Although acute rejection rates have fallen over time, how this relates to graft outcomes is not known. Using data from the ANZDATA Registry, we examined associations of rejection within six months of transplantation with graft and patient outcomes among kidney-only transplants performed between April 1997 and December 2004 in Australia and New Zealand. Associations of biopsy histology with outcomes of the rejection episode were also examined. Outcomes were examined among 4325 grafts with 1961 rejection episodes in total. Crude rejection rates have fallen by one-third over that time, but rates of graft survival are constant. The occurrence of acute rejection was associated with an increased risk of graft loss after 6 months (HR, adjusted for donor and recipient characteristics, 1.69 [1.36–2.11], p < 0.001). Late rejection (first rejection ≥90 days) was associated with higher risk of graft loss (adjusted HR 2.46 [1.70–3.56], p < 0.001). Vascular rejection was also associated with a higher risk of graft loss 2.07 [95% CI 1.60–2.68], p < 0.001. The occurrence of acute rejection is associated with an ongoing increased risk of graft loss, particularly if that episode occurred late or included vascular rejection. The reduced rates of rejection have not been associated with improved graft survival.
Patient and graft survival after renal transplantation depend on adequate immunosuppression. Rates of both graft survival and rejection have steadily improved over time. Although acute rejection is still a common occurrence, the rates of short-term graft loss due to rejection among recipients of low-risk grafts are low (1). Related to this, the traditional nexus between occurrence of rejection episodes and ultimate graft outcomes has been challenged: in the US, the recent reduction in acute rejection rates has not been associated with an improvement in graft survival (2). However, this issue has not been examined in other large populations.
We analyzed associations between various types of rejection and graft outcome using Registry data including all transplants in Australia and New Zealand. From this, the extent to which changes in rejection rates relate to changes in graft outcomes were explored.
The ANZDATA Registry collects information from all renal units throughout Australia and New Zealand on all patients treated with chronic renal replacement therapy (RRT). It includes details of all renal allografts performed in Australia and New Zealand. Full details regarding the structure and methods of this Registry are reported elsewhere (3). In summary, the collection is complete from the first RRT in Australasia in 1963, and includes all renal units in both countries. Included in the data collected is information about the underlying cause of ESRD, demographic details, the type and dose of dialysis treatment, and details about renal transplantation, including HLA matching, panel reactive antibody status, graft function and immunosuppressive drug use and dosages. Immunosuppressive drug doses (but not blood concentrations) are collected by ANZDATA. Information about the cause of death of deceased donors was used to categorize these using the OPTN “extended criteria” (4). Estimated GFR was calculated using the abbreviated MDRD equation (5).
Episodes of rejection within 6 months of the date of surgery have been collected by the Registry since 1 April 1997. For each episode, the treating physician is asked to report whether the episode was proven by a biopsy. Biopsies were processed locally and interpreted by the local histopathologist. For biopsy-proven episodes of rejection, the presence of cellular, glomerular and vascular rejection was reported. The severity of each of these elements is reported on a mild/moderate/severe scale by the treating renal unit. Although Banff classification is used for some biopsies (particularly those associated with clinical trials), most centers do not use this to code routine nontrial biopsies, hence the use of the semiquantitative scale. The type of treatment and response of each episode to treatment is also included, using the codes in Table 1. The actual data entry form can be seen at http://www.anzdata.org.au/forms/Rejection2003.pdf. Although the presence of humoral rejection has subsequently been included in rejection forms, it was not on those forms used during the period of this study. Descriptive analyses of immunosuppressive drugs used those reported as the initial immunosuppressive regimen. Choice and dose of immunosuppressive agents was determined by the treating renal units; there is no national standard protocol. During the period of the study, only clinical episodes of rejection were reported to the Registry. Biopsies were performed according the individual treating nephrologists' clinical judgment—there were no standardized criteria or indications for biopsies. Similarly, treatment of episodes of rejection was at the discretion of individual treating centers.
Table 1. ANZDATA Registry codes for response to rejection episode
1. Resolution, creatinine to baseline
2. Resolution with improvement of graft function but not to baseline
3. Resolution, but no improvement, creat <250 umoL/L
4. Resolution, no improvement, creat >250 umoL/L
5. Inadequate response, graft loss <1 month after episode
This analysis includes all grafts performed in Australia and New Zealand from 1 April 1997 until 30 June 2004, to recipients 15 years or older at the time of the transplant. Followup is available to 31 December 2004 for all grafts. Only renal allografts performed as single-organ transplants were included. Outcome measures analyzed were graft and patient survival at 12 months and serum creatinine (among functioning grafts) at 12 months, and graft and patient survival over the entire follow-up period of the study.
Graft and patient survival rates at 12 months are presented as proportions with 95% binomial confidence intervals. For the 12-month outcomes, homogeneity of relationships across different strata was tested using Mantel-Haenszel tests. Population attributable fractions were calculated using standard techniques from the rates of graft failure and rates of rejection in the cohort.
Longer term outcomes were analyzed using survival techniques, restricted to grafts which had survived 6 months. Standard techniques were used including Kaplan-Meier graphs for description of univariate relationship and Cox regression models for multivariate modeling of outcomes. Proportional hazards assumptions were assessed using graphical techniques. Cox regression used ‘robust’ techniques for calculation of standard errors, modified to account for clustering of results within centers (6).
A p-value of 0.05 was regarded as statistically significant for principal comparisons and for interactions; no formal adjustment was made for multiple comparisons.
Among the total 4325 grafts available for analysis there were 1961 episodes of rejection among 1339 grafts. Baseline characteristics are shown in Table 2; most grafts received mycophenolate and cyclosporine-based immunosuppression. The classification of the severity of cellular and vascular elements of rejection reported for the first episode of rejection is shown in Table 3.
Table 2. Descriptive statistics of cohort, by the occurrence of rejection in the first 6 months posttransplantation. eGFR calculated using the abbreviated MDRD formula
Grafts with no rejection n= 2986
Grafts with at least one episodes rejection n= 1339
Age (mean [95% CI], years)
BMI (mean [95% CI], kg/m2)
1-year graft survival
p < 0.001
Immunosuppression (at baseline)
p = 0.03
Mycophenolic acid (mofetil or EC)
ATG/OKT3 (for prophylaxis within first week)
Living related donor
Living unrelated donor
Extended criteria donors (CD only)
p = 0.02
Cold ischemic time (mean [95% CI], hours, CD only)
p = 0.3
HLA matching (mean [95% CI], mismatches)
p < 0.001
p < 0.001
Serum creatinine at 12 months (umoL/L, harmonic mean [95% CI])
p < 0.001
eGFR at 12 months (mL/min/1.73m2, mean [95% CI])
p < 0.001
Table 3. Distribution of reported severity of first episodes of rejection where transplant biopsy was performed
Severity of cellular rejection
Vascular rejection severity
A good response to treatment (where the serum creatinine returned to baseline) was seen in the majority of cases (Figure 1); however, treatment was less frequently successful for those with recurrent acute rejection (Figure 1), or vascular rejection (Figure 2). When related to the reported histopathology, there was significant variation in response rates with reported histology, particularly marked when the episode included severe vascular or glomerular rejection (Figure 3, p = 0.01 for cellular rejection, p < 0.001 for vascular and glomerular rejection). After adjustment, there was no temporal trend in the frequency of a good response to the initial rejection episode (p = 0.9).
Rejection and graft loss in the first year
These analyses were restricted to grafts performed prior to 31 December 2003 (n= 3974), for which a full 12-month follow-up was available. Among these grafts, there were 90 deaths with graft function and 232 instances of loss of graft function within 12 months of transplantation, giving an overall 12-month graft failure rate of 8%. Graft failure rates at 12 months were approximately 3–4% higher among people with at least one episode of rejection, particularly if there had been an episode of vascular rejection (Table 4). However, those with episodes of cellular but not vascular rejection did not experience higher 12-month graft failure rates.
Table 4. Prevalence of graft loss in the first year, by the presence of absence of any rejection episodes or any episode of vascular rejection in the first 6 months after transplantation
Risk difference [95% CI]
Outcomes shown for all grafts (ALL) and for primary deceased donor grafts (DD1). PAF = population attributable fraction.
There was no heterogeneity across different transplant years in the associations of rejection episodes or vascular rejection episodes with graft loss (Kaplan-Meier heterogeneity χ2 p = 0.23 and 0.96 respectively). Calculations of population attributable fractions suggest that 10–15% of all graft loss episodes in the first year are statistically attributable to the occurrence of various types of rejection; furthermore, all of this excess risk is accounted for by the excess graft loss associated with episodes of vascular rejection (Table 4).
Rejection and later graft loss
An increased rate of graft loss (from 6 months) was associated with episodes of rejection through the whole period of observation (Figure 4). This risk remained constant over time posttransplantation (Figure 5). For the occurrence of any rejection, the univariate hazard ratio (HR) for graft loss after 6 months (adjusted for donor source and graft number) was 1.80 [1.45–2.24], p < 0.001. After adjustment for age, donor characteristics (source, age and gender), HLA mismatches, maximum PRA, graft number and recipient comorbidities, the association between risk of rejection in the first 6 months and graft loss (after 6 months) remained (adjusted HR 1.69 [1.36–2.11], p < 0.001). There was no interaction between the risk of graft loss associated with rejection and year of transplant.
The risk of graft failure following a first episode of rejection varied with the time after transplantation of that rejection episode. Grafts in which the first episode of rejection occurred after 90 days had a higher rate of graft loss (adjusted HR 2.46 [1.70–3.56], p < 0.001) than those where the first episode occurred in the first 90 days after transplantation (adjusted HR 1.61 [1.27–2.04], p < 0.001]; p = 0.05 for comparison to the 90–183 day HR).
For grafts which had experienced an episode of vascular rejection, the risk for graft failure adjusted for donor source and graft number was higher than that for all-cause rejection (HR 2.21 [95% CI 1.66–2.93], Figure 4). After further adjustment for age and recipient comorbidities, the HR for graft loss after 6 months associated with the occurrence of any episode of vascular rejection in the first six months after transplantation was 2.07 [95% CI 1.60–2.68], p < 0.001. This risk also did not vary over the follow-up period (Figure 5). For grafts where the first episode was cellular rejection without elements of vascular rejection, there was an increased risk of graft loss (adjusted HR 1.38 [1.04–1.79], p = 0.02).
Function of grafts at 12 months
Among primary deceased donor grafts that survived to 12 months, the serum creatinine at 12 months was lower among those without any rejection episodes; there was no significant difference in creatinine at 12 months between those grafts with one episode and those with multiple episodes of rejection, or between those with vascular and with nonvascular rejection episodes (Figure 6).
Temporal changes in rejection and graft survival
Over the time period 1997–2004, the rate of overall rejection has reduced, as has the rate of vascular rejection (rates for primary deceased donor grafts are shown in Table 5). Assuming a linear rate of change over time, the proportion of grafts with any rejection or any vascular rejection has declined by a factor of 0.92 [95% CI 0.90–0.93] (p < 0.001), and 0.92 [95% CI 0.87–0.97] (p = 0.003) times that of the previous year respectively. However, the rate of graft failure among primary deceased donor grafts did not vary significantly with year of transplantation, whether analyzed as a linear trend (RR 0.95 [95% CI 0.88–1.01] per year, p = 0.09), or as categories (by year, testing the null hypothesis of ‘no variation between year of transplantation’).
Table 5. Rates of vascular rejection (in the first 6 months) and graft loss (at 1 year and 3 years) among primary deceased donor graft recipients in Australia and New Zealand by year of transplant
Any vascular rejection
Twelve-month graft loss rate
Three-year graft loss rate
Response to treatment and graft outcome
The response to treatment was related to longer term graft outcome: when categorized by the outcomes of the first episode of rejection, a poor response to treatment of the first episode was associated with worse long-term graft survival (Table 6). However, when that first episode contained elements of vascular rejection, this difference was not seen.
Table 6. Hazard ratios for graft loss (from 6 months) for grafts by histological classification of first rejection episode and response to treatment, compared to grafts with no rejection
All first episodes
Cellular rejection without elements of glomerular or vascular rejection
1.69[1.35–2.13], p < 0.001
1.61[1.28–2.03], p < 0.001
1.27[0.94–1.73], p = 0.1
1.25[0.90–1.72], p = 0.2
2.63[1.84–3.76] p < 0.001
2.35[1.63–3.40], p < 0.0.01
2.00[1.54–2.59], p < 0.001
1.88[1.39–2.56], p < 0.001
1.69[1.15–2.49], p < 0.01
1.74[1.11–2.72], p = 0.02
2.51[1.12–5.28], p = 0.02
2.23[1.12–4.46], p = 0.02
To evaluate the implications of a single episode of cellular rejection, graft outcomes among people with only one episode of rejection biopsied and coded as cellular (but not including vascular or glomerular elements) were compared to those with no rejection. Those grafts with a single episode did not differ statistically in their longer term risk of graft loss (univariate HR 0.98 [0.72–1.33] p = 0.9, adjusted HR 0.98 [0.71–1.34], p = 0.9).
Immunosuppressive strategies to prevent and treat acute rejection of renal allografts enabled the implementation of kidney transplantation as a viable form of treatment for end stage kidney failure in the 1960s and 1970s. Subsequent improvements in immunosuppression have led to successive reductions in rates of acute rejection and improvements in patient and graft survival that have established transplantation as the preferred form of RRT. In the current analysis of an extensive dataset, we have demonstrated a progressive reduction from 1997 in rates of rejection occurring within the first 6 months of transplantation. The rates of rejection we observed in Australia and New Zealand are higher than those in contemporary reports from the United States (7), possibly reflecting lower rates of use of IL-2 receptor antagonists and tacrolimus in Australia and New Zealand over the study period.
We have demonstrated an association between the occurrence of rejection and a higher rate of graft loss in the short term (approximately 15% of the graft loss rate statistically is attributable to acute rejection) and also the longer term, with an ongoing higher rate of graft loss for at least the first 5 years following transplantation. Furthermore, we have shown that the histological type of rejection and the response to treatment are related to outcome, and that a single episode of cellular rejection which responds well to treatment is not associated with worse graft outcome.
However, despite the reduction in rates of rejection we observed, there has not been a clear improvement in overall graft survival, either at 12 or 36 months. This is similar to a US-based registry analysis of the current era of kidney graft recipients which have suggested a disconnection of the relationship between acute rejection and graft survival, with no relationship observed between recent reductions in the incidence of acute rejection have had no significant impact on graft survival (2). There are several possible explanations. The previous episodes of rejection which have now reduced in frequency may not have been associated with graft loss (the main fall has been in rate of any rejection; whether there has been a reduction in rates of vascular rejection is less clear in our data). Alternatively, there may be a detrimental effect associated with the better prophylaxis of rejection episodes which balances the benefits of reductions in rejection rates. Statistical explanations also exist—residual confounding (i.e. incomplete adjustment for covariates, especially given the changing characteristics of donors and recipients) or type 2 error.
The attributable risk calculations emphasize the trappings of success of modern immunosuppression: the absolute rate of graft loss is now considerably lower than in decades past. The rate of all graft losses in the first 12 months statistically attributed to acute rejection is less than 4%. Hence, even if a 25% reduction in rejection rates were entirely translated to improved graft survival, the difference would be less than 1%. To an extent our results might therefore represent type 2 error, with insufficient numbers to demonstrate a difference of this magnitude (a 1% per year difference falls within the 95% confidence intervals of our temporal trend). Unfortunately, given that all grafts performed throughout both Australia and New Zealand are already included in this analysis, there is little prospect for reducing the possibility of type 2 error with data from this Registry.
The term acute rejection encompasses a range of pathologies of varying severity. For the purposes of this analysis, relatively crude divisions into cellular with no vascular component (encompassing Banff grades IA, IB and possibly borderline changes in some cases) or vascular (Banff grades IIA and higher), single or multiple episodes and good versus poor response to therapy were employed for simplicity of reporting. It is clear that different forms of acute rejection and responses to therapy are associated with different levels of risk for graft survival. In this analysis, the excess of graft loss that was associated with acute rejection was attributable statistically to vascular rejection. The reported severity of rejection was also of relevance, with lower rates of good response and higher rates of graft loss associated with more severe grades of rejection.
In contrast to vascular rejection, whether there is a higher rate of graft loss associated with cellular rejection which is mild in severity or which responds to treatment is less clear. The 1-year graft survival rates for those with only cellular rejection are the same as those among people without. Among those with a single episode of cellular rejection and a good response to treatment, we did not observe an increased rate of graft loss in the longer term rates of graft survival. Although there may be a degree of bias in the attribution of histological severity due to the nature of reporting to the Registry (the occurrence, histology, treatment and outcome are all reported together by the treating renal unit), this should not affect the relationship between reported rejection in the first 6 months and graft loss events in subsequent years. For many subjects and clinicians, the finding that single episodes of rejection in which there is no vascular component experience graft survival rates out to 5 years that are equal to those who remain free from acute rejection will be reassuring, and may temper efforts to eliminate acute rejections ‘at all costs’.
The greater risk of graft loss among those who experienced their first episode of rejection in months 4–6 after transplantation is consistent with reports from elsewhere (8, 9). There are several possible reasons for this; late rejections might signal a higher likelihood of ongoing immunological activity, or predict ongoing problems with underimmunosuppression (due to under-dose, poor absorption or poor medication compliance). Late rejection has also been shown to predict a higher likelihood of chronic allograft nephropathy (10).
There is now a substantial body of literature which suggests that grafts which survive the early period posttransplantation and have good function may be disadvantaged by calcineurin inhibitors, with better results among grafts on calcineurin-free immunosuppression regimens in studies from a variety of sources (11–13). This suggests that further improvement in longer term graft survival may not simply require intensified immunosuppression, but alternatively that avoidance of nephrotoxins might be a beneficial approach. Further long-term data are required to confirm or refute this notion (14).
In conclusion, we have shown that the occurrence of rejection in the first 6 months after transplantation is associated with ongoing increase in risk of graft loss, particularly among people who experience an episode of vascular rejection, recurrent rejection or late (4–6 months posttransplantation) acute rejection. In contrast, a single episode of acute rejection which responds to treatment with a return to baseline creatinine is not associated with an increase in graft loss. Despite the reduction in rates of rejection over time, there was no clear improvement in overall graft survival either at 12 or 36 months.
None of the authors have a direct conflict of interest in the material presented in this paper. Drs Chadban, Russ and Campbell are, or have been, members of Advisory Boards for pharmaceutical companies which manufacture immunosuppressive drugs.
The ANZDATA Registry is funded by the Australian Government Department of Health and Ageing, the New Zealand Ministry of Health and Kidney Health Australia. The Registry has also received contributions from various pharmaceutical and dialysis companies on an unrestricted basis.