..‘|t1.‘. .v.
u»: .

1.3-:
.z w

.I , .
. :Vﬁnzy

 

We

18:4:
rm? ..
mg .
3f...» 3
a“

In?

F-..
43' n V 55.
.11:
p

 

 

 

13
33:...
Ill

. it.

I .711

\- Jug-aux

 

3:, L l t
.5 .76... WuH‘.’

 

 

.. ...
.1. 5... . 1.. 31.3.13 ; 744?:

 

 

F'l

THESIS

IlllllilllllllllllIllllllll'llllillllllll

31293 017141387

This is to certify that the

thesis entitled

Eodj Composmon 1n chlc A+hit+cs

presented by

LUNA Cord/7L FOFnc Ht.

has been accepted towards fulfillment
of the requirements for

Ms K/n/Es/o LOG)’

degree in

Q”. in BM
U

 

 

Major professor

Date é/Iz 7/ 7X

0-7539 MS U is an Afﬁrmative Action/Equal Opportunity Institution

 

 

 

 

 

 

 

 

 

 

 

 

LIBRARY
MIchIgan State
Unlverslty

 

 

PLACE IN RETU
to remove this checkout
TO AVOID FINES return on

RN BOX
from your record.
or before date due.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

THE DETERMINATION OF BODY COMPOSITION IN FEMALE ATHLETES

by

Willa Corrine Fornetti

A THESIS

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Kinesiology

1998

ABSTRACT
BODY COMPOSITION IN FEMALE ATHLETES
By
Willa Fornetti
The purpose of this investigation was to determine the reliability and
validity of bioelectrical impedance (BIA) and near infrared interactance (NIR)
for estimating body composition in female athletes. Dual energy x-ray
absorptiometry (DXA) was used as the criterion measure for fat free mass
(FFM). Studies were performed on 132 college athletes (age =20.4il.5 yr).
Reliability estimates (repeat and single trial) were 0987-0997 for BIA
(resistance and reactance) and 0957-0980 for NIR (optical densities). Validity
of BIA and NIR was assessed by a double cross validation technique. Because
correlations were high (r=0969-0983), and prediction errors low, a single
equation was developed using all 132 subjects for both BIA and NIR. Also, an
equation was developed on all subjects using height and weight only. Results
from DXA analysis showed FFM =49.5:I:6.0 kg which corresponded to % body
fat ("/0 BF) of 20.4:t3.1°/o. BIA predicted FFM at 49.4159 kg (r=0981, SEE=1.1,
TE=1.1) and NIR prediction was 49.5:t5.8 kg (r=0975, SEE=1.2, TE=1.2). Height
and weight alone predicted FFM at 49.415] kg (r=0.961, SEE=1.6, TE=1.6).
When converted to. % BF values, prediction errors were ~1.8°/o for BIA and
NIR and 2.9% for height and weight. Results showed BIA and NIR to be
extremely reliable and valid techniques for estimating body composition in

college—aged, female athletes.

ACKNOWLEDGMENTS

Many individuals have attributed to the success of my thesis and graduate
education. First and foremost I’d like to extend an unending thanks to my
advisor, Dr. Jim Pivamik. Without him this thesis would not exist. I so
appreciate his guidance, advice, insight, and hard work. Also, my
appreciation to the other committee members, Dr. Jeanne Foley and Dr.
Justus Fiechtner for their critiques, input, and time. I would like to
additionally acknowledge Laurie Fruit and Cherrie Barth for performing the
DXA scans.

I would also like to thank my parents for their unconditional love and
support throughout my many years of college. My mom and dad have both
served as inspirational role models in my personal and professional life.
Also, a special thanks to Scott for listening and helping me remain sane.
Finally, a huge thanks to all my friends and fellow graduate students for their
advice, understanding and patience. I am truly blessed to be surrounded by

such wonderful people.

TABLE OF CONTENTS

List of Tables

 

List of Figures

 

List of Abbreviations

 

CHAPTER 1: INTRODUCTION

 

CHAPTER 2: LITERATURE REVIEW

 

Body Composition Literature Review Outline
Underwater Weighing

 

Single Photon Absorptiometry

 

Dual Photon Absorptiometry

 

 

Dual Energy X-ray Absorptiometry
Bioelectrical Impedance Analysis

 

Near Infrared Interactance

 

Body Composition Literature Review Summary
CHAPTER 3: METHODS

Subjects

 

Instrumentation

 

Data Collection Procedures

 

Statistical Analysis

 

CHAPTER 4: RESULTS

 

CHAPTER 5: DISCUSSION

 

Recommendations for Future Studies

 

References

 

iv

vi

vii

1-6

7-8
8-14
15-17
17-26
26-57
57-70
70-77
78-80

81
81-85
85-87
87-89
90-99
100-109
109

110-120

LIST OF TABLES

. Age and Anthropometric Subject Data (N: 132)

. Reliability Estimates for BIA and NIR Procedures (N: 132)

. Prediction Equations for FFM Using EVEN Numbered Subjects (n= 66)
. Prediction Equations for FFM Using ODD Numbered Subjects (n= 66)

. Prediction Equations for FFM Using ALL Subjects (N: 132)

LIST OF FIGURES

1a. FFM measured via DXA (kg) versus FFM predicted via BIA (kg) in female

athletes.

1b. FFM measured via DXA (kg) versus FFM predicted via NIR (kg) in female

athletes.

2. Scale weight (kg) versus DXA weight (kg) for female athletes.

vi

LIST OF ABBREVIATIONS

A - cross-sectional area

BIA - bioelectric impedance analysis
BMC - bone mineral content

BMD - bone mineral density

BMI - body mass index

Ca - calcium

Db - body density

DPA - dual photon absorptiometry
DXA - dual energy x-ray absorptiometry
FFM - fat-free mass

FM - fat mass

Ht - height

L - length

MRI - magnetic resonance imaging
NIR - near infrared interactance
OD 1 - optical density 1

OD 2 - optical density 2

R - resistance

RA - ratio of attenuation

Rst - ratio of soft tissue

SD - standard deviation

SEE - standard error of the estimate
SEM - standard error of measurement
SPA - single photon absorptiometry
TBBM - total body bone mineral
TBCa - total body calcium

TBK - total body potassium

TBW - total body water

TE - total error

UWW - underwater weighing

V - volume

Xc - reactance

Z - impedance

% BF - percent body fat

°/o CV — percent coefficient of variation
125 I - Iodine 125

153 Gd - Gadolinium 153

vii

CHAPTER 1

INTRODUCTION

Body composition analysis involves dividing the body into various
components and estimating their volumes and percentage of total mass
contained in each. Components of the body vary by classification scheme;
examples are fat mass (FM), fat-free mass (FFM), bone or soft tissue.
Information provided from body composition analysis is important for the
female athlete, as it can be indicative of health and performance status.

Excess body fat may negatively inﬂuence sport performance by
adversely affecting metabolic, mechanical, and thermoregulatory efficiency
(Houtkooper & Going, 1994). Generally, high ratios of FFM to FM are
favorable for an athlete, but too little body fat may result in the deterioration
of both health and performance (Houtkooper 8: Going, 1994). Optimal body
composition may vary among individuals in different sports (Heyward &
Stolarczyk, 1996). Body composition analysis typically includes an estimate of
an individual’s °/o body fat (°/o BF), which is calculated by dividing PM by total
body mass (i.e., weight). Some examples of reported values for college-aged
women include: 10—15% for female runners (Robinson et al., 1995), 15-17% for
female gymnasts (Kirchner, Lewis, 8: O’Connor, 1995), and 18-30% for female
non-athletes (Brozek, Kihlberg, Taylor, 8: Keys, 1963). An ongoing analysis of
body composition can provide useful information regarding the effectiveness

of sport specific training and related conditioning programs.

Knowledge of body composition is also important in helping medical
personnel in their constant surveillance of the athlete's physical and mental
health. Radical changes in body composition can be indicative of serious
health concerns. For example, the female athlete triad, consisting of
disordered eating, amenorrhea, and osteoporosis, is prevalent in many
athletes (Nattiv, Agostini, Drinkwater, 8: Yeager, 1991; Putukian, 1994; Smith,
1996). The triad has serious health implications, and may result in bone loss
or even death (Nattiv et al, 1991; Smith, 1996). Elite or competitive athletes
participating in sports in which low body weights or lean physiques are
considered advantageous are at a greater risk for developing the triad (Nattiv
et a1 1991; Putukian, 1994).

Menstrual irregularities and decreased bone mineral density are closely
related in female athletes (Fruth 8: Worrell, 1995; Lloyd et al., 1987). The
reduction of bone density, possibly due to menstrual irregularities, may put
athletes at risk of stress fractures and other injuries (Sinning 8: Little, 1987).
Therefore, accurate assessment of body composition may be a way to detect
the early signs of such health problems. For example, athletes could be
tracked individually and monitored for drastic changes in, or extremely low
°/o BF values.

Hydrodensitometry is the most widely used method for estimating %
BF (Wang, Heymsfield, Aulet, Thornton, 8: Pierson, 1989). This method has
been shown to have a measurement error of ~2—4°/o under the best conditions

(Sanborn 8: Jankowski, 1994). In hydrodensitometry, body volume is typically

estimated by weighing an individual underwater (UWW). Body volume
may also be obtained by estimating the weight of the water displaced, but this
technique is rarely used. Body density (Db) is then estimated by dividing body
mass (out of water) by body volume. Density is subsequently converted into
% BF, from which FM and FM can be calculated.

The original equation from which Db is converted into °/o BF was based
on chemical analysis performed on five Caucasian human cadavers (Siri,
1956). These cadavers included: a 42 year old female who committed suicide,
a 46 year-old male who died a week after a cerebral injury, a 35 year-old male
who died of an acute heart attack (decompensation or failure, apparently after
prolonged illness with mitral disease), a 25 year-old male who died of uremia
(blood urea 477 mg / 100 ml at death), and a 48 year-old male who died of
infective endocarditis with his body wasted and grossly edematous (Keys 8:
Brozek, 1953). Only the first two cadavers listed can be considered reasonably
”normal” since the latter were in various disease states (Keys 8: Brozek, 1953;
Brozek, Grande, Anderson, 8: Kemp, 1963). Based on cadaver analysis,
assumptions for the density of PM was 0.90 g/ cc, while FFM was 1.10 g/ cc.
Inherent in these assumptions is that FFM is at a constant hydration level of
73% (Pace 8: Rathburn, 1945).

As technology has improved, the original assumptions for UWW have
been challenged (Heyward 8: Stolarczyk, 1996; Lohman, 1992). Variations may
exist in the FEM, particularly bone mineral density. This variability may be

fairly prevalent in female athletes, depending on their training and

menstrual histories (Sinning 8: Little, 1987). Therefore, UWW may not assess
% BF accurately in female athletes. Further, UWW may not be an
appropriate criterion measure when developing body composition equations
for female athletes using various ”field” techniques (Lohman, 1984). New,
multi-component models are needed to account for interindividual
variability in FFM in order to obtain more accurate estimations of body
fatness (Heyward 8: Stolarczyk, 1996; Lohman, 1984; Withers et al., 1987).

A more valid and precise method (compared to UWW) for measuring
body composition may be dual energy x-ray absorptiometry (DXA). DXA
divides the body into three components: bone, fat free and bone free tissue,
and fat. Originally, this method was designed to analyze bone mineral
content (BMC) and bone mineral density (BMD). Precision for BMC was
high, as errors were shown to be less than 1% (Mazess, Collick, Trempe,
Barden, 8: Hanson, 1989). More recent studies have shown that DXA is
highly reproducible for both BMC and body composition (Mazess, Barden,
Bisek, 8: Hanson, 1990). Accuracy of DXA has included SEES of 29%, 1.9 kg,
and 2.7 kg for % BF, FM, and FFM, respectively (Svendsen, Haarbo, Hassager,
8: Christiansen, 1993). Studies have also shown that underwater weighing
and DXA agree well at high, moderate, and low levels of body fat (Hansen et
al., 1993).

The fact that DXA incorporates BMC when estimating % BF may be a
major advantage in comparison to UW. Bone comprises less than 5% of

FFM, but it has the highest density. Therefore, changes in BMC can have a

great effect on the average density of FFM (Wang et al., 1989). In young adult
females, variability in BMC has shown to contribute significantly to
variability in body density (Bunt et al., 1990).

Although DXA is extremely valid and precise for % BF measures, this
instrument is expensive, and is found primarily in clinical settings. Other
electronic methods, such as bioelectric impedance analysis (BIA) and near-
infrared interactance (NIR), have been developed to be used in the field. Both
machines are lightweight, portable, and require little technical training for
proper use. These two techniques are less costly than DXA and may provide
reasonably accurate results.

BIA uses an electrical current to estimate total body water by measuring
the resistance and reactance of the tissues. NIR measures the optical density
of remitted light from the biceps by indirectly measuring the tissue
composition of fat and water. These measures, along with height, weight,
age, etc. have been used to develop equations to predict FM, and % BF.

Despite the reasonable principles underlying BIA and NIR, neither
technique has been thoroughly tested for reliability and validity in female
athletes. Two equations for BIA are available for female athletes
(Houtkooper, Going, Westfall, 8: Lohman, 1989; Lukaski 8: Bolonchuk, 1987).
However, neither equation was based on a homogeneous athletic population
and HoutkOOper et al. (1989) was published only in abstract form. No

equations specifically designed for female athletes are available for NIR.

Moreover, no studies using female athletes have evaluated the validity of
either BIA or NIR using DXA as a criterion measure.

The purpose of this investigation is to determine the reliability and
validity of BIA and NIR for estimating body composition of female athletes.
The DXA technique will be used as the criterion measure. Stepwise multiple
regression will be employed to develop separate equations to predict FFM
using BIA and NIR. There will be no formal hypothesis presented for this
thesis as it is a measurement study and not hypothesis driven research. The
central questions include a) what is the degree of reliability shown by the BIA
and NIR instruments and, b) what is the degree of accuracy demonstrated by

the BIA and NIR when estimating FFM in female athletes?

CHAPTER 2

LITERATURE REVIEW

Body Composition Literature Review Outline

1. Underwater Weighing (UWW)
A. Basic Principles of UWW

B.

C.

Histological Concerns

1. Differences in fat-free mass

2. Variations of the Siri equation
Other problems with UWW

II. Single Photon Absorptiometry (SPA)
A. General Characteristics

B.
C.

Precision and Accuracy of SPA
Problems with SPA

III. Dual Photon Absorptiometry (DPA)
A. General Characteristics

B.
C.

Precision and Accuracy of DPA
Problems associated with DPA- source life variation

D. Studies comparing DPA with UWW- BMC differences
IV. Dual Energy X-ray Absorptiometry (DXA)
A. General Characteristics

.00

1. Ratio of Attenuation
2. Pixel Analysis - assumptions
Types of DXA Machines - Hologic QDR-1000 DXA
1. Formula calculations for Hologic
2. Calibrations for Hologic
a. internal calibrations
b. external calibrations
3. Assumptions for Hologic
Studies of Precision in DXA
Accuracy: In vivo analysis of DXA
1. Adult Applications
2. Pediatric Applications
Accuracy: In vitro analysis of DXA
Limitations of DXA
1. Scan table size
2. Beam hardening - studies
3. Assumptions
a. tissue surrounding bone
b. constant hydration
Studies Comparing Types of DXAs and SOftware
1. Different types of DXAs

2. Different software versions
3. Comparability of Hologic DXAs
H. Differences between DXA and DPA
1. Source Differences: x-ray vs. radionuclide
2. Precision of DXA vs. DPA
1. Studies Comparing DXA with UWW
V. Bioelectric Impedance Analysis (BIA)
A. Basic Principles of BIA
B. Advantages of Using BIA
C. Assumptions of BIA - problems with assumptions
D. Studies Examining Reliability and Validity of BIA
1. Studies assessing the sensitivity for BIA detection of body
composition changes.
2. Studies analyzing different variables
E. Sources of Error
F. BIA Prediction Models- studies with female athletes
VI. Near Infrared Interactance (NIR)
A. Basic Principles of NIR
B. Assumptions of NIR
C. Reliability and Validity of NIR
D. Sources of Error
E. NIR Prediction Models
VII. Body Composition Literature Review Summary

Underwater Weighing
B . B. 'l Elﬂlllll
The traditional "gold standard" or criterion measure for body
composition analysis has been hydrodensitometry, or more specifically,
UWW. This method indirectly estimates % BF from Db, which is the ratio of
body mass to body volume. UWW is based on Archimedes' principle that
body weight in water is equal to the weight of the water displaced during
submersion, and thus, allows for the calculation of body volume (Siri, 1956).
The formula for calculating Db via UWW is shown below.

Body density (Db) = body mass / body volume, where body volume, or

Db = Wa
Wa - WW - RLV (and intestinal gas estimate)
Dw

where Wa=weight in air, Ww=weight in water, Dw=density of water, and
RLV=residual lung volume.
Body volume is equivalent to the loss of weight in water, with the
appropriate temperature correction for the density of water (McArdle, Katch,
8: Katch, 1996). Measured density is a mean value of Db rather than the in
vivo density of tissue mass; therefore, air left in the respiratory passages and
lungs and gas in the gastrointestinal tract must be taken into account. Once
density is determined, the Siri equation (1961) is used to convert density into
percent body fat, from which FM and FFM can be calculated.

The Siri equation: % BF = (4.95 / Db) - 4.50.

PM = (Percent fat / 100) " Body mass

FFM = Body mass - Fat mass
Brozek, Grande, et a1. (1963) also formulated their own equation to calculate
% BF from Db.

% BF = (4.570 / Db) - 4.142

These two equations were based only on a few cadavers as previously
mentioned (Keys 8: Brozek, 1953). The day-to—day measurement errors for the
UW technique are small. However, inherent assumptions regarding the

density of FM and FFM for a given individual are unknown and probably

substantial, except in young white men (Roche, 1987).

H' l . l C l] mm
D' [E . ll EH 1 C I |

Recent advances in technology have led to the rejection of the
assumption that the density of FFM is constant in all individuals. FEM
consists of all tissues except fat, and includes bone, muscle, skin, and other
organs. Of particular concern is the variability of bone, which is included the
assumed FFM constant. Although bone comprises about 5% of lean body
mass, it has by far the highest density (Wang et a1. 1989).

The FFM assumption of UW, which had been formulated from the
analysis of five cadavers, has been further researched with data from the
Brussels Cadaver Study (Clarys, Martin, 8: Drinkwater, 1984). A sample of 25
cadavers, ages 55-94 years, consisted of 12 embalmed (6 male, 6 female) and 13
unembalrned (6 male, 7 female) Belgian subjects. Skeletal density varied
from 1.15-1.33 g/ cc, deviating from the assumed 1.10 g/ cc. Martin,
Drinkwater, Clarys, and Ross (1986) provided an example to demonstrate the
error in the assumption of the FEM constant. ”For a typical male of whole
body density 1.062 g / ml, predicted fat (by the Siri formula) would be 16.1%;
however if the FEM deviates by one standard deviation from the assumed
value of 1.100 g/ ml, fat prediction will range from 8.5% to 22.3%” (Martin et
al., 92, 1986).

Differences in BMC are also affected by other factors including age,

ethnicity, and athleticism. For example, Madsen, Adams, and VanLoan (1998)

10

studied the effects of physical activity, body weight/ composition, and
muscular strength on bone density. Sixty females, ages 18-26 yr, were divided
into three groups: 1) low body weight athletes involved in weight-bearing,
collegiate sports; 2) matched low body weight and sedentary; and 3) average
body weight and sedentary. Results indicated that the athletes had
significantly greater bone density at several sites. Weight-bearing exercise and
FM were shown to enhance BMC in eumenorrheic young adult women.

Variations in BMC are not accounted for in UWW. Indeed, studies
have shown that many individuals have variations in BMC that in turn will
produce incorrect results for % BF as measured by UWW (Bergsma-Kadijk,
Baumeister, 8: Deurenberg, 1996; Bunt et a1. 1990; Friedl, DeLuca, Marchitelli,
8: Vogel, 1992; Hansen et a1. 1993; Mazess, Peppler, 8: Gibbons, 1984; Wang et
a1. 1989). Based on estimates of overall variability in BMC and hydration of
the fat-free body there is theoretical error of 3-4% for predicting body fatness
in a population using UWW (Lukaski, 1987).

ll . . E l 5' . E .

To compensate for the shortcomings of UWW mentioned above,
several modifications of the Siri equation have been devised to help improve
the accuracy of estimating % BF from Db measures. For instance, Schutte et
a1. (1984) investigated the effect of bone mineral variation on the density of
lean body mass in Blacks versus Whites. They measured the density, total
body water, and anthropometric dimensions in 19 white and 15 black male

college students all matched for height, weight, and total body water. Among

11

the Blacks, the observed density was significantly greater than that predicted
from anthropometry, and the lean body mass calculated from observed
density was significantly greater than that calculated from total body water.
For the Whites, no significant differences were noted. The authors felt that a
separate formula should be used for converting density to body composition.
Based on their data, the adjusted formula for converting Db to % BE in Blacks
is: % BF = 100 * [(4.374/ Db - 3.928].

This formula assumes a FFM of 1.113 g/ cc in Blacks compared with 1.100 in
Whites.

Lohman, Boileau, and Slaughter (1984) investigated the applicability of
body composition constants for children. They showed that techniques used
to assess body composition in adults are inadequate in youngsters due to the
differences in chemical maturity between children and adults. Differences in
chemical maturity included variations in bone mineral, muscle, and level of
hydration which were all measured or estimated in the study. The result was
that Lohman et al.’s (1984) revised Siri equation for prepubescent children
became:

% BF = 100 " (5.30/ Db - 4.89)

Lohman performed a similar study to investigate the FEM assumption

for women (1992). The Lohman revised Siri equation for women became:
% BF = 100 * (5.05 / Db - 4.65)
In addition to a pure gender effect, females may also have variations in

BMC depending on their activity levels and menstrual status (Bunt et al.,

12

1990; Madsen et al., 1997; Sinning 8: Little, 1987). There are currently no
adaptations of the Siri equation for female athletes. Therefore, equations
utilizing UWW as a criterion have been applied to female athletes without
proof of their validity (Lohman, 1984).

W

In addition BMC variation, other factors can affect the accuracy of
UWW for estimating % BF. Hydration status, inability to measure residual
volume accurately, errors in intestinal gas content estimates, subjects’
cooperation and understanding of test procedures, and variations in
equipment and methodologies all contribute to errors in prediction of % BF
by UWW (Roche, 1987).

Bergsma-Kaddijk et al. (1996) investigated the hydration of young
women (aged 19-27 yr) who were weight-matched and compared with older
women (aged 65-78 yr). The authors found that total body water (TBW; as
measured via deuterium oxide dilution) from the younger group was
significantly (P<.001) greater (32.5 :t 2.9 kg) than in the older subjects (29.0 :t 2.5
kg).

The effect of experimental technique on measuring body density was
studied by Dumin and Satwanti (1982). Percent body fat for 6 women and 9
men, ages 17-51 yr, was estimated from UWW. Subjects were measured
under a variety of conditions including maximal expiration, moderate
expiration, minimal expiration, moderate inspiration, and after ingesting a

light meal (~2100 k] or ~500 kcal 1 1 /2 hours before), heavy meal (~5,000-9000

13

k] or ~1200-2200 kcal 1 1/ 2 hours before), and carbonated drink (800 m1).
Results indicated that the various combinations of food/ air in the lungs
resulted in ~1% difference in the estimated % BF content of the subjects.
Differences in % BF estimates attributed to carbonated beverages were
approximately 1.5%. The authors assumed that ingestion of carbonated .
beverages resulted in large increases in intestinal gases, but this was not
measured. The authors concluded these errors were well within the basic
errors of the method. While the measurement errors introduced in this
study were small, the authors noted that it is possible that extremes of all
three variables acting simultaneously could inﬂuence the % BF calculations
by 2-3%. The small sample size did not allow comparisons between male and
female subjects.

Friedl et al. (1992) studied the reliability and accuracy of body fat
estimates from a 4-component model devised from Db, TBW, and total body
bone mineral (TBBM) measurements. The data indicated that this four-
component model improved the accuracy of the % BF estimates. The largest
error in the multicomponent model appeared to come from the UW
method used to determine Db. The most important problem was the
assumed FFM hydration constant of 73.2%. Normal variation in the 10
subjects was 71-74%. This resulted in a decrease of 2.2 % BF and an increase of

3.4 % BF at the extremes of hydration.

l4

Single Photon Absorptiometry
GeneraLChatacteristics

Single photon absorptiometry (SPA), which was introduced in the early
19605, was developed to measure the bone mass of the appendicular skeleton.
The basic concept of this technique involves photon attenuation in vivo by
components of tissue. A radioactive source, such as iodine 125 (125 I), emits a
photon of characteristic energy. With SPA, this photon is passed transversely
across a bone and loses energy proportionally to the mineral mass of the scan
path (Mazess, 1971). After the photon passes across the bone, the photon
energy is measured with a scintillation detector-pulse height analyzer system
(Mazess, 1971).

E . . l E E SE 5

There are several sources of potential error that must be controlled so
that precision and accuracy are maintained. The largest source of error
involves the location of the correct scanning site. The magnitude of this
error is inversely proportional to the care taken in positioning the subject.
Hence, error may be lessened if the subject is carefully repositioned.

Mazess (1971) used SPA to examine the intercorrelations among bone
mass determinations at various sites on the human ulna, radius, humerus,
and femur. Four scan determinations were made at three or four locations
on various bones from the skeletal remains of 70 adult Sadlerrniut Eskimos
(30 male, 40 female). The relationships of local mineral content, determined

by SPA, were compared with weights of individual bones.

15

Scans on the different bones were highly intercorrelated (r=.80 to .90),
and correlation coefficients were even higher (r=.90 to .95) for different
locations on the same bone. Absorptometric scans were also highly correlated
with the weights (dry weight) of the bones on which they were made (r=.94).
The weights of individual bones were estimated with only a 5 to 10% error;
scans at several sites improved the accuracy.

In a follow-up investigation using similar methodology to the Mazess
(1971) study, Mazess, Judy, Wilson, and Cameron (1973) noted that variation
among seven different laboratories values were within 2% for both precision
and accuracy. These results indicated that errors associated with
absorptiometry were small, even among different laboratory sites.

BmhlemwithEEA

Problems may arise with SPA due to variations in the iodine source
(125 I). Dunn, Kan, and Wahner (1987) evaluated two commercial SPA
instruments. Several (n=42) measurements were taken over a 138 day period
using a metal bar for a standard. One SPA unit did not show any systematic
dependence of measured BMC or BMD on source activity when evaluated
over an entire source life. However, the other SPA showed measurements
which increased quadratically and gradually with time (r = .84; P< .0001) and
leveled off at about day 98. The maximum increase in BMC was ~ 2%. A
drastic change in results upon replacement of the source was apparent.

Specifically, the difference between the groups of five data points taken

16

immediately before and after source change was statistically significant
(P<.0001) in one of the SPA machines.

Another limitation of SPA is that it cannot be used for TBBM analysis.
Since the photon beam is monochromatic, only the mass of bone mineral can
be determined. Thus, SPA requires that the bone be enclosed in a constant
thickness of soft tissue (Peppler 8: Mazess, 1981). This constant thickness may
help eliminate errors due to varying mineral-nonmineral composition of the
bones (Mazess, 1971). However, this is not the case over the entire body, and
measurement errors may result.

Dual Photon Absorptiometry
Generalﬁharacteriatics

Dual photon absorptiometry (DPA), first developed in the early 19808,
utilizes a different radioactive source from SPA: isotope 153 gadolinium (153
Gd). This isotope, upon decay, has a characteristic half-life of 242 days and
emits two characteristic photon energies of 44 and 100 keV. Since DPA
consists of two photon energies, the need for constant tissue thickness across a
scan path have been eliminated and inaccessible body parts can be measured
(Peppler 8: Mazess, 1981; Watt, 1975). Hence, whole body analysis of bone
density became possible.

Having two distinct photon energies with DPA also allows the body to
be divided into two components: bone and soft tissue. The bone has a
constant value, but soft tissue varies as it consists of both fat and fat-free

tissue. These two tissues in turn produce different attenuation coefficients

17

for soft tissue at both energies (Peppler 8: Mazess, 1981). Therefore,
measurement of both TBBM and total body soft tissue analysis became
possible with DPA.

Emision and Accuracy of [212A

Although both bone and soft tissue analysis are possible, DPA is
inherently less precise than SPA. Errors arising from photon counting, taking
optimal photon energies into account, indicate that SPA is significantly better
than DPA for the measurement of TBBM (Watt, 1975).

Witt and Mazess (1978) studied the precision and accuracy of DPA using
three phantoms to simulate limbs. The phantoms consisted of
polymethylmethacrylate for soft tissue, and either an aluminum tube or
annular cavity filled with a saturated solution of dipostassium hydrogen
phosphate to simulate bone. Other phantoms were made of 11 polyethyleve
bottles, ~ 7.8 cm in diameter, filled with ethanol-water solutions of different
concentrations.

Precision was determined from repeat measurements of normal
subjects, patients, and phantoms. The upper arm and forearm of five adults
were scanned twice each week for 5 weeks; the precision of measured soft
tissue content was 2-3%. The precision of the fat fraction was about :I:.02 (0.5
to 0.7%), with SD of .02 to .04. Repositioning and subject motion were the
greatest sources of variation and limited the precision to 2-3% in normals and
3-7% in patients; instrumental variation was 0.5% for soft tissue content.

Also, DPA was found to accurately measure bone density within 1-2% for

18

phantoms and 2-4% in vivo (Witt 8: Mazess, 1978). The accuracy was 4% for
typical (i10%) ﬂuid changes observed in vivo.

Precision and accuracy of DPA were studied by Peppler and Mazess
(1981). Short-term (months) precision of TBBM was about 1.5% for isolated
skeletons and about 2% on normal human subjects. Long-term (years)
precision on skeletons was under 3%. The precision of % BF was 0.9%, which
would lead to an error of less than 1% in the TBBM. Geometry of
measurements also had minimal (and correctable) inﬂuence on the accuracy
of results. The accuracy (1 SEE) of TBBM on isolated skeletons (N = 5) was 36
g with a correlation coefficient of r=0.99. The accuracy of TBBM on the
skeletons was interpolated to be about 1-1.5% for normal adults, 2% in older
women, and 2.5% in osteoporotic females (Peppler 8: Mazess, 1981).

Gotfredsen, Jensen, Borg and Christiansen (1986) studied the in vivo
precision of the %BF and the FFM (kg) of five healthy subjects. Precision
estimates over six months were 2.5% and 2.2%, respectively. Other estimates
of the FFM were performed using both the calculations by Boddy, King, and
Hume (1972) and by skinfold thickness measurements (triceps and
subscapular). Results were then compared to DPA measurements in 50 men
and 50 women. The correlation between FFM by DPA and FFM by Boddy et
al. (1972) was significant (r=0.96, SEE = 4.4%). The authors concluded that
FFM measurements using DPA have precision and accuracy errors that are

commensurate with a reliable estimate of the gross body composition.

19

Heymsfield, Wang, Heshka, Kehayias, and Pierson (1989) compared
DPA measurements of bone mineral and soft tissue mass in vivo with the
established methods of Db, TBW, total body potassium (TBK), total body
calcium (TBCa), and nitrogen. There were 13 subjects, aged 24-94 yr. In
addition to the routine scanner calibration, the authors' approach to
evaluating soft tissue composition was to first scan seven different ethanol
and water mixtures. Second, a beef calibration was added for conversion of
the subject's ratio of soft tissue (Rst) value to % fat. Results for both the
water-ethanol and beef phantom calibrations were evaluated by five
consecutive runs. The Rst vs. percent ethanol and Rst vs. percent fat were
highly linear with minimal standard errors. The % coefficient of variation
(% CV) was calculated for each of ﬁve consecutive runs by determining the
respective Rst for 50% ethanol and fat mixtures. In both cases the %CV was
< 1%.

Other results indicated that DPA correlated significantly with all four
methods. TBBM was highly correlated with TBCa (r=.95, P<.001). Further,
the slope of TBBM vs. TBCa (0.34) regression line was similar to the that of
the TBBM vs. calcium content of ashed skeleton regression line (034-038).
DPA-measured fat (M:I:SD, 16.7149 kg) correlated significantly (r= .79-94;
P<.01) with fat established by Db (16.3:I:5.4 kg), TBW (16.0:I:4.3 kg), TBK
(17.7:I:4.6 kg, combined TBW-neutron activation (17.6159 kg), and means of
all four methods (16914.8 kg). Thus, the authors concluded that DPA offered

a new opportunity to study the human skeleton in vivo and to quantify fat by

20

a method independent from the classical assumption that bone represents a
fixed fraction of FFM.
WA

There are several potential problems/ limitations associated with DPA.
Some include: relatively high radiation dose (1/ 10 of a chest x-ray, Wang et
al., 1989), long scan time (~ 1 hr, Wang et al., 1989; 20-40 min., Gluer et al.,
1990), and difficulties in positioning which result in lowered precision.
Mazess et a1. (1990) have indicated that although the accuracy of total-body
soft-tissue composition analysis using DPA appears quite good, the precision
error has been relatively high. This paradox would argue against a solid
endorsement for its true validity.
S I 'E If . .

Another problem, which may significantly affect the precision of the
DPA measurement, is the variability of source life as 153 Gd decays. Dunn et
al. (1987) evaluated longitudinal measurements of bone mineral using four
DPA machines. The authors looked at precision over the lifetime of a source
and beyond. Repeated measures over more than one year were made on
ashed bones under 20 cm of water. All instruments were assessed for the
effectiveness of the manufacturer's recommended quality control program
for the detection of lack of stability in the instruments. Two DPA
instruments did not show any systematic dependence of measured BMC or

BMD on source activity when evaluated over an entire source life.

21

However, one DPA instrument showed strong dependence on source
strength. The 247 daily measurements of the three-bone region phantoms
showed strong linear and quadratic (P<.0001) patterns between calibration
results and time of use (source strength) for all three bone regions. Bone
mass and area density from the ashed bone also showed a strong linear
increase with decreasing source strength (P<.0001). The measured BMC
increased with a decrease in source activity. The fourth DPA instrument had
a significant linear decrease in BMD over a source life in the automatic mode
but performed better in the manual mode. Thus, the results indicate that
there may be some apparent trends with changing DPA source strength
during the useful life a radiation source.

Lindsay, Fey, and Haboubi (1987) also studied the effects of source life
on DPA measurements for BMD. They evaluated four separate specimens of
human vertebral bone consisting of vertebral bodies L2-4 still connected.
Samples were moved and repositioned between individual measurements
across two source changes. On each occasion, five measurements were made
of each specimen. During an 18 month period, BMD increased linearly an
average of 6% for the first source. The second source had an increase in BMC
of 4.7%, observed in the same time frame. At the source changes, BMD of all
specimens fell by 18-62% and 1.6—49% for the first and second sources
respectively. The vertebral measurements indicated apparent increases in
bone mass by as much as 0.6% per month as a function of source life. The

results indicated a decrease in the precision and, potentially, the accuracy.

22

Gotfredsen et al., (1986) examined the measurement of FFM and FM
using DPA. The reliability of estimating the lean mass was assessed in vitro,
using limb phantoms consisting of ox muscle, lard, and human bone; and in
vivo with duplicate measurements on five healthy subjects. The Rst value
fell with time, or with the falling activity of the source. Therefore, the
authors set up a weekly calibration system using ethanol and water mixtures.

The precision and accuracy in vitro by using limb phantoms of the lean
percent determination were analyzed to be 1.5% and 1.9%, respectively. The
accuracy in vivo of measuring the total mass of soft tissues was approximately
1.4%, thus yielding an overall accuracy error of the LBM of about 2.5%. The
precision in vivo of the lean percent and the LBM in kg of duplicate

measurements on five healthy subjects was 2.5% and 2.2%, respectively.

5 1° C . DE! 'lmlﬂil
S l’ B 1' Hit . Ell:

Bunt et al. (1990) measured the variations in BMC in 89 young adult
women, aged 25.1 i 5.3 yr. The women varied in activity levels and
menstrual status as compared to a sedentary control group. The following
BMC sites were measured: lumbar vertebrae (L2-4), radial shaft (RS), femoral
neck, and distal radius. These sites were measured and then evaluated for
their effect on body density. Theoretical differences in body density were
calculated to the :tl and :12 standard deviations of BMC for the population as

well as for several subgroups. These subgroups included: eumenorrheic

23

inactive controls, recreational runners, collegiate runners, body builders,
swimmers, and amenorrheic runners.

The differences in % BF for UWW (due to L2-4 and RS values for
BMC) ranged from an average overestimation of 1.3% for the amenorrheic
runners to an average underestimation of 1.4% for the body builders.
Individual differences were more profound, with a 3.3% fat underestimation
in body builders and a 3% fat overestimation in amenorrheic runners.
Therefore, variability in BMC attributed to variability in Db, independent of
fatness. While the impact of high or low BMC on % BF is modest for most
individuals, athletes with high or low BMC values may require adjustments
in equations used to convert Db to % BF.

Hansen et al. (1993) studied the body composition of 100
premenopausal, non-athletic females aged 28-39 yr. Bone mineral was found
to be a significant predictor in % BF models. Standard errors of estimate
decreased significantly (P <.001) from 0.0053 to 0.004 when bone mineral was
included to predict body density. The females had a lower bone density
compared to the assumptions inherent in the Siri equation (Siri, 1961). The
authors suggested that BMC should be taken into account when converting
Db to % BF.

Mazess et al. (1984) measured the FFM of 18 subjects, 14 female and 4
male, aged 23-60 yr, with both DPA and UWW. The correlation coefficient for
% BF measured from DPA compared to UWW was r=0.87. The calculated %

TBBM of FFM varied among subjects from 4.2 to 7.6%. Thus, TBBM was not

24

a constant fraction of FFM as is often assumed, nor was the percentage as low
as the assumed constant of 5%. The authors concluded that skeletal
variability, even in normal subjects, where mineral ranges only from 4 to 8%
of the FFM, may preclude use of body density as a composition indicator
unless skeletal mass is measured.

Wang et al. (1989) also emphasized the inconsistency of bone mineral
on the UW method for estimating percent body fat. Subjects included 99
males and 187 females, aged 19-94 yr. UWW assumes a constant FFM fraction
of 1.10 g/ cc, but results indicated an average (:tsd) value of 1.09:0.01 with a
range of 1.05 g/ cc to 1.13 g/ cc. Although bone mineral comprises a relatively
small component accounting for approximately 5% of the lean body, has by
far the highest density. Therefore, changes in bone mass or bone density are
of proportionately great inﬂuence on the average density of FM.

Wang et al. (1989) also evaluated measurement precision in five
normal subjects, aged 31-57 yr, by performing five measurements of UWW
and DPA within a three weeks period. The three measurements derived
from the DPA gave the higher precision, ranging from 0.97 to 0.99, with
UWW at 0.94 and 0.95.

Using DPA as a reference for BMC, the authors calculated that a 1%
change in the TBBM-FFM results in a change in density of 0.01 g/ cc. This
would result in a change of 5% in the % BF estimate when using the Siri
equation. Therefore, TBBM can significantly inﬂuence the measurement of

% BF by UWW. The higher density of the skeleton and the marked

25

intersubject variation in TBBM again make this the source of highest
variation.
Dual Energy X-ray Absorptiometry (DXA)
E l C] . .

DXA is based on the characteristic attenuation of x-rays. The dual
energy source of conventional x-rays undergoes varying degrees of
attenuation by human tissues such as bone and soft tissue. Attenuation is
defined as the reduction in intensity of the x-ray source after it interacts with a
chemical compound. Beam attenuation occurs as photons passing through
tissue are absorbed or scattered by two main types of physical interaction:
compton scattering and photoelectric effect (Pietrobelli, Formica, Wang, 8:
Heymsfield, 1996). The diminished beam intensity is directly related to the
specific chemical compound with which it interacts. The unattenuated
energy, in the form of x-ray radiation, is determined by an external detector.
BI' ill I' [BE] [122:5

When two different photons pass through an absorber, attenuation can
be expressed as a ratio (RA). The attenuation of the lower energy can be
expressed as a ratio to the attenuation observed at the higher energy. Each
element has a characteristic mass attenuation coefficient and RA value. For
example, elements found in soft tissues (e.g., Na, K, and Cl) have lower
atomic numbers, and smaller RA values compared to calcium (Ca), which is

primarily found in the bone (Pietrobelli et al., 1996).

26

Soft tissue and bone are identified by DXA-measured RA values, with
the two components being distinguished simultaneously. The mass fraction
of two components can be identified if each component's mass attenuation at
each of the two energies is known and constant (Pietrobelli et al., 1996). First,
the distinction is made between bone and non-bone; then non-bone is
subsequently divided into fat and bone-free, fat-free tissue. The theoretical
RA values calculated from elemental mass attenuation coefficients at
assumed peak energies (of 40 8: 70 keV) agree closely with RA measurements
in vitro. Theoretical and in vivo measured RA values are also strongly
associated (Pietrobelli et al., 1996).

W

As the x-ray beam is introduced into the body, the external detector
analyzes one small cross-sectional area at a time. These small cross-sectional
areas are called pixels. Pixels are first separated by computing the RA value
for every pixel in a total body DXA scan. Because the RA value for bone
mineral is much higher than for soft tissue, bone-containing pixels will be
those with the higher composite RA values. An RA value threshold must be
set to distinguish between pixels that include bone mineral and those that
consist of soft tissue alone.

The fat and lean composition of soft tissue pixels may be evaluated by
equations. The constants in the equations, RA] for pure fat and RA2 for pure
lean tissue, are established by measuring fat and lean soft tissue samples or

other materials that serve as calibration standards. Each RA for FM and FFM

27

is close to those based on an experimental phantom created from actual
animal tissues (Pietrobelli et al., 1996).

Estimating soft tissue composition in bone-containing pixels is
complex and requires some assumptions. The uniform distribution model,
applicable to bones found in regions of the lumbar spine, has been replaced by
improved algorithms such as the weighted-linear distribution model. This
weighted model takes the adjacent soft tissue composition into account when
calculating the soft tissue being analyzed. This model has improved
algorithms and is better at estimating soft tissue in bone pixels (Pietrobelli et
al., 1996).

WWW

All DXA machines share in common an x-ray source, detector, and
calibration techniques. The three types of commercial DXA machines
include: Hologic QDR, Lunar DPX, and Norland XR 26 (Jebb 8: Elia, 1993;
Lukaski, 1993). Several approaches can be used to create the two main energy
peaks. The first, Hologic, generates dual-energy x-ray spectra by pulsing the x-
ray tube kV between sequential measurement points. The Hologic QDR-1000
utilizes two energy beams, 70 and 140 keV, which are alternately pulsed
resulting in tube voltages of 45 and 110 keV (Kelly, Slovik, Schoenfeld, 8:
Neer, 1988). Lunar DPX uses a constant potential generator and a Cerium K-
edge filtration to generate photons at two energies (40 and 70 keV). The last,

Norland XR 26, also employs a constant generator, but it operates at 100 kVp

28

and employs a Samarium filter with k-edge at 46.8 keV. These machines also
have different detectors, calibration standards and techniques, and bone
thresholds for pixels and algorithms differ also (Tothill, Avenell, 8: Reid,
1994).

The principles are the same for each, but there are technical differences
in both the hardware and software such that measurements by one
instrument are not necessarily the same as those obtained by another (J ebb 8:
Elia, 1993). No specific manufacturer is preferred (Jebb 8: Elia 1993). However,
previous studies comparing the Norland DXA and other brands indicate that
Hologic and Lunar instruments typically show a closer agreement with
UWW and magnetic resonance imaging (Tothill, Hans, Avenell, McNeill, 8:
Reid, 1996).

E l C] ll' E H l . QUE-10001221!

The QDR-1000/ W's computer algorithm is based on the principle that
bone selectively attenuates high-energy x-ray photons, and that the bone
mineral content Q of any sample point can be computed from: Q = KH-L;
where H and L are the logarithms of the sample attenuation at high (140 kVp)
and low (70kVp) energies, respectively, and the constant K depends on the
tissue attenuation characteristics of the beam. In the QDR-1000 / W, K is
continuously measured using the "tissue" segment in the filter wheel
(Hologic Operator’s Manual, 1990).

Using this value of K, Q is calculated for each point scanned using the

formula given above. A histogram of the Q values is compiled. Because a

29

large portion of the image will contain soft tissue only, this histogram will
have a large peak. A threshold is chosen just above this peak, and applies
that value to discriminate, point by point in the Q image, between "bone"
points (whose Q is above threshold) and "non-bone" points (whose Q is
below threshold).

W The Hologic contains an internal reference
system consisting of a rotating wheel composed of three sections (2 sections of
epoxy-resin-based material consistent with the densities of bone and soft
tissue and one section of air). This wheel automatically determines which
points in the scan are bone, tissue, and air (air points are excluded from
analysis of soft tissue). The Hologic DXA is calibrated daily within :I:0.5%
when measuring BMC of a Hologic DPA/QDR-l anthropomorphic spine
phantom containing a known weight of calcium hydroxyapatite (CaHA)
(Hologic User’s Manual, 1990).

Thus, before the x-ray passes through the patient, the beam is filtered
through the rapidly rotating wheel. When finally intercepted by the detector,
the beam contains information about the x-ray absorbing characteristics of
both the patient and the calibration materials in the filter wheel. An A / D
converter, fed by the detector, supplies a complex digital signal to the
computer, which uses that signal both to construct the screen image and as

the basis for its computations of BMC and BMD.

3O

W External calibration for fat and lean tissue of
varying depths are performed by the tissue bar. Each scan is performed with
this tissue bar at the edge of the mattress, by the feet. Tissue points and the
soft tissue portions of the bone points are then compared to the soft tissue
calibration phantom, or the tissue bar, to determine the fat/ lean composition.
This tissue bar is included for all analyses. The manufacturers were originally
concerned with the longtime stability of the absolute rays. The tissue bar is
composed of wedges made of aluminum and lucite (polymethylmethacrylate)
calibrated against stearic acid as 100% fat, and dilute saline solution as 100%
fat-free mineral-free tissue (Hologic User’s Manual, 1990).

For correct reporting of the soft tissue values, the regions must be
adjusted to place the soft tissue in the correct regions. Regions include: head,
right arm, left arm, right leg, and left leg, and trunk.

E l' E I] H l . QUE-1000

DXA has several assumptions. The first is 17.0% brain fat, which has
been calculated by comparative studies (Hologic Operator’s Manual, 1990).
This brain fat assumption is made because the machine cannot measure the
fat due to the overlying bones of the skull and the lack of adjacent soft tissue
to estimate a value. The 17% brain fat assumption may vary with head size.
Another assumption includes the constant hydration of FFM at 73.2% water
(Pace 8: Rathburn, 1945). Finally, DXA accounts for bone marrow through the
values derived from complex algorithms in sofware (personal

communication, Hologic scientists).

31

f r ' i

Fuller, Laskey, and Elia (1992) evaluated the precision of DXA (Lunar)
for whole body and major body subregions. Subjects included 12 females and
16 males, aged 18—59 yr. Each subject was scanned twice. Following the first
scan, subjects stood and walked around before being repositioned for the
second scan. Components measured included: fat, fat-free soft-tissue, and
bone for the whole body and subregions. These components were analyzed
for precision by SDs and % CVs.

The precision (SD, % CV) for fat tissue was 0.11 kg (9.0%) for arms, 0.20
kg (3.4%) for legs, 0.25kg (4.4%) for trunk, and 0.42 kg (3.0%) for whole body.
Fat-free soft tissue had precision measures of: 0.15 kg (2.8%) for arms, 0.23 kg
(1.2%) for legs, 0.24 kg (1.0%) for the trunk, and 0.42 kg (0.8%) for whole body
analysis. Finally, bone mineral precision was 0.01 kg (2.0%) for arms, 0.02 kg
(1.4%) for legs, 0.02 (2.0%) for the trunk and 0.03 kg (0.9%) for whole body
scans. In general, the larger the size of a segment was found to be, the larger
was the associated absolute error, but the percentage error was smaller.

Going et al. (1993) studied the ability of DXA (Lunar, software 3.6) to
detect small changes in the body composition of 9 men and 8 women, aged 19-
31 yr, during a dehydration-rehydration protocol. This protocol designed to
induce changes of ~ 2% of body weight. Also during this time, the subjects
underwent repeat body composition measures over three consecutive days.

Results indicated that scale weight and total mass from DXA were

highly correlated (r> 0.99); as were estimates of FFM (r= 0.99) and % BF (r=

32

0.97) from DXA and UWW. Changes in scale weight of ~ 1.5 kg due to ﬂuid
loss and gain were highly correlated (r=.90) with both changes in total mass
and soft tissue mass from DXA. As expected, bone and FM estimates were
unaffected by changes in hydration.

In contrast to the findings for bone mass and FM, DXA had a somewhat
limited capacity to resolve small changes in FFM (r= 0.67) due to ﬂuctuations
in hydration for individual subjects. The changes in FFM detected by DXA
accounted for only ~50% of the changes in total mass during dehydration-
rehydration studies. However, mean changes in total mass, soft tissue mass,
and FFM were not significantly different (P> 0.05) from changes in body
weight. The average changes in body weight and fat free tissue weight agreed
well (509 kg). Thus, the authors concluded that DXA was able to detect small
individual changes in total mass and soft tissue mass and was also useful for
detecting group changes in FFM.

Haarbo, Gotfredsen, Hassager, and Christiansen (1991) studied the body
composition of six subjects measured six months apart. The precision errors
in vivo included the following (SD (CV%)): FM, 1.1 kg (6.4%), FM 1.4 kg
(3.1%), TBBM, 0.03 kg (1.2%), and % BF 1.6% (5.7%). None of the subjects
varied their body weights by more that 4.2 kg between measurements.

Mazess et al. (1990) measured BMC and soft-tissue composition with
DXA (Lunar) in 12 young adults (6 males and 6 females). Body composition
was analyzed on five occasions at both medium (20 minutes) and fast (10

minutes) scanning speeds. Ten scans were made over a period of 5 to 7 days

33

with an average of two scans per day. There were no significant differences in
mean results or in precision errors between the two speeds. The correlations
between determinations using the two speeds were very high ( mean r= .98, P
<.001) for both BMC and tissue composition.

Because there were no significant differences in mean values or
precision at the two scan speeds, the results were pooled. The precision error
obtained from pooling results from the two speeds for total BMC and BMD
averaged 50 g and 0.01 g/cm2, or ~ 1.8% and 0.8% respectively. Precision
errors attributed to: percent fat in soft tissue =1.4%, FM =1.0 kg, and FFM =0.8
kg. These results correspond to a relative error of 0.8% for total body BMD
and 1.5% for FFM. These preliminary results suggest that DXA can be used to
accurately estimate soft-tissue composition with better precision (1-1.5%) than
was possible with DPA.

S l' E! _1 If I . EDIE!
E l l E l' .

Haarbo et al. (1991) investigated the accuracy of DXA in vivo,
comparing the components of body composition measured by DXA to DPA,
TBK, and Db by UWW in 25 subjects (15 women and 10 men aged 23-41 yr).
The authors found agreement between % BF and FFM by DXA and the three
established modalities. Mean measurement differences were (-5.3 to -0.4%)
and (-0.7 to 2.5%) for % BF and FFM, respectively. Thus, the authors
concluded that DXA provides accuracy which is applicable to group research

studies and probably also in clinical measurements of the single subject.

34

Prior et al. (1997) examined the validity of a Hologic 1000W to evaluate
body composition against a four component model. This four component
model consisted of TBW, Db, and TBBM as described by Lohman (1986).
Subjects included 81 women (aged 20.7:t2.6 yr) and 91 men (aged 21.2:_I-2.1 yr),
of which 111 were collegiate athletes and 61 were nonathletes. The authors
also sought to determine whether accuracy is affected by gender, race, athletic
status, or musculoskeletal development in young adults.

Results indicated a high correlation between body mass via scale
weight and body mass by DXA; with r= 1.0, SEE: 0.5 kg. Also, a higher
correlation and lower errors were found between % BF by DXA and % BF
from the four component model (r= 0.94, SEE=2.8%, TB: 2.9%) compared to %
BF by UWW versus % BF from the four component model (r= 0.91; SEE:
3.4%, TB: 3.6%). Also, mean % BF values from UWW were significantly
greater than % BF from the four component model (mean diff iSD diff =
1.213.4% body mass). Differences between % BF from DXA and % BF from
the four component model were weakly related to body thickness (as reﬂected
by BMI, r: -0.34) and to the percentage of water in the FFM (r= -0.51), but were
not significantly affected by race, gender, athletic status, or musculoskeletal
development. Thus, the authors concluded that DXA body composition
estimates are accurate in young adults who vary in gender, race, athletic
status, body size, musculoskeletal development, and body fatness.

Svendsen et al. (1993) evaluated the accuracy of body composition

measurements in vivo by comparing DXA (Lunar, software 3.2) scans with in

35

vivo and chemical analysis after postmortem homogenization of seven pigs
(35-95 kg). The regression lines between these measurements were not
significantly different from the line of identity (P >0.05), the r values were
>097, and the corresponding SEES were 29%, 1.9 kg, and 2.7 kg for % BF, FM,
and FFM, respectively.

Svendsen et al. (1993) also evaluated changes in PM and FFM which
were simulated by placing 8.8 kg porcine lard on the trunk of six women. The
measurements of the simulated changes in PM and FFM by DXA were not
significantly different from those expected (P >0.05). However, both the
measured TBBM and the number of bone pixels increased by ~7% with the
lard addition. The authors could not explain the inaccuracies of the TBBM
measurements, but the fact that the lard increased the number of bone pixels
could indicate that the edge detection technique was inﬂuenced by the lard.
E 1° | . E l' l'

Brunton, Bayley, and Atkinson (1993) evaluated the DXA (Hologic
QDR-1000, software 6.0) for analysis of whole body composition in infants.
Two specific groups were targeted: small piglets (1.6 kg) and large piglets (6.0
kg). These weights were chosen to approximate the lowest midrange weight
of infants who would be of clinical interest. Precision and accuracy were
estimated by scanning piglets in triplicate to calculate CVs and comparing
DXA estimates with chemical analysis of whole carcass. Three consecutive
whole-body scans were conducted by using the pediatric whole-body software

on piglets. Generally, the piglets were not repositioned between scans unless

36

they became aroused during the scan, in which case they were given more
anesthetic and the scan was restarted.

Results included the mean CVs for all DXA measures in small piglets
and large piglets were <2.5%, except for FM, which were 6.3% and 3.5%,
respectively. The authors concluded that in large piglets, DXA provided
reasonable estimates of chemical analysis for BMC, FFM, FM; but only for
FM in small piglets. DXA overestimated fat with a 234.6% difference
between DXA and measured fat. Also BMC was underestimated by 29.7%
with DXA. The authors concluded that further improvement in accuracy is
necessary before the utilization of DXA estimates of whole-body composition
in small infants and animals weighing 1.6 kg.

Picaud, Rigo, Nyamugabo, Milet, and Senterre (1996) studied the
reproducibility, accuracy, and precision of DXA (Hologic QDR-2000, software
5.64) by scanning 13 piglets, 1471-5507 g. Three consecutive scans were
performed while the pigs were supine; there was no repositioning between
each scan. Estimates of body weight, BMC, and fat content from DXA were
compared with the results of carcass analysis in small piglets.

Reproducibility in DXA measurements from the pigs was 0.09% for
body weight, 1.95% for BMC, and 5.35% for fat content. Results also showed
that body weight determined by DXA in 13 piglets was highly correlated
(r=0.999) with scale weight. BMC determined by DXA was significantly
correlated with ash weight (r=0.955) and total calcium content (r=0.992).

Although DXA-measured fat content was significantly correlated (r=.971) with

37

chemically measured fat content, DXA-measured fat content was
overestimated by DXA in 13 piglets. The authors concluded that further
refinements for determining fat content in preterm human infants are
needed.

5 J' E? '11:. l . EDIE?

Jebb, Goldberg, Jennings, and Elia (1995) evaluated body composition
measurements with the Hologic QDR-1000W, software 5.5P and enhanced
version, for effects of depth and tissue thickness. All measurements were
made in the trunk portion of the scan; both oil and water as well as different
meat thicknesses were measured.

All data indicated a trend in the measured fat mass with depth, such
that more fat was measured at extreme depths, <10 cm and >25 cm, than at
intermediate depths. In samples of meat weighing ~ 55 kg, DXA significantly
underestimated the absolute FM compared with direct analysis by 5-8% or 1-4
kg of fat. However, the depths measured in this vitro system represent much
greater thicknesses than measured in vivo in subjects of similar volume.
Consequently, most subjects will lie in the portion of the curves in which
depth has little effect on the measured BMD or PM of the body. Greater depth
should only affect the grossly obese. The error at the low end of the range
should contribute relatively little to the overall measurement of whole-body

composition since it represents a small proportion of the total body mass.

38

Ssanlahleﬁize

One of the most limiting features is the scanning area, which is
approximately 190 cm by 60 cm. It cannot be used in the very obese, and the
accuracy with which individual segments may be distinguished declines as
body size increases. The limited scanning table also can exclude tall persons,
and limits concerning tissue thickness can affect the precision of scan modes
(Wellens et al., 1994).
Beamﬂardening

The x-ray beam analysis of DXA involves the attenuation of the lower
energy expressed as a ratio to the attenuation observed at the higher energy.
Beam hardening may result with increasing depths of tissue. It typically
occurs when lower energy photons are preferentially removed from a
radiation beam compared with higher energy photons. This in turn may lead
to a progressive shift in spectral distribution to higher effective energies with
increasing body thickness. The polyenergetic spectrum of the DXA x-ray beam
may be a potential source of error for DXA since beam hardening may result
in a non-linear measurement scale for BMC.

SI 1' l . l . . | l l l .

Blake, McKeeney, Chaya, Ryan, and Fogelman (1992) looked at the
effects of DXA beam hardening on bone density measurements. The study
consisted of bone represented by layers of aluminum of linearly increased

thickness that was scanned under water thicknesses ranging from 0-25 cm.

39

Although Hologic QDR-1000 has an internal reference wheel which provides
internal standard that continuously monitors for beam hardening, the
attenuation coefficients are only accurate for those introduced by wheel filters.
This was significantly modified by the calibration of the QDR-1000 bone
density scale with the Hologic 3-step phantom.

Beam hardening was shown to have two effects on measured BMD: 1)
at a constant true BMD, measured BMD varied with water thickness, and 2) at
constant water thickness, the BMD scale was not precisely linear. For
conditions to approximate spine and hip studies, the maximum deviation of
measured BMD from a linear scale was 0.023 g/cmz, while the root-mean-
squared deviation (0.01 g / cmz) was comparable to the measurement precision
for a spine or femoral neck scan (~ 1%). The largest error from linearity was
found to occur at the thinnest water thickensses for BMD values in the range
of 0.2 to 0.6 g/cm2. The results of this study showed that for spine and hip
studies, beam hardening was too small to affect clinical studies.

Tissue depth may be another potential source of error for DXA when
assessing body composition. Laskey, Lyttle, Flaxman, and Barber (1992)
evaluated the DXA (Lunar, 3.1) using a phantom that consisted of known but
variable amounts of lard to represent FM, water to represent FFM, and
lumbar spine phantom for BMC. Results showed that determinations of total
tissue mass were very accurate and precise but FM was slightly overestimated
(103% of the calculated value) and FFM slightly underestimated (98% of

calculated value). Also, precision was excellent for all measurements but

40

deteriorated with increasing depths (over 22 cm) of soft tissue. This study
demonstrated that in vitro bone mineral measurements are affected by depth
and composition of the subject and will be least accurate for obese subjects.
1 l' E I' S 1' B

As previously mentioned, pixels are first divided into bone and non-
bone. The non-bone pixels, containing mostly soft tissue, are subsequently
divided into fat and fat-free tissue. Bone pixels also contain some amount of
soft tissue as well. But since DXA has only two photon energies, only two
body compartments can be evaluated, independent of thickness. It has been
estimated that 40 to 45% of the 21,000 pixels in a typical whole-body scan
contain bone in addition to soft tissue, and these pixels therefore are excluded
from the calculation of values for soft tissues (Lohman, 1996). Attempts to
add a third photon have been rejected by the nature of the relationships
between attenuation, energy, and atomic number (Tothill et al., 1994). Hence,
DXA is unable to evaluate tissue thickness under or overlying bone. For
these pixels assumptions are necessary regarding the amount of soft tissue in
each pixel.

Because of the potential limitation discussed above, it is necessary for
DXA manufacturers to make assumptions about the distribution of fat. Some
software packages use the lumbar spine as the only calibration sample. The
assumption is that there is a constant amount of soft tissue surrounding the
vertebral bone This calibration in turn, provides a constant to bone pixels in

which soft tissue is unable to be divided into FM and FFM. However, there is

41

non-uniformity for the amount of soft tissue surrounding the bone, causing
unpredictable errors to occur in the determination of BMD in the lumbar
spine (Tothill 8: Pye, 1992).

For whole body scanning, more complex assumptions are necessary.
Enhanced software versions use complex algorithms to calculate bone pixel
soft tissue by taking surrounding non-bone pixels tissue amount into account.
Dual energy measurements allow the determination of the proportion of fat
in non-bone areas (60% of pixels) and these assessments are used to
extrapolate or interpolate over bone (Tothill et al., 1994).

S 1' . l' I' ll |° |°

Snead, Birge, and Kohrt (1993) studied the body composition of 113
women and 72 men, aged 21-81 yr, by DXA (Hologic QDR-1000) and UWW.
This large sample was subdivided into three smaller groups: young (21-39 yr),
middle (40-59 yr), and older aged subjects (_>_ 60 yr). The study purpose was to
determine whether % BF is overestimated in older people by UWW because
of age-related decreases in BMC. Measures of % BF by UWW and DXA,
respectively, were not different in young people [17.6 :I: 6.4 vs. 17.6 i 7.2%, NS].
Values were slightly, yet significantly different in middle-aged subjects [25.5 i
6.4 vs. 24.1 :I: 6.7%, P < 0.05]. The largest overestimation of % BF by UWW
was in older subjects [34.9 i 7.9 vs. 30.8 :I: 8.7 %, P > 0.05]. Thus, the study
results confirmed the hypothesis that UWW loses its sensitivity for % BF

estimation as BMC varies from the assumed value.

42

The authors also assessed DXA’s prediction of % BF in nine subjects
with packets of lard (2-3 kg) overlying either the trunk or thigh regions. Only
55% of the exogenous fat was identified as fat when it was in the trunk region
compared with 96% when it was positioned over the legs. These data
suggested that the age-related increase in upper body fat tissue may be
underestimated by DXA. The authors concluded that although UWW and
DXA % BF differences were small, there was a widening discrepancy with
advancing age, which possibly may be due to an underestimation of fat by
DXA in the trunk region.

Another study examining proﬁles of fat distribution was done by
Tothill et al., (1996). They compared total and regional fat by DXA (Norland,
software 2.4), UWW, and magnetic resonance imaging (MRI) in 13 pre-
menopausal women, aged 20-51 yr. Results showed poor agreement between
methods; the mean values of % BF include: DXA: 40.0%, UWW: 28.6%, and
MRI: 23.0% (although MRI excluded the head, forearms and feet, estimated
from the DXA measurements to contain 8% of the body fat). Reasons for the
differences between DXA and MRI were sought.

Evidence from in vivo comparisons and phantom measurements
suggested fat calibration errors in the Norland DXA (over a three year period)
which contributed to MRI/ DXA differences. Proﬁles of fat distribution along
the body showed variations in the DXA / magnetic resonance imaging ratio,
particularly of the chest, with the DXA pattern thought to be less accurate.

However, the authors stated that previous comparisons between the Norland

43

DXA and other brands indicate that Hologic and Lunar instruments would
show a closer agreement with MRI and UWW.
E I' [C I III! I' 1122;!

As previously mentioned, DXA assumes that 73.2% of the LBM is
water (Pace 8: Rathburn, 1945). This assumption could be a potential source
of error as FFM may differ due to ﬂuid retention or dehydration.
Hemodialysis is a convenient model for assessing the inﬂuence of changes in
hydration of the FFM on the estimates of other compartments. While bone
and fat compartments are held constant, a known amount of ﬂuid is
removed from the FM during dialysis. The salt-containing ultraﬁltrate has
the radiologic properties of pure fat free tissue (Abrahamsen, Hansen,
Hogsberg, Pedersen, 8: Beck-Neilsen, 1996; Horber, Thomi, Casez, Fonteille, 8:

Jaeger 1992).

 

Horber et al. (1992) used DXA to measure the body composition 7
patients (5 males and 2 females, aged 23-75 yr) on maintenance dialysis with
DXA (Lunar DPX, software 3.1) . Patients were scanned immediately before
and within 1 hour following hemodialysis. Results showed that FFM
decreased in all segments of the body as a consequence of the removal of 0.9
kg-4.4 kg of salt-containing ﬂuid by hemodialysis. The decrease in FFM (P <
0.01) accounted for 94 -_+-_ 5% of the changes in body weight as a consequence of
dialysis therapy. As expected, the amount of weight lost and the change in

FFM observed during dialysis were strongly correlated (r = 0.94, ; P < 0.006).

Conversely, apparent changes in PM (0.12 i 0.17 kg), and BMC (0.0 _-_I-_ 0.01 kg)
were not signiﬁcantly different from zero.

An investigation similar to the Horber et al. (1992) study was
performed by Abrahamsen et al. (1996). The authors evaluated the use of
Hologic QDR-2000 to determine whether DXA was sensitive to soft tissue
changes during hemodialysis. DXA was performed on 19 hemodialysis
patients (9 women and 10 men, aged 26-73 yr) before and after removing 0.9-
4.3L of ultraﬁltrate.

Results of reduction in FFM measured by DXA were highly correlated
with the ultraﬁltrate, as determined by the reduction in gravirnetric weight (r
= 0.975, P< 0.0001; SEE:I:233 g). Whole body BMC was estimated to be slightly
(~0.6%), yet signiﬁcantly (P<0.05) reduced after dialysis, but % BF, BMD, and
lumbar BMD did not change. Therefore, DXA accurately reﬂected the changes
in ultraﬁltrate even when a large proportion of the changes occurred in the
trunk. Once again, this study showed that the accuracy of DXA
measurements was unaffected by hydration status.

SI 1' C . I EDIE! 15E
D'EE | I [122; E

A lack of comparability of DXA results among machines of different
manufacturers because of variations in calibration procedures, photon
energies, and algorithms used to estimate body composition from the
absorption data (Wellens et a1, 1994). Tothill et al. (1994) studied the precision

and accuracy of whole body BMC in three commercially available DXA

45

machines: Hologic, Lunar, and Norland absorptiometers. Assessment of
comparability was made in vivo on 6 women and 5 men, covering a range of
size and adiposity. Precision was also evaluated in vitro through repeated
measurements of a phantom in various configurations. In vivo precision
was measured by making two consecutive measurements for each subject
with repositioning in between.

Results showed good precision for whole body BMC; CVs include 0.7%
for Hologic, 1.4% and 1.0% for Norland MkII and Norland XR36, respectively.
BMC as measured by each machine was compared by linear regression. The
correlations of BMC measurements were high, with Lunar vs. Hologic at
0.974, Norland XR 36 vs. Hologic was 0.978, and ﬁnally Norland XR 36 vs.
Lunar was 0.930. However, there were signiﬁcant differences between
instruments made by different manufacturers. Slopes of regression lines
suggested differences of calibration of up to 8%; SEE values ranged from 110 to
190 g, with maximum deviations from regression of 17%.

Several factors may account for the differences between manufacturers.
One may be that different manufacturers utilize different calibration devices.
Another may involve machine algorithm variations on the fat distribution
model. Also, regional differences in BMC, particularly in the trunk, were due
in part to Hologic having higher bone threshold than the other two
machines. Finally, the dependency of measurement on the thickness of the
subject may vary. Overall, the authors warn that these differences preclude

the interchangeability of results from different instruments; caution is needed

46

even when comparing regional versus whole body measurements from the
same machine.
Different Software Versions

Changes in computer software within a given DXA machine could also
affect measurement reliability and validity. Van Loan, Keim, Berg, 8: Mayclin
(1995) compared two different software packages, 3.4 and 3.6R, in the Lunar
DXA. Fifteen women, aged 20-40 yr, were evaluated. All women were
involved in a weight loss program and had % BF values from 26%-40%.

Total body mass measured via DXA 3.4 (76.3 kg) and 3.6R (76.5 kg) were
comparable to subjects’ scale weight (76.5 kg). BMC and BMD measures were
higher (by 5.5% and 1.8%, respectively P < 0.01) with software version 3.4 as
compared to 3.6R. Also, a larger FFM (by 1,115 g) and a smaller FM (by 1,492 g)
were observed with 3.4 compared to 3.6R. Percent fat estimated by UWW,
DXA 3.4 and DXA 3.6R were 38.1%, 39.9%, and 419%, respectively. The
authors concluded that different software versions used on a given machine
could indeed affect the measurement of total body mass, FM, FFM, and BMC
values. This is likely due to differences in algorithms among various sofware
packages.

Ellis, Shypailo, Pratt, and Pond (1994) looked at the accuracy of body
composition measurements by DXA (QDR-2000, adult whole body, single
beam, analysis versions 5.56 8: 5.57) as compared to total carcass chemical

analysis in 16 pigs with a weight range of 5-35 kg. Also, body composition

47

measurements were performed on 18 boys, aged 4-12 yr, using the same two
DXA software versions.

Results showed that both software versions accurately predicted body
weight, although there were signiﬁcant differences in the partitioning
between BMC, non-bone lean tissue, and BF compartments. All estimates of
body composition were highly correlated (r _>_ 0.99) with the results of the
direct chemical reference method. Standard errors of estimate were 226-271 g
for body weight, 387-429 g for FM, 3.5-4.3 kg for FFM, and 35.4-36.5 g for BMC.

The two different software versions produced a shift in the partitioning
of the nonbone mass of ~ 0.4 kg. Speciﬁcally, version 5.57 resulted in a
reduction of the % BF DXA estimate and an equal increase in the DXA
estimate of FFM. Furthermore, the mean BMC value for version 5.57 was
increased by ~ 65 g or 10% relative to version 5.56.

C l'l'l [H] .122“

A study examining in vitro comparability was done by Renchen,
Murano, Drinkwater, 8: Chestnut (1991). Six Hologic QDR-1000 DXA bone
densitometers located across the United States were used. Nine successive
scans were acquired on each machine using a single anthropomorphic
lumbar spine phantom, manufactured by Hologic. Values for BMC, area, and
BMD were recorded. Means, SD, and CVs were calculated for each machine.
All the CVs (BMC, area, BMD) were less than 1% (range 0.3%-0.6%). The CV
of the means at the six sites were 0.4%, 0.6%, and 0.5% for BMC, area, and

BMD, respectively. Also, the difference between the highest and lowest

48

means of the individual machines was only 1.1%, 1.31%, and 1.07% for BMC,
area, and BMD respectively. Thus, the authors concluded that these small
variations between the DXA systems (all using the same software) are
encouraging for researchers involved in multi-center trials in which data are

pooled.

D l E 2;_ 9] I.
S D'EE .2? If B 1' 1.!

There have been several source differences between DPA and DXA.
DPA employed 153 Gd, which is an isotope that emits two characteristic
photons upon decay. The natural process of radioisotope decay is not easily
controlled, making the radionuclide source in DPA unstable. As previously
discussed, problems arise due to drifting of results as the source decays. DXA,
on the other hand, utilized x-rays instead of a radioisotope as its energy
source. The x-rays generated by the DXA machine are more easily controlled,
thus providing a more stable source (Cullum, E11, 8: Ryder, 1989; Pietrobelli et
al., 1996).

Also, photon ﬂux is greater from the x-ray source in DXA than the
radioisotope source used in DPA. This increased ﬂux allows for decreased
scanning time, improved resolution, and better precision of the image
(Wahner, Dunn, Brown, Morin, 8: Riggs, 1988; Kelly et al., 1988). Precision in
vivo has been shown to be 2-3% for DPA compared to 1% for DXA (Cullum et

al., 1989). Another advantage of DXA is the smaller radiation exposure

49

compared to DPA. DPA radiation exposure for a whole body scan is ~2.0 mR
(Wang et al., 1989) in contrast with the Hologic DXA which is <1.0 mR for a
whole body scan (Hologic Operator’s Manual, 1990).

E . . [DB 9 D2; 9

Cullum et al. (1989) evaluated DPA compared to the Hologic QDR-1000
for bone density measurements. Precision of DXA was also determined in
vitro by performing 10 repeat scans on a series of rectangular slabs of different
density bone-equivalent material placed in a variety of water depths, 2.0 cm to
30.0 cm. In vivo precision was also done scanning nine patients on ten
occasions, over a period of 4-6 weeks.

Results of both the in vitro and in vivo measurements were better
than a CV of 1%. The only larger CV was from the in vitro analysis of
varying thicknesses. At 20 cm depth the CV was 0.71%, while 25 and 30 cm
were 1.43% and 3.19% respectively. Since precision was limited by the
accuracy of the detection of the bone edge and the reproducibility of object
positioning, the better resolution of DXA provided more precise results. The
authors concluded that DXA showed significant improvement over DPA in
image quality.

Gluer et al., (1990) also compared DPA with the Hologic QDR-1000 for
bone density measurements of the spine and hip. For in vitro studies, an
anthropomorphic spine phantom, lucite blocks, and variable thickness
mixtures were used as reference materials. Short-term in vivo analysis

involved two consecutive exams of 10 healthy adults, aged 21-40 yr, with

50

repositioning between scans. Long-term precision of the spine-scanning
mode was analyzed by scanning an X-Caliber (Hologic) spine phantom. This
phantom was analyzed once per day with both DPA and DXA over a nine
month period. During that time, there was one source change for the DPA
scanner.

Results from this study indicated that DXA had excellent precision in
vitro as compared to DPA. Speciﬁcally, long term precision error in vitro was
1.30% in DPA and 0.44% in DXA. However, both DPA and DXA were highly
correlated in vivo: r= 0.98 for the spine and r: 0.95 for the femoral neck.
Small differences between the two machines were explained by differences in
hardware, or resolution, and software, or bone edge deﬁnition. The scanning
time for both spine and hip measurements was reduced from 20-40 minutes
for DPA to 6-7 minutes for DXA. Thus. the authors concluded that the
precision, spatial resolution, and seaming time of DXA are significant
improvements over DPA.

Kelly et al. (1988) also evaluated the bone density of the lumbar spine
for both DPA and DXA. The lumbar spine BMD was measured by DPA
(Lunar) and by DXA (Hologic QDR-1000) in 85 patients, aged 21-78 yr. Each
patient was measured once by DPA and twice by DXA on the same day with
repositioning between patients. Long term stability was evaluated by making
serial measurements of a spine phantom of known elemental composition.

The spinal BMD measurements with DPA and DXA were linearly

related and highly correlated (r=.98) over a range of bone densities, from

51

severely osteopenic to high normal. Compared to DXA, DPA values were
three times as variable for 170 days and increased 1.0% after a software change.
Measurement time by DXA was 5-8 min. for the lumbar spine, with a
maximum radiation exposure of 3 mrem, which was significantly less than
corresponding DPA values. Therefore, the authors concluded that DXA is the
superior method for spinal BMD measurements since DXA was faster than
DPA, involved less radiation exposure, provided greater image resolution,
and better short- and long-term reproducibility.

Paciﬁci et al. (1988) assessed the short-term precision of the Hologic
QDR-1000, and also compared DPA to DXA. Short-term precision was done
by performing three DXA measurements of the lumbar spine in 19 healthy
women, aged 34 :I: 9.3 yr, over a three week period. In addition, both DPA and
DXA measurements were obtained in 52 women, aged 54.3 :I:12.5 yr.

Results of short-term precision included DXA measurements at a CV
of 1.0%. Also, a signiﬁcant correlation was found between DPA and DXA (r=
0.94; P <0.001). From the shape of the regression line, it appeared that DPA
values were approximately 6.8% higher than DXA values. Hence, a
conversion factor of -6.8% was used to convert DXA to DPA values. Both had
similar rates of bone loss; 0.66% for DXA and 0.57% for DPA. Hence, the
authors concluded that DXA was a precise method for vertebral bone mass
measurements.

Strause, Bracker, Saltman, Sartoris, and Kerr (1989) also investigated

lumbar spine bone measurements using both DPA and DXA (Hologic QDR-

52

1000). Subjects included 131 postmenopausal women, over 55 yr of age and
free from other major risk factors for osteoporosis. Bone mineral densities
were determined for 4 individual levels in the lumbar spine (Ll-L4). All
subjects were scanned by both DPA and DXA, with measurements performed
within 15 minutes of each other.

Results showed a high correlation (r= 0.918) between BMD
measurements of the lumbar by DPA and DXA. However, regression data
suggested an interlevel difference between DPA and DXA, with BMD values
derived by DPA consistently higher than those obtained by DXA. This
discrepancy may have been due to poor image resolution which in turn may
have produced inaccurate determination of vertebral edges. The authors felt
that vertebral bone of the spine may have been underestimated by DPA
computer analysis, leading to a lower BMD value from the DPA instrument.

Wahner et al. (1988) also measured the lumbar spine using DPA and
DXA. Long-term precision was assessed through spine phantom analysis
performed once weekly with the DPA instrument and once or twice daily
with the DXA instrument and over a period of 4 to 6 months. For estimation
of short-term precision, duplicate scans on the lumbar spine were made in
and 5 volunteers with DPA and 15 volunteers with DXA.

Wahner et al.’s results showed good correlation (r= 0.988) between DPA
and DXA. DPA had a longer scanning time and a lower image resolution
than the DXA system. Long-term precision had CVs for BMD of 1.5% for the

DPA instrument (n=20) and 0.4% for DXA (n=214). Short-term precision as a

53

mean percentage difference between the two scans was 1.7% for DPA (range,
0.2 to 4.4%; N: 5) and 1.0% (range, 0 to 3.1%; N: 15) for DXA. Overall, DXA
instruments showed better precision in a spine phantom and reduced
inﬂuence of thickness for patient measurement. Patients’ BMD values were
consistently lower with the DXA instrument apparently because of better
accuracy in the speciﬁc measurement area. The authors concluded that DXA
instrument is a major advance in bone mineral absorptiometry and provides
improved, yet less expensive, measurements in research and clinical
applications.

Bergsma-Kaddijk et al. (1996) measured the body composition of 20
young females (aged 19-27 yr) and 18 elderly females (aged 65-78 yr) using both
UWW and DXA (Lunar, software 3.1). The BMC of the young women was
2644 :I: 274 g, while older women had a BMC of 1987 i 234 g (P<.001). The
mineral fraction of FFM was also signiﬁcantly different (P<.001) with older
women having values of 6.0:I:0.6% and younger women having 7.5:t0.5%.
These BMC differences resulted in apparent differences in % BF as estimated
by UWW. The % BF of the young women was underestimated, while it was
overestimated in the elderly subjects.

Hansen et al. (1993) evaluated the body composition of 100
premenopausal, non-athletic women aged 28-39 yr. The authors used four
predictor methods including DXA (Lunar, 3.1), BMI, skinfolds, and BIA. The

criterion method was UWW. The primary purpose was to compare the

54

prediction of % BF from body density determined by DXA versus UWW. The
secondary purpose was to examine the effect of bone mineral on the body
density measures obtained from UWW using a multicomponent approach to
further evaluate DXA.

DXA correlated well with UWW (r=.91) and had a SEE of 2.4%. When
% BF was estimated from other methods, larger SEES were obtained: 3.0% for
skinfolds, 3.3% for BMI, and 2.9% for BIA plus weight. Also, individual body
fat values derived from UWW were corrected for bone mineral variation.
Bone mineral was found to be a signiﬁcant predictor for the % BF fat models
of DXA and skinfolds. The sum of 4 skinfolds predicted body density by
UWW with a SEE of .0068 g/ cc, but when corrected for bone mineral this
decreased to .0056. DXA also had a lower SEE when corrected bone mineral
was added, from .0053 to .004. The authors concluded that the DXA method
was extremely precise and highly correlated with FFM and % BF.

Morrison et a1. (1994) studied body composition in 31 Black and 38
White girls, aged 9-16 years. DXA measurements of FFM and % BF were
made using Hologic QDR-1000. Corresponding FFM and % BF from UWW
were calculated by the Siri equation, using the model of Lohman (1986) for
White girls only.

The two-component model of UWW significantly overestimated % BF
in both White and Black girls compared to estimates from DXA. In White
girls, % BF values measured from UWW and DXA averaged 28.4% and

25.3%, respectively. For Blacks, average % BF values for UWW and DXA

55

were 28.7% and 27.9%, respectively. The authors felt that Since the BMC
density of FM increased or approached adult status in black girls, differences
between UWW and DXA decreased. Moreover, the inability of UWW to
account for BMC in these girls can be corrected either by including it in a
multicomponent model or by using the DXA technique.

Pritchard et al. (1993) evaluated DXA as a method of measuring % BF.
Their purpose was to a) compare the precision of the Hologic QDR-1000 with
the Lunar DPX and b) compare both DXA machines to three other body
composition modalities (UWW, skinfolds, and BIA). Ten Hologic DXA scans
were performed on one subject to assess precision. The CV was 1.8% for %
BF, 0.6% for FFM, and 2.1% for FM.

Other results showed that DXA machines were highly correlated with
r= 0.986, P<0.00001. Also, based on the observation of 12 subjects, correlations
of QDR and DPX with UWW for % BF were high: r= 0.916 for QDR and r=
0.913 for DPX. The authors performed a limits of agreement estimate which
. showed a between-method difference of + 1.3% for the QDR compared with
UWW. The DPX showed a between-method difference of + 4.8% for % BF
' compared with UWW. Correlations between both DXAs and other methods
were also high. For QDR: r= 0.824 with skinfolds and r= 0.972 with BIA. For
the DPX, correlations were r = 0.923 for skinfolds and r = 0.910 for BIA. The
authors concluded that both DXAs measured % BF with greater precision
than UWW as reﬂected by the coefﬁcient of variability and the high

correlation with other methods.

56

Wellens et al. (1994) evaluated the body composition of 78 women and
50 men, aged 18-67 yr, by using DXA (Lunar, software 3.4), UWW, and TBW.
The three body composition methods were compared to determine the degree
of agreement among methods, and reasons for discrepancies were explored.

Results showed that in both men and women there were no significant
inter-method differences for % BF and FFM estimates between UWW and
DXA. On the other hand, % BF measures in women were Signiﬁcantly less
with TBW compared to UWW in women (mean difference 2.7 i4.2%) and in
men (mean difference 2.2 i. 3.2%). Thus, for each sex, statistical analysis
indicate that body composition estimates Show less agreement between TBW
and either UWW and DXA.

Bioelectric Impedance Analysis

Bioelectric impedance analysis (BIA) uses an electrical current to
estimate total body water by measuring the resistance and reactance of the
tissues. Resistance (R) is a measure of pure opposition to the current ﬂow
through the body. Reactance (X,) is the opposition to current ﬂow caused by

the capacitance produced by the cell membrane. The impedance (Z) is a
function of the resistance and reactance where Z = \/(R2 + Xcz) (Heyward 8:
Stolarczyk, 1996).

8'12"] EBI!

Thomasett (1962) was the ﬁrst to establish the basic foundation of BIA
principles. With this method, a low-level current is sent through the body

via surface electrodes. Subsequently, the BIA analyzer measures body

57

impedance, or resistance to current ﬂow. Biological tissues act as conductors
or insulators, and the ﬂow of current through the body will follow the path of
least resistance. Impedance is based upon the nature of the conduction of an
applied electrical current in an organism. An individual’s TBW can be
estimated because the electrolytes in body water are excellent conductors of
electrical current.

When TBW volume is large, current ﬂows more easily through a
given body with little resistance. On the other hand, resistance to current
ﬂow will be greater in individuals with larger FM, given that adipose tissue is
a poor conductor due to its relatively small water content. Because the water
content of FFM is relatively large (73% water), FFM can be predicted from
TBW estimates.

To measure total body impedance, a low-level excitation current (500 -
800 uA) at 50 kHz is used. At low frequencies (~1 kHz), the current passes
through the extracellular ﬂuids only. At higher currents (500 - 800 uA), the
current penetrates cell membranes and passes through the intracellular ﬂuid,
as well as the extracellular ﬂuid (Lukaski, 1987). Given that fat is a poor
conductor of electrical current, the total body impedance primarily reﬂects the
volumes of water and muscle compartments comprising the FFM and the
extracellular water volume (Kushner, 1992).

Also, impedance is a function of resistance and reactance, where

Z = \l (R2 + X3)

58

As stated earlier, R is a measure of pure opposition to current ﬂow through
the body and Xc is the opposition to current ﬂow caused by capacitance
produced by the cell membrane (Kushner, 1992). Since the size of R is much
larger than Xc when measuring whole body impedance, R is usually a better
predictor of FFM and TBW than Z (Lohman, 1992). For these reasons, the
resistance index (height (Ht2)/ R), instead of th / Z, has been used in many BIA
models to predict FFM or TBW (Lohman, 1992).
Admﬂagesﬂsingﬁm

The BIA modality offers a number of potential advantages to other
body composition techniques including the following:
1) requires minimal operator training and is easy to perform
2) is generally more comfortable than other methods and requires

minimal intrusion on an individual’s privacy
3) is safe, relatively inexpensive, and portable
4) can be used to estimate the body composition of obese individuals

(There is less observer error associated with BIA than skinfold

thicknesses, especially in the obese; Segal, Van Loan, Fitzgerald,

Hodgdon, 8: Van Itallie, 1988).

AsaiunntienmLBIA

There are a few assumptions inherent in the BIA method of body
composition analysis. One assumption is that the human body is shaped like
a perfect cylinder with a uniform length and cross-sectional area. This

assumption is not exactly true. The human body more closely resembles five

59

cylinders connected in series versus one large uniform cylinder (Kushner,
1992). Also, these cylinders are not uniform in length or cross-sectional area,
therefore, resistance to current ﬂow through these segments will differ.
Another assumption is that at a ﬁxed frequency (50 kHz), the
impedance (Z) to the current ﬂow through the body is directly related to the
length (L) of the conductor and inversely related to its cross-sectional area (A)
(Heyward 8: Stolarczyk, 1996). That is,
Z = p (L/A),
where p is the speciﬁc resistivity of the body’s tissues and is assumed to be
constant. To express this relationship in terms of Z and the body’s volume,
instead of its cross-sectional area, the equation is multiplied by L/ L.
Z=P(L/A)"(L/ L)
Since A * L is equal to volume, the equation can be rearranged to be:
V=pU/Z
Thus, the volume of the FFM or TBW of the body is directly related to L2, or
height-squared (Ht’).
W
The basic BIA assumptions are simple ideas which do not
accommodate for the complexity of the human body. The application of the
BIA equation may not be perfect because of the complex geometric shape of
the body (Van Loan, 1990). Also, the speciﬁc resistivity (p) of the body’s
tissues is not constant and has been shown to vary among body segments.

Speciﬁc resistivity differences exist due to changes in tissue composition,

60

hydration levels, and electrolyte concentration (Kushner, 1992). For example,
Chumlea, Baumgartner, and Roche (1988) reported that the speciﬁc resistivity
of the trunk is two to three times greater than that of the extremities. The
result would be that the current would ﬂow with least resistance in the limb

areas, particularly in the arms.

 

Hoffer, Meador, Simpson (1969) estimated TBW from tritium dilution
space in 20 healthy volunteers and 34 patients. Four surface electrodes were
used, two placed on the dorsal surface of the right hand and two placed on the
ﬂexor surface of the left foot; with 100 uA AC, at 100kHz. In both subject
groups, th/ Z was the best predictor (r= 0.92; P<0.001) of TBW. Thus, the
authors reported a strong relationship between total body impedance
measures and TBW, suggesting that BIA may be a valuable tool for analyzing
body composition and assessing TBW in the clinical setting.

Jackson, Pollock, Graves, and Mahar (1988) studied the reliability and
validity of BIA in the determination of body composition; they also compared
its accuracy with the results obtained by standard anthropometric methods.
Subjects included 44 women and 24 men, each having % BF measured via
UWW, skinfolds, and BIA. Each subject was tested four times by two testers
on two different days. An additional 26 men and 38 women were tested once
and combined with the data used for the reliability analysis to cross-validate

BIA estimates of % BF with UWW % BF.

61

The % BFS from BIA, skinfold, and UWW were found to be reliable
(Ru: 0.957—0987, with SEM: 0.9 to 1.5% fat). The cross-validation correlations
for BIA determinations of % BF ranged from 0.71 to 0.76, which were
Signiﬁcantly lower than those obtained using the sum of seven skinfolds
equations (rxy= 0.92 for men and 0.88 for women). Correlations between the
weight-to-height ratio for BMI and UWW % BF were 0.75 and 0.74 for men
and women, respectively. The SEE for the two BIA models ranged from 4.6%
to 6.4 % BF compared with 29% and 3.6% BF for the skinfold equations.
Thus, the authors concluded that the BIA method for measuring body
composition was comparable to the BMI method, but less accurate than
skinfold analysis, with height and weight accounting for most of the variance
in the BIA equation.

Lukaski, Johnson, Bolonchuk, and Lykken (1985) used BIA to deveIOp
models for estimated the FFM and TBW in 37 men, aged 28.8 :I: 7.1 yr.
Current source was 800 uA at 50 kHz. FFM was also assessed by UWW, with
TBW determined by deuterium dilution and total body potassium (TBK)
from whole body counting.

Results produced test-retest correlation coefﬁcients of 0.99 for a Single R
measurement; the reliability coefﬁcient for a single R measurement over 5
days was 0.99. Because R and Z were equally correlated with TBW (r= 0.99)
and Xc is not as strongly related to Z (r= 0.70), the investigators included only
R as the BIA variable in the prediction of FFM and TBW. Further, linear

relationships were found between R values and FM (r= -0.86), TBW (r= -

62

0.86), and TBK (r= -0.79). Signiﬁcant (P < 0.01) increases in the correlation
coefﬁcients were observed when the predictor Ht2 / R was regressed against
criterion measures of FFM (r= 0.98), TBW (r= 0.95), and TBK (r= 0.96). Thus,
the authors concluded that BIA is reliable and valid for the estimation of
human body composition.

Segal et al. (1988) provided validation of BIA for body composition
estimation. At four laboratories, UWW-determined body composition was
compared with BIA in 1069 men and 498 women aged 17-62 yr (% BF ranged
from 3-56%). Among the labs, some regression coefﬁcients differed, but these
differences were eliminated after adjusting for subjects’ body fatness
differences among labs. Data were pooled to derive fatness-speciﬁc equations
for predicting FFM by UWW. The resulting correlations ranged from 0.907 to
0.952 with SEES of 1.97 to 3.03 kg. Therefore, the authors concluded that BIA
was valid estimator of body composition and that the precision of predicting
FFM from impedance can be enhanced by sex- and fatness-speciﬁc equations.

u.- :. - 7.1m “7". - z: I; a m i an, mm ”In
Changes

Ross, Leger, Marin, and Roy (1989) studied the sensitivity of BIA to
detect changes in body composition. UWW and BIA techniques were used on
male subjects (n=17) before and after a 10-week diet and exercise regimen. A
control group of 20 males was also studied. The authors found that for both
FFM and % BF, the Lukaski et al. (1985) and Segal et al. (1988) BIA equations

produced values that were not significantly different (P >0.05) from UWW in

63

both pre- and postregimen values. Thus, the authors concluded that their
study findings suggest that the BIA method with these equations is a valid
means of predicting changes in % BF as measured by UWW and the Siri
equation. It is possible that 10 weeks was not sufﬁcient to effect a %BF change
that could be detected by any of the modalities used.

Kushner and Schoeller (1986) identified factors and conditions which
may require standardization during BIA testing for healthy subjects and those
in a medical setting. Some of these include: accurate measurement of height
and weight, posture (supine) and time (e.g., time of day when measurements
were taken, were they standing before, for how long?), abduction of limbs
(due to their geometric shape, the extremities contribute ~90% of whole-body
Z), consumption of food or beverage (could inﬂuence Z by changing TBW
and extracellular water volumes), and recent exercise. In terms of
positioning, the wrist electrode is the single most important site, since the
arm appears to make the largest contribution to the total measurement (Jebb
8: Elia, 1993).

Chumlea, Guo, Cockram and Siervogel (1996) studied FFM and % BF
by DXA and total-body and segmental impedance measures taken at 16
frequencies (from 5 to 1300 kHz) in a sample of white men and women aged
18-30 yr. Results showed that the segmental impedance Spectrum variables

for FM and % BF and the ratios of low- to high- frequency impedance from

the trunk were significantly associated with total body fatness as measured by
DXA.

Gleichauf and Row (1989) assessed the reliability of resistance (R) and
body composition estimates measured by BIA in 25 women during their
menstrual cycles. To examine the effect of the menstrual cycle on BIA
measures, each subject’s cycle was divided into four stages: 1) menses, 2)
follicular, 3) postovulatory, and 4) premenstrual. Weight and BIA were
measured daily for one cycle and sodium intake was recorded. The precision
of BIA was examined, and average resistance, weight, Na intake, and subjects’
calculated body composition during the four phases of menses were
compared using paired t tests.

Small, yet signiﬁcant differences were observed between phases 1 and 2
for R (P < 0.001), weight (P <0.05), and FFM (P <0.05). Differences were also
observed between phases 2 and 4 for R (P <0.05), and weight (P <0.05). No
signiﬁcant differences were observed for % BF. Changes in body weight (P
<0.001) associated with Na intake explained a signiﬁcant proportion of error
in R measures. The authors concluded that on average, BIA reliably
measures body composition variables, regardless of menstrual status.
However, for a given woman, alterations in hydration status (due to
menstrual phase) may affect her results.

Another factor that may affect reliability of BIA is skin temperature.
Gudivaka, Schoeller, and Kushner (1996) evaluated the effect of skin

temperature on multifrequency BIA and the prediction of body water

65

compartments. Skin temperature was raised over 50 minutes from a baseline
of 29.3 :I: 2.1 °C to 35.8 i 0.6 °C in 6 adults. Then, skin temperature was cooled
for 20 minutes to 26.9 i 1.3 °C. All heating and cooling was done by using
temperature controlled blankets.

Impedance has been shown to vary inversely with changes in skin
temperature across all frequencies (5-500 kHz). Thus, controlled ambient and
skin temperatures should be included in the standardization of BIA
measurements. Gudivaka et al. (1996) found that errors in predicted TBW are
<1% within an ambient temperature range of 22.3 to 27.7 °C (72.1-81.9 °F).

WM

There may be several sources of error affecting the reliability and
accuracy of BIA. Subject factors such as eating, drinking, and exercising may
alter hydration state and thereby affect impedance measures and FFM
estimates. Technician skill may also lead to incorrect measurements.
However, skill should not be a major source of error, provided that the
technician follows standardized procedures. The proximal sensor electrodes,
in particular, need to be correctly positioned at the wrist and ankle. A 1 cm
displacement of the sensor electrodes may result in a 2% change in R (Elsen,
Siu, Pineda, 8: Solomons, 1987). Other potential sources of error include
environmental factors such as ambient temperature and relative humidity.

W
The BIA method has been used to generate mathematical models

relating impedance measurement to body composition variables. BIA

66

prediction models are generally one of two types. The first is population-
speciﬁc. This type of equation is developed using a homogeneous sample in
order to predict accurately to a like population. For example, individuals
could be grouped by age, ethnicity, gender, physical activity level, level of
body fatness, etc., and speciﬁc equations developed for each. These equations
are only valid for, and should only be applied to, individuals whose physical
characteristics are representative of the population subgroup.

The other type of BIA prediction equation is one that can be
generalized to a larger population. These equations are developed for
heterogeneous populations which may vary, for example, in age, gender, and
body fatness. This approach can account for biological variability among
population subgroups by including factors such as age and gender as predictor
variables in BIA equations.

Regardless of the approach, the predictive accuracy of BIA equations
typically is improved by including body weight, along with HF, and R, and Xc
in the BIA regression model. Since many of the assumptions previously
mentioned may not be accurate, including body weight in the equation may
be one way of accounting for the complex geometric Shape of the body, as well
as individual differences in trunk size (Kushner, 1992). To date, there are i
only a few BIA equations developed against multi-component models to
derive reference measures of FFM (Guo, Roche, and Houtkooper, 1989 ;

Lohman, 1992).

67

Body weight and height are signiﬁcantly correlated with each body
composition variable. It has been argued by some investigators that height
and weight account for the majority of the BIA prediction of FFM or TBW
(Mazess, 1991). Regardless, in studies where a heterogeneous population is
represented, th / R appears to be the single best predictive variable for FFM or
TBW (Kushner, 1992).
$1.1]. E lillll

According to Heyward and Stolarczyk (1996) there are two BIA
equations that are appropriate for female athletes. One equation, developed
by Lukaski and Bolonchuk (1987) was originally formulated using males and
females with varying activity levels. Subjects included 312 individuals, aged
18-73 yr. Of these, 151 subjects were used to develop an impedance model for
predicting FFM. The rest, aged 19-50 yr, were assigned to another group
whose data were used to cross-validate the model.

Stepwise multiple regression analysis was performed to identify the
best prediction equation which follows:

FFM = (0.734 " th/ R) + (0.116 * Wt) + (0.096 * Xc) + (0.878 * Gender) - 4.03

(r2 = 0.988, SEE = 2.06 kg); gender is coded as 1=male, 0=female

The cross-validation produced a Signiﬁcant (P <0.0001) correlation between
FFM by UWW and FFM by BIA (r= 0.970, SEE=2.29 kg). Statistical analyses of
the regression coefﬁcients of this line indicated that the slope was not
signiﬁcantly different from one (F: 0.06, P = 0.81) and the intercept was

similar to zero (F: 0.02, P = 0.98). Thus, the regression line was similar to the

68

line of identity. The authors concluded that BIA provides valid and
acceptable estimates of body composition in healthy individuals.

Lukaski, Bolonchuk, Siders, and Hall (1990) used the equation
discussed above on a group of 104 male and female college athletes. All
athletes underwent UWW and BIA under controlled conditions
(measurements made two hours after consuming a light meal and no
preceding exercise) and uncontrolled conditions (measurements made
without regard to preceding exercise, level of hydration, or eating).

Results showed that the relationships between FFM measured by
UWW and BIA in both controlled (r= 0.988, SEE: 2.04 kg) and uncontrolled
(r= 0.977, SEE: 2.85 kg) groups predicted FFM with no signiﬁcant differences
(P >0.05) from the lines of identity. When comparing % BF measured by
UWW and BIA, the regression line under controlled conditions was similar
to the line of identity (r= 0.874, SEE: 2.81%). However, the relationship
between % BF by UWW and BIA under uncontrolled conditions was
different (P< 0.05) from the line of identity (r= 0.758, SEE=3.76%). Although
only athletes were used as subjects, Lukaski et al. (1990) concluded that these
ﬁndings indicated the need for controlled measurement conditions to obtain
valid body composition estimates using the TBW method in ”healthy
people”.

The second BIA equation for female athletes was developed by
Houtkooper et al. (1989). The authors actually developed their equation on a

group of 45 active and inactive females, aged 18-38 yr. FFM was determined

69

by UWW and the Siri equation, also correcting for BMC by dual and single
photon absorptiometry. Results from. an equation which apparently included
R, X, and Wt Showed an r2 = 0.83, with SEE of 1.7 kg. However, the Speciﬁc
equation was not included in the publication, which was only in abstract
form. Moreover, Heyward and Storlarczyk (1996) recommend this ”equation”
for female athletes in their book. Unfortunately, the equation included in the
book is incorrect. However, the correct formula was published correctly in
another body composition text (Lohman, 1992).
Near Infrared Interactance

Near infrared interactance (NIR), another method to estimate body
composition, is relatively new. NIR was originally used to measure the
protein, fat, and water content of agricultural products. Conway, Norris, and
Bodwell (1984) applied this technology to study human body composition.
They used an expensive, high-precision (6 nm) computerized
spectrophotometer to measure % BF in 20 males and 33 females, aged 23-65 yr.
The % BFS from NIR were compared to results from measurements of
deuterium oxide dilution (r= 0.94), skinfolds (r= 0.90), and ultrasound (r=
0.89). The authors concluded that the noninvasive NIR technique was safe,
rapid, easy to use, and may also be useful to predict % BFs in the obese.

After the Conway et al. (1984) study, a low-cost, portable, wide-Slit
spectrophotometer was designed by computer simulation for work in the
ﬁeld (Conway 8: Norris, 1987). Shortly after, a commercial NIR analyzer

called the Futrex-5000 was developed and marketed based on the results of

70

these ﬁrst two studies (Conway et al., 1984; Conway 8: Norris, 1987). The NIR
method involves a wand which emits infrared light. The wand is typically
placed over the belly of an individual’s biceps brachii muscle and optical
densities of remitted light from various tissues are measured.
E . E . . l E I 113

NIR works on the following principles. When electromagnetic
radiation strikes a material, the energy is reﬂected, absorbed, or transmitted
depending on the scattering and absorption properties of the sample. A
probe, or wand emits electromagnetic radiation to a selected site on the body,
collects interactive energy (which is the combination of reﬂected and scattered
energy), and conducts it to the detector. This remitted energy is represented
through optical density (OD) values. Subsequently, % BF may be determined
from an equation which includes these optical densities. Interactance data are
calculated by the instrument as the ratio of the energy received from a scan
site to the energy received from a calibration standard, a one centimeter thick

Teﬂon block. Thus, changes in optical density are calculated by:

A OD = OD standard - OD measured from body site
AssumptiQnLQLNIR
As with other body composition methodologies, there are assumptions
associated with the NIR technique. One assumption is that the degree of
infrared light absorbed and reﬂected is related both to the composition of the
tissues and to the speciﬁc wavelength of the near-infrared light. Conway et a1.

(1984) demonstrated that peak absorption wavelengths are 930 nm for pure fat

71

and 970 nm for pure water. Also, the Shape of the interactance curve at these
two wavelengths is a function of the amount of fat and water present in the
measured sample. The Futrex-5000 emits two wavelengths, 940 nm (resulting
in 0D,) and 950 nm (resulting in ODZ), and then measures the amount of
light reﬂected by the underlying tissues (Futrex, 1988). Therefore, OD changes

are measured by:

A OD1= OD standard - OD1 from measured site

A OD2= OD standard - OD2 from measured site

It is not clear why the manufacturer selected these wavelengths (940 nm and
950 nm) instead of the wavelengths identifying pure fat and water (930 nm
and 970 nm, respectively) (Heyward 8: Stolarczyk, 1996). Perhaps this was
done because the body is composed of neither pure fat nor pure water, but a
combination of the two.

Another concept central to current NIR methodology is that optical
densities are linearly related to subcutaneous fat at the biceps site and total
body fatness. To examine this, Quatrochi et al. (1992) examined OD as a
function of skinfold thicknesses at different sites including: pectoral,
midaxillary, biceps, triceps, subscapular, abdominal, suprailliac, thigh, and
calf. Results indicated that OD values are signiﬁcantly correlated (P < 0.05)
with skinfolds, age and % BF (via UWW) in the pectoral, biceps, subscapular,
and abdominal regions. The highest correlations included the pectoral
(skinfold- 0.61, age- 0.53, and % BF- 0.70) and biceps regions (skinfold- 0.66,

age- 0.42, and % BF- 0.71). While none of the relationships were extremely

72

high, the OD measures at the biceps appeared to correlate the best with
skinfold thickness measures.

Elia, Parkinson, and Diaz (1992) also examined different body sites for
correlations between optical densities and skinfolds at the site. Correlations
for % BF were highest for the bicep at r= -0.71 for OD1 and r= -0.75 for ODZ.
This was in comparison to the triceps (r= -0.54 for OD1 and r= -0.61 for ODZ),
thigh (r= -0.36 for OD1 and r= -0.41 for ODZ), biceps and triceps (r= -0.70 for ODl
and r= -0.75 for ODZ, and biceps, triceps and thigh (r= -0.65 for ODl and r= -0.69
for 0D,).

In addition, Hortobagyi, Israel, Houmard, McCammon, and O’Brien
(1992) examined NIR using multiple sites. Correlation coefﬁcients indicated
that the bicep was the best site for NIR measurement of optical density (when
compared to skinfold thicknesses), although no site had particularly high
correlations. Some examples of the correlations included: biceps at r= -0.67 for
ODl and r= -0.68 for ODZ, triceps at r= -0.47 for OD1 and r= -0.48 for ODZ, and
the thigh at r= -0.45 for OD1 and r= -0.40 for ODZ.

Besides the biceps being apparently the most valid body site to measure,
there is an assumption that NIR light penetrates a depth of 4 cm and is
reﬂected off the bone and back to the detector. This claim, which is made by
the manufacturer, has not been proven through research. The measured
optical density of the reﬂected light radiation is inﬂuenced by the Speciﬁc
absorption characteristics of the underlying tissue. Hence, the OD measures

of NIR are supposed to measure both subcutaneous and intramuscular fat.

73

Even if this depth penetration is true, since skin and fat layers differ among
individuals, the predictive value of resultant measures may be compromised.

Supporting this concept, Quatrochi et al. (1992) noted that the
relationship between OD and skinfolds was strongest in the lean, young
women; with the technique possibly underestimating the true fatness of
older, more obese women. Also, Elia et al. (1992) commented that NIR was
found to underestimate % BF increasingly as the degree of adiposity
increased. Thus, differences apparently exist in the layering of skin and fat in
individuals of various % BFs.

E I. I'll ll! l’l’l “1m

Quatrochi et al. (1992) examined the reliability of optical density
measures for NIR. The average reliability coefﬁcients for ODS measured
across two days ranged from r= 0.76 for rnidaxillary OD, r: 0.79 for thigh, r:
0.85 for calf, r= 0.87 for suprailiac, r= 0.88 for triceps, r= 0.93 for subscapular, r=
0.96 biceps and abdominal, r= 0.99 for pectoral.

Israel et al. (1989) evaluated the body composition techniques of NIR,
three-, and seven-site skinfold equations (Jackson 8: Pollock, 1978), and BMI
in 80 physically active males (average age was 25.8 yr). The criterion method
used was UWW. The equation for NIR was provided by the manufacturer
(Futrex, 1988). In addition to OD measures, the equation includes height,
weight, age, gender, and activity level. Results showed correlations of r= 0.79
for NIR, r= 0.90 for seven-site skinfolds, r= 0.87 for three-site skinfolds, and r=

0.68 for BMI. Mean % BFS included: and 9.8% for NIR (SEE: 4.16%), 129% for

74

UWW, 12.9% for seven-site skinfolds (SEE: 2.9%), and 12.3% for three-site
skinfolds (SEE: 3.4%). Important contributors for the NIR equation were
0D,, activity level, and weight. Age and height contributed less. From the
results of the correlations and SEES, the authors concluded that NIR did not
accurately estimate body composition in white males.

Elia et al. (1990) investigated the body composition of 15 males and 14
females, aged 18-40 yr, using the following methods: NIR, whole-body
impedance/ resistance, skinfold thicknesses, and BMI. All techniques were
compared to UW. Results showed that NIR was highly correlated with
UWW for males: FM (kg) (r=0.90, SEEzl: 2.29), % BF (r= 0.80, SEE:I: 3.10), and
FFM (kg) (r= 0.91, SEE: 2.43). For females, NIR correlations with UWW
included: FM (kg) (r= 0.93, SEE:Iz 2.35), % BF (r= 0.82, SEE:Ic 4.33), and FFM (kg)
(r= 0.61, SEE:I: 2.60). However, BMI also correlated well with UWW for males:
FM (r= 0.84, SEE:I: 2.80), % BF (r= 0.71, SEE:|: 3.66), and FFM (r= 0.88, SEEi 2.81).
For females BMI correlations with UWW included: FM (r=0.97, SEEi 2.35), %
BF (r= 0.82, SEE: 4.33), and FFM (r=0.62, SEE:Iz 2.58). Thus, the authors
concluded NIR had little or no advantage over other Simple anthropometric
techniques in evaluating body composition.

Hortobagyi et al. (1992) evaluated body composition using several
techniques including NIR, UWW, and skinfolds with seven Sites (Jackson 8:
Pollock, 1978). Subjects included 171 men, with 52 black and 119 white males
(aged 22.6:I:7.6 yr). Results Showed the reliability for repeated NIR trials

included: 0.97 for the biceps, chest, and abdomen; 0.95 for the supraillium;

75

0.87 for the subscapula; 0.73 for the thigh; 0.70 for the axilla; and 0.65 for the
triceps.

Also, mean % BF for the criterion method of UW was 13.4%. The
average % BF for NIR using a combination of eight sites was signiﬁcantly
underestimated at 12.7%, r= 0.71, SEE: 3.64%. In addition, the average % BF
by NIR using just the biceps underestimated % BF even more at 8.9%.
However, the % BF average for skinfolds was much closer at 13.7%; r= 0.94,
SEE: 2.9%. The authors summarized that when compared to the criterion of
UWW, the error of prediction for NIR could reach ~ 4%. Thus, the authors
concluded that NIR Should not be used as an alternative to skinfolds in body
composition analysis.

McLean and Skinner (1992) studied the body composition of 30 males
(aged 332:7 yr) and 31 females (aged 3116 yr) using UWW, NIR, and skinfolds
(Golding, Myers, 8: Sinning, 1982). UW was used as the criterion method.
Mean % BFS for all 61 subjects included UWW with 20.8% $8.2, NIR at 20.8%
21:55, and skinfolds at 22.0% :|:8.5. The correlation between UWW and NIR
was r= 0.81, SEE: 4.8%; while the correlation between UWW and skinfolds
was 0.94, SEE: 2.8%. Hence, skinfolds more accurately measured % BF than
NIR when compared to the UWW % BF criterion.

W

A major drawback to the ultrasonic and infrared interactance

approaches is the dependence upon regional adipose distribution to predict

total body fat (Lukaski, 1987). Also, as stated previously, NIR has been found

76

to be less accurate for obese individuals (Elia et al., 1990; Quatrochi et al., 1992)
compared to lean subjects. For example, Elia et al. (1990) found that NIR
underestimation of % BF (compared to UWW) increased as a function of
adiposity. Elia et al. (1990) also stated that this underestimation was marked
(16% body weight) in a small and separate group of grossly obese women,- BMI
> 50 kg / m2. McLean and Skinner (1992) noted that NIR overestimated % BF
in lean subjects with <8% BF and underestimated it in subjects with >30%
body fat.

Also, errors may be occur if the infrared wand is not adequately
shielded from extraneous light and the pressure exerted on the instrument
may inﬂuence the recorded spectra (Jebb 8: Elia, 1993). Other possible sources
of error may include subject factors such as Skin color and hydration status
(Elia et al., 1990; Heyward 8: Stolarczyk, 1996).

W

There has been little published research done with the NIR technique.
The studies that have been done indicate an extremely strong relationship (r=
0.99) between OD, and OD2 at the biceps Site, with only small difference
between average OD, and OD2 measurements (Israel et al., 1989). Hence, only
one OD typically enters in a prediction model. At present there are no NIR
equations based on multi-component model estimates of % BF (Heyward 8:

Stolarczyk, 1996).

77

Body Composition Literature Review Summary

Body composition has evolved with changes in technology and the
advance of scientiﬁc research. Original techniques (e.g., UWW) are constantly
being reﬁned as investigators strive to achieve more accurate measures of
body compartments. While UWW remains the popular criterion measure
among a majority of researchers, advanced techniques are being developed
with increased sensitivity to detect changes in body composition for wider
varieties of population groups.

Single photon absorptiometry was developed to quantify appendicular
bone mass. Precision and accuracy were both good but only certain Sites could
be measured versus whole body bone mineral content. Subsequently, DPA
was developed to measure total body bone mineral content. The dual photon
source also provided the ability to analyze fat mass and bone-free, fat-free
mass. This enabled clinicians and researchers to measure three components
of the body with DPA. However, the dual source was produced by 153
gadolinium, a radioactive compound. This isotope produced two
characteristic photon energies which were difﬁcult to control and standardize
over time. Thus, DPA had only average precision and accuracy.

Replacement of the radionuclide source with an x-ray source resulted
in the development of DXA. The evolution of DXA represented a major
advance in the ﬁeld of body composition analysis. Bone, fat, and fat-free,
bone-free tissue could be directly analyzed with high precision and accuracy.

This direct analysis of FFM by DXA offers a major advantage over UWW.

78

That is, the DXA technique eliminates the need to convert Db to % BF as is
the case with UWW (which compounds the measurement error in body
composition estimates).

Direct measurement of bone mass may be advantageous when
studying female athletes. Female athletes may have a wider than average
range of bone densities depending on menstrual history and training. The
UWW technique assumes a constant FFM; with bone mass remaining
constant as part of this assumption. Therefore, variations in bone density are
not taken into account with UWW, while they are accounted for by DXA.

Although DXA has been Shown to be both precise and accurate, this
instrument is expensive and limited to clinical settings. In addition, other
obstacles may include the lack of portability and the need for trained
technicians. Thus, body composition ﬁeld techniques are more desirable for
coaches, trainers, and medical staff. However, these ﬁeld techniques include
various measured variables that require the development of regression
equations to estimate body composition. Two commonly used ﬁeld
techniques include BIA and NIR. The validity of these techniques depends
on the appropriateness of the regression equation for the population being
tested. This equation should be based on a multi-component model to
account for individual differences.

Since female athletes may differ in their bone mineral content, DXA
appears to be an appropriate criterion method for this athletic population. In

the present investigation, DXA was used as the criterion method to compare

79

to body composition estimates via BIA and NIR. Specifically, stepwise

multiple regression was employed to develop separate equations (using the
double cross validation procedure) to predict FFM using BIA and NIR. The
purpose of this investigation was to determine the reliability and validity of

BIA and NIR for estimating body composition of female athletes.

80

CHAPTER 3
METHODS
Subjects

We invited athletes from all Michigan State University (MSU)
women's varsity sports to participate. These sports included basketball, crew,
cross country, ﬁeld hockey, golf, gymnastics, soccer, softball, swimming and
diving, tennis, track and ﬁeld, and volleyball. Athletes were recruited
through the MSU athletic department from each team's coaches and athletic
trainers. VanLoan and Mayclin (1992) have shown that one hundred to four
hundred athletes are needed to form an equation which would be
representative of this particular population.

The study was approved by the MSU committee on research involving
human subjects (UCRIHS) and written informed consent was obtained from
each participant. Although the medical staff expressed an interest in knowing
some of the study results, care was taken to ensure that the athletes
understood that their results would remain confidential unless they
informed us otherwise.

Instrumentation
malenergmrauhsmptiemetmtﬂm

The DXA instrument is a Hologic QDR-1000W, software 6.10. The
instrument works on the physics principle that as x-rays pass through the
body, the exiting attenuated signal is exponentially related to the path length,

tissue density, and energy of the x-ray. The x-ray source, mounted beneath

81

the patient, generates a narrow, tightly collimated beam of x-rays which pass
through the patient at rapidly switched x-ray voltage energies of 70 and 140
kVp. The transmitted intensity at each energy is measured by a radiation
detector mounted on a C-frame (movable arm) directly above the x-ray
source. During a scan, the C-arm oscillates rapidly in the transverse direction
while slowly moving longitudinally.

Participants lie supine on the scanning bed and were scanned from
head to toe in approximately 20 minutes. The instrument can be operated to
examine the whole body or any region of the body, including arms, legs, head,
and trunk. The area of the scanning table includes maximal values of 190.5
cm length and 63.0 cm width (Hologic Operator’s Manual, 1990). Line spacing
is the amount of scanning arm steps between each pass of the scanner arm,
and is ﬁxed at 1.303 cm for whole body scans. Resolution is the space between
samples along each line and is .205 cm.

As the x-ray beam is introduced into the body, the external detector
analyzes one small cross-sectional area (1 x 1 mm area) at a time. These small
cross-sectional areas are called pixels. For the analysis, each pixel in the image
is determined to be one of two components: bone or soft tissue. Spatial
threshold is the outline around bone and determines the mass of the bone
and tissue in adjacent areas.

Those pixels which do not contain bone are then re-analyzed as two
compartments containing lean and fat tissues. The distribution of the lean

and fat tissues overlying the pixels which contain bone is then estimated by

82

interpolation from the adjacent bone-free pixels. This is true of the
"enhanced" software, 6.10. Therefore, DXA divides the body into three
components: total body bone mineral, or ash; fat-free, mineral free soft tissue;
and fat.

BIA (RJL analyzer; Detroit, MI) induces a small current through the
body, providing resistance and reactance values (both measured in Ohms)
from bodily tissues. This current measures TBW which is indicative of FFM
(Hoffer et al., 1969). Individuals of a given height and weight will often
different resistance to the current, based on the fat content of their bodies
(Lukaski et al., 1985).

Procedures were performed based on methods described by Heyward
and Stolarczyk (1996) in a setting of ambient temperature (~25 °C). For the
BIA method, the athlete lay supine on a mattress, with the arms abducted 45
degrees from the legs (which are less abducted than the arms). All
measurements are performed on the right side of the body and skin electrode
sites were cleaned thoroughly with an alcohol pad.

The RJL analyzer was employed which delivers an alternating current
of 800 uA at a ﬁxed 50 kHz frequency (Heyward 8: Stolarczyk, 1996). Two

electrodes induce the current source. Placements for these current source
electrodes include: one electrode placed on the dorsal surface of the wrist so
that the upper border of the electrode bisects the head of the ulna (sensor

electrode, proximal); and the other electrode put on the dorsal surface of the

83

ankle so that the upper border of the electrode bisects the medial and lateral
malleoli. This current is detected by the sensor electrodes which are located
proxirnally to the source electrodes just described. The placement of the
sensor electrodes includes: the base of the second metacarpal and phalangeal
joints of the hand and foot. Also, there should be at least 5 cm between the
proximal and distal electrodes (Heyward 8: Stolarczyk, 1996).

The lead wires are attached to speciﬁc electrodes. Red lead wires are
attached to the wrist and ankle, with the red lead head attached to the wrist
and the black lead head attached to the hand. Black lead wires are attached to
the ankle and foot, with the red lead head attached to the ankle and the black
lead head attached to the foot. Once the subject is properly set up, current is
activated and then resistance and reactance values are recorded.
MW)

NIR (Futrex, Inc.; Gaithersburg, MD), involves a wand, emitting
infrared light, which is placed over the participant's biceps while She is seated.
The assumption is that the degree of infrared light absorbed and reﬂected is
related to both the composition of the tissues through which the light is being
passed and the Speciﬁc wavelength being emitted from the light (Conway et
al., 1984). The peak absorption wavelengths for pure fat and pure water are
930 nm and 970 nm, respectively. The Futrex-5000 emits two wavelengths,
940 nm and 950 nm from the wand and subsequently measures the amount

of light reﬂected by the underlying tissues as OD, and OD, (Futrex, 1988).

84

Measurements were made in accordance with the manufacturer’s
instructions (Futrex, 1990). Each subject was seated with her right arm
extended (palm up) and resting comfortably on a table. The NIR wand was
placed halfway between the antecubital fossa (of the elbow) and the acromion
(the armpit). The light wand was held perpendicular to the measurement
site. Caution was used to make sure that the shield ﬂaps were pressed to the
skin so that no light would penetrate and affect the optical density
measurements. The same amount of pressure was exerted on the wand for
each subject, since all measurements were performed by the same
investigator.

To obtain the optical density measurements of OD, and OD, the
”Clear” button was pushed, and the numbers 881 were entered into the NIR
instrument (Heyward 8: Stolarczyk, 1996). Also, before each NIR
measurement, the wand was calibrated with a teﬂon standard 1 cm thick.

Optical densities were determined by:

A OD, = OD standard- OD, subject

A OD, = OD standard- OD, subject

Data Collection Procedures
Testing occurred in the rheumatology clinic, located in the St.
Lawrence Health Science Pavilion (East Lansing, MI). On arriving, each
female was measured for body composition by each of the three methods:
DXA, BIA, and NIR. Trained investigators operated each station, and the

athletes rotated through the stations. First, each subject had her standing

85

height and weight measured using a pre-calibrated beam balance scale and
stadiometer. Next, each subject performed repeat BIA and NIR tests. The
order of tests was BIA, NIR, repeat BIA, repeat NIR. Finally, all subjects
underwent a Single DXA analysis. The DXA machine was calibrated daily to a
lumbar spinal phantom for bone density, and a tissue bar was also included in
every scan for soft tissue analysis of FFM.

C] l .

Athletes wore the same long t-shirt with a sports bra and shorts for all
methods of body composition except the DXA analysis. For the DXA
measurement, only the long t-shirt and sports bra were worn. It is important
that clothing be standardized and minimal for the DXA scan as this will
contribute to the attenuation (Jebb 8: Elia, 1993). Cotton, for example, has an
attenuation similar to that of fat (Jebb 8: Elia, 1993).

E I I G . , 1'

Each athlete was given a set of written guidelines to adhere to before
her designated testing date. These guidelines were distributed at either a team
meeting or in the mail. The guidelines included (Heyward 8: Stolarczyk,
1996):

1. No large meals 4 hours before the test

2. No heavy exercise 12 hours before the test

3. Urinate immediately before the test

4. No alcohol consumption 48 hours before the test

5. No diuretic medications 7 days before the test

86

6. Drink 1% of body weight, or ~2 eight ounce glasses of water, 2 hours
before the test
Statistical Analyses

Descriptive statistics (mean, standard deviation, and range) were
calculated for each varsity team studied. Variables included age (yr), height
(cm), weight (kg), body mass index (BMI), FFM (from DXA), and % BF (from
DXA). Those individuals who were athletes, yet not members of a varsity
team were included in a category termed "other". They were recruited via
personal communication with the investigator. Also, if less than 10 members
of a given varsity team were tested, then the subjects’ results were added to
the ”other” category. While team values were calculated for descriptive and
general comparison purposes, all reliability and validity analyses were
conducted on total sample as a whole.

B ,. , .,.

BIA and NIR analysis were performed twice on each subject.
Dependent variables included resistance (R) and reactance (X,) for the BIA
method (measured in ohms), and the two optical density values (OD, and
0D,) obtained with NIR (measured in run). Intraclass correlation coefﬁcients
were determined for each BIA and NIR variable, using repeated measures
analysis of variance (ANOVA) where R“. = (MSs - MSe) / MS,5 for a reliability
estimate given both trials, and Single = (MSs - MSe) / (MS, + M5,) for an
estimate given a single trial. MS, was the mean square for subjects, and MS,

was the mean square for error (Baumgartner 8: Jackson, 1987). Standard

87

errors of measurement were also calculated according to the formula SEM =
S,( ‘1 (1 - Ru.) where SEM = the standard error of measurement, Sx was the
standard deviation of the test measures, and K“. was the reliability coefﬁcient
for the test scores. The SEM values were calculated in both absolute and
relative (%) terms.

If ,. TI

The FFM value measured by DXA was used as the criterion measure.
Stepwise multiple regression was used to derive an equation for both BIA and
NIR estimates of FFM. Predictor variables entered into the BIA equation
were height, weight, R, and X,. For the NIR equation, predictor variables
included height, weight, 0D,, and 0D,. In all cases the R, X,, 0D,, and OD,
values used in the regression analysis represented the average of the repeat
test values.

Prediction equations for BIA and NIR were developed using a double
cross validation technique (Kerlinger 8: Pedhazur, 1973). That is, the subjects
were split into two samples (based on odd or even identiﬁcation numbers)
and an equation was developed for each, with the opposite group being used
to cross validate each equation. If the equations proved to be similar
(evaluated by comparison of multiple r values, and visible inspection of
graphs), groups were combined and a single equation was developed using
the entire sample.

In addition to correlation and regression techniques, error analysis was

also performed. Standard errors of estimate (SEE) were calculated (SEE = 5y *1]

88

- r2) and used as an error of prediction of DXA derived FFM using the BIA and
NIR estimates. Total error (TE) was also calculated and used as an indication
of standard deviation of error relative to the line of identity (TE = ‘1(2(y -

y’Y/n» (Lohman, 1981).

89

CHAPTER 4
RESULTS

A total 135 female athletes between the ages of 18-27 years were
recruited as subjects. The data of three subjects were dropped from the study;
two subjects were too tall for the DXA scanner and the other one was
suspected of having an eating disorder. Therefore, the ﬁnal subject total for
this study was 132 athletes. All but 14 subjects were varsity athletes currently
participating at MSU. The ”nonvarsity” athletes included four ice hockey
players from the MSU ice hockey club team and ten former competitive
athletes who were still physically active. Minimum activity for these
additional athletes included ﬁve or more days of vigorous activity per week,
30 minutes per bout of exercise.

The total number of athletes studied represented approximately 65% of
all members of women’s varsity teams during the length of the study. Very
few potential subjects (<10) who were told about the study actually refused to
participate. The major reason for nonparticipation was a time conﬂict with
the available testing schedule. Three teams, basketball, golf, and tennis were
not included due to time and scheduling conﬂicts. Eight subjects were Black,
one was Asian, and the rest were Caucasian.

Table 1 shows descriptive data for the subjects organized by teams. It
should be pointed out that diving (n=3), swimming (n=6), track (sprinters and
field events; n=8), volleyball (n=4) team data were merged with the ”other”

group due to small numbers. Also, cross country runners and track and ﬁeld

90

athletes who specialized in the 800 m run (or longer) were combined in a
group called ”distance runners”.

Reliability analysis is presented in Table 2. Reliability coefficients were
extremely high for all measures when analyzed for both multiple and single
trials. Speciﬁcally, BIA reliability ranged from Rxx = .987 - .997, while NIR
values ranged from Ru = .957 - .980. In addition to high reliability for both
multiple and single trials, SEM values were low for all variables mentioned.

To assess the validity of BIA and NIR techniques, the 132 subjects were
divided into two groups of 66, based on whether the subject identiﬁcation
number was odd or even. First, we calculated separate equations for BIA and
NIR with the even numbered subjects. Then cross-validation was performed
using the odd numbered subjects. Next, we did exactly the opposite,
formulating additional equations for BIA and NIR using the odd subjects as
the validation sample while the even subjects were used for cross-validation.
Due to a few numbers out of order, two ”odd” numbered subjects were
randomly chosen and added to the ”even” group. Thus, validity analysis was
initially performed using two groups (EVEN and ODD) of 66 subjects each.

Table 3 shows BIA and NIR equations derived from EVEN numbered
subjects. Also included in Table 3 are results of the cross validation sample
(ODD numbered subjects). That is, subjects in the ODD group were used to
predict FFM using the equation developed on the EVEN group subjects.
Table 4 shows the equations developed with the ODD group subjects, and

cross validated on their EVEN counterparts.

91

It is apparent from Tables 3 and 4 that the validity coefﬁcients (r
values), SEE, and TE were similar between EVEN and ODD samples, and that
the cross validations showed similar results. Figure 1a and 1b presents a
graphic representation of the closeness of the equations developed using the
EVEN vs. ODD subjects. It is apparent that the regression lines were virtually
identical, with deviation from the line of identities being similar for both
samples. Thus, a Single equation using all 132 subjects was developed for BIA
and NIR prediction of FFM. In addition, height and weight alone were used
to develop an equation. All three equations, validity coefﬁcients, and error

analyses can be seen in Table 5.

92

Figure 1a: FFM measured via DXA (kg) vs. FFM measured via BIA (kg)
Figure 1b: FFM measured via DXA (kg) vs. FFM measured via NIR (kg)
Legend- Even subjects are represented by hollow dots

Odd subjects are represented by ﬁlled dots

93

FFM measured via DXA (kg)

FFM measured via DXA (kg)

 

 

 

 

 

 

 

 

70
o O
60- 0’0 °
.0
00 °
0
50-
. 0
e O

40'- o,-

O
30 l ' l ' I '

3 0 4 O 5 O 6 0 7 O
FFM predicted via BIA (kg)
70
0
O
60- 000 °
00 . °

50' 0'0.

. O

O

O

40 d,

C
so . . . . . ,

3 0 4 0 5 O 6 0 7 0

FFM predicted via NIR (kg)

94

Table 1: Age and anthropometric subject data (N=132)

 

 

Sport Age Ht Wt BMI DXA Wt DXA FFM DXA % Fat
(yr) (cm) (kg) (kg) (kg)
Crew (n=22)
Mean 20.4 168.7 69.0 23.9 68.1 53.1 21.9
SD 1.9 7.4 7.8 1.6 7.7 5.4 23
Min 18.5 157.0 55.1 21.3 54.4 42.9 17.3
Max 27.2 182.0 90.9 28.1 89.5 64.4 28.1
Distance Runners (n=24)
Mean 20.3 165.1 57.3 21.0 56.4 46.0 18.3
SD 1.1 6.6 5.0 1.2 5.0 3.8 27
Min 18.6 153.0 50.0 19.1 49.2 39.9 13.2
Max 22.5 175.0 68.2 24.8 67.6 53.1 23.9
Field Hockey (n=10)
Mean 19.8 165.7 62.7 22.7 61.8 48.9 20.9
SD 1.2 7.5 10.6 24 10.6 8.5 4.1
Min 18.6 155.5 46.0 18.2 45.3 37.3 14.3
Max 21.6 180.0 81.2 26.8 80.6 64.2 27.3
Gymnastics (n=15)
Mean 19.8 157.7 57.6 23.1 56.8 45.9 19.1
SD 1.0 5.6 7.2 1.7 7.1 5.4 2.2
Min 18.5 146.5 44.5 20.4 43.8 36.4 15.9
Max 21.6 166.0 68.8 25.9 68.0 54.7 24.2
Soccer (n=10)
Mean 19.8 168.2 65.4 23.1 64.5 50.4 21.8
SD 0.9 4.9 7.7 1.7 7.7 5.8 27
Min 18.5 160.0 51.3 20.0 50.6 41.3 18.4
Max 21.3 177.5 81.2 25.8 80.2 61.8 26.3
Softball (n=17)
Mean 20.4 170.2 66.8 23.0 66.4 52.4 20.9
SD 1.4 6.6 10.0 3.0 9.0 5.9 3.9
Min 18.1 160.0 46.8 17.6 54.6 44.2 15.5
Max 23.8 181.0 89.5 30.3 88.5 66.6 27.3
Other (n=34)
Mean 21.1 166.8 63.4 22.7 62.5 49.6 20.5
SD 1.9 7.6 7.5 1.8 7.4 5.4 27
Min 18.5 152.0 49.0 19.4 48.3 38.3 14.7
Max 25.6 187.0 79.4 27.8 78.6 60.5 26.1
Total (N =132)
Mean 20.4 166.4 63.1 22.7 62.3 49.5 20.4
SD 1.5 7.6 8.7 21 8.5 6.0 3.1
Min 18.1 146.5 44.5 17.6 43.8 36.4 13.2
Max 27.1 187.0 90.9 30.3 89.5 66.6 28.1

95

 

 

Table 2: Reliability estimates for BIA and NIR procedures (N=132)

Variable &, SEM SEM% 511g SEM SEM%

R .997 3.8 0.7 .994 5.4 1.0

(ohms)

x. .993 0.8 1.3 .987 1.1 1.8

(ohms)

OD] .978 0.016 19.5 .957 0.022 27.1

(run)

OD2 .980 0.017 21.3 .961 0.024 29.6

(nm)

BIA = bioelectrical impedance analysis using resistance (R) and
reactance (Xe) measures

NIR: near-infrared interactance using two optical density (OD)
measures

R“: the intraclass correlation coefﬁcient for reliability of multiple
trials

Single: the reliability estimate of a single trial

SEM: standard error of measurement

96

Table 3: Prediction Equations for FFM using EVEN numbered subjects (n=66)

 

DXA (measured FFM = 49.5163 kg)
BIA
FFM = (0.272 * ht) + (0.461 * wt) - (0.036 * R) + (0.101 * Xe) - 11.567
predicted FFM = 49.6162 kg r = 0.964 SEE = 1.2 TE = 1.2
(cross validation using ODD numbered subjects)
predicted FFM = 49.3i5.6 r = 0.983 SEE = 1.1 TB = 1.2
N IR
FFM = (0.065 " ht) + (0.72 * wt) - (14.334 * OD2) - 5.673
predicted FFM = 49.5:t6.2 kg r = 0.984 SEE = 1.1 TE = 1.1

(cross validation using ODD numbered subjects)

 

predicted FFM = 49.6i5.5 r = 0.965 SEE = 1.5 TE = 1.7

FFM = fat free mass

BIA = bioelectrical impedance analysis using resistance (R) and
reactance (Xc) measures

N IR = near-infrared interactance using two optical density (OD)
measures

r = validity coefﬁcient

SEE = standard error of estimate

TE = total error

97

Table 4: Prediction Equations for FFM using ODD numbered subjects (n=66)

 

DXA (measured FFM = 49515.8 kg)
BIA
FFM = (0.284 * ht) + (0.38 " wt) - (0.037 " R) + (0.096 " Xc) - 7.85

predicted FFM = 49.4157 kg r = 0.983 SEE = 1.1 TB = 1.1
(cross validation using EVEN numbered subjects)

predicted FFM = 49.6163 kg r = 0.982 SEE = 1.2 TE = 1.3
NIR
FFM = (0.015 " ht) + (0.565 * wt) - (9.549 * 0D,) - 10.346

predicted FFM = 49.4156 kg r = 0.969 SEE = 1.5 TE = 1.4
(cross validation using EVEN numbered subjects)

predicted FFM = 493163 kg r = 0.981 SEE = 1.2 TB = 1.4

 

 

FFM = fat free mass

BIA = bioelectrical impedance analysis using resistance (R) and
reactance (Xc) measures

NIR - near-infrared interactance using two optical density (OD)
measures

r = validity coefﬁcient

SEE = standard error of estimate

TE = total error

98

Table 5: Prediction Equations for FFM using ALL subjects (N=132)

 

DXA (measured FFM = 49516.0 kg)
BIA
FFM = (0.282 * ht) + (0.415 * wt) - (0.037 * R) + (0.096 * Xc) - 9.734
predicted FFM = 49.4159 kg r = 0.981 SEE = 1.1 TB = 1.1
N IR
FFM = (0.111 " ht) + (0.641 " wt) - (12.397 * 0D,) - 8.423
predicted FFM = 49515.8 kg r = 0.975 SEE = 1.2 TE = 1.2
Height and Weight
FFM = (0.143 * ht) + (0.565 " wt) - 10.03

predicted FFM = 49.4157 kg r = 0.961 SEE = 1.6 TE = 1.6

 

FFM = fat free mass

BIA = bioelectrical impedance analysis using resistance (R) and
reactance (Xc) measures

NIR = near-infrared interactance using two optical density (OD)
measures

r = validity coefﬁcient

SEE = standard error of estimate

TE = total error

99

CHAPTER 5
DISCUSSION

Accurate body composition assessment is beneﬁcial for female athletes,
as it can be indicative of health and performance status. Our data provide
measured (DXA) and predicted (BIA and NIR) FFM and % BFs for the
individual sports of crew, distance runners, ﬁeld hockey, gymnastics, soccer,
softball, and other female athletes. The study purpose was to determine
whether the BIA and NIR ﬁeld techniques could provide both reliable and
valid estimates of body composition in this population.

In this study, both BIA and NIR techniques were used to predict FFM as
compared to a value obtained from a criterion measurement instrument
(DXA) that can separate fat free bone and soft tissue from fat tissue. In theory,
BIA measures the conductivity of TBW and electrolytes which are found
exclusively in the FFM. The NIR is designed to differentiate between the
water and fat contents of the various body compartments by the amount of
light that is absorbed and reﬂected from an infrared source. Both TBW and
electrolytes have been shown to be highly correlated to FFM (Kushner, 1992).

While FFM may be the appropriate variable to be measured and
predicted in this investigation, most body composition studies in the
literature have centered around an individual’s % BF. This allows a more
logical point of comparison when discussing differences among groups that

differ widely in absolute body mass.

100

As can be seen in Table 1, there was little variability in either FFM or %
BF measures when making comparisons either between and within our
different sport teams. As a group, distance runners had the lowest values
(FFM=46.013.8 kg; %BF=18.312.7%) and crew had the highest (FFM=53.115.4
kg; %BF=219123%). On the whole, FFM measured by DXA averaged 49516.0
kg and %BF averaged 20.4131 % in the 132 athletes we tested. Indeed, even
the 14 ”other” non-varsity athletes had similar values, supporting their
inclusion in our sample.

Many body composition studies have been performed with female
athletes. For instance, elite runners (training distance of 6451153
miles / week) have been measured at 14.3% with UWW (Graves, Pollock, 8:
Sparling, 1987). Picard, Kyle, Gremion, Gerbase, and Slosman (1997) used
DXA to determine body composition in elite runners (mileage values of
8216.4 km/ week) and reported average BF was 14.8%. The athletes in these
two studies had somewhat lower % BFs than our distance runners who
averaged 18.312.7%. This difference may be due to lower mileage run by
college distance runners versus highly elite runners who are more seasoned
and typically have more rigorous training schedules.

Withers et a1. (1987) investigated the % BFs of female soccer and
softball players using UWW. Their results showed an average of 22% for
soccer players and 19.1% for softball athletes. These values are consistent with
our data, which showed BF values of 21.8% for soccer players and 209% for

softball.

101

 

The body composition of gymnasts has been studied also. Sinning
(1978) found a mean value of 15.3% in champion gymnasts using UWW. In
contrast, Kirchner et al. (1995) found an average value of 17.0% for collegiate
gymnasts. Our results agree more with the Kirchner et al. study for our
gymnasts averaged 19.1 %BF. A possible reason for the difference in % BFs
may include that Sinning’s gymnasts were all collegiate champions, and were
taller and lighter than our sample.

Sinning and Wilson (1984) measured body composition (UWW and
skinfolds) in a group of 79 women who were similar to our overall study
sample. Their subjects included women who participated in a variety of
intercollegiate sports at Kent State University. The average % BF for these
athletes was 20.1%, while the average % BF for athletes in our study was
20.4%.

Although most other studies have utilized UWW as the criterion,
DXA has been shown to correlate well with UWW for eumenorrheic female
subjects (Hansen et al, 1993). Also, given the BMC assessment included in all
DXA body composition estimates, it is likely to be a more widely used
criterion measure than UWW in the future (Heyward 8: Solarczyk, 1996).
Further contributing to DXA’S validity, studies have shown high correlations
between scale weight and DXA (Going et al., 1993; Prior et al., 1997) and also
DXA and a four component model (Prior et al., 1997). It is of interest to note
that in our study, DXA values for weight were highly correlated with (r=0.999)

and only slightly above (<1%) scale weight values as seen in Figure 2.

102

 

 

 

 

 

100
90-
g 80-
I- ,-
3 70- ‘3
LLI
2‘
o 60
CD
50-_
40 r I ' I ' I ' I j 1 '
40 50 60 70 80 90 100

DXA wr (kg)

Figure 2: Scale weight (kg) vs. DXA weight (kg)
This overestimation was small and should not have contributed signiﬁcant
error to our body composition estimates.

DXA is believed to be a more sensitive measurement technique for
body composition than UWW because it is theoretically independent of
compartrnental assumptions (Lukaski, 1987). However, DXA is an expensive
device (~$60,000) and is not readily available to trainers, coaches, or even
most medical staff. Therefore, other methods are frequently utilized in ﬁeld

settings. These techniques, if shown to be reliable and valid, are more

103

 

desirable due to their reduced cost, portability, and little operational training
required.

The ﬁeld techniques employed in this study include BIA and NIR. The
reliabilities of BIA and NIR variables were assessed by calculating multiple
and single trial reliability coefﬁcients. Single trial reliability for BIA was .high
at 0.994 and 0.987 for R and X,, respectively. Also, the Single trial reliability for
NIR was also high at values of 0.957 and 0.961 for OD, and 0D,, respectively.
Previous investigations of the reliability of R in BIA studies have shown
similar results. Some of these include: 0.957 for men and 0.967 for women
(Jackson et al., 1988); 0.99 for men (Lukaski et al., 1985); and 0.907 for women
and 0.946 for men (Segal et al., 1988). Limited research has been performed
investigating the reliability of OD measures used with NIR. However, studies
have Shown reliability for the biceps site to be 0.96 (Quatrochi et al., 1992) and
0.97 (Hortobagyi et al., 1992). These values are similar to the values found in
the present study.

The high Single reliability coefﬁcients found in this study indicate that
only a Single trial is necessary when utilizing the ﬁeld techniques of BIA and
NIR. This is important for the busy sports training staff faced with caring for
hundreds of athletes on a daily basis. However, it is important to note that
the high reliability coefﬁcients obtained in this study were likely a function of
the investigator paying close attention to manufacturers’ instructions

regarding measurement techniques.

104

 

The accuracy of results in these ﬁeld techniques depends on specific
equations developed for each technique. Our equation for BIA predicted FFM
and used height, weight, resistance, and reactance. Our NIR equation also
predicted FFM and included the variables of height, weight, and 00,. OD,
was also a high predictor of FFM, but Slightly less than, and also highly
correlated with, 0D,. Consequently, it was not entered into the equation.
Equations developed for specific populations (such as a similar age group and
activity level) improve the accuracy of predicting body composition in a
comparative population. Additionally, cross-validation of these equations is
important to test for accuracy.

To illustrate the point discussed above, Sinning and Wilson (1984)
studied the validity of body composition analysis using several skinfold
equations in a group of females athletes (n=79), age 17.8 to 22.5 yr. Only two of
the nine equations tested were found to be in agreement with values obtained
via UWW. The authors pointed out that the failure of all of these equations
to be acceptable emphasized the need for further cross-validation studies.

In another study, Pichard et al., (1997) studied the relationship between
FFM and FM calculated in elite female runners with 12 different BIA
formulas reported in the literature. The criterion measure was DXA (Hologic
QDR-2000, software 5.54). There were 17 elite runners included in the study,
age 15-39 yr, and 17 female control subjects who were age and height matched.
The authors discovered that those equations that performed well in the

controls gave poor results in the elite runners. Likewise, the formulas that

105

were more predictive in elite runners were less accurate with control subjects.
This suggests that the subjects represented two distinct groups that must be
evaluated with population specific prediction equations.

In our study, we tested 132 female athletes. All had equivalently high
levels of physical activity which included daily (or twice daily) sessions of
weight training and aerobic conditioning in both the on- and off-season.

Tables 3 and 4 Show that all correlations were extremely high, and
similar for both BIA (r=0.964-0983) and NIR (r=0.956-0954) whether
validation or cross validation samples were used. Similarly, SEE and TE
values were very low and consistent in all cases (1.1-1.4 kg). Additionally, a
graph comparing the two equations (Figure 1) indicated virtually identical
lines for both BIA and NIR. This graph added further support for combining
both samples and developing a single equation on all 132 subjects. Not
surprisingly, the single sample regression lines for BIA and NIR (Table 5)
were nearly identical to all others developed from the split sample.

Pichard et al. (1997) used 12 BIA equations, including two provided by
the manufacturer (RJL Systems; Detroit, MI) and nine other equations
commonly found in the literature. The best equation for elite runners was
found to be RJL Systems-2 equation which had FFM results of r= 0.90, SEE: 2.0
kg, TE: 2.1 kg. We tested this equation with our data (N=132) which resulted
in an average FFM of 48.7 kg, r: 0.94 SEE: 2.0 kg, TE: 2.7 kg. While the

correlation using the RJL equation on our subjects was high, FFM was

106

underpredicted and error terms were nearly twice as high as those in our BIA
equaﬁon.

As previously mentioned, only two published BIA equations exist
which have Speciﬁcally included female athletes in the validation samples
(Houtkooper et al., 1989; Lukaski 8: Bolonchuk, 1987). Again, we tested both
equations using our subject data. Results for the Houtkooper et a1. (1989)
equation included: average FFM: 49.3 kg, r= 0.935, SEE: 2.0 kg, and TB: 2.2.
Also, results for Lukaski 8: Bolonchuk (1987) were: average FFM: 46.8 kg, r:
0.94, SEE: 2.0 kg, and TB: 3.4 kg. As was the case with the RJL equation, our
equations had slightly higher correlation values, and lower SEE and TE. One
possible explanation for these differences may be that our study utilized
female athletes exclusively while the other equations were developed using
both active and inactive female subjects.

As stated earlier, no published research study has formulated an
equation for female athletes using NIR to assess body composition. Authors
of other NIR studies suggested not to use the Futrex formulas provided by the
manufacturer (Heyward 8: Stolarczyk, 1996). Thus, we are the ﬁrst to develop
an equation for female athletes using NIR. From our data, the NIR technique
was Shown to be as valid as BIA for predicting FFM in this population.

Some investigators have questioned the ability of ﬁeld techniques such
as BIA and NIR to add significantly to body composition estimates.
Speciﬁcally, how much do the R, X,, and OD values add to height and weight

when developing appropriate prediction equations? To address this issue, we

107

developed an equation based only on the heights and weights of our athletes
(Table 5). Our results showed a predicted FFM value of 49.4 kg, r= 0.961, SEE:
1.7, and TE: 1.6. Given this finding, the results for BIA and NIR may appear
less impressive. The predictive ability of this height and weight regression
equation may be more of a function of the homogeneity of our population
rather than an inadequacy of the ﬁeld techniques. However, this same
homogeneity likely attributed to the accuracy of our BIA and NIR equations
as well.

The following example utilizing TE analysis may put the issue utility
of BIA or NIR vs. height and weight into perspective. For example, DXA
values averaged 62.2 kg for weight, 49.5 kg for FFM, and 20.4% for BF. The
BIA equation predicted a FFM of 49.4 kg, with TB: 1.1. The NIR equation
yielded a FFM value of 49.5 kg, with TB: 1.2 kg. Finally, the height and
weight equation predicts FM to be 49.4 kg, with TB = 1.7. All three
predictions appear very similar. However, by converting FFM predictions
into % BF values, further comparisons can be made. For example, FFM from
BIA ranges from 48.3-50.5 kg, which corresponds to a % BF of 22.3-18.8.
Compared to DXA values, the TB (deviation around the line of identity) for
BIA % BF was 19%. Similarly, NIR FFM is 48.3-50.7, with % BF 22.3-18.5.
Thus, NIR results in a 1.8% BF error. Lastly, height and weight predict FM
to be 47.7-51.1, which corresponds to a % BF 23.3-17.8. Hence, height and
weight yields a 29% error for % BF (~33% more error than BIA and NIR

equation).

108

One might argue that an increase in predictive accuracy for BF of only
1% is not worth the trouble and expense of using techniques such as BIA and
NIR. Indeed, each apparatus may cost from $2,000 - $3,000 or more. However,
the reality is that these techniques are being used all over the US. and the
world. Practically speaking, it may be difficult to convince coaches and
medical staff that this loss of accuracy is balanced by the money and time
saved. This is a decision that will certainly involve more study to completely
resolve. In the meantime, our data show that the BIA and NIR techniques
are extremely reliable and valid for estimating body composition values in
collegiate female athletes. The equations developed in this study were shown
to be slightly better predictors than using height and weight alone.

Recommendations for Future Studies

H

. Measure body composition with the additional method of UWW as a

further validation for DXA.

N

. Measure changes in body composition with DXA, BIA, and NIR. For
example, before and after a sports season, weight loss, or maturation (i.e.
freshman to senior years in college).

3. Recruit female athletes from other sports teams not included in this study.

4. Cross-validate the BIA and NIR equations on an athletic population in

other sites.

5. Perform additional reliability and validity studies with NIR.

6. Evaluate individual differences in body composition based on menstrual

patterns, type of exercise, and diet.

109

CHAPTER 6

REFERENCES

Abrahamsen B, Hansen TB, Hogsberg IM, Pedersen FB, 8: Beck-Nielsen
H. (1996). Impact of hemodialysis on dual energy x -ray absorptiometry,
bioelectrical impedance measurements, and anthropometry. American

IoumaLofﬁhmgaanmnﬂ: 80-86
Baumgartner TA 8: Jackson AS. (1987). Reliability and objectivity.

u - "111 1 - -o. 11.1 1 1. 1101111. ,1' ‘ (3rd

ed., pp 87-99). Dubuque, Iowa. Wm. C. Brown Publishers.

Bergsma-Kadijk JA, Baumeister B, 8: Deurenberg P. (1996).
Measurement of body fat in young and elderly women: comparison between a
four-component model and widely used reference methods. Bziﬂshjmimal

W 649-657

Blake GM, McKeeney DB, Chaya SC, Ryan P], 8: Fogelman1(1992).
Dual energy x-ray absorptiometry: the effects of beam hardening on bone

density measurements. W: 459-465.

Boddy C, King PC, 8: Hume R. (1972). The relation of total body
potassium to height, weight, and age in normal adults. mm
W 512-517

Brozek J, Grande R, Anderson IT, 8: Kemp A. (1963). Densitometric
analysis of body composition: revision of some quantitative assumptions.

mmmmmmmmum: 113-140.

Brozek I, Kihlberg IK, Taylor HL, 8: Keys A. (1963). Skinfold
distributions in middle-aged American men: a contribution to norms of

leanness-fatness. NW: 492-502.

Brunton JA, Bayley HS, 8: Atkinson SA. (1993). Validation and
application of dual-energy x-ray absorptiometry to measure bone mass and

body composition in small infants W
58: 839-845.

Bunt IC, Going SB, Lohman TG, Heinrich CH, Perry CD, 8: Pamenter
RW. (1990). Variation in bone mineral content and estimated fat 1n young

adult females W: 564-569

110

Chumlea WC, Baumgartner RN, 8: Roche AF. (1988). Specific
resistivity used to estimate fat-free mass from segmental body measures of

bioelectric impedance WWW: 7-15

Chumlea WC, Guo SS, Cockram DB, 8: Siervogel RM. (1996).
Mechanical and physiologic modifiers and bioelectrical impedance spectrum

determinants of body composition AmericanJoumaLoLQlinicaLNntrition
Mb. 41354225.

Clarys JP, Martin AD, 8: Drinkwater DT. (1984). Gross tissue weights in
the human body by cadaver dissection. W: 459-473.

Conway IM 8: Norris KH. (1987). Noninvasive body composition in
humans by near infrared interactance. In Ellis K], Yasumura S, 8: Morgan

WD (Eds), WW?- 163-170)- London: The

Institute of Physical Sciences in Medicine.

Conway IM, Norris KH, 8: Bodwell CE. (1984). A new approach for the
estimation of body composition: Infrared interactance. AmericanJQJimaLQf
CliniQaLNntIiﬁQnAQ 1123-1130

Cullum ID, Ell P], 8: Ryder IP. (1989). X-ray dual-photon
absorptiometry: a new method for the measurement of bone density. British

JournamfRadinogLéZ: 587-592.

Dunn WL, Kan SH, 8: Wahner HW. (1987). Errors in longitudinal
measurements of bone mineral: effect of source strength in single and dual

photon absorptiometry. W: 1751-1757.

Durnin 8: Satwanti. (1982). Variations in the assessment of the fat
content of the human body due to experimental technique in measuring body

density. AnnalLQLHumanjiologLQ: 221-225.

Elia M, Parkinson SA, 8: Diaz E. (1990). Evaluation of near infra-red
interactance as a method for predicting body composition. Enrgpganjmirnal
W‘ 113-121

Ellis K], Shypailo R], Pratt], 8: Pond WC. (1994). Accuracy of dual-
energy x-ray absorptiometry for body composition measurements in children.

Elsen R, Siu ML, Pineda O, 8: Solomons NW. (1987). Sources of
variability in bioelectrical determinations in adults. In Ellis K], Yasamura S,

& Morgan WD (Eds), MW (pp- 184-188)-

London: The Institute of Physical Sciences in Medicine.

111

Friedl KE, DeLuca IP, Marchitelli LT, 8: Vogel IA. (1992). Reliability of
body fat estimations from a four-component model by using density body
water, and bone mineral measurements.

W: 764-770

Fruth SI 8: Worrel TW. (1995). Factors associated with menstrual
irregularity and decreased bone mineral density in female athletes. JournaLQf

W112: 26-37.

Fuller NI, Laskey MA, 8: Elia M. (1992). Assessment of the
composition of major body regions by dual-energy x-ray absorptiometry

(DEXA), with special reference to limb muscle mass.
253-266.

Futrex, Inc. (1988). mm. Gaithersburg, MD.

Futrex, Inc. (1990). -
W (PP 1.-32) Gaithersburg, MD.

Gleichauf CN 8: Roe DA. (1989). The menstrual cycle’ 5 effect on the
reliability of bioirnpedance measurements for assessing body composition.

AmericaanumalnfﬂinicaLNmmionjﬂz 903-907

Gluer CC, Steiger P, Selvidge R, Elliesen-Kliefoth K, Hayashi C, 8:
Genant HK. (1990). Comparative assessment of dual-photon absorptiometry

and dual-energy radiography. W: 223-228.

Going SB, Massett MP, Hall MC, Bare LA, Root PA, Williams DP, 8:
Lohman TC. (1993). Detection of small changes 1n body composition by dual-

energy x-ray absorptiometry. W: 845-
850.

Golding LA, Myers CR, 8: Sinning WE. (1982). W9
ﬁtness. Chicago: National Board of the YMCA.

Gotfredsen A, Jensen], Borg J, 8: Christiansen C. (1986). Measurement
of lean body mass and total body fat using dual photon absorptiometry.
W: 88-93.

Graves IE, Pollock ML, 8: Sparling PB. (1987). Body composition of
elite female distance runners. W: 96-
102.

Guduvaka R, Schoeller D, 8: Kushner RF. (1996). Effect of skin
temperature on multifrequency bioelectrical impedance analysis. 15mm

Appliedﬁhxsinmlz 838-845

112

 

Guo S, Roche AF, 8: Houtkooper L. (1989). Fat-free mass in children
and young adults predicted from bioelectric impedance and anthropometric

variables. WWW: 435-443.

Haarbo I Gotfredsen A, Hassager C, 8: Christiansen C. (1991).
Validation of body composition by dual energy x-ray absorptiometry (DEXA).

Cl'uucaLEhxsinongl: 331-341

Hansen NI, Lohman TG, Going SB, Hall MC, Pamenter RW, Bare LA,
Boyden TW, 8: Houtkooper LB. (1993). Prediction of body composition in
premenopausal females from dual-energy x-ray absorptiometry. mm

W: 1637-1641

Heyward VH 8: Stolarczyk LM. (1996). W
Assessment (pp. 1-215). Champaign, IL: Human Kinetics.

Heymsfield SB, Wang F, Heshka S, Kehayias II, 8: Pierson RN. (1989).
Dual-photon absorptiometry: comparison of bone mineral and soft tissue
mass measurements in vivo with established methods. AmericanJQnmalgf
W112 1283-1289

Hoffer EC, Meador CK, 8: Simpson DC. (1969). Correlation of whole
body impedance with total body water. MW: 531-

Hologic Operator’ 5 Manual. (1990). W
WW: Waltham, MA: Hologic Inc

Horber FF, Thomi F, Casez IP, Fonteille I, 8: Jaeger PH. (1992). Impact
of hydration status on body composition as measured by dual energy x-ray
absorptiometry in normal volunteers and patients on haemodialysis. British

humaLQLRadinchﬁ: 895-900.

Hortobagyi T, Israel RG, Houmard IA, McCammon MR, 8: O’Brien KF.
(1992). Comparison of body composition assessment by hydrodensitometry,
skinfolds, and multiple site near-infrared spectrophotometry. Eiierean

IoumaLofﬂinimJSutritiQnAﬁ: 205-211

Houtkooper LB 8: Going SB. (1994). Body composition: how should it
be measured? Does it affect sport performance? WW).

Houtkooper LB, Going SB, Westfall CH, 8: Lohman TG. (1989).
Prediction of fat-free body corrected for bone mass from impedance and

anthropometry in adult females WW
2_1_:1S39 (Abstract).

113

 

Israel RG, Houmard IA, O’Brien KF, McCammon MR, Zamora BS, 8:
Eaton AW. (1989). Validity of a near-infrared spectrophotometry device for
estimation human body composition. x

ammo: 379-383.

Jackson AS 8: Pollock ML. (1978). Generalized equations for predicting
body density In men MW: 497-504

Jackson AS, Pollock ML, Graves IE, 8: Mahar MT. (1988). Reliability
and validity of bioelectrical impedance in determining body composition.

Imunalﬂpphedlhymlogyﬂ: 529-534

Iebb SA 8: Elia M. (1993). Techniques for the measurement of body
composition: a practical guide. InternatignaleirnaLefﬂhesingZ: 611-621.

Iebb SA, Goldberg GR, Jennings G, 8: Elia M. (1995).Dua1-energy x-ray
absorptiometry measurements of body composition: effects of depth and
tissue thickness, including comparisons with direct analysis. ClinigaLSeienee
88: 319-324.

Kelly TL, Slovik DM,Schoenfe1d DA, 8: Neer RM. (1988). Quantitative
digital radiography versus dual photon absorptiometry of the lumbar spine.

lonmalnLCJmaLEndocmologmiMetahohsmﬂ: 839-844.

Kerlinger EN 8: Pedhazur EJ. (1973). Explanation and Prediction.
(pp. 281-333). New York: Holt,
Rinehart and Winston, Inc.

Keys A 8: Brozek] (1953). Body fat in adult man. Rhysieiegied
M: 240-325.

Kirchner EM, Lewis RD, 8: O’Connor PI. (1995). Bone mineral density
and dietary intake of female college gymnasts. Wm
W22: 543-549.

Kushner RF. (1992). Bioelectrical impedance analysis. a review of

principles and applications. Amerigmﬁgliegummm: 199-209.

Kushner RF 8: Schoeller DA. (1986). Estimation of total body water in

bioelectrical impedance analysis AmencaanumaLoLChnicaUSumnonﬁ:
417-424.

114

 

Laskey MA, Lyttle KD, Flaxman ME, 8: Barber RW. (1992). The
inﬂuence of tissue depth and composition on the performance of the Lunar
dual-energy x-ray absoptiometer whole-body seaming mode. Bummer;

:.39-45

Lindsey R, Fey C. 8: Haboubi A. (1987). Dual photon absorptiometric
measurements of bone mineral density increase with source life. Caieiﬁed

W: 293-294

Lloyd T, Buchanan IR, Bitzer S, Waldman CI, Myers C, 8: Ford BC.
(1987). Interrelationships of diet, athletic activity, menstrual status, and bone

density 1n collegiate women W: 681-
684.

Lohman TG. (1981). Skinfolds and body density and their relation to
body fatness: a review. W: 181-225.

Lohman TG. (1984). Research progress in validation of laboratory

methods of assessing body composition. W5
Exercisejﬁ: 596-603.

LohmanTG. (1986). Applicability of body composition techniques and

constants for children and youths WWW:
325-357.

Lohman TG. (1992). Current 1ssues in exercise science series.

AdxanceLmJQdLQQmpositionAssessmem Monograph No 3 Champaign,
IL: Human Kinetics.

Lohman TG. (1996). Dual energy x-ray absorptiometry. In Roche AF,

Heymsfield SB, and Lohman TG (Eds), Humanlodxﬁmposiﬁon (pp. 63-78).
Champaign, IL: Human Kinetics.

Lohman TG, Boileau RA, 8: Slaughter MH. (1984). Body composition
in children and youth. In Boileau RA (Ed)
5eienees (pp. 29-57). Champaign, IL: Human Kinetics.

Lukaski HC. (1987). Methods for the assessment of human body

composition: traditional and new WW6:
537-556.

Lukaski HC. (1993). Soft tissue composition and bone mineral status.

evaluation by dual-energy x-ray absorptiometry.IoJirna_Qf_Nyrmrm1428:
438-443.

115

 

Lukaski HC 8: Bolonchuk WW. (1987). Theory and validation of the
tetrapolar bioelectrical impedance method to assess human body
composition. In Ellis KI, Yasumura S, and Morgan WD (Eds. ), My

Cgmmsiﬂmﬁmdies (pp. 410-414). London: The Institute of Physical
Sciences in Medicine.

Lukaski HC, Bolonchuk WW, Siders WA, 8: Hall CB. (1990). Body
composition assessment of athletes using bioelectrical impedance

measurements WWW: 434-440

Lukaski HC, Iohnson PE, Bolonchuk WW, 8: Lykken GI. (1985).
Assessment of fat-free mass using bioelectrical impedance measurements of

the human body W: 810-817

Madsen KL, Adams WC, 8: Van Loan MD. (1998). Effects of physical
activity, body weight and composition, and muscular strength on bone

density in young women WWW:
114-120.

Martin AD, Drinkwater DT, Clarys IP, 8: Ross WD. (1986). The
inconsistency of the fat-free mass: a reappraisal with implications for

densitometry. MW (pp. 92-97). Cambridge: University

Press.

Mazess RB. ( 1971) Estimation of bone and skeletal weight by direct
photon absorptiometry. W: 52-60.

Mazess RB. (1991). Do bioimpedance changes reﬂect weight, not
composition? WW: 178-179

Mazess RB, Barden HS, Bisek IP, 8: Hansen]. (1990). Dual-energy x-ray
absorptiometry for total-body and regional bone-mineral and soft-tissue

composition AmmcanJoumaLQLClinicalmmtionil: 1106-1112

Mazess R, Collick B, Trempe I, Barden H, 8: Hansen I (1989).
Performance evaluation of a dual-energy x-ray bone densitometer. Caleiﬁed

W134: 228-232.

Mazess RB, Judy PF, Wilson CR, 8: Cameron IR. (1973). Progress in
clinical use of photon absorptiometry. In Frame B, Parfitt AM, and Duncan

HM (Eds ) ChumalAspectmLMetabohcﬁoneﬂsease (PP 3743)

Amsterdam: Excerpta Medica.

Mazess RB, Peppler WW, 8: Gibbons M. (1984). Total body

composition by dual-photon (153 Gd) absorptiometry. W
W: 834-839.

116

McArdle WD, Katch F1, 8: Katch VL. (1996).

Lemiseﬁhysiologia
W (4th ed) Williams 8: Wilkins
Baltimore, MD.

McLean KP 8: Skinner IS. (1992). Validity of Futrex-5000 for body

composition determination WWW:
253-258.

Morrison IA, Khoury PR, Chumlea C, Specker B, Campaigne BN, 8:
Guo SS. (1994). Body composition measures from underwater weighing and
dual energy x-ray absorptiometry 1n black and white girls: a comparative

study AmericanlnumaLnLHumamﬂiologLé: 481-490

Nattiv A, Agostini R, Drinkwater B, 8: Yeager KK. (1991). The female
athlete triad: the inter-relatedness of disordered eating, amenorrhea, and

osteoporosis Clinicsmipmtsmedicinﬂﬁ 405-418

Pace N 8: Rathbum EN. (1945). Studies on body composition, III: the
body water and chemically combined nitrogen content in relation to fat

content humalntﬁmlcgmlﬁhemimlﬁ: 685-691

Pacifici R, Rupich R, Vered I, Fischer KC, Griffin M, Susman N, 8:
Avioli LV. (1988). Dual energy radiography (DER): a preliminary

comparative study CalciﬁeiIissueIntemationaLAa: 189-191

Peppler WW 8: Mazess RB. (1981). Total body bone mineral and lean

body mass by dual photon absorptiometry. W:
353-359.

Picaud IC, RigoJ, Nyamugabo K, MiletI, 8: Senterre]. (1996).
Evaluation of dual-energy x-ray absorptiometry for body composition
assessment in piglets and term human neonates. W
W: 157-163

Pichard C, Kyle UG, Grernion G, Gerbase M, 8: Slosman DO. (1997).
Body composition by x-ray absorptiometry and bioelectrical impedance in
female runners W2: 1527-1534

Pietrobelli A, Fromica C, Wang Z, 8: Hymsfield SB. (1996). Dual-
energy x-ray absorptiometry body composition model: review of physical

concepts AmericanloumaLQLEhysiolngLZZl: E941-E951

Prior BM, Cureton KI, Modlesky CM, Evans EM, Sloniger MA,
Saunders M, 8: Lewis RD. (1997). In vivo validation of whole body

117

 

composition estimates from dual-energy x-ray absorptiometry.]mirna1__of

Appliedlhysiologyﬁaz 623-630.

Pritchard IE, Nowson CA, Strauss BJ, Carlson IS, Kaymakci B, 8: Wark
ID. (1993). Evaluation of dual energy x-ray absorptiometry as a method of

measurement of body fat WW: 216-228

Putukian M. (1994). The female triad: eating disorders, amenorrhea,
and osteoporosis WM: 345-356

Quatrochi IA, Hicks VL, Heyward VH, Colville BC, Cook KL, Jenkins n
KA, 8: Wilson W. (1992). Relationship of optical density and skinfold
measurements: Effects of age and level of body fatness. Researghﬂiarjerly

WW: 402-409

Rencken ML, Murano R, Drinkwater BL, 8: Chesnut CH. (1991). In
vitro comparability of dual energy x-ray absorptiometry (DEXA) bone I
densitometers W: 245-248

 

Robinson TL, Snow-Harter C, Taaffe DR, Gillis D, Shaw J, 8: Marcus R.
(1995). Gymnasts exhibit higher bone mass than runners despite similar
prevalence of amenorrhea and oligomenorrhea. MW
MnemLResearchJQ: 26-35

Roche AF. (1987). Some aspects of the criterion methods for the
measurement of body composition. W: 2.09-220

Ross R, Leger L, Martin P, 8: Roy R. (1989). Sensitivity of bioelectrical
impedance to detect changes in human body composition. JenrnaigLApplied

magma: 1643-1648.

Sanbom CF 8: Jankowski CM. (1994). Physiologic considerations for
women in Sport Clinicsinipmtsmedicinela: 315-327

Schutte IE, Townsend EJ, HuggJ, Shoup RF, Malina RM, 8: Blomquist
CG. (1984). Density of lean body mass is greater in Blacks than 1n Whites.
:1647-1649.

Segal KR, Van Loan M, Fitzgerlad PI, Hodgdon IA, 8: Van Itallie TB.
(1988). Lean body mass estimation by bioelectrical impedance analysis: a four-

site cross-validation study MW: 7-14

Sinning WE. (1978). Anthropometric estimation of body density, fat,

and lean body weight 1n women gymnasts. Medieinmmmms
W: 243-249

118

 

Sinning WE 8: Little KD. (1987). Body composition and menstrual
function in athletes. WM: 34-45.

Sinning WE 8: Wilson JR. (1984). Validity of ”generalized” body
composition analysis in women athletes. r

M. 153-160.

Siri WE. (1956). The gross composition of the body. In Tobias CA and
Lawrence 11'1(Ed8), AdxmesirLBinogicaLBhysics (PP 239-280) New York

Academic Press.

Siri WE. (1961). Body composition from ﬂuid spaces and density:
analysis of methods. In Brozek] and Henschel A. (Eds), Ieehmguesfgr
(pp. 223-224). Washington, DC: National
Academy of Sciences.

Smith AD. (1996). The female athlete triad: causes, diagnosis, and E
treatment W412) 67-86 '

Snead DB, Birge SI, 8: Kohrt WM. (1993). Age-related differences in
body composition by hydrodensitometry and dual-energy x-ray

absorptiometry WW: 770-775

Strause L, Bracker M, Saltrnan P, Sartoris D, 8: Kerr E. (1989). A
comparison of quantitative dual-energy radiographic absorptiometry and dual
photon absorptiometry of the lumbar spine in postmenopausal women.

WW: 288-291.

Svendsen 0L, Haarbo I, Hassager C, 8: Christiansen C. (1993). Accuracy
of measurements of body composition by dual-energy x-ray absorptiometry 1n

vivo WW: 605-608

 

Thomasett A. (1962). Bio-electrical properties of tissue impedance

measurements. LyenMedieaLZQZ:107-118.

Tothill P, Avenell A, 8: Reid DM. (1994). Precision and accuracy of
measurements of whole-body bone mineral: comparisons between Hologic,
Lunar, and Norland dual-energy x-ray absorptiometers. Britishlqnmam

3341mm: 1210-1217.

Tothill P, Hans TS, Avenell A, McNeill F, 8: Reid DM. (1996).
Comparisons between fat measurements by dual-energy x-ray absorptiometry,
underwater weighing, and magnetic resonance imaging in healthy women.

WWW: 747-752

119

Tothill P 8: Pye DW. (1992). Errors due to non-uniform distributions

of fat m dual-x-ray absoptiometry of the lumbar spine. BritisthumaLef
W“ 807-813

Van Loan MD. (1990). Bioelectrical impedance analysis to determine
fat-free mass, total body water, and body fat. WW: 205-217.

Van Loan MD, Keim NL, Berg K, 8: Mayclin PL. (1995). Evaluation of
body composition by dual energy x-ray absorptiometry and two different

software packages WW: 587-591

Van Loan MD 8: Mayclin PL. (1992). Body composition assessment:
dual-energy x-ray absorptiometry (DEXA) compared to reference methods.
:.125-130

Wahner HW, Dunn WL, Brown ML, Morin RL, 8: Riggs BL. (1988).
Comparison of dual-energy x-ray absorptiometry and dual photon
absorptiometry for bone mineral measurements of the lumbar spine. Mayo

WW: 1075-1084

Wang J, Heymsfield SB, Aulet M, Thorton JC, 8: Pierson RN. (1989).
Body fat from body density: underwater weighing vs. dual-photon

absorptiometry. WW: E829-E834

Watt DE. (1975). Optimum photon energies for the measurement of

bone mineral and fat fractions. WWW: 265-274.

Wellens R, Chumlea WC, Guo S, Roche AF, Reo NV, 8: Siervogel RM.
(1994). Body composition in white adults by dual-energy x-ray

absorptiometry, densitometry, and total body water. AmerieanJgumgLQf
W: 547-555

Withers RT, Whittingham NO, Norton KI, LaForgiaJ, Ellis MW, 8:
Crockett A. (1987). Relative body fat and anthropometric prediction of body

density of female athletes WWW: 169-
180.

Witt RM 8: Mazess RB. (1978). Photon absorptiometry of soft tissue
and ﬂuid content: the method and its precision and accuracy. 131125191111

MW: 620-629

120

 

 

MICHIGAN smTE UNIV. LIBRARIES
1|lWI111111111111||||WWW11111111111111
31293017141387