ORIGINAL STUDY

Assessment of Bilateral Limb Lymphedema by

Bioelectrical Impedance Spectroscopy
Leigh Ward, PhD,*Þ Ann Winall, Dipl Physiol,þ Elizabeth Isenring, PhD,§ Andrew Hills, PhD,§
Sharon Czerniec, MHlthSci,Þ Elizabeth Dylke, MPhty,Þ and Sharon Kilbreath, PhDÞ

Objective: The aim of the present study was to determine if the ratio of extracellular fluid
(ECF), including the lymph, to that of intracellular fluid (ICF), as measured by bioim-
pedance spectroscopy (BIS), could be used to assess bilateral lymphedema (LE).
Background: The presence of LE is commonly determined as an increase in tissue volume
due to the presence of excess lymph relative to the volume of a comparable unaffected body
region or to comparative normative data. However, in bilateral LE of the limbs, a compa-
rable body region, the contralateral limb, is also affected, precluding normalization. An
alternative is to normalize the increase in lymph volume, as ECF, to that of ICF volume.
Methods: Extracellular/intracellular fluid ratios, expressed as the ratio of intracellular
impedance (Ri) to extracellular impedance (R0), for the limbs of 277 female and 224 male
controls were determined from an accumulated database of impedance data. Equivalent
data were obtained for an opportunistic cross-sectional sample of 37 female and 5 male
participants with bilateral LE of the legs. The ratios of Ri/R0 in the lymphedematous legs of
the affected participants were compared with the equivalent ratios in the unaffected arms of
the same participants and with those of the controls using box plots and visualized as
bivariate data using tolerance ellipses.
Results: Despite Ri/R0 ratios varying with age, sex, and limb dominance, comparison of
the ratio for affected legs (normalized to the ratio in the unaffected arms) with equivalent
ratios observed in a control population (as bivariate tolerance plots) was capable of dis-
criminating between 70% and 89% of the participants with LE.
Conclusions: Bioelectrical impedance spectroscopy and determination of Ri/R0 ratios as
indices of ECF/ICF ratios holds promise for the semiquantitative assessment of bilateral LE.
Key Words: Bilateral, Lymphedema, Bioelectrical impedance spectroscopy, BIS

Received August 16, 2010, and in revised form November 19, 2010.
Accepted for publication November 24, 2010.
(Int J Gynecol Cancer 2011;21: 409Y418)

*School of Chemistry and Molecular Biosciences, University of
Queensland, St. Lucia, Brisbane, Australia; †Breast Cancer Re-
search Group of the Faculty of Health Sciences, University of
Sydney, Sydney, Australia; ‡Lymphoedema and Laser Therapy,
Glenside, Adelaide, Australia; §Institute of Health and Biomedical
Innovation, Queensland University of Technology, Brisbane,
Australia.
Copyright * 2011 by IGCS and ESGO
ISSN: 1048-891X
DOI: 10.1097/IGC.0b013e31820866e1
Address correspondence and reprint requests to Leigh Ward, PhD,
School of Chemistry and Molecular Biosciences, University
of Queensland, St Lucia, Brisbane, QLD 4072, Australia.
E-mail: l.ward@uq.edu.au.
The present address of E Isenring: School of Human Movement
Studies, The University of Queensland, St Lucia, Brisbane,
Australia.
This study was funded, in part, by research grants from the National
Health and Medical Research Council of Australia, National
Breast Cancer Foundation, and ImpediMed Ltd.

International Journal of Gynecological Cancer & Volume 21, Number 2, February 2011 409

Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
Ward et al International Journal of Gynecological Cancer
Lymphedema (LE) will be familiar to many surgeons as a
chronic debilitating condition that may occur as a conse-
quence of axillary and inguinal surgery and/or radiotherapy
to lymph nodes. The precise incidence and prevalence is
uncertain, but it is reported to occur in between 9% and 55%
of patients receiving inguinal or ilioinguinal dissection.1Y3
This uncertainty is due largely to the lack of a universally
accepted definition for the presence of LE.2,4 Although a
staging classification for LE based primarily on clinical fea-
tures has been developed to describe the progression of the
condition,5 early detection and severity of LE are commonly
based on changes in the volume or circumference of an af-
fected limb relative to that of the contralateral control.2,5Y8
Such measurements are, however, not applicable in cases of
bilateral LE, common in postgynecological LE, where there
is no unaffected contralateral limb to act as a control.3,9
Attempts have been made to overcome this problem by es-
timating the change in volume by comparing, for example,
thigh-to-foot ratios of circumferences within the same limb.10
Clearly, such an approach will again not work where the
whole limb is affected.
Bioelectrical impedance spectroscopy (BIS) has been
used for the assessment of arm LE since the early 1990s.11
Bioelectrical impedance spectroscopy is based on the prin-
ciple that the magnitude of the opposition (impedance) to the
flow of a low-frequency alternating electric current through
the limb is inversely proportional to the volume of fluid in the
tissue. Thus, the impedance at a low frequency is a measure
of the ECF volume and, at least in the early stages of LE,
where excess lymphatic fluid accumulates, tissue impedance
decreases proportionately.12 (Strictly, the measured imped-
ance is that of the extracellular water, but because the total
extracellular fluid is approximately 95% water, the terms fluid
and water are often used interchangeably.) Because there is
wide biological variation between individuals in impedance
values, the impedance per se of a limb cannot discriminate
between an affected limb, that is, one with excess lymph,
from a normal limb.11 The impedance of an affected limb is
therefore normalized to that of the contralateral unaffected
limb and expressed as a ratio; the ratio is then compared with
normative values. The method has been shown to detect ac-
curately early-stage upper limb LE and monitor treatment
efficacy (see Ward et al12 for review) and to be suitable for
early detection of lower limb LE.13
In bilateral LE, as with volumetric measurements,
normalization to the contralateral limb is not possible. In
1996, Ward et al12 showed that the intracellular fluid volume
is largely unaffected in the early stages of LE development
and subsequently demonstrated that the ratio of extracellular
(ECF) to intracellular fluid (ICF) volumes (as the inverse ratio
of the respective resistances, Ri/R0, where Ri and R0 are the
resistances of ICF and ECF, respectively) was a sensitive
diagnostic criterion for unilateral LE posttreatment for breast
cancer.14 These authors postulated that this method could be
applied to the assessment of bilateral LE, although substan-
tiation of the technique in such cases was necessary.
The purpose of the present study was to determine the
normal variation of ECF/ICF ratios (as Ri/R0) in the limbs of
a cross section of a healthy population and to determine
410
Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
& Volume 21, Number 2, February 2011
whether this ratio could be used to discriminate individuals
with clinically ascribed bilateral LE.
METHODS
Participants
The study was an opportunistic cross-sectional design
with patients with clinically diagnosed bilateral LE identified
from patients attending an LE treatment clinic and invited to
participate in the study. Criteria to be met for inclusion were
broad, with participants being excluded only if they were
younger than 18 years, pregnant, fitted with a pacemaker or
other electrical medical device (contraindicated for BIS
measurements), or had medical conditions other than LE or
were using medications that would disturb their fluid status,
for example, diuretics.
Data were obtained for control subjects with no history
of LE from a database of impedance data maintained at The
University of Queensland. The participants had been recruited
for a variety of other studies being undertaken by the authors
for whom comparable impedance measurements had been
obtained. These data were pooled to provide the data set used
for analysis in the present study. Pooling of data was consid-
ered acceptable inasmuch as inclusion and exclusion criteria
for all of the studies were similar and the impedance mea-
surement protocols were identical in all studies.
All participants were weighed using digital scales, to
the nearest 0.1 kg, and their height measured to the near-
est 0.1 cm using a wall or stand-mounted stadiometer. All
participants were asked to state their arm and leg dominance.
The characteristics of the participants are presented in
Table 1. The mean ages of the male and female controls were
similar at approximately 43 years of age. This was consid-
erably younger than the mean ages of the LE participants at
70 and 63 years for the men and women, respectively. Cor-
respondingly, the age distribution of the LE participants was
such that all the men were aged between 60 and 80 years,
whereas the women’s ages ranged from 25.5 to 84.1 years.
The male LE participants were shorter on average by 16 cm
but 4.9 kg heavier than the male controls; consequently, their
body mass indices (BMIs) were higher: 33.7 kg/m2 compared
to 26.3 kg/m2. Although the female LE participants were the
same height as the female controls, they were considerably
heavier: average of 94.2 kg compared to 69.0 kg, with cor-
respondingly higher BMI: 34.7 compared to 25.4 kg/m2. The
number of cases of primary LE in men was 4-fold greater
than secondary LE. The ratio was 3:1 for the female partic-
ipants. There were also 7 female participants with clinically
ascribed mixed lipedema-lymphedema. Ethical approvals
were obtained from the relevant medical research ethics
committee of the participating institutions, and all partici-
pants provided written informed consent.
Measurement of Limb Impedance
The impedance of the legs and arms of all participants
was measured according to methods described in detail
elsewhere.12 Briefly, whole limb impedance (wrist to axilla
and ankle to groin) was determined according to the principle
of equipotentials, as this protocol has been shown to have
* 2011 IGCS and ESGO
 TABLE 1. Participants’ characteristics

* 2011 IGCS and ESGO

Controls Lymphedema
Males Females Males Females
Number 224 277 5 37
Age, yrs 42.6 T 17.5 (18.4Y91.0) 43.6 T 15.3 (18.3Y86.0) 70.4 T 4.2 (64.0Y74.3) 63.3 T 14.6 (25.5Y84.1)
Numbers per Decade of Age, yrs
G20 6 (2.7%) 3 (1.1%) V V
International Journal of Gynecological Cancer

20Y29.9 51 (22.8%) 56 (20.2%) V 2 (5.4%)

30Y39.9 63 (28.1%) 63 (22.7%) V V
40Y49.9 40 (17.9%) 70 (25.3%) V 5 (13.5%)
50Y59.9 30 (13.4%) 49 (17.7%) V 6 (16.2%)
60Y69.9 12 (5.4%) 15 (5.4%) 2 (40%) 8 (21.6%)
70Y79.9 6 (2.7%) 12 (4.3%) 3 (60%) 11 (29.8%)
980 16 (7.1%) 9 (3.2%) V 5 (13.5%)
Height, cm 178.2 T 7.2 (155.0 Y197.6) 164.6 T 6.7 (147.0Y187.4) 161.6 T 2.8 (157.0Y164.7) 164.8 T 8.2 (151.0Y182.1)
Weight, kg 83.5 T 13.4 (54.0Y126.9) 69.0 T 14.7 (40.4Y119.0) 88.4 T 16.4 (68.0Y111.1) 94.2 T 20.7 (62.0Y141.0)
Body mass index (BMI), kg/m2 26.3 T 3.9 (18.3Y 40.5) 25.4 T 4.8 (16.1Y 45.9) 33.7 T 5.4 (27.6Y41.8) 34.7 T 7.5 (24.2Y50.8)
Numbers per BMI range, kg/m2
Underweight (G20) 8 (3.5%) 30 (10.8%) V V
Healthy (20Y24.9) 81 (36.2%) 116 (41.9%) 3 (8.1%)
& Volume 21, Number 2, February 2011

V
Overweight (25Y29.9) 94 (42.0%) 88 (31.8%) 1 (20%) 8 (24.3)
Obese (930) 41 (18.3%) 43 (15.5%) 4 (80.0%) 25 (67.6%)
Lymphedema stage (n) 0:4:1:0 3:25:7:0*
Primary: secondary lymphedema (n) 4:1 28:9
Lipedema present 0 7
*Number of subjects per LE grades 1 to 4; 2 subjects were ungraded.

Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
BIS and Bilateral Lymphedema

411
Ward et al International Journal of Gynecological Cancer & Volume 21, Number 2, February 2011

greater precision than a ‘‘paired electrode’’ approach.12 Par-
ticipants assumed a supine position on a nonconductive ex-
amination couch, arms by their side and slightly abducted
from the body. All jewelry, except rings on fingers, was re-
moved and the skin at the electrode sites cleaned with an
alcohol wipe before the application of silverYsilver chlo-
ride gel electrodes. The electrode locations were those rec-
ommended by the manufacturer, that is, current drive
electrodes on the dorsal surface of the third metacarpal of the
hand and the dorsal surface of the third metatarsal of the foot
and the voltage sensing electrodes on the dorsum of the wrists
midway between the styloid processes and the dorsal surface
of the ankle midway between the malleoli. All measurements
were performed on each limb with the exception of a cohort
of 73 men and 103 women in the control group for whom
only measurements on the dominant side were available.
Impedance data were obtained with 2 different imped-
ance instruments. Data for 224 male controls and 277 female
controls drawn from the database had been obtained using an
SFB3 frequency-scanning bioimpedance spectrometer (range,
5-1024 kHz; SEAC-ImpediMed Ltd, Brisbane, Australia). All
other data were obtained using an SFB7 bioimpedance spec-
trometer (ImpediMed Ltd), which scans a similar range of
frequencies (3-1000 kHz). Equivalence of these instruments
has been demonstrated previously (below and in Ward15).
assuming that the resistivities of ICF and ECF are relatively
constant (see ‘‘Discussion’’ section). The ECF/ICF ratios
were calculated for each limb.
Statistical Analysis
Participant data were summarized with parametric
descriptive statistics. Individuals were classified according to
group (LE/control), sex (male/female), age (G30, 30Y39.9,
40Y49.9, 50-59.9 and 960 years) and BMI (20, 20Y24.9,
25Y29.9, 930 kg/m2). The distributions of ECF/ICF ratios, as
Ri/R0, for all limbs were then determined for subgroups.
To determine whether there were significant differences
across and between subgroups, 2- and 3-way analyses of
variance (ANOVA) were performed using the general linear
model procedures in the statistics module of Sigmaplot Ver-
sion 11.0 (Systat Software Inc, Chicago, Ill). Two-way
ANOVAs were used to determine whether there were any
overall differences between age strata and between dominant
and nondominant limbs within each sex and limbs (arm and

Data Processing and Analysis

Data from both SFB3 and SFB7 were processed
according to Cole theory12 using the manufacturer’s software
(Bioimp v 4.8.0.0, ImpediMed Ltd) to provide values for R0
and RV. R0 is the resistance at zero frequency and is equal to
that of ECF, whereas RV is the resistance at infinite frequency
equal to that of total limb water. At zero and infinite fre-
quencies, impedance (Z) and resistance (R) are equal; there-
fore, for the purposes of this manuscript, these terms are used
interchangeably. Generally, the default program settings were
used for analyses; however, for some data sets, the analysis was
optimized by adjustment of the frequency ‘‘window’’ used for
fitting the data to the Cole model guided by minimizing the
standard error of the estimate (SEE) for the curve fitting pro-
cedure.16 The SEE for all analyses was less than 1%.
Small, less than 2.5%, but significant differences in R0
and RV values between the SFB3 and SFB7 instruments have
been determined when both are tested on the participants or
electronic test circuits, an observation confirmed in a small
subset of the present subject group (data not presented).
Therefore, to facilitate pooling of data, all SFB3 data were
converted to equivalent SFB7 data using the conversion
factors provided by Ward.15
The resistance of intracellular water (ICF), Ri,, was
calculated from the relationship:
FIGURE 1. Ratio of extracellular to intracellular fluid
ðR0  RV Þ (as Ri/R0) in the arms and legs of male controls and of
Ri ¼
ðR0 jRV Þ male participants with bilateral lower limb LE stratified
Inasmuch as resistance is inversely related to volume, by age. Data presented as mean (line), 25th and 75th
the ratio of ECF/ICF is given by percentiles (box), and 5th and 95th percentiles
(whiskers). Mean values shown for each data set.
ECW Ri Upper panel: nondominant arm to dominant arm; lower
¼ ;
ICW Ro panel: nondominant leg to dominant leg.

412 * 2011 IGCS and ESGO

Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
International Journal of Gynecological Cancer & Volume 21, Number 2, February 2011 BIS and Bilateral Lymphedema

leg). Three-way ANOVAs were used to determine whether
there were any overall differences between age, sex, and be-
tween dominant and nondominant limbs for arms and leg
separately. Post hoc pairwise comparisons were performed
using the Holm-Sidak test. The effect of BMI was similarly
explored in a 2-way ANOVA.
To determine whether the ratios for the affected limbs
could be discriminated from those of the control group,
the distribution of the ECF/ICF ratios of the legs and the
arms were analyzed according to the generalized method of
Hotteling.17,18 When the ECF/ICF ratios for the arms are
plotted on polar axes against the same ratios for the legs of the
control population, they follow a bivariate Gaussian distri-
bution and a confidence interval (CI) for the correlation
between the 2 variables determines an ellipsoid shape. Con-
fidence intervals of 50%, 70%, and 95% were calculated for
the control data and presented as tolerance ellipses18 against
which the ECF/ICF ratios for the LE participants were then
compared. Tolerance ellipses were prepared using both the
data for all control subjects and those aged older than
40 years because these more closely matched the age spec-
trum of the LE participants (Table 1).
RESULTS
Control Subjects
The ratio of ECF to ICF in all 4 limbs, stratified for
limb dominance and by age decade, for the male participants
is presented in Figure 1 and for the female participants in
Figure 2. The mean ECF/ICF ratios of the arms in men
varied significantly (P G 0.001) among the different age
strata after allowing for dominance. The ratios were gen-
erally similar up to the fifth decade and then increased
progressively into the ninth decade in men. Post hoc anal-
ysis among the groups indicated that this was significant for
the 80-year-olds versus all other groups and for the 70-year-
olds versus the 30- and 40-year-olds only. Despite there
being also an overall significant, but weaker, effect (P G
0.030) of age in the arms of the female controls, there were
no individual pairwise significant differences. As for the
men, there was no effect of arm dominance or a significant
age-dominance interaction.
The effect of age on ECF/ICF ratios in the legs was the
same and highly significant (P G 0.001) in both men and
women. Post hoc analysis between age strata for each sex
indicated that significant differences (P G 0.001) were present
between 80-year-old subjects and all other age strata. The
ECF/ICF ratios of the nondominant leg were significantly
larger in both sexes (P G 0.014 for men and P G 0.005 for
women). There was no significant interaction between age
and dominance.
Three-way ANOVA revealed that the ECF/ICF ratios
were significantly higher (P G 0.001) in both the arms and
the legs of men compared to those of the women. However,
there was no significant interaction among dominance, age,
and sex.

FIGURE 2. Ratio of ECF to ICF (as Ri/R0) in the arms and legs of female controls and of female participants with
bilateral lower limb LE stratified by age. Data presented as mean (line), 25th and 75th percentiles (box), and 5th
and 95th percentiles (whiskers). Mean values shown for each data set. Upper left panel: affected leg to nondominant
to dominant; upper right panel: control leg to nondominant to dominant; lower left panel: affected arm to
nondominant to dominant; lower right panel: control arm to nondominant to dominant.

* 2011 IGCS and ESGO 413

Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
Ward et al International Journal of Gynecological Cancer & Volume 21, Number 2, February 2011

FIGURE 3. Ratio of ECF to ICF (as Ri/R0) in the arms and legs of control subjects and of participants with bilateral
lower limb LE stratified by body mass index (BMI). Data presented as mean (line), 25th and 75th percentiles (box),
and 5th and 95th percentiles (whiskers). Upper left panel: male nondominant leg; upper right panel: female
nondominant leg; lower left panel: male nondominant arm; lower right panel: female nondominant arm.

FIGURE 4. Ratio of ECF to ICF (as Ri/R0) in the arms and legs of control subjects and of participants with bilateral
lower limb LE stratified by BMI. Data presented as mean (line), 25th and 75th percentiles (box), and 5th and
95th percentiles (whiskers). Upper left panel: male dominant leg; upper right panel: female dominant leg;
lower left panel: male dominant arm; lower right panel: female dominant arm.

414 * 2011 IGCS and ESGO

Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
International Journal of Gynecological Cancer & Volume 21, Number 2, February 2011 BIS and Bilateral Lymphedema

The ratios of ECF to ICF in the nondominant limbs, stratified
by BMI, are presented in Figure 3, and the ratios of ECF to
ICF in the dominant limbs are shown in Figure 4. Two-way
ANOVA showed a significant effect (P G 0.001) of BMI,
allowing for the effects of dominance, in the arms of men
only. There was no significant effect of BMI in the legs
irrespective of sex. However, as noted earlier, the ratios were
significantly larger in the nondominant leg compared with
the dominant leg for both men and women (P G 0.016 and
P G 0.001, respectively).
Participants With Lymphedema
The small sample size of the male LE participants (n =
5 only) precluded stratification in this group by age, and the
combined group data are presented in Figure 1. The ratios in the
arms of the LE participants were slightly but not significantly
less (6%) than those observed in the comparably aged control
subjects. The ratios in the legs of the LE participants were,
however, significantly greater (P G 0.007) than those of the
comparably aged controls: 5.1 and 5.3 compared with 2.9 and
3.1 for nondominant and dominant limbs, respectively (Fig. 1).
The mean ratio in the legs of LE females was signif-
icantly greater by 29% (P G 0.001) than that of comparable
control women at all ages and irrespective of limb domi-
nance. Post hoc analysis indicated that the effect of the
presence of LE was significant at (P G 0.001) for the age
strata 40 to 49.9, 60 to 69.9, and older than 70 years of age;
(P G 0.010) for the 50 to 59.9 age group but not significant
in those aged younger than 30 years. This last observation is
not surprising given the small number of participants in this
age stratum.
The ECF/ICF ratio in the arms of the LE females was
significantly lower (P G 0.001) by an average of 17%, com-
pared with similarly aged controls (Fig. 2), a finding similar
to that observed in the men with leg LE. However, post hoc
analysis indicated no difference among age strata. There were
no other significant differences or interactions among the
variables studied.
Bivariate Analyses
Bivariate tolerance ellipses for control males and fe-
males with superimposed data points for the LE participants

FIGURE 5. Bivariate distributions of ratio of ECF to ICF (as Ri/R0) in the arms and legs of female controls and of
participants with bilateral lower limb LE. Control data presented as tolerance ellipses: 50th (V), 75th (V), and
95th (V) percentiles. Upper left panel: nondominant arm, female participants aged older than 40 years; upper
right panel: nondominant arm, female participants aged 18 to 86 years; lower left panel: dominant arm, female
participants aged older than 40 years; lower right panel: dominant arm, female participants aged 18 to 86 years.
Numbers refer to participants’ ID as discussed in text.

* 2011 IGCS and ESGO 415

Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
Ward et al International Journal of Gynecological Cancer & Volume 21, Number 2, February 2011

FIGURE 6. Bivariate distributions of ratio of extracellular to intracellular water (as Ri/R0) in the arms and legs of
male controls and of participants with bilateral lower limb LE. Control data presented as tolerance ellipses: 50th (V),
75th (V), and 95th (V) percentiles. Upper left panel: nondominant arm, male participants aged older than 40 years;
upper right panel: nondominant arm, male participants aged 18 to 86 years; lower left panel: dominant arm,
male participants aged older than 40 years; lower right panel: dominant arm, male participants aged 18 to 86 years.

are presented in Figures 5 and 6, respectively. Tolerance ellipses
are presented for participants older than 40 years only (left-
hand panels) and for all participants (right-hand panels) in each
figure. Depending on limb dominance, all but 4 (dominant) or 6
(nondominant) female LE participants fell outside the 95th
percentile for all control subjects (Figs. 5B, D). If the more
conservative tolerance ellipses for similarly aged control sub-
jects are used, then 9 of the total 37 women fell within the 95th
percentile. These data suggest that, depending on limb domi-
nance and the chosen tolerance limits, between 89% and 70%
of women were identified as having ECF/ICF ratios outside
the reference range. For the male LE participants, except for
the ECF/ICF ratio in the nondominant leg of one subject
when compared to age-matched controls (Fig. 6A), all par-
ticipants exhibited ratios that fell outside the 95th percentile
ellipse.
DISCUSSION
The assessment of bilateral limb LE presents particular
challenges to the clinician. The commonly used approach of
measuring affected limb volume excess compared to the as-
sumed normal contralateral limb is not possible. In this study,
a novel approach of referencing the volume excess in a
lymphedematous limb to another body region was explored.
Using bivariate analysis, up to 89% of the participants with a
clinically ascribed bilateral LE were correctly classified.
The approach taken was to quantify the accumulated
lymph in LE as a change in the ratio of ECF to ICF in the
limb. The present study is, to the authors’ knowledge, the first
to apply this method to the assessment of bilateral LE in the
legs. We found that the ECF/ICF ratio was increased in the
legs of those with leg LE compared with that of similarly
aged controls but that the range of ratio values found in af-
fected limbs overlapped with that observed in the healthy
population, limiting diagnostic discrimination.
Consequently, individual variation was decreased by
indexing the ECF/ICF ratio of an affected leg to the equiva-
lent ratio of the arm. The relationship between these vari-
ables, ECF/ICF ratios in the arm and leg, of healthy control
subjects is depicted as tolerance ellipses. A tolerance ellipse
is a reference for a given percentile, which gives the proba-
bility that a given datum is located from the mean vector of
the ellipse drawn for the reference population, that is, the
probability for data from participants with LE that the loci for

416 * 2011 IGCS and ESGO

Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
International Journal of Gynecological Cancer & Volume 21, Number 2, February 2011
these data fall outside the tolerance ellipse. This procedure
proved effective in classifying correctly all the male partici-
pants and up to 89% of the female participants at the 95%
probability level of having LE.
Interestingly, despite the LE participants having clini-
cally ascribed bilateral leg LE, predominantly of grades 2 and
3, comparison of ECF/ICF ratios with the reference ranges
did not consistently detect this. Six LE women had both legs
correctly identified as abnormal when compared with age-
matched control data (Figs. 5A, C). However, participants 6,
11, and 25 presented with only the dominant leg as abnormal,
whereas participants 1, 2, and 20 presented with only the
nondominant leg as having an abnormal ratio despite all
being clinically described as having bilateral LE. The possi-
bility of an original misdiagnosis of bilateral LE cannot be
excluded.
Another interesting observation was that for those
participants whose ECF/ICF ratios fell outside the reference
ranges, the loci for the ratio for each leg were not identical.
For example, subject 34 exhibited a ratio of 6.9 for the
nondominant leg compared to 4.6 for the dominant leg de-
spite having similar ratios in both arms: 3.5 and 3.6, re-
spectively (Figs. 5A, C). These results suggest that bilateral
LE may not affect both limbs equally.
A number of caveats to the conclusions drawn from this
study should be acknowledged. The relationship between the
impedance of a given volume conductor is given by
L2
V ¼Q ; was the lack of effect of dominance on the ECF/ICF ratio in the
Z arms compared with a significant effect in the legs of the con-
where V is the volume, L is the conductor length, and Q is the
specific resistivity of the conductor.12 Hence, the measured
impedance of a conductorVin this study, either the ECF or
ICFVis determined by both the resistivities of the tissue
fluids and their volumes. Thus, comparison of the Ri/R0 ra-
tios between limbs in both the LE and control conditions to
reflect accurately the volume differences assumes that the
resistivities of the ICF and ECF compartments remain
constant. The correctness of this assumption is unknown.
Indeed, accurate knowledge of the resistivities of these fluids
is lacking.19 Resistivity is primarily determined by the
concentration of electrolytes in the tissue fluids and is
exquisitely sensitive to change in their concentration. Under
normal physiological conditions, electrolyte balance is
tightly homeostatically regulated such that resistivities
remain essentially constant. Whether this is the case in
LE is unknown. As LE progresses and protein or fatty in-
filtration of the lymph occurs, resistivity will undoubtedly
alter. It is, however, noteworthy that 7 of the women with
LE presented with mixed lipedema/lymphedema yet still had
ratios falling outside the ellipses, suggesting that the pres-
ence of lipid had not affected adversely the diagnostic power
of the method. Further research is required to determine the
magnitude of such changes.
There is an assumption that effects of LE are generally
confined to a specific body region. For example, secondary
LE after inguinal surgery is usually considered to be confined
to the legs only. The procedure for indexing the ECF/ICF
ratio of an affected limbVin this case, the legVto that of
* 2011 IGCS and ESGO
Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
BIS and Bilateral Lymphedema
another body region is predicated on this assumption. The
findings from this study challenge this assumption because
the ratio was found, on average, to be significantly lower in
the arms of those with LE compared with the control subjects.
Superimposition of participant’s data on reference tol-
erance ellipses provides a simple and effective visual repre-
sentation for LE assessment. It does not, however, provide
an exact quantitative measure of the magnitude of the ECF
volume increase. The distance of a datum locus from the
ellipse is an indication of this magnitude. It may be possible,
with more data, to provide better quantification by calculation
of Kendall’s T statistic as has been used in other contexts.20
Inasmuch as it could not be assumed that the ECF/ICF
ratio is constant across all healthy individuals, we attempted
to determine the impact of factors such as age, dominance,
and BMI, as a surrogate for adiposity, on the reference range
of values. An unexpected observation was that the ratio of
Ri to R0 in the legs of older persons was increased, indicative
of expansion of the ECF in these individuals. This is sug-
gestive of a degree of edema being present and may reflect
the increasing prevalence of vascular disorders in the elderly.
It is also noteworthy that many of these older participants
were essentially sedentary institutional residents in which
pooling of ECF in the legs may occur through lowered calf
muscle pump activity due to physical inactivity. With respect
to LE assessment, these observations indicate that age-related
reference ranges are required. Although less pronounced, a
BMI effect was also noted. A further unexpected observation
trol subjects. It is generally assumed that the effects of domi-
nance, and consequently on physiology, are more pronounced
in the upper limbs than in the legs. Presumably, although arm
dominance is typically associated with a volume difference be-
tween the arms of up to 3%,11 differences in composition are
proportionate such that ECF/ICF ratios are unchanged. Why
this seems not to be the case in the legs is unknown.
This proof of concept study has demonstrated that
bioimpedance spectroscopy shows great promise for the as-
sessment of bilateral LE. The data analytic method of tol-
erance ellipses provides an effective way to display these
data. Further work is, however, required. Definitive norma-
tive ranges across a broader spectrum of ages, to allow age-
matching of patients with controls, and body habitus in
which bilateral LE commonly occurs, need to be established.
A prospective study to determine the minimum change in
ratio capable of detecting LE, that is, the detection threshold,
should also be undertaken. The successful translation of BIS
for early detection of unilateral breast cancerYrelated LE from
a research tool to routine clinical practice,21 however, gives
hope that it may eventually be equally successful for the as-
sessment of bilateral LE.
ACKNOWLEDGMENTS
The willing participation of the volunteers is grate-
fully acknowledged. The assistance of Kendall Sharpe,
Jarrod Meerkin, Simon Van Rosendal, and Julia Dyer in col-
lecting control data is also gratefully acknowledged.
417
Ward et al International Journal of Gynecological Cancer
REFERENCES
1. Moffatt CJ, Franks PJ, Doherty DC, et al. Lymphoedema: an
underestimated health problem. QJM. 2003;96:731Y738.
2. Rockson SG, Rivera KK. Estimating the population burden
of lymphedema. Ann NY Acad Sci. 2008;1131:147Y154.
3. Spillane AJ, Saw RP, Tucker M, et al. Defining lower limb
lymphedema after inguinal or ilio-inguinal dissection in
patients with melanoma using classification and regression
tree analysis. Ann Surg. 2008;248:286Y293.
4. Williams AF, Franks PJ, Moffatt CJ. Lymphoedema: estimating
the size of the problem. Palliat Med. 2005;19:300Y313.
5. Lawenda BD, Mondry TE, Johnstone PAS. Lymphedema:
a primer on the identification and management of a chronic
condition in oncologic treatment. CA Cancer J Clin. 2009;59:
8Y24.
6. Purushotham AD, Bennett Britton TM, Klevesath MB, et al.
Lymph node status and breast cancerYrelated lymphedema.
Ann Surg. 2007;246:42Y45.
7. Sander AP, Hajer NM, Hemenway K, et al. Upper-extremity
volume measurements in women with lymphedema: a
comparison of measurements obtained via water displacement
with geometrically determined volume. Phys Ther. 2002;82:
1201Y1212.
8. Stanton AW, Northfield JW, Holroyd B, et al. Validation of
an optoelectronic limb volumeter (Perometer). Lymphology.
1997;30:77Y97.
9. Brown J. A clinically useful method for evaluating
lymphoedema. Clin J Oncol Nurs. 2004;8:35Y38.
10. Hassall A, Graveline C, Hilliard P. A retrospective study of
the effects of the Lymphapress pump on lymphedema
in a pediatric population. Lymphology. 2001;34:156Y165.
11. Ward LC, Bunce I, Cornish BH, et al. Multifrequency
bioelectrical impedance augments the diagnosis and
418
Copyright © 2011 by IGCS and ESGO. Unauthorized reproduction of this article is prohibited.
& Volume 21, Number 2, February 2011
management of lymphedema in post-mastectomy patients.
Eur J Clin Invest. 1992;22:751Y754.
12. Ward LC, Kilbreath SL, Cornish BH. Bioelectrical impedance
analysis for early detection of lymphoedema. In: Weissleder H,
Schuchhardt C, eds. Lymphedema Diagnosis and Therapy.
4th ed. Essen, Germany: Viavital Verlag Gmbh Publ.;
2008:502Y517.
13. Halska MJ, Novackova M, Mala I, et al. A prospective study
of postoperative lymphedema after surgery for cervical cancer.
Int J Gynecol Cancer. 2010;20:900Y904.
14. Cornish BH, Thomas BJ, Ward LC, et al. A new technique
for the quantification of peripheral edema with application in
both unilateral and bilateral cases. Angiology. 2002;53:41Y47.
15. Ward LC. Inter-instrument comparison of bioimpedance
spectroscopic analysers. The Open Medical Device Journal.
2009;1:3Y10.
16. Ward LC, Byrne N, Rutter K, et al. Reliability of multiple
frequency bioelectrical impedance analysis: an inter-machine
comparison. Am J Hum Biol. 1997;9:63Y72.
17. Batschelet E. Circular statistics in biology. London, UK:
Academic Press; 1981.
18. Chew V. Confidence, prediction, and tolerance regions for
the multivariate normal distribution. J Am Stat Assoc.
1966;61:605Y617.
19. Ward LC, Czerneic S, Kilbreath SL. Quantitative bioimpedance
spectroscopy for the assessment of lymphedema. Breast
Cancer Res Treat. 2009;117:541Y547.
20. Smithies O, Engels WR, Devereux JR, et al. Base substitutions,
length differences and DNA strand asymmetries in the human
Gg and Ag fetal globin gene region. Cell. 1981;26:345Y353.
21. Schonholz SM. Preoperative assessment enables the early
detection and successful treatment of lymphedema. Cancer.
2009;115:909.
* 2011 IGCS and ESGO