Long-Term Deterioration of Kidney Allograft Function

Authors


*Corresponding author: Bertram L. Kasiske, kasis001@umn.edu

Abstract

Although long-term survival after kidney transplantation is critically dependent on maintaining stable allograft function, few studies have examined renal allograft function over time. Using pooled data from 10 278 consecutive transplants at five centers, we calculated slopes of estimated glomerular filtration rates (eGFR) measured after 1, 6 and 12 months in 9515, 8861 and 7359 patients surviving ≥1, ≥6 and ≥12 months, respectively. Slopes of eGFR progressively diminished for patients transplanted during 1984–1989, 1990–1993, 1994–1998 and 1999–2002 (analysis of variance p < 0.0001 and p = 0.1245 for slopes measured after 1 and 6 months, respectively). Slopes measured after 12 months were less in the most recent era: −2.2 ± 7.2 mL/min/1.73 m2/year, −2.3 ± 6.6 mL/min/1.73 m2/year, −2.4 ± 7.4 mL/min/1.73 m2/year and −1.4 ± 10.9 mL/min/1.73 m2/year, respectively, p = 0.0058. Slopes measured after 1, 6 and 12 months each were less for transplantations during 1999–2002, after adjusting for multiple transplantation characteristics (p < 0.0001). Similarly, in Cox proportional hazards analysis, the risk (95% CI) for a 25% reduction in eGFR was 0.92 (0.85–1.01), p = 0.0736 during 1990–1994; 0.94 (0.82–1.08), p = 0.4111 during 1995–1998 and 0.78 (0.64–0.95), p = 0.0110 during 1999–2002 (compared to 1984–1989). We conclude that the rate of decline in allograft function after kidney transplantation has improved, suggesting that stable, long-term function may be achievable.

Introduction

Over the past two decades there have been dramatic declines in rates of acute rejection and allograft failures that occur in the first 3–6 months after kidney transplantation (1). These gains in early outcomes have made future improvements more dependent on reductions in the rate of late allograft failure. Whether there has been significant improvement in the rate of late allograft failures is controversial. Using data from the United Network for Organ Sharing (UNOS), Hariharan and colleagues reported substantial improvements in the projected half-lives of kidney allografts, a measure of late allograft failure (2). Subsequently, Meier-Kriesche and colleagues reported that there has been little or no improvement in actual late graft failure (3). These two studies examined transplants carried out on or before 1995.

Few would argue that reducing the rate of late allograft failure depends on maintaining stable allograft function. Surprisingly, few studies have examined long-term kidney function in large numbers of unselected transplant recipients. Single centers rarely have sufficient data, and registry data rarely provide enough detail for accurate assessment of post-transplant function. The UNOS registry, for example, does not record the date the function was assessed, but reports only the ‘last known’ serum creatinine. In the current study, we pooled data from five centers in North America (each of which maintains close, long-term, patient follow-up) to examine trends in long-term kidney allograft function. Specifically, we hypothesized that there has been an improvement in the rate of decline in long-term kidney allograft function over the past two decades (4).

Methods

Patient population

We pooled data from all kidney transplantations carried out at five centers between January 1984 and December 2002. Patients transplanted with other organs were excluded, but kidney recipients who subsequently underwent pancreas transplantation were followed until pancreas transplantation, and censored at that time. Patients receiving islet transplantations were neither excluded nor censored.

Risk factors and outcomes examined

We collected data on donor, recipient and transplantation characteristics that have previously been reported to influence allograft survival. Delayed graft function was defined by the need for dialysis in the first week after transplantation. Acute rejection was defined by the need for treatment, with or without biopsy confirmation. Graft failure was defined as death, return to dialysis or retransplantation. Death-censored graft failure was defined by return to dialysis or retransplantation.

Data were collected on donor source (living or deceased) year of transplantation, prior transplantation, age, sex, ethnicity, primary cause of kidney disease, body weight, time with end-stage kidney disease prior to transplantation (in 4 of the centers), donor age, donor sex, donor race/ethnicity, number of major histocompatibility mismatches, panel-reactive antibody greater than 30% and initial immunosuppression used (analyzed by intention to treat).

We also recorded serum creatinines measured in all patients at months 1, 3, 6, 9, 12 and annually until graft failure or last follow-up. We used these serum creatinines to estimate the glomerular filtration rate (eGFR) using the 4-variable formula developed with data from the Modification of Diet in Renal Disease (MDRD) study, where eGFR in mL/min/1.73 m2= Exponent (5.228 – 1.154 × Log (creatinine in mg/dL) − 0.203 × Log (age in years) − 0.299 [if female]+ 0.192 [if Black]) (5). We also tested key findings using creatinine clearance estimated by the Cockcroft-Gault formula (eCCr), where eCCr in mL/min = ([140 − age in years]) × weight in kilograms) ÷ (72 × creatinine in mg/dL) × 0.85 [if female] (6). When body weights were missing, we assumed that the weight was the same as the most recently recorded weight. For each patient surviving with a functioning graft for ≥1 month, and having at least two eGFR values, we calculated the slope (SeGFR1) and intercept of the eGFR over time using least squares regression. Similarly, for each patient surviving with a functioning graft for ≥6 months and ≥12 months, we calculated the slopes (SeGFR6 and SeGFR12, respectively) and intercepts of the eGFR's.

Institutional review and patient privacy protection

This study was approved by the Institutional Review Board at Hennepin County Medical Center, where the analysis was carried out by the first author. Data were supplied by the individual transplant centers to the first author, who was the only person with access to the pooled data. These data were de-identified by removing all patient names, hospital numbers, birth dates and dates of transplantation.

Statistical analysis

We examined differences between groups using analysis of variance (ANOVA) for parametric data, and the Scheffe test for comparing differences between multiple groups. Differences between proportions were tested with chi square tests. Differences in survival times were examined using Kaplan-Meier curves with the log-rank test and Cox proportional hazards analyses. We examined factors associated with differences in the slopes and intercepts of eGFR using multiple linear regression. Results are expressed as mean ±SD, unless otherwise indicated. All analyses were performed with SAS Version 9.1 for the personal computer (SAS Institute, Inc., Cary, NC).

Results

Patient population

The patient population in this study is similar to the population of kidney transplant recipients throughout North America (Table 1).

Table 1.  Selected transplantation characteristics
CharacteristicsPercent
Deceased donor62.6
First transplant86.6
End-stage renal disease type 1 diabetes17.5
End-stage renal disease type 2 diabetes9.4
Age ≥60 years9.8
Donor age ≥60 years4.4
Caucasian74.2
Black20.8
Native American2.6
Preemptive transplantation5.9
0-Major histocompatibility complex mismatches9.3
Panel-reactive antibody >30%7.9
Cyclosporine used initially88.6
Tacrolimus used initially7.2
Rapamycin used initially7.8
Azathioprine used initially57.4
Mycophenolate mofetil used initially34.6
Interleukin-2 receptor antibody used initially14.5
Other antibody induction used initially50.3

Outcomes

Overall, 40% of graft failures were attributable to death of the recipient, and 60% to other causes (death censored). Of transplantations carried out during 1984–1993, 63.7% failed; Of these failures, 60% were due to death-censored graft failures, while 40% were due to deaths. Of transplantations done during 1994–2002, 25.0% failed, 58% due to death-censored graft failures and 42% due to patient death. Thus, the proportion of grafts lost to death compared to other (death censored) causes was relatively constant over time.

There were steady improvements in the survival of transplants carried out in four arbitrarily defined (a priori) transplantation eras (Figure 1). For example, 5-year Kaplan-Meier graft survival was 60.4% for transplants during 1984–1989, 67.4% during 1990–1994, 71.7% during 1995–1998 and 84.0% during 1999–2002 (log-rank p < 0.001). Even after adjusting for differences in donor and recipient characteristics, as well as initial immunosuppressive medications used (by intention to treat), there was still less graft failure. Compared to 1984–89, there was a 15–16% reduction in risk for transplants during 1990–1994 and 1995–1999, but a 33% reduction in risk for transplants during 1999–2002 (Table 2). This reduction in risk appeared to be more attributable to declines in death-censored graft failures than to reduced mortality (Table 2). There was also a remarkable 50% decrease in the relative risk (RR) for acute rejection episodes during 1999–2002 compared to 1984–1989 (Table 2). This reduction in acute rejection might explain most of the reduction in graft failure, since after taking first acute rejections into account, graft failure was not significantly different during 1999–2002 compared to 1984–2002 (Table 2).

Figure 1.

Graft survival (Kaplan-Meier) by year of transplantation. The top panel includes all patients (n = 10 278). The middle panel includes only patients who survived with a functioning graft for more than 3 months (n = 9254), and the bottom panel includes patients who survived with a functioning graft more than 1 year (n = 8649). In each case p < 0.0001 by the log-rank test.

Table 2.  Trends in adjusted outcomes after kidney transplantation
Year of transplantationAdjusted* relative risk (95% confidence interval)
All (N = 10 278)Deceased donor (N = 6431)Living donor (N = 3847)
  1. *In addition to the transplant eras indicated above, each Cox proportional hazards analysis was also adjusted for transplant center, age, donor age, ethnicity, donor ethnicity, sex, donor sex, body mass index, cause of kidney disease, years of dialysis before to transplantation, prior transplant, panel reactive antibody titer, type of initial immunosuppressive medications used (by intention to treat) and delayed graft function.

Death with function
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.80 (0.71–0.90), p = 0.00040.79 (0.68–0.91), p = 0.00110.78 (0.60–1.01), p = 0.0595
Transplant during 1995–19980.83 (0.67–1.03), p = 0.08960.87 (0.68–1.12), p = 0.28100.61 (0.39–0.97), p = 0.0376
Transplant during 1999–20020.75 (0.53–1.07), p = 0.10980.71 (0.46–1.09), p = 0.11730.75 (0.38–1.46), p = 0.3905
Death-censored graft failure
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.87 (0.79–0.96), p = 0.00650.87 (0.77–0.97), p = 0.01550.90 (0.72–1.11), p = 0.3139
Transplant during 1995–19980.86 (0.72–1.03), p = 0.10140.85 (0.69–1.05), p = 0.12460.98 (0.67–1.43), p = 0.9028
Transplant during 1999–20020.63 (0.46–0.85), p = 0.00310.63 (0.44–0.92), p = 0.01540.74 (0.40–1.36), p = 0.3327
Graft failure
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.84 (0.78–0.91), p < 0.00010.84 (0.77–0.92), p = 0.00020.82 (0.70–0.97), p = 0.0197
Transplant during 1995–19980.85 (0.74–0.97), p = 0.01950.86 (0.73–1.01), p = 0.06410.80 (0.60–1.07), p = 0.1365
Transplant during 1999–20020.67 (0.53–0.84), p = 0.00070.65 (0.49–0.86), p = 0.00280.77 (0.49–1.20), p = 0.2462
First acute rejection
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.97 (0.89–1.05), p = 0.39400.99 (0.90–1.09), p = 0.87880.88 (0.74–1.03), p = 0.1107
Transplant during 1995–19980.87 (0.76–0.99), p = 0.03770.87 (0.75–1.02), p = 0.08480.87 (0.67–1.12), p = 0.2706
Transplant during 1999–20020.50 (0.41–0.61), p < 0.00010.44 (0.34–0.57), p < 0.00010.56 (0.39–0.79), p = 0.0012
Graft failure—adjusted for first acute rejection
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.85 (0.79–0.92), p < 00010.85 (0.78–0.93), p = 0.00040.85 (0.72–1.00), p = 0.0516
Transplant during 1995–19980.90 (0.78–1.03), p = 0.12220.91 (0.78–1.07), p = 0.25260.87 (0.65–1.17), p = 0.3594
Transplant during 1999–20020.79 (0.63–1.00), p = 0.05180.80 (0.60–1.05), p = 0.11100.93 (0.59–1.45), p = 0.7353

It appears that much of the difference in graft failure between the transplant eras is due to differences in graft failures that occur in the first 3 months (Figure 1). As a result, the unadjusted, 5-year Kaplan-Meier (conditional) graft survival for patients with a functioning allograft >3 months post-transplant was 69.6%, 74.6%, 77.0% and 86.9% during 1984–1989, 1990–1994, 1995–1998 and 1999–2002, respectively (Figure 1). Among those who survived with a functioning graft for >1 year, the 5-year conditional graft survival was 74.9%, 78.4%, 80.0% and 89.4% in the four eras, respectively (log rank p < 0.0001).

We examined differences in the adjusted risk for graft failure between eras among patients who survived >3 months with a functioning allograft (Table 3). The risk for graft failure, after adjusting for basic recipient and donor characteristics, was significantly lower during 1999–2002 compared to 1984–1989 (Table 3). Further adjusting for baseline eGFR, as well as early (<3 months) and late (>3 months) acute rejections, blunted the reduction in the risk of graft failure attributable to transplant eras (Table 3).

Table 3.  Trends in adjusted outcomes of transplants that functioned for at least 3 months
Year of transplantationAdjusted* relative risk (95% confidence interval)
All (N = 9254)Deceased donor (N = 5681)Living donor (N = 3573)
  1. *In addition to the transplant eras indicated above, each Cox proportional hazards analysis was also adjusted for transplant center, age, donor age, ethnicity, donor ethnicity, sex, donor sex, body mass index, cause of kidney disease, years of dialysis before to transplantation, prior transplant, panel reactive antibody titer, type of initial immunosuppressive medications used (by intention to treat) and delayed graft function.

  2. The baseline eGFR was the mean of eGFR at 1 and 2 months or, in the 45 instances where these two values were missing, eGFR was the intercept of the regression line for subsequent eGFR values.

  3. Whether or not there was an acute rejection episode treated in the first 3 months post-transplant.

  4. #Whether or not there was an acute rejection episode treated after 3 months (analyzed as a time-dependent covariate).

  5. First occurrence of a 25% decline in eGFR from the baseline eGFR (analyzed as a time-dependent covariate).

  6. GFR = glomerular filtration rate estimated using the 4-variable Modification of Diet in Renal Disease equation (5).

Death with function
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.79 (0.70–0.90), p = 0.00050.78 (0.67–0.91), p = 0.00130.79 (0.60–1.04), p = 0.0954
Transplant during 1995–19980.79 (0.62–1.00), p = 0.04690.80 (0.61–1.05), p = 0.10220.62 (0.38–1.01), p = 0.0560
Transplant during 1999–20020.75 (0.51–1.12), p = 0.15570.60 (0.37–0.98), p = 0.04210.82 (0.40–1.70), p = 0.5939
Death-censored graft failure
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.85 (0.75–0.95), p = 0.00520.84 (0.73–0.96), p = 0.01330.85 (0.67–1.07), p = 0.1677
Transplant during 1995–19980.75 (0.60–0.94), p = 0.01250.78 (0.61–1.01), p = 0.05930.69 (0.43–1.11), p = 0.1245
Transplant during 1999–20020.54 (0.37–0.78), p = 0.00120.56 (0.36–0.87), p = 0.00980.52 (0.25–1.09), p = 0.0839
Graft failure
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.83 (0.76–0.90), p < 0.00010.82 (0.74–0.91), p = 0.00010.80 (0.67–0.96), p = 0.0162
Transplant during 1995–19980.78 (0.66–0.91), p = 0.00200.80 (0.67–0.96), p = 0.01860.66 (0.47–0.92), p = 0.0153
Transplant during 1999–20020.63 (0.48–0.82), p = 0.00070.58 (0.42–0.80), p = 0.00100.66 (0.39–1.12), p = 0.1222
Graft failure: adjusted for baseline eGFR
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.84 (0.77–0.92), p = 0.00010.84 (0.76–0.93), p = 0.00060.83 (0.69–0.99), p = 0.0346
Transplant during 1995–19980.84 (0.72–0.99), p = 0.03770.87 (0.72–1.04), p = 0.13280.73 (0.52–1.03), p = 0.0706
Transplant during 1999–20020.67 (0.51–0.88), p = 0.00340.62 (0.45–0.86), p = 0.00410.72 (0.42–1.21), p = 0.2088
Graft failure: adjusted for baseline eGFR+ rejection <3 months
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.84 (0.77–0.91), p < 0.00010.83 (0.75–0.92), p = 0.00040.82 (0.69–0.98), p = 0.0280
Transplant during 1995–19980.84 (0.71–0.98), p = 0.02880.86 (0.71–1.04), p = 0.11300.73 (0.52–1.03), p = 0.0695
Transplant during 1999–20020.68 (0.52–0.89), p = 0.00460.63 (0.46–0.87), p = 0.00540.73 (0.43–1.23), p = 0.2314
Graft failure: adjusted for baseline eGFR+ rejection <3 months+ rejection >3 months#
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.87 (0.80–0.95), p = 0.00240.83 (0.75–0.92), p = 0.00040.86 (0.72–1.03), p = 0.1104
Transplant during 1995–19980.88 (0.75–1.03), p = 0.11140.86 (0.71–1.04), p = 0.11300.74 (0.53–1.05), p = 0.0879
Transplant during 1999–20020.74 (0.57–0.98), p = 0.03210.63 (0.46–0.87), p = 0.00540.77 (0.45–1.29), p = 0.3189
Graft failure: adjusted for baseline eGFR+ rejection <3 months+ rejection >3 months#+ decreased 25% eGFR
Transplant during 1984–19891.00 (reference)1.00 (reference)1.00 (reference)
Transplant during 1990–19940.91 (0.83–0.99), p = 0.03130.90 (0.81–1.00), p = 0.04140.86 (0.72–1.03), p = 0.1089
Transplant during 1995–19980.90 (0.76–1.05), p = 0.18540.92 (0.77–1.11), p = 0.39530.76 (0.53–1.07), p = 0.1153
Transplant during 1999–20020.78 (0.60–1.02), p = 0.06990.71 (0.52–0.99), p = 0.04180.85 (0.50–1.44), p = 0.5503

The risk of reaching a 25% reduction in eGFR declined during 1999–2002 compared to 1984–1989. In a fully-adjusted, Cox proportional hazards analysis, the RR for a 25% reduction in eGFR (compared to 1984–1989) was 0.92 (0.85–1.01), p = 0.0736 during 1990–1994; 0.94 (0.82–1.08), p = 0.4111 during 1995–1998 and 0.78 (0.64–0.95), p = 0.0110 during 1999–2002. When including a 25% or more decline in eGFR as a time-dependent covariate, the RR associated with graft failure during 1999–2002 was only 22% lower than that for transplants during 1984–1989, and in fact was no longer statistically different compared to the earlier era (Table 3). This suggests that reduced acute rejection and moderation in declines in graft function together account for much, if not all, of the lower risk for late (>3 months) graft failure during 1999–2002.

The best clinical correlates of late allograft failure

Among patients who survived with a functioning allograft for at least 3 months, reaching a 25% reduction in eGFR was the strongest (highest χ2) risk factor associated with subsequent graft failure. After adjusting for multiple recipient and donor characteristics, the RR of graft failure associated with a 25% reduction in eGFR was 3.67 (3.40–3.96); χ2= 1123, p ≤ 0.0001. In contrast, the RR associated with baseline function (each mL/min/1.73 m2 in eGFR) was 0.978 (0.975–0.980); χ2= 277, p =≤0.0001. For acute rejection in the first 3 months post-transplant RR = 1.56 (1.44–1.70); χ2= 108, p ≤ 0.0001, and for acute rejection after the first 3 months RR = 3.28 (2.98 – 3.60); χ2= 602, p ≤ 0.0001.

The results were even more striking for death-censored graft failure, where the RR associated with reaching a 25% reduction in eGFR was 7.85 (7.01–8.80); χ2= 1123, p ≤ 0.0001, while for baseline eGFR RR = 0.965 (0.961–0.969); χ2= 354, p ≤ 0.0001, for acute rejection in the first 3 months RR = 2.13 (1.88–2.40); χ2= 145, p ≤ 0.0001 and for acute rejection after the first 3 months RR = 5.34 (4.70–6.06); χ2= 669, p ≤ 0.0001.

Changes in allograft function assessed by eGFR and interval changes in eGFR

There appeared to be little change in the means of eGFR at different times after transplantation, but eGFRs were highest in the most recent era (Table 4). Between baseline and 3 months, function increased in transplants carried out during 1984–1989 and 1990–1994, while function declined slightly in transplants carried out during 1995–1998 and 1999–2002 (Table 4). There were no significant differences in interval changes in eGFR between 3 and 6 months or 6 months and 1 year (Table 4). Between 1 and 2 years, function declined less in transplants carried out during 1984–1989 and 1990–1994, compared to transplants carried out during 1995–1998 and 1999–2002 (Table 4).

Table 4.  Difficulty in interpreting function when different patients are analyzed at different times (due to attrition from graft failures) with different baseline function
Estimated and interval changes in glomerular filtration rates
EraBaseline3 months6 months1 year2 years
  1. Values are mean ± standard deviations. p-values are by ANOVA. All had ≥2 years follow-up. Any shared superscript indicates p > 0.05 by Sheffe's test (lack of any shared superscript indicates p < 0.05).

Estimated glomerular filtration rates (mL/min/1.73 m2)
1984–198946.1 ± 16.8a47.0 ± 16.5a47.3 ± 16.3a47.6 ± 17.1a48.0 ± 18.4a
 (N = 2325) (N = 2153) (N = 2086) (N = 1847) (N = 1751)
1990–199446.6 ± 18.2a48.3 ± 18.4a49.4 ± 17.1b49.0 ± 17.7a50.2 ± 19.1b,c
 (N = 2768) (N = 2286) (N = 2495) (N = 2005) (N = 2128)
1995–199852.6 ± 20.2b53.6 ± 19.5b52.4 ± 18.5c51.8 ± 18.9b49.8 ± 18.5b
 (N = 2414) (N = 2160) (N = 2169) (N = 1957) (N = 1924)
1999–200254.8 ± 18.4c55.1 ± 18.4b54.6 ± 18.1d53.7 ± 17.6c51.8 ± 18.4c
 (N = 2254) (N = 2118) (N = 2037) (N = 1767) (N = 1255)
p < 0.0001p < 0.0001p < 0.0001p < 0.0001p < 0.0001
Interval changes in estimated glomerular filtration rates (mL/min/1.73 m2)
1984–19890.65 ± 11.8a,b−0.80 ± 12.0a−0.35 ± 13.0a−0.06 ± 14.1a,b
 (N = 2152) (N = 2009) (N = 1809) (N = 1626)
1990–19941.48 ± 11.6a−0.65 ± 12.2a−1.15 ± 12.0a0.40 ± 13.3a
 (N = 2283) (N = 2137) (N = 1977) (N = 1789)
1995–1998−0.17 ± 12.5b−1.50 ± 12.3a−0.76 ± 12.2a−2.77 ± 13.5c,d
 (N = 2159) (N = 2039) (N = 1909) (N = 1772)
1999–2002−0.13 ± 11.7b−0.42 ± 13.0a−1.00 ± 11.9a−1.42 ± 13.0b,d
 (N = 2117) (N = 1986) (N = 1726) (N = 1157)
p < 0.0001p = 0.0339p = 0.2088p < 0.0001

Changes in allograft function assessed by linear regression slopes of eGFR

For each transplant, there were 9.3 ± 4.2 (range: 2–23) creatinines to calculate slopes after 1 month, 7.1 ± 3.9 (range: 2–20) creatinines to calculate slopes after 6 months and 6.2 ± 3.6 (range: 2–18) creatinines to calculate slopes after 12 months. Slopes measured after 1 month were based on only two values in 222 of 9515 (2.3%). Slopes after 6 months were based on two values in 618 of 8681 (7.1%). Slopes after 12 months were based on two eGFR values in 1080 of 7359 (14.7%). We calculated the R2 values for each slope that included more than two eGFR values. The mean ± SD for the R2 for regression relationships yielding the slopes measured after 1, 6 and 12 months, respectively, were 0.39 ± 0.29, 0.44 ± 0.32 and 0.44 ± 0.32. Thus, the ‘average’ regression relationship was a reasonably good fit.

Baseline function, as the intercept of regression with eGFR versus time measured after 1 month, improved in recent years (Table 5). Similarly, eGFR measured at 1 month post-transplant was 46.0 ± 18.0a mL/min/1.73 m2 (N = 2221), 46.3 ± 8.6a mL/min/1.73 m2 (N = 2683) and from 52.5 ± 21.4b mL/min/1.73 m2 (N = 2384) to 54.7 ± 19.7c mL/min/1.73 m2 (N = 2229) during 1984–1989, 1990–1994, 1995–1998 and 1999–2002, respectively (shared superscripts indicate p < 0.05 by ANOVA and with Scheffe's test for multiple comparisons).

Table 5.  Baseline and rate of decline in estimated glomerular filtration rate by year of transplantation (unadjusted)
Year of transplantBaseline and rate of decline in eGFR
Measured after 1 monthMeasured after 6 monthsMeasured after 12 months
  1. Mean ± SD baseline (intercept) and rate of decline (slope) in estimated glomerular filtration rate over time, determined by least squares fitted regression lines for each patient. Included in the analysis are 9515, 8861 and 7359 patients who survived with a functioning graft for at least 1, 6 and 12 months, respectively, and had at least two serum creatinine measurements. Any shared superscript indicates p > 0.05 between eras by Sheffe's test (lack of any shared superscript indicates p < 0.05).

  2. eGFR = glomerular filtration rate estimated using the 4-variable Modification of Diet in Renal Disease formula (5).

Baseline (intercept) = mean ± SD mL/mn/1.73 m2
1984–198949.0 ± 15.9 (2280)a51.3 ± 18.2 (2073)a53.7 ± 19.0 (1873)a
1990–199449.9 ± 16.7 (2679)a52.9 ± 18.9 (2465)a55.3 ± 21.1 (2242)a
1995–199853.8 ± 18.5 (2345)b54.3 ± 20.0 (2167)b55.1 ± 22.3 (1998)a
1999–200255.0 ± 18.6 (2211)b55.5 ± 22.4 (1976)b54.6 ± 22.3 (1246)a
p < 0.0001p < 0.0001p = 0.0789
Rate of decline (slope) = mean ± SD mL/min/1.73 m2/year
1984–1989−5.1 ± 32.4 (2280)a−2.8 ± 14.5 (2073)a−2.2 ± 7.2 (1873)abc
1990–1994−4.5 ± 32.0 (2679)a−3.1 ± 15.4 (2465)a−2.3 ± 6.6 (2242)ab
1995–1998−3.2 ± 32.4 (2345)ab−2.8 ± 13.7 (2167)a−2.4 ± 7.4 (1998)b
1999–2002−0.9 ± 35.5 (2211)b−1.9 ± 21.1 (1976)a−1.4 ± 10.9 (1246)c
p < 0.0001p = 0.1245p = 0.0058

The intercepts for eGFR versus time (in least squares regression) were highest in patients who were alive with a functioning graft at last follow-up. Among those surviving ≥1 month, intercepts were 54.5 ± 16.4a mL/min/1.73 m2 (n = 5737), 49.2 ± 18.3b mL/min/1.73 m2 (n = 1654) and 46.7 ± 18.9b mL/min/1.73 m2 (n = 2124) for those who remained alive, died or required retransplantation/returned to dialysis, respectively (shared superscripts indicate p > 0.05). The corresponding values for those surviving ≥6 months were 54.6 ± 18.2a mL/min/1.73 m2 (n = 5495), 50.7 ± 19.1b mL/min/1.73 m2 (n = 1424) and 52.3 ± 24.8b L/min/1.73 m2 (n = 1762), respectively. The values for those surviving ≥12 months were 54.9 ± 19.3a,b mL/min/1.73 m2 (n = 4770), 53.3 ± 22.4a mL/min/1.73 m2 (n = 1207) and f, respectively.

Slopes (SeGFR1, SeGFR6 and SeGFR12) were significantly steeper among patients who required retransplantation or returned to dialysis. Among those surviving ≥1 month SeGFR1 was −0.49 ± 19.6a mL/min/1.73 m2/year (n = 5737), 0.40 ± 49.7a mL/min/1.73 m2/year (n = 1654) and −14.7 ± 42.3b mL/min/1.73 m2/year (n = 2124) for those who remained alive, died or required retransplantation or return to dialysis, respectively. The corresponding SCCr6 values for those surviving ≥6 months were −0.5 ± 11.5a mL/min/1.73 m2/year (n = 5495), −0.5 ± 14.2a mL/min/1.73 m2/year (n = 1424) and −11.2 ± 25.4b mL/min/1.73 m2/year (n = 1762), respectively. The corresponding SCCr12 values for those surviving ≥12 months were −0.6 ± 5.8a mL/min/1.73 m2/year (n = 4770), −1.2 ± 7.6b mL/min/1.73 m2/year (n = 1207) and −8.2 ± 10.8c mL/min/1.73 m2/year (n = 1382), respectively.

There was marked variability in SeGFR1, somewhat less variability in SeGFR6 and even less variability in SeGFR12 (Table 5). Unadjusted SeGFR1 and SeGFR12 were reduced for transplants performed during 1999–2002 compared to 1984–1989 (Table 5). After adjusting for multiple recipient and donor characteristics, slopes were significantly less steep during 1999–2002 compared to 1984–1989 (Table 6). If slopes were also adjusted for immunosuppressive medications, the improvement in recent transplantation eras was less (Table 7).

Table 6.  Independent effects of transplant era and other variables on the rate of decline in function
VariablesRegression coefficients for slopes of eGFR*
Measured
after 1 month
Measured
after 6 months
Measured
after 12 months
  1. *Multiple linear regression coefficients (and 95% confidence intervals) for the slopes (least-squares-fitted regression lines for eGFR versus time after transplantation in each patient) of eGFR in mL/min/1.73 m2. Coefficients in each column are additive, so that the rate of change in eGFR for an ‘average’ patient would be the sum of all applicable coefficients for that patient, including the baseline in the first row and the regression intercept in the last row.

  2. eGFR = glomerular filtration rate estimated using the 4-variable Modification of Diet in Renal Disease formula (5); CKD = chronic kidney disease; GN = glomerulonephritis; MHC = major histocompatibility complex; BMI = body mass index.

Decline in eGFR for each 10 mL/ min/1.73 m2) in baseline eGFR−6.8 (−7.2–−6.4),−5.0 (−5.2–−4.9),−2.3 (−2.4–−2.2),
 p < 0.0001 p < 0.0001 p < 0.0001
Transplant in 1990–94 vs. 1984–892.3 (0.5–4.1),1.2 (0.4–2.0),0.6 (0.2–1.0),
 p = 0.0106 p = 0.0039 p = 0.0028
Transplant in 1995–98 vs. 1984–896.8 (4.8–8.7),2.7 (1.8–3.6),0.7 (0.2–1.1),
 p < 0.0001 p < 0.0001 p = 0.0024
Transplant in 1999–02 vs. 1984–8910.1 (8.0–12.1),4.3 (3.4–5.2),1.7 (1.2–2.2),
 p < 0.0001 p < 0.0001 p < 0.0001
Transplant in center 1 vs. 3−3.5 (−6.2–−0.8),−0.1 (−1.2–1.1),−0.4 (−1.0–0.2),
 p = 0.0098 p = 0.9245 p = 0.2439
Transplant in center 2 vs. 31.7 (−1.3–4.7),−0.8 (−2.1–0.5),−0.8 (−1.4–−0.1),
 p = 0.2674 p = 0.2474 p = 0.0211
Transplant in center 4 vs. 31.8 (−0.4–3.9)1.9 (0.9–2.9),0.6 (0.1–1.1),
 p = 0.1128 p = 0.0002 p = 0.0197
Transplant in center 5 vs. 3−0.6 (−3.4–2.1),0.9 (−0.3–2.1),0.1 (−0.5–0.8),
 p = 0.6623 p = 0.1303 p = 0.6391
Deceased vs. living donor−3.4 (−4.9–−1.9),−2.3 (−3.0–−1.6),−1.1 (−1.4–−0.8),
 p < 0.0001 p < 0.0001 p < 0.0001
Black vs. Caucasian1.7 (−0.3–3.7),0.9 (0.0–1.8),−0.4 (−0.9–0.1),
 p = 0.1025 p = 0.0457 p = 0.1052
Asian vs. Caucasian8.2 (3.2–13.3),0.6 (−1.6–2.9),1.1 (0.0–2.2),
 p = 0.0014 p = 0.5908 p = 0.0537
Donor black vs. Caucasian−2.1 (−4.6–0.4),−1.6 (−2.7–−0.4),−0.8 (−1.4–−0.2),
 p = 0.1029 p = 0.0071 p = 0.0065
CKD type 1 diabetes vs. GN1.2 (−0.6–3.0),2.1 (1.3–2.9),0.8 (0.4–1.2),
 p = 0.2002 p < 0.0001 p < 0.0001
Each MHC mismatch−0.7 (−1.1–−0.3),−0.5 (−0.7–−0.3),−0.3 (−0.4–−0.2),
 p = 0.0009 p < 0.0001 p < 0.0001
Age 0–29 years vs. 30–44 years0.8 (−1.1–2.6),−0.5 (−1.3–0.4),−0.5 (−1.0–−0.1),
 p = 0.4013 p = 0.2664 p = 0.0105
Age 45–59 years vs. 30–44 years0.0 (−1.5–1.5),0.6 (0.0–1.3),0.5 (0.2–0.9),
 p = 0.9595 p = 0.0647 p = 0.0034
Age ≥60 years vs. 30–44 years2.5 (0.1–4.8),0.5 (−0.5–1.6),0.7 (0.2–1.3),
 p = 0.0389 p = 0.3423 p = 0.0064
Donor age 0–29 years vs. 30–44 yearsp < 0.0001p < 0.0001p < 0.0001
 p < 0.0001 p < 0.0001 p < 0.0001
Donor age 45–59 years vs. 30–44 years−5.6 (−7.3–−3.8),−3.6 (−4.4–−2.8),−1.7 (−2.1–−1.3),
 p < 0.0001 p < 0.0001 p < 0.0001
Donor age ≥60 years vs. 30–44 years−10.1(−13.3–−6.8),−7.4 (−8.9–−6.0),−3.6 (−4.3–−2.8),
 p < 0.0001 p < 0.0001 p < 0.0001
Donor male vs. female0.9 (−0.4–2.3),1.3 (0.7–1.9),0.5 (0.2–0.8),
 p = 0.1752 p < 0.0001 p = 0.0012
BMI <18.5 vs. 18.5–29.9 kg/m23.3 (−0.1–6.8),3.1 (1.6–4.7),1.4 (0.7–2.2),
 p = 0.0580 p < 0.0001 p = 0.0003
BMI ≥30.0 vs. 18.5–29.9 kg/m2−1.4 (−3.0–0.2),−1.7 (−2.4–−1.0),−0.8 (−1.1–−0.4),
 p = 0.0903 p < 0.0001 p < 0.0001
Intercept (mL/min/1.73 m2)31.5 (28.3–34.7),24.1 (22.8–25.5),11.0 (10.3–11.7),
 p < 0.0001 p < 0.0001 p < 0.0001
Table 7.  The independent effects of transplant era on the rate of decline in function
VariablesRegression coefficients for slopes of eGFR*:
Measured
after 1 month
Measured
after 6 months
Measured
after 12 months
  1. *Multiple linear regression coefficients (and 95% confidence intervals) for the slopes (least-squares-fitted regression lines for eGFR versus time after transplantation in each patient) of eGFR in mL/min/1.73 m2.

  2. eGFR = glomerular filtration rate estimated using the 4-variable Modification of Diet in Renal Disease formula (5).

Not adjusted for other variables
Transplant during 1990–1994 vs. 1984–19891.2 (−0.6–3.0)0.4 (−0.4–1.2)0.3 (−0.1–0.7)
 p = 0.1818 p = 0.2911 p = 0.1901
Transplant during 1995–1998 vs. 1984–19894.8 (2.9–6.6)1.4 (0.5–2.2)0.1 (−0.3–0.5)
 p < 0.0001 p = 0.0013 p = 0.6362
Transplant during 1999–2002 vs. 1984–19897.8 (6.0–9.7)2.7 (1.9–3.6)0.9 (0.5–1.4)
 p < 0.0001 p < 0.0001 p < 0.0001
Adjusted for pre-transplant variables (as in Table 2)
Transplant during 1990–1994 vs. 1984–19892.3 (0.5–4.1)1.2 (0.4–2.0)0.6 (0.2–1.0)
 p = 0.0106 p = 0.0039 p = 0.0028
Transplant during 1995–1998 vs. 1984–19896.8 (4.8–8.7)2.7 (1.8–3.6)0.7 (0.2–1.1)
 p < 0.0001 p < 0.0001 p = 0.0024
Transplant during 1999–2002 vs. 1984–198910.1 (8.0–12.1)4.3 (3.4–5.2)1.7 (1.2–2.2)
 p < 0.0001 p < 0.0001 p < 0.0001
Adjusted for pre-transplant variables + initial immunosuppressive agents
Transplant during 1990–1994 vs. 1984–19892.2 (0.4–4.1)0.8 (0.0–1.7)0.6 (0.2–1.0)
 p = 0.0196 p = 0.0581 p = 0.0034
Transplant during 1995–1998 vs. 1984–19894.0 (1.1–6.9)1.6 (0.3–2.9)0.6 (0.0–1.3)
 p = 0.0074 p = 0.0172 p = 0.0564
Transplant during 1999–2002 vs. 1984–19894.0 (0.3–7.8)1.2 (−0.5–2.9)1.2 (0.3–2.1)
 p = 0.0359 p = 0.1720 p = 0.0065

We examined whether the differences in slopes across the different eras were due to differences in some, but not all, of the five centers. We did this by including center × era interaction terms in multiple linear regression analyses. Although changes in slopes were different in some centers, there was still an effect of the transplant era that was independent of changes in slope in different eras in the individual centers (data not shown). We also limited the analysis of slopes to cases with at least three eGFR determinations, and the results were similar (data not shown). Finally, we examined whether the results were unique to eGFR calculated using the MDRD equation. In multiple linear regression analyses that included all of the variables shown in Table 6, the effects of transplantation era on slopes of eCCr or inverse serum creatinines, measured after 1, 6 and 12 months, were similar to those found for eGFR (data not shown).

Discussion

With low rates of acute rejection and early allograft failure, it has become difficult to measure progress in kidney transplantation. Although there is a relatively constant rate of late graft failure, the number of failures is small enough to require very long-term follow-up to measure differences attributable to therapies or other factors. By relying on the fact that the rate of late failures tends to be log-linear, it is tempting to calculate graft half-lives, and thereby project long-term outcomes. However, the reliability of this technique has recently been questioned (3).

Until new surrogate markers for late graft failure are defined, kidney function remains the best candidate for predicting outcomes. Although kidney function at a fixed time after transplantation is a strong predictor of outcomes, it is not as predictive as measuring decline in function over time (7). Kidney function at different times after transplantation appears to be stable (Table 4), but this may be deceptive when patients with graft failure drop out in each time interval. Changes in eGFR in short intervals can be examined, but again patient attrition makes comparisons between intervals over time problematic (Table 4). In contrast, assuming that declines in eGFR are linear, measuring the rate of decline over time as slopes of least squares-fitted regression lines assign a unique value to each patient and is therefore appealing. However, eGFR may not always decline in a linear fashion (Table 4) (8). An alternative method, treating a particular threshold of chronic decline in function as an event that can occur at any time after baseline function is established, may be a better way to predict outcomes than either measuring function at a fixed time post-transplant or measuring slopes (7).

In the present study, both slopes of eGFR over time and the occurrence of a 25% decline in eGFR (after baseline eGFR was established during the first 2 months post-transplant) indicate that long-term graft function is more stable in patients transplanted recently compared to patients transplanted several years ago. In particular, between 1984 and 1989 and 1999 and 2002, decline in kidney function measured as slope of eGFR lessened. This improvement is evident even after adjusting for multiple pre-transplantation characteristics (Table 6). Interestingly, after adjusting for differences in immunosuppressive medications used after transplantation, the difference in the rate of declining kidney function in different transplant eras was less (Table 7). This suggests, but does not prove that recent improvements in the rate of decline in function may be due to new immunosuppressive medication regimens or other associated treatment strategies.

In a recent study by Gourishankar and colleagues, the authors concluded that for deceased donor kidney recipients the rate of change of creatinine clearance beyond the first 6 months after transplantation improved between 1990 and 2000 (4). They required a minimum of five serum creatinine measurements to calculate slope, and as a result only patients who survived at least 18 months were included in their analysis. Therefore, their results are most comparable to the slopes measured after 12 months in the current study that included only patients surviving at least 24 months (since at least two creatinine clearance determinations ≥12 months were required to calculate a slope). In both studies, the rate of decline in function was less in the most recent era. It should be noted that the 429 patients studied by Gourishankar and colleagues were also included in the present study.

Examining the rate of decline in creatinine clearance (slopes) may be more useful for describing what has happened than predicting what is going to happen (7). It is not surprising that patients who returned to dialysis or required retransplantation had a greater rate of decline in eGFR than patients who had a functioning kidney at last follow-up. However, it is noteworthy that patients who died with a functioning graft had lower baseline function, but did not have progressively declining function.

We have previously reported that a pre-defined threshold of a decline in kidney function is a better predictor of graft failure than declining function measured by slope (7). In the present study, the best clinical correlate of graft failure was a decline in eGFR of 25%. A decline in eGFR of 25% was more strongly associated with graft failure than baseline eGFR, or the occurrence of either early or late acute rejection episodes. The relative risk of graft failure associated with a 25% decline in eGFR was 3.7, while the relative risk associated with death-censored graft failure was 7.9.

There was a substantial amount of inter-patient variability in the rate of declining kidney function measured by slopes of eGFR. Indeed, the standard deviations of the slopes were several-fold greater than the slopes themselves (Table 5). This variability was somewhat less for slopes measured after 6 months, and less still for slopes measured after 12 months (Table 5). We did not examine intra-patient variability in the rate of decline, or the pattern of declining kidney function. However a previous, small study found that this too is quite variable after kidney transplantation (8). It is possible that a statistical method for taking into account changes in of the rate of decline in eGFR over time would more accurately reflect declining function after transplantation than the simple linear model used here. Whether the best model for describing declines in function after transplantation is a single straight line, a combination of two hinged lines, or some other non-linear approach is not known and deserves additional investigation.

There were relatively few clinical features associated with declines in kidney function measured as slopes (Table 6). This may be due to the large amount of variability in the decline in function, and it is possible that a more accurate measure of kidney function would reveal a greater number of clinical correlates to declining function. We did not find that estimating creatinine clearance with the Cockcroft-Gault formula, or simply using inverse serum creatinine to measure slopes provided any advantages over eGFR calculated by the MDRD formula (data not shown). In any case, a search for factors associated with inter-patient differences in the rates of the decline in kidney function may ultimately yield important clues into the pathogenesis of graft failure.

The rate of decline in function was similar among the five centers (Table 6). The rate of decline was greater at Center 1 versus Center 3 measured after the first month. The decline was also slightly greater measured after 12 months in Center 2, but it tended to be better after both 6 and 12 months in Center 4. Perhaps, more interesting is the lack of difference in the decline in function between the centers in this study, despite their very different approaches to long-term immunosuppression. For example, at Fairview University Medical Center the tradition has been to maintain adequate blood levels of calcineurin inhibitors in the late post-transplantation period (9), while at Hennepin County Medical Center the majority of patients were withdrawn from cyclosporine after the first year (10). At the University of Manitoba, protocol biopsies have been used routinely to detect and treat subclinical acute rejection (11), while this has not been the practice at other centers. Despite these differences in management strategies, the rates of declining kidney function appeared to be more alike than different.

Not surprisingly, the clinical correlates of declining kidney function (Table 6) are similar to those of death-censored graft failure (1). For example, older recipients, who may have better adherence to medications and less acute rejection, had less decline in kidney function (Table 6). Kidneys from older donors, on the other hand, had a greater decline in function, as did patients with deceased donor kidneys and patients with more major histocompatibility mismatches.

There are a number of important limitations to this study. Serum creatinines were measured at regular, albeit relatively infrequent intervals. This wide interval in the measurement of creatinine limited our ability to detect rapid changes in function. In addition, serum creatinine is a relatively crude measure of glomerular filtration rate. In this study, we used the MDRD formula to calculate eGFR with serum creatinines, but the Crockcroft-Gault formula for estimating creatinine clearance yielded very similar results (data not shown). Although still other formulas could reduce inter-patient variability and improve the correlation between serum creatinine and true glomerular filtration rate, these formulas will not likely overcome the fact that factors other than kidney function can affect serum creatinine levels. Differences between laboratories in calibration of serum creatinine could affect the results of this study. However changes in eGFR should be less affected by differences in laboratories (assuming that methodologies did not change over time) than comparisons of eGFR between centers. Ultimately, better measures of glomerular filtration rate may improve our understanding of kidney function after kidney transplantation.

In summary, by pooling data derived from a large number of patients at five transplant centers, we demonstrated significant improvement in the rate of decline in long-term kidney allograft function in recent years. The rate of decline of −0.9 ± 35.5 mL/min/1.73 m2/year, −1.9 ± 21.1 mL/min/1.73 m2/year and −1.4 ± 10.9 mL/min/1.73 m2/year, measured after 1, 6 and 12 months, respectively, may be slightly greater than the expected −1.0 mL/min/year decline in creatinine clearance attributed to aging in normal individuals (12). These rates of decline appear to be less than the −2.2 mL/min/1.73 m2/year decline measured in African Americans with hypertensive nephropathy (13), and substantially less than the −3.8 to −10 mL/min/1.73 m2/year decline measured in patients with non-proteinuric and proteinuric kidney diseases (14–16). These findings are encouraging, and may indicate that further improvement in stability of long-term kidney function and allograft survival is achievable.

Ancillary