Trends in the Causes of Death among Kidney Transplant Recipients in the United States (1996–2014)

In 2015, there were 124,114 newly reported cases of end-stage kidney disease (ESKD) in the United States [1]. Kidney transplantation offers the best survival for selected patients who develop ESKD. Although the number of deceased donor kidney transplants increased from 12,279 in 2015 to 13,501 in 2016 [2], there is substantial organ shortage. As of December 21, 2015, approximately 84,000 patients on dialysis were on the waiting list for a transplant, and 8% of the patients listed in 2013 died before they could receive a transplant [1]. Hence, it is vital to maximize graft and patient survival in patients who have undergone kidney transplantation.

Death with graft function (DWGF), defined as death of a recipient without prior graft failure, re-transplantation, or return to dialysis, leads to the simultaneous loss of the recipient’s life and the functioning kidney. Identifying cause of death patterns in these patients will help develop strategies to improve patient and graft survival by focusing resources on specific causes. Available evidence on this subject is limited to single-center studies, or outdated data from patients who received older induction and maintenance transplant medications [3-5]. It is unclear if factors leading to mortality have changed in the last 2 decades, especially with the introduction of newer technologies and immunosuppressive agents. Mortality and contributions from various causes could have been affected by a myriad of factors over the years. According to the annual data report of Scientific Registry of Transplant Recipients, 18.4% of kidney transplant recipients (KTRs) in 2016 were ≥65 years of age compared to around 10% in 2004. Older patients and those with diabetes are at higher risk of cardiovascular disease and infections. A significant proportion of KTRs have pre-transplant dialysis durations exceeding 5 years. In addition to change in patient characteristics, there were significant changes in induction and maintenance protocols to minimize acute rejection episodes over the past 2 decades. The use of T-cell depleting antibody for induction has increased from ∼50% in 2005 to 70% in 2016 [2]. Compared to interleukin-2 receptor antibodies, T-cell depleting antibodies are associated more with opportunistic infections [6] and malignancies.

Therefore, we hypothesized that the 1- and 10-year mortality rates would have declined, whereas the causes of death could have changed over time. To address these hypotheses, we examined the 1- and 10-year mortality rates in KTRs with functioning graft in the United States over almost 2 decades of observation. In addition, we studied the causes of death post-transplantation and examined the trends in 1- and 10-year cause-specific mortality among KTRs.

We conducted the present study using the United States Renal Data System (USRDS), which is the national registry of patients with ESKD and includes almost all patients requiring renal replacement therapy – either dialysis or a kidney transplant. We restricted the study cohort to adult (age ≥18) first-time KTRs who received their allograft between January 1, 1996 and December 31, 2014. Patients who received kidney and pancreas transplantation, any solid organ transplantation or bone marrow transplantation prior to the kidney transplantation were excluded.

Death events reported in the USRDS were obtained from several sources, including the Centers for Medicare and Medicaid Services (CMS) Medicare Enrollment Database, CMS forms 2746 to 2728, Organ Procurement and Transplantation Network (OPTN) transplant follow-up forms, CROWNWeb database, Social Security Death Master File, and inpatient claims. For the USRDS ESKD patient cohort, the cause of death is determined from the Death Notification form (CMS-2746). We classified the cause of death into categories according to the primary cause of death documented in the USRDS “patients” file: cardiovascular, infectious, malignancy and others (online suppl. Table 1; for all online suppl. material, see www.karger.com/doi/10.1159/000495081). A large proportion of deaths had a primary cause of death reported as unknown or this information completely missing in the “patients” file. We classified them as separate categories.

We used the calendar year of transplantation as exposure. Demographics including age, sex, race, and health-related characteristics of recipients such as hypertension, diabetes, and several other comorbidities were abstracted from the “patients” and Medicare Evidence Report files in the USRDS. Demographic characteristics of donors were obtained from the transplant files in USRDS along with additional information from OPTN.

We described the characteristics of recipients at the time of transplantation. Means and SDs or medians and interquartile ranges were used for continuous variables and absolute numbers and percentages were used for categorical variables. Recipients were followed for death event with a functioning graft. We studied mortality in both the short term (1-year) and long-term (10-year) duration. In the 1-year mortality analysis, patients were censored when they lost graft function or at the end of 1-year after the first transplantation, whichever was earlier. We divided the number of death-events by the number of recipients at baseline to calculate the proportion of deaths in the 1-year period. In the 10-year mortality analysis, we limited patient population to KTRs between January 1, 1996 and December 31, 2005 to allow 10-year follow-up time for patients in each transplantation year. They were censored when they lost graft function or at the end of 10 years after the initial transplantation, whichever came first.

We described the proportion of deaths among recipients by donor type (deceased, living and unknown), cause of ESKD (diabetes, hypertension, glomerulonephritis and others), and recipient age (< 65 vs. ≥65 years). Proportions of each category of the cause of death were described among death events. We estimated the cumulative incidence of all-cause mortality during the first year after transplantation, then cumulative incidence of each cause-specific death, considering death from other causes as competing events. We estimated the hazard ratio (HR) of all-cause death according to transplantation year with year 1996 as the referent (HR of 1) from the Cox proportional hazard model. Then, we applied the cause-specific hazard regression model to estimate the cause-specific HR for each cause of death, considering death due to other causes as competing events. HRs were adjusted for age (18–29, 30–49, 50–69, ≥70), sex, and race (White, Black, Asian, other). To test for the linear trend of all-cause and cause-specific mortality over time, we included the calendar year as a continuous variable in the adjusted model. We repeated the analyses described above for 10-year mortality data as well.

Due to concern that the use of anti-lymphocyte polyclonal antibodies (i.e., depleting agents) may increase the risk of malignancy and thus mortality, we stratified the 1- and 10-year mortality analyses by the use of depleting agents. Patients who received induction by anti-lymphocyte globulin, anti-thymocyte globulin, and Moromunab (OKT3) were included in the “depleting agents” group, while patients who did not receive these agents were included in the “non-depleting agents” group. As death due to malignant disease in the first year after kidney transplantation is a rare event, we collapsed transplantation years into three 5-year periods, using 1996–2001 as a reference group to evaluate 1 year mortality, stratified by depleting agents.

The study was approved by an institutional review board at Baylor College of Medicine (protocol #H-36408), and an active Data Use Agreement with the National Institutes of Diabetes and Digestive and Kidney Diseases (NIDDK) was in place. All analyses were conducted using SAS version 9.4 (SAS Institute, Inc. Cary, NC, USA).

Of 429,717 patients who were recorded to have received a kidney transplant in the USRDS, 156,231 patients were excluded; their first kidney transplantation occurred outside the study period (January 1, 1996 to December 30, 2014; Fig. 1). A total of 210,327 KTRs who had transplantation from 1996 to 2014 were included in the study of 1-year mortality and 94,384 (transplanted between 1996 and 2005) were studied to assess 10-year mortality.

Table 1 describes the baseline characteristics of the study population. The mean age of KTRs was 50.5 years (SD 13.4), with the majority being Caucasian (65.6%) and male (60.7%). Diabetes was the leading reported cause of ESKD (27% of patients), followed by glomerulonephritis (25%) and hypertension (21.2%). The narrower sample of KTRs from 1996 to 2005 that was studied for 10-year mortality is also described in detail in Table 1.

Of the 210,327 patients, 6,774 (3.2%) died within 1 year after receiving a kidney transplant; 756 (11.2%) patients’ cause of death was reported as unknown, while 1,959 (28.9%) patients cause of death was missing (Table 2). Among KTRs with a known cause of death, 1 quarter died from cardiovascular causes (24.7%), followed by infectious (15.2%) and malignant causes (2.9%). Other known causes constitute 17.1% of the known causes of death.

To assess the impact of etiology of ESKD on the mortality rate and cause of death, we subdivided the KTR population according to the leading causes of ESKD (Table 2). Of note, 5% of the patients who had ESKD secondary to diabetes died in the first-year post-transplant, with the majority (29.9%) of deaths secondary to cardiovascular causes.

Figure 2 shows the trends in 1-year all-cause and cause-specific mortality for the patients who received kidney transplantation from 1996 to 2014. All-cause mortality decreased over the years, from a mortality of 4.1% in 1996 to 2.5% in 2014 (online suppl. Table 2). The results also show a decline in mortality due to cardiovascular and infectious causes. However, mortality due to malignancy did not decrease.

Ten-year mortality data is presented in Table 3. Almost 2 thirds (61.5%) of patients had an unknown or missing cause of death. Death from cardiovascular causes constituted the majority (14.6%) of known causes, followed by infectious (8.2%) and malignant causes (5.4%). Other known causes were reported for 10.3% of deaths.

Figure 3 examines the trend of 10-year all-cause and cause-specific mortality for the transplantation years 1996–2005. Ten-year all-cause mortality was 28.1% for transplantation year 1996, while it was 27.0% for transplantation year 2005 (online suppl. Table 3). Ten-year cardiovascular mortality showed a downtrend since 2002. By contrast, 10-year mortality due to malignant causes was higher (1.86%) for transplantation year 2005 than it was for year 1996 (1.53%).

Cox-proportional hazard models showed a decline over time in adjusted 1- and 10-year all-cause mortality as shown in Tables 4 and 5. To formally consider competing events, we used cause-specific hazard models. Table 4 shows the model for 1-year cause-specific death. Using 1996 as the referent year, the HRs for CV and infectious mortality have steadily and significantly (p value for trend < 0.001) declined and the respective HRs were 0.38 (CI 0.28–0.51) and 0.25 (CI 0.17–0.38) in 2014, respectively. However, the HR for malignant causes remained comparable (albeit with wide confidence limits) during the same period with an HR value of 0.86 (CI 0.33–2.23) for 2014. The cause-specific hazard model for 10-year cause-specific death showed a similar pattern (Table 5). A significant linear trend was observed for 1-year mortality due to all causes, cardiovascular, infections and other causes (p < 0.0001); and for 10-year mortality due to all causes, cardiovascular and infections causes (p < 0.0001).

When stratified according to the use of depleting agents, no significant interaction between time and depleting drug use was observed. As expected, we observed more use of depleting drugs in induction among kidney transplant patients over time, 15.5% in 1996, 38.0% in 2005 and 50.7% in 2014 (Data not shown). When including the interaction term of transplant year and use of depleting drugs in the model, we did not find significant interaction between them for 1-year (p = 0.550) and 10-year (p = 0.142) malignant cause of death. One- and ten-year mortality decreased due to cardiovascular and infectious causes, but not due to malignant causes, in both, depleting as well as non-depleting agent groups (online suppl. Table 4, 5).

From this comprehensive analysis of national data spanning almost 2 decades, we conclude that the mortality of first-time KTR with a functioning allograft declined considerably over the study period. Cardiovascular disease remained the leading cause of death followed by death due to infections and malignancies. These results were consistent, both in analyses focusing on short-term and in those focusing on long-term follow-up, as well as among both younger patients and those older than 65 years. Our findings extend previous observations conducted in other studies that used considerably older data and had relatively smaller sample sizes [3, 7, 8]. Ojo et al. [3] examined the UNOS/Scientific Registry of Transplant Recipient and USRDS databases to study DWGF among patients who received their first kidney transplantation from 1988 to 1997. This study included 86,502 patients and showed an overall 31% reduction in DWGF among patients who received a kidney transplant from 1993 to 1997, compared to patients who were transplanted from 1988 to 1992. El Husseini et al. [7] studied patient and graft survival in first-time KTRs in UNOS database from 2000 to 2014. However, their objective was to compare survival in this group to repeat kidney transplants and kidney-after-nonrenal solid organ transplant, and they studied death with a functioning graft as well as death with graft loss. As of date, our study is the most comprehensive that has analyzed DWGF among more than 200,000 primary KTRs in the USRDS database over the last 2 decades.

The preponderance of death due to cardiovascular disease is not surprising, as KTRs have multiple cardiac risk factors, including diabetes and hypertension pre- and post-transplantation, resulting in cumulative burden of heart and vascular disease over time. Albeit good allograft function improves uremic milieu, most traditional risk factors of atherogenesis still persist after successful transplantation. However, compared to a KTR who underwent kidney transplant in 1996, the 1- and 10-year cardiovascular death rates have improved in recent years, not much different from the general population. The decline among KTRs may be due to overall improvement in cardiac care, or be related to evolving immunosuppressive regimens over last 2 decades [9, 10]. Cyclosporine has been largely replaced by tacrolimus, and steroid minimization strategies are common. Compared to tacrolimus, cyclosporine causes more hypertension and lipid abnormalities [11, 12], but tacrolimus causes islet cell damage and post-transplant diabetes [13, 14]. The use of steroids in the post-transplant period is associated with a dose-dependent increase in mortality [15]. Steroid minimization strategies reduce the risk of post-transplant diabetes [16], although there is not enough evidence to suggest that this plays a significant role in the reduction of cardiovascular mortality. Also, reduction in acute rejection episodes minimizes the need for high-dose pulse steroids.

Similar to cardiac disease, short- and long-term infection-related death rates have also declined. This observation is encouraging as attempts to reduce acute rejection rates may lead to a parallel increase in mortality related to opportunistic infections and malignancies. With the introduction of newer potent transplant medications, acute rejection rates within the first-year post-transplant have declined significantly to the lowest level of around 8% in 2014 recipients [17]. The use of T-cell depleting antibody for induction has increased from ∼50% in 2005 to around 70% in 2015. Also, cyclosporine and azathioprine have been replaced by tacrolimus and mycophenolate respectively [2]. At the same time, the proportion of patients aged 65 years and older has significantly increased among transplant recipients. When compared to young-age patients, elderly patients are more prone to infection-related complications [18]. The observed reduction in infection-related deaths could be due to reduced exposure to T-cell depleting antibodies and pulse-dose steroids as acute rejection rate continues to improve. It could also be secondary to increasing vigilance and aggressive treatment of infections as well as reduction in invasive viral infections [19]. We acknowledge that we cannot deduce from this data that the incidence of infections has reduced in KTRs, mainly due to censoring for graft failure and partially due to missing causes of death. In fact, the follow-up of FAVORIT trial actually demonstrated that infections are the leading cause of death in the late post-transplant period (4–10 years after transplant), especially in diabetic patients [20].

Unlike cardiovascular and infection-related mortality, 1- and 10-year mortality due to malignancy has not decreased over time. These findings are consistent with studies from other countries that have examined mortality trends in KTRs [21, 22]. Lack of decline in mortality due to malignancy could be multifactorial, including but not limited to aging patient population, prolonged exposure to immunosuppressive medications due to longer graft survival and reduced number of deaths from cardiovascular and infectious causes. In another analysis of the USRDS data (1990–2004), patients who were < 50 years had a higher standardized mortality ratio due to higher risk of succumbing to their malignancy. On the other hand, competing risks of other causes of death in older patient populations reduce their risk of dying from cancer [23]. However, our study did not show a reduced mortality in older patient population and in fact, people above 65 years of age had a higher mortality due to cancer. In our sensitivity analyses, mortality due to malignant causes did not decline over the years in patients who received depleting agents during induction, as opposed to mortality secondary to cardiovascular and infectious causes. Further studies are required to better investigate the relationship between use of depleting agents and mortality due to malignant causes.

Being a population-based study with few selection criteria, its results are broadly generalizable to KTRs in the United States. USRDS data are drawn from a multitude of sources, including CMS, OPTN, and the ESRD networks. As it analyzes the mortality over the last 20 years, the results reliably reflect the impact of our current practices in transplant, including pre-operative evaluations and contemporary immunosuppression regimens. Our study carries the inherent limitations that are associated with studies involving large national datasets and registries [24]. Missingness of key data is one of them, and according to data obtained from USRDS registry, the cause of death is either missing or unknown in about 40% for patients who die within the first year with a functioning graft. Despite a low overall death rate of 3.2% in this population, delineating a precise cause of death may help develop strategies to reduce immediate post-transplant mortality. During 10-year follow-up, when KTRs are less frequently followed in a transplant center, the information on the cause of death is missing or unknown in an even higher proportion of patients.

Although alarming, a high fraction of missing or unknown cause of death was also shown by previous studies involving national registries [3, 7, 23]. Missing data does not seem to have affected the contribution of various causes of death, as our results are generally consistent with other studies of mortality where cardiovascular disease, infection and malignancy are the 3 leading causes of death in patients with KTRs [3, 7]. A large retrospective study in an integrated health system in the United States (with no missing data) had similar conclusion regarding the causes of mortality [8]. Studies done outside the United States, including Spain [21], Australia and New Zealand (ANZDATA) [22], and India [25] have shown very similar patterns.

In summary, cardiovascular mortality remains the leading cause of death in KTRs at 1 and 10 years post-transplant and surpasses infection and malignancy at all times. Although mortality due to cardiovascular and infectious causes has been trending down, there is still a substantial proportion of transplant recipients, especially patients over 65 years of age with a deceased donor, who remain at increased risk of death due to these 2 factors. Remarkably, the mortality due to malignancy has essentially remained unchanged and may become the most significant cause of death with a functioning graft as the transplant population continues to age and the cumulative exposure to immunosuppressive medications continues to rise. Moreover, we are losing a large proportion of KTRs without adequately delineating the factors that lead to mortality in this high-risk patient population. The current study not only gives a unique insight into the various causes of death in patients with kidney transplant, but it also highlights the significant gaps that currently exist in our knowledge of transplant mortality. This void needs to be filled through improved reporting and collection of mortality data in KTRs.

This work was conducted under a data use agreement between Dr. Winkelmayer and the NIDDKs. An NIDDK officer reviewed this manuscript for research compliance and approved of its submission for publication. Data reported herein were supplied by the USRDS. Interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as official policy or interpretation of the US government.

The results presented in this paper have not been published previously in whole or part, except in the abstract format – presented as a poster at the 2017 Kidney Week held in New Orleans, LA on November 3, 2017.

The authors disclose no conflicts of interest relevant to this manuscript. The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government.

This work was supported in part by Career Development Award #IK2 BX002912 from the United States (U.S.) Department of Veterans Affairs Biomedical Laboratory Research and Development Program to JP.