Association Between Duration of Delayed Graft Function, Acute Rejection, and Allograft Outcome After Deceased Donor Kidney Transplantation

Abstract

Background 

Prolonged duration of delayed graft function (DGF) may be associated with adverse allograft outcomes, but the association between threshold duration of DGF, acute rejection and long-term allograft loss remains undefined. We aimed to determine the impact of DGF duration on allograft outcomes and to assess whether this association was mediated by acute rejection.

Methods 

Using data from the Australian and New Zealand Dialysis and Transplant Registry, Cox proportional modeling was used to determine the association between quartiles of DGF duration, acute rejection at 6 months and death-censored graft loss (DCGL). Mediation analysis was conducted to determine whether acute rejection was a causal intermediate between DGF and DCGL.

Results 

Of 7668 deceased donor kidney transplants between 1997 and 2014, 1497 (19.5%) recipients experienced DGF requiring dialysis. The median (interquartile range) duration of DGF was 7 (9) days, with 25% requiring dialysis for 14 days or longer. Among recipients who had experienced a DGF duration of 1 to 4 days, the adjusted hazard ratio for duration of 5 to 7, 8 to 13, and 14 days or longer were 1.13 (95% confidence interval [CI], 0.83-1.55; P = 0.43), 1.44 (95% CI, 1.08-1.91; P = 0.013), and 1.99 (95% CI, 1.50-2.65; P < 0.001), respectively, for acute rejection; and were 1.10 (95% CI< 0.73-1.67; P = 0.64), 1.45 (95% CI, 1.00-2.11; P = 0.05) and 1.60 (95% CI, 1.10-2.31; P = 0.01), respectively, for DCGL. On average, 8% of the effects between DGF duration and DCGL were explained by acute rejection.

Conclusions 

There was a direct dose-dependent effect between DGF duration and DCGL, with acute rejection explaining less than 10% of the effects between DGF duration and DCGL. Future research identifying other potential modifiable mediators that lies in the causal pathway between DGF duration and allograft loss is essential.

In Brief

This study shows that there is a direct dose-dependent effect between the duration of delayed graft function and death-censored graft loss (DCGL). Surprisingly acute rejection explains <10% of the impact of delayed graft function duration and DCGL.

The incidence of delayed graft function (DGF) after kidney transplantation has increased substantially over time, after the increased utilization of kidneys from marginal donors (higher Kidney Donor Profile Index) and donation after circulatory death (DCD).¹ Between 2000 and 2009, the incidences of DGF exceeded over 20% for brain death donors and 50% for DCD transplants in Australia.² Similar rates of DGF have been reported in the United States and the United Kingdom.^3,4

Delayed graft function may have an unfavorable impact on allograft outcomes, such as premature allograft loss, attributed to excess inflammatory response, increased hypoxic injury and damage to endothelial and tubular cells triggered by ischemic-reperfusion injury.^5-8 In a meta-analysis of 33 studies totaling 151 594 kidney transplant recipients, DGF was associated with over a 40% greater risk of allograft loss.⁴ Although an association between DGF and early graft loss was found, uncertainties exist as to the threshold duration of DGF associated with adverse allograft outcomes.^9,10 In addition, prior studies have reported higher incidence rates of acute rejection during DGF. The reasons for the observed increased risk may be multifactorial, with indirect activation of the innate and adaptive immunity through ischemic-reperfusion injury likely a key driver for acute rejection and subsequent graft dysfunction.^11-15 This study aimed to examine the associations between DGF status, DGF duration and allograft outcomes; and to determine if the relationship between DGF duration and allograft loss was mediated by acute rejection in deceased donor kidney transplant recipients.

MATERIALS AND METHODS

Study Population

Patients with end-stage kidney disease (ESKD) who received a first deceased donor kidney transplant and experienced DGF between 1997 and 2014 in Australia and New Zealand were included. Patients who had received live donor kidney transplants, multiple-organ grafts, and those who had received prior grafts were excluded. In Australia and New Zealand, deceased donor kidneys are typically allocated to ABO-compatible transplant candidates with negative complement-dependent cytotoxicity (CDC) T-cell crossmatch.

Exposure Factor

Delayed graft function was defined as requiring any dialysis within 72 hours posttransplant. The duration of DGF was defined as the number of days that dialysis was required, with Australia and New Zealand Dialysis and Transplant (ANZDATA) registry collecting the dates of the last posttransplant dialysis in recipients who had experienced DGF. Kidney transplant recipients with a DGF duration that was either not recorded (n = 305) or recorded as being 0 days (n = 11) or longer than 35 days (n = 77) were excluded from this study (total n = 393). Even though the commonly accepted definition for DGF is dialysis within 7 days posttransplantation,¹⁶ the definition used by the ANZDATA registry for DGF is dialysis within 72 hours posttransplantation, which captured all recorded cases of DGF (even if DGF was defined as dialysis within 7 days). Of the 5778 recipients who did not experienced DGF, only 3 (0.05%) had dialysis after 2 weeks after transplantation (which were unlikely to represent DGF) and therefore were not included in the DGF cohort.

Data Collection

The following data were extracted from the registry: donor age, sex, and DCD status; recipient age, sex, duration of waiting time, ethnicity, ESKD cause, body mass index and comorbidities at time of transplant (diabetes, coronary artery disease, and peripheral vascular disease); immunological factors of sensitization status (ie, percentage panel-reactive antibodies [%PRA]) and HLA mismatches; transplant era, initial immunosuppression at time of transplant (calcineurin inhibitor, none, cyclosporin, or tacrolimus; antimetabolite, none, azathioprine, or mycophenolic acid; and prednisolone) and use of induction therapy (interleukin-2-receptor, T cell–depleting antibodies, or rituximab).

Clinical Outcomes

The outcomes of this study included acute rejection at 6 months (recorded as biopsy-proven), death-censored graft loss (DCGL), and overall graft loss.

Statistical Analyses

Data were expressed as number (proportion), mean (standard deviation [SD]), and median (interquartile range [IQR]), where appropriate. Associations between DGF duration and baseline characteristics were analyzed using χ² test, analysis of variance or Kruskal-Wallis test. Restricted cubic spline was used to determine the linearity of the association between DGF duration and allograft outcomes. The associations between DGF status and outcomes and between DGF duration and outcomes were examined using the adjusted Cox proportional hazard regression analysis. Covariates with P values less than 0.10 in the unadjusted analyses or covariates with established biological relationships with outcomes (eg, ethnicity, donor, and recipient age) were included in the multivariable-adjusted analyses. The proportional hazard assumptions of all models were checked graphically by plotting the Schoenfeld residuals. Results were expressed as hazard ratio (HR) with 95% confidence interval (95% CI). A sensitivity analysis was undertaken to determine the association between DGF duration and graft loss after excluding allografts that have failed within the first month posttransplant.

Mediation analysis allowing for exposure-mediator interactions was used to test whether acute rejection was a mediator between the exposure (DGF duration) and allograft outcomes (DCGL and overall graft loss), with the proportion and total effects that were mediated by both the direct (ie, the part of the exposure [DGF duration] that was not mediated by acute rejection) and indirect effects (ie, the part of the exposure [DGF duration] that was mediated by acute rejection) determined (Figure S1, SDC,https://links.lww.com/TP/B585). A mediator is the variable that often lies in the causal pathway and account for the direction and magnitude of the effect between the exposure factor (ie, duration of DGF) and the outcomes of interest (DCGL and overall graft loss).^17,18Figure S1 (SDC,https://links.lww.com/TP/B585) shows diagrammatically the differences between covariates, moderator, and mediator. The associations between DGF duration and acute rejection at 6 months and between acute rejection at 6 months and DCGL or overall graft loss were examined using adjusted Cox regression analyses.

Statistical evaluations were performed by STATA version 11 and SAS 9.4. P values less than 0.05 were considered statistically significant. The SAS macro used for the mediation analysis is shown as an online appendix by Valeri and VanderWeele.¹⁹

RESULTS

Study Population

There were 7668 deceased donor transplants performed during the study period, of which 5778 (75.4%) recipients did not experienced DGF and 1890 (24.6%) had experienced DGF (393 [20.8%] were excluded leaving 1497 [79.2%] recipients available for analysis). Of the 393 recipients excluded from the study, there were a higher proportion of recipients who had experienced DCGL compared with the study cohort (49.4% vs 16.8%, P < .001). However, the proportion of patients experiencing acute rejection at 6 months was similar (31.8% vs 33.3%, P = 0.264). Recipients in the excluded cohort were more likely to have received older donor kidneys (mean [SD] donor age, 49.6 [15.3] vs 47.6 [16.3] years; P = 0.024), but recipient age (mean [SD], 46.8 [13.1] vs 45.4 [13.5] years; P = 0.059), total ischemic time (mean [SD] 13.7 [4.9] vs 14.0 [4.9] hours; P = 0.285), HLA mismatches (mean [SD], 3.6 [1.6] vs 3.6 [1.7] mismatches; P = 0.638), and proportion of DCD kidneys (20.6% vs 23.9%, P = 0.167) were similar to the study cohort.

Of the 1497 primary deceased donor kidney transplant recipients who had experienced DGF during the study period, the median (IQR) duration of DGF was 7 (9) days, with 25% of recipients who had experienced DGF had required dialysis for at least 14 days posttransplant. The median (IQR) graft and patient follow-up periods were 4.2 (6.3) years and 4.9 (7.3) years, resulting in 8564 and 9587 graft- and patient-years, respectively. The relationship between DGF duration, DCGL, and overall graft loss were linear (Figure S2 for DCGL, SDC,https://links.lww.com/TP/B585). Delayed graft function duration was categorized into quartiles (1-4 days, 5-7 days, 8-13 days, and ≥14 days) and an incremental 5-day interval for analysis.

The characteristics of the entire cohort stratified by DGF status are shown in Table S1 (SDC,https://links.lww.com/TP/B585). Recipients who did not experience DGF were younger, less likely to have received DCD kidneys, less likely to have received induction therapy, and were more likely to have received younger donor kidneys, compared with recipients who had experienced DGF. The majority of the recipients with or without DGF had received calcineurin inhibitor (96% vs 98%; with over 50% of recipients initiated on tacrolimus), mycophenolic acid (91% vs 89%), and prednisolone (97% vs 97%).

Table 1 shows the baseline characteristics stratified by DGF duration in quartiles. Recipients with prolonged DGF were more likely to have received DCD kidney transplants, be highly sensitized, have waited longer on dialysis, have not received induction therapy, and have received donor kidneys with a longer mean total ischemic time. The proportion of recipients with prolonged DGF had reduced over successive eras (1997-2002, 58%; 2003-2008, 52%; 2009-2014, 43% required over 7 days of dialysis; P < 0.001). The characteristics of the excluded (n = 393) cohort with DGF were similar, with mean (SD) recipient age (50.9 [12.9] vs 49.6 [13.5], P = 0.075), mean (SD) number of HLA mismatches (3.6 [1.6] vs 3.6 [1.7], P = 0.638) and proportion of recipients who had received DCD kidneys (21% vs 24%, P = 0.167) not significantly different to the included cohort. The proportions of recipients who had experienced acute rejection in the excluded and included cohorts were similar (31% and 29%, P = 0.231).

TABLE 1: Baseline characteristic of deceased donor kidney transplant recipients with DGF stratified by duration of posttransplant dialysis between 1997 and 2014 (n = 1497)

Causes of DCGL

Chronic allograft nephropathy/interstitial fibrosis and tubular atrophy was the predominant cause of DCGL across quartiles of DGF duration (1-4 days, 66%; 5-7 days, 66%; 8-13 days, 61%; ≥14 days, 60%), followed by acute rejection (1-4 days, 4%; 5-7 days, 15%; 8-13 days, 11%; ≥14 days, 15%), and recurrent or de novo glomerulonephritis (1-4 days, 4%; 5-7 days, 2%; 8-13 days, 11%; ≥14 days, 8%; P value for trend = 0.230) (Figure 1).

FIGURE 1: Proportion of cause-specific death-censored graft loss in kidney transplant recipients who had experienced delayed graft function (DGF), stratified into quartiles. Causes of death-censored graft loss included CAN, acute rejection, recurrent or de novo GN, vascular complications, BKVAN, nonadherence, and other causes. BKVAN, BK viral allograft nephropathy CAN, chronic allograft nephropathy; GN, glomerulonephritis.

Association Between DGF, DCGL, and Overall Graft Loss

Compared with recipients who did not experience DGF, recipients with DGF were more likely to develop DCGL (adjusted HR, 1.34; 95% CI, 1.15-1.56; P < 0.001) and overall graft loss (adjusted HR, 1.28; 95% CI, 1.15-1.43; P < 0.001). The Kaplan-Meier failure curves for DCGL and overall graft loss are shown in Figure 2A and B, respectively.

FIGURE 2: Unadjusted Kaplan-Meier failure curves for death-censored graft loss (DCGL) (A) and overall graft loss (B), stratified by delayed graft function (DGF) status. Log-rank P values for both less than 0.001.

Association Between DGF Duration and DCGL

Compared with DGF duration of 1 to 4 days, the adjusted HR (95% CI) for DCGL for each category of DGF duration in ascending order were 1.10 (0.73-1.67; P = 0.638), 1.45 (1.00-2.11; P = 0.05), and 1.60 (1.10-2.31; P = 0.01), respectively (Table 2 and Figure 3). For every 5-day increase in DGF duration, the adjusted HR (95% CI) for DCGL was 1.11 (1.02-1.20; P = 0.011). Other covariates associated with DCGL are shown in Table 2. The Kaplan-Meier failure curves for DCGL stratified by quartiles of DGF duration are shown in Figure 4A.

TABLE 2: Associations between duration of DGF, DCGL, and overall graft loss

FIGURE 3: Forest plots showing the adjusted hazard ratio with 95% confidence interval (CI) of the association between quartiles of delayed graft function (DGF) duration of 1-4 days, 5-7 days, 8-13 days, and ≥14 days, overall graft loss and death-censored graft loss (DCGL).

FIGURE 4: Unadjusted Kaplan Meier failure curves for death-censored graft loss stratified by quartiles of delayed graft function (DGF) duration of 1-4 days, 5-7 days, 8-13 days and ≥14 days. A, The curves for the entire cohort (log-rank, P = 0.023). B, The curves for the cohort after excluding failed allografts within 1 month post-transplant (log-rank, P = 0.012).

Association Between DGF Duration and Overall Graft Loss

Compared with a DGF duration of 1 to 4 days, the adjusted HR (95% CI) for overall graft loss for each category of DGF duration in ascending order were 1.25 (0.95-1.65; P = 0.113), 1.24 (0.95-1.62; P = 0.12), and 1.44 (1.11-1.88; P = 0.006), respectively (Table 2 and Figure 3). For every 5-day increase in DGF duration, the adjusted HR (95% CI) for overall graft loss was 1.08 (1.01-1.14; P = 0.014). Other covariates associated with overall graft loss are shown in Table 2.

Association Between DGF Duration and Acute Rejection at 6 Months

The majority of the acute rejections (70%) occurred within the first month posttransplant. Of those who had experienced acute rejection within 6 months posttransplants, 9%, 26%, 54%, and 74% of acute rejection episodes occurred during DGF durations of 1 to 4 days, 5 to 7 days, 8 to 13 days, and ≥14 days, respectively. Compared with a DGF duration of 1 to 4 days, the adjusted HR (95% CI) for categories of DGF duration in ascending order were 1.13 (0.83-1.55; P = 0.428), 1.44 (1.08-1.91; P = 0.013), and 1.99 (1.50-2.65; P < 0.001), respectively, independent of HLA mismatches, %PRA, ethnicity, donor, and recipient age. For every 5-day increase in DGF duration, the adjusted HR (95% CI) for acute rejection was 1.17 (1.10-1.25; P < 0.001). The Kaplan-Meier failure curves stratified by quartiles of DGF duration for acute rejection at 6 months are shown in Figure S3 (SDC,https://links.lww.com/TP/B585).

Association Between Acute Rejection at 6 Months, DCGL, and Overall Graft Loss

Compared with recipients who did not develop acute rejection at 6 months, those who developed acute rejection at 6 months were more likely to experience DCGL and overall graft loss with adjusted HR (95% CI) of 1.56 (1.02-2.04; P = 0.001) and 1.36 (1.12-1.65; P = 0.002), respectively, independent of HLA mismatches, %PRA, ethnicity, donor, and recipient age.

Mediation Effects of Acute Rejection Between DGF Duration, DCGL, and Overall Graft Loss

Table 3 shows the estimates of the direct, indirect, and total effects between DGF duration and DCGL and overall graft loss. The proportion of the effect between DGF duration and DCGL mediated by acute rejection was 0.084 (ie, 8.4% of the effect between DGF duration and DCGL is explained by acute rejection). Similarly, the proportion of the effect between DGF duration and overall graft loss mediated by acute rejection was 0.087 (ie, 8.7% of the effect between DGF duration and overall graft loss was explained by acute rejection).

TABLE 3: Estimates from the counterfactual framework of causality model of the direct and indirect effects of DGF on DCGL, and overall graft loss

Sensitivity Analysis

There were 17 allografts (1.1% of study cohort) that had failed within the first month posttransplant with 5 cases attributed to vascular complications, 2 cases due to acute rejection, 2 cases due to hemolytic uremic syndrome, and 8 cases due to other/unknown causes (3 [18%] of 17 allograft failure within 3 days of transplant). After exclusion of these allografts, the associations between quartiles of DGF duration or every 5-day incremental increase in DGF duration and outcomes remained unchanged. Compared with recipients who had experienced DGF duration of 1 to 4 days, those who had experienced DGF duration of at least 14 days were more likely to develop DCGL and overall graft loss, with adjusted HR (95% CI) of 1.65 (1.13-2.41; P = 0.003) and 1.48 (1.14-1.93; P = 0.012), respectively. For every 5-day increase in DGF duration, the adjusted HR (95% CI) for DCGL and overall graft loss were 1.12 (1.03-1.21; P = 0.006) and 1.09 (95% CI, 1.03-1.15; P = 0.005), respectively. The Kaplan-Meier failure curves for DCGL stratified by quartiles of DGF duration are shown in Figure 4B.

DISCUSSION

Recipients of deceased donor kidney transplants who have developed DGF were more likely to experience graft loss compared with those who did not develop DGF. More importantly, we have shown a linear association between duration of DGF requiring dialysis and graft loss, such that prolonged DGF beyond 7 days posttransplant was associated with a more than 40% greater risk of DCGL. We have also found that acute rejection explained less than 10% of the effects between DGF duration and allograft outcomes, suggesting other mechanistic pathways, apart from acute rejection alone, may have influenced the long-term graft outcomes in deceased donor kidney transplant recipients.

The causative pathway resulting in DGF is often multifactorial, including immunological and nonimmunological factors that are present during organ procurement and immediately posttransplantation.^3,8 Irrespective of the causative event(s) that result in DGF, the relationship between the severity of DGF and allograft outcome remains uncertain. One prior study reported that DGF duration exceeding 6 days was associated with adverse long-term allograft survival compared to DGF of less than 6 days, whereas patients with DGF duration of less than 6 days had similar allograft survivals to those without DGF.¹⁰ In another study of 1412 kidney transplant recipients, duration of DGF (defined as requiring dialysis within the first week posttransplant) exceeding 15 days was associated with over threefold increased risks of DCGL and overall graft loss at 12 months posttransplant compared to recipients without DGF. Recipients who had experienced DGF of between 1 to 15 days had similar allograft survival up to 12 months posttransplant compared to those without DGF.⁹ This study evaluated short-term allograft survival and did not establish a linear relationship between duration of DGF and adverse allograft survival. Our present study suggested an inverse linear relationship between duration of DGF and graft survival, with an apparent DGF threshold effect for DCGL and overall graft loss of over 7 days.

The ANZDATA registry does not collect data on the number of dialysis sessions required or the reasons for initiating dialysis posttransplant, but it is likely that recipients who had received dialysis for a short duration posttransplant (ie, 1-4 days) were because of electrolyte (eg, hyperkalemia) and/or water imbalances and therefore may not necessarily represent “true” DGF. Nevertheless, differences between studies are likely to reflect the varying definitions of DGF and durations of follow-up period. Even though there have been multiple definitions of DGF proposed over time, it has been shown in previous studies that no single definition of DGF is superior compared with other definitions in predicting the risk of adverse graft outcome after kidney transplantation with the use of dialysis within the first week of transplant the most preferred and accepted definition.^16,20 Our study findings also highlight the complexity in estimating the clinical significance between DGF and allograft outcome and that there is a clear dose-dependent effect between duration of DGF and allograft loss.

Previous epidemiological studies have shown an independent association between DGF and acute rejection, presumably from a reduction in immunosuppression during periods of DGF.^11,13,21,22 In a cohort study of 645 kidney transplant recipients, the 5-year cumulative probabilities of acute rejection in recipients with and without DGF were 23% and 16%, respectively.¹¹ The influence of acute rejection on the effects between DGF on graft outcome remains unknown. Several studies have shown the additive adverse effects of DGF and acute rejection on graft survival, whereas a small single-center study suggested that the adverse association between DGF and graft survival was only evident in those who had developed acute rejection.^13,23 In this study, the effects of DGF on allograft loss were not totally mediated by acute rejection, suggesting the observed effects between DGF and allograft loss may result from other unfavorable effects of ischemic-reperfusion injury rather than a sole intervening pathological factor, such as acute rejection, that occurs after transplantation. The ANZDATA registry does not collect details of the change in immunosuppression after DGF, and therefore, we were unable to explore in detail those recipients who subsequently developed rejection-related graft loss or whether a reduction in the intensity of overall immunosuppression as a result of DGF was directly responsible for the development of acute rejection and subsequent graft loss. Nevertheless, the small number of recipients who had experienced acute rejection may have led to erroneous inference, and therefore, we were unable to generate reliable estimates with certainty.

Our study has several strengths and limitations. The prospective nature and the completeness of the data set would suggest that selection and ascertainment biases of the exposure and study factors were minimized. Nevertheless, selection bias remained possible because there were likely to be systematic differences in the management of kidney transplant recipients with varying DGF durations between transplanting centers. Even though we had adjusted for multiple confounding factors that were collected and recorded in the registry data set, there were other unmeasured residual confounders, such as changes in the dose, intensity, and types of immunosuppressive agents, hemodynamic instability, changes to the dose/types of immunosuppressive regimens during periods of DGF and use of nephrotoxins or exposure to contrast, which were not collected by the ANZDATA registry but may have potentially modified the association between the exposure factor and outcomes. It must be acknowledged that ANZDATA registry do not verify the accuracy of the data regarding the reporting or nonreporting of acute rejection to the registry, nor does it verify the accuracy of the reporting of allograft loss or DGF. However, the dates of all acute rejection episodes and allograft loss were reported to the registry.

CONCLUSIONS

There was a direct dose-dependent effect between the duration of DGF and allograft loss, such that DGF duration of at least 14 days was consistently associated with DCGL and overall graft loss. Acute rejection accounts for approximately 10% of the effects between DGF and graft loss. Future research focusing to identify other potential modifiable mediators that lie in the causal pathway between duration of DGF and allograft outcome is critical. Nevertheless, clinical interventions that aim to prevent or reduce the duration of DGF per se may potentially improve allograft outcomes in susceptible kidney transplant recipients at risk of developing DGF posttransplantation.