Reduced survival and quality of life following return to dialysis after transplant failure: the Dialysis Outcomes and Practice Patterns Study

Abstract

Background

Although dialysis after kidney transplant failure (TF) is common, the outcomes of these patients remain unclear. We compared outcomes of TF patients with transplant-naïve (TN) patients wait-listed for kidney transplantation.

Methods

We used data from the Dialysis Outcomes and Practice Patterns Study (DOPPS), including laboratory markers and health-related quality of life (HR-QOL). Mortality and hospitalization of participants with one prior TF versus TN patients were compared using the Cox regression analysis. HR-QOL physical and mental component summary scores (PCS and MCS) were examined using linear mixed models, and clinical practices were compared using logistic regression.

Results

Compared with TN patients (n = 2806), TF patients (n = 1856) were younger (48 versus 51 years, P = 0.003), less likely to be diabetic (18 versus 27%, P < 0.0001) and to use a permanent surgical vascular access {adjusted odds ratio (AOR): 0.85 [95% confidence interval (CI): 0.70–1.03], P = 0.10}, particularly within the first 3 months after TF [AOR 0.45 (0.32–0.62), P < 0.0001]. TF patients also had lower PCS [mean difference −2.56 (−3.36, −1.75), P < 0.0001] but not MCS [−0.42 (−1.34, 0.50), P = 0.37]. All-cause mortality [adjusted hazard ratio (AHR): 1.32 (95% CI: 1.05–1.66), P = 0.02], especially infection-related [AHR 2.45 (95% CI: 1.36–4.41), P = 0.01], was higher among TF patients.

Conclusions

TF patients have reduced QOL and higher mortality, particularly due to infections, than TN patients. Interventions to optimize care before and after starting dialysis remain to be identified and applied in clinical practice.

Introduction

Compared with chronic dialysis, kidney transplantation offers longer life expectancy and improved quality of life (QOL), physical functioning and vocational abilities [1, 2]. Advances in kidney transplantation have translated into greater improvements in short-term kidney allograft survival relative to long-term graft survival [3]. Therefore, many patients will experience kidney transplant failure (TF) and will require initiation of dialysis. In the USA, return to dialysis after kidney TF represents the cause of dialysis initiation in 4.1% of incident dialysis patients, and 16% of patients wait-listed for kidney transplantation have a history of kidney TF [4]. As the kidney transplantation rates in developed countries increase, and with a fixed duration of graft survival, the absolute numbers of patients returning to dialysis after kidney TF are expected to increase.

A better understanding of the outcomes of TF patients is necessary. In North American registries, high mortality rates have been described in patients returning to dialysis after TF [5–7]. When the survival of these patients is compared with those with ongoing graft function, the annual adjusted death rate is 3-fold greater among TF patients [8].

Less clear are the outcomes of TF patients when compared with transplant-naïve (TN) patients initiating dialysis for the first time. Patients initiating dialysis after TF are a selected group whom at one point were placed on a transplant waiting list and received a kidney transplant. Restriction of the comparator TN group only to those patients wait-listed for kidney transplantation provides the opportunity to minimize selection bias by the restriction of the comparator group to a similar transplant-eligible population. Our primary objective was to evaluate the impact of kidney TF on mortality and hospitalization in an international cohort of patients enrolled in the Dialysis Outcomes and Practice Patterns Study (DOPPS). Secondary objectives were to determine the association between a TF and infection-related adverse outcomes, achievement of performance targets on dialysis therapy and health-related QOL (HR-QOL).

Materials and methods

Data source

This study used data from DOPPS 1 (1996–2001), 2 (2002–04) and 3 (2005–08). Adults (≥18 years of age) receiving long-term in-center hemodialysis (HD) were randomly selected from all participating facilities across all DOPPS phases [308 dialysis facilities in DOPPS 1 (n = 17 034), 322 dialysis facilities in DOPPS 2 (n = 12 839) and 300 dialysis facilities in DOPPS 3 (n = 11 361)]. Patients in DOPPS 2 and 3 were enrolled from the same countries as in DOPPS 1 with the addition of Australia, Belgium, Canada, New Zealand and Sweden (Appendix 1). The DOPPS sampling plan and study methods have been previously published [9]. Institutional review boards approved the DOPPS and informed patient consent was obtained in accordance with local requirements.

Study population

Figure 1 demonstrates the derivation of the study cohort. Patients with a history of greater than one kidney transplant, and those in whom the time from TF to DOPPS enrollment was missing were excluded. The comparator group consisted of TN HD patients who were on a waiting list for kidney transplantation at DOPPS enrollment.

Fig. 1.

Assembly of the study cohort.

Outcomes

The primary outcomes were mortality and hospitalization events due to (i) any cause, (ii) cardiovascular disease and (iii) infection. Definitions for cause-specific mortality and hospitalization are shown in Appendix 2. Hospital admissions, diagnoses and major procedures were recorded during study follow-up.

Secondary outcomes were time to first infectious complication defined as time to first of either infection-related hospitalization or infection-related death, achievements of clinical practice recommendations and HR-QOL. Achievements of clinical practice targets were based on accepted practice guidelines over the course of the study and included (i) use of arteriovenous fistula (AVF) or graft (versus catheter), (ii) hemoglobin 11–13 g/dL (versus else), (iii) albumin >4.0 g/dL (versus ≤4.0 g/dL), (iv) Kt/V >1.2 (versus ≤1.2), (v) phosphorus >5.5 mg/dL (versus ≤5.5 mg/dL) and (vi) PTH >500 pg/mL (versus ≤500 pg/mL). HR-QOL was measured at study enrollment with the SF-36 Health Survey using standard scoring procedures [10, 11]. The SF-36 measures eight separate scales of HR-QOL: physical functioning, role physical, bodily pain, general health, mental health, role emotional, social functioning and vitality. The two general summary scales were also computed: the physical component summary (PCS) and mental component summary (MCS) [12]. We evaluated three scales of patient health-related concerns by using the KDQOL-SF [13]: (i) symptoms/problems, (ii) effects of kidney disease on daily life and (iii) burden of kidney disease. Scales were scored from 0 to 100 points, with higher scores representing better HR-QOL. Depressive symptoms were assessed at study enrollment by the short, 10-item version of the Center for Epidemiological Studies Depression Screening Index (CES-D). Each response item is scored from 0 to 3 points. A summary CES-D score (0–30 points) is derived, with higher scores indicating greater depressive symptoms. The cut-off value of ≥10 was used for the summary CES-D score as an indicator of possible clinical depression [14]. Physician-diagnosed depression within the past 12 months was obtained from the baseline medical questionnaire.

Covariates

Demographic, comorbidity, laboratory and vascular access-related data were collected at the time of study entry. Dialysis vintage was the time since TF for TF patients and the time since first ever dialysis for TN patients. Thirteen summarised comorbid conditions are described in Appendix 3. Laboratory values included hemoglobin, serum albumin, calcium, phosphorus, parathyroid hormone (PTH) and ferritin. Three types of vascular access were AVF, arteriovenous graft (AVG) and central venous catheter (CVC). Education was classified as the highest level of education received: less than high school, high school or above high school.

Statistical analysis

Descriptive statistics evaluated differences in baseline characteristics between TN and TF patients. TF patients were stratified into subgroups based on time since kidney TF (defined as <3, 3–12 and >12 months from kidney TF to DOPPS enrollment). For each of the outcomes (mortality, hospitalizations, time to first infection, achievement of clinical practices and HR-QOL), we tested for overall differences between TF and TN patients. For HR-QOL, we also assessed trends across TF patient subgroups.

The associations of transplantation status (TF versus TN) with all-cause, infection-related and cardiovascular-related mortality, time to first infectious complication and time to first hospitalization were examined using the Cox proportional hazards regression. Time at risk began at study entry. For hospitalization, follow-up was censored at the earliest time point: death, departure from study, kidney transplantation or change in dialysis modality. For mortality, follow-up was censored at the earliest time point: 7 days after departure from the study, or change in renal replacement modality. Models were adjusted for demographic information, body mass index (BMI), vintage, 13 summarised comorbid conditions, serum albumin and catheter use; stratified by country and study phase; and used the sandwich covariance estimator to control for clustering by facility. The proportional hazards assumption was checked graphically and using time-by-covariate interactions. When non-proportional hazards were found, stratification by the corresponding covariate was performed as a sensitivity analysis. To better characterize the impact of wait-list status on outcomes after kidney TF, we performed an additional sensitivity analysis in which hazard ratios for all-cause mortality and hospitalization among TF patients were estimated separately by wait-list status for repeat kidney transplantation at study enrollment.

Logistic regression, using generalized estimating equations to adjust for clustering by facility, was used to identify associations between transplantation status and achievement of clinical practice guidelines [15–17]. Models were adjusted for demographic information, 13 summarised comorbid conditions, serum albumin (except among analyses pertaining to serum albumin as the outcome variable), vascular access (except among analyses pertaining to vascular access type as the outcome variable), study phase and country. Linear mixed models with the same adjustments, with random effects for facility, were used to (i) assess the differences in HR-QOL and depression measures between TN and TF patients, and (ii) test for trends across three TF patient subgroups.

Statistical analyses were performed using SAS software, version 9.2 (SAS Institute, Cary, NC). The STROBE Statement guidelines were followed for reporting observational studies [18].

Results

Baseline patient characteristics

Among 41 234 DOPPS participants, 4.5% (n = 1856) had a history of a first kidney TF, while 6.8% (n = 2806) had no prior TF and were wait-listed for kidney transplantation at the time of DOPPS enrollment, 25% (n = 466) of TF patients were wait-listed for repeat transplantation, while 37.5% (n = 696) did not have information on waiting list status. Table 1 lists the distribution of patient characteristics for TN wait-listed patients, all TF patients and TF patients stratified by the time since TF. TF patients tended to be younger, have a lower BMI and of greater dialysis vintage (all P < 0.05). TF patients had a lower prevalence of diabetes, but a higher prevalence of congestive heart failure, cancer, psychiatric disorders and recurrent cellulitis/gastrointestinal bleed (all P < 0.05). TF patients were less likely to use an AVF (P = 0.06) or AVG (P = 0.01) as HD vascular access and more likely to use a CVC (P = 0.0003). There was a lower mean hemoglobin and serum albumin among TF patients, but higher PTH levels and serum ferritin relative to TN patients (all P < 0.05).

Table 1.

Patient characteristics at study enrollment

	TN patients (n = 2806)	TF patients (n = 1856)	P-valuea	TF patients			P-valueb
				<3 months (n = 313)	3–12 months (n = 299)	>12 months (n = 1244)
Age (years, mean ± SD)	51.1	48.3	0.0032	47.4	49.9	48.1	0.22
Male (%)	61.6	61	0.66	64.2	59.2	60.7	0.58
Black (%)	15.1	10.6	0.04	15.3	9.7	9.6	0.90
Weight (kg)	73.5	66.5	<0.0001	71.2	68.3	65.0	<0.0001
BMI (kg/m2, mean ± SD)	25.4	23.5	<0.0001	24.2	23.7	23.3	0.0030
Vintage (years, mean ± SD)c	3.6	4.2	<0.0001	–	–	–	–
Comorbidities (%)
Coronary artery disease	39.5	33.1	0.69	27.9	36.2	33.7	0.55
Congestive heart failure	21.5	28.6	<0.0001	28.9	32.9	27.5	0.64
Other cardiac disease	21.9	33.1	<0.0001	26.6	35.7	34.0	0.11
Diabetes	26.9	17.5	<0.0001	26.0	24.2	13.7	<0.0001
Hypertension	80.6	79.5	0.51	83.2	84.8	77.4	0.03
Cerebrovascular disease	7.8	8.4	0.71	9.4	9.5	8.0	0.51
Peripheral vascular disease	15.0	15.8	0.06	15.9	15.9	15.7	0.95
Cancer	6.0	7.7	0.0035	5.5	7.1	8.4	0.19
Lung disease	6.2	6.7	0.35	5.2	6.7	7.1	0.17
Gastrointestinal bleed	3.9	4.7	0.08	6.2	6.1	4.0	0.07
Neurological disease	7.4	8.9	0.004	5.8	8.1	9.8	0.02
Psychiatric disorder	14.9	21.5	<0.0001	22.4	22.6	21.1	0.61
Recurrent cellulitis/gangrene	4.1	7.0	0.0002	7.8	5.1	7.3	0.44
Vascular access (%)
Arteriovenous fistula	70.1	63.9	0.06	52.0	60.9	67.7	0.0030
Graft	15.2	18.5	0.01	14.8	15.8	20.1	<0.0001
Catheter	14.7	17.6	0.0003	33.2	23.2	12.2	<0.0001
Laboratory values (mean ± SD)
Hemoglobin (g/dL)	11.6	11.1	<0.0001	9.8	11.2	11.3	<0.0001
Albumin (g/dL)	3.9	3.7	<0.0001	3.4	3.7	3.8	<0.0001
Calcium (mg/dL)	9.4	9.5	0.35	9.3	9.4	9.6	0.0001
Phosphorus (mg/dL)	5.8	5.8	0.25	5.7	5.9	5.7	0.33
PTH (pg/mL)	325	371	0.03	478	333	360	0.04
Ferritin (ng/mL)	423	426	0.01	325	428	445	0.01

Model adjusted for country and study phase and accounted for facility clustering.

^aTest for difference between TF and TN patients in adjusted models.

^bTest for trend across TF patient subgroups in adjusted models.

^cTime since first ever dialysis from TN patients, and time since TF for TF patients.

BMI, body mass index; SD, standard deviation; PTH, parathyroid hormone.

Prevalence of TF and TN patients by country

Figure 2 shows the prevalence of patients with a history of TF among (i) all patients and (ii) patients wait-listed for kidney transplant. The prevalence of prior TF ranges from 13.3% in Spain to 1.1% in Japan, with 4.3% in the USA. Among wait-listed patients, the prevalence of prior TF ranges from 31.9% in France to 4.6% in Japan, with 11.8% in the USA.

Fig. 2.

Percentage of patients by country with a history of transplant failure and wait-listed for kidney transplantation at DOPPS enrollment (Phases 1–3). SP, Spain; UK, United Kingdom; SW, Sweden; FR, France; ANZ, Australia–New Zealand; CA, Canada; GE, Germany; BE, Belgium; USA, United States; IT, Italy; JP, Japan.

Mortality and hospitalizations

Among TN patients, 225 deaths occurred over a median of 1.4 years. After a median of 1.69 years, 246 TF patients died. Among TN patients, 1139 had a hospitalization after a median of 0.66 years; among TF patients, there were 1004 hospitalizations after a median of 0.58 years of follow-up. After a median of 1.2 years, 833 TN patients had a kidney transplant, while 233 TF patients had a repeat kidney transplant after a median of 1.5 years. Elevated mortality hazards were found for TF patients (versus TN patients) with and without adjusting for key covariates (Figure 3). Compared with TN patients, the adjusted hazards for TF patients were 32% higher for all-cause mortality, 53% higher for cardiovascular-related mortality and 145% higher for infection-related mortality (all P < 0.05). In exploring differences in infection-related mortality between TF and TN patients, using a forward-selection process in the multivariable model, albumin had the highest impact on the unadjusted or crude hazards ratio (HR) followed by dialysis vintage. In contrast, adjustment for vascular access type did not significantly alter the crude HR. Yet, within the fully adjusted final model, compared with AVF/AVG use, CVC use was independently associated with a 131% increased risk of infection-related mortality. We found elevated hazards of hospitalization events for TF versus TN patients (Figure 4). In sensitivity analyses comparing wait-listed TF and TN patients, we observed an increased trend in all-cause hospitalization in the wait-listed TF group [adjusted hazard ratio (AHR) 1.14, 95% confidence interval (CI) = 0.99–1.33, P = 0.07], but no difference in all-cause mortality (AHR 0.93, 95% CI = 0.62–1.39, P = 0.7).

Fig. 3.

Mortality: TF versus TN. All models stratified by country and study phase, and accounted for facility clustering. *Model adjusted for age, sex, race, BMI, time since initiation of HD or TX failure, 13 summary comorbid conditions, albumin and catheter use.

Fig. 4.

Time to first hospitalization: TF versus TN. All models stratified by country and study phase, and accounted for facility clustering. *Model adjusted for age, sex, race, BMI, time since initiation of HD or TX failure, 13 summary comorbid conditions, albumin and catheter use.

Restricted to TF patients, we evaluated the impact of the duration of renal allograft function and the duration of dialysis prior to renal allograft on mortality and hospitalization due to any causes (model adjusted for key covariates). Neither measure was associated with mortality or hospitalization due to any causes (all P > 0.15).

Infection-related complications

The most common causes of infection-related hospitalizations included septicemia (n = 114) and pneumonia (n = 79), while for infection-related mortality, septicemia (n = 25) and infections related to gangrene (n = 16) were most common. TF patients experienced a greater rate of infection-related complications (either hospitalization or death; Figure 5), a difference that was apparent soon after dialysis initiation and persisted after multivariable adjustment (AHR 1.45, 95% CI = [1.16–1.80], P = 0.001).

Fig. 5.

Infection-related hospitalization or death: TF versus TN. All models stratified by country and study phase, and accounted for facility clustering. *Model adjusted for age, sex, race, BMI, time since initiation of HD or TX failure, 13 summary comorbid conditions, albumin and catheter use.

HR-QOL scores

HR-QOL scores of TN and TF patients are presented in Table 2. The adjusted difference (AD) was greater than 3 points (positive AD for lower score in TF patients) for physical functioning, role physical, general health and bodily pain scales (P < 0.05 for all), and considered clinically significant [19]. The AD was 2.5 points for PCS (P < 0.0001). For mental scales, social functioning (AD = 5.2, P = 0.02) and vitality (AD = 6.7, P = 0.0002) were lower among TF patients. TF patients had lower scores for health-related symptoms/problems (AD = 3.0, P = 0.02) and higher prevalence of physician-diagnosed depression [adjusted odds ratio (AOR) = 1.42, P = 0.003].

Table 2.

QOL score and depression symptoms

	TN patients (n = 2806)	TF patients (n = 1856)	P-valuea	TF patients			P-valueb
				<3 months (n = 313)	3–12 months (n = 299)	>12 months (n = 1244)
Physical component summary	39.6	37.1	<0.0001	36.4	36.5	37.4	0.81
Physical functioning	54.2	47.3	<0.0001	46.2	47.7	47.4	0.22
Role physical	42.3	34.7	0.0007	25.9	30.8	37.5	0.03
General health	46.9	42.4	0.0005	41.0	40.4	43.1	0.88
Bodily pain	66.3	60.8	0.0001	59.1	59.4	61.4	0.93
Mental component summary	46.5	44.8	0.51	43.5	43.9	45.3	0.34
Mental health	65.1	61.5	0.36	59.2	61.5	62.0	0.38
Role emotional	59.6	55.8	0.74	49.5	54.2	57.6	0.49
Social functioning	65.8	60.6	0.02	60.6	56.1	61.6	0.50
Vitality	45.5	38.8	0.0002	33.7	38.5	40.0	0.20
Burden	40.4	38.2	0.68	40.4	35.9	38.3	0.0022
Effects	61.5	57.3	0.06	60.0	55.8	57.1	0.0004
Symptoms	75.3	72.3	0.02	71.0	70.8	72.9	0.77
CES-D ≥ 10 (%)	36.7	41.7	0.11	38.5	49.5	40.6	0.80
Depression (%)	9.1	14.0	0.0027	13.1	15.5	13.9	0.12

Models adjusted for age, sex, race, BMI, 13 comorbidities, albumin, country and study phase and accounted for facility level clustering.

^aTest for difference between TF and TN patients in adjusted models.

^bTest for trend across TF patient subgroups in adjusted models.

QOL, quality of life; TN, transplant naïve; TF, transplant failure; CES-D, Centers for Epidemiologic Studies Depression scale.

We examined differences in HR-QOL scores across TF patient subgroups stratified by the time since TF (<3, 3–12 and >12 months). Our results revealed a trend toward the lowest scores among patients enrolled into DOPPS within 3 months from TF, with scores improving over time among those enrolled into DOPPS within 3–12 months after TF and further improving among those with a history of TF but enrolled within DOPPS at least 12 months after TF. The trend was attenuated after adjusting for patient characteristics, country and study phase (P < 0.05 for role physical, burden and effects).

Achievement of clinical practice guidelines

Compared with TN patients, TF patients were less likely to have serum albumin >4.0 g/dL (AOR = 0.67, 95% CI = 0.56–0.80, P = 0.0001) and more likely to have Kt/V >1.2 (AOR = 1.39, 95% CI = 1.10–1.75, P = 0.01) and PTH >500 pg/mL (AOR = 1.45, 95% CI = 1.20–1.74, P = 0.0001). TF patients were less likely to use an AVG or AVF (versus CVC) as vascular access (AOR = 0.85, 95% CI = 0.70–1.03, P = 0.10). TF patients enrolled in DOPPS within 3 months of TF (n = 313), had lower AVG and AVF use (AOR = 0.45, 95% CI = 0.32–0.62, P < 0.0001), were less likely to have hemoglobin of 10–12 g/dL (AOR = 0.54, 95% CI = 0.40–0.73, P < 0.0001), serum albumin >4.0 g/dL (AOR = 0.28, 95% CI = 0.19–0.43, P < 0.0001) and Kt/V >1.2 (AOR = 0.34, 95% CI = 0.22–0.53, P < 0.0001), and more likely to have PTH >500 pg/mL (AOR = 2.32, 95% CI = 1.59–3.38, P < 0.0001).

Discussion

The extensive DOPPS data set allowed us to examine the impact of a prior kidney TF on clinical outcomes and QOL of patients on chronic HD in a large, international cohort. Despite younger age and a lower prevalence of diabetes, patients with a history of TF had reduced survival and reduced QOL compared with wait-listed TN patients. We also found that QOL differences were less apparent in those with a longer time since TF and that survival differences between TF and TN patients were greatest for infection-related mortality.

Previous research examining the mortality of TF patients has yielded conflicting results [6, 7]. Several registry-based observational studies have examined the outcomes of patients with a history of TF; however, these studies were restricted to North American cohorts [5–7]. Data from The Canadian Organ Replacement Register (CORR) demonstrated no survival differences between TF and TN dialysis patients [6]. In this study, the comparator group consisted of all incident dialysis patients, as waiting list data were unavailable. It has been well documented that wait-listed dialysis patients have a lower risk of death than those not yet listed [1]. Moreover, case-mix adjustment was performed using comorbidities obtained at the time of kidney transplantation, not at the time of kidney TF, underestimating accrued comorbidities over the course of renal transplantation.

Using US data from the Scientific Registry of Transplant Recipients, Rao et al. [7] demonstrated a greater mortality risk among TF patients compared with TN patients who were wait-listed for kidney transplantation. While Rao et al. did employ the use of a wait-listed TN comparator group in the second study, other factors, such as the baseline differences in survival between Canadian and US dialysis patients, as well as the limited comorbidity adjustment in the US study may also account in part for the different results obtained. Similar to the study by Rao et al., we performed additional sensitivity analyses restricting survival and hospitalization comparisons between both wait-listed TF and TN patients. In doing so, we saw a trend of a 17% increased risk of hospitalization which persisted among TF wait-listed patients compared with TN wait-listed patients. Unlike the study by Rao et al., we did not see an increased risk of death among TF wait-listed patients compared with TN wait-listed patients. The increased risk of death among TF patients was largely seen among TF patients not wait-listed for kidney TF at study enrollment compared with TN wait-listed patients. These findings may relate to the limited power in our analysis to detect such differences owing to the low mortality rate of wait-listed patients coupled with fewer (<25%) TF patients wait-listed for repeat kidney transplantation. The low wait-list rate may relate to differences in timing and eligibility of wait-listing across DOPPS countries. Alternatively the paper by Rao et al. characterized comorbidities among TF patients at the time of transplantation, while in the present analysis, they were characterized at the time of TF. Therefore, it is possible that accounting for the accrual of comorbidities over the transplant duration may have attenuated survival differences between TF and TN wait-listed patients.

There are several reasons why CVC use may be higher among TF patients. TF patients have experienced a period of HD prior to receiving a kidney transplant. For some patients, options for a surgical vascular access may have been exhausted during the pretransplant HD period, with limited opportunities for a repeat AVF or AVG. The higher rates of CVC use among TF patients may be a proxy for suboptimal chronic kidney disease management prior to starting dialysis. Despite being managed by transplant nephrologists in the predialysis period, TF patients may be referred late for dialysis evaluation as a result of: (i) fragmentation of care between the kidney transplant and dialysis centers, (ii) an overemphasis on preservation of renal allograft function, and an underemphasis of predialysis care, (iii) patient-induced delays including reluctance to accept the need for dialysis and (iv) unanticipated and rapid loss of kidney allograft function [20]. Among patients with native kidney function decline, multidisciplinary predialysis care has been demonstrated to improve the use of surgical HD vascular access, and improve bone mineral metabolism parameters at dialysis initiation [21], while reducing morbidity and mortality upon dialysis initiation [22]. Consistent with previous observations, despite younger age, TF patients within the first 3 months of dialysis initiation were less likely to achieve clinical practice guidelines on HD and had lower serum albumin, lower hemoglobin and greater PTH compared with TN patients [20]. These parameters improved among patients with a longer period of dialysis after TF. The provision of more comprehensive predialysis care among patients with failing renal allografts may improve achievement of practice guidelines and clinical outcomes as well.

The adverse outcomes of TF patients relative to TN patients may also reflect the deleterious effects of prolonged and chronic immunosuppression exposure over the duration of graft function. Maintenance immunosuppression regimens carry short- and long-term increased risks of cancer, infection and cardiovascular complications [23, 24]. Whether or not immunosuppression exposure after TF may further increase the risks of adverse events has not been well studied.

The presence of a failed renal allograft may be an ongoing source of chronic inflammation, an established risk factor for adverse events among end-stage renal disease patients [25, 26]. We observed several features of chronic inflammatory syndrome including a lower BMI, anemia and elevated ferritin and hypoalbuminemia, features that are often accompanied by elevated systemic inflammatory markers as well as erythropoietin resistance [27]. However, preliminary observational data suggested that transplant nephrectomy may improve erythropoietin sensitivity, correct hypoalbuminemia and is associated with improved survival [27–29]. Further prospective data are required to confirm this finding before definitive conclusions can be drawn.

Depression and reduced QOL are prevalent among HD patients compared with the general population [30]. Kidney transplantation is associated with significant improvements in QOL [2]. Not surprisingly, patients returning to dialysis after kidney TF reported inferior QOL and greater physician-diagnosed depression compared with wait-listed TN patients. The reduction in QOL among TF patients was largest for patients initiating dialysis <3 months after TF. This represents a challenging transitional period as TF patients adapt to the loss of autonomy and significant QOL improvements associated with kidney transplantation. While the QOL deficit lessened over time relative to TN patients, it is possible that the improvements in QOL may reflect the impact of survivor bias; namely, QOL may be greater in TF patients healthy enough to survive to 1 year or more post-failure. Greater differences between PCS relative to MCS between TF and TN patients may relate to functional limitations not only due to the adverse effects of uremia, but may be compounded by the observed elevations in inflammatory markers among TF patients, indicating chronic systemic inflammation and the potential deleterious effects of immunosuppression exposure on muscle strength and exercise tolerance.

There are limitations worth noting in the present study. As with all analyses of observational data, a threat to validity is confounding based on unmeasured facility- and patient-level characteristics. Variables including exposure to multidisciplinary predialysis care and the acuity of dialysis initiation and levels of kidney function at dialysis initiation would have been of interest in understanding the mortality differences between TF and TN patients. Vascular access attempts prior to the start of dialysis were not recorded. Information regarding cause of renal TF, immunosuppression exposure and the use of transplant nephrectomy was limited but may have impacted the survival of TF patients. Furthermore, QOL data were available for a subset of patients and limited to one point in time. Assessment of the impact of the time since TF on HR-QOL would have ideally been addressed using serial assessments of QOL.

Notwithstanding these limitations, we have demonstrated that compared with TN patients, TF patients have reduced survival and QOL. Kidney care practitioners need to familiarize themselves with the medical and psychosocial challenges of this unique and growing patient population. Interventions aimed at reducing the morbidity of TF patients are needed. The role of modifiable practices to improve the outcomes of TF patients including the use of multidisciplinary predialysis care, the method and rapidity of immunosuppression reduction after TF and the use of transplant nephrectomy are questions which need to be answered.

Appendix 1.

International composition of the DOPPS

DOPPS 1 (1996–2001)	DOPPS 2 (2002–04)	DOPPS 3 (2005–08)
France	Australia and New Zealand	Australia and New Zealand
Germany	Belgium	Belgium
Italy	Canada	Canada
Japan	France	France
Spain	Germany	Germany
UK	Italy	Italy
USA	Japan	Japan
	Spain	Spain
	Sweden	Sweden
	UK	UK
	USA	USA

Appendix 2. Definitions of cause-specific mortality and hospitalization

	Causes of death	Hospitalization
		Diagnoses	Procedures
Cardiovascular-related	Myocardial infarction (acute)	Hypertension	Cardiac catheterization
	Pericarditis (including cardiac tamponade)	Angina	Coronary angioplasty
	Atherosclerotic heart disease	Chest pain (MI ruled out)	Coronary bypass graft (CABG)
	Cardiomyopathy	Acute myocardial infarction (MI)	Valve repair or replacement
	Cardiac arrhythmia	Cardiac arrest/sudden death	Cardioversion
	Cardiac arrest	Congestive heart failure	AICD (defibrillator) placement
	Valvular heart disease	Cardiomyopathy	Pacemaker placed
	Pulmonary edema due to exogenous fluid	Valvular heart disease	Pericardial procedure
	Congestive heart failure	Atrial fibrillation	Other cardiac procedures
	Cerebrovascular accident (including intracranial hemorrhage)	Other arrythmia	Carotid endarterectomy
		Pericarditis and/or tamponade	Evacuation of hematoma
		Hypotension	Angiogram
		Other cardiac diagnosis	Arterial bypass surgery
		TIA	Amputation
		Stroke (CVA)	Aortic aneurysm repair
		Subdural hematoma	Wound debridement
		Claudication/rest pain	Other vascular procedures
		Ulcer of extremity
		Gangrene
		Aortic aneurysm
		Deep vein thrombosis
		Other vascular access diagnosis
Infection-related	Septicemia due to vascular access	Pneumonia	Abscess drainage
	Septicemia due to peritonitis	Septicemia	Antibiotic therapy
	Septicemia due to peripheral vascular disease (gangrene)	Endocarditis	Other infection procedures
	Septicemia (other)	AIDS/HIV
	Pulmonary infection (bacterial)	Urinary tract infection
	Pulmonary infection (fungal)	Wound infection
	Pulmonary infection (other)	Abscess
	Viral infection (CMV)	Meningitis
	Viral infection (other)	Cellulitis/soft tissue infection
	Tuberculosis	Osteomyelitis
	AIDS	Viral infection
	Infections (other)	Fungal infection
	Hepatitis B	Fever or chills (source unknown)
	Other viral hepatitis	Other infection diagnosis
	Fungal peritonitis

Appendix 3. Summary comorbid conditions collected at DOPPS study entry

Summary comorbid condition
Coronary artery disease (CAD)
Congestive heart failure (CHF)
Other cardiovascular disease
Cancer (other than skin)
Cerebrovascular disease
Diabetes
Gastrointestinal bleeding in prior 12 months
Hypertension
Lung disease
Neurological disease
Peripheral artery disease (PAD)
Psychiatric disorder
Recurrent cellulitis or gangrene