Content uploaded by Beth Piraino
Author content
All content in this area was uploaded by Beth Piraino on Apr 03, 2015
Content may be subject to copyright.
Comparison of Infectious Complications between Incident
Hemodialysis and Peritoneal Dialysis Patients
Nabeel Aslam,*
a
Judith Bernardini,* Linda Fried,*
†
Renee Burr,* and Beth Piraino*
*Renal and Electrolyte Division, Department of Medicine, University of Pittsburgh School of Medicine, and
†
VA
Pittsburgh Health Care System, Pittsburgh, Pennsylvania
The impact of dialysis modality on infection, especially early in the course of dialysis, has not been well studied. This study
compared infection between hemodialysis (HD) and peritoneal dialysis (PD) from the start of dialysis and evaluated factors
that have an impact on infection risk. In this observational cohort study, all incident dialysis patients (n ⴝ 181; HD 119 and
PD 62) at a single center from 1999 to 2005 had data collected prospectively beginning day 1 of dialysis. Excluded were those
with any previous ESRD therapy. Infection rates were evaluated using multivariate Poisson regression. Overall infection rates
were similar (HD 0.77 versus PD 0.86/yr; P ⴝ 0.24). Only HD patients had bacteremia (0.16/yr), and only PD patients had
peritonitis (0.24/yr). Bacteremia that occurred <90 d after start of HD was 0.44/yr, increased compared with overall rate of
0.16/yr (P < 0.004). HD catheters, used in 67% of patients who started HD, were associated with a strikingly increased rate of
bacteremia. Peritonitis <90 d was 0.22/yr, no different from the overall rate. Modality was not an independent predictor of
overall infections (PD versus HD: relative risk 1.30; 95% confidence interval 0.93 to 1.8; P ⴝ 0.12) using multivariate analysis.
PD and HD patients had similar infection rates overall, but type of infection and risk over time varied. HD patients had an
especially high risk for bacteremia in the first 90 d, whereas the risk for peritonitis for the PD cohort was not different over
time. These results support the placement of permanent accesses (fistula or PD catheter) before the start of dialysis to avoid
use of HD catheters.
Clin J Am Soc Nephrol 1: 1226 –1233, 2006. doi: 10.2215/CJN.01230406
I
nfection is the second leading cause of death among dial-
ysis patients, accounting for 33 deaths per 1000 patient
years in the prevalent US Renal Data System (USRDS)
cohort for 2001 through 2003 (1). Septicemia, which is poorly
defined, accounts for 79.7% of infectious deaths (1). Dialysis
patients who are hospitalized for bacteremia/septicemia have a
relative risk (RR) for death of 7 for the first 6 mo after the event
compared with those without hospitalization for septicemia,
and the increased risk persists for at least 48 mo (2). The
adjusted mortality rate for the 6 mo after hospitalization for
septicemia is 120.1 per 1000 patient-years, compared with ap-
proximately 10 per 1000 patient-years in those without hospi-
talization for septicemia.
The rates of hospital admissions for septicemia are higher in
hemodialysis (HD) patients than peritoneal dialysis (PD) patients
(2,3). HD compared with PD as an initial modality doubles the risk
for hospitalization for septicemia (3). The hospitalization rates for
septicemia in PD have remained relatively constant since 1997,
whereas the rates for hospitalization in HD patients have doubled
between 1991 and 2001, because of a striking increase in the
admission rate for vascular access infection (2). However, the
recorded rate of death for septicemia is higher for PD patients
(29.3 per 1000 patient-years) than for HD patients (26.1 per 1000
patient-years) for the most current period for which data are
available (1). These results seem inconsistent and confusing.
Most previous reports that used USRDS data did not include
the critical initial period of time on dialysis, because USRDS
collection of information does not begin until the 90th day of
dialysis. This is a time when a high proportion of patients are
using HD catheters, which are known to increase the risk for
bacteremia (4,5). Excluding this period of time will underesti-
mate the rate of infections.
The use of HD catheters has been shown repeatedly to be an
independent predictor of death in HD patients (6 –9). Data from
elderly patients with preexistent Medicare, permitting tracking
of information from day 1 of HD, indicate that 15.1% of patients
with an HD catheter die in the first 90 d compared with 6.7%
with a fistula, with a hazard ratio of death with an HD catheter
of 2.15 (9). Therefore, the risk for bacteremia and death in HD
are especially high in the first 3 mo of dialysis.
A common perception is that PD is associated with a higher
risk for infection compared with HD. However, the impact of
dialysis modality on the rate and particularly type of infection
has not been well studied, especially during the first few
months of dialysis. Using a dialysis registry that tracks patients
from the first day of dialysis, we designed this study to com-
pare the rates and type of infections in patients who were
incident to either HD or PD, controlling for important variables.
Received April 14, 2006. Accepted September 11, 2006.
Published online ahead of print. Publication date available at www.cjasn.org.
a
N.A.’s current affiliation is Fatima Memorial Medical and Dental College, La
-
hore, Punjab Province, Pakistan.
Address correspondence to: Dr. Beth Piraino, 3504 5th Ave., Suite 200, Pittsburgh,
PA 15213. Phone: 412-383-4899; Fax: 412-383-4898; E-mail: piraino@
pitt.edu
Copyright © 2006 by the American Society of Nephrology ISSN: 1555-9041/106-1226
Materials and Methods
Since 1999, we have had dialysis registries to collect information
prospectively on patients who undergo both HD and PD at Dialysis
Clinics Inc. (DCI) of Oakland dialysis center, an urban unit with phy-
sicians who are affiliated with the University of Pittsburgh Medical
Center. The dialysis registries are approved by the university’s Institu-
tional Review Board and DCI administrative review office, and the
patients provided informed consent at the start of dialysis to collect
data prospectively for research purposes. None of the PD patients and
one of the HD patients (⬍1%) declined participation in the registry. We
retrieved data from these prospective databases on incident patients
from January 1, 1999, to September 1, 2006.
We excluded 63 patients who were on HD and 57 who were on PD
and had any previous ESRD therapy, including failed renal transplants.
This was done to focus on patients who initiated dialysis as a first ESRD
therapy and to avoid confounding a potential impact of one therapy on
another and the access related to that modality. Eight patients who
were incident to ESRD during the period studied were excluded be-
cause all were started on HD (but not at our center) and transferred to
PD at our center within 3 mo of initiation of dialysis (range 0.33 to 1.90
mo) and therefore were on two modalities. To evaluate any potential
bias, we ran an additional analysis that included these eight patients in
the PD group. Modality was defined as that started on the first day of
dialysis and time at risk began at the outpatient dialysis treatment for
HD and the first day of training for PD. The majority of time on PD was
on automated PD (APD; 92%), with the remaining time on continuous
ambulatory PD (CAPD).
Follow-up continued to September 1, 2005. Individuals were cen-
sored at transfer to another modality (three on HD and seven on PD),
improvement in residual renal function that led to discontinuation of
dialysis (two on HD and one on PD), transplantation (four on HD and
15 on PD), permanent transfer to another center on the same modality
(five on HD and zero on PD), and death (33 on HD and seven on PD).
Temporary transfers to another modality, transient cessation of dialysis
as a result of partial recovery of renal function, no dialysis as a result of
catheter malfunction, or time in a rehabilitation center were not reasons
for censoring when the patient returned to the original modality in ⬍1
mo (six each on HD and PD). However, the time off the dialysis therapy
was excluded as were infections during this short period for both HD
and PD because the focus of this article is on infections on one modality
versus the other.
Prospective data that were collected included demographics, comor-
bid conditions at dialysis initiation, serum albumin at the start of
dialysis, infection episodes, and hospitalizations as a result of infec-
tions. Comorbidity status at the start of dialysis was assessed using
Charlson Comorbidity Index (CCI) (10), scored by an experienced
physician assistant for HD and an experienced research nurse for PD. A
careful review of the patient’s medical history and physical examina-
tion was used to determine the correct score. Diabetes was defined as
elevated blood glucose and a history of diabetes on the basis of review
of the history of the patient regardless of the use of insulin or hypo-
glycemic medications. The guide for scoring the CCI is in Appendix 1.
Infections that were included in the registry were those related to the
access regardless of whether hospital admission was required and all
infections that required hospital admission.
Infections were classified as catheter infections, either exit site (de-
fined as drainage, erythema, or exit-site pain) or tunnel infection (de-
fined as swelling or pain/tenderness with or without erythema over
the catheter tunnel) for both HD and PD. Pneumonia was defined as
radiologic evidence of pneumonia in the clinical setting of pneumonia.
Peritonitis was defined as cloudy effluent with ⱖ100/
l white cells
with ⱖ50% of these polymorphonuclear cells. Cellulitis was defined as
infection of subcutaneous tissue that required hospitalization. Bactere-
mia and fungemia were defined by positive blood cultures. “Other
infections” that required hospital admission included infected limb
ulcers.
Infectious rates were calculated by dividing the total number of
infections by total time at risk in the HD or PD cohorts and expressed
as number of infections per dialysis year. We also expressed rates as the
number of episodes per 1000 d at risk because this is the way it
generally is expressed in HD. All hospitalizations as a result of infec-
tions were included in the analysis as well as all access-related infec-
tions. We also determined the rate of infection during the first 90 d of
dialysis, both HD and PD.
Protocols to reduce the risk for infections were used in both the HD
and PD programs. All patients with HD catheters had antibiotic cream
placed at the exit site after cleansing of the site at the end of dialysis as
part of routine HD catheter care, done by the staff. All HD patients with
a catheter were asked to keep this dry and specifically to refrain from
showers. All PD patients were instructed to wash the PD catheter exit
site using antibacterial soap in the shower and then after the shower
use a cotton swab to put antibiotic cream around the catheter at the exit
site. PD patients were told to avoid tub baths.
Statistical Analyses
Variables (PD versus HD) were compared using t test, Mann-Whitney
U test, or
2
test, as appropriate. Poisson regression analysis was used
to assess the impact of various factors on outcome of interest (total
Table 1. Demographic characteristics of the HD and PD cohorts
a
Characteristic HD PD P
No. of patients 119 62
Follow-up on dialysis (mo; median 关range兴) 18 (0.3 to 75) 15 (0.13 to 72) 0.96
Age at start of dialysis 59 ⫾ 16 55 ⫾ 17 0.15
Men (%) 57 44 0.12
White (%) 48 84 ⬍0.001
Diagnosis diabetes (%) 54 46 0.35
Albumin at dialysis start (g/dl) 3.2 ⫾ 0.6 3.7 ⫾ 0.6 ⬍0.001
CCI (median 关range兴) 6 (2 to 14) 6 (2 to 14) 0.23
HD catheter as first access (%) 67 0
PD catheter as first access (%) 0 100
a
CCI, Charlson Comorbidity Index; HD, hemodialysis; PD, peritoneal dialysis.
Clin J Am Soc Nephrol 1: 1226 –1233, 2006 Infectious Complications in HD and PD 1227
infections and infections that required hospitalization). Because age
and diabetes are part of CCI, they were not analyzed as separate
covariables. Two different models were used for multivariate Poisson
regression to evaluate the impact of variables (one with age and dia-
betes and the other with CCI along with other variables) because age
and diabetes are part of CCI. Interactions of dialysis modality with age,
race, gender, diabetes, and CCI were tested. Because an interaction was
found between modality and diabetes, separate models also were run
in those with and without diabetes. Number Crunch Statistical System
software (Dr. Jerry Hintze, Kaysville, UT) was used for statistical anal-
ysis.
Results
A total of 181 incident dialysis patients (119 on HD and 62 on
PD) met the inclusion criteria. The demographics of the two
groups of de novo dialysis patients are shown in Table 1. The
HD cohort was slightly older, had more black patients, and had
a lower serum albumin level at the start of dialysis. Sixty-seven
percent of the HD patients began dialysis using an HD catheter.
The median comorbidity index was the same for the two
groups at the start of dialysis. Cardiac disease was present in 44
and 57% (P ⫽ 0.10), malignancy in 11 and 19% (NS), pulmonary
disease in 15 and 13% (NS), and systemic immune disease in 3
and 7% (NS) for HD and PD, respectively. Peripheral vascular
disease was present more frequently in the PD patients (10 and
22%; P ⫽ 0.03) whereas liver disease was more frequent in the
HD patients (18 and 3%; P ⬍ 0.003).
Infection rates for each group are summarized in Table 2 and
Figure 1. The overall infection rates were the same in the HD
and PD cohorts. However, the rates for various types of infec-
tions were different in HD and PD patients. Bacteremia and
fungemia occurred only in HD patients, whereas peritonitis
occurred only in PD patients.
The rate of bacteremia was much higher during the first 90 d
of HD than during the overall time at risk in HD patients (P ⬍
0.004). Forty percent of all episodes of bacteremia during the
observation period occurred in the first 90 d on HD. The risk for
bacteremia fell significantly after 90 d from 0.42 per year to 0.12
per year (P ⫽ 0.001). The risk for infections varied for HD by the
type of access, with the highest risk for acute dialysis catheters
and tunneled dialysis catheters (Figure 2). In contrast, in PD
patients, the peritonitis rate was 0.22 episodes per year at risk
during the first 90 d, not different from the overall rate of 0.24
per year at risk for the total time for the PD cohort.
Organisms that caused dialysis-related infections were sim-
ilar in the HD and PD cohorts. Gram-positive bacteria ac-
counted for 62% of the bacteremic episodes in the HD patients
(rate 0.10 episodes per year at risk) and 46% of the peritonitis
episodes (0.11 per year at risk). The rate of Staphylococcus aureus
bacteremia was 0.01 per year for HD, and the rate of S aureus
peritonitis was 0.04 per year for PD (P ⫽ 0.12). Gram-negative
organisms accounted for 22% of bacteremic episodes in HD
Figure 1. Comparison of infection rates in hemodialysis (HD; f)
and peritoneal dialysis (PD; u). Data are episodes per 1000 d at
risk.
Table 2. Infections in the HD and PD cohorts
Parameter HD PD P
Patients infection-free 54 (45%) 25 (40%) NS
Infections, total per time at risk (median 关range兴) 1 (0 to 14) 1 (0 to 10) NS
Infection rates per year at risk
bacteremia/fungemia, overall 0.16 0 ⬍0.0001
bacteremia in the first 90 d 0.47 0 ⬍0.0001
peritonitis 0 0.24 ⬍0.0001
peritonitis in the first 90 d 0 0.22 ⬍0.001
catheter exit-site and tunnel infections 0.35 0.41 NS
total infection rate per dialysis year 0.77 0.86 NS
Hospitalization rates per year at risk
admission for bacteremia 0.10 0 ⬍0.001
admission for peritonitis 0 0.19 ⬍0.001
admission for cellulitis 0.06 0 ⬍0.001
admission for pneumonia 0.07 0.02 ⬍0.001
admission for other serious infections 0.13 0.21 0.04
all admissions for infection 0.29 0.42 0.02
total admissions including noninfectious 2.4 1.4 ⬍0.0001
1228 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 1: 1226 –1233, 2006
(0.04 per year at risk) and 54% of peritonitis episodes (0.13 per
year for PD).
Overall rates of catheter infection (exit site and tunnel) were
similar for HD and PD patients (0.35 versus 0.41 per year; P ⫽
0.20, respectively). S. aureus catheter infections for HD were
0.05 per year. This was comparable to 0.07 per year for PD (P ⫽
0.30).
Rates of infection for CAPD versus APD were not different.
There were no episodes of peritonitis on CAPD compared with
0.26 per year on APD (P ⫽ 0.14). Catheter infection rates were
similar, 0.35 per year on CAPD and 0.42 per year on APD (P ⫽
0.43).
Table 3 shows the results of the multivariate analysis. In the
first model, CCI (which includes both age and diabetes) was
used, whereas in the second model, age and diabetes were used
in place of CCI. In both models, a decreased serum albumin at
the start of dialysis was a strong predictor of subsequent infec-
tion. Black patients were at decreased risk for infection in both
models. Dialysis modality (PD versus HD) was not an indepen-
dent risk for infection; neither was gender. In the second model,
which separates diabetes and age, younger age was a risk factor
for infection as was the presence of diabetes.
We ran an additional analysis to include the eight incident-
to-dialysis patients who had 3 mo or less on HD before PD. This
constituted 11% of PD patients. Inclusion of these few patients
did not change the results of the study because there were no
differences in overall infection rates or hospitalization between
PD (now 70 patients) and HD (119 patients).
Interactions
There was not a significant interaction of modality with age,
albumin, or CCI. However, significant interactions were seen
with modality and diabetes (P ⫽ 0.02). In addition, there were
important interactions in infection risk for modality and gender
(P ⫽ 0.04).
The interaction between modality and diabetes for infection
risk was explored further. Patients who had diabetes and were
on PD had higher infection rates than patients who had diabe-
tes and were on HD (1.28 versus 0.84/yr; P ⬍ 0.004). Con-
versely, patients without diabetes had a lower risk for infection
on PD than on HD (0.51 versus 0.69/yr; P ⬍ 0.03). Separate
models were run in patients with and without diabetes. Ad-
justment for albumin, age, race, gender, and dialysis modality
was NS in individuals with diabetes (RR 1.13; 95% confidence
interval [CI] 0.76 to 1.67; P ⫽ 0.55) or in individuals without
diabetes (RR 0.67; 95% CI 0.37 to 1.21; P ⫽ 0.18).
The interaction between modality and gender (P ⫽ 0.04) for
infection risk also was evaluated further. Men who were on PD
had a significantly higher rate of infection than women who
were on PD (1.07 versus 0.71/yr; P ⬍ 0.009). In contrast, infec-
Figure 2. Rates of bacteremia by access in the HD patients.
Table 3. Multivariate Poisson regression analysis models for overall infection rate
a
Parameter RR 95% CI P
Model 1: Using albumin, race, gender, modality, and CCI
albumin, for each
g/dl
0.62 0.50 to 0.77 ⬍0.001
race (black) 0.62 0.47 to 0.82 0.008
modality (PD) 1.13 0.82 to 1.55 0.45
gender (male) 1.05 0.83 to 1.41 0.73
CCI, for each point 1.01 0.81 to 1.35 0.67
Model 2: Using albumin, race, gender, modality, diabetes, and age
age, per 10-yr
increase
0.87 0.80 to 0.95 0.002
albumin, for each
g/dl
0.66 0.53 to 0.82 ⬍0.001
diabetes 1.48 1.13 to 1.94 0.004
gender (male) 1.03 0.80 to 1.33 0.82
race (black) 0.58 0.44 to 0.77 ⬍0.001
modality (PD) 0.98 0.70 to 1.36 0.90
a
CI, confidence interval; RR, relative risk.
Clin J Am Soc Nephrol 1: 1226 –1233, 2006 Infectious Complications in HD and PD 1229
tion rates were the same in men and women who were on HD.
When stratified by gender and controlling for age, race, and
diabetes, significant differences were seen in infection rate by
modality for men (PD versus HD: RR 1.92; 95% CI 1.13 to 3.27;
P ⫽ 0.02) and women (PD versus HD: RR 0.61; 95% CI 0.39 to
0.96; P ⫽ 0.03).
White patients had a similar infection rate on PD and HD
(0.99 versus 0.97/yr). However, black patients had a lower rate
of infections on PD than on HD (0.44 versus 0.68; P ⬍ 0.002).
This held true with the multivariate analysis as well.
Hospitalization for Infections
The all-cause admission rate was 2.4 per dialysis year for HD
and 1.4 per dialysis year for PD (P ⬍ 0.001). However, PD
patients had a higher rate of infection-related hospitalization
(0.42 versus 0.29 hospitalization per year at risk; P ⬍ 0.001). Two
models were used for multivariate Poisson regression analysis
for infections that required hospitalization, one with CCI, albu-
min, gender, race, and modality and a second model using age,
diabetes, albumin, gender, race, and modality (Table 4). Albu-
min and modality were significant predictors in both models.
CCI was significant in the first model as were age and diabetes
in the second model. There were no significant interactions of
race or diabetes with modality for the risk for hospitalization
for infection. However, there was a trend for interaction of
gender with mode for hospitalization for infection (P ⫽ 0.06).
Men who were on PD had a higher admission rate for infection
compared with men who were on HD (0.59 versus 0.32 per
dialysis year), whereas women who were on PD had similar
rates of infection as women who were on HD (0.29 versus 0.24
per dialysis year).
Discussion
This study demonstrated that patients who start dialysis on
HD compared with those who start on PD have similar overall
rates of infection, controlling for comorbidity, race, diabetes,
age, gender, and initial serum albumin level. However, there
were marked differences in both the type of infections and the
risks during the first 90 d of dialysis among patients who were
on these two modalities. Only HD patients had bacteremia, and
only PD patients had peritonitis. Both of these complications
likely are related to the dialysis access or contamination during
connection.
The risk for bacteremia in HD patients was strikingly in-
creased during the first 90 d of dialysis. The early risk for
bacteremia in HD patients very likely is related to the use of HD
catheters as the initial access in 67% of our HD patients, similar
to that reported in the CHOICE study (8). HD catheters are
widely known to increase the risk for bacteremia (4,5). These
results support the need for a fistula-first protocol so that all
patients who want HD have a fistula placed before commence-
ment of dialysis (5). If HD catheter use could be avoided with
careful planning of dialysis initiation, then it is possible that the
risk for infection actually may be lower for HD compared with
PD. In some cases, this may require placing an HD graft if a
primary fistula cannot be achieved in a timely manner. For
patients who desire PD, a prudent approach would be to place
a PD catheter in a timely manner and avoid a period on HD
with an HD catheter. This approach is very likely to decrease
the risk for bacteremia.
Our study refutes the common misconception that PD is
associated with an overall higher rate of infection as compared
with HD. In particular, we found bacteremia to be a complica-
tion almost solely of HD. These results are consistent with those
of Abbott and Agodoa (11), who examined 327,993 patients
who initiated dialysis between January 1992 and June 1997 for
risk for hospitalization for bacterial endocarditis. HD patients
had an RR of 17.86 (95% CI 6.62 to 48.90) compared with the
general population in 1996. This was in contrast to PD patients,
who had an insignificant increased risk compared with the
general population (10.54; 95% CI 0.71 to 158.13).
When septicemia is examined, as opposed to endocarditis,
the evidence is less clear. Powe et al. (12) examined data from
the USRDS and found that 11.7% of HD patients and 9.4% of
PD patients had during 7 yr at least one episode of septicemia.
Table 4. Multivariate Poisson regression analysis for infections that required hospitalization
Parameter RR (95% CI) P
Model with CCI
albumin (per 1-g/dl increase) 0.38 (0.27 to 0.53) ⬍0.001
CCI (per point increase) 0.91 (0.82 to 1.01) 0.07
gender (male) 1.45 (0.96 to 2.20) 0.07
race (black) 0.61 (0.39 to 0.96) 0.03
modality (PD) 1.96 (1.20 to 3.21) 0.008
Model with age and diabetes
age (per 10-yr increase) 0.84 (0.73 to 0.96) 0.01
albumin (per 1-g/dl increase) 0.40 (0.28 to 0.57) ⬍0.001
diabetes 1.46 (0.96 to 2.21) 0.07
gender (male) 1.45 (0.96 to 2.19) 0.07
race (black) 0.60 (0.38 to 0.95) 0.03
modality (PD) 1.73 (1.03 to 2.91) 0.04
1230 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 1: 1226 –1233, 2006
Part of the problem with examining septicemia from the
USRDS is the lack of clarity regarding the definition of septi-
cemia as pointed out recently by O’Seaghdha and Foley (13).
Sepsis and septicemia may be confused with peritonitis by the
coding doctors. The USRDS does not provide a definition of
septicemia, and it is not clear that doctors are using this only for
bacteremia. To study this further through the USRDS success-
fully will require abandoning the term “septicemia” and replac-
ing it with more specific terms, such as bacteremia and perito-
nitis.
In a center with a low incidence of both bacteremia in HD
patients and peritonitis in PD patients, the overall rate of in-
fection was similar in the two cohorts. Our center rate for
catheter-related bacteremia in HD of 1.2 per 1000 d (Figure 1) is
well below recent published rates of 2.5 to 5.5 per 1000 patient
days (14). Likewise, our center rate for peritonitis of 0.24 per
patient year (one episode per 50 dialysis months) is a low rate
but certainly achievable by other centers (15). The low rates of
bacteremia and peritonitis at our center may be due to the close
attention to infectious risk, patient and staff training, and pro-
phylactic use of exit-site antibiotic cream to prevent exit-site
infection in both HD and PD patients (16,17).
We found a low serum albumin at the start of dialysis,
younger age, race, and the presence of diabetes to be the
strongest risk factors for overall infections, regardless of mo-
dality and controlling for other comorbid conditions. Our find-
ings are consistent with other studies showing the presence of
diabetes, low serum albumin, and temporary vascular access to
be risk factors for septicemia in HD patients (10,18,19). Hy-
poalbuminemia also has been shown to be a risk factor for
subsequent peritonitis (20 –22). This relationship between hy-
poalbuminemia and infection in dialysis patients is an associ-
ation rather than a proven causal relationship. Precisely how a
low serum albumin level increases the risk for infection in both
HD and PD patients is uncertain.
We found a higher rate of admissions for infection in the PD
patients compared with the HD patients. PD patients were
likely to be admitted to the hospital for management of infec-
tions, especially limb ulcers and peritonitis. Vascular access
may make it less likely to hospitalize HD patients for treatment
of limb ulcers, because patients can receive intravenous antibi-
otics during dialysis. However, the overall hospitalization rate
was lower among PD patients. Many dialysis-related infections
are treated as outpatients. Studies that use only admissions for
Charlson Comorbidity Index
a
Score Condition Definition
1 Myocardial infarct Not ECG changes only
Congestive heart failure With dyspnea, responded to treatment
Peripheral vascular disease Includes untreated aneurysm ⱖ6cm
Cerebrovascular disease CVA with minor or no residual, TIA
Dementia Includes cognitive deficits
Chronic pulmonary disease Mild, moderate, and severe
Connective tissue disorder SLE, polymyositis, MCTD, polymyalgia rheumatica, RA
Ulcer disease With or without bleeding
Mild liver disease Cirrhosis without portal HTN, chronic hepatitis (includes C)
Diabetes
b
(without end-organ disease)
Not diet-controlled alone
2 Hemiplegia
Moderate or severe renal disease With retinopathy or neuropathy, or nephropathy, or juvenile
onset, or brittle diabetic
Diabetes
b
(with end-organ damage)
Dialysis dependent
Any tumor Without mets, first treatment in last 5 yr
Leukemia Acute, chronic, lymphocytic, and polycythemia vera
Lymphoma Hodgkin’s lymphosarcoma, Waldenström’s
macroglobulinemia, myeloma, and all lymphomas
3 Moderate or severe liver disease Cirrhosis with portal HTN with or without bleeding,
⫾ variceal bleeding
6 Metastatic solid tumor
AIDS Not just HIV⫹
Age score: For each decade ⬎40 yr of age, a score of 1 is added to the above score
1 50 to 59.9
2 60 to 69.9
3 70 to 79.9
4 80 to 89.9
a
ECG, electrocardiogram; SLE, systemic lupus erythematosus.
b
Assign onlya1ora2fordiabetes.
Clin J Am Soc Nephrol 1: 1226 –1233, 2006 Infectious Complications in HD and PD 1231
infections will not determine accurately the infection rates for
the two modalities.
There were interesting interactions between infection risks
on the two modalities and both race and diabetes. Black pa-
tients had a lower rate of infection on PD than on HD. Histor-
ically, in the United States, black patients are less likely to be
placed on PD for somewhat unclear reasons, but some litera-
ture suggests a higher risk for infection on PD compared with
other patients (23). Most of this literature is older; with new
technology and the decrease in coagulase-negative infections,
black patients may no longer have a higher risk for infection.
Our results suggest that PD may be underused in black patients
and that in regard to infection risk, PD is a good choice for this
population. However, the total number of black patients who
were on PD in our study is small, so further studies should
confirm our finding.
Patients with diabetes seem to have a higher risk for infection
on PD than on HD. This result was independent of serum
albumin level at the start of dialysis. These results are consis-
tent with the finding of Sarnak and Jaber (24), who found that
patients who have diabetes and are on PD had a strikingly
higher risk for death from “sepsis” than patients who did not
have diabetes and were on PD. We also found a lower rate of
infection on PD for individuals without diabetes. It is interest-
ing to speculate as to whether this may explain the results of
some observational studies showing a higher risk for death
with increasing time on dialysis with PD compared with HD
for patients with diabetes. More studies will be needed in this
area.
There are limitations to this study. The study is based on a
single center. Our unit is an academic dialysis unit, and our
overall infection rate is low in both PD and HD compared with
many other centers. As with all observational studies, there
may be confounding by selection bias. We use exit-site antibi-
otic prophylaxis to decrease the rate of both peritonitis (PD)
and bacteremia (HD), and this may be a different approach
than in other programs. Our numbers limit the evaluation of
the interactions between race and diabetes with infections on
modalities. Whether our results could be generalized to other
sites or units with higher infection rates is not known.
Conclusion
This study provides evidence that dialysis modality is not an
independent predictor of overall infection rate in a cohort of
incident dialysis patients but is a strong predictor of the type
of infection and the difference in risk during the first 90 d of
dialysis. Patients should be informed that there is a high risk for
bacteremia when starting dialysis using an HD catheter and
that the risk of peritonitis in association with a PD catheter
during the first 3 mo of PD is considerably lower than the risk
for bacteremia with an HD catheter. Choice of modality should
not be dictated by the concern of higher infection rate on PD;
rather, it should be based on the individual patient’s prefer-
ence.
Acknowledgments
Preliminary results were presented as a poster at the annual meeting
of the American Society of Nephrology; Philadelphia, PA; October 30
through November 14, 2002; and as an update at the annual meeting of
the American Society of Nephrology; Philadelphia, PA; November 8
through 13, 2005.
B.P. and J.B. are consultants and on the Speakers Bureau for Baxter
Healthcare Corporation. We do not believe that these represent a
conflict of interest for the subject matter of this article.
References
1. US Renal Data System: USRDS 2005 Annual Data Report:
Atlas of End-Stage Renal Disease in the United States, Be-
thesda, National Institutes of Health, National Institute of
Diabetes and Digestive and Kidney Diseases, 2005
2. US Renal Data System: USRDS 2004 Annual Data Report:
Atlas of End-Stage Renal Disease in the United States, Be-
thesda, National Institutes of Health, National Institute of
Diabetes and Digestive and Kidney Diseases, 2004
3. Foley RN, Guo H, Snyder JJ, Gilbertson DT, Collins AJ:
Septicemia in the United States dialysis population 1991 to
1999. J Am Soc Nephrol 15: 1038 –1045, 2004
4. Hoen B, Paul-Dauphin A, Hestin D, Kessler M: EPIBACDIAL:
A multicenter prospective study of risk factors for bacteremia
in chronic hemodialysis patients. J Am Soc Nephrol 9: 869 –876,
1998
5. Oliver MJ, Rothwell DM, Fung K, Hux JE, Lok CE: Late
creation of vascular access for hemodialysis and increased
risk of sepsis. J Am Soc Nephrol 15: 1936 –1942, 2004
6. Pastan S, Souchie M, McClellan WM: Vascular access and
increased risk of death among hemodialysis patients. Kid-
ney Int 62: 620 –626, 2002
7. Dhingra R, Young EW, Hulbert-Shearon TE, Leavey SF,
Port FK: Type of vascular access and mortality in US
hemodialysis patients. Kidney Int 60: 1443–1451, 2001
8. Astor BC, Eustace JZ, Powe NR, Klag MJ, Fink NE, Coresh
J; the CHOICE Study: Type of vascular access and survival
among incident hemodialysis patients: The Choices for
Healthy Outcomes in Caring for ESRD (CHOICE) Study.
J Am Soc Nephrol 16: 1449 –1455, 2005
9. Xue JL, Dahl D, Ebben JP, Collins AJ: The association of
initial hemodialysis access type with mortality outcomes in
elderly Medicare ESRD patients. Am J Kidney Dis 42: 1013–
1019, 2003
10. Charlson ME, Pompei P, Ales KL, MacKenzie CR: A new
method of classifying prognostic comorbidity in longitudi-
nal studies: Development and validation. J Chron Dis 40:
373–383, 1987
11. Abbott KC, Agodoa LY: Hospitalizations for bacterial en-
docarditis after initiation of chronic dialysis in the United
States Nephron 91: 203–209, 2002
12. Powe NR, Jaar B, Furth SL, Hermann J, Briggs W: Septice-
mia in dialysis patients: Incidence, risk factors, and prog-
nosis. Kidney Int 55: 1081–1090, 1999
13. O’Seaghadha CM, Foley RN: Septicemia, access, cardiovas-
cular disease, and death in dialysis patients. Perit Dial Int
25: 534–540, 2005
14. Allon M: Dialysis catheter-related bacteremia: Treatment
and prophylaxis. Am J Kidney Dis 44: 779 –791, 2004
15. Li PK, Law MC, Man C, Szeto CC, Cheng YL, Wong TY,
Leung CB, Wang AY, Lui SF, Yu AW: Comparison of
1232 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 1: 1226 –1233, 2006
clinical outcome and ease of handling in two double-bag
systems in continuous ambulatory peritoneal dialysis: A
prospective, randomized, controlled, multicenter study.
Am J Kidney Dis 40: 373–380, 2002
16. Jaber BL: Bacterial infections in hemodialysis patients:
Pathogenesis and prevention Kidney Int 67: 2508–2519,
2005
17. Bernardini J, Bender F, Florio T, Sloand J, PalmMontal-
bano L, Fried L, Piraino B: Randomized double blinded
trial of exit site cream for prevention of exit site infection
in peritoneal dialysis patients. J Am Soc Nephrol 16: 539 –
545, 2005
18. Jaar BG, Hermann JA, Furth SL, Briggs W, Powe NR:
Septicemia in diabetic hemodialysis patients: Compari-
son of incidence, risk factors, and mortality with non-
diabetic hemodialysis patients. Am J Kidney Dis 35: 282–
292, 2000
19. Abbott KC, Agodoa LY: Etiology of bacterial septicemia in
chronic dialysis patient in the United States. Clin Nephrol
56: 124–131, 2001
20. Wang Q, Bernardini J, Piraino B: Albumin at the start of
peritoneal dialysis predicts the development of peritonitis.
Am J Kidney Dis 41: 664 –669, 2003
21. Lim CT, Wong K, Foo MW: The impact of topical mupiro-
cin on peritoneal dialysis infection in Singapore General
Hospital. Nephrol Dial Transplant 20: 2202–2206, 2005
22. Chow KM, Szeto CC, Leung CB, Kwan BC, Law MC, Li
PK: A risk analysis of continuous ambulatory peritoneal
dialysis-related peritonitis. Perit Dial Int 25: 374–379,
2005
23. Farias MG, Soucie JM, McClellan, Mitch WE: Race and the
risk of peritonitis: An analysis of factors associated with
the initial episode. Kidney Int 46: 1392–1396, 1996
24. Sarnak MJ, Jaber BL: Mortality caused by sepsis in patients
with end-stage renal disease compared with general pop-
ulation. Kidney Int 58: 1758 –1764, 2000
Clin J Am Soc Nephrol 1: 1226 –1233, 2006 Infectious Complications in HD and PD 1233
