Content uploaded by Janet Brunton
Author content
All content in this area was uploaded by Janet Brunton on Jan 13, 2015
Content may be subject to copyright.
ABSTRACT Precision and validity of dual-energy x-ray ab
sorptiometry (DXA) for analysis of whole-body composition in
infants were assessed by 1) scanning piglets in triplicate to cal
culate CVs, and 2) comparing DXA estimates with chemical
analysis of whole carcass. The mean CVs for all DXA measures
in small piglets and large piglets were < 2.5%, except for fat
mass, which were 6.3% and 3.5%, respectively. In large piglets
DXA provided reasonable estimates of chemical analysis for
bone mineral content (BMC), lean body mass, and fat mass, but
only for lean body mass in small piglets. DXA overestimated fat
by twofold and underestimated BMC by a third in small piglets.
Scans of prematurely born infants (n = i7) at term and at 3, 6,
and i2 mo corrected age demonstrated that changes in BMC,
lean body mass, and fat mass can be quantitated by DXA. How
ever, further refinement of DXA technology is necessary before
reliable measures of BMC and fat mass in small infants are at
tainable. Am J Cli,, Nutr i993;58:839—45.
KEY WORDS Dual-energy x-ray absorptiometry, body
composition, premature infants, bone mineral content, total body
fat, piglets
Introduction
The influence of nutrient intake on body composition in grow
ing low-birth-weight infants is essential knowledge for optimiz
ing nutritional management for the support of growth and de
velopment. Until recently, indirect methods of measuring body
composition were not easily applied to small infants, but the lat
est generation of whole-body absorptiometers using dual-energy
x-ray absorptiometry (DXA) may be a useful tool for this pedi
atric population.
The application of DXA for quantitative assessment of fat
mass, lean body mass, and bone mass should provide measures
with precision and accuracy while being safe and noninvasive
for use in infants and small children. Previous generations of
densitometers using radionucides were too slow and subjected
infants to radiation exposures that were of concern (i). DXA uses
a constant potential x-ray energy source, resulting in an increased
scan speed with greater resolution and lower radiation exposure
compared with dual-photon absorptiometry using ‘¿53Gd(2). To
tal-body-potassium counting is not sensitive enough to use in
small infants (3) and methods of electrical conductivity have not
been validated in this population. Skinfold-thickness measure
meats at one or more sites have been used to estimate total body
fat, but these are difficult to measure reliably in very small in
fants. Single-photon absorptiometry is a precise method for mea
suning bone mineral content (BMC) at peripheral individual bone
sites (4), but may not reflect total skeletal bone mineral mass in
infants prone to osteopenia.
To establish DXA as a useful clinical and research tool, the
sensitivity of the technology in detecting small changes in body
composition must be determined. Validation of DXA for use in
infants has been limited to repeated measures of small tissue
phantoms or excised bone (5, 6); precision has not been estab
lished by repeated in vivo measures. Our objectives were first to
establish the precision and accuracy of DXA measurements of
whole-body BMC, fat mass, and lean body mass in young piglets,
which are similar in body composition to infants (7). Second,
serial DXA scans of prematurely born infants were conducted at
term and at 3, 6, and i2 mo corrected age to 1) determine the
feasibility of the measurement technique in young infants; 2)
assess the sensitivity of DXA in quantifying small changes in
BMC, lean body mass, and fat mass in rapidly growing infants;
3) compare DXA measures of radial bone mineral density (BMD)
to radial BMD determined by single-photon absorptiometry
(SPA); and 4) determinethe accuracy of DXA in predictingtotal
body weight.
Materials and methods
Piglets
Piglets ranging from 2 to 20 d of age were removed from the
sow at the Arkell Research Farm (University of Guelph, Guelph,
Ontario)andtransportedto McMasterUniversity CentralAnimal
1 From the Department of Pediatrics, McMaster University, Hamilton,
and the Department of Nutritional Sciences, University of Guelph,
Guelph, Ontario, Canada.
2 Supported by grants from The Ministry of Health of Ontario (SAA)
and the Medical Research Council of Canada(SAA) and by a Career
Scientist Awardat McMasterUniversity from the Ministryof Healthof
Ontario (SAA).
3 Address reprint requests to SA Atkinson, Department of Pediatrics,
McMaster University 1200 Main Street West, Hamilton, Ontario, Canada
L8N3Z5.
Received December 15, 1992.
Accepted for publicationJune 21, 1993.
Am J Clin Nutr 1993;58:839—45.Printed in USA. ©1993 American Society for Clinical Nutrition 839
Validationandapplicationof dual-energyx-ray
absorptiometryto measurebonemassand body
compositioninsmallinfants1@
Janet A Brunton, Henry S Bayley, and Stephanie A Atkinson
by guest on July 12, 2011www.ajcn.orgDownloaded from
840 BRUNTON ET AL
Facility. All piglets were housed in stainless steel cages, in ac
cordance with the Guide to the Care and Use of Experimental
Animals (8). Infrared lamps were used to maintain an ambient
temperature of 30 °C.The piglets were weighed on an electronic
scale accurate to 0.1 g (Sartorius, Goettingen, Germany). Two
specific weight groups were targeted for study: 1.6 kg (small
piglets) and 6.0 kg (large piglets). These weights were chosen to
approximate the lowest and midrange weight of infants who
would be ofclinical interest. Piglets that were within 5% of either
targeted weight were given 5% glucose water only, and killed
later the same day. Piglets that were underweight were weaned
to formula designed to meet their nutrient requirements starting
with half-strength feeding and progressing to full-strength feed
ing by the third day. Piglets weighing < 1.5 kg were intermit
tently gavage fed formula or glucose water until they were feed
ing independently. Once the piglets had reached the targeted
weight, formula feeding was stopped and glucose water was pro
vided until the scan measurement occurred later that day. The
final formula feeding was provided 24 h before the scan mea
surement and consisted of half of the piglet's daily requirement.
If the piglets were studied on the day of arrival to the laboratory,
they were killed@ 8 h after removal from the sow.
To facilitate the scan procedure, piglets were anesthetized with
an intraperitoneal injection of 25 mg pentobarbital/kg body wt
(65 g/L). This was necessary to ensure that the animal remained
completely motionless for the DXA scan.
DXA scan procedure
Scans were performed by using a Hologic QDR-1000/W
(Hologic Inc, Waltham, MA). The principles of dual-photon ab
sorptiometry to estimate tissue composition are described in de
tail elsewhere (9). The Hologic system uses fast kV@switching
of an x-ray tube to generate both high- and low-energy photon
beams (140 and 70 kV@,).The beam is transmitted in a rectilinear
raster throughout the entire body. Measurements of photon trans
mission are made for the high- and low-energy beams along the
scan path. At all measurement locations the beam passes through
a wheel containing tissue and bone simulants for continuous in
ternal calibration during the scan. The relative attenuation of the
two photon beams can be related to the mass of either component
in a two-component system. Consequently, if the body is consid
ered to consist of bone mineral and soft tissue, then total-body
bone mineral mass can be measured. If a second dual-photon
analysis is performed at sites that contain only soft tissue, then
soft tissue mass can be divided into lean tissue mass and fat mass.
Triplicate whole-body scans were conducted by using the pe
diatric scan mode. The piglets were placed uncovered on the scan
bed on their stomachs in the spread-eagle position, with the legs
extended from the body. Generally, scans were performed with
out repositioning but if the piglets started to arouse during scan
ning, more anesthetic was given and the scan restarted. On com
pletion of the scans the piglets were killed with an injection of
sodium pentobarbital. The whole carcass was immediately frozen
for subsequent chemical analysis. To determine whether fluid
shifts that occur postmortem would influence the DXA estimate
of soft tissue, six piglets (three small, three large) were frozen in
the scanning position and rescanned. This variable was important
for future studies in animals.
Individual scans were analyzed by using the Pediatric Whole
Body software (version 6.01, Hologic Inc.). The soft tissue anal
ys@sin the pediatric mode assumes a lean tissue hydration factor
of 86%. Large piglets have a lean tissue water content of@ 69%
(7). One scan for each of the large piglets was reanalyzed by
using 69% as the lean tissue hydration factor.
To determine the influence of covers and blankets in the sub
ject scan field, one piglet was scanned three times with 1) a flan
nd sheet under and over it in a single layer (total blanket weight
92 g), 2) a flannel sheet under and over it in a single layer with
the piglet also wrapped in a flannel receiving blanket (total blan
ket weight 220 g), and 3) a double layer of flannel under and
over it with the piglet wrapped in a receiving blanket (total blan
ket weight 265 g). This experiment was conducted after an up
grade to the software had been installed (version 6.02).
Carcass preparation and analysis
The whole frozen carcasses were individually ground to a
coarse homogenate (small piglets: Hobart model M803, Don
Mills, Ontario; large piglets: Autio model 801CH25, Astoria,
OR) then reground to ensure a fine homogeneous mixture. Two
samples of tissue homogenate from each piglet were weighed to
0.1 g, frozen, and then lyophilized. Total body water of each
animal was determined by the difference between the wet tissue
weight and the lyophilized weight. The lyophilized tissue was
reground in a blender to facilitate sampling.
Total carcass BMC was determined by weighing@ 0.5 g dried
tissue to a reproducibility of 0.1 mg. The weighed tissue was
heated to 500 °Cin a muffle furnace for 72 h, dried in a desiccator
to a constant weight, then reweighed to determine ash weight.
The CVs of triplicate samples from each piglet were 7.0% in the
small piglets and 7.6% in the large piglets. Total carcass lean
body mass was determined by the following equation:
Lean body mass
= total carcass nitrogen X 6.25 + total body water
Nitrogen was analyzed by the micro-Kjeldahl method (10) by
using 0.5 g dry tissue. The (Vs for repeated samples for the
nitrogen analysis were 2.6% and 5.7% in the small and large
piglets, respectively. Total carcass fat mass was determined by
the lipid-extraction method adapted from Folch et al (11). Ap
proximately 1.5 g dried tissue was homogenized (Polytron;
Brinkman, Lucerne, Switzerland) with a mixture of chloroform,
water and methanol for 5 mm. Separation of chloroform and
methanol layers was facilitated by centrifugation at 10 000 X g
for 10 min at 4 °C.The chloroform layer was drawn off and
evaporated in a nitrogen stream. Samples were heated in an oven
at 100 °Cfor 0.5 h then placed in a desiccator. The lipid was then
weighed. The CVs by this method were 2.2% and 2.9% in the
small and large piglets, respectively.
Infant scans
Infants (birth weight < 1500 g) who were diagnosed with
bronchopulmonary dysplasia (BPD)were recruited from the neo
natal intensive care unit at the Children's Hospital at Chedoke
McMaster, Hamilton, Ontario, as part of an ongoing clinical nu
trition trial. The study was approved by the Research Project
Advisory Committee and informed parental consent obtained.
During routine visits to the Growth and Development Follow-up
Clinic at term and at 3, 6, and 12 mo corrected age, serial DXA
measures were conducted. Infants were laid on the scan bed on
their stomachs. When sleeping quietly, the infants were scanned
once by using the pediatric scan mode. No sedation was used.
by guest on July 12, 2011www.ajcn.orgDownloaded from
MeanMean
CV%5Rangeg%%Small
piglets (n=10)Weight15720.20.0-0.3BMC272.30.6-4.3Lean
mass13710.80.3—1.6Fat
mass1746.32.8—12.2Large
piglets (n=10)Weight59840.20.0-0.5BMC1161.50.1—3.4Lean
mass52170.60.3—1.0Fat
mass6513.30.9—5.7
841
BODY COMPOSITION BY DXA
TABLE 1
Precision of whole-body scans by dual-energyx-ray absorptiometry
in small and large piglets
bone mass compared for the two methods. A comparison of BMD
of the radius was of interest to determine whether DXA values
could be compared with previously published values for infants,
obtained by using SPA technology. All regional analyses were
conducted by one person with a mean intraobserver CV of 4.0%
(determined from repeated analyses of scans of three infants of
varying size). Infant scans were analyzed as two groups (< 5 kg
and 5 kg body weight) to determine whether infant size influ
enced the relationship between the methods.
Statistics
CVs were calculated for the triplicate measures. Differences
between mean DXA estimates for the groups of pigs and mea
sured values were determined by paired t tests (Minitab, version
7.1; Minitab Inc, Pittsburgh). The relationship between the two
methods was determined by regression analyses (Fig P, 6.0;
Biosoft, Ferguson, MO). The level of significance for all tests
wasP < 0.05.
Results
The mean weights and CVs for whole body and individual
body compartments as measured by DXA for each group are
presented in Table 1. FIgure 1 shows linear regressions of whole
body composition determined by DXA compared with the scale,
or chemically analyzed carcass weights in the small piglets. Only
total-body weight (Fig 1, A) and lean body mass (Fig 1, C) men
sured by the two methods were significantly correlated. DXA
determined BMC (Fig 1, B) and fat mass (Fig 1, D) were not
Fromtriplicatescanson individualpiglets.
During the scan infants were covered by a single layer of flannel
blanket and were wearing only a disposable diaper. Each infant
was weighed to 1 g on an electronic balance. Mean values of the
measurements of fat mass of infants were compared with litera
ture values for full-term infants (12) at similar corrected ages to
compare percent body fat values obtained by DXA vs those ob
tained by skinfold-thickness measurement.
Triplicate measures of BMD of the left radius at the one-third
distal site were conducted by using SPA (Norland, Fort Atkinson,
WI). The pediatric software allowed the DXA scans to be re
gionally analyzed so that the radius could be isolated and the
A 1800 B 46
42
36
30
24
y—fl.9+ 0.102*
r—0.16
P.O...
y.—5&3 + 1.03*
t—1.00
p<0.01
:@ @@oo
@ 1500
@ 1500
0
1400 T—@1@,@ ‘¿,
1400 1500 1500 1700 @80O
ScsI. Weight (g)
24 @0 36 42 46
Total Ash (9)
D 250
200
@ 150
0 100
50
C 1550@
@I
@ 1350
1200
y—162+ 0.617*
r.O.92
p<O.O1 S
•¿! a
55 5
Sy.171 + 0.053*
r.O.OI
-i@ 1@0 150 200 250
Total Fit (a)
1200 1350 1500 1550
Total L.an (g)
FIG 1. Linear regression comparing dual-energy x-ray absorptiometry (DXA)-determined total body weight (A),
bone mineralcontent (BMC) (B), lean body mass (C), and fat mass (D) with measuredvalues in small piglets. Hatched
line represents regression line, solid line is the line of identity.
by guest on July 12, 2011www.ajcn.orgDownloaded from
PercentDXAMeasuredPdifferencetgg%Small
piglets (n10)Weight1572
±76
(1426—1720)@1575
±73
(1431—1710)0.130.2Bone
mineral27 ±2.2
(25—32)38
±3.3
(31-43)<
0.0129.7Lean
tissue1371 ±62.8
(1261—1509)1456 ±70.6
(1335—1599)<
0.015.9Fat
tissue174 ±20.9
(139-205)52
±22.5
(27—89)<
0.01234.6Large
piglets (n =10)Weight5984±211
(5704-6415)5894±208(5630-6293)<0.011.5Bone
mineral116 ±14.0
(98—141)116
@ 19.9
(90-147)0.850.6Lean
tissue5217 ±203.3
(4866—5612)5151 ±188.5
(4779—5505)<
0.011.3Fat
tissue651 ±82.7
(485—736)480 ±78.8
(325—585)< 0.0135.6
842 BRUNTON ET AL
correlated with the chemically analyzed values in the small pig
lets. Figure 2 represents linear regressions comparing the two
methods in the large piglets. DXA-estimated total body weight,
BMC, total lean body mass, and fat mass were all significantly
correlated with the measured values.
Comparisons between DXA-determined weights and tissue
analysis for body weight, BMC, lean body mass, and fat mass
for both small and large piglets are presented in Table 2. In the
small piglets, DXA-determined values were significantly differ
ent from measured values for all indexes except total body
weight. DXA consistently underestimated total bone and lean
body mass and overestimated fat mass. In large piglets, DXA
marginally overestimated the total body weight and total lean
body mass. Total fat mass was overestimated by approximately
one-third.
In the large piglets, when the lean tissue hydrationassumption
was changed from 86% to 69%, further errorsoccurred in the
soft tissue measurements. Overestimation of total fat mass in
creased the error from 35% to@ 57%. Lean body mass was un
derestimated by < 1% when compared with chemically analyzed
values.
Table 3 representsthe results of the three scans with different
coverings. Additional layers of flannel consistently increased the
DXA-determined weight by 33% of the measured blanket
weight. The addition of blankets reduced the precision of the
measurementscomparedwith scans in small piglets (Table 1) for
total body weight (CV 1.7 vs 0.2%) and lean body mass (CV 1.9
vs 0.8%).
There were no significant differences between live and frozen
piglets for DXA estimates of weight (small 1565 ±54 vs 1569
TABLE 2
The accuracyof dual-energyx-ray absorptiometry(DXA)-estimated
weights vs measured tissue weights in small and large piglets
* Determined by paired t test.
t PercentdifferencebetweenmeasuredvalueandDXAestimate.
tir ±SD;rangeinparentheses.
±57 g; large5984 ±207 vs5984 ±200 g), BMC (small25.7
±1.4 vs 25.6 ±2.8 g; large 114.8 ±5.0 vs 118.6 ±9.4 g), fat
mass (small 172 ±29 vs 136 ±24 g; large 624 ±101 vs 614
A esoo
3 5200
.5
2
@ 5900
5600
C
; 5400
a
.
-I
@ 5100
4600
B 150
.@ 130
@ 110
0
90
D 750
@3650
a
U- 550
0 450
350
y—48.9+ 0.572x
r—O.61
p.cO.O1
UU
y——310+ 1.07*
r.1.0O
p'CO.Ol
5600 5900 6200 6500 90 110 130 150
Scala W.ight (a) Total Ash (a)
U@ U
I.
U
y.—145.9 + 1.04*
r—0.96
p'c0.01 y@232+ 0.868*
r-O.$3
p'cO.01
4800 5100 5400 5700
Total L.an (a)
350 450 550 650 750
Total Fat (g)
FIG 2. Linear regression comparing dual-energy x-ray absorptiometry (DXA>-determined total body weight (A),
bonemineralcontent(BMC)(B), leanbody mass(C),andfat massif)) to measuredvaluesin largepiglets.Hatched
line representsregressionline, solid line is the line of identity.
by guest on July 12, 2011www.ajcn.orgDownloaded from
Scan
1@Scan 2tScan
34i±SDCVgggg%Weight1749178718091782
±30.41.7BMC42.440.541.341.4 ±0.962.3Lean1592163916491627 ±30.41.9Fat115108119114 ±5.64.9
WeightCorrected
agegwkBirth
(n = 17)847 ±15224.6 ±1.6tFollow-up
ageTerm
(n = 9)3089 ±4224.3 ±1.9k3mo(n=14)4675±82415.1±1.9j6mo(n=10)5924±95226.8±0.8*l2mo(n=4)8140±129151.8±1.8*
843
BODY COMPOSITION BY DXA
DXA-determined BMC and the ashed weight of fresh chicken
bones. In Clan's study, DXA underestimated BMC by > 50%,
which is similar to our results in the small piglets.
Ashing whole-carcass tissue homogenate will result in a slight
overestimation of BMC because of the inclusion of nonbone min
eral, but this should not influence the results significantly. The
discrepancy between our study results and previously published
in vitro findings may be partially explained by different instru
mentation. Braillon et al (5) used the same device on bones that
were similar in size to those found in our piglets, but the scan
type and software used in the study were not clearly defined.
Thus, it is difficult to compare their findings with ours, deter
mined by using the pediatric whole-body scan mode.
When the body weight ofan animal (and presumably an infant)
is —¿6 kg, both the accuracy and precision of DXA estimates of
BMC are acceptable. At 1.6 kg body wt, DXA consistently un
derestimated BMC by a range of 17-40%. The failure of this
device to accurately measure BMC in small piglets is likely mul
tifactorial. The partial volume effect of determining bone content
in pixels containing bone and soft tissue will be exaggerated in
very small animals. As well, skeletal density varies; therefore,
regions of low density may be below the threshold for detection
resulting in artificially low whole-body BMC. The small and
large piglets had 38 ±3.3 and 116 ±19.9 g whole-body BMC,
respectively. Healthy full-term infants have whole-body BMC
comparable to that of the small piglets (13). The infants we stud
ied had an average DXA-estimated whole-body BMC of58 ±18
g at full-term corrected age. Infants at 3 mo corrected age had
almost double this mass (112 ±26 g), a value similar to that of
the large piglets. Our results illustrate the difficulties that will be
encountered when DXA is used for in vivo measurement of
whole-body composition in small infants in early life.
The small piglet is an ideal model for studying soft tissue com
position of preterm infants. Both have lean tissue hydration of
@ 80% and both have low body-fat stores that are well under
10% (7, 13). DXA-estimated total body fat was the measure with
the most variability and least accuracy. In small piglets, DXA
overestimated total body fat by 2.5-fold. In absolute terms, this
corresponds to@ 120 g fat. DXA calculates a fat fraction for the
pixels containing no bone. This is extrapolated to total body fat
from the soft tissue mass. In small piglets and preterm infants,
their distinct body composition (ie, minimal body fat) may be
beyond the capabilities of the current pediatric software. A fur
ther explanation for the large error in estimation of total body fat
may be the absolute thickness of the animal being measured. In
TABLE4
Infant characteristics5
TABLE 3
The influenceof flannelblankets in the scan field on body
compositionmeasurementsof a single piglet weighing1717g by
dual-energy x-ray absoptiometry
S Single flannel sheet under and over piglet; blanket weight 92 g.
t Singleflannelsheetunderandover pigletwith pigletin flannelre
cciving blanket; blanket weight 220 g.
t Doubledflannelsheetunderandoverpigletwithpigletinreceiving
blanket;blanketweight 265 g.
±107 g), or leanbodymass(small1366 ±32 vs 1409 ±42 g;
large 5247 ±147 vs 5252 ±144 g). The CVs of triplicate scans
on the frozen animals were similar to those calculated for the
same animals when scanned alive.
The characteristics of the infants are presented in Table 4. A
total of 37 scans were conducted in 17 infants. There was a high
correlation between DXA-estimated total body weight and the
scale-measured weight of infants weighing < 5 kg (y = 0.024
+ 1.039x,r = 0.997,P < 0.01) and 5 kg (y = 0.206 + 1.014x,
r 0.995, P < 0.01).
The correlation between the DXA and SPA measures of radial
BMD in infants weighing < 5 kg and 5 kg is depicted in Fig
ure 3. There was no correlation between methods in the smaller
infants, but there was a significant correlation in infants weight
ing 5 kg. In most infants weighing 5 kg the DXA-deter
mined radial BMD was higher than the SPA measurement.
Figure 4 depicts the proportion of fat and fat-free body mass
(FFBM) as determined by DXA in infants scanned during follow
up visits. Compared with reference infants (12), the infants with
BPD had lower body weights and a higher proportion of total
body weight as fat at all ages except 12 mo.
Discussion
Measures of whole-body composition in infants, especially
those of low birth weight, require an accurate tool. For longitu
dinal measures the precision of measurement must be small
enough to allow detection of small differences in BMC and soft
tissue compartments.
Our data established that repeated whole-body measures of
BMC by DXA were repeatable with CVs < 2.5% in live animals
as small as 1.6 kg, but in these animals the quantitation of BMC
by DXA was not accurate compared with total body ash. Other
investigators have assessed the accuracy of DXA-determined
BMC but used in vitro rather than in vivo models. In one study
by Braillon et al (5), femurs excised from premature infants were
submersed in water, scanned in triplicate (Hologic QDR-1000),
then ashed. In contrast with our findings, they found a high cor
relation between BMC measured by DXA and ash weight, but
DXA consistently overestimated BMC. Chan (6) used a Norland
XR-26 (Fort Atkinson, WI) and found a high correlation between
S f@ SD.
t Postconceptionalage at birth.
t Agecountedfromterm(40wkpoatconceptionalage).
by guest on July 12, 2011www.ajcn.orgDownloaded from
844 BRUNTON ET AL
B
.300
.250
E
U
2 .200
0
< .150
0
.100
of@@@0
A
0250@ y0.132 + 0.163*
idj 0.200@ p0.46 U
:@
0
z U
@ 0.150 UU U@
0U
0.100
0L@@ I
0 0.100 0.150 0.200 0.250
SPA BMD (g/cm2)
A
A
A A y.0.106 + 0.548*
A r—0.82
p'CO.Ol
.150 .200
SPA BMD (g/cm@)
.250 .300
FIG 3. Linear regression comparing dual-energy x-ray absorptiometry (DXA)-determined radial bone mineral density
(BMD) to single-photon absorptiometry (SPA)—determined radial BMD in infants weighing < 5 kg (n = 16) (A), and
5 kg (n = 14) (B).
vivo studies of tissue phantoms mimicking various tissue thick
nesses showed the greatest accuracy between 10 and 20 cm (14,
15). At thicknessesoutside this range, the attenuation error
caused by the fractional crossover of the high- and low-energy
beams increases. A large proportion of the total body in a 1.6-kg
piglet was < 10 cm thick in the scanning position, particularly
in the appendages. Specific calibration to account for this effect
is likely required when small animals or infants are measured.
Although the precision and accuracy of DXA was better in the
large piglets than in the small piglets, DXA still overestimated
total body fat by = 35%. DXA measures of large piglets would
also be subject to the error introduced by minimal tissue thickness
but to a lesser degree than in the small piglets.
The 6-kg piglet is not an ideal model for an infant of the same
size because of differences in body fat content and regional dis
FIG 4. Dual-energy x-ray absorptiometry (DXA) estimates of body
composition in prematurely born infants (DXA) (n = 17) at specific
corrected ages during the first year of life compared with skin-fold-thick
ness estimatesof body composition in a referencepopulation(Ret)(12).
The valuesrepresentpercentbody fat.Term (n = 9),3 mo (n = 14),6
mo (n = 10), and 12 mo (n = 4). FFBM, fat-freebody mass.
tribution. Large piglets were 8.2 ±1.3% fat by carcass analysis;
an infant of the same weight has = 25% fat (12). The distribution
of fat in infants would likely influence the accuracy, ie, a high
proportion of fat in the appendages could be overestimated be
cause of the low tissue thickness.
The hydration of the lean tissue decreases in both piglets and
infants during early neonatal life (7, 13). Theoretically, the lean
tissue hydration assumption should be altered in the software
program to account for this change in body hydration. In practice,
the use of a lower hydration factor (69% vs 86%) increased the
error in the soft tissue analysis by further overestimating total
body fat. The relationship between the ratio of attenuation co
efficients at the low and high energy levels (RM) and the percent
fat value of soft tissue will be altered as the tissue hydration
changes. Further study is required to determine the effect of hy
dration state on DXA estimates of soft tissue, because a change
in the lean tissue hydration factor did not improve accuracy.
The numerous potential sources of error in DXA technology
dictate that all DXA instruments must be validated before reliable
in vivo measures of whole-body composition can be achieved in
small infants. Previous technological refinements of the pediatric
software program for the QDR-1000/W resulted in improved pre
cision in small piglets; the CVs for total fat decreased from 8.6%
to 6.3% when scans were reanalyzed with upgraded software
(version 6.01). Further improvement in accuracy is necessary be
fore DXA estimates of whole-body composition in infants or
animals weighing < 6 kg can be interpreted with confidence.
Appropriate blanket coverings and positioning will influence
DXA estimates, thus standardization of the DXA scanning pro
cedure is essential.
DXA scanning to determine whole-body composition has
proven to be a feasible technology for use in infants. Precision
was not assessed because repeated measures unnecessarily cx
pose infants to further radiation and would require sedation for
the infants to remain still for a prolonged period of time. In our
experience, scanning of infants without the use of sedation is
certainly feasible; 40 infant scans were attempted, of which 37
were completed successfully. The average scan time for whole
body-composition analysis of an infant is 8 mm.
Term 3Mo 6Mo l2Mo
by guest on July 12, 2011www.ajcn.orgDownloaded from
BODY COMPOSITION BY DXA 845
Analysis of the accuracy of DXA in infants resulted in similar
findings to those for the piglets. DXA consistently overestimated
total body weight by@ 3.5%, a value slightly greater than in the
piglets (0.2-4.5%). This is likely due to the use of blanket coy
erings and disposable diapers on the infants during scanning,
which were not used on the piglets. The use of flannel coverings
in the scan field increases the DXA-determined weight of the
subject by@ 33% of the weight of the blanket in small piglets.
Unpublished data from our laboratory showed that in large pig
lets the same blanket coverings increased the DXA-determined
weight, representing 50% of the blanket weight. Therefore, the
use of a correction factor to accommodate blanket use must be
done with caution.
We analyzed the infant bone-density-scan data in two groups
(those weighing < 5 kg and those 5 kg) based on the differ
ences in accuracy of DXA observed in the large piglets vs that
in the small piglets. In infants weighing < 5 kg, the mean DXA
and SPA-determined BMD were 0.211 and 0.192 g/cm2, respec
tively. The greater value for radial BMD by DXA over SPA is
as expected because the DXA analysis includes more dense cor
tical bone in its measurement than would be present at the distal
one-third radius site. There was no association between the DXA
and SPA measures in infants weighing < 5 kg. In the smaller
infants it was more difficult to isolate the radius for the regional
analysis by DXA. The potential error in isolation of the radius in
the DXA scan field likely contributes to the lack of agreement
between measures by DXA and SPA.
If measures of body fat in infants taken by using DXA tech
nology are validated, this technique could become the standard
for body-composition analysis because hydrostatic weighing is
not ethical in this population and skinfold-thickness measures do
not represent total body fat (16). Body-composition data for a
reference population of infants used two-site skinfold-thickness
measures to estimate total body fat (12). Compared with these
reference data, our infants had lower weights (by@ 1 kg) and
had a greater percent of total body fat (Fig 4). The largest din
crepancy between our measures of percent body fat and the pub
lished values (12) was in the smallest infants. If our values are
corrected by the 35% overestimation of fat compared with chem
ical analysis that we observed in the large piglets, then the percent
body fat of our infants is similar to the reference population at
term and at 3 and 6 mo corrected age. Further study is required
to determine whether body composition of term infants is a rca
sonable reference standard of comparison for prematurely born
infants when they reach similar corrected ages.
DXA technology is noninvasive and relatively safe for use in
small infants and children. We believe that further software im
provements could result in a technology that has the potential to
be the standard for estimation of body composition in small in
fants. N
Wegratefullyacknowledgetheassistanceof RobertBertolo,Michelle
Whelan, and Cohn Webber, and especially appreciate the cooperation of
the parents of the infants who were scanned.
References
1. Johnson J, Dawson-Hughes B. Precision and stability of dual energy x
rayabsorptio.netrymaasuremants.CalcifTissuehit 1991;49:174-8.
2. Russell-AuletM, Wang J, Thornton J, et al. Comparisonof dual
photonabsorptiometrysystemsfor total body bone and soft tissue
measurement: dual energy x-rays versus gadolinium 153. J Bone
Miner Res 1991;6:411—5.
3. Forbes GB. Assessment and Significanceof body composition in
children. Iii SchOrch B, Scrimshaw NS, eds. Activity, energy cx
penditure, and energy requirements of infants and children.
Lausanne, Switzerland: International Dietary Energy Consultancy
Group,1989:71—80.
4. Grccr FR, Lane J, Weiner S, Mazess RB. An accurate and repro
ducible absorptiomctric technique for determining bone mineral
contentin newborninfants@PediatrRes 1983;17:259-63.
5. BraillonPM,SalleBL,BrunetJ, GlorieuxPH,DelmasPD,Meunier
P1.Dualenergyx-rayabsorptiometiymeasurementof bone mineral
content in newborns: validation of technique. Pediatr Rca
1992;32:77—80.
6. ChanG. Performanceof dual energyx-rayabsorptiometryin eval
uating bone, lean body mass and fat in pediatric subjects. J Bone
Miner Rca 1992;7:369-74.
7. MannersMi, McCreaMR.Changesin theChemicalcompositionof
sow reared piglets during the 1st month of life. Br J Nutr
1963;17:495—513.
8. CanadianCouncil on Animal Care. Guide to the care and use of
experimental animals. Vol 2. Institute of Laboratory Animal
Resources of the National Research Council, 1984.
9. PepplerWW, MazessRB. Total body bone mineraland lean body
mass by dual photon aborptiometry. I. Theory and measurement pro
cedure. CalcifTissue 1st 1981;33:353-9.
10. Association of Official Analytical Chemists Official methods and
analysis. 10th ed Washington, DC: Association of Official Analyt
ical Chemists 1965:15-6.
11. Folch3, Lees M, Stanley OHS.A simple methodfor isolationand
purification of total lipids from animal tissue. J Biol Chem
1957;226:497—509.
12. Fomon SJ, Haschke F, Ziegler BE, Nelson SE. Body composition
of referencechildrenfrom birth to age 10 years. Am J Cliii Nutr
1982;35:1169—75.
13. Widdowson EM. Changes in body composition during growth. In:
Davis JA, Dobbing J, eds. Scientific Foundation of Pediatrics.
London: Heinman, 1974:330-42.
14. HaarboJ, GotfredsenA, HassagerC, Otristiansen C. Validationof
body composition by dual energy x-ray absorptiomctry (DXA). (in
Physiol1991;11:331—41.
15. Mazess RB, Barden HS, Black JP, Hanson J. Dual-energy x-ray ab
sorptiometry for total-body and regional bone-mineral and soft-tin
sue composition.Am J Clin Nutr 19@51:1106-12.
16. Forbes GB. Human body composition: growth, aging, nutrition and
activity. New York Springer-Verlag, 1987.
by guest on July 12, 2011www.ajcn.orgDownloaded from
