SEARCH

SEARCH BY CITATION

Keywords:

  • Adolescence;
  • graft survival;
  • HLA matching;
  • kidney allocation;
  • kidney graft function;
  • pediatric kidney transplantation;
  • waiting time

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

As HLA matching has been progressively de-emphasized in the American deceased donor (DD) kidney allocation algorithm, concerns have been raised that poor matching at first transplant may lead to greater sensitization and more difficulty finding an acceptable donor for a second transplant should the first transplant fail. We compared proportion of total observed lifetime with graft function after first transplant, and waiting times for a second transplant between individuals with different levels of HLA mismatch (MM) at first transplant. We studied patients recorded in the United States Renal Data System (1988–2009) who received a first DD transplant at age ≤21 years (n = 8433), and the subgroup who were listed for a second DD transplant following first graft failure (n = 2498). Compared with recipients of 2–3 MM first grafts, 4–6 MM graft recipients spent 12% less of their time and 0–1 MM recipients 15% more time with a functioning graft after the first transplant (both p < 0.0001); 4–6 MM recipients were significantly less likely (hazard ratio [HR] 0.87 [95% confidence interval 0.76, 0.98]; p = 0.03), and 0–1 MM recipients more likely (HR 1.26 [0.99, 1.60]; p = 0.06) to receive a second transplant after listing. The benefits of better HLA matching at first transplant on lifetime with graft function are significant, but relatively small.


Abbreviations
CI

confidence interval

DD

deceased donor

HR

hazard ratio

IPW

inverse probability weighting

IQR

interquartile range

LD

living donor

MM

mismatch

OPO

organ procurement organization

OPTN

Organ Procurement and Transplantation Network

PRA

panel reactive antibody

USRDS

United States Renal Data System

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

HLA matching is not only a well-known predictor of renal allograft survival [1-3], but may have effects extending beyond the life of the first graft. For example, poorer HLA matching at first transplant was associated with higher panel reactive antibodies (PRAs) at second transplant among adults and children [4, 5]. Increased sensitization leads to longer waiting times for a second graft and a higher risk of loss of the second graft [4, 6, 7]. As a result, sensitized patients tend to spend a greater proportion of their lives on dialysis—and therefore at substantially higher risk of death [8] and disability—than unsensitized patients. Poorer HLA matching was also recently shown to be significantly associated with a higher risk of death with a functioning graft [9].

Transplants with a high degree of mismatching have become increasingly prevalent over time [10, 11] in the United States as a result of sequential changes to allocation algorithms. Priority was eliminated for HLA A matching in 1995, and for HLA B matching in 2003 [12, 13]. In late 2005, Share35, the deceased donor (DD) kidney allocation policy for recipients <18 years old in the United States, was implemented in an effort to reduce excessive waiting times for children. Share35 gives children priority for kidneys from DDs under 35 years old, but effectively removes HLA matching from the allocation algorithm—with the important exception of 0 mismatch (MM), which still receives top priority. Share35 has succeeded in reducing waiting times for children [10, 11]. However, the long-term impact of poorer HLA matching is not known.

The impact of HLA matching on the ability to maintain graft function, and to obtain another graft in the event of a graft loss, may be more important in young recipients than older recipients. Pediatric and young adult kidney transplant recipients have a substantially higher risk of death-censored graft failure than other age groups; graft failure rates peak as young recipients traverse the interval between 17 and 24 years of age [14]. In addition, young recipients have a longer potential lifespan than older recipients simply because of their young age; they will require a functioning graft for many decades. These factors combine to greatly increase the chances that young recipients will require a second graft at some point in their lives. Therefore, it is important to consider the impact of HLA matching at the first transplant not only on survival of the first graft, but both waiting time for a second transplant and total lifetime with graft function.

We hypothesized that better HLA matching at first transplant would be associated with a greater proportion of total subsequent lifetime spent with graft function, and with a shorter waiting time for a second transplant. The first aim of the present study was to compare different levels of HLA MM at first DD renal transplant with respect to the proportion of the total observed lifetime spent with graft function after the first graft. The second aim was to compare waiting times for a second DD transplant by HLA MM at first transplant.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Data source and population

This was a retrospective cohort study of individuals recorded in the United States Renal Data System (USRDS) database who received a first renal transplant from a DD in the United States at ≤21 years of age between 1988 and 2009, and were followed until 2009. Analyses addressing the first aim included all recipients of a first DD transplant. We excluded 312 individuals with missing HLA information (3.6%). Analyses addressing the second aim included only those listed for a second DD transplant. Information on donor and recipient organ procurement organizations (OPOs) and Organ Procurement and Transplantation Network (OPTN) regions was obtained from the OPTN, after approval by the American Health Resources and Services Administration.

Statistical analysis

Primary exposure variable

HLA ABDR MM was the primary exposure for both aims, and was categorized as 0–1 MM, 2–3 MM or 4–6 MM. Analyses were also conducted using HLA DR MM as the exposure (0, 1 or 2 MM).

Matchability index

“Matchability,” defined as the probability of obtaining a graft with 0 or 1 HLA MM, was considered an important potential confounder in the associations between HLA MM at first transplant and both proportion of observed lifetime with graft function and waiting time for a second transplant; highly “matchable” patients may be more likely to get a well-matched first transplant, and to have shorter waiting times for a second transplant. Therefore, we developed national and OPTN region-specific “matchability indices” based on the empirical distributions of blood groups and HLA A, B and DR loci in all donors from 1988 to 2009 (n = 105 341) nationally, and in donors from each OPTN region in the same interval, respectively. An OPO-specific matchability index was not calculated because the relatively small numbers of donors in each OPO did not allow robust estimates of blood group and HLA frequencies. Generation of the matchability index is described in detail in the Supplemental Appendix. Briefly, we calculated the frequency in the national donor pool, and in the donor pool for each OPTN region, of blood groups and of each possible un-ordered quadruplet at the A and B loci and each possible un-ordered pair of alleles at the DR locus, taking antigen splits and equivalents into account [15]. A national matchability index, expressed as the probability of obtaining a 0 or 1 MM (out of 6) graft per 1000 donors nationwide [16], was calculated for each patient [16, 17]; a regional matchability index, relevant to the transplant OPTN region, was calculated in an analogous fashion. The United Network for Organ Sharing rules governing blood groups were respected. PRA was not included in the matchability index calculation because it was missing in 15% of patients; we adjusted for PRA separately.

Proportion of observed lifetime with graft function

The observation time of each patient was cut into 3-month intervals, and graft status was coded as functioning or not (failed, on dialysis) in each 3-month interval. Graft status was allowed to vary over time. All patients started observation with function of a DD graft. Graft status changed to not functioning after a graft failure, but returned to functioning after a second transplant. Patients were censored at the time of a third transplant; this censoring excluded ∼5% of all observed experience. We used logistic regression, with generalized estimating equations to adjust for clustering, to estimate the association between HLA ABDR MM at first transplant and proportion of total observed lifetime with graft function. Proportion of total observed lifetime with graft function captures both survival of the first and subsequent grafts and waiting times for subsequent transplants. The generalized estimating equation is a method for repeated measures, where the repeated measure is graft status in each 3-month interval; this approach takes into account the time elapsed since first transplant. The model was adjusted for potential confounders, including OPTN region-specific matchability index. Recipient OPO was not included as a potential confounder because there was no association between HLA MM at first transplant and OPO: the HLA MM distribution was almost identical across OPOs, with a median HLA MM in the 4–6 MM category in every OPO except one. Missing data were imputed using multiple imputation methods [18] based on the joint distributions of all other variables in the model. Multiple imputation is the recommended method of dealing with missing data, and avoids the serious bias that may result if cases with missing data are simply excluded [18]. Transplant era categories were based on changes in immunosuppression practices over time [19]. Socioeconomic status, estimated using median household income by zipcode, was classified by quartile within the U.S. Census data (1999) [20].

In order to minimize the potential for bias due to censoring of patients at death, we used inverse probability of censoring weighting (IPW) [21]. This method uses logistic regression to estimate the probability of being censored due to death within each time interval, then weights the person-time of patients who were not censored due to death by the inverse of their probability of survival (see Supplemental Appendix for details on IPW and the logistic regression models). Such weighting creates a pseudo-population without death censoring such that the weighted population is no longer a biased sample [21]. Models were also fit without IPW.

The same approach was taken for models comparing proportion of observed lifetime with graft function by DR MM; these were adjusted for the same covariates as the ABDR MM models, and for HLA MM at the A and B loci.

Waiting times comparison

We used Kaplan–Meier plots and robust multivariate Cox models to compare the waiting times for a second transplant between patients with different levels of HLA MM at the first transplant. Only patients listed for a second DD transplant were included. Waiting times were compared from the time of listing. Patients were censored at 10 years after listing, end of observation, living donor (LD) transplant or death. Observation was censored at 10 years after listing because hazards for death were nonproportional across HLA MM categories after ∼12 years; 90% of patients had received a second transplant, ended observation or had died by 10 years postlisting. The model was adjusted for potential confounders. Missing data were imputed, as described above. We used IPW in these models to minimize the potential for bias due to censoring of patients at death (see Supplemental Appendix for details). We also fit the models without IPW. The same approach was taken for models comparing time to second transplant after listing by DR MM; these were adjusted for the same covariates as the ABDR MM models, and for HLA MM at the A and B loci.

Data analyses were performed using Statistical Analysis Software 9.2 (SAS Institute, Cary, NC); a p-value <0.05 was considered statistically significant. The study was approved by the Research Ethics Board at the Montreal Children's Hospital.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Patient characteristics

Table 1 summarizes the characteristics of the two cohorts (first DD transplant recipients, and those listed for second transplant) by HLA MM at first transplant. The characteristics of those in each cohort with missing HLA data are also shown. The 8433 first DD graft recipients were followed for a median of 7.9 years (interquartile range [IQR] 3.5, 14.1), with a total of 75,288 person-years of observation. The 2498 patients listed for a second transplant were followed for a median of 1.9 years (IQR 0.6, 4.0) between listing and transplant or censoring. A disproportionate number of those with 0–1 and 2–3 MM at first transplant, compared with 4–6 MM, were White; as the degree of MM increased, the proportion of non-White recipients increased. More recipients with 4–6 MM were transplanted in the most recent era. In addition, a smaller proportion of those with 0–1 MM were in the lowest income quartile, compared with those with 2–3 and 4–6 MM. Recipients of 0–1 MM grafts had the highest and recipients with 4–6 MM had the lowest matchability indices. PRA at first transplant was not missing at random; more values were missing from individuals with 2–3 MM than from other HLA MM categories.

Table 1. Patient characteristics by HLA MM category
HLA MM at first transplantAll first transplant recipients (n = 8433)Patients listed for a second DD transplant (n = 2498)
0–1 MM2–3 MM4–6 MMMissing HLA0–1 MM2–3 MM4–6 MMMissing HLA
  • CAKUT, congenital anomalies of the kidneys or urinary tract; DD, deceased donor; FSGS, focal and segmental glomerulosclerosis; GN, glomerulonephritis; IQR, interquartile range; LD, living donor; MM, mismatch; PRA, panel reactive antibody.

  • Categorical variables are presented as proportions (%). Continuous variables are presented as medians with IQR.

  • 1

    Socioeconomic status (quartiles corresponded to median household income in the patient's zipcode of ≤$35,170 (lowest), $35,170–$42,171 (Mid-low), $42,171–$52,372 (mid-high), and ≥$52,372 (highest).

n52718616045312146698165465
Male (%)306 (58.1)1090 (58.6)3417 (56.5)161 (51.6)82 (56.2)405 (58.0)911 (55.1)27 (41.5)
Race (%)
White453 (86.0)1303 (70.0)3499 (57.9)211 (67.6)127 (87.0)450 (64.5)893 (54.0)44 (67.7)
Black47 (8.9)415 (22.3)1825 (30.2)56 (18.0)18 (11.0)211 (30.2)630 (38.1)15 (23.1)
Other27 (5.1)143 (7.7)721 (11.9)45 (14.4)3 (2.1)37 (5.3)131 (7.9)6 (9.2)
Age at first transplant (years)
0–9 (%)94 (17.8)393 (21.1)1505 (24.9)95 (30.5)24 (16.4)141 (20.2)372 (22.5)19 (29.2)
10–16 (%)172 (32.6)732 (39.3)2516 (41.6)124 (39.7)39 (26.7)259 (37.1)658 (39.8)28 (43.1)
17–21 (%)261 (49.5)736 (39.6)2024 (33.5)93 (29.8)83 (56.9)298 (42.7)824 (37.7)18 (27.7)
Median (IQR)16.9 (11.7, 19.4)15.6 (11.0, 18.4)15.1 (10.0, 17.7)14.3 (8.0, 17.3)17.8 (13.5, 19.7)15.9 (11.1, 18.8)15.6 (10.8, 18.1)14.4 (9.2, 17.2)
Era (%)
1988–1993161 (30.6)732 (39.3)1265 (20.9)66 (21.1)66 (45.2)394 (56.5)661 (40.0)28 (43.1)
1994–1998145 (27.5)460 (24.7)1118 (18.5)42 (13.5)51 (34.9)203 (29.1)466 (28.2)20 (30.8)
1999–2004142 (26.9)365 (19.6)1586 (26.2)95 (30.5)24 (16.4)85 (12.2)390 (23.6)15 (23.1)
2005–200979 (15.0)304 (16.3)2076 (34.3)109 (34.9)5 (3.40)16 (2.3)137 (8.3)2 (3.1)
LD source second transplant (%)37 (7.0)120 (6.5)235 (3.9)14 (4.5)21 (14.4)76 (10.9)165 (10.0)11 (16.9)
Socioeconomic status1 (%)
Lowest88 (16.7)411 (22.8)1447 (23.9)83 (26.6)24 (16.4)147 (21.1)380 (23.0)15 (23.1)
Mid-low107 (20.3)384 (20.6)1123 (18.6)45 (14.4)36 (24.7)146 (20.9)324 (19.6)12 (18.5)
Mid-high130 (24.7)429 (23.1)1447 (23.9)71 (22.8)34 (23.3)170 (24.4)405 (24.5)17 (26.2)
Highest168 (31.9)528 (28.4)1633 (27.0)67 (21.5)47 (32.2)208 (29.8)452 (27.3)19 (29.2)
Missing34 (6.4)109 (5.9)395 (6.5)46 (14.7)5 (3.4)27 (3.8)93 (5.6)2 (3.1)
Duration dialysis before first transplant (months)
Median (IQR)12.6 (4.2, 23.6)13.9 (5.4, 25.8)14.6 (5.3, 27.0)12.1 (0.0, 26.2)13.4 (4.9, 29.2)13.6 (5.8, 25.1)14.9 (6.5, 26.4)15.4 (3.8, 21.8)
National matchability index
Median (IQR)44.6 (14.8, 73.7)11.4 (2.6, 37.1)4.1 (0.8, 18.2)N/A40.8 (14.8, 64.9)10.0 (2.5, 37.5)3.3 (0.6, 15.5)N/A
Regional matchability index
Median (IQR)35.5 (11.5, 66.4)7.5 (1.2, 28.9)2.4 (0.2, 12.9)N/A30.1 (12.1, 56.2)6.2 (0.8, 29.0)1.7 (0.1, 11.7)N/A
PRA at first transplant (%)
0%346 (65.7)1161 (62.4)4161 (68.8)69 (22.1)87 (59.6)385 (55.2)963 (58.2)23 (35.4)
1–10%53 (10.1)220 (11.8)752 (12.4)14 (4.5)12 (8.2)80 (11.5)244 (14.8)7 (10.8)
11–50%29 (5.5)86 (4.6)233 (3.9)4 (1.3)6 (4.1)27 (3.9)65 (3.9)3 (4.6)
>50%9 (1.7)21 (1.1)63 (1.0)0 (0)1 (0.7)6 (0.9)13 (0.8)0 (0)
Missing90 (17.1)373 (20.0)836 (13.8)225 (72.1)40 (27.4)200 (28.7)369 (22.3)32 (49.2)
Primary disease (%)
CAKUT105 (19.9)404 (21.7)1515 (25.1)56 (18.0)28 (19.2)134 (19.2)325 (19.7)15 (23.1)
GN169 (32.1)490 (26.3)1383 (22.9)59 (18.9)49 (33.6)215 (30.8)476 (28.8)17 (26.2)
FSGS54 (10.3)226 (12.1)848 (14.0)30 (9.6)18 (12.3)106 (15.2)260 (15.7)6 (9.2)
Other103 (19.5)376 (20.2)1253 (20.7)78 (25.0)27 (18.5)135 (19.3)359 (21.7)18 (27.7)
Unknown48 (9.1)205 (11.0)618 (10.2)39 (12.5)17 (11.6)70 (10.3)161 (9.7)4 (6.2)
Missing48 (9.1)160 (8.6)428 (7.1)50 (16.0)7 (4.8)38 (5.4)73 (4.4)5 (7.7)
Blood group (%)
O253 (48.0)927 (49.8)3114 (51.5)138 (44.2)65 (44.5)358 (51.3)833 (50.4)38 (58.5)
A197 (37.4)662 (35.6)1898 (31.4)104 (33.3)52 (35.6)240 (34.4)540 (32.7)19 (29.2)
B64 (12.1)203 (10.9)786 (13.0)32 (10.3)25 (17.1)72 (10.3)217 (13.2)6 (9.2)
AB13 (2.5)69 (3.7)244 (4.0)7 (2.2)4 (2.7)28 (4.0)64 (3.8)2 (3.1)
Missing0 (0)0 (0)3 (0.05)31 (9.9)0 (0)0 (0)0 (0)0 (0)
Donor age
Median (IQR)25 (18, 40)22 (16, 36)21 (15, 31)18 (9, 28)29 (19, 43)24 (15, 40)23 (15, 39)16 (9, 30)
Missing (%)2 (0.4)3 (0.2)8 (0.1)4 (1.3)1 (0.7)1 (0.1)4 (0.2)1 (1.5)
Donor/recipient weight ratio
Median (IQR)1.4 (1.0, 2.0)1.6 (1.1, 2.6)1.6 (1.1, 2.7)1.7 (1.1, 2.8)1.2 (0.9, 1.6)1.5 (1.0, 2.4)1.6 (1.0, 2.5)1.6 (1.1, 2.3)
Missing (%)176 (33.4)797 (42.8)1453 (24.0)113 (36.2)71 (48.6)420 (60.0)718 (43.4)33 (50.8)

Proportion of observed lifetime with graft function

Figure 1 illustrates the proportion of patients who had a functioning graft in each 3-month interval (among those with observation time in that interval) following the first transplant, by HLA MM category. The area under each “curve” created by the histograms for each HLA MM category effectively represents the total proportion of all patient-years with graft function since the first transplant. Table 2 shows the results of unadjusted and adjusted models (without and with IPW) used to formally compare the proportion of the observed experience with graft function since the first transplant between HLA MM categories. An interaction between HLA MM and age at transplant was considered, but was not significant (p = 0.2). Compared with recipients of 2–3 MM first grafts, recipients of 4–6 MM grafts spent 12% less of their time with a functioning graft after the first transplant (p < 0.0001), and recipients of 0–1 MM grafts spent 15% more of their time with a functioning graft after the first transplant (p < 0.0001). Results were virtually identical when models were fitted without IPW, and when the national matchability index was used instead of regional. The hazard ratios (HRs) associated with the other covariates in the models are presented in Table 3.

image

Figure 1. Proportion of observed experience with graft function by HLA MM at first transplant. Each bar represents the proportion of the total observed patient experience in a 3-month interval during which there was graft function. Successive 3-month intervals since the first transplant are shown. Green bars represent those with 0–1 ABDR HLA mismatch (MM) at first transplant, blue bars those with 2–3 MM and red bars those with 4–6 MM.

Download figure to PowerPoint

Table 2. Relative likelihood of having graft function by HLA MM at first transplant
 HR (95% CI) associated with HLA ABDR MM
 0–1 MM2–3 MM4–6 MM
Unadjusted1.16 (0.91, 1.47); p = 0.2Reference0.79 (0.71, 0.89); p = 0.0001
Adjusted, no IPW1.16 (1.11, 1.20); p < 0.0001Reference0.88 (0.87, 0.90); p < 0.0001
Adjusted, with IPW1.15 (1.10, 1.20); p < 0.0001Reference0.88 (0.86, 0.90); p < 0.0001
 HR (95% CI) associated with HLA DR MM
 0 MM1 MM2 MM
  1. CI, confidence interval; HR, hazard ratio; IPW, inverse probability weighting; MM, mismatch; PRA, panel reactive antibody.

  2. Models were adjusted for sex, race, recipient age (including a quadratic age term), socioeconomic status, transplant era (1988–1993, 1994–1998, 1999–2004, 2005–2009), donor source (time-dependent), donor age, donor/recipient weight ratio, primary diagnosis, blood group, duration of dialysis before first transplant, regional matchability index, PRA and multiplicative PRA by age at transplant and PRA by age at transplant-squared (p = 0.6) interaction terms. The model for HLA DR MM was adjusted for HLA AB MM in addition to the other variables listed above. The lack of difference in results with and without IPW reflects the fact that death rates were similar between those with 0–1 HLA MM (1.5 deaths per 100 person-years), 2–3 MM (1.7 deaths per 100 person-years) and 4–6 MM (1.6 deaths per 100 person-years) at first transplant.

Unadjusted1.12 (0.98, 1.29); p = 0.1Reference0.84 (0.75, 0.95); p = 0.005
Adjusted, no IPW1.00 (0.98, 1.02); p = 0.9Reference0.84 (0.82, 0.85); p < 0.0001
Adjusted, with IPW0.99 (0.97, 1.02); p = 0.8Reference0.83 (0.82, 0.85); p < 0.0001
Table 3. Relative likelihood of having graft function associated with covariates
CovariateModel with inverse probability weighting HR (95% CI)
  • CAKUT, congenital anomalies of the kidneys or urinary tract; CI, confidence interval; DD, deceased donor; FSGS, focal and segmental glomerulosclerosis; HR, hazard ratio; PRA, panel reactive antibody.

  • 1

    The significant PRA at first transplant by age at transplant interaction indicates that the magnitude of the association between PRA at first transplant and proportion of subsequent lifetime with graft function decreased with increasing age at first transplant.

Age at first transplant (per 1-year increment)0.84 (0.83, 0.85); p < 0.0001
Age at first transplant-squared1.005 (1.005, 1.006); p < 0.0001
Male (vs. female)1.06 (1.05, 1.08); p < 0.0001
Race (vs. White)
Black0.47 (0.46, 0.48); p < 0.0001
Other1.22 (1.17, 1.26); p < 0.0001
Socioeconomic status (vs. lowest)
Mid-low0.98 (0.95, 1.01); p = 0.1
Mid-high1.07 (1.04, 1.09); p < 0.0001
Highest1.22 (1.19, 1.25); p < 0.0001
Era of transplant (vs. 1988–1993)
1994–19981.45 (1.40, 1.50); p < 0.0001
1999–20042.35 (2.27, 2.44); p < 0.0001
2005–20094.34 (4.12, 4.57); p < 0.0001
Living donor (vs. DD) for subsequent transplant1.46 (1.39, 1.54); p < 0.0001
Primary disease (vs. CAKUT)
Glomerulonephritis0.94 (0.92, 0.97); p < 0.0001
FSGS0.81 (0.78, 0.83); p < 0.0001
Other1.07 (1.04, 1.10); p < 0.0001
Unknown0.90 (0.87, 0.94); p < 0.0001
Blood group (vs. O)
A1.06 (1.04, 1.08); p < 0.0001
B1.07 (1.04, 1.10); p < 0.0001
AB1.37 (1.30, 1.43); p < 0.0001
Donor/recipient weight ratio (per 1 unit increment)1.04 (1.03, 1.05); p < 0.0001
Duration of dialysis before transplant (per 1-year increment)0.94 (0.93, 0.95); p < 0.0001
Donor age (per 10-year increment)0.992 (0.991, 0.993); p < 0.0001
PRA at first transplant (per 10 unit increment)0.81 (0.78, 0.85); p < 0.0001
PRA × age at transplant interaction11.001 (1.0003, 1.002); p = 0.005
Regional matchability index (per 10 units higher)1.04 (1.03, 1.05); p < 0.0001

We conducted a sensitivity analysis to determine whether changes over time in allocation policy—and therefore in waiting times—influenced our results. Grafts with poorer matching tend to fail sooner than grafts with good matching [2, 3]. Therefore, 4–6 MM grafts are more likely to fail, and be listed for a second transplant, in a year closer in time to the first transplant than grafts with better matching (HLA MM was negatively correlated with time to re-listing; p < 0.0001). If waiting times for second transplant increased over the observation period, then poor matching would be associated with shorter waiting times after second listing, resulting in underestimation of the magnitude of the effect of HLA MM at first transplant on proportion of lifetime with graft function. In contrast, if waiting times decreased over the observation interval, then poorer matching would be associated with longer waiting times after second listing, resulting in overestimation of the magnitude of the effect of HLA MM at first transplant on proportion of lifetime with graft function. Waiting times for a second transplant increased between 1988 (median 1.4 years [IQR 0.4, 3.6]) and 2001 (median 2.1 years [IQR 0.7, 4.0]), and decreased thereafter. Therefore, we repeated the analysis censoring all observation at the end of 2001 to ensure that shorter waiting times in the most recent years were not driving the association between HLA MM and proportion of waiting time with function. The results were unchanged (HR for 0–1 MM 1.18 [95% confidence interval (CI) 1.14, 1.23], p < 0.0001; HR for 4–6 MM 0.88 [0.86, 0.89], p < 0.0001).

The level of DR MM was also independently important. Compared with patients with 1 DR MM at first transplant, those with 2 DR MM had a significantly smaller proportion of their experience with graft function (HR 0.83 [0.82, 0.85]; p < 0.0001); there were no differences for those with 0 DR MM (HR 0.99 [0.97, 1.02]; p = 0.8).

HLA MM at first transplant and waiting time for second DD transplant

As illustrated in Figure 2, waiting times for a second transplant were progressively longer with higher degrees of HLA ABDR MM at first transplant. The results of the corresponding Cox models are presented in Table 4. Interactions between age at listing and HLA MM (p = 0.5) and between age at listing and PRA at first transplant (p = 0.4) were not significant. Compared with patients with 2–3 MM at the first transplant, those with 4–6 MM at the first transplant were significantly less likely to get a second transplant (HR 0.87 [0.76, 0.98]; p = 0.03) and therefore waited longer, whereas those with 0–1 MM were more likely to receive a second transplant (HR 1.26 [0.99, 1.60]; p = 0.06) and therefore had shorter waiting times—though this did not reach statistical significance. The HR associated with the other covariates in the models are presented in Table 5.

image

Figure 2. Time to second transplant from listing by HLA MM at first transplant. Kaplan–Meier curves illustrating the cumulative probability of receiving a second transplant with increasing time after listing, censored at 10 years. The solid line represents those with 0–1 ABDR HLA mismatch (MM) at first transplant, the dotted line those with 2–3 MM and the dashed line those with 4–6 MM. The number of patients still waiting for a second transplant at 0, 1, 2, 5 and 10 years after listing is indicated in the table below the plot.

Download figure to PowerPoint

Table 4. Relative “hazards” of retransplant among those listed for a second DD transplant
 HR (95% CI) associated with HLA ABDR MM
 0–1 MM2–3 MM4–6 MM
Unadjusted1.30 (1.03, 1.62); p = 0.02Reference0.77 (0.68, 0.86); p < 0.0001
Adjusted, no IPW1.15 (0.91, 1.46); p = 0.2Reference0.82 (0.72, 0.92); p = 0.001
Adjusted, with IPW1.26 (0.99, 1.60); p = 0.06Reference0.87 (0.76, 0.98); p = 0.03
 HR (95% CI) associated with HLA DR MM
 0 MM1 MM2 MM
  1. CI, confidence interval; DD, deceased donor; HR, hazard ratio; IPW, inverse probability weighting; MM, mismatch; PRA, panel reactive antibody.

  2. Models were adjusted for: age at listing (≤10 years, 11–17 years, ≥18 years), sex, race, socioeconomic status, era of listing (based on allocation policies), primary diagnosis, blood group, duration of dialysis before first transplant, PRA at first transplant and regional matchability index. The model for HLA DR MM was adjusted for HLA AB MM in addition to the other variables listed above. The slightly different results obtained in models without IPW likely reflect the higher death rate on the waiting list for those with 2–3 MM at first transplant (3.9 deaths per 100 person-years) than for those with 0–1 MM (1.6 deaths per 100 person-years) or 4–6 MM (3.2 deaths per 100 person-years). When observation was not censored at 10 years, results of unadjusted models and adjusted models without IPW were similar.

Unadjusted1.23 (1.07, 1.42); p = 0.003Reference0.89 (0.79, 1.01); p = 0.09
Adjusted, no IPW1.09 (0.95, 1.26); p = 0.2Reference0.86 (0.76, 0.98); p = 0.02
Adjusted, with IPW1.09 (0.94, 1.27); p = 0.2Reference0.98 (0.86, 1.11); p = 0.8
Table 5. Relative likelihood of re-transplant associated with covariates among those listed for a second DD transplant
CovariateModel with inverse probability weighting HR (95% CI)
  1. CAKUT, congenital anomalies of the kidneys or urinary tract; CI, confidence interval; DD, deceased donor; FSGS, focal and segmental glomerulosclerosis; HR, hazard ratio; PRA, panel reactive antibody.

Age at re-listing (vs. 11–17 years)
≤10 years1.95 (1.65, 2.30); p < 0.0001
≥18 years0.75 (0.66, 0.86); p < 0.0001
Male (vs. female)0.88 (0.78, 0.98); p = 0.02
Race (vs. White)
Black0.88 (0.77, 0.997); p = 0.04
Other0.85 (0.66, 1.09); p = 0.2
Socioeconomic status (vs. lowest)
Mid-low0.92 (0.76, 1.10); p = 0.3
Mid-high0.99 (0.83, 1.17); p = 0.9
Highest1.24 (1.06, 1.46); p = 0.007
Period of re-listing (vs. 1988–1994)
1995–20021.00 (0.86, 1.15); p = 0.9
2003–20050.63 (0.53, 0.76); p < 0.0001
2006–20090.53 (0.43, 0.65); p < 0.0001
Primary disease (vs. CAKUT)
Glomerulonephritis1.03 (0.88, 1.23); p = 0.7
FSGS1.09 (0.89, 1.33); p = 0.4
Other1.47 (1.24, 1.74); p < 0.0001
Unknown1.11 (0.89, 1.40); p = 0.4
Blood group (vs. O)
A1.50 (1.33, 1.70); p < 0.0001
B0.88 (0.72, 1.07); p = 0.2
AB1.98 (1.51, 2.61); p < 0.0001
Duration of dialysis before first transplant (per 1-year increment)0.91 (0.87, 0.94); p < 0.0001
PRA at first transplant (per 10 unit increment)0.95 (0.90, 1.02); p = 0.1
Regional matchability index (per 10 unit higher)1.10 (1.08, 1.13); p < 0.0001

There was no significant association between HLA DR MM at first transplant and waiting time for a second transplant after correction for censoring due to death. Compared with patients with 1 HLA DR MM at first transplant, the HR associated with 0 DR MM at first transplant was 1.09 [0.94, 1.27]; p = 0.2), and with 2 DR MM at first transplant was 0.98 [0.86, 1.11]; p = 0.8).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Physicians making decisions about the acceptability of kidneys offered for transplantation to young recipients must consider not only the immediate urgency for transplantation and the likelihood of prolonged function of the kidney being offered, but the potential impact of their decision on the patient's ability to receive a second transplant should the first graft eventually fail. These difficult decisions are made even more difficult by the lack of evidence indicating which factors should be taken into account in these decisions.

We have demonstrated that poorer overall HLA matching at the first transplant is associated with longer waiting times for a second transplant, and with a smaller proportion of the lifetime after first transplant being spent with graft function. HLA DR MM at the first transplant also predicted the proportion of the lifetime after first transplant with graft function, but not waiting time for second transplant, suggesting that the level of DR MM primarily affects graft survival.

These are potentially important findings in light of the progressive de-emphasis of HLA matching in the allocation of DD kidneys over time [9, 12, 13]. Pediatric recipients are at particularly high risk of receiving a graft mismatched at 5–6 alleles since the inception of Share35 [10, 11, 22]. Our findings suggest that poorer overall HLA matching at first transplant may result in children and young adults with end-stage renal disease spending a smaller proportion of their lives with graft function than they would have had with better matching.

However, conclusions about the net effect of this de-emphasis of HLA matching would be premature. Lesser emphasis on HLA matching has had important benefits, including a reduction in unacceptable disparities in access to organs between minority and majority ethnic groups [11, 23], and shorter waiting times for children [11]. The positive impact of greater HLA matching at first transplant on lifetime with graft function after the first transplant may be offset, or even overshadowed, by the longer waits for transplant that would likely accompany a greater emphasis on HLA matching. We could not assess the association between HLA MM at first transplant and proportion of lifetime with graft function since onset of end-stage renal disease because HLA MM at first transplant was partly determined by the waiting time for the first transplant. Such an analysis would be strongly biased toward an association between poorer HLA matching and a smaller proportion of lifetime with graft function. This is because allocation algorithms awarded increasing priority points for longer waiting times such that individuals with few or no priority points for HLA matching were only offered grafts provided they had waited long enough, whereas those with good HLA matching had enough points based on matching to be offered grafts, without needing many points based on waiting time.

It is important to recognize that while Share35 has resulted in shorter waiting times for both White and minority patients, waiting times remain significantly shorter for Whites [11]. Furthermore, whereas degree of HLA MM has not changed for minorities since inception of Share35, White patients have poorer HLA matching than in the past [11]. It may be possible to develop an alternative allocation strategy that gives some priority for HLA matching, but avoids long waiting times for those with uncommon HLA profiles by incorporating an index of “matchability.” This approach is used successfully by Eurotransplant [17, 24-27]. The Eurotransplant algorithm balances a preference for well-matched kidneys against the likelihood of obtaining a good match, thereby maintaining HLA matching as part of the algorithm without unfairly disadvantaging patients for whom it is difficult to find a good match. Although some US programs make an effort to do this locally, current allocation algorithms in the United States do not take matchability into account. Incorporating matchability may result in better HLA matching overall, and finally even out racial disparities in waiting times [28].

However, it must be recognized that increased emphasis on HLA matching for pediatric recipients would likely result in longer waiting times for some children [29]. At what point the disadvantages of longer waiting outweigh the advantages of better HLA matching is not clear. Furthermore, the tipping point may vary depending on the age of the recipient. As reflected in organ allocation policies around the world, most agree that prolonged periods of dialysis are particularly undesirable for children. This is because childhood is considered a “critical window” of vulnerability during which the adverse effects of end-stage renal disease—including impaired growth and neurodevelopment, and disruptions in education—may have lifelong effects [22, 30]. Late adolescence and early young adulthood may constitute another critical window of vulnerability, since this is a period of continued neurodevelopment, identity development, and education and vocational training [31]. Indeed, serious illness during this period may contribute to these patients' “failure to launch” as independent, contributing members of adult society.

Ours is not the first study to consider the impact of HLA MM at first transplant on outcomes other than survival of the first graft. Opelz and Dohler found associations between poorer HLA DR matching and posttransplant lymphoproliferative disorder in pediatric recipients [1] and between poorer overall HLA matching and death with graft function [9]. Meier-Kriesche et al [4] showed a higher risk of elevated PRA at listing for second transplant among adults with poorer overall HLA matching at first transplant. In contrast, Gritsch et al [30] found no association between HLA A, B or DR MM at first transplant and PRA at listing for second transplant in a group of 313 pediatric recipients in whom PRA was available at both initial transplant and listing for second transplant. However, this study had several important limitations. First, they only considered the effect of each of the A, B and DR alleles separately—not the effect of overall HLA MM. Second, the small sample size may have limited their ability to detect significant associations. Finally, the analysis excluded patients for whom PRA at first or second listing was missing; data that cannot be confirmed to be missing completely at random may introduce important biases [18]—and we have shown that PRA is not missing completely at random. Among American pediatric kidney recipients listed for a second transplant, Gralla et al showed both larger increases in PRA between first transplant and second listing, and longer waiting times for those with poorer matching [5]. However, this study was limited by exclusion of patients with missing PRA data, lack of standardization of PRA testing across centers and over time and inclusion of recipients of both DD and LD first transplants; the effects of HLA MM may be different in DD and LD. Failure to adjust for matchability may also have resulted in overestimation of the effects of HLA MM on waiting time for second transplant.

A strength of our study is the evaluation of outcomes of concrete importance to transplant recipients, rather that the intermediate outcome of PRA. However, this study has some limitations. The relatively short follow-up period meant that much of the association between HLA MM and proportion of time with graft function was driven by the well-known association between HLA MM and survival of the first graft. Figure 1 indicates that outcomes only begin to diverge between the 2–3 and 4–6 HLA MM categories after about 12 years, when failures are more prevalent. Furthermore, the waiting times analysis confirms that matching at first transplant does have an impact on events after first graft failure. The long interval of observation (1988–2009) introduces the possibility that changes in waiting times over time may have influenced our results. Our sensitivity analysis addressing this issue suggests that our estimates of the impact of HLA MM may represent underestimates of the true effect.

The inevitable trade-offs between waiting time and HLA MM must be carefully weighed [29]. Simulation studies are needed to understand how greater emphasis on HLA matching in allocation algorithms may influence the total lifetime with graft function from the onset of end-stage renal disease. We found a relatively small effect of HLA MM at first transplant on the proportion of observed lifetime after first transplant with graft function. The potential small benefit of greater HLA matching may not be justifiable during the critical developmental intervals of childhood and adolescence because longer waits for a first graft may be required if HLA matching were to be prioritized in allocation. Further studies are needed to evaluate this question.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Dr. Foster and Dr. Platt, members of the McGill University Health Centre Research Institute (supported in part by the Fonds de la Recherche en Santé du Québec [FRSQ]), received salary support from the FRSQ. This work was supported in part by Health Resources and Services Administration contract 234-2005-370011C. Data were supplied by the USRDS. The content is the sole responsibility of the authors alone and does not necessarily reflect the views or policies of the Department of Health and Human Services or the US government. This manuscript was not prepared in any part by a commercial organization. The work described in this manuscript was not funded in any part by a commercial organization.

Disclosure

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
ajt12643-sm-0001-SuppData-S1.doc76KAppendix: Calculating the Matchability Index.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.