Nephrol Dial Transplant (2003) 18: 15851591

DOI: 10.1093/ndt/gfg225

Original Article

Interactions between dialysis-related volume exposures, nutritional

surrogates and mortality among ESRD patients
Lynda Anne Szczech1,2, Donal N. Reddan1,2, Preston S. Klassen1,2, Joseph Coladonato1,2,
Benjamin Chua2, Edmund G. Lowrie1,2,3, J. Michael Lazarus3 and William F. Owen Jr1,2,4
1

Department of Medicine, Division of Nephrology, 2Institute for Renal Outcomes Research and Health Policy,
Duke University Medical Center, Durham, NC, 3Fresenius Medical CareNorth America, Inc., Lexington, MA and
4
Baxter International Healthcare, Waukegan, IL, USA

Correspondence and offprint requests to: Lynda Anne Szczech, MD,

MSCE, Duke University Medical Center, Department of Medicine,
Division of Nephrology, 2400 Pratt Street, Room 7060, Durham,
NC 27705, USA. Email: szcze001@mc.duke.edu

among patients with creatinine <7.26, which failed to

remain signicant for patients whose creatinine was
7.26 mg/dl. Increasing IDWL% is associated with a
greater mortality risk among patients with greater postdialysis weight, greater body mass index and lower
serum sodium measurements.
Conclusions. This study conrms and extends the
ndings of the deleterious association between
increasing IDWL% and mortality among patients
with diabetes mellitus and among subgroups based on
serum creatinine and body weight. The putative
deleterious effect of dialysis-related volume expansion
on mortality must be interpreted in the context of the
patients diabetic and nutritional status.
Keywords: end-stage renal disease; haemodialysis;
mortality; nutrition; weight

Introduction
During haemodialysis, ultraltration of uid at excessive rates from patients with end-stage renal disease
(ESRD) may precipitate episodes of muscle cramping
and/or hypotension [1]. Additionally, increasing interdialytic weight gain is associated with hypertension
among ESRD patients [2,3], which may contribute to
cardiac dysfunction [4]. Fluid restriction is, therefore,
prescribed for patients to limit the amount of uid that
needs to be removed during each treatment. Although
interdialytic weight gain reects uid and food intake,
increasing interdialytic weight gain has been characterized as a measure of non-compliance and a deleterious
outcome [57]. Supporting this contention is the
observed relationship between increasing interdialytic
weight gain and an increased risk of mortality [6,7].
Competing with the putative increasing mortality
risk of volume expansion is the association between

2003 European Renal AssociationEuropean Dialysis and Transplant Association

Downloaded from http://ndt.oxfordjournals.org/ by guest on February 12, 2012

Abstract
Background. Interdialytic weight gain is used as a
surrogate for volume expansion in haemodialysis
patients and as an indicator of non-compliance.
Increased weight gain is associated with both a
greater mortality risk and better nutrition indices.
This analysis characterizes the association between
dialysis-related volume expansion and mortality in the
context of its interaction with nutritional surrogates.
Methods. All patients receiving haemodialysis through
Fresenius Medical CareNorth America during 1998
were included. The percentage reduction in weight or
intradialytic weight loss (IDWL%) was dened as the
difference between the average of pre- and post-dialysis
weights from the last 3 months of 1997 expressed
as a percentage of post-dialysis weight. Associations
between IDWL% and clinical and demographic
variables were estimated using linear regression. The
association between mortality risk and IDWL% was
estimated using Cox proportional hazards regression.
Results. Younger age, male gender, the presence of
diabetes mellitus, decreasing cholesterol, post-dialysis
weight and pre-dialysis blood pressure (systolic and
pulse pressure) were associated with increased
IDWL%. Increasing IDWL% was associated with
increasing phosphorus, creatinine, albumin, potassium
and urea reduction ratio. Increasing IDWL% was
signicantly associated with mortality at 1 year [hazard
ratio (HR) 1.07, P 0.003]. Among patients with
diabetes mellitus, increasing IDWL% was associated
with a mortality HR of 1.03 (P 0.02). Among patients
without diabetes mellitus, increasing IDWL% was not
associated with an increased mortality risk. Increasing
IDWL% is associated with a greater mortality risk

1586

Subjects and methods

Data were taken from the routine analytical les of
Fresenius Medical CareNorth America, Inc. (FMC;
Lexington, MA) for the calendar year 1998 [9,13,14]. The
nal sample was comprised of all patients 18 years and
older receiving three times weekly haemodialysis who were
prevalent on January 1, 1998 and either lived the entire year
on dialysis or died. Urea reduction ratio (URR) and other
laboratory measurements for the last 3 months of the prior
year were averaged. All measurements were determined in a
single laboratory (Spectra Laboratories, Rockleigh, NJ).
Only patients with complete demographic and laboratory
data were included in the cohort.
IDWL was dened as the difference between the average
of pre-dialysis weights and the average of post-dialysis
weights from each dialysis treatment received during the
last 3 months of 1997. Relative intradialytic weight loss
(IDWL%) was expressed as a percentage of IDWL relative
to estimated dry weight as approximated by average postdialysis weight. Pulse pressure was dened as the difference
between pre-dialysis seated systolic and diastolic blood
pressures. The presence or absence of diabetes mellitus was
dened considering it as a co-morbid condition rather than
as the cause of renal failure. Dialysis vintage was dened as
the time period between initiation of renal replacement
therapy and January 1, 1998. Because vintage was positively
skewed, with the majority of patients clustered among lower
values, the variable was transformed to a more normal
distribution using its square root.
The mean and standard deviation for IDWL and
IDWL% were calculated for groups of patients dened by
categories of age, gender, race and presence or absence of

diabetes mellitus. Differences between groups were compared using analysis of variance (ANOVA) and the
Students t-test.
Associations between IDWL% and clinical and demographic variables were estimated using linear regression in
both univariate and multivariable analyses. The fully
adjusted, multivariable linear regression model was built
using both forward and backward stepwise elimination
methods. Variables tested for signicance included case-mix
variables (age, gender, race and the presence of diabetes
mellitus), clinical variables (pre-dialysis systolic blood
pressure, pre-dialysis pulse pressure and the difference
between pre- and post-dialysis systolic blood pressure,
dialysis vintage), post-dialysis weight (as a surrogate for
estimated dry weight), biochemical markers of nutrition
(creatinine, albumin, cholesterol), other laboratory measurements (sodium, phosphorus, haematocrit, glucose, haemoglobin A1c, potassium) and a measurement of dialysis
adequacy (URR). Entry and elimination criteria were set at
a value of P 0.05.
Cox proportional hazards regression was used to estimate
the association between IDWL% and mortality risk in both
unadjusted and adjusted analyses. The model was built
using both forward and backward stepwise elimination
methods testing the variables described above. Interactions
between IDWL% and nutritional markers such as albumin
and creatinine as well as case-mix variables (age, gender,
race, diabetes mellitus), serum sodium and anthropometric
attributes (post-dialysis weight, body mass index) were
tested in separate models. Additional analyses were
performed stratifying on values of those variables that
interacted signicantly with IDWL% incorporating all
variables found to be signicantly associated with mortality
in the full main effects model. For binomial variables (such
as diabetes mellitus), the analysis was repeated stratied
on the presence and absence of the condition represented by
the variable. For continuous variables (such as creatinine),
the analysis was stratied into groups divided by quartiles
of the variable of interest. Because serum sodium was
positively skewed, the analysis was stratied into the
clinically signicant groupings of 135 and 135 meq/l.
Quartiles of patients based on selected variables provided
subsamples with 80% power to detect the association
between IDWL% and mortality demonstrated among the
entire cohort.
All P-values reported are two-sided, and all condence
intervals (CIs) reported are 95% intervals. All analyses were
performed using SAS (version 6.12, SAS Institute Inc.,
Cary, NC).

Results
The demographic and clinical characteristics of patients
in this cohort are described in Table 1. The cohort was
composed of 47.7% patients with diabetes mellitus,
51.4% were male, 50.4% were black, and they had a
mean age of 60.2 years. IDWL and IDWL% for clinical
subgroups dened by gender, age, race and the presence
or absence of diabetes mellitus are shown in Table 2.
Weight loss dened by both IDWL and IDWL% was
greater among younger patients (P < 0.0001 and

Downloaded from http://ndt.oxfordjournals.org/ by guest on February 12, 2012

increasing interdialytic weight gain and better nutrition as measured by normalized protein catabolic rate
(nPCR) and serum albumin [5,7,8]. Patients with
increasing albumin and nPCR have lower mortality
rates [912]. Increased protein-calorie intake is usually
accompanied by larger intake of foodstuffs and
beverages reected by a larger interdialytic weight
gain. While patients with greater interdialytic weight
gain have better serum albumin measurements [5,8]
and greater nPCR [5,7], the impact of these
nutritional factors on the mortality risk associated
with interdialytic weight gain has not been fully
explored. Given the positive associations between a
greater interdialytic weight gain and laboratory
surrogates of nutrition, the relationship between an
elevated weight gain and mortality may be affected by
an individual patients nutritional status. However,
due to small sample size, earlier studies were limited
by their inability to incorporate biochemical and
anthropometric markers of nutritional status and so
examine clinically relevant subgroups of patients [7].
Because intradialytic weight loss represents the
component of each patients volume exposure over
which health care providers have impact by ultraltration, this study was undertaken to explore the
relationship between intradialytic weight loss
(IDWL), nutritional parameters and survival.

L. A. Szczech et al.

Volume exposures and ESRD mortality

1587

Table 1. Description of clinical cohort

% or mean (SD)

Total number of patients

Gender
Male
Female
Age (years)
Race
Black
White
Asian
Native American
Other
Diabetes mellitus
Present
Absent
Laboratory measurements:
Creatinine (mg/dl)
Albumin (g/dl)
Phosphorus (mg/dl)
Potassium (meq/l)
Glucose (mg/dl)
Haemoglobin A1c (%)a
Cholesterol (mg/dl)
Haematocrit (%)
URR (%)
Post-dialysis weight (kg)
Body mass index
Pre-dialysis systolic blood
pressure (seated) (mmHg)
Pulse pressure (mmHg)
Dialysis vintage (days)
Serum sodium (meq/l)

44 114

51.4%
48.6%
60.2 (15.1)
50.4%
42.6%
0.7%
0.7%
4.9%
47.7%
52.3%
9.6
3.9
5.8
4.9
151.5
7.4
171.1
33.3
68.6
72.3
26.4
154.2

(3.3)
(0.4)
(1.6)
(0.7)
(73)
(1.6)
(41.8)
(3.3)
(7.4)
(18.7)
(7.8)
(20.3)

74.9 (15.0)
1261 (1337)
136.8 (3.3)

Among patients with diabetes mellitus.

Table 2. Average IDWL% and IDWL by clinical subgroups

Variable
Entire cohort
Gender
Male
Female
Age (years)e
1840 (reference group)
4160
6180
>80
Race
Black
White
Diabetes mellitus
Present
Absent

IDWL% (SD)

IDWL (SD)

3.9% (1.58)

2.75 kg (1.15)

3.99% (1.53)
3.77% (1.63)a

2.99 kg (1.19)
2.49 kg (1.06)a

4.49%
4.10%
3.67%
3.42%

3.14 kg
3.08 kg
2.54 kg
2.10 kg

(1.78)
(1.58)a
(1.46)a
(1.58)a

(1.24)
(1.21)c
(1.03)a
(0.92)a

3.89% (1.52)
3.84% (1.61)b

2.84 kg (1.16)
2.67 kg (1.15)a

3.91% (1.51)
3.88% (1.63)d

2.86 kg (1.13)
2.66 kg (1.17)a

Using Students t-test: aP < 0.0001; bP 0.009; cP 0.01; dP 0.04.

Using ANOVA: eP < 0.0001 for both IDWL and IDWL%

P < 0.0001) and among patients with diabetes mellitus

(P 0.04 and P < 0.0001). Weight loss was also greater
among men as compared with women (P < 0.0001 and
P < 0.0001) and among blacks as compared with
whites (P 0.009 and P < 0.0001).
The clinical and demographic variables associated
with increasing IDWL% in the multivariable model are
shown in Table 3. Increasing IDWL% was signicantly

Downloaded from http://ndt.oxfordjournals.org/ by guest on February 12, 2012

Variable

associated with decreasing age (P 0.0001), black race

(P 0.02), male gender (P 0.0001), the presence of
diabetes mellitus (P 0.0001), decreasing post-dialysis
weight (P 0.0001), decreasing pre-dialysis pulse
pressure (P 0.002), increasing pre-dialysis systolic
blood pressure (P 0.0001) and increasing dialysis
vintage (P 0.0001). Increasing IDWL% was also
signicantly associated with decreasing cholesterol
(P 0.0001), increasing phosphorus (P 0.0001),
potassium (P 0.0001), creatinine (P 0.002) and
URR (P 0.0001). Among patients with diabetes
mellitus, increasing IDWL% was also associated with
increasing haemoglobin A1c levels (parameter estimate
0.12, P 0.0001, data not shown).
The association between IDWL% and mortality risk
is shown in Table 4. Each increment in IDWL of 1% of
post-dialysis weight was associated with an increase in
the hazard of death of 7% over 1 year [hazard ratio
(HR) 1.07, 95% CI 1.021.012, P 0.003] in the fully
adjusted model. Clinical and demographic variables
associated with an increasing mortality risk include
older age, male gender, the presence of diabetes
mellitus, decreasing pre-dialysis systolic blood pressure,
decreasing post-dialysis weight and increasing dialysis
vintage (all P 0.0001, except pre-dialysis systolic
blood pressure where P 0.0002). Laboratory variables associated with a greater mortality risk include
increasing phosphorus, decreasing albumin, decreasing
haematocrit, decreasing creatinine, decreasing URR,
increasing cholesterol, decreasing sodium and increasing potassium (all P 0.0001, except for cholesterol,
sodium and potassium, where P 0.003, 0.006 and
0.003, respectively).
The associations between diabetes mellitus, creatinine, body mass index, serum sodium and postdialysis weight on mortality interacted signicantly
with IDWL% in separate survival models (P 0.005,
0.0009, 0.0001, 0.005 and 0.02, respectively). This
indicates that the association between IDWL% and
mortality was different among patients with and
without diabetes mellitus and among patients with
increasing values of serum creatinine, body mass
index, serum sodium and post-dialysis weight. In fully
adjusted models stratied on the presence and
absence of diabetes mellitus, increasing IDWL%
was associated with increased mortality among
patients with diabetes (HR 1.03, 95% CI
1.0051.05, P 0.02). Among patients without diabetes mellitus, no signicant association was demonstrated (HR 1.01, 95% CI 0.991.04, P 0.28).
Among patients with diabetes mellitus, neither
glucose nor haemoglobin A1c were signicant
predictors of mortality or interacted with the effect
of IDWL% on mortality (data not shown). Among
patients with serum sodium of 135 meq/l, increasing
IDWL% was associated with an increased risk of
mortality (HR 1.04, 95% CI 1.011.08, P 0.01).
Among patients with serum sodium of 136 meq/l,
increasing IDWL% failed to demonstrate a signicant
association with mortality (HR 1.00, 95% CI
0.971.03, P 0.97).

1588

L. A. Szczech et al.

Table 3. The clinical and demographic variables associated with IDWL% (multivariable model)
Variable

Parameter estimate

F-value

P-value

Post-dialysis weight (per 1 kg increase)

Age (for each increment of 1 year)
Potassium (per 1 meq/dl increase)
Diabetes mellitus (present vs absent)
Phosphorus (per 1 mg/dl increase)
Sex (male compared with female)
Dialysis vintage (per increase in 1 year
of the square root term)
URR (per 1% increase)
Cholesterol (per 1 mg/dl increase)
Pre-dialysis systolic blood pressure
Creatinine (per 1 mg/dl increase)
Pulse pressure (per 1 mmHg increase)
Race (black compared with white)

0.021
0.019
0.32
0.42
0.11
0.39
0.008

1640.2
645.7
505.1
456.9
332.0
311.8
184.0

0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001

0.015
0.0022
0.0067
0.012
0.0042
0.046

136.8
101.7
48.1
9.4
8.7
5.5

0.0001
0.0001
0.0001
0.002
0.002
0.02

Table 4. Predictors of mortality in the Cox proportional hazards model

RR

95% CI

2

IDWL% (per increase of 1%)

Age (per year increase)
Albumin (per 1 g/dl increase)
Dialysis vintage (per increase
in 1 year of square root term)
Phosphorus (per 1 mg/dl increase)
Creatinine (per 1 mg/dl increase)
Pre-dialysis systolic blood
pressure (seated) (mmHg)
Post-dialysis weight (per 1 kg increase)
URR (per 1% increase)
Gender (female compared with male)
Haematocrit (per 1% increase)
Diabetes mellitus (present vs absent)
Decline in systolic blood pressure
(pre-to post-) (per increase in 1 mmHg)
Cholesterol (per 1 mg/dl increase)
Potassium (per 1 meq/l increase)
Sodium (per 1 meq/l increase)

1.07
1.03
0.46
1.013

1.021.12
1.021.03
0.420.50
1.01021.017

9.0
353.4
297.0
143.1

0.003
0.0001
0.0001
0.0001

1.14
0.91
0.992

1.111.17
0.900.93
0.990.994

126.8
118.0
65.8

0.0001
0.0001
0.0001

0.99
0.99
0.80
0.97
1.18
0.995

0.980.99
0.980.99
0.740.87
0.960.98
1.081.28
0.9920.997

60.5
36.9
31.8
29.2
14.7
14.2

0.0001
0.0001
0.0001
0.0001
0.0001
0.0002

0.999
1.09
0.98

0.9981.000
1.031.15
0.970.99

9.1
9.0
7.5

0.003
0.003
0.006

The association between increasing IDWL% and

mortality among patients described by quartiles of
creatinine, body mass index and post-dialysis weight
are shown in Table 5. While patients in the lowest
quartile (creatinine <7.26) showed a similar relationship between increasing IDWL% and mortality to the
cohort overall (HR 1.04, P 0.002), patients whose
creatinine was >7.26 mg/dl did not experience the
same increased mortality risk associated with increasing IDWL%.
The associations between increasing IDWL% and
mortality among patients described by quartiles of
both body mass index and post-dialysis weight are
similar. Patients whose post-dialysis weights were
among the greater two quartiles experienced an
association between IDWL% and mortality
(HR 1.07 and 1.08, P 0.002 and 0.007, respectively). Patients whose weights were in the lower
two quartiles did not experience an increased risk
of mortality associated with increasing IDWL%.

Table 5. Mortality for IDWL% for patients grouped by quartiles

of creatinine, body mass index and post-dialysis weight

Creatinine (mg/dl)
0.57.26
7.279.29
9.3011.59
11.6
Body mass index
<21.73
21.7325.2
25.229.85
>29.85
Post-dialysis weight (kg)
59.3
59.3169.98
69.9882.15
82.16

2

Risk
ratio

95% CI

1.04
1.02
1.02
0.97

1.021.07
0.991.05
0.981.06
0.911.03

1.03
1.02
1.08
1.115

0.991.07
0.981.07
1.021.14
1.051.19

2.4
0.7
6.7
11.6

0.11
0.41
0.009
0.0006

1.002
1.02
1.07
1.08

0.981.03
0.981.05
1.031.12
1.021.13

0.03
1.02
10.2
7.4

0.87
0.31
0.002
0.007

9.5
1.0
0.73
1.17

P-value

0.002
0.31
0.39
0.99

Downloaded from http://ndt.oxfordjournals.org/ by guest on February 12, 2012

Variable

Volume exposures and ESRD mortality

Fig. 1. Mortality risk ratio among patients stratied by quartiles of

weight and creatinine.

Discussion
This study explores the interaction between IDWL%
and nutrition and their effects on mortality.
Consistent with prior research, patients with increasing IDWL% tended to be younger and have
increasing biochemical markers suggesting better
somatic nutrition, including higher serum creatinine
concentration, and markers of greater protein-caloric
intake, including phosphorus, potassium, glucose and
cholesterol [15]. While increasing IDWL% was
associated with an increase in mortality, this relationship was not homogeneous among clinical subgroups.
Increasing IDWL% was associated with a much
greater risk in mortality among patients with diabetes
mellitus as compared with patients without diabetes
mellitus. Additionally, the strength and signicance of
the mortality risk associated with increasing IDWL%
declined among patients with increasing serum
creatinine such that patients with the greatest serum

creatinine values experienced no signicant increase in

death risk associated with increasing IDWL%.
Interestingly, the association between IDWL% and
mortality was also strongest among patients with the
greatest weights and declined at lesser post-dialysis
weights and among patients with lower serum sodium
measurements.
Exposure to volume changes and the relative risk
of mortality may be assessed and analysed in a
number of ways. Weight gain has been reported as
the percentage gain between dialysis treatments
relative to estimated dry weight [5,7], weight gain
greater than or less than 5.7% [6], weight gain greater
than or less than 5.0% [16] and weight gain expressed
in litres [5]. For the purposes of this analysis, we
chose to express this concept as intradialytic weight
loss. It is unclear which arithmetic expression most
accurately reects the risks and physiological
responses to volume expansion and resultant ultraltration requirements. Clinical practice, however, is
often driven by the pre- to post-dialysis weight
change or ultraltration volume required for a
particular dialysis treatment. In this manner, IDWL
reects a compromise between the clinicians perception of the patients volume exposure and their
haemodynamic tolerance of its minimization.
Towards understanding the haemodynamic components reected by IDWL, we included a number of
measures of blood pressure to attempt to reect the
changes seen during ultraltration. These include predialysis measure of pulse pressure, pre-dialysis systolic
blood pressure and the fall in systolic blood pressure
seen before and after dialysis. In models controlling for
these factors, IDWL% continued to have an association with mortality. While these measures do not
entirely describe the haemodynamic changes associated
with ultraltration and how they vary based on cardiac
function, their lack of signicant interaction suggests
that the effect of IDWL% on mortality is not entirely
mediated by these mechanisms.
This is the rst study to demonstrate the
heterogeneous relationship between IDWL% and
mortality among patients with increasing serum
creatinine concentration. Increasing creatinine in the
setting of adequate dialysis is thought to reect a
greater muscle mass and subsequently to be a somatic
marker of better nutrition [12]. As a laboratory
surrogate of somatic nutrition, increasing serum
creatinine is associated with decreased mortality risk
[12]. Patients with a greater serum creatinine
experienced less of an association between IDWL%
and mortality, suggesting that the ability to defend
against decreasing intravascular volume status may be
protected among patients with better nutrition.
Alternatively, patients with better nutrition may
have less co-morbid disease and subsequently are
less affected by larger weight gains, or nally
maintenance of better nutritional status mandates
greater uid removal. Counterintuitive was the
nding that patients with both increasing post-dialysis
weight and increasing body mass index experienced a

Downloaded from http://ndt.oxfordjournals.org/ by guest on February 12, 2012

The mortality risk ratios associated with IDWL% for

patients stratied on quartiles of both creatinine and
weight are shown in Figure 1. The risk associated
with increasing IDWL% was least among patients
with highest creatinine and lowest weight, increasing
as creatinine declined and weight increased. Lower
creatinine for any category of weight was associated
with an increased mortality risk.
The association between IDWL% and mortality
did not interact with gender, race, blood pressure
measurements (pre-dialysis systolic blood pressure,
pre-dialysis pulse pressure or decline in pre- to postdialysis systolic blood pressure) or other laboratory
measurements (albumin, potassium, phosphorus,
cholesterol). The relationships between IDWL and
mortality are similar to those presented for IDWL%
and mortality in direction, magnitude and signicance
(data not shown).

1589

1590

research using a prospective, randomized design

will be required to conrm this relationship.
Additionally, while residual renal function may
affect the weight gain between dialysis treatments,
data reecting individual urine outputs were not
available for this or other similar analyses [5,6].
While the individual components of weight gain
cannot be assessed by this and other studies [5,6], the
combined effect of intradialytic oral intake and
residual renal function is assessed periodically by
the rounding nephrologist and is the net physiological
effect examined here. Finally, the effect of classes
of antihypertensive medications on this relationship
between IDWL% and mortality was not assessed.
Any protective or deleterious effect that may be
conferred through a decrement in systemic vascular
resistance or heart rate by the individual classes of
antihypertensives will need to be explored in future
prospective investigations.
The current study is the rst to demonstrate that
the association between IDWL% and mortality risk
declines among patients with increasing serum creatinine concentrations. It further demonstrates that the
association between IDWL% and mortality is different
among patients with and without diabetes mellitus.
Prospective investigations will be necessary to dene
the optimal IDWL% for each individual patient and
whether a change in IDWL% subsequently will alter
mortality risk. Clinical recommendations for suggested
IDWL% should be made in the context of clinical
factors reective of nutrition and the presence of
diabetes mellitus. With recent recognition that an
expanded dialysis-associated volume can also reect a
greater state of nutrition, acute weight changes may not
be a robust marker of non-compliance. The putative
deleterious effect of dialysis-related volume exposure
on mortality must be interpreted in the context of the
patients nutritional status.
Acknowledgments. L.A.S. is supported by grant DK02724-01A1
from the National Institutes of Health. D.N.R. was supported by
fellowship grant awards from the National Kidney Foundation and
Dr James Clapp. P.S.K. was supported by an American Kidney
Fund Clinical Scientist in Nephrology Fellowship Award. J.C. was
supported by a National Research Service Award T32 HS0079-03
from the Agency of Healthcare Research and Quality. B.C. is
supported by a fellowship award from the National Medical
Research Council, Singapore. This study was presented in abstract
form at the American Society of Nephrology Annual Meeting,
October 2000.

References
1. Cannella G, Picotti GB, Mioni G, Cristinelli L, Maiorca R.
Blood pressure behaviour during dialysis and ultraltration.
A pathogenic hypothesis on hemodialysis-induced hypotension.
Int J Artif Org 1978; 1: 6975
2. Fishbane S, Natke E, Maesaka JK. Role of volume overload in
dialysis-refractory hypertension. Am J Kidney Dis 1996; 28:
257261
3. Rahman M, Fu P, Sehgal AR, Smith MC. Interdialytic
weight gain, compliance with dialysis regimen, and age are

Downloaded from http://ndt.oxfordjournals.org/ by guest on February 12, 2012

stronger association between IDWL% and mortality

risk. The difference between creatinine and both body
mass index and weight may reect the importance of
body composition rather than size in the relationship
between IDWL% and mortality.
While the relationship between albumin and
mortality was similar to that previously reported
[9,12,17], the effect of IDWL% on mortality did not
vary based on a patients albumin measurement. This
may reect uctuations in albumin related to
intercurrent illness or other inammatory states that
limit the ability of albumin to represent only the
nutritional state in the short term [18,19]. The
relationship between IDWL% and mortality varied
based on a patients serum sodium concentration.
While serum sodium is affected by many factors
including diet, dialytic parameters and residual renal
function, the use of serum sodium may provide a
measure to compare volume and nutritional status
(hyponatraemia secondary to hypervolaemia).
These ndings are consistent with prior reports
demonstrating that the relationship between increasing IDWL% and mortality is different in the presence
and absence of diabetes mellitus [7]. In a prospective
study of 283 patients, Kimmel et al. [7] demonstrated
that patients without diabetes mellitus did not
experience an increased mortality risk with increasing
IDWL%. While interpretation of this study is limited
by its small sample size (n 164 without diabetes
mellitus), the nding that IDWL% is not signicantly
associated with an increased mortality risk among
patients without diabetes mellitus was conrmed here
using a cohort with greater power to examine this
clinical subgroup. Among patients with diabetes
mellitus, this association may be related to the
relationship between increased IDWL% and poor
glycaemic control, which portends a poorer survival
[20]. Alternatively, patients with diabetes mellitus may
also have a greater prevalence of cardiovascular
disease and other co-morbidities, not assessed in this
analysis, which may augment the risk associated with
increasing IDWL%.
While the relationship between IDWL% and
mortality is signicant, the strength of this relationship
relative to other predictors of mortality must be noted.
The 2 for IDWL% is comparable with that of
cholesterol and potassium and only slightly less than
that of diabetes mellitus. Relative to the predictive
power of albumin and age as assessed by the 2 term for
each variable, IDWL% accounts for only a small
amount of the variance in the model. Given this,
clinical and research efforts should be directed toward
those factors that may be, in some part, actionable,
such as blood pressure, divalent cations and nutritional
surrogates, to attempt to improve mortality risk.
While this study demonstrates heterogeneity in the
association between IDWL% and mortality among
patients with ESRD, its conclusions should be
interpreted in the setting of certain limitations. As a
retrospective cohort study, these associations cannot
be ascribed a causeeffect relationship. Further

L. A. Szczech et al.

Volume exposures and ESRD mortality

4.

5.

6.

7.

8.

9.

11. Acchiardo SR, Moore LW, Latour PA. Malnutrition as the

main factor in morbidity and mortality of hemodialysis
patients. Kidney Int 1983; 16: S199S203
12. Lowrie EG, Lew NL. Death risk in hemodialysis patients: the
predictive value of commonly measured variables and an
evaluation of death rate differences between facilities. Am J
Kidney Dis 1990; 15: 458
13. Owen WF, Chertow G, Lazarus JM, Lowrie EG. The dose of
hemodialysis: mortality responses by race and gender. J Am
Med Assoc 1998; 280: 16
14. Szczech LA, Lowrie EG, Li Z, Lew N, Lazarus JM, Owen WF.
Changing hemodialysis thresholds for optimal survival. Kidney
Int 2001; 59: 738745
15. Ifudu O, Uribarri J, Rajwani I et al. Relation between
interdialytic weight gain, body weight and nutrition in
hemodialysis patients. Am J Nephrol 2002; 22: 363368
16. Testa A, Beaud JM. The other side of the coin: interdialytic
weight gain as an index of good nutrition. Am J Kidney Dis
1998; 5: 830834
17. Goldwasser P, Mittman N, Antignani A et al. Predictors of
mortality in hemodialysis patients. J Am Soc Nephrol 1993; 3: 1613
18. Gitlin JD, Colten HR. Molecular biology of the acute phase
plasma proteins. In: Pick E, Landy M, eds. Lymphokines, Vol.
14. Academic Press, San Diego, CA: 1987; 123153
19. Kaysen GA. Biological basis of hypoalbuminemia in ESRD.
J Am Soc Nephrol 1998; 9: 23682376
20. Ifudu O, Dulin AL, Friedman EA. Interdialytic weight gain
correlates with glycosylated hemoglobin in diabetic hemodialysis patients. Am J Kidney Dis 1994; 23: 686691
Received for publication: 27.9.02
Accepted in revised form: 6.3.03

Downloaded from http://ndt.oxfordjournals.org/ by guest on February 12, 2012

10.

independent predictors of blood pressure in hemodialysis

patients. Am J Kidney Dis 2000; 35: 257265
Harnett JD, Kent GM, Barre PE, Taylor R, Parfrey PS. Risk
factors for the development of left ventricular hypertrophy in a
prospectively followed cohort of dialysis patients. J Am Soc
Nephrol 1994; 4: 14861490
Sherman RA, Cody RP, Rogers ME, Solanchich JC.
Interdialytic weight gain and nutritional parameters in
chronic hemodialysis patients. Am J Kidney Dis 1995; 25:
579583
Leggat JE, Orzol SM, Hulbert-Shearon TE et al. Noncompliance
in hemodialysis: predictors and survival analysis. Am J Kidney
Dis 1998; 32: 139145
Kimmel PL, Varela MP, Peterson RA et al. Interdialytic
weight gain and survival in hemodialysis patients: effects of
duration of ESRD and diabetes mellitus. Kidney Int 2000; 57:
11411151
Sezer S, Ozdemir FN, Arat Z, Perim O, Turan M, Haberal M.
The association of interdialytic weight gain with nutritional
parameters and mortality risk in hemodialysis patients. Renal
Failure 2002; 24: 3748
Owen WF, Lew NL, Liu Y, Lowrie EG, Lazarus JM. The urea
reduction ratio and serum albumin concentration as predictors
of mortality among patients undergoing hemodialysis. N Engl J
Med 1993; 329: 10011006
Parker TF, Laird NM, Lowrie EG. Comparison of the study
groups in the National Cooperative Dialysis Study and a
description of morbidity, mortality, and patient withdrawal.
Kidney Int 1983; 13: S42S49

1591