Journal Logo

Original Clinical Science—General

Late Graft Loss After Kidney Transplantation: Is “Death With Function” Really Death With a Functioning Allograft?

Gaston, Robert S. MD1; Fieberg, Ann MS2; Helgeson, Erika S. PhD2; Eversull, Jason MD1; Hunsicker, Lawrence MD3; Kasiske, Bertram L. MD4; Leduc, Robert PhD2; Rush, David MD5; Matas, Arthur J. MD6; for the DeKAF Investigators*

Author Information
doi: 10.1097/TP.0000000000002961



With development of new immunosuppressive agents and major advances in clinical care, short-term outcomes after kidney transplantation are substantially better.1,2 Today, the primary challenge is to improve long-term outcomes. One problem limiting advances is uncertainty regarding the causes and clinical course of late graft failure, characterized as death (or death with function, DWF) and death-censored graft failure (DC-GF). Traditionally, each has accounted for approximately half of late losses.3 Improved graft survival in recent decades is almost entirely because of less DC-GF, primarily in the first year post-transplant.4 There has not been a similar reduction in DWF, at least in part due to expanding candidacy among an aging, more chronically ill population.5 Regardless, rates of DWF from 6 months to 10 years post-transplant are relatively unchanged and remain a major factor contributing to static long-term outcomes.

An important consideration is whether DC-GF and DWF represent differing or related phenotypes. The most common causes of DC-GF are immunologic injury and recurrent disease, with a myriad of less common causes (eg, BK virus nephropathy) also playing a role.6 In contrast, DWF is most often attributed to cardiovascular disease, malignancy, and infection.7 Yet, in both chronic kidney disease and kidney transplantation, declining GFR is clearly linked with increased mortality risk.8,9 Does DWF as an outcome merely reflect declining graft function? Are DWF and DC-GF different manifestations of a similar process, or distinct clinical outcomes that affect different populations, reflect the impact of different risk factors, and ultimately will require different approaches to ameliorate? In other words, is “DWF” really death with allograft function? In this report, utilizing the unique resource generated by the Long-term Deterioration of Kidney Allograft Function study (DeKAF), we sought to answer these questions in the current era.


The DeKAF project originated as a National Institutes of Health–funded prospective, multicenter, observational investigation of late allograft failure, conducted at 7 transplant centers in the United States and Canada.10 Since 2013, its continuation has been made possible via funding from a consortium of industry sources. A detailed description of DeKAF can be found at (NCT00270712). Institutional Review Board approval was obtained at all participating sites, with informed consent obtained from each patient at the time of enrollment. All kidney or kidney-pancreas transplant recipients undergoing transplantation at 7 participating centers between October 1, 2005 and April 20, 2011 were eligible for enrollment in the prospective cohort. After obtaining informed consent, donor and recipient demographics were obtained. Early events (acute rejection [AR] and delayed graft function [DGF]) were recorded, and each patient established a “baseline” serum creatinine (Cr) level: the mean of 3 consecutive Cr values obtained on or about day 90. Follow-up Cr values were collected every 6 months to a year and whenever any “event” (eg, rejection, infection) occurred. Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation.11 An index biopsy (IBx) was defined before study initiation as the first for-cause biopsy occurring >90 days post-transplant and precipitated by (1) an unexplained and persistent ≥ 25% increase of serum Cr over baseline (irrespective of the interval of time between baseline and the 25% increase and in the absence of potential confounding factors) or (2) new-onset proteinuria, defined as an albumin/Cr ratio ≥ 0.2 or a protein/Cr ratio >0.5. Proteinuria was defined as any spot urine protein/Cr ≥0.5 or any spot urine albumin/Cr ≥ 0.25. All information was recorded in a centralized database.

The current study includes only those patients with a functional allograft at 90 days post-transplant. DC-GF was defined as a documented return to dialysis or retransplantation. DWF was defined as death that occurred in recipients not undergoing retransplantation or return to dialysis. The reported primary causes of each death and graft failure were independently classified by 3 DeKAF investigators utilizing any information provided by the center (including documented narratives, death certificates, and hospital discharge notes). Any disagreements were resolved by a fourth investigator.

Separate univariable and multivariable survival models, stratified by clinical center, were used to assess factors associated with DC-GF and DWF. Time zero for this analysis was the date baseline serum Cr level was established (usually day 90 post-transplant). For the model with DWF as an outcome, individuals were censored at the date of return to dialysis or retransplantation. The variables examined in the analyses were recipient age, race, history of diabetes, history of hypertension history of cardiovascular disease (excluding hypertension), donor age, donor type (living/deceased), years on dialysis pretransplant, DGF, AR before day 90, and eGFR. Cox proportional hazards models were used to estimate the hazard ratio (HR) and associated 95% confidence interval (CI) for the univariable models. Because of intermittent creatinine measurements, the association between eGFR and outcome was evaluated using a joint model which combined a linear mixed effect model for eGFR trajectory and a Cox model for censored survival outcomes. Specifically, eGFR trajectory was modeled using a longitudinal linear mixed-effects model with subject-specific intercept and slope with unstructured covariance. For the univariable longitudinal model, only time was included as a covariate. In the multivariable longitudinal model, all previously mentioned variables were included as covariates. The joint survival process was modeled using a time-dependent relative risk model with piecewise constant baseline risk function as previously described.12

Descriptive statistics were used to examine characteristics of patients who experienced DC-GF, DWF, or who did not experience either graft failure or death during the course of the study, denoted “maintained function” (MF). Means were reported with their SD and medians were reported with interquartile ranges. Categorical variables were compared between groups using χ2 tests, and continuous measures were compared using ANOVA tests. The eGFR trajectory for each of the 3 groups was illustrated using locally estimated scatterplot smoothing (loess) curves. All statistical analyses were performed by A.F., R.L., and E.S.H. (University of Minnesota) using SAS 9.4 and R 3.2.3.


There were 3587 patients >18 years of age who enrolled prospectively at the time of transplant, survived at least 90 days post-transplant with a functioning allograft, and had a baseline Cr established around day 90. Mean (SD) follow-up was 5.2 (1.8) years, with a median (interquartile range) of 5.2 (3.8–6.6) and ranging out to 8.1 years. DC-GF was experienced by 295 (8.2%) recipients; 350 (9.8%) died with a functioning allograft (DWF); and 2942 (82.0%) were alive with a functioning kidney transplant at last follow-up (MF).

Different risk factors evaluated at time of transplantation were identified for DC-GF in comparison to DWF (Table 1). In the multivariable model, DC-GF was significantly associated with younger age at transplant (HR, 0.95 for each additional year of age; 95% CI, 0.94-0.96; P < 0.001), African American race (HR, 1.94; 95% CI, 1.37-2.76, P < 0.001), cardiovascular disease (HR, 1.47; 95% CI, 1.05-2.06, P = 0.024), and hypertension (HR, 1.54; 95% CI, 1.00-2.38; P = 0.048). In contrast, DWF was associated with older age at transplant (HR, 1.04; 95% CI, 1.03-1.05, P < 0.001), longer time on dialysis before transplant (HR, 1.06; 95% CI, 1.03-1.09; P < 0.001), preexisting diabetes (HR, 1.50; 95% CI, 1.19-1.88; P < 0.001), cardiovascular disease (HR, 1.62; 95% CI, 1.28-2.04; P < 0.001), and having received a kidney from a deceased donor (HR, 1.35; 95% CI, 1.06-1.74, P = 0.017). Likewise, compared with those who MF during the study, those with DWF were older, more likely to have a deceased donor, had spent longer time on dialysis, and were more likely to have diabetes or cardiovascular disease (Table 2, P < 0.001 for all comparisons).

Multivariable analyses of risk factors for DC-GF and DWF
Characteristics of patients at the time of transplantation, by group

Early events (<90 days) after transplantation also differed for DC-GF and DWF (Table 1). Not surprisingly, DC-GF was significantly associated with at least one early episode of AR (HR, 1.50; 95% CI, 1.03-2.18; P = 0.035), whereas the association between early AR and DWF was not statistically significant (HR, 1.20; 95% CI, 0.82-1.77; P = 0.354). Table 3 compares the proportion of patients experiencing AR and DGF early (<90 days) post-transplant. Early AR occurred less frequently in the DWF and MF groups (≈8%) than in the DC-GF group (16%) (P < 0.001). In contrast, the proportion of patients who experienced DGF in the DWF and DC-GF groups (16%) was higher than in the MF group (7%) (P < 0.001).

Events occurring early (<90 d) and at baseline (90 d) post-transplant, by group

Events occurring beyond 90 days post-transplant are outlined in Table 4. Median time to DC-GF was 2.9 years, and to DWF was 3.3 years. Proteinuria was substantially more common among DC-GF patients (72%) than among either MF (25%) or DWF patients (36%). Only 15% in the MF group underwent IBx after day 90 for episodic graft dysfunction, as compared to 23% in the DWF group and 68% in the DC-GF group. Additionally, the histologic findings reported in the index biopsies differed substantially by group (Table 5). Notably, AR (either cellular- or antibody-mediated) was reported in 48.3% of those with DC-GF, versus, 29.9% of the MF group and 18.5% of the DWF group.

Clinical events occurring beyond 90 d post-transplant, by group
Primary diagnosis at index biopsy (local pathology, by group)

Although worsening renal function was associated with both DWF and DC-GF (Table 1) the HR for a 1 unit change in eGFR was much closer to 1.0 for DWF compared with DC-GF (HR, 0.99; 95% CI, 0.98-0.99 versus HR, 0.86; 95% CI, 0.84-0.87). Retrospectively comparing eGFR at day 90 (Table 3) between groups, we find similar average values between those that died with function and those that MF for the duration of the study (56 versus 58 mL/min/1.73 m2), which was higher than the average eGFR value in those that experienced graft failure (52 mL/min/1.73 m2). The median last available eGFR values (Table 4) were also similar between those that MF (54.9 mL/min/1.73 m2) and those that died during the study (44.6 mL/min/1.73 m2), and much higher than the median eGFR value for those that experienced DC-GF (11.2 mL/min/1.73 m2). Graphically examining the eGFR trends reveals a clear decline in eGFR before DC-GF that was not seen among those who MF or died during the study (Figure 1A–C). Individuals with graft loss experienced a sharp decline in function approximately 7 months before graft failure, whereas the slopes of eGFR were relatively stable in the MF group and DWF group.

Line graphs of estimated glomerular filtration rate (eGFR) over time from 90 days post-transplant until event (death-censored graft failure [DC-GF] or death with function [DWF]) or last available measurement (maintained function [MF]), with overlaid loess smoothed curve, by group. A, DC-GF. B, DWF. C, MF.

Causes of DC-GF and DWF are shown in Table 6. DC-GF was most commonly attributed to nonadherence (21%) and rejection (20%), with only 11.5% unknown. DWF was most often attributed to malignancy (15%), infection (14%), and cardiac events (13%). A cause of death could not be ascertained in 151 (43.1%) of subjects who died. The slope of eGFR before death between those with known and unknown causes was not significantly different (−1.80 mL/min/1.73 m2/y in the known cause group; −2.83 mL/min/1.73 m2/y in the unknown-cause group, P = 0.23), indicating the clinical course of the unknown-cause patients was reasonably represented in the larger group (Figure S1, SDC,

Causes of death-censored graft failure and death with function


Improved outcomes in kidney transplantation in recent years largely reflect the impact of reduced early graft failure, with little change in subsequent rates of long- term graft loss.1,2 The DeKAF study was established to better define the variables responsible for this long-term attrition and to create a framework for developing new approaches to prevent it. Although the major focus of DeKAF is differentiating causes of DC-GF, our multicenter data, derived in the current era, offer important insights into all events impacting late graft failure. In the current era, DWF and DC-GF continue to each account for approximately half of late graft losses, with clear differences in baseline demographics and clinical course in recipients having DWF compared with DC-GF. These data indicate that improving long-term kidney transplant survival beyond current standards will require different approaches to deal with each outcome.

Given that in both the general population and kidney transplant recipients, elevated Cr levels—even a mild elevation—are associated with increased mortality, an important question has been whether DWF is a manifestation of graft dysfunction or is, in itself, a different entity.8,9 West et al13 noted, in patients transplanted a quarter-century ago, that the majority of patients dying with DWF had serum Cr levels <2 mg/100 mL in the year preceding death. However, in a more recent analysis of data from the Patient Outcomes in Renal Transplantation study, Kasiske et al14 reported that reduced 12-month eGFR was associated with increased HR for subsequent mortality. Lorenz et al7 similarly showed a correlation between lower eGFR at 1 year and increased mortality risk, though ultimately concluding, “…the vast majority of patients experiencing DWF have well-preserved allograft function before death.” Our data, in a substantially larger, more diverse, and current, cohort confirm this observation. In a registry analysis from 2000, Ojo et al15 commented “whether patients in the study actually died with graft function as opposed to death because of impairment of graft function is open to question.” The DeKAF study, while not refuting a relationship between compromised eGFR and mortality, answers that open question: for most patients, “death with function” really is death with a functioning allograft.

Understanding and comparing transplant outcomes is complex. Early in the history of clinical transplantation, recipient outcome data were presented simply as patient survival and graft survival. In 1996, West et al13 demonstrated that there were different risk factors for DWF and DC-GF, suggesting that these distinctions be considered when analyzing transplant outcomes. A quarter-century later, it remains concerning that DWF is so often a part of important studies and yet so poorly characterized. Lorenz et al,7 in a single-center study, were unable to identify cause of death in one-third of their cases. Though ascertaining causes of death was not the focus of DeKAF design, we were unable to identify causes in 43.1% of cases in this prospective study. Neither centers nor registries reliably collect these data. This may reflect that most deaths occur distant from transplant centers, under the care of other physicians and hospitals, sometimes at home. Likewise, this deficiency is neither limited to transplantation nor easily remediable.16,17 Although net and relative survival methods could be used to examine if survival is lower than would be expected from an appropriate reference population, not fully understanding causes of death in patients with DWF poses a significant impediment in understanding how to prevent them.

Our analysis clearly indicates that DWF and DC-GF represent distinct phenotypes. The current data demonstrate that pretransplant characteristics (eg, age, dialysis vintage, and donor source) exert a substantially greater impact on risk of DWF than early posttransplant events, including DGF, AR, or renal function at 90 days. Although we find a univariate association between DGF and DWF, after adjusting for pretransplant characteristics, this association dissipates, indicating the relationship between DGF and DWF may be confounded by pretransplant characteristics. Compared with those who did well, patients destined to experience DWF demonstrated near-identical baseline renal function, absence of episodic dysfunction and proteinuria, and slope of eGFR over time, indicating that mortality risk is largely independent of allograft function. Lorenz et al7 have similarly shown that increasing recipient age, diabetes, and prior dialysis were risk factors for DWF; in their analysis, including <25% deceased donors, donor source was not significant. Additionally, among the 24% of patients that died with function who had undergone allograft biopsy within 1-year of death, 68% had normal or mild histologic changes comparable to those who MF.

We have previously reported that, in DeKAF, DC-GF was associated with both early posttransplant events (early rejection) and, episodic dysfunction after 3 months, with greater long-term impact of the latter.18 A unique aspect of the DeKAF Prospective Cohort is data collected around an “Index biopsy,” performed in response to development of new-onset proteinuria or a 25% elevation in serum Cr level over baseline occurring >90 days post-transplant. In the current report, not only did patients with DC-GF undergo more episodes necessitating biopsy than those who died or MF but also IBx diagnoses were predominantly immunologic (AR, cell- or antibody-mediated) in patients with DC-GF, differing substantially from findings in the other 2 groups (Table 5).

There are limitations to this analysis. Defining study groups based on outcomes poses statistical challenges in analyzing the buildup to those events. With lengthier follow-up, those who MF at last encounter might transition to a different group. Furthermore, within each group, there is variability in clinical courses, including changes in eGFR over time. However, we found the outcomes and groupings consistent across the application of several different statistical approaches and the mean duration of follow-up (>5 y) to be clinically meaningful. Additionally, though there is substantial interindividual variability, including change points in eGFR, the volume of subjects studied, and the size and consistency of the differences among groups support the validity of the observations.

If DWF is a distinct long-term outcome in kidney transplant recipients, there are at least 3 important clinical implications. First, improving long-term outcomes will require differing approaches to prevent DWF and DC-GF. With pretransplant characteristics driving DWF, better outcomes can only come from different selection criteria, improved care of recipients with risk factors for DWF, or both. Our data indicating a deceased donor source as a risk factor for DWF could influence allocation algorithms. However, maintaining balance between efficacy and efficiency, with data indicating benefits of transplantation over dialysis extend even to candidates with high mortality risk, may limit opportunities for improvement.19 Nonetheless, added emphasis on living donation for candidates at greatest risk may be warranted. Recent data on patient care are encouraging. Lam et al5 analyzed outcomes of patients transplanted in Ontario between 1994 and 2009 and divided into 4-year cohorts. During that interval, there was an increase in the length of time on dialysis, in the number of recipients at least 65 years old (from 5.7 to 17%), in pretransplant diabetes (from 21% to 29%), and in pretransplant coronary artery disease (from 22% to 37%). When adjusting for preexisting risk, there was a 30% reduction in death or major cardiovascular events within 3 years post-transplant, indicating vastly improved patient care.

Second, these phenotypes likely need to be considered in developing inclusion and exclusion criteria for clinical trials. Although one might argue that the traditional Food and Drug Administration composite endpoint for efficacy failure (AR, graft failure, death, and loss to follow-up) captures the overall impact of a novel therapeutic agent on outcomes in the patient population at risk, the differing phenotypes defined by our data indicate that potential benefits (on graft or patient survival) may be obscured.20 The importance of this definition is illustrated in the Belatacept Evaluation of Nephroprotection and Efficacy as First Line Immunosuppression Trial (BENEFIT) of belatacept, where in the first posttransplant year biopsy-proven acute rejection occurred with almost twice the frequency as in a cyclosporine-based control group despite comparable rates of graft failure and death in both low-intensity belatacept-treated patients and controls.21 Subsequently, there was little additional rejection: at 3 years, overall efficacy remained comparable. By 7 years, however, there was demonstrable superiority among belatacept-treated patients.22,23 Although belatacept-treated patients suffered more rejection, there were fewer graft failures and deaths, with less DWF a substantial component of better outcomes. Would this trial have produced the same results in cohorts more homogeneous for pretransplant risk of either immunologic graft failure or DWF? Might a differential impact on outcomes have been evident sooner, or not at all, in better-defined populations?

Finally, these data may offer support for reexamining endpoints in performance metrics for individual centers. It is well documented that attempts to respond to scorecards with poor outcomes may inappropriately alter selection criteria.24,25 Although the Scientific Registry of Transplant Recipients segregates DC-GF and DWF in the examination of long-term kidney transplant outcomes, outcomes metrics in program-specific reports combine all deaths within a reporting period regardless of whether they were preceded by graft failure. Given the strong impact of pretransplant characteristics (patient selection) on risk of DWF and posttransplant events (patient management) on DC-GF, it may be of greater benefit to report the timing of death relative to graft failure for center self-evaluation and quality improvement.

In summary, DWF after kidney transplantation really is death with a functioning kidney, with relatively stable eGFR and a phenotype distinctly different from patients who lose their grafts for other reasons (DC-GF). These differing clinical phenotypes have important implications for patient selection, posttransplant management, clinical trial design, and choice of outcomes metrics.


We thank Stephanie Taylor for help in preparation and submission of the article.


J. Michael Cecka: UCLA Immunogenetics Laboratory

Fernando Cosio: Mayo Clinic (Rochester)

Robert S. Gaston: University of Alabama at Birmingham

Sita Gourishankar: University of Alberta (Edmonton)

Joseph Grande: Mayo Clinic (Rochester)

Lawrence Hunsicker: University of Iowa (Iowa City)

Bertram L. Kasiske: Hennepin County Medical Center (Minneapolis)

Roslyn B. Mannon: University of Alabama at Birmingham

Arthur J. Matas: University of Minnesota (Minneapolis)

David Rush: University of Manitoba (Winnipeg)


1. Lamb KE, Lodhi S, Meier-Kriesche HU. Long-term renal allograft survival in the United States: a critical reappraisal. Am J Transplant. 2011; 113450–462
2. Hart A, Smith JM, Skeans MA, et al. OPTN/SRTR 2017 Annual Data Report: Kidney. Am J Transplant. 2019; 19 Suppl 2:19–123
3. Pascual M, Theruvath T, Kawai T, et al. Strategies to improve long-term outcomes after renal transplantation. N Engl J Med. 2002; 3468580–590
4. Hart A, Smith JM, Skeans MA, et al. OPTN/SRTR 2016 annual data report: kidney. Am J Transplant. 2018; 18Suppl 118–113
5. Lam NN, Kim SJ, Knoll GA, et al. The risk of cardiovascular disease is not increasing over time despite aging and higher comorbidity burden of kidney transplant recipients. Transplantation. 2017; 1013588–596
6. El-Zoghby ZM, Stegall MD, Lager DJ, et al. Identifying specific causes of kidney allograft loss. Am J Transplant. 2009; 93527–535
7. Lorenz EC, El-Zoghby ZM, Amer H, et al. Kidney allograft function and histology in recipients dying with a functioning graft. Am J Transplant. 2014; 1471612–1618
8. Meier-Kriesche HU, Baliga R, Kaplan B. Decreased renal function is a strong risk factor for cardiovascular death after renal transplantation. Transplantation. 2003; 7581291–1295
9. Go AS, Chertow GM, Fan D, et al. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med. 2004; 351131296–1305
10. Gourishankar S, Leduc R, Connett J, et al. Pathological and clinical characterization of the ‘troubled transplant’: data from the dekaf study. Am J Transplant. 2010; 102324–330
11. Levey AS, Stevens LA, Schmid CH, et al.; CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration)A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009; 1509604–612
12. Rizopoulos DD. JM An R package for the joint modelling of longitudinal and time- to-event data. J Stat Soft. 2010; 3591–33
13. West M, Sutherland DE, Matas AJ. Kidney transplant recipients who die with functioning grafts: serum creatinine level and cause of death. Transplantation. 1996; 6271029–1030
14. Kasiske BL, Israni AK, Snyder JJ, et al.; Patient Outcomes in Renal Transplantation (PORT) InvestigatorsThe relationship between kidney function and long-term graft survival after kidney transplant. Am J Kidney Dis. 2011; 573466–475
15. Ojo AO, Hanson JA, Wolfe RA, et al. Long-term survival in renal transplant recipients with graft function. Kidney Int. 2000; 571307–313
16. Yin D, Morris CR, Bates JH, et al. Effect of misclassified underlying cause of death on survival estimates of colon and rectal cancer. J Natl Cancer Inst. 2011; 103141130–1133
17. Olubowale OT, Safford MM, Brown TM, et al. Comparison of expert adjudicated coronary heart disease and cardiovascular disease mortality with the National Death Index: results from the Reasons for Geographic and Racial Differences in Stroke (REGARDS) Study. J Am Heart Assoc. 2017; 65e004966
18. Gaston RS, Fieberg A, Hunsicker L, et al. Late graft failure after kidney transplantation as the consequence of late versus early events. Am J Transplant. 2018; 1851158–1167
19. Gill JS, Schaeffner E, Chadban S, et al. Quantification of the early risk of death in elderly kidney transplant recipients. Am J Transplant. 2013; 132427–432
20. Burckart GJ, Amur S, Goodsaid FM, et al. Qualification of biomarkers for drug development in organ transplantation. Am J Transplant. 2008; 82267–270
21. Vincenti F, Charpentier B, Vanrenterghem Y, et al. A phase III study of Belatacept-Based Immunosuppression Regimens Versus Cyclosporine in Renal Transplant Recipients (BENEFIT Study). Am J Transplant. 2010; 103535–546
22. Vincenti F, Larsen CP, Alberu J, et al. Three-year outcomes from BENEFIT, a randomized, active-controlled, parallel-group study in adult kidney transplant recipients. Am J Transplant. 2012; 121210–217
23. Vincenti F, Rostaing L, Grinyo J, et al. Belatacept and long-term outcomes in kidney transplantation. N Engl J Med. 2016; 3744333–343
24. Schold JD, Srinivas TR, Howard RJ, et al. The association of candidate mortality rates with kidney transplant outcomes and center performance evaluations. Transplantation. 2008; 8511–6
25. Schold JD, Buccini LD, Srinivas TR, et al. The association of center performance evaluations and kidney transplant volume in the United States. Am J Transplant. 2013; 13167–75

Supplemental Digital Content

Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.