Delays in expanded criteria donor (ECD) kidney placement increases cold ischemia times (CIT) potentially leading to discard. The effect of increased CIT on ECD kidney transplant outcomes is unknown. We evaluated paired ECD kidneys (derived from the same donor transplanted to different recipients) from the SRTR registry transplanted between 1995 and 2009 (n = 17 514). To test the effect of CIT, we excluded paired transplants with the same CIT (n = 3286). Of 14 230 recipients (7115 donors) the median difference in CIT was 5 h (Q1 = 3 h, Q3 = 9 h). Delayed graft function (DGF) was significantly more likely between pairs with greater CIT (35% vs. 31%, p < 0.001) including substantially higher rates for CIT differences ≥15 h (42%). Overall graft loss was not significantly different between recipients with higher CIT relative to paired donor recipients with lower CIT (p = 0.47) or for pairs with differences of 1–3 h (p = 0.90), 4–9 h (p = 0.41), 10–14 h (p = 0.36) or ≥15 h (p = 0.10). Results were consistent in multivariable models adjusted for recipient factors. Although increasing cold ischemia time is a risk factor for DGF among ECD kidney transplants, there is no effect on graft survival which may suggest an important utility for donor kidneys that may not currently be considered viable.
As the disparity between kidney need and availability continues to widen, efforts to reduce the discard of deceased donor kidneys are critically important, particularly for expanded criteria donor (ECD) kidneys which are the type of kidneys most likely to be discarded. In 2009, 44% of ECD kidneys recovered in the United States were not transplanted compared to 10% of standard criteria donor (SCD) and 23% of donation after cardiac death (DCD) recoveries (1). High ECD kidney discard rates have not improved over time. An analysis of kidneys recovered between October 1999 and June 2005 demonstrated that more than 9000 recovered kidneys were determined to be unsuitable for transplantation, including 41% (5139 of 12 536) of ECDs (2). The proportion of discards in the United States appear high in contrast to that seen in the Eurotransplant Seniors Program, which aims to keep transport time short and preferentially offers kidneys from donors over age 65 to consented patients over age 65, resulting in a low discard rate of about 5% for these older donor kidneys (3).
Local discard rates of ECD kidneys vary from 14% to 60% depending upon where the kidneys were procured (2); suggesting that local practices in different donor service areas play an important role in whether these kidneys are transplanted or not (4). Placement of ECD kidneys to centers outside of the local donor service area offers the potential to reduce discard; however, placement of locally rejected kidneys is often delayed because of the time required to find an accepting center (5). This may lead to prolonged cold ischemia time (CIT), which may contribute to the ultimate discard of the kidney. Based on the response-to-injury hypothesis (6–8), early injury from cold ischemia may set the stage for indolent, yet chronically progressive, damage leading to higher rates of chronic graft loss. Kidneys from older donors have been thought to be even more susceptible to the negative impact of long cold ischemic time (9–12). Clinicians may be reluctant to accept ECD kidneys with already long CIT for fear of an additive deleterious effect in conjunction with CIT (13).
CIT has clearly been shown to be an independent risk factor for delayed graft function (DGF) (14–16) and DGF is clearly associated with inferior graft survival (14–18); however, the direct impact of CIT on graft survival is unclear (14–24). The etiology of DGF is multifactorial with some causes of DGF potentially more directly associated with outcomes (e.g. donor quality, technical failures) compared to other causes of DGF which may or may not have a direct causal relationship with long-term graft failures (e.g. CIT).
To determine the risks associated with transplantation of ECD kidneys with long CIT, we analyzed national registry data for patient and graft survival of adult transplant recipients of ECD kidneys pairs. An analysis of mate kidneys from the same donor is optimal to control for the predominant effects of donor quality while illustrating the effects of CIT. Additionally, we evaluated associations of CIT with the incidence of DGF and acute rejection using the same paired donor approach.
Materials and Methods
We utilized 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 United States, submitted by the members of the Organ Procurement and Transplantation Network (OPTN). The Health Resources and Services Administration (HRSA), US Department of Health and Human Services provides oversight to the activities of the OPTN and SRTR contractors.
Data submitted to the SRTR were accessed to identify all kidney alone transplant recipient pairs with a common ECD deceased donor in the United States between January, 1995 and October, 2009. Exclusions included recipients of kidneys from deceased donors whose contralateral kidneys were not transplanted (n = 3731) and paired transplants with the same CIT (less than 1 h difference; n = 3286).
A Cox regression model was fitted to compute covariate-adjusted patient loss, graft loss and death-censored graft loss hazard ratios (aHR). CIT and donor variables were not included in these models based on the paired kidney study design. The following recipient variables were included in these models: age (continuous), gender, race (African American, Hispanic, Caucasian, other), cause of end-stage renal disease (ESRD) (glomerulonephritis, diabetes, polycystic kidney disease, hypertension, other), retransplantation (yes/no), time on dialysis pre- and postplacement on the waiting list and prior to transplantation (categorized) (25), number of HLA-A, B and DR mismatches (0–6), panel-reactive antibody (PRA) level (0, 1–10,11–30, >30), body mass index (BMI), year of transplant (continuous), insurance status (private, Medicare, Medicaid, other), educational attainment (missing, high school, some college, college) and machine perfusion (yes, no, missing). The appropriate functional form of model covariates was determined by exploratory data analysis in unadjusted models and perceived impact on clinical meaningfulness.
Paired donors were categorized using a donor identifier available in the registry. Recipient outcomes were compared between paired donors according to (1) any CIT difference, and (2) CIT differences (delta CIT) of 1–3, 4–9, 10–14 and ≥15 h which were grouped based on the statistical distribution. ECD was defined as deceased donations deriving from donors with ages >60 years or >50 years with any two of the following donor criteria: (1) donor serum terminal creatinine >1.5 mg/dL, (2) hypertension or (3) death from cerebrovascular accident. All other donations were considered SCD. The definition of acute rejection was based on reported treatment for acute rejection as recorded in standard UNOS follow-up forms. DGF was defined as the reported need for dialysis within 1 week post-transplantation. Transplants were classified as nonlocal, if the donor and recipient transplant center were not in the same organ procurement organization (OPO).
The odds of DGF and acute rejection were assessed using a multivariable generalized estimating equation adjusted for recipient factors with paired donors considered a repeated measure in the models. The dependence of observations derived from the kidneys from the same donor was accounted for in Cox models with adjustment of the standard error of the hazard ratio (sandwich estimator). Overall kidney graft survival (defined as time to graft loss or death) plots were generated from Kaplan–Meier models. Relevant characteristics of the donor, the recipient and the graft between patient groups were compared using McNemar's test for categorical variables and paired t-tests for continuous variables based on paired observations.
We also investigated the overall graft outcomes in three strata of the population: (1) among ECD donors that were also DCDs, (2) by delta-CIT stratified by the CIT of the first (shortest CIT) transplant to investigate whether the effect of additional hours was similar depending on the extent of CIT of the initial transplant and (3) by absolute CIT time differences (0–10, 11–20, 21–30 > 30 h) with the earliest as the reference group.
All analyses were performed using SAS software, version 9.2 (SAS Institute, Inc., Cary, NC, USA). Statistical significance was identified by a p-value of less than 0.05 and all confidence intervals also used a 95% threshold. This study was approved by the Cleveland Clinic Institutional Review Board.
The study population consisted of 14 230 ECD kidney transplant recipient pairs reported to the SRTR between January, 1995 and October, 2009. The median difference in CIT was 5 h (25th percentile = 3 h, 75th percentile = 9 h). Donor characteristics of the study population included 10.0% African American and 51.9% female; 85.6% had a cerebrovascular cause of death, 66.8% had a history of hypertension and 10.6% had a history of diabetes.
Compared to donors that were excluded due to equivalent CIT, there were no statistically significant differences among any of these characteristics. The DGF incidence in the paired kidneys with the same CIT time was similar (32%), to the lower CIT group (31%) and significantly lower than the higher CIT group (35%). Overall graft survival (8 year = 40%) was slightly lower compared to both the lower (42%) and higher (41%) CIT group and acute rejection at 1 year was lower in the same CIT group (14%) compared to the lower and higher CIT groups which were both 18%.
The relevant characteristics of recipients are presented in Table 1. The shorter CIT group comprised significantly higher proportions of pre-emptive recipients and a lower proportion of kidneys from nonlocal donors and fewer 0 HLA mismatched transplants compared to the long CIT group; the two groups were similar in terms of the other characteristics evaluated.
Table 1. Recipient characteristics by cold ischemia time (CIT) groups
Shorter CIT group (n = 7115)
Longer CIT group (n = 7115)
CIT 1–3 h (n = 2018)
CIT 4–9 h (n = 3345)
CIT 10–14 h (n = 938)
CIT 15+ h (n = 814)
1Testing the difference between the long- and short-CIT groups (McNemar's test for paired observations).
2Testing for differences across CIT group levels.
BMI = body mass index; CIT = cold ischemia time; PRA = panel reactive antibodies; HLA = human leukocyte antigens; OPO = organ procurement organization.
Age (mean ± sd)
55.7 ± 12.3
55.6 ± 12.2
55.6 ± 12.6
55.6 ± 12.1
55.8 ± 12.3
55.0 ± 11.8
Primary diagnosis of diabetes
BMI > 30 kg/m2
PRA > 30%
CIT in hours (mean ± sd)
16.2 ± 7.5
23.3 ± 9.1
19.2 ± 7.3
22.2 ± 7.5
27.0 ± 7.8
34.0 ± 10.6
HLA mismatches > 3
0 antigen HLA mismatch
Kidney machine perfusion
Delayed graft function
DGF was statistically significantly more likely between pairs with longer CIT. As displayed in Figure 1, recipients receiving the paired donor with longer CIT experienced DGF 35% of the time compared to 31% among recipients of the shorter CIT (p-value < 0.001). This relationship also demonstrated a graded, dose-dependent effect, utilizing the more detailed grouping levels wherein DGF increased from 31% for the lowest CIT (reference) group to 42% for the longest CIT (≥15 h delta CIT) group. These results were consistent in the multivariable generalized estimating equations using the two and five level CIT groupings (Table 2).
Table 2. Adjusted likelihood of delayed graft function by the difference in cold ischemia time between paired donors
Parameter (reference = short CIT group)
Adjusted odds ratio
95% Confidence limits
Pr > |Z|
Generalized estimating equation was adjusted for age (continuous), gender, race (African-American, Hispanic, Caucasian, other), cause of end-stage renal disease (ESRD) (glomerulonephritis, diabetes, polycystic kidney disease, hypertension, other), retransplantation (yes/no), time on dialysis pre- and postplacement on the waiting list and prior to transplantation (categorized) (25), number of HLA-A, B and DR mismatches (0–6), panel-reactive antibody (PRA) level (0, 1–10,11–30, >30), body mass index, year of transplant (continuous) and insurance status (private, Medicare, Medicaid, other), educational attainment (missing, high school, some college, college) and machine perfusion (yes, no, missing) with donors considered a random effect.
CIT 1–3 h longer
CIT 4–9 h longer
CIT 10–14 h longer
CIT 15+ h longer
Graft and patient survival
Overall graft loss rates were not significantly different between recipients with higher CIT relative to the paired donor recipients with lower CIT (Figure 2A, p = 0.47). Moreover there was no association with graft survival between pairs with delta-CITs of greater than 1–3 h (Figure 2B, p = 0.90), 4–9 h (Figure 2C, p = 0.41), 10–14 h (Figure 2D, p = 0.36) or ≥ 15 h (Figure 2E, p = 0.10). Results were consistent in multivariable models adjusted for recipient factors (Table 3). Overall patient death rates were not significantly different using the two and five level CIT groupings (Table 3).
Table 3. Cox proportional hazard models for overall graft loss and patient death by CIT groups
Overall graft loss
Study group (reference level)
95% Confidence interval
Study group (reference level)
95% Confidence interval
All models adjusted for age (continuous), gender, race (African-American, Hispanic, Caucasian, other), cause of end-stage renal disease (ESRD) (glomerulonephritis, diabetes, polycystic kidney disease, hypertension, other), retransplantation (yes/no), time on dialysis pre- and post-placement on the waiting list and prior to transplantation (categorized), number of HLA-A, B and DR mismatches (0–6), panel-reactive antibody (PRA) level (0, 1–10,11–30, >30), body mass index, year of transplant (continuous), insurance status (private, Medicare, Medicaid, other), educational attainment (missing, high school, some college, college) and machine perfusion (yes, no, missing). Standard errors adjusted for dependence of donor pairs using sandwich estimator.
Longer CIT group (shorter CIT group)
Longer CIT group (shorter CIT group)
1–3 h longer CIT group (shorter CIT group)
1–3 h longer CIT group (shorter CIT group)
4–9 h longer CIT group (shorter CIT group)
4–9 h longer CIT group (shorter CIT group)
10–14 h longer CIT group (shorter CIT group)
10–14 h longer CIT group (shorter CIT group)
15+ h longer CIT group (shorter CIT group)
15+ h longer CIT group (shorter CIT group)
In the subset of ECD transplants that were also classified as DCD (n = 498), there was no statistically significant difference in time to overall graft loss associated with the kidney with longer CIT (AHR = 1.17, 95% confidence interval [CI] 0.84–1.63).
We examined the outcome of overall graft survival based on delta-CIT stratified by the CIT of the first (shortest CIT) transplant. Among patients in whom the first transplant occurred with less than the median CIT (16 h), there was no significant difference in the risk adjusted time to overall graft loss (AHR = 0.95, p = 0.20, mean CITs = 10.6 and 18.5 h respectively). Among transplants with 16–26 h of CIT (26 h representing the 90th percentile, n = 5798), there was an increased hazard for overall graft loss associated with the longer CIT time but this was not statistically significant (AHR = 1.08, p = 0.07, mean CITs = 10.6 and 18.5 h, respectively). Similarly, among transplants in which CIT was in the highest 10th percentile (>26 h) and 5th percentile (>30 h), there was also no significant difference in time to overall graft loss for the longer CIT kidney (AHR = 1.04, p = 0.62, n = 1422, mean CITs = 31.2 and 37.3 h respectively) and (AHR = 1.11, p = 0.45, n = 700, mean CITs = 35.1 and 41.4 h, respectively), respectively.
In order to potentially identify a threshold level of absolute CIT at which graft outcome does diminish, we further evaluated graft outcomes of paired kidney transplants performed with a CIT 0–10 h relative to mate kidneys transplanted with CITs of 11–20, 21–30 and > 30 h. Compared to the 0–10 group, the overall survival was significantly better for the 11–20 group (AHR 0.80, 95% CI 0.68–0.94; 5-year graft survival (GS) 58%[0–10 h] vs. 66%[11–20h], n = 886 each arm) and not significantly different for the 21–30 h group (AHR 1.05, 95% CI 0.79–1.40; 5-year GS 59%[0–10 h] vs. 58%[21–30h], n = 272 each arm). The sample size was not sufficient for a multivariate analysis of the 0–10 h versus > 30 h group; however, the 5-year GS was not worse in the group with the longer CIT (60%[0–10 h] vs. 66%[>30 h], n = 66 in each arm). Death censored graft loss was not significantly different for the 0–10 h compared to the 11–20 h group (AHR 0.87, 95% CI 0.70–1.09; 5-year GS 75% vs. 79%, n = 886) or the 21–30 h group (AHR 1.02, 95% CI 0.67–1.55; 5-year GS 75% vs. 74%, n = 272). Five-year graft survival was 75% in the 0–10 group and 74% in the >30 h group (p = 0.57, n = 66). There were no differences in the recipient characteristics depicted in Table 1 between the four groups except for a higher proportion of patients with PRA > 30% in the 0–10 CIT group compared to the 11–20 CIT group and a higher proportion of recipients with PRA >30% and HLA mismatch > 3 in the 21–30 group relative to the 0–10 group.
The proportion of patients that were treated for acute rejection within 1 year by the CIT group is displayed in Figure 3. Based both by the two group levels of shorter and longer CIT and by the more detailed groups, there was no statistically significant difference in the proportion of acute rejection. In the multivariable generalized estimating equations these results were consistent, neither the two group levels (p-value = 0.80) nor the more detailed levels (p-value = 0.54) demonstrated a significant association with acute rejection.
Center level effects
To evaluate for center-level effects we included centers as an additional random effect in the outcomes models for DGF and AR. In both cases, there was evidence of center-level variation, however, the point estimates and statistical significance of the CIT group remained consistent. In the case of the graft survival, adjustment for center-level use of the short versus long CIT time revealed no evidence that centers with a higher proportion of long CITs were associated with worse survival (AHR = 1.00, 1.00–1.01, per 10% more long CIT vs. short CIT, p = 0.54), and most relevant to this study there was no significant effect on the estimate for the effect of CIT.
One of the concerns of the current allocation system for deceased donor kidneys is that increased CIT, incurred while attempting to place ECD donor kidneys, may have a deleterious impact on transplant outcome and thereby inhibit acceptance rates. Using SRTR data of paired ECD kidney transplants between 1995 and 2009, we found that although DGF was significantly higher with increasing durations of CIT, graft and patient survival was similar regardless of the difference in CIT between the paired kidneys and there was no association of CIT with acute rejection.
Previous large multivariable analyses have generally reported worse graft survival associated with increasing CIT. The Collaborative Transplant Study found greater risks of graft failure with CITs > 18 h (18) and an analysis of US Renal Data System found a continuous worsening of outcome associated with each 6-h increase in ischemia (14). A major limitation of these studies is the likely inability of multivariate models to accurately adjust for important donor quality confounders which may not always be codified in standard registries, a problem that is ameliorated with the use of a paired kidney analysis. Additionally, baseline renal functionality is a strong predictor of long-term renal function (26), yet currently measured variables such as serum creatinine and estimated glomerular filtration rate may not fully account for baseline renal functionality.
Studies specifically comparing outcomes of mate kidneys from the same donor transplanted sequentially to different recipients have generally been small or single center reports and have reported conflicting results. Giessing et al. (19) found no difference in graft survival among 80 mate kidneys with average CITs of 8.3 and 14.3 h in each group. Kyllonen et al. (20) reported significantly better 1-year graft survival in the group with short average CITs (22 vs. 28 h) from a cohort of 816 paired kidney transplants. Giblin et al. (21) reported significantly worse 1-, 5- and 10-year graft survival in the longer CIT group in a cohort of 520 paired kidneys transplanted on average with 19.9 and 25.7 h of CIT at a single center. However, this analysis was not adjusted for important recipient factors such as race and BMI.
In previous paired analyses of nonmandatory and mandatorily shared kidneys the longer CIT due to sharing has not been shown to impair graft survival (22–24). Lim et al. (17) noted that recipients of either shipped or locally transplanted well-matched renal allografts have similar graft and patient survivals. We previously reported overall similar overall graft survival for 441 regional and 311 national nonmandatory share kidneys relative to locally transplanted mates (23). Stegall et al. (24) found no significant difference in the survival of payback kidneys compared with locally transplanted mismatched kidneys.
This paired analysis of ECD kidney recipients demonstrates that despite a strong association of CIT with DGF, graft survival was similar regardless of CIT differences up to 15 h over the CIT of the first transplanted mate kidney. We also found a lack of decrement in graft survival for CIT strata up to 30 h relative to those transplanted at 0–10 h. We hypothesize that the DGF caused by prolonged CIT is due to acute tubular necrosis which is largely sublethal damage and recoverable (27); and therefore does not impact long-term function. These findings are similar to those seen in graft outcomes in kidney transplantation of DCD kidneys from standard criteria donors. In spite of the terminal warm ischemia inherent in the DCD donor organ recovery process, resulting in high DGF rates, it has been shown that short- and medium-term graft survival rates are similar between donation after brain death standard criteria donors and DCD donor kidney grafts, suggesting that ischemic injury is likely to be a reversible lesion (28).
The national sharing system in the United States is an important method for ECD kidney placement. Most discarded kidneys are not rejected by one but by many transplant centers. Our finding that prolonged CIT is not associated with graft survival among kidney pairs has important implications for transplant centers considering utilization of ECD kidneys with anticipated prolonged CITs. In particular, kidneys currently discarded due to CIT still provide a significant benefit to patients. The perception that these kidneys are too high risk which may not be fully supported by empirical evidence, may have led to discard, while results of this study could potentially highlight the utility of these organs.
Our analysis demonstrated a significant increase in DGF for kidneys with longer CITs than their mates. Incremental increases in DGF were also seen with increasing duration of CIT from the reference kidney. Three other kidney transplant paired analyses also found an association between CIT and DGF (15,16,20), whereas one study found DGF rates did not differ between mate kidneys transplanted with average CITs of 19.9 h versus 25.6 h (21). In that study, DGF was low in both groups (3.8% vs. 7.3%, respectively) and the low overall DGF was attributed to the average young donor age (35.4 years). In a subgroup analysis of Eurotransplant Seniors Program recipients, Giessing et al. (19) found no impact of the 5-h longer CIT on DGF; the lack of difference may be due to the small sample size (n = 20) or to the extremely short CITs in both groups due to local allocation. Although the findings of our study suggest that CIT-induced DGF is not an independent risk factor for graft failure in ECD kidney transplants, other consequences of DGF noted in previous studies are substantial including prolonged hospitalization, higher cost of transplantation, increased complexity of management of immunosuppressive drugs and an adverse effect on the rehabilitation potential of transplant recipients (29–32).
Acute rejection was not associated with CIT in our study. These findings are in accordance with the findings of some other studies (16,19), but not in others (18,21). Giblin et al. (21) found acute rejection rates to be significantly lower for first kidneys compared to second kidneys transplanted (22.3% vs. 28.1%, p < 0.01) in an unadjusted paired analysis at a single center. Opelz et al. (18) demonstrated a 20% increase in rejection risk on nonpaired multivariate analysis only when kidneys were preserved for more than 36 h and suggested that prolonged cold storage results in increased allograft immunogenicity. Others suggest a correlation of CIT with graft fibrosis (33,34). Kim et al. (33) studied the effects of transient ischemia in mice and demonstrated increased production of reactive oxygen species resulting in fibrosis. Kuypers et al. (34) reported that DGF solely due to ischemia (in the absence of rejection) correlates strongly with the presence of fibrosis in the renal allograft at 3 months posttransplant.
Our results are subject to the limitations inherent in observational data. Because recipients are often not randomly selected to receive kidneys, it is possible that they are in some unmeasured way systemically less (or more) healthy than recipients of kidneys with long CIT kidneys. A recipient selection bias may exist in which transplant centers that are utilizing a kidney with long CIT may select a ‘healthier’ recipient than what would be selected for a kidney with shorter CIT. For example, the differences in recipient characteristics suggest a recipient selection bias that may explain the poorer overall graft survivals in the 0–10 CIT group compared to the 11–20 h group. This explanation also seems likely, since death censored graft survival was not significantly different. There is the possibility for residual confounding as a result of recipient- or center-related factors not included in the OPTN data. Our analyses included many but not all of the factors that may confer risks at or after transplantation such implantation technique, anastomosis time, recipient anatomy, immunosuppression type and dosing, recipient anatomic abnormalities, preservation solution type utilized during machine perfusion and length of machine perfusion. The paired kidney analysis allows for the adjustment for most donor factors, but it is not possible to capture anatomical abnormalities in one of the kidneys (multiple vessels, extent of atherosclerosis, injured vasculature) that could lead to technical difficulties, independent of recipient factors. Additionally, there may be combinations of donor and/or recipient characteristics or interactions between factors that influence outcomes but are not captured in our multivariate models. Lastly, there may be a CIT threshold at which graft outcomes begins to deteriorate that would not be detected in analysis due to the paucity of cases with CITs beyond 30 h. Potential issues relating to the determination of acute rejection include missing or incomplete data, reporting bias, sampling and technique errors, measures of quantification and subjective interpretation.
There has been extensive focus in the field of transplantation on recovery and placement of all possible donor organs. Although increasing CIT is a risk factor for DGF in our analysis, there is no effect of CIT on graft survival suggesting that DGF due to CIT may not be associated with diminished long-term outcomes of ECD kidneys. These data may suggest important opportunities to increase transplant rates of previously discarded organs.
The data reported here have been supplied by the Minneapolis Medical Research Foundation as the contractor for the Scientific Registry of Transplant Recipients (SRTR). The interpretation and reporting of these data are the responsibility of the author(s) and in no way should be seen as an official policy of or interpretation by the SRTR or the US Government.
The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.