Introduction

Racial and ethnic disparities in pediatric kidney transplantation are well described. Black children are less likely to receive a transplant (¹), are less likely to be on the transplant waiting list (²), and have overall longer wait times to transplantation (³) compared with White children. Lower transplant access is linked to a 64% higher mortality rate for Black compared with White children, regardless of income (¹).

On December 4, 2014, the Kidney Allocation System (KAS) was implemented to reduce racial and ethnic disparities and increase access for sensitized candidates (⁴). Specific changes include (1) preregistration dialysis time added to calculated wait time, (2) prioritization for sensitized candidates using a sliding scale for calculated panel reactive antibodies (cPRAs), and (3) prioritization of deceased donor kidneys with a Kidney Donor Profile Index (KDPI) <35% for pediatric candidates (⁴). Preliminary evidence indicated that the new KAS improved transplant access (^5678–9), but racial and ethnic disparities persisted for highly sensitized candidates (¹⁰). There was also concern that longer ischemia times from the mandatory sharing rules would lead to higher rates of delayed graft function (⁶,⁸,¹¹). Many of these studies had minimal follow-up time after the KAS policy change and focused mostly on adult populations.

Accordingly, this study explores how the 2014 KAS change affected access to transplant and transplant-related outcomes in children across racial and ethnic groups with 5 years of follow-up time to evaluate the current policy landscape and guide future improvements in equity. We hypothesized that the 2014 KAS decreased time to transplantation across all racial and ethnic groups but was associated with worse short-term transplant outcomes.

Materials and Methods

Data Sources and Study Population

This study used data from the Scientific Registry of Transplant Recipients (SRTR). The SRTR data system includes data on all donors, waitlisted candidates, and transplant recipients in the United States, submitted by the members of the Organ Procurement and Transplantation Network (OPTN) under the administration of the Health Resources and Services Administration, US Department of Health and Human Services (¹²).

We included all kidney transplant candidates in the SRTR database who were activated on the transplant list at <18 years of age between January 1, 2008 and December 31, 2019. Those who were listed for multiorgan transplant, received a prior organ transplant, or received a living donor transplant were excluded. To eliminate outliers who were on dialysis for an extended period of time, we excluded candidates who were initiated on dialysis prior to January 1, 2003 (¹) (Figures 1 and 2). This study was approved by the institutional review board of Stanford University.

Exposures, Outcomes, and Covariates

The exposures of interest were the 2014 KAS policy change and race and ethnicity. We defined the pre-2014 KAS era as up to December 3, 2014 and the post-2014 KAS era as on or after December 4, 2014 (Figures 1 and 2). Racial and ethnic groups were consolidated into the following mutually exclusive categories: non-Hispanic White (White), Black (included patients of Black race and any ethnicity), Hispanic (not White), and other (which includes other races not included in the previous categories and any ethnicity). Outcomes of interest were time from activation to deceased donor transplantation, time from dialysis initiation to deceased donor transplantation, occurrence of delayed graft function defined as the need for dialysis 1 week after transplantation, and time to graft loss.

Recipient demographic characteristics examined were age at listing, payer, blood type, dialysis days, underlying diagnosis, and cPRA. Donor and transplant characteristics examined were KDPI (manually calculated for all recipients and generated using the 2015 scaling factor) and cold ischemic time.

Statistical Analyses

All recipient, donor, and transplant characteristics were reported as numbers and percentages for categorical variables and medians and interquartile ranges for continuous variables. Kaplan–Meier curves were generated to show time from activation to transplantation, time from dialysis initiation to transplantation, and time to graft loss after transplantation, stratified by racial and ethnic group in both eras.

Parametric accelerated failure time (AFT) models were used to examine how KAS affected time to transplantation for different racial and ethnic groups. We chose AFT models over the traditional Cox proportional hazards models as the proportional hazards assumption was not met for the transplantation outcome. AFT models generate time ratios, which express covariates’ effect on differences in survival (e.g., a time ratio of 1.3 means a 30% longer survival time per one unit higher value of the covariate). AFTs require a specified distribution to fit the underlying hazard of the model. Goodness of fit was determined by a visual examination of Cox–Snell residuals compared with the cumulative hazard of the sample (¹³). We determined that the log-logistic distribution was an appropriate choice for every model (Supplemental Figure 1).

We ran two sets of AFT models with two different time variables. The first modeled time from initial activation on the transplant list to deceased donor transplant, where candidates temporarily transitioned out of the model if they were deactivated and transitioned back in upon reactivation. The second modeled time from initiation of dialysis until transplantation. We tested each model for an interaction between race and ethnicity and era, and a likelihood ratio test determined significance. To assess if racial disparities changed in each era, we ran regressions for each era with race and ethnicity as the exposure of interest. Then, to assess the effect of the 2014 KAS, we ran regressions for each racial and ethnic group with KAS as the exposure of interest (¹⁴). In addition to crude analyses, we adjusted for patient- and transplant-related confounders, donor service area, and year of activation or dialysis start year using multivariable analyses. Candidates were censored by death (98 candidates), end of the data collection period (1178 candidates), or removal from the list (666 candidates). Because AFT models estimate a cause-specific hazard function, they are well suited for the consideration of competing risks (e.g., death, removal from waiting list) (¹⁵). Because of the effect that KAS could have on waiting list time with the addition of back-calculated preregistration dialysis, candidates were restricted to stay within their own era of activation or initiation of dialysis (Figure 2).

We performed a secondary analysis on those who were transplanted to explore racial and ethnic disparities in short-term transplant outcomes. We used logistic regression to determine disparities in odds of delayed graft function, where univariable and multivariable models were run separately for the two eras. We then performed log-rank tests to compare time to graft loss after transplant in each era. Data analysis was conducted using Stata 14 (¹⁶). The amount of missing data was 7% or less for all variables (Table 1). Donor service area was missing for 1% of the cohort. Multivariable analyses were restricted to those with complete data for all variables.

Results

Patient Characteristics

Among the 7496 children studied, 2848 (38%) were White, 1738 (23%) were Black, 2347 (31%) were Hispanic, and 563 (8%) were in the other racial and ethnic category. Other children of color had the lowest proportion of children transplanted (72% versus 75%–79%). Black and Hispanic children were more likely to have public insurance compared with White children or other children of color. Black and Hispanic children had the longest time on dialysis (648 days for Black children and 636 days for Hispanic children versus 537 days for White children and 586 days for other children). Black (3%) or other (5%) children of color were most likely to have a cPRA>80% compared with White (1%) or Hispanic (2%) children. Black children who were transplanted had the highest proportion of delayed graft function (9% versus 6%–7%) (Table 1). Demographic data and transplant characteristics where children were divided on the basis of activation era have similar differences among racial and ethnic groups (Supplemental Table 1). Characteristics of those with complete data were very similar to those with missing data (Supplemental Table 2) and, so, were generalizable to the remainder of the cohort.

Time from Activation to Deceased Donor Transplantation

In both eras, Black and Hispanic children and other children of color had longer median times from activation to deceased donor transplant compared with White children. Overall, median times to transplant were substantially longer post-KAS (Figure 3, Table 2). Black and Hispanic children and other children of color had lower transplant rates per 1000 person-years compared with White children, and these rates were overall lower in the post-KAS era compared with the pre-KAS era (Table 2).

In univariable AFT models exploring associations of race and ethnicity within each era, Black and Hispanic children and other children of color had significantly longer wait times compared with White children in both eras. In the multivariable analysis, these findings were largely attenuated. Black children had a 14% longer wait time (time ratio, 1.14; 95% confidence interval [95% CI], 1.00 to 1.29) pre-KAS and a 16% longer wait time (time ratio, 1.16; 95% CI, 1.01 to 1.32) post-KAS. Hispanic children had a 12% longer wait time (time ratio, 1.12; 95% CI, 0.99 to 1.28) pre-KAS and a 13% longer wait time (time ratio, 1.13; 95% CI, 1.00 to 1.28) post-KAS. Other children of color had an 8% longer wait time (time ratio, 1.08; 95% CI, 0.87 to 1.33) pre-KAS and a 17% longer wait time post-KAS (time ratio, 1.17; 95% CI, 0.96 to 1.41) (Table 2).

In multivariable AFT models exploring associations of era within each group, every racial and ethnic group had significantly longer wait times post-KAS compared with pre-KAS. White children waited 169% longer (time ratio, 2.69; 95% CI, 2.15 to 3.36), Black children waited 197% longer (time ratio, 2.97; 95% CI, 2.27 to 3.91), Hispanic children waited 229% longer (time ratio, 3.29; 95% CI, 2.56 to 4.22), and other children of color waited 173% longer (time ratio, 2.36; 95% CI, 1.63 to 4.56) (Table 2).

Time on Dialysis to Deceased Donor Transplantation

In both eras, Black and Hispanic children and other children of color had longer times on dialysis to transplant compared with White children. Overall, median times were shorter in the post-KAS era (Figure 3, Table 3). Black and Hispanic children and other children of color had lower rates of transplantation per 1000 person-years compared with White children. Rates were overall higher in the post-KAS era (Table 3).

In univariable AFT models exploring associations of race and ethnicity within each era, Black and Hispanic children and other children of color had significantly longer times on dialysis compared with White children in both eras. In the pre-KAS era in multivariable models, these findings are attenuated but still showed differences (time ratio, 1.20; 95% CI, 1.09 to 1.32 for Black children; time ratio, 1.13; 95% CI, 1.02 to 1.24 for Hispanic children; time ratio, 1.17; 95% CI, 1.01 to 1.36 for other children). In the post-KAS era, there were no significant differences in time on dialysis for Black and Hispanic children and other children of color compared with White children in multivariable models (Table 3).

In multivariable AFT models exploring associations of era within each group, Black and Hispanic children experienced longer times on dialysis post-KAS compared with pre-KAS (time ratio, 1.31; 95% CI, 1.08 to 1.59 for Black children; time ratio, 1.27; 95% CI, 1.07 to 1.52 for Hispanic children) (Table 3).

Delayed Graft Function and Graft Loss

In the pre-KAS era, Black children had higher odds of delayed graft function compared with White children (odds ratio, 1.46; 95% CI, 1.08 to 1.96) in unadjusted analysis. This finding was no longer significant in multivariable analysis (odds ratio, 1.17; 95% CI, 0.85 to 1.60). There were no significant associations between race and ethnicity and delayed graft function in the post-KAS era in univariable or multivariable analyses (Table 4). Black and Hispanic children had longer time to graft loss post-KAS compared with pre-KAS (log-rank P=0.03 and P=0.01, respectively) (Figure 4).

Discussion

This retrospective cohort study of pediatric kidney transplant recipients is the first to assess how the 2014 KAS policy affected racial and ethnic disparities in deceased donor transplantation with 5 years of follow-up. After the 2014 KAS policy change, racial and ethnic disparities in time from activation and time on dialysis to deceased donor transplant began to exist for Black and Hispanic children and other children of color compared with White children, but this finding is largely explained by differences in patient demographic– and transplant-related factors as these findings are attenuated in multivariable analyses. We also found improvements in odds of delayed graft function for Black children and longer time to graft loss for Black and Hispanic children post-KAS.

Preliminary studies were optimistic that the 2014 policy change improved disparities. A study of transplant rates 1 year after the change showed a 21% higher rate of transplants for Black patients and a 9% higher rate of transplants for Hispanic patients (⁹). Another study using a difference-in-difference analysis showed narrowing of gaps in transplant rates among patients of color compared with White patients 20 months after the policy change (⁷). For children specifically, a quasiexperimental time series study in children 15 months after the KAS policy change estimated a 46% higher rate of transplants for Black children (⁵). Our study similarly shows that 5 years after the policy change, racial and ethnic disparities are not exacerbated as a result of the 2014 KAS.

Prior studies have shown that for adults with cPRA>80%, racial and ethnic disparities still exist post-KAS (¹⁰). Our results show that disparities in access to transplantation for children in univariable analysis are explained by patient factors as opposed to race and ethnicity itself. Because race and ethnicity are social constructs, as opposed to biologic differences among individuals, future work should focus on the effect of KAS on patient-specific barriers for children of color, such as cPRA, geography, or blood type, to help improve these disparities. Additionally, as studies have described lower acceptance rates for pediatric patients post-KAS (¹⁷), how organ offer and acceptance rates differ by racial and ethnic group post-KAS should be explored.

This study indicates that the 2014 KAS may be decreasing the availability of transplants for children overall. A study 15 months after the KAS policy change showed that children 0–6 years of age had a 21% lower likelihood of transplant (⁵). Another study of children and adults 1 year after the policy change reported that pediatric transplant rates were higher compared with adults but decreased after the policy change—a finding that leveled off after a bolus effect (⁸). Decreasing access to transplantation for children overall will further disparities for already vulnerable children.

Previous studies can help explain why children have longer wait times post-KAS. One reason is that KDPI does not accurately assess pediatric donors as no donor below age 6 years old is assigned a KDPI<35 (¹⁸). As donors with a KDPI<35 are prioritized to children, acceptable pediatric donors are preferentially being offered to adult recipients. One study found that children received 34% fewer pediatric donors, with a substantial increase in >35-year-old donors (¹¹). Another study 3 years post-KAS found that children were 79% less likely to receive offers of donors ages <18 and 18–34 years old (¹⁷). Changes in acceptance patterns could be contributing as kidneys from donors 18–34 years old and KDPI<35% were 23% less likely to be accepted post-KAS, whereas donors 35+ years old with KDPI<35% were three-fold more likely to be accepted (¹⁷). Finally, prioritization of ever increasing numbers of adult multiorgan transplants over pediatric kidney-only recipients may be contributing (¹⁹).

Adding preregistration dialysis time to candidate wait time improved racial and ethnic disparities for those on dialysis prior to transplant as there is now equity in time on dialysis to transplantation post-KAS in multivariable analysis. This is especially important as previous studies describe longer times on dialysis to transplantation for Black or Hispanic children, resulting in higher mortality (¹,²⁰). Our results are contrary to studies that found longer time on dialysis prior to transplantation post-KAS for children (²¹) and adults (⁷). These studies only followed patients up to 2 years; therefore, the results were likely due to an initial bolus effect where preregistration dialysis time was added to transplant waiting time post-KAS, making wait times appear longer.

Initial studies of the 2014 KAS policy change were concerning for higher rates of delayed graft function post-KAS (⁸). A study of the OPTN data up to 2016 found a 69% higher odds in delayed graft function in recipients <10 years of age post-KAS compared with pre-KAS explained by age or size of donors and pretransplant dialysis duration (¹¹). Separately, another study reported higher likelihood of delayed graft function post-KAS for all adults but more so for those of Black or Hispanic race and ethnicity (⁶). Our study finds that with prolonged follow-up time, delayed graft function did not differ among racial and ethnic pediatric groups. We also found that time to graft loss improved for Black and Hispanic children after KAS.

There are important limitations of this study. Although this is a comprehensive dataset of US pediatric transplants, sample sizes were low for some racial and ethnic groups. Our data are limited to 5 years after policy change, so longer-term transplant outcomes cannot be evaluated. We used insurance status as a proxy of socioeconomic status as we lacked more granular information of income. Finally, as only patients on dialysis subsequently listed for transplantation were included in this study, our findings are not generalizable to the whole kidney failure population.

Assessing the 2014 KAS policy change does not address racial and ethnic disparities that occur throughout the transplant listing process. The pediatric nephrology community needs to remain diligent in considering all aspects of kidney failure care, steps to transplantation, and consequences of systemic racism and provider biases in this process to truly address racial and ethnic disparities: from early identification of CKD to access to subspecialty care, activation, and preemptive and living related donors (²²). A study found that with adherence being equal, referral to transplantation varied by race and ethnicity (²³). How racism independently affects outcomes as a social determinant of health should be the focus of future studies (²⁴). Additionally, removal of donor service areas and integration of proximity points to transplant centers were implemented on March 15, 2021. It is too early to comment on how this policy change will affect racial and ethnic disparities, but this deserves attention in future studies.

In conclusion, wait times for children from activation to deceased donor transplantation were equitable before and after implementation of KAS, whereas racial and ethnic disparities for time on dialysis to transplantation and odds of short-term graft loss improved in equity after KAS, after adjusting for patient- and transplant-related factors. These improvements occurred without causing disparities in delayed graft function. Wait times overall for children have increased since the policy change, requiring discussion among the pediatric nephrology community and policy makers to implement further changes to prevent vulnerable children from having increased barriers to pediatric transplantation.

Disclosures

X.S. Cheng reports receiving honoraria from ClarityCo and Medscape Education and receiving research funding from the American Heart Association and the National Institutes of Health. P.C. Grimm reports consultancy agreements with Eloxx and Horizon; receiving research funding from Alexion Pharmaceuticals; receiving honoraria from Eloxx, Horizon, and Recordati; and serving as a board member of the Improve Renal Outcomes Collaborative. J.C. Tan reports serving as an associate editor of Clinical Transplantation, serving as a scientific advisor or member of American Journal of Kidney Diseases, and serving on the American Society of Transplantation Scientific Review Board. J.C. Tan's spouse is employed by and has ownership interest in GNE/Roche. All remaining authors have nothing to disclose.

Funding

The John M. Sobrato Gift Fund was used for purchase of the SRTR database. J.R. Krissberg was a Stanford Maternal and Child Health Research Institute Tashia and John Morgridge Endowed Postdoctoral fellow at the time this study was conducted. M.B. Kaufmann is supported by Agency for Healthcare Research and Quality grant T32HS026128.