See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/7488671

Estimation of total –body and limb muscle mass in hemodialysis patients by

using multi frequency bioimpedance spectroscopy

Article  in  American Journal of Clinical Nutrition · December 2005

DOI: 10.1093/ajcn/82.5.988 · Source: PubMed

CITATIONS READS

58 67

8 authors, including:

George A Kaysen Steven B Heymsfield

University of California, Davis Pennington Biomedical Research Center
307 PUBLICATIONS   13,098 CITATIONS    1,022 PUBLICATIONS   59,330 CITATIONS   

SEE PROFILE SEE PROFILE

Charoen Kaitwatcharachai Martin Kuhlmann

Hat Yai Hospital Vivantes Klinikum im Friedrichshain
18 PUBLICATIONS   568 CITATIONS    165 PUBLICATIONS   3,736 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Body Composition View project

Dialysis Outcomes View project

All content following this page was uploaded by George A Kaysen on 11 March 2014.

The user has requested enhancement of the downloaded file.

Estimation of total-body and limb muscle mass in hemodialysis
patients by using multifrequency bioimpedance spectroscopy1–3
George A Kaysen, Fansan Zhu, Shubho Sarkar, Steven B Heymsfield, Jack Wong, Charoen Kaitwatcharachai,
Martin K Kuhlmann, and Nathan W Levin
ABSTRACT
Background: Skeletal muscle mass can be measured noninvasively
with magnetic resonance imaging (MRI), but this is time-consuming
and expensive.
Objective: We evaluated the use of multifrequency bioimpedance
spectroscopy (BIS) measurements of intracellular volume (ICV) to
model total-body skeletal muscle mass (TBMM) and limb skeletal
muscle mass in hemodialysis patients.
Design: TBMM was measured by MRI in 20 male and 18 female
hemodialysis patients with a median (range) age of 54 y (33–73 y),
weight of 78.9 kg (43.2–120 kg), and body mass index (BMI; in
kg/m2) of 27.3 (19.4 – 46.6). We measured total body water (TBW)
by using D2O dilution, extracellular volume (ECV) as bromide
space, and ICV as TBW minus bromide space. Total body potassium
(TBK) measured as 40K was used as an independent model of
TBMM. BIS was used to measure whole-body TBW (ankle to wrist)
and TBW in the arms and legs. BIS-estimated ICV was used to
construct models to calculate limb muscle mass and TBMM. The
latter was compared with models derived from isotopic methods.
Results: BIS yielded a model for TBMM [TBMM ҃ 9.52 ѿ 0.331
҂ ICV ѿ 2.77 (male) ѿ 0.180 ҂ weight (kg) Ҁ 0.133 ҂ age] (R2 ҃
0.937, P 쏝 0.0001) as precise as TBK-measured TBMM [TBMM ҃
1.29 ѿ 0.00453 ҂ TBK (mEq) ѿ 1.46 (male) ѿ 0.144 ҂ weight (kg)
Ҁ 0.0565 ҂ age] (R2 ҃ 0.930, P 쏝 0.0001) or isotopic methods. BIS
models were also developed for measuring leg and arm muscle mass.
Conclusion: BIS provides an estimate of TBMM that correlates well
with isotopic methods in approximating values obtained by MRI and
can be used to estimate limb muscle mass. Am J Clin Nutr 2005;
82:988 –95.
KEY WORDS Body composition, body cell mass, nutrition,
extracellular fluid, total body water, bioimpedance spectroscopy,
hemodialysis patients
INTRODUCTION
The evaluation of nutritional status in dialysis patients is both
important and quantitatively difficult. Changes in body compo-
sition consistent with malnutrition are frequent in long-term he-
modialysis patients and are powerful predictors of mortality (1,
2). Anthropometric measurements of nutritional status fre-
quently are operator dependent. Formulas using those measure-
ments to calculate total body water (TBW) are imprecise and in
poor agreement with direct measurements provided by isotope
988 Am J Clin Nutr 2005;82:988 –95. Printed in USA. © 2005 American Society for Nutrition
dilution, especially in the obese and in dialysis patients. Body
mass index (BMI) is associated with a low relative risk of mor-
tality in hemodialysis patients (3). Body composition may differ
in patients having the same BMI. Creatinine can be used to model
the relative effects of fat and muscle mass (3), but the availability
of an independent method of assessing muscle mass could have
value in distinguishing the contributions of muscle and fat to
Downloaded from ajcn.nutrition.org by guest on June 12, 2013
morbidity associated with different BMIs.
Magnetic resonance imaging (MRI) provides a precise mea-
sure of the volume of fat, muscle, and bone compartments,
depending on the number of slices (4). Volumes are then
transformed to masses. Muscle mass is 1.04 times muscle
volume (5, 6).
We designed this study to establish whether regression models
could be developed to estimate whole-body muscle mass as well
as limb skeletal muscle mass in hemodialysis patients by using
bioimpedance spectroscopy (BIS) and whether the results ob-
tained were not significantly different from those obtained by
using other standard and accepted techniques. Measurement of
the quantity of potassium in the TBK also defines the intracel-
lular volume (ICV) (7). The clinical use of measurement of TBK
is limited, however, by the availability of instrumentation. Iso-
topic methods also provide precise evaluations of intracellular
and extracellular compartments. Estimation of TBW by using the
dilution of isotopically labeled water (D2O) combined with eval-
uation of extracellular space by using bromide allows the calcu-
lation of ICV by subtraction.
Because of its availability and simplicity, bioelectrical imped-
ance analysis (BIA) has significant potential as a complement to
standard anthropometric techniques for assessing the nutritional
status of dialysis patients (8). Whereas single-frequency BIA
1
From the Renal Research Institute, New York, NY (GAK, FZ, SS, CK,
MKK, and NWL); the Division of Nephrology, Department of Medicine,
University of California, Davis, Davis, CA (GAK); the Research Service, VA
Northern California Health Care System, Mather, CA (GAK); and St Lukes
Hospital, New York, NY (SBH and JW).
2
Supported by a grant from the Renal Research Institute, by grant no.
PO1-DK42618 from the National Institutes of Health (to SBH), and by the
Research Service of the US Department of Veterans Affairs.
3
Reprints not available. Address correspondence to GA Kaysen, Division
of Nephrology, University of California, Davis, Genome and Biomedical
Sciences Facility, 451 East Health Sciences Drive, Davis, CA 95616. E-mail:
gakaysen@ucdavis.edu.
Received September 22, 2004.
Accepted for publication July 20, 2005.
 TOTAL-BODY AND LIMB MUSCLE MASS MEASURED WITH BIS
provides data that are useful for predicting mortality in popula-
tions (9, 10), the output data (phase angle and reactance) are not
readily comparable to other methods that provide estimates of
physiologically recognizable compartments, such as ICV, extra-
cellular volume (ECV), and fat-free mass (8). Furthermore, BIA
relies on a single frequency, 50 kHz (11).
Isotopic methods are restricted to measurement of whole-body
fluid volumes. In contrast, both MRI and BIS (depending on
electrode placement) can measure regional tissue compartments
or fluid distribution. Thus, we also sought to ascertain whether
limb muscle mass as measured by MRI could be modeled by
measurements of ICV in the arms and legs.
SUBJECTS AND METHODS
Subjects
Forty-two hemodialysis patients (19 F, 23 M) aged 쏜18 y were
chosen so as to encompass a wide range of BMIs and ages. All
subjects had been on maintenance hemodialysis for 욷3 mo be-
fore the study.
All subjects provided written informed consent to their par-
ticipation. The institutional review boards of St Luke’s, Beth
Israel, and the University of California, Davis, hospitals ap-
proved the study protocol.
Methods
On the day of a regularly scheduled hemodialysis session, the
subjects were studied 앒3 h before initiation of the dialysis treat-
ment. The study began with the administration of D2O and so-
dium bromide, as described below. After the administration of
these isotopes, the subject was then placed in the 4-pi counter. On
exiting the counter, the patient was placed in a recumbent posi-
tion as described below for measurement of bioimpedance. After
this measurement, the patient underwent an MRI scan to measure
muscle mass. The patient was then transported to the dialysis
unit, where blood was obtained before the initiation of dialysis to
measure D2O and bromide.
After an overnight fast, body weight was measured to the
nearest 0.1 kg (Weight Tronix, Pinbrook, NJ), and height was
measured to the nearest 0.5 cm by using a stadiometer (Holtain,
Crosswell, United Kingdom). A sample of blood (5 mL) was
obtained for measurement of baseline deuterium and bromide.
D2O at a dose of 500 mg/kg body wt and 0.06 g of 4 mol NaBr/L
were then given by mouth. Whole-body MRI scans were pre-
pared by using a 1.5 Tesla scanner (6X Horizon; General Elec-
tric, Milwaukee, WI) to evaluate muscle mass (12). The 4-pi
whole-body counter was used to measure 40K (13).
Both whole-body (wrist to ankle) and segmental BIS (SBIS,
Xitron 4200; Xitron, San Diego, CA) measurements that used
multiple frequencies from 5 kHz to 1000 kHz were recorded
before dialysis to include defined segmental BIS in the leg and
arm on the nonaccess side of the body in those patients having an
arm access. In those patients with a catheter for dialysis access,
the dominant arm was chosen. Two electrodes were placed on the
hand, wrist, and foot for injecting current at a frequency between
5 kHz and 1 MHz. Two electrodes were placed on the wrist and
ankle for whole-body BIS measurement, and another 2 elec-
trodes were placed on the shoulder and greater trochanter to
measure the fluid compartments in the arm (between electrodes
989
at wrist and shoulder) and leg (between electrodes at great tro-
chanter and ankle). A switch was used to transfer data accumu-
lation from the whole body and segments to the Xitron device (4,
12). Raw data derived from these measurements included resis-
tance, reactance, phase angle, and impedance. A program pro-
vided by Xitron based on the Cole-Cole model was used to
calculate segmental and wrist-to-ankle extracellular and intra-
cellular resistance (RE and RI, respectively (14). Calculations of
wrist-to-ankle ECV and ICV were based on the variables RE, RI,
and body weight and included constants for resistivity, body
geometry, and body density (14). Previously published equations
(14, 15) were used to calculate ECV and ICV (see below). TBW
was calculated as the sum of ECV and ICV by using the whole-
body measurements alone.
Patients were then transported to the hemodialysis unit and
again weighed at 3 h after the initial dose of D2O and bromide;
blood for measurement of both D2O and bromide was obtained
before dialysis. At that time, immediately before the initiation of
dialysis, an additional 5 mL blood was drawn for measurement of
both D2O and bromide so that their respective volumes of dis-
tribution could be determined by isotope dilution (16). The dose
Downloaded from ajcn.nutrition.org by guest on June 12, 2013
concentration in the collected specimen is measured on a single-
frequency infrared spectrophotometer after the specimen is lyo-
philized. The TBW volume is calculated by dividing the dose by
the net D2O concentration in specimen. The precision for the
TBW measurement by this method is ⫾ 1.5%. Measurements of
bromide concentration are carried out on an HPLC system (16).
The CV of measurement of bromide dilution is 1.4%.
Wrist-to-ankle (whole-body) ECV, ICV, and TBW
(ECVѿICV) were calculated as follows (14):
ECV ⫽ KECV(H2 冑W/RE)2/3 (1)
where H is body height in cm, W is body mass in kg, and RE is
given in ⍀. KECV is a factor related to body geometry, density,
and resistivity, according to the following equation (14):
KECV ⫽ 1/1000(K2B␳ECV
2
/DB)1/3 (2)
where ␳ECV is the resistivity of extracellular fluid, DB is body
density considered as a constant value (kg/L), and KB is a coef-
ficient of body height related to geometric limb size (14). KB is
considered a constant on the basis of the assumption that there are
relatively uniform proportions between limbs and total body
height (14). Intracellular fluid volume was calculated by using
the following equation (14):
1 ⫹ (ICV/ECV)5/2 ⫽ (RE ⫹ RI/RI)
⫻ (1 ⫹ K␳ICV/ECV) (3)
where ␳ICV is the resistivity of intracellular fluid, k␳ is the ratio of
intracellular to extracellular fluid resistivity (k␳ ҃ ␳ICV/␳ECV),
and RI is given in ⍀. ␳ICV was 273.9 ⍀ cm for males and 264.9 ⍀
cm for females, whereas ␳ECV was 40.5 ⍀ cm for males and 30 ⍀
cm for females.
Segmental ECV and ICV were calculated for the arms and legs
by using the following equations:
ECVi ⫽ 2/1000(␳E ⫻ Li2/REi) (4)
and
ICVi ⫽ 2/1000(␳I ⫻ Li2/RI,i) (5)
990 KAYSEN ET AL

where ␳ represents factors for extracellular (␳E ҃ 47 ⍀ cm) and TABLE 1

intracellular (␳I ҃ 273.9 ⍀ cm) resistivity, Li represents segmen- Body composition of 38 hemodialysis patients measured immediately
tal length, REi represents segmental extracellular resistance, and before hemodialysis1
RI represents segmental intracellular resistance (the subscript i 10th 90th
represents the arm or leg) (15). Variable Median percentile percentile Range
Segmental ECV and ICV of the arms and legs were derived by
Age (y) 53.5 40.8 69 33–73
applying the respective parts of equations 4 and 5 separately.
Height (m) 1.695 1.53 1.80 1.49–1.85
Weight (kg) 78.9 57.6 102 43.2–120
Establishing error in BIS measurement of ICV in BMI (kg/m2) 27.5 23.1 37.2 19.4–46.6
hemodialysis patients BSA (m2) 1.90 1.55 2.16 1.34–2.35
When performed 3 times a week, hemodialysis is generally TBK (mEq) 3282 2232 4191 2020–4345
TBW (L)2 40.6 29.4 49.7 28.3–54.2
conducted after either a 1- or 2-d interval. It is possible, therefore,
TBW (L)3 41.4 25.4 52.1 22.3–57.7
that fluid volumes may be affected by the span of time between ECV (L)4 19.0 10.6 23.7 6.7–28.6
dialysis treatments, during which an artifact may be introduced. ECV (L)3 19.2 13.4 23.2 11.2–26.8
To establish the precision of ICV measurement with BIS in ICV (L)5 21.6 14.7 27.5 13.3–28.6
dialysis patients and to ascertain whether ICV measurements ICV (L)6 22.8 15.7 29.4 14.6–33.9
were sensitive to the time elapsed before treatment, we measured ICV (L)3 22.0 12.7 28.8 10.5–31.7
ICV repeatedly in 13 hemodialysis patients. Measurements were TBMM (kg)7 24.1 15.4 33.0 12.6–38.4
all conducted within the 2 h before initiation of hemodialysis but Arm muscle mass (kg)7 3.85 2.18 5.22 2.08–5.72
either 1 or 2 d after the previous treatment. Four of the patients Leg muscle mass (kg)7 11.3 7.7 16.0 6.1–16.1

Downloaded from ajcn.nutrition.org by guest on June 12, 2013

were women. 1
BSA, body surface area; TBK, total body potassium; TBW, total body
water; ECV, extra cellular volume; ICV, intracellular volume; TBMM, total-
Statistical analysis body muscle mass.
2
Measured with deuterium oxide dilution.
Models were created by using a forward stepwise multiple 3
Measured with bioimpedance spectroscopy.
regression model (JMP software, version 5.0.1; SAS Inc, Cary, 4
Measured with sodium bromide dilution.
NC). The dependent variable was muscle mass measured by 5
Measured with potassium dilution.
MRI. Models were created by using age; sex; weight; BMI; 6
Measured with deuterium oxide minus sodium bromide.
height; body surface area; diabetic status; and either TBW mea- 7
Measured with magnetic resonance imaging.
sured with D2O dilution, ECV measured by using bromide dilu-
tion, or ICV measured by using the difference between these
variables (isotopic method), TBK, and BIS estimates of com-
dilution (Table 1) and by using BIS were significantly corre-
partmental fluid volumes. In the latter instance, BIS estimates of
lated. (R2 ҃ 0.785, P 쏝 0.0001). No bias was evident by
TBW, ICV, and ECV were initially all entered as independent
Bland-Altman analysis.
variables. Recognizing that ECV, unlike ICV, may vary in dial-
ECV measurements calculated from the bromide volume of
ysis patients independently of body cell mass because of fluid
distribution and from whole-body BIS also were not significantly
retention between dialysis treatments, a BIS model was forced
with the use of ICV only and compared with the best-fit model
using all fluid compartments. TABLE 2
To estimate the independent contribution of BIS measure- Models for estimation of total-body muscle mass (TBMM)1
ments in the absence of sex, weight, age, BMI, and height, BIS Model for TBMM
estimates of ICV only were added, and other variables were
deliberately omitted. MRI measurement of muscle volume was BIS TBK2 Isotope
multiplied by 1.04 to calculate muscle mass in kilograms (5, 6).
Constant 9.52 1.29 10.5
Variables that did not contribute significantly to the regression ICV coefficient3 0.331 0.00453 0.101
model were not included in the final regression output. Male sex 2.77 1.46 3.32
Weight (kg) 0.180 0.144 Ҁ0.0565
Age (y) Ҁ0.113 Ҁ0.0565 Ҁ0.113
RESULTS SEE (kg) 1.85 1.84 2.00
R2 0.937 0.930 0.916
Complete measurements were available in 38 of the subjects
P 쏝0.0001 쏝0.0001 쏝0.0001
(20 men, 18 women) (Table 1). The patients were from the
multiethnic area of Upper Harlem in New York City: 33 patients 1
n ҃ 38. BIS, bioimpedance spectroscopy; TBK, total body potassium;
were African American, 3 were Hispanic, and 2 were Asian. ICV, intracellular volume. The dependent variable was muscle mass mea-
Twelve of the patients were diabetic. sured by magnetic resonance imaging. Best-fit multiple regression models
were used for the correlation between TBMM measured by magnetic reso-
Several models were fitted with the MRI measurement of
nance imaging and modeled by using sex, weight, and age and ICV measured
muscle volume by using either isotopically derived data (D2O
by using BIS or as D2O Ҁ bromide distribution volumes.
and bromide), TBK data, or data derived from anthropometric 2
The model using measurement of TBK was based on milliequivalents
measures and BIS (Table 2). Total mean (앐SD) skeletal muscle of TBK rather than adjustment to volume. Equations are formulated as
as measured by MRI was 24.0 앐 1.1 kg (range: 12.6 –38.4 kg) TBMM ҃ constant ѿ a ҂ ICV ѿ b ҂ sex ѿ c ҂ weight ѿ d ҂ age, where
(Table 1). Muscle represented 30.6% of body weight (range: a is the ICV coefficient, b is sex, c is weight, and d is age.
20.9 Ҁ41.8%). TBW measurements obtained by using D2O 3
Values are mL, except TBK, which is mEq.
 TOTAL-BODY AND LIMB MUSCLE MASS MEASURED WITH BIS
FIGURE 1. Correlation between total-body muscle mass (TBMM) mea-
sured by magnetic resonance imaging (MRI) and a regression model devel-
oped by using ankle-to-wrist multifrequency bioimpedance spectroscopy
(BIS), age, sex, and weight (whole-body BIS model). The equation used was
TBMM (kg) ҃ 9.52 ѿ 0.331 ҂ BIS-derived ICV (L) ѿ 2.77 (male) ѿ 0.180
҂ weight (kg) – 0.113 ҂ age (y). The results of this equation give a value of
SE ҃ 1.85 kg (R2 ҃ 0.937, P 쏝 0.0001). The dependent variable was TBMM
measured by using MRI. ICV, intracellular volume.
different, and they correlated with one another (Table 1) (P 쏝
0.0001, R2 ҃ 0.501). No bias was evident. ICV calculated on the
basis of TBK did not differ significantly from that measured with
whole-body BIS. Whole-body BIS-measured ICV was (x៮ 앐
SEM) 21.0 앐 0.92 L, and that calculated from TBK was 20.9 앐
0.72 L (x៮ difference: Ҁ0.28 앐 0.41 L) (P ҃ 0.532). The corre-
lation between the 2 values was 0.889 (P 쏝 0.0001).
TBW in dialysis patients may vary independently of intracel-
lular water because of failed renal function and the resulting
expansion of ECV. Furthermore, ECV, even in a given dialysis
patient, will vary according to the time since the previous dialysis
treatment. Accordingly, we based our BIS model on measure-
ments of ICV alone. With the use of this measure, the best-fit
multiple regression model included sex, weight, and age (Ta-
ble 2 and Figure 1). The coefficients of the regression models
are presented in tabular form. Median TBMM estimated by the
BIS model was 27.2 kg (x៮ 앐 SEM: 26.0 앐 1.0 kg; range:
15.4 –35.9 kg).
Bland-Altman analysis found bias (R2 ҃ 0.259, P 쏝 0.001)
(Figure 2). Median TBMM estimated by the TBK model was
25.5 kg (x៮ 앐 SEM: 24.3 앐 1.02 kg; range: 14.9 –33.6 kg).
Bromide space (ECV) was eliminated from the regression
model when the D2O-measured volume of distribution was in-
cluded. Much as was seen with BIS analysis, when TBW (vol-
ume of distribution of D2O) was intentionally omitted from the
regression model, ICV calculated as the difference between D2O
and bromide distribution volume (ICV bromide) provided a
model that did not differ significantly from that achieved by the
other methods (Table 2). Median TBMM estimated by the iso-
tope dilution model using ICV was 26.7 kg (x៮ 앐 SEM: 25.5 앐 0.9
kg; range: 17.3–35.2 kg).
The correlation between TBMM measured with MRI and
measured by using the BIS model described above did not differ
significantly from that obtained by using TBK plus the same
demographic variables (ie, weight, sex, and age) or from that
obtained with the model that used isotope dilution plus age, sex,
and weight (Table 2). Thus, ICV measured either with BIS or
with isotope dilution provided a model that closely correlated
991
FIGURE 2. Bland-Altman plot of the difference between magnetic res-
onance imaging (MRI) and bioimpedance spectroscopy (BIS) estimates of
total-body skeletal muscle mass (TBMM) model versus the average value for
total body muscle computed by the model and TBMM measured with MRI
[x៮ 앐 SD estimate: 2.64 앐 2.03 kg (R2 ҃ 0.259, P 쏝 0.001)].
Downloaded from ajcn.nutrition.org by guest on June 12, 2013
with TBMM measured with MRI. In patients studied to establish
the precision of BIS measurements and the variability introduced
by time since the previous dialysis treatment, weight was 77.7 앐
15.9 kg (range: 56.2–102.4 kg), BMI was 27.0 앐 0.9 (range:
19.4 – 46.6), and ICV was 18.5 앐 5.0 L (range: 10.8 –26.7 L).
ICV was measured 3 times in each patient. The CV of the mea-
surement of ICV was 3.78%. ICV was independent of the length
of time after the previous dialysis. ICV measured 1 d after a
dialysis treatment was 18.6 앐 1.54 L, and that measured 2 d after
treatment was 18.5 L 앐 1.61 L (P ҃ 0.599, paired t test).
Understanding that sex, weight, and age and could predict
TBMM without additional information on TBW or ICV, we
modeled TBMM versus TBK (mEq), isotope dilution methods,
and the measurement of ICV by using BIS alone to establish the
specific precision of each volume measurement and by including
BIS in the absence of other information, as shown in the follow-
ing equation:
TBMM(kg) ⫽ 2.074 ⫹ 1.064
⫻ BIS-derived ICV(L) (6)
where ICV is measured with whole-body BIS (SE ҃ 3.28 kg;
R2 ҃ 0.783, P 쏝 0.001) (Figure 3).
Bland-Altman analysis using only BIS-derived data for esti-
mation of muscle mass showed no significant bias (R2 ҃ 0.0365).
Thus, regression models that correlate with TBMM measured
with MRI can be obtained by using TBK, isotope dilution, and
whole-body BIS measurements. However, it may be possible to
use segmental measures of fluid compartments with BIS— but
not with either TBK or isotope dilution—to estimate arm or
leg muscle mass. To test this hypothesis, we developed mod-
els relating BIS fluid compartment measurements of the arms
and legs to MRI measurements of skeletal muscle mass in
these extremities.
To avoid possible variability in the ECV compartment in di-
alysis patients, the final models for the arms and legs were cal-
culated with the use of BIS measurement of ICV. Total leg
muscle measured with MRI could be modeled with the use of sex,
992 KAYSEN ET AL

FIGURE 3. Correlation between total-body muscle mass (TBMM) mea-

sured by magnetic resonance imaging (MRI) and a regression model devel-
oped by using ankle-to-wrist multifrequency bioimpedance spectroscopy
(BIS) measurement of intracellular volume (ICV) alone with no demographic
variables. The equation used was TBMM (kg) ҃ 2.074 ѿ 1.064 ҂ BIS-
derived ICV (L). The results of this equation give a value of SE ҃ 3.28 kg
(R2 ҃ 0.78, P 쏝 0.001). Bland-Altman analysis found no bias [SD ҃ 0.94 앐
3.11 kg (R2 ҃ 0.0365, P ҃ 0.15); data not shown]. The dependent variable
FIGURE 4. Correlation between leg muscle mass measured by magnetic
resonance imaging (MRI) and a regression model developed by using ankle-
to-hip multifrequency bioimpedance spectroscopy (BIS), age, sex, and weight.
The equation used was MRI leg skeletal muscle (kg) ҃ 4.53 ѿ 0.5005 앐 ICV
(L) ѿ 1.535 (male) ѿ 0.0839 ҂ weight (kg) Ҁ 0.0483 ҂ age (y). The results of
this equation give a value of SE ҃ 1.16 kg (R2 ҃ 0.890, P 쏝 0.001). The

Downloaded from ajcn.nutrition.org by guest on June 12, 2013

dependent variable was leg skeletal muscle mass measured by MRI. ICV, intra-
was skeletal muscle mass measured by using MRI. cellular volume.

age, weight, and segmental ICV by using an ankle-to-hip elec-
trode (Table 3, Figure 4).
Median leg muscle mass as measured with MRI was 11.3 kg
(range: 6.1–16.1 kg) (Table 1), and total leg muscle mass as
measured by using the BIS model was 12.3 kg (x៮ 앐 SEM:
13.0 앐 0.5 kg; range: 7.5–18.5 kg). Bland-Altman analysis
showed some degree of bias (R2 for the regression ҃ 0.178, P ҃
0.01) (Figure 5). When demographic and anthropometric data
were not used, BIS modeling of the leg yielded a strong corre-
lation, but bias was significantly reduced, as seen in Figure 6
and in the following equation:
Bland-Altman analysis showed no significant bias and an in-
tercept of nearly zero (R2 ҃ 0.088, P ҃ 0.073). Analysis of arm
composition gave comparable results (Table 3). Total arm mus-
cle mass measured by using MRI was 3.9 kg (range: 2.1–5.7 kg),
and that measured by using the BIS model was 4.3 kg (x៮ 앐 SEM:
4.00 앐 0.14 kg; range: 2.7–5.6 kg), as shown in Figure 7 and the
following equation:
MRI arm muscle mass (kg) ⫽ 1.69 ⫹ 0.301 ⫻ ICV(L)
⫹ 0.603 male ⫹ 0.0234 ⫻ weight (kg)

MRI leg muscle mass (kg) ⫽ 4.44 ⫹ 1.21 ⫺ 0.123 ⫻ age (y) (8)
where the relation was SE ҃ 0.306 kg (R ҃ 0.933, P 쏝 0.0001).
2
⫻ ICV (L) (7)
where the relation was SE ҃ 2.03 kg (R2 ҃ 0.63, and P 쏝 0.001).

TABLE 3
Models for estimation of segmental muscle mass measured by intracellular
volume (ICV) as measured by bioimpedance spectroscopy1
Compartment measured

Leg Arm

Constant 4.53 1.69

ICV coefficient (L) 0.5005 0.301
Male sex 1.535 0.603
Weight (kg) 0.0839 0.0234
Age (y) Ҁ0.0483 Ҁ0.123
SEE (kg) 1.16 0.306
R2 0.890 0.933
P 쏝0.001 쏝0.001
1
n ҃ 38. The dependent variable was muscle mass measured by mag-
netic resonance imaging. Best-fit multiple regression models were used for
FIGURE 5. Bland-Altman plot of the difference between magnetic res-
the correlation between total limb muscle mass measured by magnetic res- onance imaging (MRI) and bioimpedance spectroscopy (BIS) estimates
onance imaging and modeled by using sex, weight, age, and ICV measured of leg skeletal muscle mass model versus the average value for leg muscle
by bioimpedance spectroscopy. Equations are formulated as TBMM ҃ con- mass computed by the intracellular volume (ICV) model and leg muscle
stant ѿ a ҂ ICV ѿ b ҂ sex ѿ c ҂ weight ѿ d ҂ age, where a is the ICV mass measured by MRI [x៮ 앐 SD estimate: Ҁ 0.98 앐 1.16 kg (R2 ҃ 0.178,
coefficient, b is sex, c is weight, and d is age. P ҃ 0.01)].
 TOTAL-BODY AND LIMB MUSCLE MASS MEASURED WITH BIS
FIGURE 6. Correlation between total leg muscle mass measured by
magnetic resonance imaging (MRI) and a regression model developed by
using only the bioimpedance spectroscopy (BIS) measure of intracellular
volume (ICV). The equation used was MRI leg muscle (kg) ҃ 4.44 ѿ 1.21 ҂
ICV (L). The results of this equation give a value of SE ҃ 2.03 kg (R2 ҃ 0.63,
P 쏝 0.001). Bland-Altman analysis found no bias [x៮ 앐 SD ҃ Ҁ0.5 앐 1.9 kg
(R2 ҃ 0.088, P ҃ 0.073); data not shown]. The dependent variable was skeletal
muscle mass measured with MRI.
Bland-Altman analysis showed significant bias (R2 ҃ 0.448,
P 쏝 0.001) (Figure 8). As we observed with the leg, when
demographic and anthropometric variables were removed from
the relation, a model that correlated total arm muscle estimated
by BIS with MRI measurement of muscle mass was established,
as shown in Figure 9 and the following equation:
MRI arm muscle mass (kg) ⫽ ⫺ 0.774 ⫹ 1.1
⫻ ICV(L) (9)
where the relation was SE ҃ 0.63 kg (R2 ҃ 0.69, P 쏝 0.001), but,
in this instance, the relation was without significant bias (R2 ҃
0.0636, P ҃ 0.128).
DISCUSSION
Skeletal muscle mass represents a large fraction of ICV. Thus,
it is possible to construct models using isotopic or bioelectrical
FIGURE 7. Correlation between arm muscle mass measured by magnetic
resonance imaging (MRI) and a regression model developed by using wrist-
to-shoulder multifrequency bioimpedance spectroscopy (BIS) measurement
of intracellular volume (ICV). The equation used was MRI arm muscle mass
(kg) ҃ 1.69 ѿ 0.301 ҂ ICV (L) ѿ 0.603 (male) ѿ 0.0234 ҂ weight (kg) Ҁ
0.123 ҂ age (y). The results of this equation give a value of SE ҃ 0.306 kg
(R2 ҃ 0.933, P 쏝 0.001). The dependent variable was arm muscle mass
measured with MRI.
993
FIGURE 8. Bland-Altman plot of the difference between magnetic res-
onance imaging (MRI) and bioimpedance spectroscopy (BIS) estimates of
arm muscle mass versus the average value for arm muscle mass measured
with MRI and computed by using a BIS model estimate of arm muscle mass
Downloaded from ajcn.nutrition.org by guest on June 12, 2013
[x៮ 앐 SD ҃ Ҁ0.1 앐 0.59 kg (R2 ҃ 0.448, P 쏝 0.001)].
measures of ICV that model MRI measures of muscle mass.
Because of the nature of isotopic measurements, limb ICV can-
not be measured easily or at all. In contrast, BIS lends itself to
measurements of both whole-body fluid volumes and volumes of
defined anatomic compartments—in this case, limb ICV. This
study shows that, in a cohort of mostly African American hemo-
dialysis patients, multifrequency BIS can be used to estimate
TBMM in addition to muscle mass of both the arms and legs.
Because potassium is largely an intracellular cation, skeletal
muscle mass can be estimated by measuring TBK in healthy
adults (17). Dialysis patients may accumulate potassium, and
thus one may question the validity of TBK in predicting muscle
mass in these patients. Because 쏜95% TBK is intracellular at a
concentration 앒40 times that in the extracellular compartment,
the modest increase in extracellular potassium that might be
encountered in dialysis patients should have little effect on the
FIGURE 9. Correlation between total arm muscle mass measured by
magnetic resonance imaging (MRI) and a regression model developed by
using only the bioimpedance spectroscopy (BIS) measurement of intracel-
lular volume (ICV). The equation used was MRI arm muscle mass (kg) ҃
Ҁ0.774 ѿ 1.1 ҂ ICV (L). The results of this equation give a value of SE ҃
0.63 (R2 ҃ 0.69, P 쏝 0.001). Bland-Altman analysis found no bias [x៮ 앐 SD:
0.0056 앐 0.49 kg (R2 ҃ 0.0636, P ҃ 0.128); data not shown]. The dependent
variable was muscle mass measured with MRI.
994 KAYSEN ET AL
use of this measure (18). A second potential problem that might
confound results in patients with chronic kidney disease is that of
alteration of intracellular electrolyte composition. Cotton et al
(19) reported that patients with creatinine clearances 쏝 6.3 mL/
min had low resting transmembrane potential differences and
low intracellular potassium. However, hemodialysis 3 times/wk
was sufficient to restore intracellular potassium to a concentra-
tion that did not differ significantly from that in normal subjects
(155 앐 4.9 and 155 앐 4.3 mEq/L, respectively).
Empirical models that rely on measures of TBW have been
developed to describe total protein mass (20). Hemodialysis pa-
tients, however, accrue extracellular fluid between dialysis treat-
ments. Although we were able to construct a model of precision
whose results did not differ significantly between the use of
isotope dilution and BIS measures of TBW, it is possible that the
variability of ECV in dialysis patients may make such models
vary as functions of the time since the previous dialysis treat-
ment. Models of TBMM developed by using BIS measures of
ICV had regression coefficients that did not differ significantly
from those obtained by using TBK or isotope dilution estimates
of ECV. Similarly, measurement of ICV in either the arms or legs
provided models that correlated well with measured limb muscle
mass.
We have developed regression models by using multifre-
quency BIS with regression coefficients similar to those obtained
by using isotopic techniques when they were compared with
independently measured values for muscle mass. In addition, we
have established that segmental measurements of limb muscle
mass provided by multifrequency BIS estimates of ICV correlate
closely with MRI measures of limb muscle mass. Bias in the
relations is introduced by including various demographic or an-
thropometric measures in the regression model, not by using the
BIS measures of volume. Estimates of limb muscle mass ob-
tained by using BIS volume measurements are without notable
bias. Whereas isotopic methods are not capable of analyzing
compartmental muscle masses, BIS is of sufficient quality to
provide this information. This information could allow for quan-
titative structural evaluation of rehabilitation efforts after injury
to associated joints. Furthermore, we have established that ICV
does not vary significantly with intradialytic periods, which
would be of clinical interest.
Whereas BMI and adiposity have been found to be associ-
ated with low mortality among dialysis patients, indexes as-
sociated with lean body mass, such as volume of distribution
of urea, independently predict survival (21, 22). Similarly,
TBK, a measure of ICV, positively predicts survival in
chronic disease (23–25).
Because visceral volume increases at a lower rate than does
muscle mass (26), muscle mass contributes a greater fraction of
tissue mass as body cell mass increases (27). This is of some
importance, because the resting energy expenditure of skeletal
muscle is significantly less than that of organ tissues (28), and,
thus, energy needs as a fraction of the body mass would be
expected to decline as body cell mass increases (27, 29). Indexes
associated with greater muscle mass likewise predict survival
among patients with end-stage renal disease (30).
All of the models we developed that use measures of ICW
alone implicitly assume that a fixed fraction of body cell mass is
composed of muscle within a group of persons with a similar
body cell mass. The addition of weight, sex, and age provides
more information; however, all measurements other than MRI
must be used with the understanding that muscle mass may differ
from that obtained by using the model in subjects whose body
morphometry differs widely from average values.
Measurement of lean body mass in dialysis patients is con-
founded by the expansion of extracellular fluid, a component of
lean body mass. More precise measures of ICV should provide a
better measure of somatic protein stores (body cell mass rather
than lean body mass, which includes ECV and bone) than does
measurement of lean body mass alone. Anthropometric measure-
ments are limited by operator variability and precision (31). BIA
has been found to be useful in monitoring the nutritional status of
dialysis patients (8). However, the use of BIA in dialysis patients
has mostly relied on measurement of phase angle, and those
studies have not been extended to segmental analysis (8, 32).
Nutritional status can be evaluated longitudinally by the mon-
itoring of serum albumin. In contrast to the measurement of
serum protein, the estimation of intracellular mass presents more
of a challenge. Measurement of TBK by using 40K or of total
body nitrogen by using neutron activation provides excellent
measures of body cell mass (33); however, both procedures are
time-consuming, expensive, and not widely available. Evalua-
Downloaded from ajcn.nutrition.org by guest on June 12, 2013
tion of TBW by isotope dilution of extracellular space using
bromide (34, 35), chloride (34), inulin (36), or sulfate (34) also is
not easily applicable clinically. MRI also provides a measure of
specific extracellular and intracellular compartments (28, 37).
Measurement of the quantity of potassium either in the total body
or regionally also defines the volume of intracellular water (7).
We have calibrated whole-body BIS with other measures of
body composition, specifically 40K, MRI muscle, and isotope
dilution techniques of intracellular water and extracellular water
that correlates sufficiently well to isotopic measures to be useful
in describing body composition. In contrast to isotopic and MRI
measurements, BIS measurements are noninvasive and inexpen-
sive, and the procedure is portable. The precision of BIS in
estimating ECV and ICV is sufficiently good (published CV:
앒3– 6%; 38, 39) to be useful in longitudinal analysis of individ-
ual patients by providing the capability of measuring muscle
mass regionally or segmentally in longitudinal studies, in reha-
bilitation after surgery or injury, and after training in the gym-
nasium.
GAK conceived of and designed the project, designed the protocol, did the
regression modeling, and participated in manuscript preparation. FZ super-
vised the acquisition of the bioimpedance spectroscopy (BIS) data; analyzed
the BIS data; developed the concept of regional BIS analysis, including the
mathematical model to measure intracellular volume, extracellular volume,
and total water in individual body segments; and participated in manuscript
preparation. SS recruited patients, and SS and CK performed the BIS mea-
surements. SBH analyzed the magnetic resonance imaging data. JW super-
vised and performed the total body potassium measurements as well as
measurements of both D2O and bromide volumes of distribution. MKK
supervised the acquisition of data for the last 20 patients studied and partic-
ipated in manuscript preparation. NWL participated in the study concept and
design and participated in manuscript preparation. None of the authors had
any personal or financial conflict of interest.
REFERENCES
1. Lowrie E, Lew N. 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 – 82.
2. Salahudeen AK, Dykes P, May W. Risk factors for higher mortality at the
highest levels of spKt/V in haemodialysis patients. Nephrol Dial Trans-
plant 2003;18:1339 – 44.
 TOTAL-BODY AND LIMB MUSCLE MASS MEASURED WITH BIS
3. Johansen KL, Young B, Kaysen GA, Chertow GM. Association of body
size with outcomes among patients beginning dialysis. Am J Clin Nutr
2004;80:324 –32.
4. Shen W, Wang Z, Tang H, et al. Volume estimates by imaging methods:
model comparisons with visible woman as the reference. Obes Res
2003;11:217–25.
5. Heymsfield SB, Gallagher D, Kotler DP, Wang Z, Allison DB, Heshka
S. Body-size dependence of resting energy expenditure can be attributed
to nonenergetic homogeneity of fat-free mass. Am J Physiol Endocrinol
Metab 2002;282:E132– 8.
6. Snyder WS, Cook MJ, Nasset ES, Karhausen LR, Howells GP, Tipton
IH. Report of the Task Group on Reference Man. Oxford, United King-
dom: Pergamon, 1975.
7. Wang Z, St-Onge MP, Lecumberri B, et al. Body cell mass: model
development and validation at the cellular level of body composition.
Am J Physiol Endocrinol Metab 2004:286:E123– 8.
8. Fein PA, Gundumalla G, Jorden A, Matza B, Chattopadhyay J, Avram
MM. Usefulness of bioelectrical impedance analysis in monitoring nu-
trition status and survival of peritoneal dialysis patients. Adv Perit Dial
2002;18:195–9.
9. Di Iorio B, Cillo N, Cirillo M, De Santo NG. Charlson Comorbidity
Index is a predictor of outcomes in incident hemodialysis patients and
correlates with phase angle and hospitalization. Int J Artif Organs 2004;
27:330 – 6.
10. Chertow GM, Lowrie EG, Wilmore DW, et al. Nutritional assessment
with bioelectrical impedance analysis in maintenance hemodialysis pa-
tients. J Am Soc Nephrol 1995;6:75– 81.
11. Janssen I, Heymsfield SB, Baumgartner RN, Ross R. Estimation of
skeletal muscle mass by bioelectrical impedance analysis. J Appl Physiol
2000;89:465–71.
12. Lee RC, Wang Z, Heo M, Ross R, Janssen I, Heymsfield SB. Total-body
skeletal muscle mass: development and cross-validation of anthropo-
metric prediction models. Am J Clin Nutr 2000;72:796 – 803.
13. Pierson RN Jr, Lin DHY, Phillips RA. Total body potassium in health:
effects of age, sex, height, and fat. Am J Physiol 1974;226:206 –12.
14. De Lorenzo A, Andreoli A, Matthie J, Withers P. Predicting body cell
mass with bioimpedance by using theoretical methods: a technological
review. J Appl Physiol 1997;82:1542–58.
15. Zhu F, Schneditz D, Wang E, Levin NW. Dynamics of segmental ex-
tracellular volumes during changes in body position by bioimpedance
analysis. J Appl Physiol 1998;85:497–504.
16. Schoeller DA. Hydrometry. In: Roche AF, Heymsfield SB, Lohman TG,
eds. Human body composition. Champaign, IL: Human Kinetics, 1996:
25– 44.
17. Wang Z, Zhu S, Wang J, Pierson RN Jr, Heymsfield SB. Whole-body
skeletal muscle mass: development and validation of total-body potas-
sium prediction models. Am J Clin Nutr 2003;77:76 – 82.
18. Blumberg A, Roser HW, Zehnder C, Muller-Brand J. Plasma potassium
in patients with terminal renal failure during and after haemodialysis;
relationship with dialytic potassium removal and total body potassium.
Nephrol Dial Transplant 1997;12:1629 –34.
19. Cotton JR, Woodard T, Carter NW, Knochel JP. Resting skeletal muscle
membrane potential as an index of uremic toxicity. A proposed new
method to assess adequacy of hemodialysis. J Clin Invest 1979;63:
501– 6.
20. Beddoe AH, Streat SJ, Hill GL, Knight GS. New approaches to clinical
assessment of lean body tissues. In: Roche AF, ed. Body composition
assessments in youth and adults. Columbus, OH: Ross Laboratories,
1985:65–72.
21. Wolfe RA, Ashby VB, Daugirdas JT, Agodoa LY, Jones CA, Port FK.
Body size, dose of hemodialysis, and mortality. Am J Kidney Dis 2000;
35:80 – 8.
View publication stats
995
22. Gotch FA, Levin NW, Port FK, Wolfe RA, Uehlinger DE. Clinical
outcome relative to the dose of dialysis is not what you think: the fallacy
of the mean. Am J Kidney Dis 1997;30:1–15.
23. Kotler DP, Tierney AR, Wang J, Pierson RN Jr. Magnitude of body-
cell-mass depletion and the timing of death from wasting in AIDS. Am J
Clin Nutr 1989;50:444 –7.
24. Kamat AM, Shock RP, Naya Y, Rosser CJ, Slaton JW, Pisters LL.
Prognostic value of body mass index in patients undergoing nephrec-
tomy for localized renal tumors. Urology 2004;63:46 –50.
25. Le Blanc K, Ringden O, Remberger M. A low body mass index is
correlated with poor survival after allogeneic stem cell transplantation.
Haematologica 2003;88:1044 –52.
26. Gallagher D, Belmonte D, Deurenberg P, et. al. Organ-tissue mass mea-
surement allows modeling of REE and metabolically active tissue mass.
Am J Physiol 1998;275:E249 –58.
27. Muller MJ, Bosy-Westphal A, Kutzner D, Heller M. Metabolically ac-
tive components of fat-free mass and resting energy expenditure in
humans: recent lessons from imaging technologies. Obes Rev 2002;3:
113–22.
28. Bosy-Westphal A, Reinecke U, Schlorke T, et al. Effect of organ and
tissue masses on resting energy expenditure in underweight, normal
weight and obese adults. Int J Obes Relat Metab Disord 2004;28:72–9.
29. Illner K, Brinkmann G, Heller M, Bosy-Westphal A, Muller MJ. Met-
abolically active components of fat free mass and resting energy expen-
diture in nonobese adults. Am J Physiol Endocrinol Metab 2000;278:
Downloaded from ajcn.nutrition.org by guest on June 12, 2013
E308 –15.
30. Beddhu S, Pappas LM, Ramkumar N, Samore M. Effects of body size
and body composition on survival in hemodialysis patients. J Am Soc
Nephrol 2003;14:2366 –72.
31. Pollock ML, Jackson AS. Research progress in validation of clinical
methods of assessing body composition. Med Sci Sports Exerc 1984;
16:606 –15.
32. Johansen KL, Kaysen GA, Young BS, Hung AM, da Silva M, Chertow
GM. Longitudinal study of nutritional status, body composition, and
physical function in hemodialysis patients. Am J Clin Nutr 2003;77:
842– 6.
33. Wang Z, Shen W, Kotler DP, et al. Total body protein: a new cellular
level mass and distribution prediction model. Am J Clin Nutr 2003;78:
979 – 84.
34. Kim J, Wang Z, Gallagher D, Kotler DP, Ma K, Heymsfield SB. Extra-
cellular water: sodium bromide dilution estimates compared with other
markers in patients with acquired immunodeficiency syndrome. JPEN J
Parenter Enteral Nutr 1999;23:61– 6.
35. Janssen YJ, Deurenberg P, Roelfsema F. Using dilution techniques and
multifrequency bioelectrical impedance to assess both total body water
and extracellular water at baseline and during recombinant human
growth hormone (GH) treatment in GH-deficient adults. J Clin Endo-
crinol Metab 1997;82:3349 –55.
36. Krejcie TC, Henthorn TK, Gentry WB, et al. Modifications of blood
volume alter the disposition of markers of blood volume, extracellular
fluid, and total body water. J Pharmacol Exp Ther 1999;291:1308 –16.
37. Johnson JA, Albu JB, Engelson ES, et al. Increased systemic and adipose
tissue cytokines in patients with HIV-associated lipodystrophy. Am J
Physiol Endocrinol Metab 2004;286:E261–71.
38. Bedogni G, Malavolti M, Severi S, et al. Accuracy of an eight-point
tactile-electrode impedance method in the assessment of total body wa-
ter. Eur J Clin Nutr 2002;56:1143– 8.
39. Sun SS, Chumlea WC, Heymsfield SB, et al. Development of bioelec-
trical impedance analysis prediction equations for body composition
with the use of a multicomponent model for use in epidemiologic sur-
veys. Am J Clin Nutr 2003;77:331– 40.