Comparison of Continuous Ambulatory Peritoneal Dialysis

and Hemodialysis Patient Survival With Evaluation of

Trends During the 198051
Christopher B. Nelson, Friedrich K. Port,2 Robert A. Wolfe, and Kenneth E. Guire

whereas diabetic
HD mortality rates increased (P =
C.B. Nelson, F.K. Port, R.A. Wolfe, K.E. Guire, The Mich- 0.06). Among men had higher
diabetics, mortality
igan Kidney Registry, Ann Arbor, Ml rates than women (ITT RxHx; RR = I .22 to I .27; P <
-

F.K. Port, Department of Epidemiology, School of Public 0.001) and white patients had higher mortality rates
Health, University of Michigan, Ann Arbor, Ml than black patients (ITT-RxHx, PR = 1.34 to 1.44; P
< 0.001). Differences in mortality by sex and race
R.A. Wolfe, K.E. Guire, Department of Biostatistics.
were not found among nondiabetics, but mortality
School of Public Health, University of Michigan. Ann
Arbor, MI did increase significantly with age in all groups. For
nondiabetics, results were unchanged when RxHx
F.K. Port, Department of Internal Medicine, School of
censoring criteria were used, whereas for diabetics,
Medicine, University of Michigan, Ann Arbor, Ml
mortality rates for CAPD and HD users became more
(J. Am. Soc. Nephrol. 1992; 3:1147-1 155)
similar (P> 0.10). Median survival of diabetic CAPD
and HD patients differed by I 2 months (HD, 24 months
versus CAPD, 36 months) in the ITT model. The lower
ABSTRACT mortality rates of young diabetic and glomerulone-
To evaluate the mortality of continuous ambulatory phritic patients using CAPD may be due to differential
peritoneal dialysis (CAPD) patients relative to he- distribution of comorbid conditions and bias in the
modialysis (HD) patients, all Michigan residents 20 to selection of a dialytic modality.
59 yr of age who initiated therapy for ESRD during the Key Words: ESRD, patient survival, dialysis, continuous ambu-
1980s (N= 4,288) were studied. The study population latory peritoneal dialysis, hemodialysis
was stratified by primary renal diagnosis (glomeru-
lonephritis,
lyses
hypertension,
were conducted
diabetes,
within each
other), and ana-
group by Cox
O nly a few
tion around
years after
1 980, several
its
studies
widespread
from both the
introduc-

proportional hazards methods controlling for age,

United States and Europe described clinical popula-
race, sex, and year in which chronic dialysis was tions treated by continuous ambulatory peritoneal
initiated. Intent-to-treat (ITT) and treatment history dialysis (CAPD) and evaluated factors affecting tech-
(RxHx) censoring criteria were used. For patients with nique and patient survival ( 1 -6). In the batter part of
hypertension or other reported causes of ESRD, there the 1980s, long-term results from the National (U.S.)
was no significant difference in CAPD and HD patient CAPD Registry and the European Dialysis and Trans-
mortality (relative risk (RR) = 0.99 and 1.05, respec- plant Association (EDTA) registry were published as
tively). In the ITT analysis, both glomerulonephritic were long-term results from mubticenter studies done
(RR = 0.73; P = 0.10) and diabetic patients using in the United States, the United Kingdom, and Italy

CAPD experienced mortality rates lower than their (7- 1 1). These studies were effective in describing the
respective CAPD populations and their dialysis ex-
HD counterparts. Among diabetics, this difference
penience (types and causes of infection, hospitaliza-
ranged from a PR of 0.40 to 0.70, being lowest for
tion, patient and treatment survival rates) and gen-
younger diabetics and statistically significant (P
erated a consensus that CAPD was a satisfactory
0.05) for ages 20 to 52 yr. Evaluation of mortality therapy. although peritonitis rates and technique
trends showed a significant (P < 0.01) decrease in failure were recognized as major concerns.
diabetic CAPD mortality rates during the decade, More recently, clinical researchers have focused on
the mortality rates of CAPD and hemodlabysis (HD)
I Received March 30, 1992. Accepted July 1, 1992. patients to determine the relative efficacy of CAPD
2 Correspondence to Dr. F.K. Port, The Michigan Kidney Registry, 3 15 West Huron, (12-27). DescrIption of CAPD populations found
Suite 340, Ann Arbor, Ml 48103. them to have a greater number and more severe
1046-6673/0305-1 14750300/0
Journal of the American Society of Nephrology
comorbid conditions than HD populations. Yet, none
Copyright © 1992 by the American Society of Nephrology of these studies identified a statistically significant

Journal of the American Society of Nephrology 1147

Comparison of CAPD and HD Patient Survival

difference in patient survival rates, leading many and August 3 1 1 989, and who were using
. In-center
researchers to conclude that CAPD was a comparable HD or home CAPD at day 1 20 of ESRD. Patients
therapy to HD. receiving other modes of dialysis on day 1 20 of ESRD
Analyses of trend data for both HD and CAPD were excluded from the analysis (5%) as were pa-
patients have been presented by Wolfe et at. (25). tients who recovered renal function subsequent to
Although lacking detailed information on comorbid ESRD diagnosis (3%) 1321 on who had a prior trans-
conditions, their analysis of data from Michigan plant (0. 1 %). Patients beginning ESRD therapy be-
showed that, among young diabetics, patient survival fore January 1 1 980. were
. excluded because CAPD
was significantly better for the HD population, when was not widely available in Michigan before that
adjusted to 1980, and that no significant differences date.
existed for young, nondiabetic patients. However, Because no information was available regarding
they also found that, among diabetics, the mortality comonbid conditions present at the onset of ESRD,
rates of successive HD cohorts were increasing sig- the population examined for this analysis was re-
nificantly as the decade progressed, whereas for stnicted to persons 20 to 59 yr of age. This excludes
successive CAPD cohorts, the mortality rates were older patients who exhibit a higher prevalence and
decreasing slightly. As a result, differences In the severity of comorbid conditions and higher rates of
mortality rates of CAPD and HD cohorts Initiating natural mortality, as well as pediatric patients whose
therapy around the middle of the decade were found disease etiology differs from that of adults. Almost
to be statistically insignificant. half of incident ESRD patients in the state of Michi-
In this article, we examine the experience of the gan are between the ages of 20 and 59 yr.
Michigan ESRD population initiating In-center HD or Because there was an insufficient number of non-
home CAPD therapy between 1 980 and 1 989. We white/nonblack ESRD patients (<3% of the popula-
evaluated the survival of CAPD patients relative to tion) to draw meaningful conclusions from an analy-
HD patients using intent-to-treat (ITT) and treatment sis stratified beyond two race categories. only white
history (RxHx) censoring criteria. The ITT analysis and black patients are included in this study.
evaluates the experiences of patients given their
mode of therapy on day 1 20 of ESRD, Ignoring sub-
sequent changes In dialytic modality (28,29). Results Classification of Treatment Modality
of this analysis are useful for counseling on the Each patient was classified as using either in-cen-
subsequent mortality of patients who were using a ter HD or home CAPD on the basis of their experience
given dialytic modality at a specific point in the ESRD on day 1 20 of ESRD. This point was chosen to avoid
time course. The RxHx analysis evaluates the expe- the relatively large percentage of CAPD and HD pa-
riences of patients only while on the modality they tients who change their mode of dialysis during the
were using at day 1 20 of ESRD and might he used as first months of ESRD therapy. In Michigan, Wolfe et
an estimate of CAPD efficacy relative to HD. at. (25) have shown that during the first 4 months of
therapy, the proportion of HD patients changing their
METHODS mode of dialysis in any one month declines dramati-
cabby from over 4.5% in the first month to under 1.5%
The Michigan Kidney Registry in the fifth month. For patients receiving CAPD,
The Michigan Kidney Registry (MKR) contains in- there is an average monthly change of over 2% during
formation on all ESRD patients treated by dialysis or the first to fourth months, after which the percent
transplantation in the state of Michigan since 1973. change per month Is less than 2%.
Although no cormorbid information Is collected, the
primary renal diagnosis as well as detailed demo-
Statistical Methods
graphic and treatment information is available. A
detailed description of the Michigan ESRD population The Cox proportional hazards regression model
and an analysis of Michigan ESRD incidence by type was used to evaluate the mortality of CAPD patients
of diabetes have been published elsewhere (30,3 1). relative to HD patients (33). This model gives an
Identification of cases and data collection have been estimate, the relative risk (RR), of the death rate in
estimated to be over 99% complete and 97% accurate one group relative to the death rate In a baseline
(31). group while adjusting for the influence of other in-
dependent variables Included in the model. In the
following analyses, the mortality rates of CAPD pa-
Inclusion Criteria
tients are compared with those of HD patients with
The MKR dataset used in this analysis contains adjustment for age, race (white relative to black), sex
information on 4,288 Michigan residents who initi- (male relative to female), and year of first ESRD
ated therapy for ESRD between January 1 , 1980, (relative to 1989).

1148 Volume 3 ‘ Number 5 ‘ 1992

 Nelson et al

The point estimate of the RR Indicates the magni- transplant is attempted. Overall, mean length of fol-
tude of the effect. An RR greater than 1 indicates a low-up was 2 1 (±22) months (range. 4 to 113
higher rate of death as compared with the baseline months).
group, an RR less than 1 indIcates a lower rate, and The censoring of survival times for patients who
an RR equal to 1 Indicates no difference in the death changed their mode of therapy subsequent to day
rates of the comparison and baseline groups. Ninety- 1 20 of ESRD was done in two different ways. In the
five percent confidence intervals reflect both the ITT analysis, changes In dialytic therapy were ig-
magnitude and precision of the RR estimate. They nored. In the RxHx analysis, survival times were
are interpreted as the range of values that contains censored when the patient changed dialytic therapy.
the true RR, with only a 5% chance of it falling In a variation of the RxHx analysis, survival times
outside of this range (34). were censored 30 days after the change of therapy
Because the primary renal diagnosis variable did (RxHx + 30) with the reasoning that mortality within
not fulfill the proportional hazards assumption of the this window of time should be associated with the
model, we conducted analyses stratified by this vail- original classification of therapy. Various other au-
able. Therefore, statistical tests reflect the experl- thors have censored follow-up at changing of therapy
ence of patients within a specific diagnostic category (13,18,19,20,28), 60 days after changing of therapy
only. Statistical analyses were performed by the pro- (22,29), and ignoring changes of therapy (25).
cedure PHREG of SAS v6.06 (35).

RESULTS
Censoring Criteria: ITT and RxHx Analyses
Table 1 shows the number of deaths in the study
In each of the analyses, time was measured from population by censoring criteria and primary diag-
day 1 20 of ESRD until death, transplantation, loss nosis. Within each primary diagnosis, the greatest
to follow-up, or December 3 1 , 1 989, whichever came number of deaths occurred with the ITT censoring
first. By censoring at the time of transplant, we are criteria as a result of the longer period of follow-up.
modeling survival while on dialysis and before a Accordingly, the fewest deaths occurred with the

TABLE I . Number of deaths in the study population by censoring criteria and primary diagnosis

Glomerulonephritis Hypertension Diabetes Other

Censoring (N= 798) (N= 993) (N= 1458) (N= 1039)
Criteria0
N % N % N % N %

ITT 148 18.5 308 31.0 613 42.0 285 27.4

RxHx 121 15.2 266 26.8 501 34.4 231 22.2
RxHx+30 134 16.8 277 27.9 566 38.8 258 24.8

0 ITT. ignores changes of therapy: PxHx, censors at change of therapy; RxHx + 30, censors 30 days after change of therapy.

50

40

- 30
1)

0_ 20

10

0
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989
Year of ESRO Onset

ALL #{149} GLOM HYP #{176} DIAB OTHER

Figure 1 . Percentage of patients using CAPD at day 120 of ESRD, overall and among those with the specified cause of ESRD,
by year of ESRD onset. GLOM, glomerulonephritis; HYP, hypertension; DIAB, diabetes.

Journal of the American Society of Nephrology I 149

Comparison of CAPD and HD Patient Survival

RxHx criteria with censoring at change of therapy. varies from 46% in diabetics to 27% among hyperten-
Diabetics had the highest unadjusted mortality rates, sives, whereas among black patients, CAPD use does
and glomerubonephritics had the lowest. not vary widely across diagnostic groups. Overall,
approximately 1 9% of black patients use CAPD.
There are no significant differences in the proportion
Patient Population and Use of CAPD
of men and women who are using CAPD or in the
Figure 1 shows the percentage of dialysis patients mean age of CAPD and HD patients. Although there
using CAPD, overall and among those with the spec- are no clear trends, CAPD use does appear to vary by
ified primary diagnosis. by year in which therapy age.
was initiated. During the early 1 980s, there was a
steady increase in the proportion of patients using
Risk of Dying for CAPD Patients Relative to HD
CAPD. but since 1983, overall levels of CAPD use
Patients
have increased only slightly. By 1 989, 36% of all
dialysis patients chose this mode of therapy. The diagnosis-specific risk of death for CAPD pa-
Demographic characteristics of the study popula- tients relative to HD patients is shown in Table 4.
tion stratified by primary renal diagnosis are shown These risks were estimated with ITT censoring cr1-
in Table 2. The mean age of the population at initia- teria and are adjusted for age. race, sex, and, among
tion of therapy was 44 yr (± 1 1 yr:
range, 20 to 59 yr), diabetics only, year of ESRD onset (adjusted to 1989).
and one third to nearly one half of patients were Among nondiabetic groups (Table 4), differences
women. In contrast to the other patient groups. a between CAPD- and HD-adjusted mortality rates were
majority of patients with hypertensive ESRD (76%) marginal (gbomerubonephnitis, RR = 0.73: 95% con-
are black. Statistical analysis did not identify any fidence interval [CI]: 0.47 to 1 .07: P = 0. 10) or non-
significant demographic changes over the course of existent (hypertension. RR = 0.99: other, RR = 1.05)
the decade in the study population. and did not vary by age.
Differences between the diagnosis-defined groups Among diabetics, CAPD patients were shown to
in CAPD use are in part explained by the proportion have lower adjusted mortality rates than HD patients
of white patients using this mode of therapy (Table (RR = 0.40 to 0.70). As shown more clearly in Figure
3). Among white patients. the proportion using CAPD 2, this difference was greatest among younger pa-

TABLE 2. Demographic characteristics of the study population by primary diagnosis

. . . . All
Glomerulonephritis Hypertension Diabetes Other .
Diagnoses

Mean Age (±SD) 41 (12) 46 (10) 44 (1 1) 43 (1 1) 44 (11)

Race (% Black Patients) 32 76 33 30 42
Sex (% Female) 37 36 46 48 43
N 798 993 1458 1039 4288

TABLE 3. Percentage of patients using CAPD at day 120 of ESRD by age, race, sex, and primary diagnosis

.All
Glomerulonephritis Hypertension Diabetes Other
Diagnoses

Age (yr)
20-29 20 24 44 21 28
30-39 28 25 40 22 30
40-49 37 21 43 26 32
50-59 30 17 28 31 26
Race
Black 20 19 18 21 19
White 33 27 46 28 36
Sex
Female 30 21 36 24 29
Male 28 20 37 28 29
All Groups 29 21 36 26 29

1150 Volume 3 ‘ Number 5 . 1992

 Nelson et al

TABLE 4. Diagnosis-specific risk of death for CAPD relative to HD (Rx) and for age, race, and sex as estimated
by an ITT analysis

Covariate (Unit) Glomerulonephritis Hypertension Diabetes#{176} Other

Rx (CAPD versus HD) O.73b 0.99 0.40 - 0.70c 1.05

Age (per 10 yr) 1#{149}14d(HD); 1.37d (CAPD)
1.55d 1.35d 1.13
Race (White versus Black) 1.09 0.84 1.44d 0.87
Sex (Male versus Female) 1.07 0.95 1.22#{176} .93

0 Also adjusted for year of ESPD onset. with reference to 1989.

b p< 0.10.
C p< 0.05 for persons 52 yr of age and younger (see Figure 2).
d p< 0.01.

P<0.05.

1.20

1.00 HO

Q) 0.80
> .70
0) 0.60

0.40

0.20
20-29 30-39 40-49 50-59
Age at Onset of ESRD
Figure 2. Diabetics only: risk of death (and 95% Cl) for CAPD relative to HD by age at onset of ESRD and adjusted for race,
sex, and year of ESRD onset.

tients and was significant (P 0.05) through the age to 1.43: P 0.016). In nondiabetic patient groups,
of 52 yr. For CAPD patients 20 to 29 yr of age, there were no significant differences in mortality
mortality rates were 60% lower than those of HD associated with race and sex. These results changed
patients in the same age category (RR = 0.40: 95% CI little when the first three cohorts (1 980 to 1 982) of
= 0.23 to 0.70: P < 0.01). For patients 30 to 39 yr of patients were excluded from the analysis.
age. CAPD mortality rates were than 52%
HD lower The average mortality trends across the 1 980s
mortality rates (RR = 0.48: 95%to 0.76: P
CI = 0.31 showed no statistically significant pattern for diag-
< 0.0 1), for 40 to 49 yr olds, CAPD mortality rates nostic groups other than diabetics. For diabetic pa-
were 42% lower (RR = 0.58; 95% CI = 0.39 to 0.86: tients (Figure 3), the death rates of CAPD cohorts
P < 0.01), and for 50- to 59-yr-old CAPD patients, have been decreasing on average by 9% per year (RR
mortality rates were 30% lower (RR = 0.70: 95% CI = 0.91: 95% CI:0.85 to 0.96: P < 0.001), whereas
= 0.46 to 1.05: P 0.06). those of HD cohorts appeared to be increasing in the
Of the demographic variables (Table 4), older age middle of the decade but have fallen again recently.
was associated with an increased mortality rate in The overall linear trend for HD cohorts shows a slight
all diagnostic groups. However, among diabetics, increase of 4% per year (RR = 1 .04: 95% CI: 1 .00 to
mortality rates rose significantly fasten (P = 0.03) 1 .08: P 0.06).
among CAPD patients than among HD patients
(CAPD-RR: 1.37/10 yr of age: 95% CI:1.20 to 1.56: P
< 0.001: HD-RR: 1.14/10 yr of age: 95% CI:1.03 to
Comparison of Risks Estimated by ITT and RxHx
1 .26: P < 0.001). Diabetics also showed significant
Analyses
differences in mortality rates for white versus black Comparison of results from ITT and RxHx analyses
patients (RR = 1 .44: 95% CI: 1 .20 to 1 .72: P < 0.00 1) (Table 5) show that among gbomerubonephritic pa-
and males versus females (RR = 1 .22: 95% CI: 1.04 tients the difference In mortality of CAPD and HD

Journal of the American Society of Nephrology 1151

Comparison of CAPD and HD Patient Survival

CAPD
Chng.in#{241}.ckp.ryr=-9%
,._ Testof Ir.,tp = 0.001

0 \HD
Changeinriskp.ryr- +4%
TestofTr.nd.p < 0.06

1980-81 1982.83 1984-85 1986.87 1988.89

Year of ESROOnset

Figure 3. Diabetics only: change in CAPD and HD mortality risk by year of incidence, 1980 to 1989, and adjusted for age,
race, and sex. Each increment represents a 22% higher risk.

TABLE 5. Comparison of the diagnosis-specific risk of death for CAPD relative to HD (Rx) and for age, race,
and sex with ITT and RxHx censoring criteria0

Glomerulonephritis Hypertension Diabetesb Other

Factor Unit
ITT RxHx RxHx+30 ITT RxHx RxHx+3O ITT RxHx RxHx+30 ITT RxHx RxHx+30

Rx CAPD versus HD O.73c 0.72 O.65d 0.99 0.99 0.98 0.40 1.11 1.12 1.05 1.04 1.06
Age per lOyr 1.55 1.54 1.60 1.35 1.3& 1.34w 1.14’(HD) 1.37(CAPD) lIT 1.18 1.13d 1.12c 1.12d
Race (Whiteversus 1.09 0.97 1.12 1.19 1.17 1.20 1.44’ 1.34’ 1.42 0.87 0.79c 0.85
black)
Sex (Maleversus 1.07 1.17 1.21 0.95 1.05 1.01 1.22d 1.27 1.26t 0.93 0.97 0.93
female)

a For definitions of ITT, RxHx, and RxHx, see footnote to Table 1.

b Also adjusted for year of ESPD onset, baseline 1989.
C p< 0.10.
0 p< 0.05.

. p< 0.05 for persons 52 yr of age and younger.

I p< 0.01.

users emerges when follow-up Is censored 30 days For all groups and all censoring criteria, mortality
after the first change of therapy, RxHx + 30 (RR = increased significantly (P < 0.01) with age. but only
0.65: 95% CI 0.43 to 0.99; P < 0.05). When ITT among diabetics was there a significant (P < 0.05)
censoring criteria are used, the RR is marginally difference in mortality rates by race or sex.
significant (RR = 0.73: 95% CI = 0.47 to 1.07: P =
0.06), and when follow-up Is censored at change of
therapy (RxHx). the RR is statistically insignificant
(RR = 0.72: 95% CI 0.47 to 1 . 1 2, P = not signifi-
Survival and Hazard Rate Curves
cant). The difference in CAPD and HD patient survival
Among diabetic patients. RxHx estimates of the RR among gbomerubonephnitic and diabetic patients Is
show no significant difference between CAPD and illustrated in Figure 4. These curves were modeled
HD mortality rates but are contrastingly different with ITT censoring criteria and are adjusted to a 45-
from the ITT estimate, which showed a significantly yr-old white man with ESRD onset in 1 989. 45-yr-
(P 0.05) lower mortality rate for CAPD patients old white men being the most prevalent patients in
relative to HD patients up through the age of 52 yr. the study population.
For hypertensive and other patients. there was very Among diabetics, median survival for HD patients
little difference in the magnitude, direction, or sig- was 23.8 months, whereas median survival for CAPD
nificance of the three estimates. patients was 36. 1 months, a difference of 1 2 months

1152 Volume 3 ‘ Number 5 ‘ 1992

 ... .. :.,., L .. Nelson et aI

100
90
80
0)
C 70
>
.; 60

c 50
C
.2 40
t
a
g 20

10

0
4 6 8 1012141618202224262830323436384042444648505254565860

Months of ESRD

. GLOM-CAPD -0--- GLOM-HD #{149} DIAB-CAPD -#{176}- DIAB-HD

Figure 4. Survival of glomerulonephritic (GLOM) and diabetic (DIAB) patients using CAPD and HD. Curves are for a 45-yr-old
white man initiating therapy in 1989 and were generated by a Cox model with ITT censoring criteria. Follow-up began at
day 120 of ESRD.

in favor of CAPD patients. One-, 3-, and 5-yr survival stnicted the study population to patients between 20
rates for diabetic HD users were 80, 33, and 12%, and 59 yr of age. This age group is thought to have
respectively. Among diabetic CAPD users, 1 -, 3-, and fewer and less severe comorbid conditions than do
5-yr survival rates were 87, 50, and 27%. Among the older dialysis patients and constitutes about half
gbomerubonephritic patients using HD, 1 -. 3-. and 5- of the Michigan ESRD population incident during the
yr survival rates were 94. 72, and 57%, respectIvely, 1 980s. In this way. we reduced the influence of these
and for CAPD users, they were 95, 79. and 67%. unmeasured comorbid conditions on our analysis.
Other studies have reported patient selection based
on comorbid conditions ( 1 5, 1 7,36), and one study has
DISCUSSION
shown that the use of a dialytic modality can be
The MKR is a unique source of data that can be facility dependent (37). In addition, some preexisting
used In the evaluation of questions such as the ones comonbid conditions are significantly associated with
addressed here. Because It contains information on increased mortality rates (12,17-19.21,22,28,29).
all patients initiating dialysis in the state regardless In a previous analysis of data from the MKR, Wolfe
of the facility where they receive care or their type of et at. (25) described trends in diabetic CAPD and HD
insurance, this dataset avoids the biases that may mortality rates with ITT censoring criteria and raised
result when study participants are recruited from a the hypothesis that if these trends continued, mor-
single or selected facilities, from insurer databases, tality rates for young diabetic HD cohorts may be-
or change facilities. come higher than those for young diabetic CAPD
The MKR also lends itself to studies that require cohorts by the late 1 980s. The 1 99 1 USRDS Annual
barge numbers of participants. Although the most Report showed evidence of this at the national level.
efficient design for evaluating the relative efficacy of In an analysis of mortality rates by age, race, and
CAPD would be a prospective randomized clinical primary diagnosis, the USRDS reported that mortal-
trial, It is difficult to conduct such a study because ity rates tended to be higher for young HD than for
of the unique advantages each modality has to offer. young CAPD patients, with few exceptions (38).
For this reason, researchers are limited to the analy- Our ITT analysis of the Michigan ESRD population
sis of existing patient cohorts with special attention supports this hypothesis also, showing that during
being paid to potential biases in the data. A major the 1 980s mortality rates for diabetic CAPD cohorts
disadvantage of techniques used to control bias, such have decreased at a nearby constant rate whereas the
as restriction or stratification of the study popula- mortality rates of diabetic HD cohorts have been
tion, is the reduction in sample size and as a conse- increasing, albeit less consistently. As a result, by
quence. the loss of statistical power. The use of the 1989, diabetic CAPD patients 20 to 59 yr of age had
MKR database allows us to evaluate a barge number significantly lower mortality rates than diabetic HD
of patients and therefore maximizes the statistical patients Initiating therapy In the same year.
power of the analyses. Among nondiabetics in our study, there was evi-
Because the MKR has no information regarding dence that the mortality of patients using CAPD has
comorbid conditions present at the onset of ESRD, been better than or comparable to that of patients
we took advantage of its large sample size and re- using HD: patients with a renal diagnosis of gbomer-

Journal of the American Society of Nephrology 1153

Comparison of CAPD and HD Patient Survival

ubonephritis who were using CAPD had a mortality mode of therapy at the fourth month of ESRD. Among
rate 27% below that of their counterparts using HD: young diabetics, bong-term mortality was shown to
among patients with ESRD attributed to hyperten- be lower for those patients using CAPD on day 120
sion or other causes, the relative risks were nearly of ESRD but was not significantly different, and
1 .0. The marginal significance of the relative risk might be slightly higher, during the time from day
among glomerulonephnitic patients is most likely the 1 20 of ESRD until the first change of therapy. Among
result of bow statistical power. as reflected in the young nondiabetics, CAPD patients with glomerubo-
broad CI, leading us to believe that a more efficient nephnitls as the primary cause of ESRD had lower
design would show a significant difference In CAPD mortality rates than their HD counterparts. whereas
and HD mortality rates in this group. Overall, there those with hypertension and other reported causes
are fewer patients with a primary renal diagnosis of of ESRD showed no difference in mortality. If p0551-
glomerubonephritis and only a small proportion, ble, future studies should account for dialysis ade-
18.5%, of these patients died. These results are also quacy and factors that might affect the use of CAPD
consistent with the 1991 USRDS findings (38). or HD such as preexisting comorbid conditions, the
Comparison of results with ITT and RxHx censor- clinicians’ recommendations, and factors important
ing criteria found them to be similar among the non- to the patient.
diabetic diagnostic groups. However, among dia-
ACKNOWLEDGMENTS
betics, results of the ITT and RxHx analyses were
This study was supported in part by a grant from the Baxter Extra-
contrastingly different. This may indicate that the
mural Grants Program.
diabetic CAPD population has a lower long-term mor-
tality rate but may experience similar, and perhaps REFERENCES
slightly higher, mortality rates than their HD coun- 1 . Kjellstrand CM: Introduction to a workshop on
terparts If only the time before a patient’s first morbidity and mortality in hemodialysis, hemo-
change of dialytic therapy Is considered. filtration and continuous ambulatory penitoneal
dialysis. J ASAIO 1983:6:167-169.
The results of this study must be interpreted with
2. Blagg CR. Wahi PW, Lamers JY: Treatment of
caution and do not imply a causal relationship be- chronic renal failure at the Northwest Kidney
tween the use of CAPD and lower mortality rates. A Center, Seattle, from 1960-1982. J ASAIO
more valid evaluation of this relationship would nec- 1983:6:170-175.
3. Khanna R, WU G, Vas 5, Oreopoulos DG: Mor-
essanily consider conditions existing at the onset of
tality and morbidity on continuous ambulatory
ESRD as well as dialysis adequacy. Results of a re- penitoneal dialysis. J ASAIO 1 983;6: 197-204.
cently completed multicenter, case-mix study con- 4. Wing AJ, Broyer M, Brunner FP, et at.: The
ducted by the USRDS suggest that differences in the contribution of continuous ambulatory penito-
survival of young diabetic CAPD and HD populations neal dialysis in Europe. J ASAIO 1983:6:214-
219.
is attributable to a lower count of comorbid condi- 5. Noiph KD, Pyle WK, Hiatt M: Mortality and
tions among CAPD patients (39). morbidity In continuous ambulatory peritoneal
However, despite the possibility of biased selection dialysis: Full and selected registry populations.
of dialytic modality as a result of preexisting comor- J ASAIO 1983:6:220-226.
6. Anderson JE, Sturgeon D, Lindsay J, Schiller
bid conditions and patient preferences, the existence
A: Use of continuous ambulatory penitoneal di-
of a real difference in therapy efficacy cannot be alysis in a nursing home: Patient characteristics,
completely ruled out. Better management of CAPD technique success, and survival predictors. Am
therapy relative to HD therapy may account for some J Kidney Dis 1990:16:137-141.
7. Lindblad AS, Novak JW, Noiph KD. The USA
of the difference in mortality, whereas the avoidance
CAPD registry: Characteristics of participants
of sudden, drastic hemodynamic and biochemical and selected outcome measures for the period
changes as well as the improved control over gly- January 1 , 1981 through August 31 1987. .In:
cemia may also contribute to lower mortality rates Nolph KD, ed. Penitoneal DiaFysis. 3rd Ed. Bos-
among CAPD patients (24). ton: Kluwer Academic Publishers: 1989:389-
413.
In contrast to the findings among diabetic ESRD 8. Golper TA, Geerlings W, Seiwood NH, Brunner
patients under the age of 60 yr. older diabetic pa- FP, wing AJ. Peritoneal dialysis results in the
tients using CAPD have a higher rate of mortality EDTA registry. In: Nolph KD, ed. Peritoneal Di-
than do those using HD. This preliminary finding alysis. 3rd Ed. Boston: Kluwer Academic Pub-
lishers: 1989:414-428.
from the MKR (40) has been confirmed by the USRDS
9. Nissenson AR, Gentile DE, Soderbiom RE,
case-mix study (39). The issues of patient selection Oliver DF, Brax C: Morbidity and mortality of
criteria and unmeasured comorbidity remain unre- continuous ambulatory penitoneal dialysis: Re-
solved, however, in both studies. gional experience and long-term prospects. Am
J Kidney Dis 1986:7:229-234.
In summary, the use of home CAPD in Michigan
10. Ataman R, Burton PR, Gokal R, Brown CB,
has grown such that by 1 989 one third of incident Marsh FP, Walls J: Long-term CAPD-some UK
ESRD patients 20 to 59 yr of age were using this experience. Clin Nephrol 1988:30:71-S75.

I I 54 Volume 3 ‘ Njmber 5 ‘ 1992

 Nelson et al

1 1 . Tarchini R, Segoloni GP, Gentile MG, et al.: pies of choice: In-center hemodialysis vs. contin-
Long-term results of CAPD in Italy: A report from uous ambulatory peritoneal dialysis at home. Am
the Italian CAPD study group. Clin Nephrol J Kidney Dis 1990:15:433-440.
1 988:30:S68-S70. 26. Armitage P. Exclusions, losses to follow-up and
12. Bovbjerg RR, Diamond LH, Held PJ, Pauly MV: withdrawals in clinical trials. In: Shapiro SH,
Continuous ambulatory penitoneal dialysis: Pre- Louis TA, eds. Clinical Trials: Issues and Ap-
liminary evidence in the debate oven efficacy and proaches. New York: Dekker: 1983:99-113.
cost. Health Affairs 1983:12:96-102. 27. Maiorca R, Vonesh EF, Cavalli P. et al. : A mul-
1 3. Prowant B, Nolph KD, Dutton 5, et al.: Actu- ticenter, selection-adjusted comparison of pa-
anal analysis of patient survival and dropout tient and technique survivals on CAPD and he-
with various end-stage renal disease therapies. modialysis. Penitoneal Dial Int 1 99 1 : 1 1 : 1 18-
Am J Kidney Dis 1983:3:27-31. 127.
14. Kurtz SB, Johnson WJ. A four-year comparison 28. Feinstein AR. Clinical Epidemiology: The Archi-
of continuous ambulatory penitoneal dialysis tecture of Clinical Research. Philadelphia: WB
and home dialysis: A preliminary report. Mayo Saunders: 1985.
Clin Proc 1984:59:659-662. 29. Serkes KD, Blagg CR, Noiph KD. Vonesh EF,
1 5. Capelli JP, Camiscioli TC, Vallorani RD. Bo- Shapiro F: Comparison of patient and technique
beck JD: Comparative analysis of survival on survival in continuous ambulatory penitoneal di-
home hemodialysis. in-center hemodialysis and alysis (CAPD) and hemodialysis: A multicenter
chronic penitoneal dialysis (CAPD-IPD) thera- study. Penitoneab Dial Int 1 990; 10:15-19.
pies. Dial Transplant 1985:14:38-52. 30. Weller JM, Wu SC, Ferguson CW, Hawthorne
16. Rubin J, Barnes T, Bower J: Morbidity and VM: End-stage renal disease in Michigan: Inci-
mortality in CAPD and home hemodiabysis. J dence, underlying causes, prevalence and mo-
ASAIO 1985:8:22-27. dalities of treatment. Am J Nephrol 1985:5:84-
1 7. Charyton C, Spinowitz BS, Galler M: A compar- 95.
ative study of continuous ambulatory penitoneal 3 1. Cowie CC, Port FK, Wolfe RA, Savage PJ, Moll
dialysis and center hemodialysis: Efficacy, com- PP. Hawthorne VM: Disparities in incidence of
plications. and outcome in the treatment of diabetic end-stage renal disease according to
end-stage renal disease. Arch Intern Med race and type of diabetes. N Engl J Med
1986;146:1 138-1143. 1989:321:1074-1079.
1 8. Burton PR, Walls J: Selection-adjusted compar- 32. Sekkarie MA, Port FJ, Wolfe RA, et al. : Recov-
ison of life-expectancy of patients on continuous ery from end-stage renal disease. Am J Kidney
ambulatory penitoneal dialysis, haemodialysis, Dis 1990:15:61-65.
and renal transplantation. Lancet 1987:2:1115- 33. Cox DR: Regression models and life tables (with
1119. discussion). J R Stat Soc 1972:34:197-220.
1 9. Gokal R, King J, Bogle 5, et at.: Outcome In 34. Rothman KJ. Modern Epidemiology. Boston: Lit-
patients on continuous ambulatory penitoneal tle, Brown and Company: 1986:119-125.
dialysis and haemodlalysis: 4-year analysis of a 35. SAS Institute INC. SAS/STAT Software: The
prospective multicentre study. Lancet 1987: PHREG Procedure Preliminary Documentation.
1:1105-1109. Release 6.06. Cary, NC: SAS Institute Inc. : 1991.
20. Mailloux LU, Bellucci AG, Mossey RT, et al.: 36. Deber RA: The impact of selected patient char-
Predictors of survival in patients undergoing di- acteristics on practitioners’ treatment recom-
alysis. Am J Med 1988:84:855-962. mendations for end-stage renal disease. Med
2 1 . Maiorca R, Cancarini G, Manili L, et al.: CAPD Care 1985:23:95-109.
is a first class treatment: results of an eight-year 37. Smith MD, Hong BA, Michelman JE, Robson
experience with a comparison of patient and AM: Treatment bias in the management of end-
method survival in CAPD and hemodialysis. Clin stage renal disease. Am J Kidney Dis 1983:3:21-
Nephrol 1 988:30:S3-S7. 26.
22. Majorca R, Vonesh E, Cancarini GC, et al.: A 38. U.S. Renal Data System. USRDS 1991 Annual
six-year comparison of patient and technique Data Report. Bethesda, MD: The National Insti-
survivals in CAPD and HD. Kidney Int tutes of Mental Health, National Institutes of
1988:34:518-524. Diabetes and Digestive and Kidney Diseases: Au-
23. Nolph KD: Comparison of continuous ambula- gust 1991.
tory penitoneal dialysis and hemodialysis. Kid- 39. Held PJ. The latest information from the
ney Int 1988:33:5123-5131. USRDS. XII Annual Conference on Penitoneal
24. Diaz-Buxo JA: Continuous ambulatory penito- Dialysis. Seattle, Washington. February 1992.
neal dialysis (CAPD) and hemodlalysis: Pride and 40. Nelson CB, Port FK, Wolfe RA, Guire KE: Di-
prejudice. Peritoneal Dial mt i 990; 10:5-7. alysis patient survival: Evaluation of CAPD ver-
25. Wolfe RA, Port FK, Hawthorne VM, Guire KE: sus HD using 3 techniques. Penitoneal Dial Int
A comparison of survival among dialytic thera- 1992:12(Suppl 1):144.

Journal of the American Society of Nephrology I 155