estimated GFR versus measured nuclear GFR

Our primary objective was to compare serum creatinine-estimated GFR to each subject’s measured nuclear GFR. Serum creatinine values were obtained clinically and measured with an enzymatic assay (Ortho Vitros Fusion, Rochester, NY) in the CCHMC clinical laboratory. In subjects with more than one measurement, we examined the serum creatinine value closest to the date the nuclear GFR was performed. In the entire cohort, GFR was separately estimated with the original Schwartz formula[15] and the new “bedside” Chronic Kidney Disease in Children (CKiD) formula [16]. For subjects ≥18 years of age, we also estimated GFR using the Modification of Diet in Renal Disease (MDRD) formula[17] and the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) creatinine-only formula [18]. The GFR estimating equations are shown in Table 1.

Table 1.

GFR estimating equations

Age <18 years
Creatinine-based formulas
Original Schwartz	k*height (cm)/Scr, where k is 0.7 for males 13–18 years of age, 0.45 for children <1 year of age, and 0.55 otherwise.
Bedside CKiD	0.413*height (cm)/SCr
Cystatin C-based formula
CKiDcys	70.69*(cys)−0.931
Combined cystatin C and creatinine formula
Full CKiD	39.1(height (m)/SCr)0.516*(1.8/cys)0.294*(30/BUN)0.169*(1.099)male*(height (m)/1.4)0.188
Age ≥18 years
Creatinine-based formulas
MDRD	175*Scr−1.154*age−0.203*0.742(if female)*1.212(if African-american)
CKID-EPIcreat	141*(minimum of Scr/k or 1)α*(maximum of Scr/k or 1)−1.209*0.993age*1.018 (if female)x1.159(if black), where k is 0.7 for females and 0.9 for males and α is −0.329 for females and −0.411 for males.
Cystatin C-based formula
CKD-EPIcys	133*(minimum of cys/0.8 or 1)−0.499*(maximum of cys/0.8 or 1)−1.328*0.996age*0.932 (if female)
Combined cystatin C and creatinine formula
CKD-EPIcreat-cys	135*(minimum of Scr/k or 1)α*(maximum of Scr/k or 1)−0.601*(minimum of cys/0.8 or 1)−0.375*(maximum of cys/0.8 or 1)−0.7110.995age*0.969 (if female)x1.08(if black), where k is 0.7 for females and 0.9 for males and α is −0.248 for females and −0.207 for males.

GFR in ml/min/1.73m² cys, Cystatin C (mg/L); BUN, blood urea nitrogen (mg/dL); cm, centimeters, m, meters; SCr, serum creatinine (mg/dL); CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; MDRD, Modification of Diet in Renal Disease; CKiD, Chronic Kidney Disease in Children

Secondary Analysis: Cystatin C-estimated GFR versus measured nuclear GFR

We included comparisons using cystatin C-estimated GFR as a secondary analysis because there was a lag between cystatin C testing and the nuclear GFR measurement in some subjects (see below). Plasma had been collected from each subject prior to HCT and was stored at −80C prior to cystatin C testing. Cystatin C concentrations were measured in the CCHMC Division of Nephrology clinical laboratory using particle-enhanced immunonephelometry (Siemens Healthcare Diagnostics Products, Germany). To compare the performance of cystatin C and creatinine-estimated GFR to the nuclear GFR, we used the creatinine measure closest in time to the cystatin C result. In the entire cohort, GFR was separately estimated with the CKiD cystatin C only formula and the Full CKiD formula combining cystatin C, creatinine, and blood urea nitrogen [7]. For subjects ≥18 years of age, we also estimated GFR using the CKD-EPI cystatin C only and the CKD-EPI combined creatinine and cystatin C formulas (Table 1) [19].

Statistical analyses

Descriptive statistics for continuous variables were reported as medians (interquartile ranges, IQR) and n (percentages) for categorical variables. The GFR values were reported as means (standard deviations). To compare the estimating equations to nuclear GFR estimation, mean bias and 95% limits of agreement were calculated according to the methods of Bland-Altman [20]. We also computed each equation’s accuracy as the proportion of the estimated GFR values within ±10% and ±30% of the gold standard, which are accepted criteria for evaluating equation performance [13]. In addition, we calculated the sensitivity and specificity of each equation to detect a decreased nuclear GFR (i.e. abnormal kidney function), defined as <90 ml/min/1.73m² in those ≥2 years old [13]. For children <2 years old, we used published age-based normal values [21] to define abnormal kidney function as less than two standard deviations below the mean. We also examined a GFR cutoff of <50–60 ml/min/1.73m² as this threshold is used for dosing adjustment of medications used in HCT conditioning including fludarabine, melphalan, platinum drugs, etoposide, carmustine, and cytarabine [22, 23]. Finally, we examined the association between a decreased pre-transplant nuclear GFR and post-HCT renal outcomes (defined separately as AKI, dialysis, or death in the first 100 days) using univariate logistic regression.

Estimating equation performance was examined in the cohort as a whole and separately using pediatric and adult-specific formulas in those <2 and ≥18 years of age. Analyses were conducted using SAS (version 9.3, SAS Institute, Cary, North Carolina) and STATA software (version 12, College Station, Texas). A p-value <0.05 was considered statistically significant.

RESULTS

Subjects

For the primary analysis, 77/95 (81.1%) subjects had a creatinine value obtained clinically within 3 days of nuclear GFR testing and were included. In the remaining 18 subjects, there was a median of 7 days (IQR 5–13 days) between the nuclear GFR and serum creatinine testing. One of these 18 subjects was excluded due to admission for fever, leaving 94 subjects included in the primary analysis. For the secondary analysis, we excluded subjects because: 6 had their cystatin C obtained after the start of conditioning, 7 had an intra-subject creatinine standard deviation of ≥0.1 between measures, and/or 10 were admitted to the hospital between nuclear GFR and cystatin C testing. This left a total of 73 subjects included in the secondary analysis.

The clinical, demographic, and laboratory characteristics of the included subjects are shown in Table 2. The median age of the 94 subject cohort was 5.9 years (IQR 2.8–15.1 years) with the youngest subject 0.26 years and the oldest subject 30.5 years. Most of the subjects (78/94, 83.0%) were <18 years of age, with 19/94 (20.2%) subjects <2 years of age. Most of the study population were male, Caucasian, and diagnosed with a malignancy. Approximately 8% of the subjects were receiving treatment with corticosteroids at the time of cystatin C testing, about half had received prior chemotherapy, and 3% had received a prior HCT.

Table 2.

Clinical and demographic characteristics of the study cohort

	Primary analysis: Creatinine vs Nuclear GFR	Secondary analysis: Cystatin C vs Nuclear GFR
	N=94	N=73
Age (years)	5.9 [2.8–15.1]	6.7 [3.0–15.1]
Male gender	60 (63.8%)	44 (60.3%)
Primary Disease
Malignancy	36 (38.3%)	31 (42.5%)
Immunodeficiency	32 (34.0%)	24 (32.9%)
Bone marrow failure	22 (23.4%)	15 (20.6%)
Genetic/metabolic	3 (3.2%)	2 (2.7%)
Benign hematological	1 (1.1%)	1 (1.4%)
Race
Caucasian	79 (84.0%)	63 (86.3%)
African American	12 (12.8%)	9 (12.3%)
Other	3 (3.2%)	1 (1.4%)
Height (cm)	112.6 [87.0–156.6]	121.7 [91.2–158.4]
Weight (kg)	22.7 [13.7–55.7]	24.9 [14.3–56.8]
Prior chemotherapy	46 (48.9%)	39 (53.4%)
Prior HCT	3 (3.2%)	2 (2.7%)
Serum creatinine closest to nuclear GFR testing (mg/dL	0.4 [0.3–0.6]	0.4 [0.3–0.6]
Serum creatinine closest to cystatin C testing (mg/dL)	-	0.4 [0.3–0.6]
Blood urea nitrogen closest to cystatin C testing (mg/dL)	-	11 [8–14]
Cystatin C (mg/L)	-	0.69 [0.61–0.78]
Steroid therapy at time of cystatin C	-	6 (8.2%)

Data shown as median [interquartile range] or n (%)

GFR, glomerular filtration rate

Timing of the GFR evaluations

The nuclear GFR was performed a median of 19 days (IQR 13–32 days, range 1–172 days) prior to the start of conditioning chemotherapy. Cystatin C testing was performed on frozen plasma samples that had been collected a median of 2 days (IQR 0–7 days) prior to the start of conditioning chemotherapy. The serum creatinine and blood urea nitrogen values closest to the cystatin C testing were obtained clinically a median of 0 days (IQR 0–1 days, range 0–15 days) from cystatin C testing.

Performance of the GFR estimating equations

The mean cohort (n=94) nuclear GFR was 107.4 ± 24.7 ml/min/1.73m². The mean GFR was 107.9 ± 26.3 ml/min/1.73m² in the 78 subjects <18 years of age, 104.5 ± 14.5 ml/min/1.73m² in the 16 subjects ≥18 years of age, and 110.5 ± 14.5 ml/min/1.73m² in the 19 subjects <2 years of age. Using a cutoff of <90 ml/min/1.73m² for those ≥2 years of age and age-based normative values for those <2 years of age, 18/94 (19.1%) subjects had abnormal kidney function prior to transplant.

Primary Analysis: Creatinine-estimated GFR versus the nuclear GFR

Table 3 summarizes the performance of each estimating equation in the entire cohort. The creatinine-based original Schwartz and bedside CKiD formulas overestimated the nuclear GFR with a significant mean bias (95% confidence interval) of 57.4 (49.0–65.8) and 14.1 (7.1–21.1) ml/min/1.73m², respectively. Although the mean bias of the bedside CKiD formula was better than the original Schwartz formula, its performance was not optimal as its accuracy was moderate with 67.0% of the estimates within ±30% of the measured GFR.

Table 3.

Performance of pediatric GFR estimating equations in the entire cohort

Formula	Mean GFR ± SD	Mean Bias (95% CI)	95% LOA	30% accuracy† (%)	10% accuracy† (%)
Primary Analysis: Creatinine formulas (n=94)
Nuclear GFR	107.4 ± 24.7	-	-	-	-
Original Schwartz	164.8 ± 43.5*	57.4 (49.0–65.8)	−22.7–137.5	22.3	9.6
Bedside CKiD	121.4 ± 34.7*	14.1 (7.1–21.1)	−53.0–81.1	67.0	29.8
Secondary Analysis: Cystatin C formulas (n=73)
Nuclear GFR	108.7 ± 23.5	-	-	-	-
CKiD cystatin C only	100.2 ± 20.7*	−8.6 (−13.6– −3.5)	−51.0–33.8	87.7	41.1
Full CKiD	105.1 ± 18.1	−3.6 (−8.4–1.2)	−43.6–36.4	87.7	42.5

GFR, glomerular filtration rate; SD, standard deviation; CI, confidence interval; LOA, limits of agreement; CKiD, Chronic Kidney Disease in Children study

GFR in ml/min/1.73m²

^*Significantly different from nuclear GFR (p<0.05)

^†30% accuracy is the percentage of estimated GFR values falling within 30% of the nuclear GFR and 10% accuracy is the percentage of estimated GFR values falling within 10% of the nuclear GFR

The original Schwartz formula detected none of the 18 subjects with abnormal kidney function prior to transplant, (sensitivity 0%) and the bedside CKiD detected 4/18 (sensitivity 22.2%). There were only 2 subjects with a pre-HCT nuclear GFR <60 ml/min/1.73m². One was 4.5 months old and would therefore be classified as having normal kidney function based on age. The other subject had a nuclear GFR of 48.6 ml/min/1.73m² that would not have been detected by either creatinine-based GFR formula (sensitivity 0%).

Secondary Analysis: Cystatin C and creatinine-estimated GFR versus the nuclear GFR

The performance of CKiD GFR estimating equations including cystatin C alone or in combination with creatinine (Full CKiD) are shown in Table 3. Overall, equations including creatinine and cystatin C performed better than equations with creatinine alone. The Full CKiD formula had the best mean bias at −3.6 (−8.4–1.2) ml/min/1.73m², which was not significantly different than the nuclear GFR. The CKiD cystatin C only formula performed comparably to the Full CKiD formula in terms of accuracy, although it did slightly underestimate renal function (significant mean bias of −8.6 ml/min/1.73m²). Both the Full CKiD and cystatin C only formulas had good accuracy, with 87.7% of values falling within ±30% of the nuclear GFR and >40% falling with ±10% of the nuclear GFR. However, they did not adequately detect abnormal kidney function, with a sensitivity of 60.0% for the cystatin C only formula and a sensitivity of 40.0% for the Full CKiD.

Performance of the estimating equations in children <2 and ≥18 years of age

The CKiD formulas were originally validated in children ≥2 years of age [16]. As our cohort contained a subset of subjects outside this age range, we separately examined the performance of the GFR estimating equations in those <2 years of age and ≥18 years of age. In the primary analysis, among the children <2 years of age (n=19), the original Schwartz and bedside CKiD formulas again overestimated the nuclear GFR with a significant mean bias of 53.6 (22.0–85.3) and 24.9 (1.0–48.7) ml/min/1.73m², respectively. In the secondary analysis (n=13), both the CKiD cystatin C only and Full CKiD formulas significantly underestimated the nuclear GFR with a significant mean bias of −20.6 (−38.6– −2.6) and −18.8 (−33.4– −4.2) ml/min/1.73m², respectively. No formula, creatinine or cystatin C-based, identified the 2 young children with abnormal kidney function.

In the primary analysis, among the subjects ≥18 years of age (n=16), the pediatric-specific bedside CKiD performed better than both the original Schwartz and the two adult-specific formulas (MDRD and CKD-EPI creatinine). Specifically, the bedside CKiD had a mean bias of −7.3 (−18.4–3.9) ml/min/1.73m² that was not significantly different from the nuclear GFR, an acceptable ±30% accuracy of 87.5%, and correctly identified 2/3 (sensitivity of 66.7%) of the subjects with a nuclear GFR <90 ml/min/1.73m². The original Schwartz, MDRD, and CKD-EPI equations significantly overestimated GFR with mean biases of 45.6 (30.8–60.5), 20.1 (8.1–32.0), and 21.3 (12.0–30.6) ml/min/1.73m², respectively. In the secondary analysis (n=13), the pediatric-specific cystatin-C based formulas performed better than the two adult CKD-EPI cystatin C formulas. Specifically, the Full CKiD had a mean bias of −4.6 (−13.9–4.7) ml/min/1.73m² that was not significantly different from the nuclear GFR, while the CKiD cystatin C only formula again slightly underestimated nuclear GFR with a mean bias of −11.4 (−21.1– −1.7) ml/min/1.73m². Both formulas were accurate within ±30% of the nuclear GFR in all cases and correctly identified 1/2 (sensitivity of 50.0%) of the subjects with a nuclear GFR <90 ml/min/1.73m². Both the CKD-EPI cystatin C formulas overestimated nuclear GFR, with mean biases of 14.9 and 19.4 ml/min/1.73m², although they had a ±30% accuracy of >75%.

Association between pre-HCT nuclear GFR and post-HCT outcomes

Among the 95 subjects in our cohort, 50 (52.6%) developed AKI in the first 100 days after HCT, as defined by at least one serum creatinine value ≥2 times each subject’s baseline value. A total of 6 (6.3%) subjects required dialysis and a total of 10 subjects (10.5%) died within the first 100 days after transplant. An abnormal baseline nuclear GFR was not significantly associated with an increased odds (OR [95% confidence interval]) for developing post-HCT AKI (OR 0.50 [0.18–1.44]), needing dialysis (OR 2.28 [0.38–13.55], or dying (OR 3.38 [0.84, 13.56) in the first 100 days after HCT.

DISCUSSION

We evaluated the performance of the most current GFR estimating equations in a cohort of children and young adults prior to HCT. We observed that although most of the cohort had normal kidney function prior to HCT, almost 1 in 5 subjects had a nuclear GFR less than normal for age before starting transplant. Formulas relying on the serum creatinine (original Schwartz formula and bedside CKiD) significantly overestimated the nuclear GFR and had a poor sensitivity for detecting abnormal kidney function as measured by the nuclear GFR. Equations including cystatin C performed better than those including only creatinine, although they still failed to detect an abnormal GFR in a significant percentage of subjects. Finally, pediatric-specific formulas performed better than adult-specific formulas in the subset of subjects >18 years of age.

Our work expands on the few studies examining GFR prior to HCT. Kletzel et al compared the original Schwartz formula to a DTPA-measured GFR in 95 children. They recommended using the Schwartz formula, even though it overestimated the nuclear GFR by a mean 14 ml/min/1.73m² [24]. Qayed et al compared the original Schwartz and the bedside CKiD formulas to a DTPA-measured GFR in 107 children >1 year of age. The Schwartz formula overestimated nuclear GFR by 28 ml/min/1.73m² and the bedside CKiD underestimated the GFR by 10 ml/min/1.73m² and they therefore concluded that nuclear GFR testing should be preferred for most patients prior to HCT [25]. Hazar et al compared cystatin C (Filler equation) and a nonstandard creatinine formula to the DTPA-measured GFR in 34 children. The cystatin C-estimated GFR was similar to the nuclear GFR while the creatinine-based GFR overestimated kidney function [10]. Finally, in 16 children undergoing autologous HCT, we found that although cystatin C-based equations underestimated nuclear GFR, they outperformed the creatinine-based bedside CKiD formula, which significantly overestimated GFR [6]. Notably, these studies consistently demonstrated the limitations of creatinine-based methods to accurately estimate GFR, with poor sensitivity and correlation with nuclear GFR in this patient population.

To our knowledge, this is the largest study estimating kidney function with cystatin C in the HCT population. Cystatin C is a small housekeeping protein produced by all nucleated cells in the body at a constant daily rate. It is freely filtered by the glomerulus, independent of muscle mass, minimally protein bound, and metabolized completely by tubular cells after filtration, making it an attractive endogenous marker to estimate GFR [21]. Cystatin C has proven superior to creatinine in detecting impaired GFR, especially in specific patient populations [26]. For example, cystatin C has demonstrated significantly improved sensitivity to identify renal impairment in children and adults receiving chemotherapy for malignancy [27, 28]. Similarly, our results demonstrate that the CKiD cystatin C only equation in particular outperformed all creatinine-based equations in identifying decreased GFR in children undergoing HCT. Furthermore, both the CKiD cystatin C only and Full CKiD formulas offered improved accuracy compared to creatinine-based equations. However, although our results suggest that cystatin C may be a more useful marker than creatinine for monitoring kidney function, at the current time, cystatin C-based methods do not offer sufficient accuracy to supplant formal nuclear GFR testing prior to HCT.

We estimated GFR with the most recent creatinine and cystatin C-based equations recommended by the 2012 KDIGO consensus guidelines [13]. These equations have been validated in large populations of children and adults, many of whom have primary kidney disease [7, 16, 19]. However, as with any estimating equation developed through statistical modelling, the equations will perform best in the original population and may not be applicable to other groups of patients [5]. For example, the CKiD equations were generated using data from children with CKD and a median iohexol-measured GFR of 40 ml/min/1.73m² [16], far lower than the average kidney function in our cohort. The most recent CKD-EPI equation was developed in a cohort of adults with a mean age of almost 50 years [19], possibly explaining why adult estimating equations did not perform well in our young adult subjects.

It remains unknown if measuring GFR should become routine practice prior to HCT. While GFR estimation methods using endogenous markers (creatinine and/or cystatin C) are becoming more standardized and widely available [7], individual transplant centers may not have the ability to measure GFR in all their patients. Cost should also be a consideration, as we previously reported that nuclear GFR testing is at least 10 times more expensive than estimating GFR [6]. Importantly, more evidence is needed to determine if screening for pre-HCT kidney dysfunction will help to identify those at highest risk for poor outcomes. While we did not observe that a low pre-transplant nuclear GFR predicted later AKI, dialysis, or death in the first 100 days, our sample size was relatively small and we were likely underpowered to detect significant associations. Finally, it seems reasonable to measure GFR in patients receiving certain conditioning agents, such as platinum chemotherapy [2]. Accordingly, a recent review recommended that cyclophosphamide, fludarabine, and melphalan also require dose adjustments in patients with kidney dysfunction. However, the authors noted that more research in this area is needed as the available literature, especially in children, primarily includes small, retrospective reports [29].

Several limitations of our study deserve mention. First, our cohort included a relatively small number of young adults and children <2 years of age, so the findings may be less applicable to these age groups. We used DTPA-measured GFR as the gold standard, similar to previous reports in patients being evaluated for HCT [3, 24, 25]. However, unlike EDTA (which is not available in the US), DTPA may not perfectly correlate with inulin clearance, perhaps due to differences in the compound’s formulation [30]. As more data becomes available on the use of iohexol, a non-radioactive, safe to administer contrast agent currently used by the CKiD research study, it may also become the marker of choice in the clinical setting [7]. Because our study was a cross-sectional analysis of an existing cohort, there was a significant time lag between some of the measurements, potentially introducing bias if kidney function was not stable. This was less of an issue in comparing creatinine-based GFR with the nuclear GFR as the lag was minimal. However, in the analyses including cystatin C, the lag was longer. We tried to account for this by excluding subjects with “unstable” kidney function between measures. Finally, only 1 subject had a pre-HCT GFR <60 ml/min/1.73m², limiting our ability to draw conclusions about which GFR method is optimal to detect patients requiring dose-adjustment to their conditioning chemotherapy.

In conclusion, we observed that creatinine-based GFR estimating equations performed poorly in children prior to HCT. Cystatin C-based equations performed better, but still had low sensitivity for detecting an abnormal GFR before transplant. There are currently no clinical guidelines for the most effective method to assess kidney function prior to HCT in children or adults. We suggest that formal assessment of kidney function should be considered in children and young adults needing an accurate measure of GFR prior to HCT to properly dose certain medications used for conditioning. We recommend that each center develop a protocol for assessing kidney function prior to HCT with input from the transplant team, nephrology, and nuclear medicine. Only prospective studies designed to measure creatinine, cystatin C, and formal GFR testing at the same time (or on the same sample) and correlating this information with defined clinical endpoints will definitively answer these questions and hopefully improve kidney-related outcomes for this high-risk population.