Mortality in Dialysis Patients: Analysis of the Causes of Death Lionel U. Mailloux, MD, Alessandro G. Bellucci, MD, Barry M. Wilkes, MD, Barbara Napolitano, MS, Robert T. Mossey, MD, Martin Lesser, PhD, and Peter A. Bluestone, MD • The objective of this study was to identify the causes of death in maintenance dialysis patients who survived at least 90 days and were monitored during a 16-year period. Of 532 patients starting dialysis, 222 died. The causes of death were grouped into six categories: cardiac, infectious, withdrawal from dialysis, sudden, vascular, and "other." The greatest number of deaths were due to infections, followed by withdrawal from dialysis, cardiac, sudden death, vascular, and other. The risk of dying increased for the first 4 years of dialysis, decreased in years 5 through 10, and had a second increase at 11 years. The mortality during the first 4 years consisted largely of infectious and cardiac deaths. The late peak of deaths was mainly due to infections and withdrawal from dialysis. Overall, infections accounted for more than 36% of all deaths. Withdrawal occurred in 21.2% of the patients and was most common in patients over age 61. Notably, cardiac deaths accounted for only 14.4% of all deaths and no patient died from a cardiac cause after surviving on dialysis more than 8.5 years. We conclude that infection is the leading cause of death in our dialysis patient population. Withdrawal from dialysis was a common cause of death, especially in older patients. Cardiac mortality was not as frequent as anticipated and occurred mainly in patients on dialysis less than 4 years, suggesting that it is the result of preexisting disease. © 1991 by the National Kidney Foundation, Inc. INDEX WORDS: Causes of death; cardiac death; infection; withdrawal from dialysis.
I
N RECENT YEARS, the widespread availability of dialysis for the treatment of end-stage renal disease has resulted in a changing patient population. In our program, as in others, more older patients with multiple medical problems have begun dialysis during the last 8 years. I -4 Shorter survival on dialysis is recognized to occur in patients with diabetes mellitus, renal vascular disease, and advancing age. 1,2,5,6 Previous studies in dialysis patients have found that cardiac disease is the most frequent cause of death. 4,7-14 This is not surprising, since previous studies have linked long-term dialysis to accelerated atherosclerosis and cardiac mortality.3,15-25 These studies are largely based on registries and often involve multiple centers with different patterns of practice. The Division of Nephrology at North Shore University Hospital has maintained a clinical database for all dialysis patients over the age of From the Division of Nephrology and Hypertension, Department of Medicine, and the Division of Biostatistics, Department of Research, North Shore University Hospital-Cornell University Medical College; andfrom the Departments of Medicine and Public Health, Cornell University Medical College, Manhasset, NY. Presented in part at the 22nd Annual Meeting of the American Society of Nephrology, Washington, DC, December 1989. Address reprint requests to Lionel U. Mailloux, MD, FACP, Division of Nephrology and Hypertension, North Shore University Hospital-Cornell University Medical College, 300 Community Dr, Manhasset, NY 11030. © 1991 by the National Kidney Foundation, Inc. 0272-6386/91/1803-0005$3.00;0 326
15 since the inception of the program in January 1970. 1 We have reviewed our database to calculate annual gross mortality rates, to determine the changing pattern of mortality, and to evaluate the risk factors for mortality on dialysis. We report our 16-year experience. MATERIALS AND METHODS
Patient Population From 1970 through 1985. demographic and clinical data of all 532 patients with end-stage renal disease aged above IS years and surviving on dialysis over 90 days were kept on a computerized database. Table I shows median age and the renal diagnoses by the years the dialysis treatments were started. The North Shore dialysis unit is based in a tertiary care University Hospital in a suburban community. In our community, over 98% of the patients have health insurance before starting dialysis. The same physicians have been caring for these patients since 1970. All dialysis treatments are performed using generally accepted protocols to maintain normal blood pressures, dry weights, and a sense of well-being. Our hemodialysis prescription has consisted of three treatments weekly, with blood flows of at least 250 mL/min dependent on dialyzer membrane, dialysate flows of 500 mL/min, and 4.0 to 5.0 hours' duration. No "rapid dialysis" protocol has been used in our units. Intermittent peritoneal dialysis treatments were performed using 2-L exchanges for 20 hours biweekly. The patients on continuous ambulatory peritoneal dialysis (CAPD) performed at least four 2-L exchanges daily. None of the patients received erythropoietin. To evaluate patients who switched from one form of dialysis treatment to another, we arbitrarily adopted the following convention. If a patient ended one form of treatment (ie, switched from hemodialysis to CAPD), the patient was "transferred alive" from the hemodialysis group and entered as a new patient into the
American Journal of Kidney Diseases, Vol XVIII, No 3 (September), 1991: pp 326-335
327
CAUSES OF DEATH IN DIALYSIS PATIENTS
Table 1. Median Age, Renal Diagnosis, and Year Started Start Years
1970-1973 (n = 63)
Median age (yr) Diagnosis (%) Chronic glomerulonephritis Diabetes mellitus Renal vascular Polycystic kidneys Other'
1974-1977 (n = 152)
1977-1981 (n = 135)
1982-1985 (n = 182)
47
52
54
60
41.3 6.4 7.9 11.1 33.3
31.0 18.4 7.9 9.2 33.5
22.2 29.0 6.7 7.4 34.7
24.2 25.0 16.5 9.3 25.0
, Other renal diagnoses include transplant rejection, amyloidosis, tubulointerstitial nephritis, chronic pyelonephritis, obstructive uropathy, collagen vascular disease, miscellaneous renal disorders, and malignant hypertension.
CAPD group as if the patient had just started CAPD treatments. A specific renal diagnosis was entered into the database for each patient starting dialysis, and these diagnoses were then grouped into five categories: chronic glomerulonephritis, polycystic kidney disease, diabetes mellitus, renal vascular disease, and "other" as previously described. l Throughout the 20-year North Shore University Hospital program, the diagnosis of diabetic renal disease has been made on the basis of fasting hyperglycemia, retinopathy, nephrotic syndrome, hypertension, and renal insufficiency, regardless of whether patients were taking insulin. Large kidneys and, in some cases, renal biopsy material were available to further assist with the diagnosis. Therefore, we feel that stratifying diabetic patients into type I and type 2 would be somewhat arbitrary, so all diabetics with chronic renal failure are considered together, providing they meet the above clinical criteria. A diagnosis of renal vascular disease was made in patients with clinical evidence of systemic atherosclerotic vascular disease by a combination of imaging findings, such as a positive renal angiogram or suggestive radionuclide studies with unequal renal size, low level proteinuria « 1.0 g), and benign urinary sediment. Although these patients had hypertension, none had malignant hypertension with papilledema or end-organ damage. The diagnosis of malignant hypertension was reserved
for those patients with a hypertensive crisis, grade IV retinopathy, and end-organ damage. The cause of death was established by the nephrologists caring for the patient at the time of death. The 15 causes of death were compressed into six categories: cardiac, infectious, withdrawal, sudden, vascular, and "other." The same definition for each cause of death has been in use throughout the study period and is presented in Table 2. Sixty-seven (30.2%) of the deceased underwent an autopsy, with the clinical diagnosis confirmed in all but two cases.
Statistical Methods All statistical analyses were performed on a mainframe computer with the SAS software package (SAS-Institute, Cary, NC). The time until death was measured as the time from the start of the first dialysis until death. The annual gross mortality rate for a year, GMRy, was calculated by the formula: GMRy = n deceasedy/(Jan Iy n/2 + Dec 31y n/2), where Jan Iy n/2 is half the number of dialysis patients alive on Jan Iyand similarly for Dec 31y n/2, and y is the year of interest. Cause-specific survival distributions, corresponding to the six cause of death groupings, were estimated using the productlimit method. 26 A patient's survival time was considered censored for a given cause of death if the patient was still alive,
Table 2. Definitions of the Causes of Death Infectious Withdrawal Cardiac
Sudden Vascular Other
• Death either from sepsis with proven bacteremia or as the sequelae of acute fulminant hepatitis. • Death with manifestations of uremia because of withdrawal from dialysis. Underlying medical conditions should not have been active, leading to rapid deterioration with imminent death. • Death due to the manifestations of a well-documented acute myocardial infarction (typical chest pain associated with electrocardiographic changes and enzyme elevations) or from overt congestive cardiac failure. • Sudden death without explanation, but not related to a preexisting cardiac condition or in a patient with known hyperkalemia. • Death as the result of either a cerebrovascular aCCident, mesenteric artery thrombosis, ruptured aneurysm, or generalized ischemic ("low flow") state. • All other causes of death including respiratory failure, hyperkalemia, gastrointestinal bleed, and other miscellaneous causes (trauma, etc).
328
MAILLOUX ET AL
WITHDRAWAL
21%
OTHER
9%
VASCULAR
CARDIAC
15%
8% SUDDEN
10%
died of another cause other than the one under study, received a transplant, or was transferred alive to another unit. For example, when analyzing cardiac deaths, the survival time of a patient who died of infections was considered censored. Survival distributions for different patient groupings, eg, startyear groups, age groups, renal diagnoses, etc, were compared using the log rank test. Cause-specific multivariate analyses were performed using the Cox proportional hazards regression model. Variables that proved to be either significant on univariate analysis or were believed, a priori, to be prognostic for survival, were entered into the model and eliminated using the backward elimination procedure. Renal diagnosis and start-year groupings were transformed into dummy variables for this analysis. (Dummy variables were used for start-year groups, even though they were ordinal in nature, because we could not assume a linear, or even a monotonic, relationship between the log hazard ratio and these ordinal variables.) A total of seven possible prognostic variables were introduced into the model: startyear group 1970 to 1973,start-yeargroup 1974 to 1977, startyear group 1978 to 1981, diabetes mellitus, chronic glomerulonephritis/polycystic kidney disease, vascular disease, and age at start of dialysis. Start-year group 1982 to 1985 and "other" renal diagnoses were implicitly entered into the model as a result of the dummy variable scheme. Cause-specific survival time was the dependent variable. A variable was deleted from the model if it did not meet the 5% significance level. To summarize the results of the Cox regression model, relative risks were computed relative to 1982 to 1985 for start years, age below 40 years, and "other" for renal diagnosis. The hazard function was used to describe the likelihood of dying just after time t, given that the patient had survived through time t. The hazard function plot provided a visual picture of the risk of all deaths and cardiac death over time. However, we recognize that the relatively small number of deaths in any specific cause group may limit formal inference from these curves.
Fig 1. Distribution of all 222 deceased dialysis patients presented as a percentage of the total:
RESULTS
There were 532 chronic dialysis patients over the age of 15 years between January 1, 1970 and December 31,1985, who survived at least 90 days on dialysis. There were 293 males, 239 females, 478 whites, 47 blacks, and 7 of other races. There were 425 patients on in-center hemodialysis, 52 on home hemodialysis, 46 on CAPD, and 9 on in-center peritoneal dialysis. Nine patients switched modality (3 from in-center hemodialysis to CAPD, 3 from CAPD to in-center hemodialysis, 2 from in-center hemodialysis to intermittent peritoneal dialysis, and 1 CAPD to home hemodialysis); and 77 patients were transferred out of the dialysis groups alive because of renal transplantation. Two hundred twenty-two (41.7%) of the 532 patients died through 1988. The death rate was lowest in the 52 home dialysis patients (7.7%) and highest (44%) in the in-center peritoneal dialysis patients. The causes of death, in order of frequency, were infections, withdrawal, cardiac, sudden death, vascular, and other (Fig 1). Visual inspection of crude cause-specific death rates compared across start-year groupings showed an increased risk of cardiac death from 5.6% in 1970 to 1973 to 25.6% in 1982 to 1985 (Table 3). There was a decrease in deaths due to infections from 44.4% in 1970 to 1973 to 21.6% in 1982 to 1985. To adjust for the shorter follow-
329
CAUSES OF DEATH IN DIALYSIS PATIENTS
Table 3. Causes of Death and Start-Year Groups Start-Year Groups
1970-1973
1974-1977
1978-1981
1982-1985
Causes of Death
(%)
(%)
(%)
(%)
Infectious Withdrawal Cardiac Sudden Vascular Other
44.4 13.9 5.6 16.7 11.1 8.3
42.9 26.0 9.1 7.8 6.5 7.8
36.2 20.7 19.0 10.3 5.2 8.6
21.6 19.6 25.6 9.8 11.8 11.7
n = 36
n = 77
n = 58
Deceased
up of the later start-year groupings (1978 to 1981 and 1982 to 1985), especially in the last grouping where follow-up time was 5 years or less, the analysis was redone using only those cases with 5 years or less of follow-up. The same trend was observed: an increase in cardiac deaths and a decrease in deaths due to infections. The same trend was also observed if follow-up was restricted to 3 years or less, as well as to 2 years or less. The relationship between age at start of dialysis and causes of death is depicted in Fig 2. Fifty-six percent of the patients over 61 years of age died during the time of study, while 79% of those entering dialysis below age 40 were still alive. These younger patients had a higher death rate from infection, while deaths due to withdrawal from dialysis increased significantly in patients above 61 years of age. The patients withdrawing from 50
n
=
51
dialysis were older at death, with a median age of69.6 years, compared with the five other causes of death (P < 0.0018, Kruskal-Wallis test) (Table 4). In 1970, the first year of the program, there were only two patients at risk and no deaths. For the years 1971 through 1988, the overall dialysis gross mortality rate was 14.3%, with a range of 0% to 19%. Dialysis mortality rates for our center from 1970 to 1988 and for the United States from 1979 to 1988 are presented in Table 5. The hazard function for death due to all causes increased during the first 3 years and peaked at 4 years, with a subsequent decrease and leveling off between years 5 through 10, at which time a second peak was observed at 11 years (Fig 3). Visual inspection of the cause-specific hazard functions for the causes of death showed that the
Percent
40
30
20 10
0
Fig 2. Different causes of death compared by age at start of dialysis.
-
=61
am
CARDIAC
CJ
OTHER
YRS
330
MAILLOUX ET AL Table 4. Causes of Death Versus Age at Death* Age (yr) n
Infectious Withdrawal Cardiac Sudden Vascular Othert
81 47 33 23 18 20
Mean ± SO
Median
Range
± ± ± ± ± ±
60.7 69.6 64.8 60.2 62.7 52.0
24-84 15-85 43-80 32-84 38-77 23-79
57.7 65.3 63.5 58.7 60.3 53.0
1.4 2.1 1.9 2.5 2.8 4.0
• P < 0.0018, Kruskal-Wallis test.
t
Other deaths include hyperkalemia in 10 patients, respiratory failure in 4, gastrointestinal bleeding in 1, and miscellaneous causes in 5.
first peak was composed mainly of cardiac and infectious deaths, while the late peak was withdrawal and infectious deaths. Univariate cause-specific survival analyses yielded nearly identical results to those obtained using the multivariate Cox proportional hazards regression model. Therefore, results for only the Cox model are reported. The Cox proportional hazards regression model identified several variables that were significantly associated with the risk of death for each cause of death grouping and for all deaths. Table 6 summarizes the significant variables for cardiac, infectious, withdrawal, and all causes of death. The risk of a cardiac death increased steadily as start year increased from 1970 to 1985 and with advancing age. The risk of cardiac death was significantly lower in patients with chronic glomerulonephritis or polycystic kidney disease than for those patients with diabetes mellitus, vascular, or other renal diagnoses. The hazard function for cardiac deaths showed that there was an early peak at 48 months, a decreasing risk through 8.5 years, and no deaths thereafter (Fig 4). There was a significant and steady increase in the risk of death from infections with increasing age at the start of dialysis (Table 6). For infectious causes of death, there was a steady decrease in the risk of death across the start years (Table 3). Patients with chronic glomerulonephritis or polycystic kidney disease showed a decreased risk of infectious deaths (Tables 6 and 7). The oldest patients were at significantly increased risk of death due to withdrawal from dialysis (Tables 4 and 6). However, the risk of death due to withdrawal was not different for any start-
year group and was lower in patients with polycystic kidney disease and chronic glomerulonephritis and other renal diagnoses compared with diabetes and renal vascular disease (Table 6). Patients with diabetes mellitus and renal vascular disease accounted for 52% of all the withdrawals from dialysis (Table 7). For all causes of death, it was observed that the start-year groups 1970to 1973, 1974 to 1977, and 1982 to 1985 shared the same risk of death. The start-year group 1978 to 1981 had a lower risk of death than the other three start-year Table 5. Gross Mortality Rates in Dialysis Patients by Year: 1970 Through 1988
Year
NSUH-CUMC (%)
US (%)
1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988
0 13.3 5.6 17.4 16.4 17.2 16.8 17.4 11.8 18.8 11.3 10.8 12.6 17.8 16.1 19.0 15.4 17.9 17.3
NA NA NA NA NA NA NA NA NA 20.6 20.9 20.5 20.1 20.6 20.7 22.2 22.8 23.4 24.3
Abbreviation: NA, data not available.
331
CAUSES OF DEATH IN DIALYSIS PATIENTS HAZ""" 0.030
o.oz. O.OZII 0.024 0.OZ2 O.OZO
0.011 0.0111 0.01" O.OlZ
0.010 0.0011 0.0011 0.004 O.OOZ
Fig 3. Hazard function for death due to all causes presented in 12-month intervals.
....~-,-~....~............~......,...,...,...,~,..............,,........~--r:=:;:::;:::,........
0.000 'r~,..............,,....
--'HS 011 DIALYSIS
groups. Risk of death for all causes of death was once again associated with increasing age. Each of the four renal diagnostic groups carried with it a different risk of death, with diabetes mellitus having the highest risk, followed by renal vascular disease, other, and then polycystic kidney disease and chronic glomerulonephritis.
DISCUSSION
Much attention has been devoted to survival of dialysis patients, but few analyses of the causes of death have been published. Most of the previously published data have been derived from patient registries where detailed clinical information was often lacking. The current study was
Table 6. Associated Risk of Death for Each Cause of Death: The Cox Proportional Hazards Regression Model Causes of Death Variable
Cardiac
Infectious
Withdrawal
All
1.00 2.39t 5.70t
1.00 1.65t 2.72t
1.00 3.46§ 11.94§
1.00 1.97§ 3.90§
0.18' 0.23t 0.32' 1.00
2.03' 1.67' 1.00 1.00
1.00 1.00 1.00 1.00
1.00 1.00 0.66t 1.00
1.00 1.00 0.21t 1.00
1.00 1.00 0.35§ 1.00
3.00t 3.08t 1.00 1.00
1.93+ 1.63' 0.52+ 1.00
Age
60 yr Start years
1970-1973 1974-1977 1978-1981 1982-1985 Renal diagnosis Diabetes mellitus Renal vascular PKD/CGN Other
NOTE. Entries and corresponding p values represent the relative risk of cause specific death, using as reference levels
1982 to 1985 for start years, less than 40 years for age at start, and "other" for renal diagnoses. These reference groups are all labeled with a relative risk of 1.00. Other factors shown with a relative risk of 1.00 had a risk of death that did not differ significantly from the reference level. Abbreviation: PKD/CGN, polycystic kidney disease and chronic glomerulonephritis . • p < 0.05. t p < 0.01.
+
P < 0.001. § P < 0.0001.
MAILLOUX ET AL
332
Table 7. Causes of Death and Renal Diagnoses
performed to investigate the clinical parameters that were indicative of long or short survival in patients with end-stage renal disease. Our definitions for the initial renal diagnoses 1 and causes of death were set 20 years ago and have been strictly adhered to throughout the program's existence. Furthermore, the clinical diagnoses for the cause of death were substantiated in one third of the patients by autopsies, being confirmatory in more than 97% of them. Several important findings emerged from the analysis. The dialysis patient population has aged significantly over the past 20 years, with the median starting age in our program having increased from 47 years in the 1970 to 1973 start group to 60 years in 1982 to 1985. 1 Not surprisingly, age has been shown to be a major risk factor for shorter survival on dialysis. 1,2,5,6,24,25,27-31 In addition, the number of patients with underlying diseases that are generally recognized as having poor prognosis (ie, diabetes mellitus, vascular disease) has increased in recent years. 1,5,7,21,25,32,33 Overall, the most common cause of death was infectious, followed by withdrawal from dialysis, cardiac, sudden, vascular, and other (Fig 1). Much attention has been paid to possible causes of the increasing dialysis mortality rates in the United States, including the changing dialysis patient population, demographics, renal diagnoses, and the dialysis prescription,z,5,21.25 We have not decreased our dialysis time nor changed our dialysis prescription despite an increased use of more permeable membranes and larger surface area dialyzers in our unit. Since our patient trends have paralleled the national figures, we feel that
Diagnoses
Deceased (n) Causes of death (%) Infectious Withdrawal Cardiac Vascular Sudden Other
CGN
PKD
Other
Diabetes
RVD
40
10
76
63
33
35 20 10 8 10 17
40 10 0 20 20 10
45 18 16 4 6 11
35 22 16 8 16 3
21 30 21 16 6 6
Abbreviations: CGN, chronic glomerulonephritis; PKD, polycystic kidney disease; diabetes, diabetes mellitus; RVD, renal vascular disease.
maintaining longer dialysis treatments pver the years may have resulted in our lower gross mortality rates. The importance of maintaining dialysis treatment time has recently been suggested by Held et al. 34 The risk of death as observed from the hazard function demonstrated two major peaks, one each at 48 and 132 months (Fig 3). The early peak consisted largely of patients dying of cardiovascular and infectious causes; the late peak was composed mainly of withdrawal from dialysis or infection. Although previously not recognized, the late peak of mortality is not completely unexpected, since long-term dialysis may lead to other poorly understood complications including secondary amyloidosis, neuropathy, bone disease, etc. Undoubtedly, there is the possibility that preexisting disease may contribute to the early
HAZARQ
0.020
0.018 0.016 0.014
0.012 0.010 0.008 0.006 0.004
0 . 0 0 2 l _ -__
12
2.
36
.8
60
72
14
95
108
MONTHS ON DIALYSIS
120
132
...
155
168
180
Fig 4. Hazard function for cardiac deaths presented in 12-month intervals.
CAUSES OF DEATH IN DIALYSIS PATIENTS
mortality rate. Since patients were not exhaustively studied at the outset for cardiac disease, eg, cardiac catheterization, it is not possible to exclude that preexisting disease contributed to mortality. We also used the common convention of excluding patients who died before the 90th day of dialysis, eliminating some early deaths and acute renal failure, although time to death was calculated from the first dialysis treatment. Infection was the most common cause of death throughout the study period and is consistent with the data of Collins et al. 33 None of the infectious deaths in our population were from arteriovenous access or graft infections. Beyond 6 years of dialysis, infections were the most common cause of death in patients starting dialysis under the age of 60, and were equal to the number of deaths caused by withdrawal from dialysis in patients over 61 years of age (Fig 2). Infectious deaths are generally due to common pathogens, but may also occur as a result of unusual pathogens. 35 Withdrawal from dialysis was our second most frequent cause of death, accounting for 29% of all deaths in patients over 61 years of age. Together, withdrawal from dialysis and infection were the most common causes of death in the oldest patient group. More than half of the withdrawals were in patients over 61 years of age at the time of entry to dialysis (Fig 2). The proportion of deaths due to infections appears to decrease in patients over 61 years of age as there is an increase in deaths by withdrawal. The observation that withdrawal is not an uncommon cause of death is supported by Neu and Kjellstrand who found a large proportion of withdrawals in diabetic dialysis patients. 36-39 An older starting age, diabetes mellitus, and renal vascular disease seemed to be important risk factors for this cause of death. Nine of the 47 withdrawals had a malignancy and six patients had significant malnutrition with continuing catabolism. Withdrawal from dialysis has not been discussed frequently in the medical literature, but probably occurs more often than reported. Social and cultural reasons, as well as attitudes toward living and dying, may explain the higher rates of withdrawal in certain populations. 36-39 Dialysis patients who were otherwise in good physical and mental health did not withdraw from dialysis. The withdrawals occurred in pa-
333
tients with systemic diseases, mostly diabetes and advanced age. These were debilitated individuals who would not have been considered to be in immediate danger of dying. Cardiac disease was the third most common cause of mortality. Although previous studies have emphasized cardiac disease as the leading cause of mortality, cardiac mortality was not the major cause of death in our patient population (Fig 1, Table 3). The reasons for this difference are not entirely apparent, but the widespread treatment of hypertension and occlusive cardiovascular disease by aggressive medical and surgical approaches in our patient population may be contributing factors. However, lethal cardiac events did occur in 15% of the patients. Although there has been an increase in cardiac mortality in the later start-year groups, the pattern of these deaths has not changed. Generally, these events took place within the first 4 years (Fig 4) and were the major contributor to the early mortality. Patients in the oldest age group had more cardiac deaths than other patients (Tables 4 and 6). The observation that a large portion of cardiac deaths occurred early suggests that the cardiac diseases predated the onset of chronic dialysis and were not caused by dialysis (Fig 4). These data are supported by some studies,4,6,9,IO,12,19,27,32,33 but differ from others. 3,1l,12,15.18,22 Early cardiac deaths formed a significant portion of deaths in the patients reported by Lundin et al 19 and Collins et al. 33 Myocardial ischemia and infarction have been reported as the most frequent causes of lieath in diabetics between 1980 to 1985 in the European Dialysis and Transplant Association Registry.4 Cardiac deaths have been the most frequent cause of mortality in a large CAPD program in France, and account for 52.4% of deaths in ESRD Network 4, 50% of deaths in ESRD Network 25, and 51% of deaths in 41 diabetic dialysis patients in Cleveland.4,7,1l-14,3~,34 The methods to determine the cause of death in those studies have not been specifically defined. If data were obtained from death certificates, which often state cardiopulmonary arrest as the proximate cause of death, 1hcorrect conclusions could easily be drawn. 4o The systematiC evaluation of causes of death in our dialysis population has led to several interesting new findings. However, it is difficult to speculate whether our findings in a surburban
334
MAILLOUX ET AL
population can be applied to other dialysis populations. Using the statistical approach adopted in our study, it will be possible to study prospectively larger populations of dialysis patients to determine factors that contribute to survival and mortality.
ACKNOWLEDGMENT We acknowledge the on-going support that Lawrence Scherr, MD, MACP, Director of the Department of Medicine, has provided through the years. We also appreciate the excellent care given these difficult patients by the house officers, renal fellows, and dialysis staff.
REFERENCES 1. Mailloux LU, Bellucci AG, Mossey RT, et al: Predictors of survival in patients undergoing dialysis. Am J Med 84:855862, 1988 2. Collins AJ, Hanson G, Umen A, et al: Changing risk factor demographics in end-stage renal disease patients entering hemodialysis and the impact on long-term mortality. Am J Kidney Dis 15:422-432, 1990 3. Schaefer K, Asmus G, Quellhorst E, et al: Optimum dialysis treatment for patients over 60 years with primary renal disease. Survival data and clinical results from 242 patients treated either by hemodialysis or hemofiltration. Proc Eur Dial Transplant Assoc Eur Ren Assoc 21 :510-523, 1984 4. Brunner FP, Brynger H, Chaillah S, et al: Renal replacement therapy in patients with diabetic nephropathy, 19801985. Report from the European Dialysis and Transplant Association Registry. Nephrol Dial Transplant 3:585-595, 1988 5. Hull AR, Parker TF III: Introduction and summary: Proceedings from the Morbidity, Mortality and Prescription of Dialysis Symposium. Dallas, TX 1989. Am J Kidney Dis 15:375-383, 1990 6. Shapiro FL, Umen A: Risk factors in hemodialysis patient survival. Am Soc Artif Intern Organs J 6:167-184, 1983 7. ESRD Network 25 Annual Reports for 1984,1985, New York, NY 8. Bryan CA, The National Dialysis Registry: Final Report 6/67-8/76. Artificial Kidney-Chronic Uremia Program. Bethesda, MD, National Institutes of Arthritis, Metabolism and Digestive Diseases, National Institutes of Health, 1976 9. Perry RJ, Griffiths W, Dextraze P, et a1: Elevated nicotine levels in patients undergoing hemodialysis. A role in cardiovascular mortality and morbidity? Am J Med 76:241-246, 1984 10. Degoulet P, Legrain M, Reach I, et al: Mortality risk factors in patients treated by chronic hemodialysis. Report of the Diaphane Collaborative Study. Nephron 31:101-110, 1982 II. Khauli RB, Steinmuller DR, Novick AC, et al: A critical look at survival of diabetics with end stage renal diseasetransplantation versus dialysis therapy. Transplantation 41: 598-602, 1986 12. Kindler J, Sieberth HG, Hahn R, et al: Does atherosclerosis caused by dialysis limit this treatment? Proc Eur Dial Transplant Assoc 19:168-174,1982 13. Verger C, Faller B, Ryckelynck JP, et al: Bilan de cinq annees de dialyse peritoneale continue ambulatoire (DPCA)Etude multicentrique de 345 patients. Nephrologie 6:1-6, 1985
14. Nissensen AR, Gentile DE, Soderblom RE, et al: Morbidity and mortality of continuous ambulatory peritoneal dialysis: Regional experience and long-term prospects. Am J Kidney Dis 7:229-234, 1986 15. Lowrie EG, Lazarus JM, Hampers CL, et al: Cardiovascular disease in dialysis patients. N Eng! J Med 290:737738, 1974 16. Lindner A, Charra B, Sherrard DJ, et al: Accelerated atherosclerosis in prolonged maintenance hemodialysis. N Eng! J Med 290:697-701, 1974 17. Lazarus JM, Lowrie EG, Hampers CL, et al: Cardiovascular disease in uremic patients on hemodialysis. Kidney Int 7:167-175,1975 18. Bagdade JD, Albers J: Plasma high-density lipoproteins concentrations in chronic hemodialysis and renal transplantation patients. N Eng! J Med 296:1436-1439, 1977 19. Lundin AP III, Adler AJ, Feinroth MV, et al: Maintenance hemodialysis-Survival beyond the first decade. JAMA 244:38-40, 1980 20. Burton PR, WallisJ: Selection-adjusted comparison of life-expectancy of patients on continuous ambulatory peritoneal dialysis, hemodialysis, and renal transplantation. Lancet 1:1115-1119, 1987 21. Eggers PW: Mortality rates among dialysis patients in Medicare's end-stage renal disease program. Am J Kidney Dis 15:414-421, 1990 22. Port FK: Mortality and causes of death in patients with end-stage renal failure. Am J Kidney Dis 15:215-217, 1990 23. Port FK, Garcia JR, Held PJ, et a1: Changes in the causes of death among ESRD patients in the US. Proc Am Soc Nephrol 22:95A, 1989 (abstr) 24. United States Renal Data System: USRDS 1990 Annual Data Report. U.S. Department of Health and Human Services. Bethesda, MD, The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, August 1990 25. Held PJ, Garcia JR, Blagg CR, et al: Trends in mortality rates among dialysis and transplant patients in the US. Proc Am Soc Nephrol 22:5IA, 1989 (abstr) 26. Lee ET: Statistical Methods for Survival Data Analysis. Belmont, CA, Lifetime Learning Publications, 1980 27. WestIie L, Umen A, Nestrid S, et al: Mortality, morbidity, and life satisfaction in the very old dialysis patient. Trans Am Soc ArtifIntern Organs 30:21-30, 1984 28. Husebye DG, WestIie L, Styrvoky n, et al: Psychological, social and somatic prognostic indicators in old patients undergoing long term dialysis. Arch Intern Med 147:19211924, 1987 29. Kjellstrand CM, Koppy K, Umen A, et al: Hemodialysis of the elderly, in Oreopoulos DG (ed): Proceedings of the International Symposium on Geriatric Nephrology, To-
CAUSES OF DEATH IN DIALYSIS PATIENTS
ronto 1985. The Hague, The Netherlands, Nijhoff, 1986, pp 135-145 30. Kjellstrand CM, Logan GM: Racial, sexual and age inequality in chronic dialysis. Nephron 45:257-263, 1987 31. Eggers PW, Connerton R, McMullan M: The Medicare experience with end-stage renal disease: Trends in incidence, prevalence, and survival. Health Care Financ Rev 5:69-88, 1984 32. Weller JM, Wu SCH, Ferguson WC, et al: End-stage renal disease in Michigan. Am J NephroI5:84-95, 1985 33. Collins A, Umen A, Odar-CederliefI, et al: Changing death patterns with time on chronic dialysis [CHDJ-lnfection is the specific lethal complication ofCHD. Proc Am Soc NephroI25:94A, 1988 (abstr) 34. Held PJ, Levin NW, Bovbjerg RR, et al: Mortality and duration of hemodialysis treatment. JAMA 265:871-875, 1991
335 35. Mossey RT, Sondheimer J: Listeriosis in patients with long term hemodialysis and transfusional iron overload. Am J Med 79:397-400, 1985 36. Husebye DG, Kjellstrand CM: Old patients and uremia: Rates of acceptance to and withdrawal from dialysis. lnt J Artif Organs 10: 166-172, 1987 37. Neu S, Kjellstrand CM: Stopping long-term dialysisAn empirical study of withdrawal oflife supporting treatment. N Eng! J Med 314:14-20,1986 38. Roberts JC, Kjellstrand CM: Choosing death. Withdrawal from chronic dialysis without medical reason. Acta Med Scand 223:181-186, 1988 39. Kjellstrand CM: Giving life-Giving death. Ethical problems of high technology medicine. Acta Med Scand (suppl) 725:1-88,1988 40. Kircher T, Nelson J, Burdo H: The autopsy as a measure of accuracy of the death certificate. N Eng! J Med 313: 1263-1269, 1985
I
N RECENT YEARS, the widespread availability of dialysis for the treatment of end-stage renal disease has resulted in a changing patient population. In our program, as in others, more older patients with multiple medical problems have begun dialysis during the last 8 years. I -4 Shorter survival on dialysis is recognized to occur in patients with diabetes mellitus, renal vascular disease, and advancing age. 1,2,5,6 Previous studies in dialysis patients have found that cardiac disease is the most frequent cause of death. 4,7-14 This is not surprising, since previous studies have linked long-term dialysis to accelerated atherosclerosis and cardiac mortality.3,15-25 These studies are largely based on registries and often involve multiple centers with different patterns of practice. The Division of Nephrology at North Shore University Hospital has maintained a clinical database for all dialysis patients over the age of From the Division of Nephrology and Hypertension, Department of Medicine, and the Division of Biostatistics, Department of Research, North Shore University Hospital-Cornell University Medical College; andfrom the Departments of Medicine and Public Health, Cornell University Medical College, Manhasset, NY. Presented in part at the 22nd Annual Meeting of the American Society of Nephrology, Washington, DC, December 1989. Address reprint requests to Lionel U. Mailloux, MD, FACP, Division of Nephrology and Hypertension, North Shore University Hospital-Cornell University Medical College, 300 Community Dr, Manhasset, NY 11030. © 1991 by the National Kidney Foundation, Inc. 0272-6386/91/1803-0005$3.00;0 326
15 since the inception of the program in January 1970. 1 We have reviewed our database to calculate annual gross mortality rates, to determine the changing pattern of mortality, and to evaluate the risk factors for mortality on dialysis. We report our 16-year experience. MATERIALS AND METHODS
Patient Population From 1970 through 1985. demographic and clinical data of all 532 patients with end-stage renal disease aged above IS years and surviving on dialysis over 90 days were kept on a computerized database. Table I shows median age and the renal diagnoses by the years the dialysis treatments were started. The North Shore dialysis unit is based in a tertiary care University Hospital in a suburban community. In our community, over 98% of the patients have health insurance before starting dialysis. The same physicians have been caring for these patients since 1970. All dialysis treatments are performed using generally accepted protocols to maintain normal blood pressures, dry weights, and a sense of well-being. Our hemodialysis prescription has consisted of three treatments weekly, with blood flows of at least 250 mL/min dependent on dialyzer membrane, dialysate flows of 500 mL/min, and 4.0 to 5.0 hours' duration. No "rapid dialysis" protocol has been used in our units. Intermittent peritoneal dialysis treatments were performed using 2-L exchanges for 20 hours biweekly. The patients on continuous ambulatory peritoneal dialysis (CAPD) performed at least four 2-L exchanges daily. None of the patients received erythropoietin. To evaluate patients who switched from one form of dialysis treatment to another, we arbitrarily adopted the following convention. If a patient ended one form of treatment (ie, switched from hemodialysis to CAPD), the patient was "transferred alive" from the hemodialysis group and entered as a new patient into the
American Journal of Kidney Diseases, Vol XVIII, No 3 (September), 1991: pp 326-335
327
CAUSES OF DEATH IN DIALYSIS PATIENTS
Table 1. Median Age, Renal Diagnosis, and Year Started Start Years
1970-1973 (n = 63)
Median age (yr) Diagnosis (%) Chronic glomerulonephritis Diabetes mellitus Renal vascular Polycystic kidneys Other'
1974-1977 (n = 152)
1977-1981 (n = 135)
1982-1985 (n = 182)
47
52
54
60
41.3 6.4 7.9 11.1 33.3
31.0 18.4 7.9 9.2 33.5
22.2 29.0 6.7 7.4 34.7
24.2 25.0 16.5 9.3 25.0
, Other renal diagnoses include transplant rejection, amyloidosis, tubulointerstitial nephritis, chronic pyelonephritis, obstructive uropathy, collagen vascular disease, miscellaneous renal disorders, and malignant hypertension.
CAPD group as if the patient had just started CAPD treatments. A specific renal diagnosis was entered into the database for each patient starting dialysis, and these diagnoses were then grouped into five categories: chronic glomerulonephritis, polycystic kidney disease, diabetes mellitus, renal vascular disease, and "other" as previously described. l Throughout the 20-year North Shore University Hospital program, the diagnosis of diabetic renal disease has been made on the basis of fasting hyperglycemia, retinopathy, nephrotic syndrome, hypertension, and renal insufficiency, regardless of whether patients were taking insulin. Large kidneys and, in some cases, renal biopsy material were available to further assist with the diagnosis. Therefore, we feel that stratifying diabetic patients into type I and type 2 would be somewhat arbitrary, so all diabetics with chronic renal failure are considered together, providing they meet the above clinical criteria. A diagnosis of renal vascular disease was made in patients with clinical evidence of systemic atherosclerotic vascular disease by a combination of imaging findings, such as a positive renal angiogram or suggestive radionuclide studies with unequal renal size, low level proteinuria « 1.0 g), and benign urinary sediment. Although these patients had hypertension, none had malignant hypertension with papilledema or end-organ damage. The diagnosis of malignant hypertension was reserved
for those patients with a hypertensive crisis, grade IV retinopathy, and end-organ damage. The cause of death was established by the nephrologists caring for the patient at the time of death. The 15 causes of death were compressed into six categories: cardiac, infectious, withdrawal, sudden, vascular, and "other." The same definition for each cause of death has been in use throughout the study period and is presented in Table 2. Sixty-seven (30.2%) of the deceased underwent an autopsy, with the clinical diagnosis confirmed in all but two cases.
Statistical Methods All statistical analyses were performed on a mainframe computer with the SAS software package (SAS-Institute, Cary, NC). The time until death was measured as the time from the start of the first dialysis until death. The annual gross mortality rate for a year, GMRy, was calculated by the formula: GMRy = n deceasedy/(Jan Iy n/2 + Dec 31y n/2), where Jan Iy n/2 is half the number of dialysis patients alive on Jan Iyand similarly for Dec 31y n/2, and y is the year of interest. Cause-specific survival distributions, corresponding to the six cause of death groupings, were estimated using the productlimit method. 26 A patient's survival time was considered censored for a given cause of death if the patient was still alive,
Table 2. Definitions of the Causes of Death Infectious Withdrawal Cardiac
Sudden Vascular Other
• Death either from sepsis with proven bacteremia or as the sequelae of acute fulminant hepatitis. • Death with manifestations of uremia because of withdrawal from dialysis. Underlying medical conditions should not have been active, leading to rapid deterioration with imminent death. • Death due to the manifestations of a well-documented acute myocardial infarction (typical chest pain associated with electrocardiographic changes and enzyme elevations) or from overt congestive cardiac failure. • Sudden death without explanation, but not related to a preexisting cardiac condition or in a patient with known hyperkalemia. • Death as the result of either a cerebrovascular aCCident, mesenteric artery thrombosis, ruptured aneurysm, or generalized ischemic ("low flow") state. • All other causes of death including respiratory failure, hyperkalemia, gastrointestinal bleed, and other miscellaneous causes (trauma, etc).
328
MAILLOUX ET AL
WITHDRAWAL
21%
OTHER
9%
VASCULAR
CARDIAC
15%
8% SUDDEN
10%
died of another cause other than the one under study, received a transplant, or was transferred alive to another unit. For example, when analyzing cardiac deaths, the survival time of a patient who died of infections was considered censored. Survival distributions for different patient groupings, eg, startyear groups, age groups, renal diagnoses, etc, were compared using the log rank test. Cause-specific multivariate analyses were performed using the Cox proportional hazards regression model. Variables that proved to be either significant on univariate analysis or were believed, a priori, to be prognostic for survival, were entered into the model and eliminated using the backward elimination procedure. Renal diagnosis and start-year groupings were transformed into dummy variables for this analysis. (Dummy variables were used for start-year groups, even though they were ordinal in nature, because we could not assume a linear, or even a monotonic, relationship between the log hazard ratio and these ordinal variables.) A total of seven possible prognostic variables were introduced into the model: startyear group 1970 to 1973,start-yeargroup 1974 to 1977, startyear group 1978 to 1981, diabetes mellitus, chronic glomerulonephritis/polycystic kidney disease, vascular disease, and age at start of dialysis. Start-year group 1982 to 1985 and "other" renal diagnoses were implicitly entered into the model as a result of the dummy variable scheme. Cause-specific survival time was the dependent variable. A variable was deleted from the model if it did not meet the 5% significance level. To summarize the results of the Cox regression model, relative risks were computed relative to 1982 to 1985 for start years, age below 40 years, and "other" for renal diagnosis. The hazard function was used to describe the likelihood of dying just after time t, given that the patient had survived through time t. The hazard function plot provided a visual picture of the risk of all deaths and cardiac death over time. However, we recognize that the relatively small number of deaths in any specific cause group may limit formal inference from these curves.
Fig 1. Distribution of all 222 deceased dialysis patients presented as a percentage of the total:
RESULTS
There were 532 chronic dialysis patients over the age of 15 years between January 1, 1970 and December 31,1985, who survived at least 90 days on dialysis. There were 293 males, 239 females, 478 whites, 47 blacks, and 7 of other races. There were 425 patients on in-center hemodialysis, 52 on home hemodialysis, 46 on CAPD, and 9 on in-center peritoneal dialysis. Nine patients switched modality (3 from in-center hemodialysis to CAPD, 3 from CAPD to in-center hemodialysis, 2 from in-center hemodialysis to intermittent peritoneal dialysis, and 1 CAPD to home hemodialysis); and 77 patients were transferred out of the dialysis groups alive because of renal transplantation. Two hundred twenty-two (41.7%) of the 532 patients died through 1988. The death rate was lowest in the 52 home dialysis patients (7.7%) and highest (44%) in the in-center peritoneal dialysis patients. The causes of death, in order of frequency, were infections, withdrawal, cardiac, sudden death, vascular, and other (Fig 1). Visual inspection of crude cause-specific death rates compared across start-year groupings showed an increased risk of cardiac death from 5.6% in 1970 to 1973 to 25.6% in 1982 to 1985 (Table 3). There was a decrease in deaths due to infections from 44.4% in 1970 to 1973 to 21.6% in 1982 to 1985. To adjust for the shorter follow-
329
CAUSES OF DEATH IN DIALYSIS PATIENTS
Table 3. Causes of Death and Start-Year Groups Start-Year Groups
1970-1973
1974-1977
1978-1981
1982-1985
Causes of Death
(%)
(%)
(%)
(%)
Infectious Withdrawal Cardiac Sudden Vascular Other
44.4 13.9 5.6 16.7 11.1 8.3
42.9 26.0 9.1 7.8 6.5 7.8
36.2 20.7 19.0 10.3 5.2 8.6
21.6 19.6 25.6 9.8 11.8 11.7
n = 36
n = 77
n = 58
Deceased
up of the later start-year groupings (1978 to 1981 and 1982 to 1985), especially in the last grouping where follow-up time was 5 years or less, the analysis was redone using only those cases with 5 years or less of follow-up. The same trend was observed: an increase in cardiac deaths and a decrease in deaths due to infections. The same trend was also observed if follow-up was restricted to 3 years or less, as well as to 2 years or less. The relationship between age at start of dialysis and causes of death is depicted in Fig 2. Fifty-six percent of the patients over 61 years of age died during the time of study, while 79% of those entering dialysis below age 40 were still alive. These younger patients had a higher death rate from infection, while deaths due to withdrawal from dialysis increased significantly in patients above 61 years of age. The patients withdrawing from 50
n
=
51
dialysis were older at death, with a median age of69.6 years, compared with the five other causes of death (P < 0.0018, Kruskal-Wallis test) (Table 4). In 1970, the first year of the program, there were only two patients at risk and no deaths. For the years 1971 through 1988, the overall dialysis gross mortality rate was 14.3%, with a range of 0% to 19%. Dialysis mortality rates for our center from 1970 to 1988 and for the United States from 1979 to 1988 are presented in Table 5. The hazard function for death due to all causes increased during the first 3 years and peaked at 4 years, with a subsequent decrease and leveling off between years 5 through 10, at which time a second peak was observed at 11 years (Fig 3). Visual inspection of the cause-specific hazard functions for the causes of death showed that the
Percent
40
30
20 10
0
Fig 2. Different causes of death compared by age at start of dialysis.
-
=61
am
CARDIAC
CJ
OTHER
YRS
330
MAILLOUX ET AL Table 4. Causes of Death Versus Age at Death* Age (yr) n
Infectious Withdrawal Cardiac Sudden Vascular Othert
81 47 33 23 18 20
Mean ± SO
Median
Range
± ± ± ± ± ±
60.7 69.6 64.8 60.2 62.7 52.0
24-84 15-85 43-80 32-84 38-77 23-79
57.7 65.3 63.5 58.7 60.3 53.0
1.4 2.1 1.9 2.5 2.8 4.0
• P < 0.0018, Kruskal-Wallis test.
t
Other deaths include hyperkalemia in 10 patients, respiratory failure in 4, gastrointestinal bleeding in 1, and miscellaneous causes in 5.
first peak was composed mainly of cardiac and infectious deaths, while the late peak was withdrawal and infectious deaths. Univariate cause-specific survival analyses yielded nearly identical results to those obtained using the multivariate Cox proportional hazards regression model. Therefore, results for only the Cox model are reported. The Cox proportional hazards regression model identified several variables that were significantly associated with the risk of death for each cause of death grouping and for all deaths. Table 6 summarizes the significant variables for cardiac, infectious, withdrawal, and all causes of death. The risk of a cardiac death increased steadily as start year increased from 1970 to 1985 and with advancing age. The risk of cardiac death was significantly lower in patients with chronic glomerulonephritis or polycystic kidney disease than for those patients with diabetes mellitus, vascular, or other renal diagnoses. The hazard function for cardiac deaths showed that there was an early peak at 48 months, a decreasing risk through 8.5 years, and no deaths thereafter (Fig 4). There was a significant and steady increase in the risk of death from infections with increasing age at the start of dialysis (Table 6). For infectious causes of death, there was a steady decrease in the risk of death across the start years (Table 3). Patients with chronic glomerulonephritis or polycystic kidney disease showed a decreased risk of infectious deaths (Tables 6 and 7). The oldest patients were at significantly increased risk of death due to withdrawal from dialysis (Tables 4 and 6). However, the risk of death due to withdrawal was not different for any start-
year group and was lower in patients with polycystic kidney disease and chronic glomerulonephritis and other renal diagnoses compared with diabetes and renal vascular disease (Table 6). Patients with diabetes mellitus and renal vascular disease accounted for 52% of all the withdrawals from dialysis (Table 7). For all causes of death, it was observed that the start-year groups 1970to 1973, 1974 to 1977, and 1982 to 1985 shared the same risk of death. The start-year group 1978 to 1981 had a lower risk of death than the other three start-year Table 5. Gross Mortality Rates in Dialysis Patients by Year: 1970 Through 1988
Year
NSUH-CUMC (%)
US (%)
1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988
0 13.3 5.6 17.4 16.4 17.2 16.8 17.4 11.8 18.8 11.3 10.8 12.6 17.8 16.1 19.0 15.4 17.9 17.3
NA NA NA NA NA NA NA NA NA 20.6 20.9 20.5 20.1 20.6 20.7 22.2 22.8 23.4 24.3
Abbreviation: NA, data not available.
331
CAUSES OF DEATH IN DIALYSIS PATIENTS HAZ""" 0.030
o.oz. O.OZII 0.024 0.OZ2 O.OZO
0.011 0.0111 0.01" O.OlZ
0.010 0.0011 0.0011 0.004 O.OOZ
Fig 3. Hazard function for death due to all causes presented in 12-month intervals.
....~-,-~....~............~......,...,...,...,~,..............,,........~--r:=:;:::;:::,........
0.000 'r~,..............,,....
--'HS 011 DIALYSIS
groups. Risk of death for all causes of death was once again associated with increasing age. Each of the four renal diagnostic groups carried with it a different risk of death, with diabetes mellitus having the highest risk, followed by renal vascular disease, other, and then polycystic kidney disease and chronic glomerulonephritis.
DISCUSSION
Much attention has been devoted to survival of dialysis patients, but few analyses of the causes of death have been published. Most of the previously published data have been derived from patient registries where detailed clinical information was often lacking. The current study was
Table 6. Associated Risk of Death for Each Cause of Death: The Cox Proportional Hazards Regression Model Causes of Death Variable
Cardiac
Infectious
Withdrawal
All
1.00 2.39t 5.70t
1.00 1.65t 2.72t
1.00 3.46§ 11.94§
1.00 1.97§ 3.90§
0.18' 0.23t 0.32' 1.00
2.03' 1.67' 1.00 1.00
1.00 1.00 1.00 1.00
1.00 1.00 0.66t 1.00
1.00 1.00 0.21t 1.00
1.00 1.00 0.35§ 1.00
3.00t 3.08t 1.00 1.00
1.93+ 1.63' 0.52+ 1.00
Age
60 yr Start years
1970-1973 1974-1977 1978-1981 1982-1985 Renal diagnosis Diabetes mellitus Renal vascular PKD/CGN Other
NOTE. Entries and corresponding p values represent the relative risk of cause specific death, using as reference levels
1982 to 1985 for start years, less than 40 years for age at start, and "other" for renal diagnoses. These reference groups are all labeled with a relative risk of 1.00. Other factors shown with a relative risk of 1.00 had a risk of death that did not differ significantly from the reference level. Abbreviation: PKD/CGN, polycystic kidney disease and chronic glomerulonephritis . • p < 0.05. t p < 0.01.
+
P < 0.001. § P < 0.0001.
MAILLOUX ET AL
332
Table 7. Causes of Death and Renal Diagnoses
performed to investigate the clinical parameters that were indicative of long or short survival in patients with end-stage renal disease. Our definitions for the initial renal diagnoses 1 and causes of death were set 20 years ago and have been strictly adhered to throughout the program's existence. Furthermore, the clinical diagnoses for the cause of death were substantiated in one third of the patients by autopsies, being confirmatory in more than 97% of them. Several important findings emerged from the analysis. The dialysis patient population has aged significantly over the past 20 years, with the median starting age in our program having increased from 47 years in the 1970 to 1973 start group to 60 years in 1982 to 1985. 1 Not surprisingly, age has been shown to be a major risk factor for shorter survival on dialysis. 1,2,5,6,24,25,27-31 In addition, the number of patients with underlying diseases that are generally recognized as having poor prognosis (ie, diabetes mellitus, vascular disease) has increased in recent years. 1,5,7,21,25,32,33 Overall, the most common cause of death was infectious, followed by withdrawal from dialysis, cardiac, sudden, vascular, and other (Fig 1). Much attention has been paid to possible causes of the increasing dialysis mortality rates in the United States, including the changing dialysis patient population, demographics, renal diagnoses, and the dialysis prescription,z,5,21.25 We have not decreased our dialysis time nor changed our dialysis prescription despite an increased use of more permeable membranes and larger surface area dialyzers in our unit. Since our patient trends have paralleled the national figures, we feel that
Diagnoses
Deceased (n) Causes of death (%) Infectious Withdrawal Cardiac Vascular Sudden Other
CGN
PKD
Other
Diabetes
RVD
40
10
76
63
33
35 20 10 8 10 17
40 10 0 20 20 10
45 18 16 4 6 11
35 22 16 8 16 3
21 30 21 16 6 6
Abbreviations: CGN, chronic glomerulonephritis; PKD, polycystic kidney disease; diabetes, diabetes mellitus; RVD, renal vascular disease.
maintaining longer dialysis treatments pver the years may have resulted in our lower gross mortality rates. The importance of maintaining dialysis treatment time has recently been suggested by Held et al. 34 The risk of death as observed from the hazard function demonstrated two major peaks, one each at 48 and 132 months (Fig 3). The early peak consisted largely of patients dying of cardiovascular and infectious causes; the late peak was composed mainly of withdrawal from dialysis or infection. Although previously not recognized, the late peak of mortality is not completely unexpected, since long-term dialysis may lead to other poorly understood complications including secondary amyloidosis, neuropathy, bone disease, etc. Undoubtedly, there is the possibility that preexisting disease may contribute to the early
HAZARQ
0.020
0.018 0.016 0.014
0.012 0.010 0.008 0.006 0.004
0 . 0 0 2 l _ -__
12
2.
36
.8
60
72
14
95
108
MONTHS ON DIALYSIS
120
132
...
155
168
180
Fig 4. Hazard function for cardiac deaths presented in 12-month intervals.
CAUSES OF DEATH IN DIALYSIS PATIENTS
mortality rate. Since patients were not exhaustively studied at the outset for cardiac disease, eg, cardiac catheterization, it is not possible to exclude that preexisting disease contributed to mortality. We also used the common convention of excluding patients who died before the 90th day of dialysis, eliminating some early deaths and acute renal failure, although time to death was calculated from the first dialysis treatment. Infection was the most common cause of death throughout the study period and is consistent with the data of Collins et al. 33 None of the infectious deaths in our population were from arteriovenous access or graft infections. Beyond 6 years of dialysis, infections were the most common cause of death in patients starting dialysis under the age of 60, and were equal to the number of deaths caused by withdrawal from dialysis in patients over 61 years of age (Fig 2). Infectious deaths are generally due to common pathogens, but may also occur as a result of unusual pathogens. 35 Withdrawal from dialysis was our second most frequent cause of death, accounting for 29% of all deaths in patients over 61 years of age. Together, withdrawal from dialysis and infection were the most common causes of death in the oldest patient group. More than half of the withdrawals were in patients over 61 years of age at the time of entry to dialysis (Fig 2). The proportion of deaths due to infections appears to decrease in patients over 61 years of age as there is an increase in deaths by withdrawal. The observation that withdrawal is not an uncommon cause of death is supported by Neu and Kjellstrand who found a large proportion of withdrawals in diabetic dialysis patients. 36-39 An older starting age, diabetes mellitus, and renal vascular disease seemed to be important risk factors for this cause of death. Nine of the 47 withdrawals had a malignancy and six patients had significant malnutrition with continuing catabolism. Withdrawal from dialysis has not been discussed frequently in the medical literature, but probably occurs more often than reported. Social and cultural reasons, as well as attitudes toward living and dying, may explain the higher rates of withdrawal in certain populations. 36-39 Dialysis patients who were otherwise in good physical and mental health did not withdraw from dialysis. The withdrawals occurred in pa-
333
tients with systemic diseases, mostly diabetes and advanced age. These were debilitated individuals who would not have been considered to be in immediate danger of dying. Cardiac disease was the third most common cause of mortality. Although previous studies have emphasized cardiac disease as the leading cause of mortality, cardiac mortality was not the major cause of death in our patient population (Fig 1, Table 3). The reasons for this difference are not entirely apparent, but the widespread treatment of hypertension and occlusive cardiovascular disease by aggressive medical and surgical approaches in our patient population may be contributing factors. However, lethal cardiac events did occur in 15% of the patients. Although there has been an increase in cardiac mortality in the later start-year groups, the pattern of these deaths has not changed. Generally, these events took place within the first 4 years (Fig 4) and were the major contributor to the early mortality. Patients in the oldest age group had more cardiac deaths than other patients (Tables 4 and 6). The observation that a large portion of cardiac deaths occurred early suggests that the cardiac diseases predated the onset of chronic dialysis and were not caused by dialysis (Fig 4). These data are supported by some studies,4,6,9,IO,12,19,27,32,33 but differ from others. 3,1l,12,15.18,22 Early cardiac deaths formed a significant portion of deaths in the patients reported by Lundin et al 19 and Collins et al. 33 Myocardial ischemia and infarction have been reported as the most frequent causes of lieath in diabetics between 1980 to 1985 in the European Dialysis and Transplant Association Registry.4 Cardiac deaths have been the most frequent cause of mortality in a large CAPD program in France, and account for 52.4% of deaths in ESRD Network 4, 50% of deaths in ESRD Network 25, and 51% of deaths in 41 diabetic dialysis patients in Cleveland.4,7,1l-14,3~,34 The methods to determine the cause of death in those studies have not been specifically defined. If data were obtained from death certificates, which often state cardiopulmonary arrest as the proximate cause of death, 1hcorrect conclusions could easily be drawn. 4o The systematiC evaluation of causes of death in our dialysis population has led to several interesting new findings. However, it is difficult to speculate whether our findings in a surburban
334
MAILLOUX ET AL
population can be applied to other dialysis populations. Using the statistical approach adopted in our study, it will be possible to study prospectively larger populations of dialysis patients to determine factors that contribute to survival and mortality.
ACKNOWLEDGMENT We acknowledge the on-going support that Lawrence Scherr, MD, MACP, Director of the Department of Medicine, has provided through the years. We also appreciate the excellent care given these difficult patients by the house officers, renal fellows, and dialysis staff.
REFERENCES 1. Mailloux LU, Bellucci AG, Mossey RT, et al: Predictors of survival in patients undergoing dialysis. Am J Med 84:855862, 1988 2. Collins AJ, Hanson G, Umen A, et al: Changing risk factor demographics in end-stage renal disease patients entering hemodialysis and the impact on long-term mortality. Am J Kidney Dis 15:422-432, 1990 3. Schaefer K, Asmus G, Quellhorst E, et al: Optimum dialysis treatment for patients over 60 years with primary renal disease. Survival data and clinical results from 242 patients treated either by hemodialysis or hemofiltration. Proc Eur Dial Transplant Assoc Eur Ren Assoc 21 :510-523, 1984 4. Brunner FP, Brynger H, Chaillah S, et al: Renal replacement therapy in patients with diabetic nephropathy, 19801985. Report from the European Dialysis and Transplant Association Registry. Nephrol Dial Transplant 3:585-595, 1988 5. Hull AR, Parker TF III: Introduction and summary: Proceedings from the Morbidity, Mortality and Prescription of Dialysis Symposium. Dallas, TX 1989. Am J Kidney Dis 15:375-383, 1990 6. Shapiro FL, Umen A: Risk factors in hemodialysis patient survival. Am Soc Artif Intern Organs J 6:167-184, 1983 7. ESRD Network 25 Annual Reports for 1984,1985, New York, NY 8. Bryan CA, The National Dialysis Registry: Final Report 6/67-8/76. Artificial Kidney-Chronic Uremia Program. Bethesda, MD, National Institutes of Arthritis, Metabolism and Digestive Diseases, National Institutes of Health, 1976 9. Perry RJ, Griffiths W, Dextraze P, et a1: Elevated nicotine levels in patients undergoing hemodialysis. A role in cardiovascular mortality and morbidity? Am J Med 76:241-246, 1984 10. Degoulet P, Legrain M, Reach I, et al: Mortality risk factors in patients treated by chronic hemodialysis. Report of the Diaphane Collaborative Study. Nephron 31:101-110, 1982 II. Khauli RB, Steinmuller DR, Novick AC, et al: A critical look at survival of diabetics with end stage renal diseasetransplantation versus dialysis therapy. Transplantation 41: 598-602, 1986 12. Kindler J, Sieberth HG, Hahn R, et al: Does atherosclerosis caused by dialysis limit this treatment? Proc Eur Dial Transplant Assoc 19:168-174,1982 13. Verger C, Faller B, Ryckelynck JP, et al: Bilan de cinq annees de dialyse peritoneale continue ambulatoire (DPCA)Etude multicentrique de 345 patients. Nephrologie 6:1-6, 1985
14. Nissensen AR, Gentile DE, Soderblom RE, et al: Morbidity and mortality of continuous ambulatory peritoneal dialysis: Regional experience and long-term prospects. Am J Kidney Dis 7:229-234, 1986 15. Lowrie EG, Lazarus JM, Hampers CL, et al: Cardiovascular disease in dialysis patients. N Eng! J Med 290:737738, 1974 16. Lindner A, Charra B, Sherrard DJ, et al: Accelerated atherosclerosis in prolonged maintenance hemodialysis. N Eng! J Med 290:697-701, 1974 17. Lazarus JM, Lowrie EG, Hampers CL, et al: Cardiovascular disease in uremic patients on hemodialysis. Kidney Int 7:167-175,1975 18. Bagdade JD, Albers J: Plasma high-density lipoproteins concentrations in chronic hemodialysis and renal transplantation patients. N Eng! J Med 296:1436-1439, 1977 19. Lundin AP III, Adler AJ, Feinroth MV, et al: Maintenance hemodialysis-Survival beyond the first decade. JAMA 244:38-40, 1980 20. Burton PR, WallisJ: Selection-adjusted comparison of life-expectancy of patients on continuous ambulatory peritoneal dialysis, hemodialysis, and renal transplantation. Lancet 1:1115-1119, 1987 21. Eggers PW: Mortality rates among dialysis patients in Medicare's end-stage renal disease program. Am J Kidney Dis 15:414-421, 1990 22. Port FK: Mortality and causes of death in patients with end-stage renal failure. Am J Kidney Dis 15:215-217, 1990 23. Port FK, Garcia JR, Held PJ, et a1: Changes in the causes of death among ESRD patients in the US. Proc Am Soc Nephrol 22:95A, 1989 (abstr) 24. United States Renal Data System: USRDS 1990 Annual Data Report. U.S. Department of Health and Human Services. Bethesda, MD, The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, August 1990 25. Held PJ, Garcia JR, Blagg CR, et al: Trends in mortality rates among dialysis and transplant patients in the US. Proc Am Soc Nephrol 22:5IA, 1989 (abstr) 26. Lee ET: Statistical Methods for Survival Data Analysis. Belmont, CA, Lifetime Learning Publications, 1980 27. WestIie L, Umen A, Nestrid S, et al: Mortality, morbidity, and life satisfaction in the very old dialysis patient. Trans Am Soc ArtifIntern Organs 30:21-30, 1984 28. Husebye DG, WestIie L, Styrvoky n, et al: Psychological, social and somatic prognostic indicators in old patients undergoing long term dialysis. Arch Intern Med 147:19211924, 1987 29. Kjellstrand CM, Koppy K, Umen A, et al: Hemodialysis of the elderly, in Oreopoulos DG (ed): Proceedings of the International Symposium on Geriatric Nephrology, To-
CAUSES OF DEATH IN DIALYSIS PATIENTS
ronto 1985. The Hague, The Netherlands, Nijhoff, 1986, pp 135-145 30. Kjellstrand CM, Logan GM: Racial, sexual and age inequality in chronic dialysis. Nephron 45:257-263, 1987 31. Eggers PW, Connerton R, McMullan M: The Medicare experience with end-stage renal disease: Trends in incidence, prevalence, and survival. Health Care Financ Rev 5:69-88, 1984 32. Weller JM, Wu SCH, Ferguson WC, et al: End-stage renal disease in Michigan. Am J NephroI5:84-95, 1985 33. Collins A, Umen A, Odar-CederliefI, et al: Changing death patterns with time on chronic dialysis [CHDJ-lnfection is the specific lethal complication ofCHD. Proc Am Soc NephroI25:94A, 1988 (abstr) 34. Held PJ, Levin NW, Bovbjerg RR, et al: Mortality and duration of hemodialysis treatment. JAMA 265:871-875, 1991
335 35. Mossey RT, Sondheimer J: Listeriosis in patients with long term hemodialysis and transfusional iron overload. Am J Med 79:397-400, 1985 36. Husebye DG, Kjellstrand CM: Old patients and uremia: Rates of acceptance to and withdrawal from dialysis. lnt J Artif Organs 10: 166-172, 1987 37. Neu S, Kjellstrand CM: Stopping long-term dialysisAn empirical study of withdrawal oflife supporting treatment. N Eng! J Med 314:14-20,1986 38. Roberts JC, Kjellstrand CM: Choosing death. Withdrawal from chronic dialysis without medical reason. Acta Med Scand 223:181-186, 1988 39. Kjellstrand CM: Giving life-Giving death. Ethical problems of high technology medicine. Acta Med Scand (suppl) 725:1-88,1988 40. Kircher T, Nelson J, Burdo H: The autopsy as a measure of accuracy of the death certificate. N Eng! J Med 313: 1263-1269, 1985
