Kidney Exchange Match Rates in a Large Multicenter Clearinghouse

Abstract

Kidney paired donation (KPD) can facilitate living donor transplantation for candidates with an incompatible donor, but requires waiting for a match while suffering the morbidity of dialysis. The balance between waiting for KPD versus desensitization or deceased donor transplantation relies on the ability to estimate KPD wait times. We studied donor/candidate pairs in the National Kidney Registry (NKR), a large multi-center KPD clearinghouse, between 10/2011-9/2015 using a competing risk framework. Among 1894 candidates, 52% were male, median age was 50 years, 66% were white, 59% had blood type O, 42% had PRA>80, and 50% obtained KPD through NKR. Median times to KPD ranged from 2 months for candidates with ABO-A and PRA 0, to over a year for candidates with ABO-O or PRA 98+. Candidates with PRA 80-97 and 98+ were 23% (95% CI: 6-37%) and 83% (78-87%) less likely to be matched than PRA 0 candidates. ABO-O candidates were 67% (61-73%) less likely to be matched than ABO-A candidates. Candidates with ABO-B or ABO-O donors were 31% (10-56%) and 118% (82-162%) more likely to match than those with ABO-A donors. Providers should counsel candidates about realistic, individualized expectations for KPD, especially in the context of their alternative treatment options.

INTRODUCTION

Kidney paired donation (KPD) can improve compatibility and facilitate transplantation between incompatible living donor/candidate pairs (1–5). However, there are other options for these candidates, including desensitization (6–9) as well as awaiting deceased donor kidney transplantation (DDKT). Although the Kidney Allocation System increased access to DDKT for highly sensitized candidates (10), waitlist mortality remains high, and some phenotypes do not benefit from the new allocation priorities (11). With these considerations in play, it is critical for providers and candidates to understand wait times for KPD so they can make informed treatment decisions.

Prior reports of wait times in small KPD programs suggested that their benefit may be limited to candidates with easy-to-match blood types and antibody profiles (12, 13). These findings were reported before the establishment of large multicenter KPD registries that have expanded rapidly in the past decade due to evidence that the transport of living donor kidneys is safe and feasible (14, 15). For example, the National Kidney Registry (NKR), the largest KPD registry in the US by volume, has facilitated over 2000 KPD matches since its inception in 2008, and continues to facilitate more matches each year (16, 17). Unfortunately, the only estimates of individual wait times in large KPD programs came from simulations performed by our group that suggested that almost all patients with panel-reactive antibody (PRA) <80% would match within a few months, but those with PRA >80 or blood type O would experience significantly longer wait times (18). However, these simulations were based on many assumptions, including an older paradigm of KPD (19). More recent reports suggest the landscape of living donor kidney transplantation has evolved differently than our simulations assumed: willingness to participate in KPD differs by race (20), and there is evidence of survival benefit of desensitization and incompatible transplantation as compared to remaining on the waitlist (8, 9). Thus, an up-to-date estimate of wait times in the current environment of KPD, based on observational data rather than simulation, is essential.

Using NKR data, we studied time to KPD for candidates and their paired donors, stratified by ABO blood type and PRA. To describe likelihood of KPD match, we modeled time to KPD in a competing risk framework accounting for alternative outcomes candidates faced while waiting in the registry, such as DDKT or death.

METHODS

Study Population

We studied kidney transplant candidate/potential donor pairs who registered in the National Kidney Registry (NKR) between October 2011 and September 2015. The registering potential donor (who entered the NKR with the candidate, but was incompatible with the candidate for direct donation) is referred to as a “paired donor” or “donor” for simplicity. Only 139 (7.3%) candidates registered with more than one potential donor. For those donors who underwent KPD within the NKR, the donor who participated in KPD was the donor listed. For other candidates, we listed the primary donor registered with the candidate. Recipients in a KPD chain who received a kidney facilitated by the chain without having their own incompatible donor (i.e. the last recipient in the chain who was drawn from the deceased donor waiting list and was never awaiting a match in KPD) were excluded. This study was approved by the Johns Hopkins Medicine Institutional Review Board.

Outcomes after Registration

Our primary outcome was KPD transplantation facilitated by NKR. We treated other removals from the NKR as competing events. Reasons for removal included: (1) identifying a compatible living donor outside of KPD, (2) transplantation via another KPD registry, (3) DDKT, (4) unavailability of incompatible paired donor (i.e. donor candidate no longer willing or able to donate), (5) candidate ineligibility (death or deteriorating condition) and (6) other/unspecified reason for removal.

Time to KPD

We defined time to KPD as the time from NKR registration to the recipient’s transplant date. Candidates who survived free of the defined outcomes were administratively censored on September 30, 2015 and were considered to be still waiting in the NKR. We used Kaplan-Meier curves to describe time to KPD within strata of candidate and donor ABO and candidate PRA categories. We used competing risk analysis as described by Fine and Gray to model likelihood of KPD transplant (21). This method allows modeling adjusted subhazard ratios within the subdistribution of the outcome of KPD transplant, with other outcomes considered competing events.

Statistical Analysis

Adjusted subhazard ratio (aSHR) confidence intervals are reported per the method of Louis and Zeger (22). We used a two-sided α of 0.05 to indicate a statistically significant difference. Statistical analysis was performed using Stata 14.2/MP for Linux (College Station, Texas). Figures were prepared using R 3.4 (the R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Study Population

During the study period, 1894 candidate/donor pairs registered in the NKR. The median (IQR) age of candidates was 50 years (39–60), 48% were female, and the median (IQR) body mass index (BMI) was 27 (23–31). The majority were Caucasian (66%) and 59% had blood type O. Candidate PRA ranged from 0%-100%; 42% had a PRA>80% and 79 (4.2%) reported PRA of 100%. Donors had a median (IQR) age of 45 years (35–53) and were predominantly female (62%) and ABO-A (47%, Table 1).

Table 1.

Characteristics of NKR registrants

	Candidate	Donor
Age at registration, med (IQR) years	50 (39-60)	45 (35-53)
Female	898 (48%)	1168 (62%)
BMI, med (IQR)	27 (23-31)	26 (24-29)
Race
Caucasian	1241 (66%)	1310 (70%)
Black	272 (14%)	232 (12%)
Asian	92 (5%)	78 (4%)
Latino	200 (11%)	185 (10%)
Other	79 (4%)	79 (4%)
ABO
A	450 (24%)	888 (47%)
B	278 (15%)	333 (18%)
AB	40 (2%)	98 (5%)
O	1126 (59%)	573 (30%)
PRA
0	524 (29%)	–
1-79	528 (29%)	–
80-97	323 (18%)	–
98+	432 (24%)	–

Outcomes after Registration

Among the 1894 candidates, 956 (50%) were eventually transplanted in an NKR-facilitated KPD, 102 (5.4%) were transplanted via another KPD registry, 178 (9.4%) underwent DDKT, 138 (7.3%) found a non-NKR mediated compatible living donor, 76 (4.0%) were removed due to donor ineligibility or unavailability, 129 (6.8%) were removed for other or unknown reasons, 90 (4.8%) died or were too sick for transplant, and 225 (12%) remained in the NKR at the end of the study (Figure 1). In all, 72.5% of the registered candidates ultimately received a kidney transplant during our study period.

Figure 1.

Candidate registry removal reasons

Of the 956 NKR-facilitated KPD matches, 25 (2.6%) were ABO incompatible. Of these, 8 were A2 to O, 1 A1 to O, 8 A to O without A subtyping available, 2 A2B to B, 1 AB to A, 1 AB to B without A subtyping available, 3 B to O, and 1 B to A.

Likelihood of NKR-facilitated KPD

Matching into a KPD transplant was the most frequent outcome among NKR registrants. At 3, 6, and 12 months, 32%, 38%, and 46% of candidates underwent NKR-facilitated KPD (Figure 2). More recent registration was also associated with a 17% increased chance of NKR KPD per year (aSHR: _1.101.17_1.24, p<0.001, Table 2). Each decade of candidate age was associated with a 5% increased chance of NKR KPD (_1.0021.05_1.10, p=0.04). Relative to candidates with a PRA of 0%, candidates with a PRA between 80-97% had a 23% lower chance of matching into an NKR KPD (_0.630.77_0.94, p=0.01), and candidates with PRA >97% had an 83% lower chance of NKR KPD (_0.130.17_0.22, p<0.001, Table 2). Candidates with ABO O had a 67% lower chance of matching relative to ABO A candidates (_0.270.33_0.39, p<0.001). Conversely, candidates whose donor had ABO blood type B or ABO O had a 31% and 118% higher chance of matching respectively (ABO B: _1.101.31_1.56, p<0.01; ABO O: _1.822.18_2.62, p<0.001, Table 2).

Figure 2. Stacked Plot of NKR Registrant Candidate Outcomes.

Other includes: other, unknown, and unavailable donor; Other LDKT includes other KPD and found compatible donor; Died includes too sick. The numbers in the graph represent the percentage of registrants who experienced that outcome at that time. For example, at 1 year, 46% of registrants were matched into an NKR KPD, and 9.2% received another LDKT, etc.

Table 2.

Factors associated with matching into a KPD through NKR

	Adjusted Subhazard Ratio	p-value
Year of registration		<0.001
Candidate age (per 10 years)		0.04
Candidate PRA
0	1 (ref)
1-79		0.08
80-97		0.01
98+		<0.001
Candidate ABO
A	1 (ref)
B		0.5
AB		0.6
O		<0.001
Donor ABO
A	1 (ref)
B		<0.01
AB		0.1
O		<0.001

Time to NKR-facilitated KPD

Median times to KPD ranged from 2 months for candidates with ABO type A and PRA 0 to over a year for candidates with blood type O or PRA 98+ (Figure 3). With respect to donor ABO type and candidate PRA, median times to KPD ranged from 2.5 months for candidates with PRA 0 and a donor with blood type AB, to over a year for candidates with PRA 98+ (Figure 4). With respect to candidate and donor ABO types, median times to KPD ranged from 2.3 months for candidates with blood type A and donors with blood type AB to over a year for candidates with blood type O (Figure 5).

Figure 3. Cumulative Incidence and Median Time to KPD in NKR by Candidate ABO Type and PRA.

Median times to KPD, in months, and number of KPDs are shown within each cumulative incidence curve pane.

Figure 4. Cumulative Incidence and Median Time to KPD in NKR by Paired Donor ABO Type and Candidate PRA.

Median times to KPD, in months, and number of KPDs are shown within each cumulative incidence curve pane.

Figure 5. Cumulative Incidence and Median Time to KPD in NKR by Candidate and Donor ABO Type.

Median times to KPD, in months, and number of KPDs are shown within each cumulative incidence curve pane.

Time to Removal from the NKR for DDKT or Other KPD Match

For the 178 (9.4%) candidates who left the registry due to DDKT, the median (IQR) time to removal from the registry was 10.3 months (4.3-20.0). For the 102 (5.4%) candidates who left the registry due to matching in another KPD program, the median (IQR) time to removal from the registry was 8.3 months (4.9-18.4).

DISCUSSION

In this study of almost 2000 candidate-donor pairs in the largest multicenter KPD clearinghouse in the United States, we found that 50% received an NKR-facilitated KPD transplant, 22% received a living donor or deceased donor kidney transplant outside of NKR, and 5% died waiting. Candidates with PRA 0 and blood type A experienced the shortest median wait time to KPD at 2 months, while candidates with blood type O, PRA 98+, and those with donors with blood type A all experienced median wait times of more than a year.

The general trends in wait times identified in this observational study are consistent with the trends described by our early simulation-based estimates: candidates with higher PRA and candidates with blood type O experienced longer wait times to KPD match (18). However, these new observational data suggest that KPD wait time is more varied than we previously simulated: more than a third of registered candidates left the NKR for alternate reasons that were not KPD. Prior simulations either could not account for these competing risks, or had to assume that non-KPD facilitated transplants were solely due to failure to match (18, 23). Furthermore, there is variability in living donor quality which might affect acceptance of matches in KPD programs (24), and these results provide insight into matches which were accepted and proceeded to transplantation. The benefit of these new real-world observational data over prior simulations is the ability to capture behavior, real compatibility testing, and the contribution of each of the competing risks experienced by candidates.

Some limitations of our study warrant further discussion. First, although our study analyzed the largest KPD registry in the US, KPD, in general, remains underutilized (1, 12, 25), and therefore these wait times and outcomes may not be applicable to future KPD opportunities if KPD expands substantially. As these data were reported by centers registering their candidates and paired donors as part of an administrative database, we do not have additional clinical granularity beyond the covariates reported, although the data available to us are considered highly reliable in that they were required to facilitate KPD matches for the patients; similarly, we were not limited by any missing values of those covariates we report, as registration to the NKR system cannot be completed until all fields are entered.

In conclusion, we have characterized the time to KPD in the largest multicenter KPD clearinghouse in the US. We have also identified donor and candidate characteristics associated with longer time to KPD, particularly candidate blood type O, PRA 98+ and donor blood type A. Finally, we explored the variety of other reasons that candidate-donor pairs left the registry. These observational data offer insight into the likelihood and timeline of KPD match, facilitating informed dialogue between providers and individual candidates about KPD and the candidate’s specific options, which might include incompatible transplantation.

Abbreviations

aSHR

adjusted subhazard ratio

BMI

body mass index

DDKT

deceased donor kidney transplant

IQR

interquartile range

KPD

kidney paired donation

NIDDK

National Institute of Diabetes and Digestive and Kidney Diseases

NKR

National Kidney Registry

PRA

panel reactive antibody