A donor risk index for graft loss in pediatric living donor kidney transplantation

Abstract

Pediatric kidney transplant (KT) candidates often have multiple potential living donors (LD); no evidence-based tool exists to compare potential LD, or to decide between marginal LD vs deceased donor (DD) kidney transplantation (KT). We developed a pediatric living kidney donor profile index (P-LKDPI) on the same scale as the DD KDPI by using Cox regression to model the risk of all-cause graft loss as a function of living donor characteristics and DD KDPI. HLA-B mismatch (aHR per mismatch=_1.041.27_1.55), HLA-DR mismatch (aHR per mismatch=_1.021.23_1.49), ABO incompatibility (aHR=_1.203.26_8.81), donor systolic blood pressure (aHR per 10mm Hg=_1.011.07_1.18), and donor eGFR (aHR per 10 ml/min/1.73m²=_0.880.94_0.99) were associated with graft loss after LDKT. Median (IQR) P-LKDPI was −25 (−56-12). 68% of donors had P-LKDPI<0 (less risk than any DD kidney) and 25% of donors had P-LKDPI>14 (more risk than median DD kidney among pediatric KT recipients during study period). Strata of LDKT recipients of kidneys with higher P-LKDPI had a higher cumulative incidence of graft loss (39% at 10 years for P-LDKPI≥20, 28% for 20>P-LKDPI≥−20, 23% for −20>P-LKDPI≥−60, 19% for P-LKDPI<−60 (log rank p<0.001)). The P-LKDPI can aid in organ selection for pediatric KT recipients by allowing comparison of potential LD and DD kidneys.

1. Introduction

Several studies have shown an association between donor characteristics and graft survival in deceased donor kidney transplantation (DDKT) and living donor kidney transplantation (LDKT) in adults (1–6). The kidney donor risk index (KDRI) and related kidney donor profile index (KDPI) were created to quantify the risk of graft loss due to donor factors in adult DDKT (7). Recently, a living donor KDPI (LKDPI) was created to quantify the risk of graft loss due to donor factors in adults undergoing LDKT, on the same scale as the DD KDPI (8). Thus, the LKDPI aids in organ selection by allowing for comparison of living donor (LD) kidneys to each other and to deceased donor (DD) kidneys.

There is no risk index to quantify the risk of graft loss based on donor characteristics in children undergoing LDKT. Based on Organ Procurement and Transplantation Network (OPTN) data from 2016, 34.2% of kidney transplants received by pediatric patients <18 years of age in the United States were from LD and these patients received 4.4% of all LDKT (9). A pediatric LDKT risk index that combines multiple donor factors to create a summary score of relative donor quality on the same scale as the DD KDPI would provide a clinically interpretable framework for quantifying donor risk. It would be useful in choosing a donor for pediatric patients who have multiple potential LD. Furthermore, the current Kidney Allocation System (KAS) in the United States preferentially allocates high-quality DD kidneys to pediatric patients, so wait times for DD kidneys tend to be shorter for children compared to adults (10). Thus, a pediatric LDKT risk index on the same scale as the DD KDPI would be helpful in making the decision to choose a DD kidney when a potential LD is available. Pediatric patients and families could also use a LDKT risk index to evaluate offers made through kidney paired donation (KPD) (11).

Several recipient factors that affect graft survival differ between children and adults, including immune factors, body size, potential for growth and development, and risk of primary viral infections (12). These differences suggest that the impact of LD characteristics may vary between children and adults. For example, the LKDPI includes a variable for donor recipient weight ratio among donors with a ratio < 0.9, and this variable is irrelevant for the many pediatric KT recipients who are smaller than their adult LD (8). Therefore, the LKDPI is not suitable for use in pediatric KT recipients.

To address the absence of a pediatric-specific LDKT risk index, we conducted a retrospective cohort study of first-time KT recipients in the United States using national registry data. The aims of this study were to determine LD characteristics associated with graft loss among pediatric patients undergoing LDKT, to create a pediatric-specific LD KDPI (P-LKDPI) on the same scale as the DD KDPI, and to assess the predictive validity of the P-LKDPI.

2. Materials and Methods

2.1. Study Population

The study population included 7,155 first-time kidney-only transplant recipients<18 years of age in the United States between January 1, 2005 and December 31, 2015. We excluded recipients with missing donor or recipient characteristics (N=366). We excluded LD recipients whose donors had an estimated glomerular filtration rate<60 ml/min/1.73m² (N=50) or whose donor BMI was outside of the range of 17 to 45 kg/m² (N=351) as these values were assumed to be miscoded. Characteristics of LDKT and DDKT recipients at the time of transplant were compared using rank-sum tests for continuous variables and chi-square tests for binary variables.

The study used data from the Scientific Registry of Transplant Recipients (SRTR). The SRTR data system includes data on all donor, wait-listed candidates, and transplant recipients in the US, submitted by the members of the Organ Procurement and Transplantation Network (OPTN), and has been described elsewhere (13). The Health Resources and Service Administration (HRSA), U.S. Department of Health and Human Services provides oversight to the activities of the OPTN and SRTR contractors.

2.2. Calculation of deceased donor KDPI

The DD KDPI is a percentile score ranging 0%−100% based on 10 donor factors in which a lower KDPI is associated with longer graft survival (14). The percentile is calculated by dividing the KDRI of a particular kidney by the median KDRI of the previous year and mapping the relative risk to a percentile score. Since the KDPI is based on the quality of DD kidneys offered during the prior calendar year, the risk of graft loss associated with a particular KDPI may change over time if the distribution of the quality of kidneys offered changes over time. For this study, we used DD kidneys offered in 2015 as the reference to calculate KDPI in order to ensure consistency in KDPI.

2.3. Model of Donor Risk

We used Cox proportional hazard regression to create a single model of the risk of all-cause graft loss as a function of living donor characteristics for LDKT recipients and deceased donor KDPI for DDKT recipients, adjusted for recipient characteristics including age, sex, race, peak panel reactive antibody (PRA), years of renal replacement therapy (RRT), cause of end stage renal disease (ESRD), and transplant year. LD characteristics used in exploratory models included age, race (modeled as black vs. other), body mass index, body surface area (BSA), donor recipient BSA ratio, ABO incompatibility, HLA-A mismatches, HLA-B mismatches, HLA-DR mismatches, systolic blood pressure (SBP), diastolic blood pressure, eGFR as calculated by the CKD-EPI equation (15), biologic relatedness to the recipient, cigarette use, Epstein-Barr virus serostatus, and cytomegalovirus serostatus. Our definition of ABO incompatibility included A2 incompatible (A2i) donor/recipient pairs (A2 to B, A2 to O, A2B to B transplants, N=4); in a sensitivity analysis, reclassifying these pairs as compatible did not change inference. Continuous variables were modeled empirically using Martingale residual plots. The LD characteristics included in the final model were determined by selecting the model with the lowest Akaike information criterion (AIC). In order to model the risk of graft loss on a scale that included both LDKT and DDKT recipients, all LD characteristics were set to 0 for DDKT recipients and KDPI was set to 0 for LDKT recipients. We used an indicator variable in the model to identify a patient as a LDKT or DDKT recipient.

2.4. Calculation of P-LKDPI

We used the model of donor risk including LD characteristics, KDPI, indicator variable for LD versus DD, and recipient characteristics to calculate the P-LKDPI. In order to calculate the P-LKDPI so that the risk from LDKT for a given P-LKDPI score is equivalent to the risk from DDKT for the same KDPI score after adjusting for recipient characteristics, we divided the β coefficients (log hazard ratio) for each LD characteristic in the model by the β coefficient for KDPI and summed these values. We performed a Kaplan-Meier analysis with log-rank test to assess the risk of all-cause graft loss by strata of P-LKDPI, (<−60, −60 to −20, −20 to 20, >20), to determine whether higher P-LKDPI was associated with increased risk of graft loss.

2.5. P-LKDPI Validation

To determine whether the risk of graft loss for a given P-LKDPI in LDKT recipients was equivalent to the same KDPI in DDKT recipients, we used Kaplan-Meier survival analysis and log-rank tests to compare the cumulative incidence of all-cause graft loss between LDKT recipients and DDKT recipients within a given stratum of P-LKDPI/KDPI (1-20, 21-40, 41-60, 61-100). To minimize confounding by recipient characteristics, we used inverse probability weighting to balance recipient characteristics including recipient age at transplant, sex, race, cause of ESRD, peak PRA, years on RRT, and insurance status between LDKT and DDKT recipients within a given P-LKDPI/KDPI stratum. We compared the risk of all-cause graft loss of LDKT recipients to DDKT recipients within each P-LKDPI/KDPI stratum using Cox regression, adjusting for recipient characteristics, to further assess whether the risk of graft loss for LDKT and DDKT recipients in the same P-LKDPI/KDPI stratum was similar. We calculated Harrell’s C-statistic for P-LKDPI for LDKT recipients and for KDPI for DDKT recipients without recipient factors. 95% confidence intervals for the C-statistic were calculated with a bootstrap procedure repeated 200 times.

2.6. Temporal validation of the P-LKDPI

To assess predictive validity of the P-LKDPI, we performed temporal validation by calculating the P-LKDPI for first-time pediatric kidney-only transplant recipients in the US from 1999-2004 who had all donor information needed to calculate the P-LKDPI. We performed a Kaplan-Meier analysis with log-rank test to assess the risk of all-cause graft loss by strata of P-LKDPI to determine whether a higher P-LKDPI was associated with an increased risk of graft loss in this validation cohort.

2.7. Statistical analysis

All statistical analyses were conducted using Stata version 14 (Stata Corp., College Station, TX).

3. Results

3.1. Study population

Among the 7,155 pediatric KT recipients between 2005-2015, 2,649 (37%) underwent LDKT and 4,506 (63%) underwent DDKT. Compared to DDKT recipients, LDKT recipients were more likely to be younger (median [interquartile range; IQR] age 11 [4-15] years versus 13 [8-16] years, p<0.001), to be male (60.5% versus 57.4%, p=0.01), to be non-black race (90.4% versus 75.1%, p<0.001), to have private insurance (58.8% versus 30.3%, p<0.001), to have lower peak PRA (median [IQR] PRA 0 [0-3] versus 0 [0-5], p<0.001), to have fewer years on RRT (median [IQR] years 0.3 [0-1.1] versus 1.1 [0.3-2.1], p<0.001), and to have ESRD from congenital anomalies of the kidneys and urinary tract (34.2% versus 31.3%, p<0.001) (table 1).

Table 1:

Baseline characteristics of pediatric DDKT and LDKT recipients in the study population

	DDKT recipients n = 4,506	LDKT Recipients n = 2,649	p
Median (IQR) age	13 (8 - 16)	11 (4 - 15)	<0.001
Median (IQR) peak PRA	0 (0 - 5)	0 (0 - 3)	<0.001
Median (IQR) years RRT	1.1 (0.3 - 2.1)	0.3 (0 - 1.1)	<0.001
% Female	43%	40%	0.01
% Black	25%	10%	<0.001
% Private Insurance	30%	59%	<0.001
Cause of Renal Disease (%)			<0.001
CAKUT	31%	34%
FSGS	14%	11%
Other glomerular	18%	15%
Other	37%	40%

DDKT, deceased donor kidney transplantation; LDKT, living donor kidney transplantation; IQR, interquartile range; PRA, panel reactive antibody; RRT, renal replacement therapy; CAKUT, congenital anomalies of the kidneys and urinary tract; FSGS, focal segmental glomerulosclerosis

3.2. Model of donor risk

The LD characteristics included in the single model of DDKT and LDKT recipients, selected by AIC, were age among donors above age 50, female sex, number of HLA-B mismatches, number of HLA-DR mismatches, ABO incompatibility, eGFR, and SBP (Tables 2 and 3). Among DDKT recipients, increasing KDPI was associated with increased risk of graft loss (adjusted hazard ratio [aHR] per 10-unit increase in KDPI=1.06, 95% confidence interval [CI] 1.01-1.10, p=0.01). Among LDKT recipients, factors significantly associated with increased risk of graft loss included increased number of HLA-B mismatches (aHR per mismatch=1.27, 95% CI 1.04-1.55, p=0.02), increased number of HLA-DR mismatches (aHR per mismatch=1.23, 95% CI 1.01-1.47, p=0.04), the presence of ABO incompatibility (aHR=3.26, 95% CI 1.20-8.81, p=0.02), and increased SBP (aHR per 10mmHg increase=1.08, 95% CI 1.02-1.49, p=0.03). Although not significant, increased donor age above age 50 (aHR per 10 years=1.07, 95% CI 0.99-1.14, p=0.07) and female donor sex (aHR=1.19, 95% CI 0.97-1.46, p=0.10) remained in the model based on AIC. Conversely, increased eGFR was associated with decreased risk of graft loss in LDKT recipients (aHR per 10 unit increase=0.94, 95% CI 0.88-0.99, p=0.03).

Table 2:

Distribution of living donor characteristics included in the model of donor risk

Living Donor Characteristic
Median (IQR) age in years	37 (31-43)
% Female	58%
% HLA-B mismatch
0	14%
1	70%
2	16%
% HLA-DR mismatch
0	21%
1	66%
2	13%
% ABO incompatible	0.50%
Median (IQR) Systolic BP (mm Hg)	120 (110-128)
Median (IQR) eGFR (ml/min/1.73 m2)	105 (91-116)

BP, blood pressure; eGFR, estimated glomerular filtration rate

Table 3:

Donor characteristics associated with risk of all-cause graft loss in pediatric KT recipients, adjusted for recipient characteristics

Donor Characteristic	Adjusted Hazard Ratio (95% Confidence Interval)	p
DD: KDPI (per 10 units)	1.06 (1.01-1.10)	0.01
LD: Baseline	0.37 (0.11-1.21)	0.1
LD: Age per 10 years (past age 50)	1.07 (0.99-1.14)	0.07
LD: Female	1.19 (0.97-1.46)	0.1
LD: HLA-B mismatch (per mismatch)	1.27 (1.04-1.55)	0.02
LD: HLA-DR mismatch (per mismatch)	1.23 (1.02-1.49)	0.03
LD: ABO incompatible	3.26 (1.20-8.81)	0.02
LD: Systolic BP (per 10 mmHg)	1.09 (1.01-1.18)	0.03
LD: eGFR (per 10 units)	0.94 (0.88-0.99)	0.02

“Baseline” compares a LDKT recipient to a DDKT recipient of a kidney with KDPI = 0

KT, kidney transplant; DD, deceased donor; LD, living donor; KDPI, kidney donor profile index; BP, blood pressure; eGFR, estimated glomerular filtration rate

3.3. Calculation of P-LKDPI

Using the LD characteristics from the model of donor risk, we calculated the P-LKDPI.

P-LKDPI = −184.93 + 11.87*[(age-50) if age > 50] + 32.30 if female sex + 44.11*(number of HLA-B mismatches) + 38.83*(number of HLA-DR mismatches) + 218.63 if ABO incompatibility + 1.61*SBP – 1.24*eGFR. Online calculator available at http://transplantmodels.com/plkdpi/.

In this study population, the median LKDPI among LDKT recipients was −25 with an interquartile range of −56 to 12. The range of P-LKDPI was −200 to 290 (Figure 1). The median KDPI among DDKT recipients was 14. 68% of LD (N=1,801) had a P-LKDPI<0 indicating lower risk than any DD kidney. 25% of LD (N=652) had a P-LKDPI>14, indicating greater risk than the median DD kidney among pediatric DDKT recipients during the study period. 15% (N=393) of LD had a P-LKDPI>35 which is notable as pediatric candidates receive priority for DD kidneys with KDPI<35 under the current KAS (10). Moreover, 11% of LD (N=284) had a P-LKDPI>50, indicating greater risk than the median DD kidney among all DDKT recipients during the study period. Among LD with a P-LKDPI>50, median [IQR] age was 42 [34-50], median [IQR] systolic blood pressure was 125 [118-136], median [IQR] eGFR was 91 [82-101], 67% (N=190) were female, 74% (N=210) had two HLA-B mismatches, 66% (N=188) had two HLA-DR mismatches, and 4% (N=12) were ABO incompatible with their recipient.

Figure 1:

Distribution of P-LKDPI of transplanted living donor kidneys and KDPI of transplanted deceased donor kidneys in 7,155 pediatric kidney transplant recipients

3.4. P-LKDPI Validation

LDKT recipients of kidneys with higher P-LKDPI had an increased risk of all-cause graft loss (Figure 2). Patients in the highest stratum of P-LKDPI (P-LKDPI>20) had a greater incidence of graft loss compared to those in the lowest stratum of P-LKDPI (P-LKDPI<−60) (19% versus 10% at 5 years, 39% versus 19% at 10 years, log-rank p<0.001). Within strata of P-LKDPI/KDPI, there were no statistically significant differences in risk of graft loss between LDKT and DDKT after adjusting for recipient characteristics (Table 4). Likewise, in a matched analysis, there was no significant difference in cumulative incidence of graft loss in LDKT or DDKT recipients within the same stratum of P-LKDPI/KDPI (Figure 3). The C-statistic for the P-LKDPI was 0.57 (95% CI 0.54 to 0.60) as compared to the C-statistic for the KDPI which was 0.53 (95% CI 0.51 to 0.55).

Figure 2: Cumulative all-cause graft loss in pediatric LDKT recipients by strata of P-LKDPI.

Recipients with a higher P-LKDPI had a higher risk of all-cause graft loss.

Table 4:

Comparison of KDPI and P-LKDPI in pediatric KT recipients

KDPI/P-LKDPI Strata	Adjusted Hazard Ratio (95% Confidence Interval)	p
1-20	0.90 (0.65-1.24)	0.5
21-40	0.89 (0.59-1.32)	0.6
41-60	1.22 (0.77-1.95)	0.4
61-100	0.95 (0.52-1.74)	0.9

The hazard ratio is the hazard of all-cause graft loss associated with LDKT (vs. DDKT) in each stratum of KDPI or P-LKDPI, adjusted for recipient but not donor characteristics

Figure 3:

Cumulative all-cause graft loss in DDKT and LDKT recipients across strata of KDPI/P-LKDPI in recipient populations matched on age at transplant, sex, race, cause of ESRD, peak PRA, years on renal replacement therapy, and insurance status.

3.5. Temporal validation of the P-LKDPI

In a population of pediatric LDKT recipients not used to generate the P-LKDPI model, higher P-LKDPI was associated with higher risk of all-cause graft loss (Figure 4). Patients in the highest stratum of P-LKDPI (P-LKDPI>20) had a higher incidence of graft loss compared to those in the lowest stratum of P-LKDPI (P-LKDPI<−60) (21% versus 15% at 5 years, 39% versus 23% at 10 years, log-rank p<0.001). The C-statistic of the P-LKDPI in this validation set was 0.56 (95% CI, 0.53 to 0.58).

Figure 4: Cumulative all-cause graft loss by strata of P-LKDPI in validation set of pediatric LDKT recipients who received KT between 1995 and 2004.

Recipients with a higher P-LKDPI had a higher risk of all-cause graft loss.

4. Discussion

In this national study of pediatric kidney transplant recipients, we developed a risk index for pediatric LDKT recipients, the P-LKDPI, on the same scale as the DD KDPI that allows for comparison of multiple LD kidneys and for comparison of LD kidneys to DD kidneys. Strata of LDKT recipients of kidneys with higher P-LKDPI had a higher incidence of graft loss in a cohort of pediatric KT recipients used to generate the model and in a validation cohort of patients. Furthermore, there was no significant difference in risk of graft loss between LDKT recipients and DDKT recipients within the same P-LKDPI/KDPI stratum, indicating that similar P-LKDPI and KDPI scores are associated with comparable risk.

LDKT in children has been shown to be associated with longer graft survival compared to DDKT (16–18). Not surprisingly, the majority of LD (68%) had a P-LKDPI <0 indicating lower risk for graft loss than all DD kidneys. However, overlap exists between KDPI and P-LKDPI with a quarter of LD having a P-LKDPI above the median KDPI for pediatric DDKT recipients during the study period, indicating greater risk of graft loss than over half of DD kidneys used for pediatric DDKT. Furthermore, over ten percent of LD had a P-LKDPI above the median KDPI for all DDKT recipients. These findings challenge the common belief that a LD kidney is always superior to a DD kidney. Although the implications of rejecting a LD kidney for a DD kidney with a lower KDPI must be considered given the scarcity of DD kidneys, many pediatric patients will require repeated KT and thus it is important to optimize graft survival in this population.

Because higher P-LKDPI is associated with increased risk of graft loss, the P-LKDPI can be used to compare kidneys from multiple LD. Thus, the P-LKDPI could be useful for organ selection in pediatric patients who have multiple potential LD. Furthermore, because the P-LKDPI was developed on the same scale as the KDPI, it can be used to compare kidneys from LD to kidneys from DD. This may be useful when a pediatric patient receives a DD kidney offer but has a potential LD undergoing evaluation at the same time or when families are evaluating potential LD or DD kidney offers as part of KPD. Of note, our model does not consider the risks of remaining on dialysis after rejecting a LD kidney to wait for a DD kidney with a lower KDPI.

Although P-LKDPI was strongly associated with graft survival following LDKT, the C-statistic, a measure of the discriminatory ability of a model, was relatively low at 0.57. The C-statistic of the LKDPI generated in a cohort of adult LDKT recipients was similar at 0.59 (8). The low C-statistic is not unexpected. The C-statistic of a donor risk index will depend partly on accurate modeling of donor risk, but also on variability of donor quality and the extent to which graft loss is driven by donor quality versus other factors (19). A recent publication on the relationship between the C-statistic and the accuracy of program-specific evaluations demonstrated this by showing that the C-statistic is not associated with model accuracy and that models with low C-statistics can lead to effective risk stratification (20). Because LD undergo a rigorous screening process prior to kidney donation, most LD are healthy with few medical comorbidities (21); the C-statistic of a donor index would improve if unfit individuals were cleared to donate. Furthermore, post-transplant graft survival is largely determined by recipient characteristics, which cannot be predicted by a donor index. Nonetheless, we have demonstrated substantial differences in graft survival across strata of P-LKDPI; 10-year risk of graft loss for recipients from donors with P-LKDPI>20 is more than double the risk for recipients from donors with P-LKDPI<−60. The strong and graded association between P-LKDPI and graft loss makes it a useful tool for risk stratification despite the low c-statistic.

Notably, the DD KDPI, a tool currently used in clinical practice, had a lower C-statistic than P-LKDPI among pediatric LDKT recipients in this study. The KDPI was developed among adult DDKT recipients and the C-statistic of the KDPI for graft loss among adult DDKT recipients has been reported to be near 0.6 in previous studies (8, 22). The lower C-statistic of the KDPI for graft loss among pediatric DDKT recipients in this study suggests the need for pediatric-specific risk indices to assess organ quality for transplantation. The P-LKDPI was scaled to the DD KDPI to provide a clinically interpretable framework that would allow comparison of LD and DD kidney offers. It is important to note that the model used to create the P-LKDPI would rank LD candidates the same, even if we had not chosen to scale the risk score to the DD KDPI. Thus, LD with the greatest predicted risk would continue to have the greatest risk even if the P-LKDPI was scaled differently.

In the P-LKDPI, LD characteristics associated with increased risk of graft loss include HLA-B mismatches, HLA-DR mismatches, ABO incompatibility, and increasing systolic blood pressure. Conversely, higher eGFR among LD was associated with a decreased risk of graft loss. These findings are generally consistent with previous studies looking at the association between individual donor factors and graft survival in KT recipients. Similar to our finding of an association between HLA-mismatch and graft loss, two retrospective cohort studies, one of pediatric DDKT recipients using the International Collaborative Transplant Study database and one of pediatric LDKT recipients using the U.S. Renal Data System database, found a strong association between increasing number of HLA mismatches and increased risk of graft loss (23–24). Published literature on ABO incompatible kidney transplantation in pediatric KT recipients is limited to small, single-center case series that report similar graft survival in recipients of ABO incompatible kidneys and recipients of ABO compatible kidneys (25–26). However, these findings should be interpreted with caution given the small number of KT recipients and short follow-up time included in these studies. In a larger, registry-based study of both pediatric and adult KT recipients, the cumulative incidence of graft loss was significantly higher in recipients of ABO incompatible kidneys than recipients of ABO compatible kidneys at 1, 3, 5, and 10 years following transplant (27). Little is known about the association between donor blood pressure and eGFR and graft survival in pediatric KT recipients. However, consistent with our results, studies done in adult KT recipients have shown increased donor SBP to be associated with increased risk of graft loss and increased donor eGFR to be associated with decreased risk of graft loss (6, 8, 28).

Donor age and donor sex remained in our final model of donor risk despite having borderline significant associations with graft loss. Previous studies in pediatric KT recipients have shown increased donor age to be associated with graft loss. Analysis of data on pediatric DDKT recipients from the North American Pediatric Renal Trials and Collaborative Studies showed donor age above 50 to be associated with a 1.8-fold increased risk of graft loss compared to donor age 20 to 25 (29). Similarly, a retrospective analysis of pediatric LDKT recipients using the OPTN database showed that donor age above 55 was associated with a greater than 50% increased risk of graft loss compared to donor age 18 to 34 (30). The relatively weak association between donor age and risk of graft loss in our study may be explained by our adjustment for donor eGFR, as a causal association between donor age and recipient graft loss is at least partially mediated by donor renal function. The relationship between donor sex and graft survival is inconsistent in studies in adult KT recipients (5, 31–32). However, this association has not been evaluated in pediatric KT recipients and deserves attention based on the results of our study. Previous studies have suggested that the lower number of nephrons in female kidneys compared to male kidneys and the possibility that female kidneys express more HLA antigens than male kidneys, as shown in animal studies, may explain the association between donor female sex and worse graft survival (31, 33).

To our knowledge, this study is the first to construct a risk index based on donor characteristics for pediatric LDKT. Development of the P-LKDPI was strengthened by using a national cohort that included all pediatric kidney transplant recipients, which allowed for a relatively large sample size. However, several limitations should be considered. Data were obtained from an administrative database and thus we were only able to examine donor characteristics reported in the database. There may be donor characteristics that would improve prediction of recipient graft loss that we could not evaluate. Additionally, this database depends on quality control methods of individual transplant centers, which may lead to missing data, inaccurate data, and incompletely ascertained outcomes. Fortunately, these limitations should generally bias our results toward the null and we were still able to identify associations between multiple LD characteristics and graft survival. Although we cannot be sure that the P-LKDPI will be generalizable to future pediatric LDKT recipients, it is encouraging that it was associated with graft survival in an earlier cohort of patients not used to develop the model.

In conclusion, the P-LKDPI can be used to compare LD kidneys to each other and to DD kidneys in pediatric KT recipients. Thus, the P-LKDPI may help in the selection of a LD when multiple LD are available, may help to choose a DD kidney when a potential LD exists, and may help to evaluate offers through KPD.

Abbreviations:

aHR

adjusted hazard ratio

AIC

Akaike information criterion

BSA

body surface area

DD

deceased donor

DDKT

deceased donor kidney transplantation

eGFR

estimated glomerular filtration rate

ESRD

end stage renal disease

HRSA

Health Resources and Service Administration

KAS

Kidney Allocation System

KDPI

kidney donor profile index

KDRI

kidney donor risk index

KPD

kidney paired donation

KT

kidney transplantation

LD

living donor

LDKT

living donor kidney transplantation

OPTN

Organ Procurement and Transplantation Network

P-LKDPI

pediatric living kidney donor profile index

PRA

panel reactive antibody

RRT

renal replacement therapy

SBP

systolic blood pressure

SRTR

Scientific Registry of Transplant Recipients.