Abstract

Background

Increasing numbers of elderly patients face decisions about the management of end-stage kidney disease. Improved understanding of contemporary patient and practice factors influencing prognosis may assist decision making for individual patients and their care providers.

Methods

This is a prospective registry study using multivariable proportional hazards models.

A total of 1781 patients aged ≥75 years at dialysis initiation recorded in ANZDATA, the Australia and New Zealand renal replacement registry, between January 2002 and December 2005.

The patient characteristics were demographic and comorbid conditions. The practice characteristics were late referral, access at dialysis initiation and intended dialysis modality (modality established by 90 days).

The study outcome was mortality censored at 31 December 2007 or at recovery of renal function (of at least 30 days), transplantation or loss to follow-up.

Results

Median follow-up was 2.3 years (interquartile range 1.1–3.3 years) during which time, 65% of the patients died. Baseline factors independently associated with mortality were older age [hazard ratio (HR) 1.24 for 5-year increase, 95% confidence interval (CI) 1.13–1.36], body mass index <18.5 (HR 1.78, 95% CI 1.33–2.38), number of comorbidities (one comorbidity HR 1.38, 95% CI 1.13–1.69; two comorbidities HR 1.55, 95% CI 1.27–1.89; three or more comorbidities HR 1.89, 95% CI 1.55–2.31), late referral (HR 1.19, 95% CI 1.02–1.39), peritoneal dialysis as intended modality (HR 1.26, 95% CI 1.08–1.47) and unprepared access (HR 1.43, 95% CI 1.23–1.67).

The limitations of the study were the observational nature of the analysis, potential selection bias introduced through analysis of a group who actually commenced dialysis and the potential confounding from unmeasured factors or dichotomous reporting of comorbidities.

Conclusions

Within the elderly cohort, other patient characteristics have a greater association with mortality than 5-year age increments. Even after consideration of patient characteristics, practice factors have a striking impact on the survival of elderly patients commencing dialysis. In the absence of randomized studies, efforts to enhance the identification and preparation of elderly patients for dialysis may improve outcomes within current settings.

Introduction

The numbers of older people commencing dialysis are increasing globally [1, 2] with ongoing debate about the utility of such treatment in this age group [3]. There are sizeable variations in the proportion of elderly patients among dialysis populations ranging from 16% in Finland [4], approximately one quarter in the USA [5] and UK [4], 33% in Japan [6] and 40% in France and Greece [4]. This variation would appear to reflect more than differences in disease prevalence given estimates that significant numbers of elderly people with end-stage kidney disease (ESKD) are not managed with renal replacement therapy (RRT) [7]. Outcomes for elderly patients receiving dialysis therapies are poor, with actuarial life expectancy of a 75-year-old on dialysis only one-third of that of a similarly aged person not requiring dialysis [8]. Whilst age and comorbid conditions are established factors for higher mortality [1, 9], less is known about the impact of practice variables such as late referral and unprepared access.

Understanding the ratio of benefit to harm from dialysis therapy is particularly germane for the elderly [3] because of the limited life expectancy and considerable burden of treatment [10] in older people receiving dialysis. This study analyses a large contemporary cohort of elderly people initiating dialysis in order to understand patient and practice factors associated with survival.

Materials and methods

Population

Patients aged ≥75 years who started RRT between January 2002 and December 2005 inclusive were identified from the Australian and New Zealand Dialysis and Transplant Registry (ANZDATA). The ANZDATA registry includes data on all people treated with chronic RRT (dialysis or kidney transplantation) in Australia and New Zealand.

Measures

Patients were followed until the ANZDATA census of 31 December 2007. Patients were censored at transplantation, loss to follow-up or recovery of renal function (survival of >30 days following the cessation of RRT). Demographic and clinical information were collected at the initiation of RRT including age, gender, ethnicity, height, weight, history of co-morbid conditions (coronary heart disease, cerebrovascular disease, peripheral vascular disease, chronic lung disease, diabetes and cancer), cause of renal disease and past or current smoking. Age was classified in 5-year brackets to reflect the presentation of information used by many registries [11]. Body mass index (BMI) was calculated according to World Health Organization categories [12]. Practice characteristics included the presence or absence of late referral (defined as referral <3 months prior to dialysis initiation); dialysis access preparation (prepared dialysis access being the use of an arteriovenous fistula, arteriovenous graft or peritoneal dialysis catheter at dialysis initiation; unprepared access being the use of a central venous catheter at dialysis initiation); dialysis modality at first dialysis and intended dialysis modality (defined as the modality at 90 days following initiation with last dialysis modality assigned to those not dialysing at 90 days). The primary outcome was death. Cause of death was classified as cardiac, vascular, infectious, malignancy, cancer, social and other causes.

Statistical analysis

Survival and mortality rates were analysed using Kaplan–Meier analysis. Variables were tested using a log-rank test in univariable analyses. A co-morbid score was created by the numerical addition of the number of comorbidities individually significant on univariable analyses. Collinearity was defined as a Spearman's coefficient >0.4. Only one of any collinear variables was eligible for inclusion in the multivariable model building process with selection based on clinical grounds. It was anticipated that collinearity may be found between diabetes as a comorbidity and diabetes as a cause of renal disease. An a priori decision was made to select diabetes as a comorbidity as a candidate in the multivariable model (if it met other eligibility criteria for the model) as many cases of ESKD attributed to diabetes do not have biopsy confirmation and thus the presence of diabetes as a comorbidity is likely to be assessed more accurately. In fact, only 13% of Australian patients diagnosed with diabetes as a cause of ESKD had pathological confirmation in 2009 [11]. The assumption of proportional hazards was checked graphically using the log cumulative hazard plots for death according to the subgroups in the variables examined in the study. A P-value ≤0.05 was regarded as significant in all tests. Confidence intervals (CIs) were calculated with 95% probability. Age, gender and all single variables significant on univariable analyses were potential candidates in the final multivariable model. Comorbid score was chosen for inclusion over the individual comorbidities unless the inclusion of the latter significantly improved the fit of the model. Age and gender were added initially, followed by other patient characteristics and lastly by practice variables. Additional variables, beyond age and gender, were retained if their omission from the model significantly worsened the fit of the model according to the likelihood ratio. Backward and forward stepwise selection methods were performed in sensitivity testing. Sensitivity analyses were conducted to test the impact of design decisions on the final model. The univariable and final multivariable models were repeated in those who survived 90 days following dialysis commencement as a sensitivity analysis. A further sensitivity model included diabetes as a cause of renal disease and not as a comorbidity. The area under the receiver operator curve for the model was calculated with the pre-specification that a c-statistic of ≥0.7 was required as the threshold for the development of a risk equation [13]. Risk tertiles were derived from the multivariable model. Analyses were performed using Stata software, version 11 (StataCorp LP, College Station, TX, 2010).

Results

Characteristics of the cohort

A total of 1781 patients aged ≥75 years began chronic RRT over the 4 years from 2002 to 2005 and were followed for a median of 2.3 years [interquartile range (IQR) 1.1–3.3]. The majority (1650, 92.6%) were aged between 75 and 85 years with only 19 (1.1%) aged over 90 years at the commencement of dialysis. During the follow-up period, 1150 patients died at a median of 1.50 (IQR 0.66–2.50) years, 1 patient was transplanted, 1 was lost to follow-up and 20 were censored at recovery of renal function. The vast majority of patients were Caucasian (n = 1622, 91.1%). There were smaller numbers of patients of non-Caucasian ethnicity including Chinese (n = 38, 2.1%), Indian (n = 22, 1.2%), Maori (n = 20, 1.2%) and Aboriginal (n = 8, 0.4%) patients who were combined for subsequent analyses. Baseline clinical characteristics (Table 1) are notable for the high burden of comorbidities in this group, with 91% of patients having at least one, and almost half having three or more, comorbidities. Some form of vascular disease was present in ∼70% of patients.

Table 1.

Baseline characteristics of patients aged ≥75 years initiating dialysis in Australasia between January 2002 and December 2005

Patient characteristics Number (percentage) 
Total number 1781 
Age (median) 79 (IQR 77–81) 
Gender (male) 1103 (61.9) 
Ethnicity (Caucasian) 1622 (91.1) 
BMI categories  
 BMI <18.5 79 (4.4) 
 BMI 18.5–25 832 (46.8) 
 BMI 25–29 606 (34.1) 
 BMI ≥30 261 (14.7) 
Comorbid conditions  
 No comorbid conditions 151 (8.5) 
 Any cardiovascular disease 1230 (69.1) 
 Coronary heart disease 1063 (59.7) 
 Peripheral vascular disease 579 (32.5) 
 Cerebrovascular disease 422 (23.7) 
 Chronic lung disease 366 (20.6) 
 Diabetes 552 (31.0) 
 Smoking history (former/current) 947 (53.2) 
 Cancer 392 (22.0) 
Cause of renal disease  
 Diabetic nephropathy 335 (18.8) 
 Glomerulonephritis 278 (15.6) 
 Renovascular disease 592 (33.2) 
 Other 576 (32.3) 
Practice variables  
 Treatment type at initiation (haemodialysis) 1328 (74.6) 
 Treatment type at 90 days (haemodialysis) 1238 (69.5) 
 Late referral to nephrologist 489 (27.5) 
 Prepared access at first usea 821 (54.4) 
Patient characteristics Number (percentage) 
Total number 1781 
Age (median) 79 (IQR 77–81) 
Gender (male) 1103 (61.9) 
Ethnicity (Caucasian) 1622 (91.1) 
BMI categories  
 BMI <18.5 79 (4.4) 
 BMI 18.5–25 832 (46.8) 
 BMI 25–29 606 (34.1) 
 BMI ≥30 261 (14.7) 
Comorbid conditions  
 No comorbid conditions 151 (8.5) 
 Any cardiovascular disease 1230 (69.1) 
 Coronary heart disease 1063 (59.7) 
 Peripheral vascular disease 579 (32.5) 
 Cerebrovascular disease 422 (23.7) 
 Chronic lung disease 366 (20.6) 
 Diabetes 552 (31.0) 
 Smoking history (former/current) 947 (53.2) 
 Cancer 392 (22.0) 
Cause of renal disease  
 Diabetic nephropathy 335 (18.8) 
 Glomerulonephritis 278 (15.6) 
 Renovascular disease 592 (33.2) 
 Other 576 (32.3) 
Practice variables  
 Treatment type at initiation (haemodialysis) 1328 (74.6) 
 Treatment type at 90 days (haemodialysis) 1238 (69.5) 
 Late referral to nephrologist 489 (27.5) 
 Prepared access at first usea 821 (54.4) 

aData available on 1509 patients (84.7%) of the total group as this information was only collected after October 2003.

Access preparation for dialysis could be determined in 84.7% (n = 1509) of patients of whom 46% commenced dialysis without permanent access. The majority of patients (74.6%) received haemodialysis for their first treatment. Only a few patients (n = 124) commencing haemodialysis subsequently changed to peritoneal dialysis within the first 90 days.

Mortality

The majority of elderly patients commencing dialysis (1150 patients or 64.6%) died during the total follow-up period with 1- and 2-year survival rates of 76.9 and 59.3% respectively (Figure 1). Causes of death are illustrated in Table 2.

Fig. 1.

Survival of patients aged ≥75 years initiating dialysis in Australasia between January 2002 and December 2005 (Kaplan–Meier curves) with 95% CIs compared with survival of 75- and 80-year-olds from the general Australian population [23]. Inset: proportion alive at 1, 2, 3 and 4 years following commencement of dialysis compared to general population.

Fig. 1.

Survival of patients aged ≥75 years initiating dialysis in Australasia between January 2002 and December 2005 (Kaplan–Meier curves) with 95% CIs compared with survival of 75- and 80-year-olds from the general Australian population [23]. Inset: proportion alive at 1, 2, 3 and 4 years following commencement of dialysis compared to general population.

Table 2.

Causes of death of patients aged ≥75 years initiating dialysis in Australasia between January 2002 and December 2005

Cause of death Number of patients Percentage 
Cardiaca 457 39.7 
Socialb 230 20.0 
Vascularc 182 15.8 
Infection 120 10.4 
Malignancyd 113 9.8 
Othere 48 4.2 
Cause of death Number of patients Percentage 
Cardiaca 457 39.7 
Socialb 230 20.0 
Vascularc 182 15.8 
Infection 120 10.4 
Malignancyd 113 9.8 
Othere 48 4.2 

aDefined as death due to myocardial ischaemia, cardiac arrest, cardiac failure of any cause, pulmonary oedema, pericarditis, hyperkalaemia and withdrawal due to cardiovascular disease.

bDefined as withdrawal due to psychosocial reasons, patient refused treatment, suicide and therapy ceased.

cDefined as death due to pulmonary embolus, cerebrovascular accident, haemorrhage, bowel infarction, withdrawal due to cerebrovascular disease, withdrawal due to peripheral vascular disease and withdrawal due to dialysis access difficulties.

dDefined as death due to malignancy and withdrawal due to malignancy.

eDefined as death due to hepatic failure, pancreatitis, bone marrow failure, cachexia, respiratory failure, sclerosing peritonitis, accidental death, unknown and other cause.

Analysis of separate variables

Both patient and practice variables were associated with higher mortality on univariable analyses (Table 3). Increasing age was associated with higher mortality (Supplementary Figure 1). Increasing numbers of comorbid conditions conferred a significant and graduated increase in mortality risk (Figure 2) with a single comorbidity conferring a 43% risk (P < 0.001), two comorbidities 61% (P < 0.001) and three or more comorbidities doubling the risk of death compared to the absence of reported comorbidities (P < 0.001). Individual comorbid conditions including coronary heart disease [hazard ratio (HR) 1.40, 95% CI 1.24–1.58], peripheral vascular disease (HR 1.42, 95% CI 1.26–1.60), cerebrovascular disease (HR 1.37, 95% CI 1.20–1.56), chronic lung disease (HR 1.35, 95% CI 1.18–1.56), diabetes (HR 1.14, 95% CI 1.01–1.29) and underweight status (HR 1.56, 95% CI 1.20–2.01) were individually associated with increased risk (Supplementary Figure 1). Categories of renal disease [diabetic nephropathy (HR 1.01, P = 0.90), glomerulonephritis (HR 0.81, P = 0.11), renovascular disease (HR 1.25, P = 0.06) and other causes (HR 1.00, P = 0.97)] were not associated with increased mortality. Gender [male (HR 1.02, P = 0.69)], smoking, former (HR 0.99, P = 0.89) or current (HR 1.06, P = 0.69), or cancer history prior to RRT commencement (HR 1.03, P = 0.65) were also not associated with increased mortality risk.

Fig. 2.

Survival of patients aged ≥75 years initiating dialysis in Australasia between January 2002 and December 2005 according to comorbid score (calculated as the numerical addition of baseline comorbid conditions significant on univariable analyses: coronary heart disease, cerebrovascular disease, peripheral vascular disease, chronic lung disease and diabetes).

Fig. 2.

Survival of patients aged ≥75 years initiating dialysis in Australasia between January 2002 and December 2005 according to comorbid score (calculated as the numerical addition of baseline comorbid conditions significant on univariable analyses: coronary heart disease, cerebrovascular disease, peripheral vascular disease, chronic lung disease and diabetes).

Table 3.

Cox proportional univariable and multivariable analyses of patient and practice variables on survival in patients aged ≥75 years initiating dialysis in Australasia between January 2002 and December 2005a

 Univariable analysis
 
Multivariable analysis
 
Variable HR (crude) 95% CI P-value HR (crude) 95% CI P-value Wald chi-square 
Age—5-year increment 1.27 1.17–1.38 <0.001 1.24 1.13–1.36 <0.001 19.80 
Gender—male versus female 1.02 0.91–1.16 0.689 1.04 0.91–1.19 0.585 0.30 
Ethnicity—Caucasian versus non-Caucasian 1.13 0.92–1.40 0.247 1.22 0.96–1.55 0.105 2.62 
Number of comorbid conditions versus no comorbid conditions        
 1 1.43 1.19–1.70 <0.001 1.38 1.13–1.69 0.002 9.99 
 2 1.61 1.34–1.92 <0.001 1.55 1.27–1.89 <0.001 18.66 
 ≥3 2.00 1.68–2.40 <0.001 1.89 1.55–2.31 <0.001 39.31 
BMI versus BMI        
 18.5–24.9 (Normal weight)        
 <18.5 (Underweight) 1.56 1.21–2.01 0.001 1.78 1.33–2.38 <0.001 15.21 
 25–29.9 (Overweight) 0.95 0.83–1.08 0.414 0.96 0.83–1.12 0.627 0.24 
 ≥30 (Obese) 1.02 0.86–1.22 0.818 1.05 0.86–1.27 0.653 0.20 
Intended dialysis modalityb        
 Peritoneal dialysis versus haemodialysis 0.95 0.84–1.08 0.471 1.26 1.08–1.47 0.003 8.82 
Late referral to nephrologistc 1.44 1.27–1.64 <0.001 1.19 1.02–1.39 0.025 5.02 
Unprepared access at first dialysisd 1.46 1.28–1.66 <0.001 1.43 1.23–1.67 <0.001 20.61 
 Univariable analysis
 
Multivariable analysis
 
Variable HR (crude) 95% CI P-value HR (crude) 95% CI P-value Wald chi-square 
Age—5-year increment 1.27 1.17–1.38 <0.001 1.24 1.13–1.36 <0.001 19.80 
Gender—male versus female 1.02 0.91–1.16 0.689 1.04 0.91–1.19 0.585 0.30 
Ethnicity—Caucasian versus non-Caucasian 1.13 0.92–1.40 0.247 1.22 0.96–1.55 0.105 2.62 
Number of comorbid conditions versus no comorbid conditions        
 1 1.43 1.19–1.70 <0.001 1.38 1.13–1.69 0.002 9.99 
 2 1.61 1.34–1.92 <0.001 1.55 1.27–1.89 <0.001 18.66 
 ≥3 2.00 1.68–2.40 <0.001 1.89 1.55–2.31 <0.001 39.31 
BMI versus BMI        
 18.5–24.9 (Normal weight)        
 <18.5 (Underweight) 1.56 1.21–2.01 0.001 1.78 1.33–2.38 <0.001 15.21 
 25–29.9 (Overweight) 0.95 0.83–1.08 0.414 0.96 0.83–1.12 0.627 0.24 
 ≥30 (Obese) 1.02 0.86–1.22 0.818 1.05 0.86–1.27 0.653 0.20 
Intended dialysis modalityb        
 Peritoneal dialysis versus haemodialysis 0.95 0.84–1.08 0.471 1.26 1.08–1.47 0.003 8.82 
Late referral to nephrologistc 1.44 1.27–1.64 <0.001 1.19 1.02–1.39 0.025 5.02 
Unprepared access at first dialysisd 1.46 1.28–1.66 <0.001 1.43 1.23–1.67 <0.001 20.61 

ac-statistic for multivariable model 0.614.

bDialysis modality at 90 days following initiation with last dialysis modality assigned to those not dialysing at 90 days.

cReferral <3 months prior to dialysis initiation.

dUse of central venous catheter at dialysis initiation.

Late referral (HR 1.44, 95% CI 1.27–1.64) and unprepared access (use of a central venous catheter at dialysis initiation) (HR 1.46, 95% CI 1.28–1.66) were associated with increased mortality (Table 3, Supplementary Figure 2). Intended dialysis modality (peritoneal dialysis compared with haemodialysis at 90 days) was not associated with a difference in survival in univariable analysis (HR 0.95 95% CI 0.84–1.08, P = 0.471) (Table 3, Supplementary Figure 2). There was no interaction between access preparation and intended dialysis modality. When access preparation was examined by intended dialysis modality, prepared haemodialysis patients had significantly lower mortality than prepared peritoneal dialysis patients (HR 1.28 95% CI 1.07–1.53), unprepared haemodialysis patients (HR 1.65 95% CI 1.40–1.95) and unprepared peritoneal dialysis patients (HR 1.73 95% CI 1.26–2.36) (Figure 3). When analyses were repeated excluding the 110 patients who died within the first 90 days of dialysis, the results were similar (Supplementary Figure 3).

Fig. 3.

Survival of patients aged ≥75 years initiating dialysis in Australasia between January 2002 and December 2005 according to access preparation and intended dialysis modality (defined below). Prepared access defined as use of arteriovenous graft, arteriovenous fistula or peritoneal catheter at dialysis initiation and unprepared access defined as use of temporary vascular catheter. Intended dialysis modality defined as the modality at 90 days following initiation with last dialysis modality assigned to those not dialysing at 90 days. HD, haemodialysis; PD, peritoneal dialysis.

Fig. 3.

Survival of patients aged ≥75 years initiating dialysis in Australasia between January 2002 and December 2005 according to access preparation and intended dialysis modality (defined below). Prepared access defined as use of arteriovenous graft, arteriovenous fistula or peritoneal catheter at dialysis initiation and unprepared access defined as use of temporary vascular catheter. Intended dialysis modality defined as the modality at 90 days following initiation with last dialysis modality assigned to those not dialysing at 90 days. HD, haemodialysis; PD, peritoneal dialysis.

Collinearity

The only evidence of collinearity was between diabetes as a comorbidity and diabetes as a cause of renal disease (Spearman's coefficient 0.71). All other collinear values were <0.22, well below the 0.4 eligibility cut-off for inclusion in the model.

Multivariable analysis

Age, underweight status, number of comorbidities, late referral, unprepared access and peritoneal dialysis at 90 days were all independent predictors of mortality in the multivariable model (Table 3). Unprepared access and late referral were associated with increased mortality risks of 43 and 19%, respectively, even after adjustment for patient characteristics. The multivariable model demonstrated that a comorbid score of three or more and being unprepared at first dialysis had an association with mortality that was not weaker than that of age (Wald score 39.31, 20.61 and 19.80, respectively, with higher Wald scores indicating greater contribution to the overall model).

Sensitivity analyses were conducted which confirmed the multivariable model was not influenced by the decisions made in its construction. In analyses limited to those who survived the first 90 days of dialysis, the results were unchanged (Supplementary Table). In a multivariable model that included diabetes as a cause of renal disease whilst excluding it from the comorbid score, renal disease due to diabetes was not an independent predictor (P = 0.874) and the overall model was unchanged from that presented in Table 3 (data not shown). Sensitivity analyses, which included comorbidities separately rather than as a comorbid score, confirmed that individual comorbidities were separately significant (other than peripheral vascular disease which was near-significant with a P-value of 0.06) whilst the rest of the model was essentially unchanged (data not shown).

Despite the inclusion of multiple patient and practice characteristics, some of the variation in survival remains unexplained. A risk score was not created as the model did not meet our pre-specified criterion for sufficient predictive power of a receiver operator curve score of 0.7. Nonetheless, risk tertiles were derived from the model to test its discrimination. Tertiles derived from the model separated elderly dialysis patients according to their mortality outcomes with those identified in the highest tertile of risk observed to have a median survival of 664 days compared with a median survival of 1257 days in those in the lowest third of risk (Figure 4).

Fig. 4.

Patient and practice variables discriminate low, medium and high risk of mortality in elderly dialysis patients. Observed survival in Kaplan–Meier survival curves of tertiles of risk defined by the multivariable model.

Fig. 4.

Patient and practice variables discriminate low, medium and high risk of mortality in elderly dialysis patients. Observed survival in Kaplan–Meier survival curves of tertiles of risk defined by the multivariable model.

Discussion

Our results confirm the poor survival of elderly patients starting dialysis (median survival 2.3 years) and highlight the independent effect of both patient and practice characteristics on survival in a large contemporary elderly dialysis cohort. Factors present at the time of starting dialysis that predict poorer survival include increasing age, greater number of comorbidities, being underweight, peritoneal dialysis as the intended dialysis modality, late referral and commencing dialysis without prepared access. The multivariable model using these variables describes a clinically meaningful difference in survival of elderly patients with a 1.6-year difference in median survival between the lowest and highest risk tertiles. The predictive variables identified in this study can assist clinicians in estimating prognosis in this group in whom determination of dialysis benefit and harm is essential. Integrated assessment of elderly patients approaching the decision for RRT should be informed by an appraisal of age, comorbidities, BMI, urgency of presentation and readiness for RRT.

Prepared access rates remain low in the elderly [14] with postulated barriers including patient [15], physician, surgical facility and health care system factors [16]. Our results argue against the traditional hypothesis that late referral and unprepared dialysis commencement are simply markers of illness and other characteristics associated with poor prognosis [17]. An analysis of elderly US Medicare patients surviving at least 90 days suggested the use of central venous catheters at haemodialysis commencement was associated with an increased 1-year mortality relative risk of 1.7 compared with simple arteriovenous fistulae although the analysis did not account for comorbid conditions other than diabetes and late referral [14]. A previous Australasian report of all-age haemodialysis patients did account for a large number of comorbid conditions and for late referral in a propensity score analysis and also demonstrated increased mortality associated with catheter use [18]. Our analysis of elderly patients treated with all dialysis modalities supports the hypothesis that the association of unprepared access and mortality is not explained by confounding as incorporation of multiple patient characteristics and late referral reduced the associated risk only minimally from 46 to 43%. The implication of these findings is that modifications to practice to increase access preparation rates may improve survival for the elderly dialysis cohort.

Survival for elderly people requiring dialysis is markedly less than that of similarly aged patients with cancer (observed median 5-year survival of 23.6% in dialysis compared with 51% in cancer [19]). Our study confirms the high burden of comorbid disease experienced by elderly people commencing dialysis in Australasia, with the total number of comorbidities being as important as individual conditions in predicting survival in elderly people commencing dialysis. Comorbid disease burden predicts survival in both elderly dialysis patients [20–23] and dialysis patients in general [9, 24] in most but not all [25] studies. The presence of two to three comorbid conditions in US patients over the age of 65 increased mortality risk by 31% compared to zero to one comorbid conditions [1]. Analyses of a small UK cohort choosing between dialysis or non-dialysis management showed dialysis was not associated with increased survival for those aged >75 with two or more comorbidities [20]. Indeed, for those surviving to the age of 75 and commencing dialysis in this current Australasian cohort, the presence of comorbid burden appeared to have a stronger effect on survival than age increments at the time of commencement of RRT. Some conditions traditionally associated with higher mortality were not prominent in our model with smoking and cancer not associated with increased mortality whilst diabetes had a relatively small impact. We believe these results likely reflect the effect of survival bias in this cohort that only includes those who both survive to the age of 75 and are commenced on RRT programmes.

The difficulty of identifying individual elderly patients who will do well or poorly with dialysis therapy has long been recognized [9]. Despite its comprehensive nature, our model failed to explain much of the variability in survival and did not meet our pre-specified threshold of accuracy for the generation of a risk score. Future attempts to study outcomes should consider better characterization of the variables studied and/or inclusion of other aspects, which may more completely describe mortality risk.

This study is the most detailed of the large analyses of survival of elderly patients entering dialysis due to the inclusion of both patient and practice characteristics, augmenting the findings from other large studies [1, 21, 26, 27]. Having used a treated patient register, this study reflects outcomes for those patients who actually commence dialysis—a group who select, and are selected for, dialysis. The purpose of these analyses was to provide information pertinent to decisions around the commencement of dialysis. In using a patient register, we could not directly measure modality ‘choice’ but instead have used the surrogate measure of the modality patients were actually transferred to within the first 90 days of dialysis. It is theoretically possible that in so doing, we have introduced bias although the results were unchanged when restricted to only those patients surviving 90 days. Our analysis is limited by the dichotomous reporting of most variables and cannot account for unmeasured factors or for residual confounding associated with severity of measured diseases. Comorbid conditions that we were unable to include but that have been shown to affect survival of elderly dialysis patients include hypoalbuminaemia [21, 25], hyperphosphataemia [25] and functional status [1, 25].

This is a comprehensive contemporary evaluation of the outcomes of Australasian dialysis patients aged ≥75 years indicating that both patient and practice variables predict survival in a large cohort of elderly patients. This analysis indicates that appraisal of characteristics beyond age alone will improve prognostication for elderly patients with ESKD. Additional research is needed to better predict outcomes and permit the development of a prognostic score that will further assist patients and their families in making treatment decisions. In the absence of randomized data, this analysis suggests that practice changes that facilitate the timely preparation of elderly chronic kidney disease patients for dialysis may improve outcomes.

Supplementary data

Supplementary data is available online at http://ndt.oxfordjournals.org. Figures 1-4 are available in colour online.

Conflict of interest statement

None declared.

Acknowledgements

C.F. was supported by a National Health and Medical Research Council (NHMRC) Medical and Dental Postgraduate Scholarship. A.C. was supported by an NHMRC Principal Research Fellowship. V.P. was supported by an Australian Heart Foundation Career Development Award. M.J. and M.G. were supported by Jacquot Research Establishment Awards & Jacquot Research Fellowship Awards respectively. M.J. was also supported by a Royal Australasian College of Physicians Servier Staff Post Doctoral Fellowship.

The ANZDATA Registry is funded by the Australian Organ and Tissue Donation and Transplantation Authority, The New Zealand Ministry of Health and Kidney Health Australia.

Disclaimer. The data reported here have been supplied by the Australia and New Zealand Dialysis and Transplant Registry. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as an official policy or interpretation of the Australia and New Zealand Dialysis and Transplant Registry.

References

1
Kurella
M
Covinsky
KE
Collins
AJ
, et al.  . 
Octogenarians and nonagenarians starting dialysis in the United States
Ann Intern Med
 , 
2007
, vol. 
146
 (pg. 
177
-
183
)
2
Krishnan
M
Lok
CE
Jassal
SV
Epidemiology and demographic aspects of treated end-stage renal disease in the elderly
Semin Dial
 , 
2002
, vol. 
15
 (pg. 
79
-
83
)
3
Renal Physicians Association
Shared Decision-Making in the Appropriate Initiation of and Withdrawal from Dialysis
 , 
2010
2nd edn
Rockville, MD
Renal Physicians Association
4
ERA-EDTA Registry Annual Report 2009
2011
Amsterdam, The Netherlands
Academic Medical Center, Department of Medical Informatics
5
U.S. Renal Data System
USRDS 2009 Annual Data Report: Atlas of End-Stage Renal Disease in the United States
 , 
2009
Bethesda, MD
National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases
6
Nakai
S
Suzuki
K
Masakane
I
, et al.  . 
Overview of regular dialysis treatment in Japan (as of 31 December 2008)
Ther Apher Dial
 , 
2010
, vol. 
14
 (pg. 
505
-
540
)
7
Australian Institute of Health and Welfare
End-Stage Kidney Disease in Australia: Total Incidence, 2003–2007
 , 
2011
Canberra, Australia
AIHW
 
Cat. no. PHE 143
8
USRDS
USRDS 2008: Annual Data Report: Atlas of Chronic Kidney Disease and End-Stage Renal Disease in the United States
 , 
2008
Bethesda, MD
National Institute of Diabetes and Digestive and Kidney Diseases
9
Chandna
SM
Schulz
J
Lawrence
C
, et al.  . 
Is there a rationale for rationing chronic dialysis? A hospital based cohort study of factors affecting survival and morbidity
BMJ
 , 
1999
, vol. 
318
 (pg. 
217
-
223
)
10
Carson
RC
Juszczak
M
Davenport
A
, et al.  . 
Is maximum conservative management an equivalent treatment option to dialysis for elderly patients with significant comorbid disease?
CJASN
 , 
2009
, vol. 
4
 (pg. 
1611
-
1619
)
11
McDonald
S
Excell
L
Livingston
B
The ANZDATA Registry
The 33rd Annual Report
ANZDATA Registry Report 2010 Australia and New Zealand Dialysis and Transplant Registry
 , 
2010
Adelaide, South Australia
ANZDATA Registry Report
12
Obesity: preventing and managing the global epidemic
Report of a WHO consultation
World Health Organ Tech Rep Ser
 , 
2000
, vol. 
894
 (pg. 
i
-
xii
1–253
13
Ohman
EM
Granger
CB
Harrington
RA
, et al.  . 
Risk stratification and therapeutic decision making in acute coronary syndromes
JAMA
 , 
2000
, vol. 
284
 (pg. 
876
-
878
)
14
Xue
JL
Dahl
D
Ebben
JP
, et al.  . 
The association of initial hemodialysis access type with mortality outcomes in elderly Medicare ESRD patients
Am J Kidney Dis
 , 
2003
, vol. 
42
 (pg. 
1013
-
1019
)
15
Xi
W
MacNab
J
Lok
CE
, et al.  . 
Who should be referred for a fistula? A survey of nephrologists
Nephrol Dial Transplant
 , 
2010
, vol. 
25
 (pg. 
2644
-
2651
)
16
Roca Tey
R
Vascular access for haemodialysis: an unresolved issue
Nefrologia
 , 
2010
, vol. 
30
 (pg. 
280
-
287
)
17
Baer
G
Lameire
N
Van Biesen
W
Late referral of patients with end-stage renal disease: an in-depth review and suggestions for further actions
NDT Plus
 , 
2010
, vol. 
3
 (pg. 
17
-
27
)
18
Polkinghorne
KR
McDonald
SP
Atkins
RC
, et al.  . 
Vascular access and all-cause mortality: a propensity score analysis
J Am Soc Nephrol
 , 
2004
, vol. 
15
 (pg. 
477
-
486
)
19
AIHW
Australia's Health 2008
 , 
2008
Canberra
AIHW
 
Cat. no. AUS 99
20
Murtagh
FEM
Marsh
JE
Donohoe
P
, et al.  . 
Dialysis or not? A comparative survival study of patients over 75 years with chronic kidney disease stage 5
Nephrol Dial Transplant
 , 
2007
, vol. 
22
 (pg. 
1955
-
1962
)
21
Couchoud
C
Moranne
O
Frimat
L
, et al.  . 
Associations between comorbidities, treatment choice and outcome in the elderly with end-stage renal disease
Nephrol Dial Transplant
 , 
2007
, vol. 
22
 (pg. 
3246
-
3254
)
22
Lamping
DL
Constantinovici
N
Roderick
P
, et al.  . 
Clinical outcomes, quality of life, and costs in the North Thames Dialysis Study of elderly people on dialysis: a prospective cohort study
Lancet
 , 
2000
, vol. 
356
 (pg. 
1543
-
1550
)
23
Joly
D
Anglicheau
D
Alberti
C
, et al.  . 
Octogenarians reaching end-stage renal disease: cohort study of decision-making and clinical outcomes
J Am Soc Nephrol
 , 
2003
, vol. 
14
 (pg. 
1012
-
1021
)
24
Davies
SJ
Phillips
L
Naish
PF
, et al.  . 
Quantifying comorbidity in peritoneal dialysis patients and its relationship to other predictors of survival
Nephrol Dial Transplant
 , 
2002
, vol. 
17
 (pg. 
1085
-
1092
)
25
Jassal
SV
Douglas
JF
Stout
RW
Prognostic markers in older patients starting renal replacement therapy
Nephrol Dial Transplant
 , 
1996
, vol. 
11
 (pg. 
1052
-
1057
)
26
Jassal
SV
Trpeski
L
Zhu
N
, et al.  . 
Changes in survival among elderly patients initiating dialysis from 1990 to 1999
CMAJ
 , 
2007
, vol. 
177
 (pg. 
1033
-
1038
)
27
Jager
KJ
van Dijk
PC
Dekker
FW
, et al.  . 
The epidemic of aging in renal replacement therapy: an update on elderly patients and their outcomes
Clin Nephrol
 , 
2003
, vol. 
60
 (pg. 
352
-
360
)

Comments

0 Comments