A COMPARISON
MEASURE
OF TWO NONINVASIVE
TOTAL
BODY
DANIEL
LIPID
TECHNIQUES
TO
IN LIVE BIRDS
D. ROBY
Cooperative
WildlifeResearch
Laboratory
andDepartmentof Zoology,
SouthernIllinoisUniversity,Carbondale,
Illinois62901 USA
AI3STRACT.--Total
bodyelectricalconductivity(TOBEC)and near infraredinteractance(IRI)
are noninvasivetechniqueswith potential utility for monitoringbody compositionof freeranging birds or captive experimentalsubjects.I used live Northern Bobwhites(Colinus
virginianus)
to validatethesemethodsfor estimationof total body lipid. Total body electrical
conductivityand IRI measurements
were comparedwith body compositionas determined
by solvent extractionof carcasses.
I measuredIRI at threebodysites(subscapular,
pectoralis,and subfemoral)on 42 bobwhites.
Only the subscapular
and pectoralissitesyielded IRI measurements
that were significantly
correlatedwith percentbody lipid, but IRI from thesesitesexplainedonly 17.5%and 10.0%
of the variation in percentbody lipid, respectively.The causeof the low correlationsis
unknown, but it was not a function of differencesin subjectpostureor light-wand position
between replicate measureson the samesubject.
Total body electricalconductivitymeasurementsexplained92% of the variation in lean
body massin a sampleof 62 bobwhites.Measurementswere affectedby small differencesin
subjectmovementand positionin the measurementchamber.There was no significantdifferencein TOBECof subjectswith or without aluminum leg bands.Measurementsobtained
from the same subjectswith two different SA-1 TOBEC analyzerswere not significantly
different. The residualsabout the regressionof TOBECagainstlean body masswere not
correlatedsignificantlywith body temperature,but were significantlycorrelatedwith hydration. Total body electricalconductivityexplained99.1%of the variation in lean body mass
when only thosedatafrom normallyhydratedand properlyrestrainedsubjects
were included
(n = 38). Estimatedbody lipid from TOBEC measurementswas highly correlatedwith total
body lipid (R = 0.96) as determined by solvent extraction.The 95% confidencelimits from
the inverseregressionprocedurefor the estimateof total body lipid from TOBECmeasurementswere +4.75 g at 25 g lipid, +4.85 g at 10 g lipid, and +5.62 g at 50 g lipid. Estimated
total body lipid from TOBEC measurementswas on averagewithin 10%of total body lipid
from solvent extraction.Although IRI analysisfailed to yield a suitableindex to body composition,TOBECanalysisprovidedan accurateestimateof total body lipid in live bobwhites.
Received
30 July 1990,accepted
4 December
1990.
BIROSare highly active, obligate homeotherms that maintain high body temperatures
(38-42øC) and have few options for reducing
energy expenditurerates.Survival during periodsof negativeenergybalance,suchaswinter
or migration (Blem 1980),is related to the quantity of fat storedduring periods of energy surplus. In some species,reproductive successis
dependent on the parents' ability to recover
from the rigorsof winter and migrationand to
deposit sufficientenergy reservesto meet demands for courtship, egg production, incubation, and brood-rearing(King 1973).Energy is
an important currencyof fitnessfor birds, and
lipid depots are the primary form of energy
reproduction,molt, migration, and other components of avian life histories (reviewed in
Ricklefs 1974, Walsberg 1983, Blem 1990). Seasonal patterns of lipid deposition and catabolism identify thoseperiodswhen energyis limiting and how foraging behavior is adjustedto
meet these constraints(Blem 1976, Drobney
1980).Suchstudieshave provided considerable
insightinto the factorsresponsiblefor the evo-
lution of avian life-historytraits.
Considerable
information
on the nutritional
requirementsof birds and the impact of environmental contaminants has been gained
through controlled experiments with captive
stock. In these studies,a sample of subjectsis
reserves (Drent and Daan 1980, Blem 1990).
maintainedon a particulardietary regimefor a
Measurements
of bodycomposition
havebeen prescribedperiod, after which the subjectsare
usedto evaluatethe energeticcostof growth, sacrificedand body composition is analyzed
509
The Auk 108:509-518. July 1991
510
DANIEL
D. ROSY
[Auk,Vol. 108
(Scott1973).Periodicweighingshavebeen used
to nondestructively
trackbodycomposition
over
the courseof an experiment,but total body fat
stituentsis a function of their higher water and
electrolyte content (Pethig 1979). The differ-
pling precludesrepeated measurementof the
sameindividual to ascertaintemporal variation
in fat depots.
Over the last15yr, ornithologistshavesought
wavelength range of 600-2,500 nm (Rosenthal
encebetween total body massdeterminedby
may be poorly correlatedwith total body mass weighing and lean body massestimatedby TO(Blem 1990: 71).
BEC providesan estimateof total body lipid.
Proximateanalysisto determine total body Subjectsare placedin a chambersurroundedby
lipid (percent of total body mass)is a funda- a solenoid,which altersthe electromagneticinmental methodfor evaluatingphysicalcondi- ductanceof the coil. The changein inductance
tion and nutritional status of animals (Dobush is determined primarily by the subject'selecet al. 1985).This method is time-consumingand trical conductivity.A major advantageof the
expensive,and it requireskilling the subjectfor techniqueis that measurements
canbe obtained
analysis.Despitethe importanceof body-com- rapidly and easily. Validation studiesto date
positiondata as an index to physiologicalcon- indicate that accuracyof TOBEC estimatesof
dition, destructive methods are not feasible
lean body masscanbe high if careis taken to
where endangeredspeciesor valuablezoospec- insurethat subjectsare (1) properly positioned
imensare concerned.Even in the caseof larger, in the chamber,(2) not hyperthermic,and (3)
stablepopulations,there is growing publiccon- normally hydrated (Braccoet al. 1983,Walsberg
cern over the ethicsof sacrificinga sufficiently 1988, Keim et al. 1988, Castro et al. 1990).
largesampleof individualsto draw statistically
Near infrared interactance(IRI) analysisuses
valid conclusions. In addition, destructive samlow-energy electromagneticradiation in the
1986).The methodisbasedon the principlethat
differencesin the physicalcharacterof hydrogen bondsbetweenwater, protein, carbohy-
to devise nondestructive
methods to determine
drates,and lipids result in differencesin near
physicalconditionand energyreservesof birds. infraredabsorption
thatarecorrelated
with body
Bothinvasiveand noninvasivetechniqueshave composition (Futrex, Inc., 1988). The IRI techbeen investigated, including blood chemistry nique was initially developedby the U.S. De-
(LeMaho et al. 1981), labeled water dilution
partment of Agriculture in 1965 to measurethe
space(Nagy and Costa1980),fat scoring(Kre- compositionof foodstuffs.The technologyhas
mentzand Pendleton1990),bodymassadjusted beenusedwidely to controlqualityin the grain
for morphological measurements(Johnsonet industry(Rosenthal1977;Norris 1983a,b, 1985),
al. 1985),and ultrasonictechniques(Baldassarre to measurefat content of raw meats(Lanza 1983),
et al. 1980). All of these indirect techniques, and to measurehumanbody composition,parhowever, have limited ability to monitor pre- ticularly percentbody fat (Conway et al. 1984,
cisely temporal variation in lipid reserves.
Conway and Norris 1987). In theselatter studIn the last few years,two new noninvasive ies, IRI data from a seriesof body siteswere
techniqueswith the potential to measureac- comparedby linear regressionwith indepencurately total body lipid have attractedatten- dent estimates
of bodyfat obtainedby standard
tion in the field of human clinical nutrition
clinical techniques.The r2 of the regressionof
(reviewedin Lukaski1987).The totalbodyelec- resultsfrom IRI analysison deuteriumdilutiontrical conductivity (TOBEC) method was first predictedfat was 0.83, with a SEEof 3.1%(Condeveloped for use in analyzing the chemical wayandNorris1987).ConwayandNorris(1987)
compositionof foodstuffs,specificallyground concludedthat IRI analysishad potentialas a
meat and live hogs (Harker 1973, Domermuth field methodfor measuringbody composition,
et al. 1976).Subsequently,
the methodwasmod- but further validationwasnecessary
beforethe
ified for clinical studiesof human body com- techniquecouldbe widely applied. Regressed
position (Harrison and Van Itallie 1982; Presta againsta direct measureof body composition,
et al. 1983a, b; Harrison 1987, Van Loan et al.
such as solvent extraction, IRI data would re-
1987).The TOBEC method relies on the major
differencein conductivitybetween lipids and
other body constituentsto estimate total lean
body mass(Van Loan and Mayclin 1987). The
much higher conductivity of lean body con-
liably assess
the accuracyof the technique.
Near infrared interactancefat analyzershave
recently been developedthat are compactand
considerablymore portable than instruments
required for TOBECanalysis.The Futrex-5000
July1991]
Noninvasive
BodyFatMeasurement
51 !
computerized spectrophotometeris a small, inexpensive,battery-poweredIRI instrumentwith
the potential to estimatebody compositionover
a wide range of subjectbody sizes.This instrument usesa near infrared emitting and detecting "wand" placedon the skin surfaceduring
measurements. Consequently, subjects large
tained on their respectivediets for a minimum of 12
weeks(early Octoberto early January)before the ini-
enoughfor placementof the light wand on the
conditions
tiation
of TOBEC
and IRI measurements.
Most
sub-
jectswere measuredin midwinter, when bobwhites
normallyhavemaximalbodyfat reserves(Robel1972).
In addition, 11 wild bobwhites were live-trapped
at Crab Orchard National Wildlife Refuge in midNovember and kept in a separatepen under the same
described
above. Wild
bobwhites
were fed
body surfacecan be measured.The instrument
the high-energy diet describedabove for at least 12
calculates interactance
at two different waveweeksbefore measurementof body composition.
Totalbodyelectricalconductivity.--Preliminarytrials
lengthsby dividing the signal obtainedat the.
body site by the signal from a reflectancestan- indicatedthat ambient temperature,wind, and lightdard (1-cm thick Teflon block). The radiation ing conditionshada significanteffecton both TOBEC
intensityemittedby the light wand equalsa 60- and IRI measurements.Consequently, subjectswere
moved indoors for measurements.Each subjectwas
watt light bulb at 6 m, penetratesca. 1 cm into
restrainedin a nylon stockingand weighed on a Fishtissues,and is consideredsafefor live subjects er top-loadingbalance(+0.01 g). We useda protocol
(Futrex, Inc., 1988).
similar to that describedby Walsberg(1988) to meaI attemptedto validate the IRI techniqueon sure total body electricalconductivitywith the SA-1
live avian subjects.Walsberg(1988)and Castro SmallAnimal BodyCompositionAnalyzer(EM-SCAN,
et al. (1990) previously validated the TOBEC Inc., Springfield, Illinois). Before measurement,the
technique to estimatelean body massin a va- calibration of the instrument was checkedby reading
riety of avian species,but neither determined
(1) the reference number, (2) the data number with
the accuracy of the TOBEC technique to esti-
the sample chamber empty, and (3) the data number
with the calibration tube insertedin the samplecham-
mate total body lipid within a single species.
This is crucialif the techniqueis to be usedto
track changesin body compositionof individual subjectsover time, a major advantageof
nondestructivetechniques.
METHODS AND MATERIALS
Subjects.--NorthernBobwhites(Colinusvirginianus)
ber. Subjectmeasurements
were takenby (1) reading
the referencenumber, (2) inserting the empty sample
tube into the sample chamber and reading the resultingdata number,(3) insertingthe sampletube
with animal into the samplechamberand reading the
resulting data number, (4) reiterating steps2 and 3
sixtimes,and (5) readingthe referencenumberagain.
An index of lean body masswas calculatedas
EM-SCAN number = [(N• - N2)/R]C,
werechosenfor studybecause
theyaresmallenough
to fit into a commerciallyavailableTOBECapparatus where EM-SCAN number is an index of lean body
and large enough to permit IRI measurementat several different body sites.I purchased72 bobwhites
from a local breederin early October1989.Individualswere assignedrandomly to six groupsof 12, and
each group was housed in separate, raised outdoor
pens(4.9 x 1.2 x 1.2m high). Two-thirds of eachpen
was covered, and windbreaks were on 3 sides. Each
pen had a water containerand feed tray at each end
andevergreenboughsfor cover.Freshwaterandfood
were provided ad libitumon a daily basis.Two pens
of subjects(24 individuals) were placedon each of 3
differentdietaryregimes.One groupreceiveda high
protein (minimum 24.0%)game-birdstarterfeed (Siemet GamebirdStarter-Grower).The secondgroupreceived a lower protein (13.0%)game-birdmaintenancefeed (SiemerGamebirdMaintenance).The third
group receiveda high-energydiet of a 2:1 mixture
(v:v) of medium crackedcorn (ca. 8% protein) and
game-birdmaintenancefeed. The three dietary regimeswere designedto producesubjects
with a range
of bodylipid levels,but metabolizableenergycontent
of eachdiet was not measured.Subjectswere main-
mass,N• is the averagedatavalue for the six iterations
when the animal was not in the chamber, N2 is the
average data value for the six iterations when the
animal was in the chamber,R is the averageof the
two reference numbers, and C is a normalization
con-
stant specificto each instrument. This protocol was
designedto minimize error associatedwith variation
in the position of the subject in the chamber. Immediately following completionof the TOBEC measurements,I measureddeep proventricularbody temperature with a BAT-12 thermocouplethermometer
(ñ0.1øC).
After TOBECanalysis,an aluminum leg band was
attachedto 29 of the subjects
and the TOBECprotocol
was repeated.In another subsetof subjects(n = 12),
the TOBEC protocol was repeated with a secondSA-1
Analyzer (identicalmodel) to determineconsistency
of measurements
between
instruments.
Nearinfraredinteractance.--After
completionof the
TOBEC protocol, I used the Furrex-5000(Furrex, Inc.,
Gaithersburg,Maryland) computerized spectrophotometer to measure near infrared interactance (IRI).
512
DANIEL
D. ROB¾
[Auk,Vol. 108
TABLE
1. Morphometricsof Northern Bobwhitesusedin validationsof the total body electricalconductivity
and near infraredinteractancenondestructive
bodycompositionanalysistechniques.
œ_+SD
Range
n
Total body mass(g)
Parameter
228.3 _+25.0
171.8-276.9
62
Game farm-reared
237.7 _+ 15.8
206.8-276.9
51
Wild-caught
185.2 _+8.3
171.8-197.5
Total body lipid (% live mass)
Game farm-reared
High-energy diet
Intermediate
9.5 _+2.7
4.9-19.2
62
9.7 _+ 2.9
4.6-19.2
51
5.9-17.6
14
10.6 +_3.1
diet
Low-energy diet
Males
Females
9.7 _+ 3.4
4.6-19.2
16
9.1 _+2.3
5.8-14.2
21
5.9-19.2
4.6-14.6
27
24
8.9 +_ 1.7
7.1-11.9
11
9.3 _+ 1.9
8.6 _+ 1.6
7.5-11.9
7.1-10.5
10.8 _+ 3.0
8.4 _+ 2.3
Wild-caught
Males
Females
5
6
Total body lipid (g)
21.6 _+7.5
10.4-50.2
62
Game farm-reared
22.8 _+ 7.7
11.3-50.2
51
Wild-caught
16.4 _+3.7
10.4-22.6
11
In a room
darkened
to minimize
interference
from
at the two wavelengthsfor eachbody site. Immediately after the IRI protocol,deep proventriculartemperaturewas measuredand subjectswere quickly and
toralis, and subfemoral). Before measurement, each humanely sacrificedby thoracic compression.Carsite was plucked to provide bare skin for placement casseswere weighed to the nearest0.01 g, placed in
of the light wand.The threesiteswere chosenpartly double plastic bags and frozen at -20øC.
Carcassanalysis.--Carcasses
were partially thawed,
becausethey coincide with apterylae, which minimized the need for extensive plucking of contour weighed,plucked,and reweighedto determinefeathambient light, I measuredIRI at three different anatomical locationson each subject(subscapular,pec-
feathers. Before IRI measurements, an outline of the
light wand was drawn on the skin at each site with
a felt-tipped pen to insure that replicateswere taken
at the same site. I measuredby firmly pressingthe
end of the light wand againstthe subject'sskin within
the drawn
outline.
I scanned
each site 4 times
in a
seriesof 6 replicates,for a total of 24 scansper site.
Beforetaking each replicateset of measurements,
I
insertedthe probe into a Teflon optical standardand
took 4 reference measurementsat each wavelength.
Considerable
care was taken
to maintain
consistent
subjectposture and light-wand placement between
replicatesand subjects.
The Futrex-5000canbe programmedto output "optical density" (OD) values at 940 and 950 nm wavelengths. Optical density values are equivalent to log
1/interactance that vary linearly with body composition.Thesevalueswere standardizedby taking the
differencebetweenthe averageof the referencemeasurements and the baseline reference (measured at
the beginning of the study) at the respectivewavelengths, and subtractingthis difference from the average subjectmeasurement.This yielded a corrected
OD value at 940 nm and at 950 nm for eachbody site.
In clinical nutrition applications, both the corrected
OD values
and the difference
between
corrected
OD
er massby subtraction. Crop contents were then removed and weighed to the nearest 0.01 g. Plucked
whole carcasses were air-dried
at 60øC in a convection
oven to constantmass.Dried carcasses
were weighed
to determine water content by subtraction, then
ground and homogenizedby passingthem 6 times
through a meat grinder.
Extraction
thimbles
were dried at 60øC for at least
24 h, weighed to the nearest0.1 mg on a Sartorious
analyticalbalance,and loadedwith 2-3 g of dry infusorialearth(FisherScientific)to preventclogging
of thimble poresduring extraction.Threealiquots(23 g) of the dried carcass
homogenatewere weighed
(_+0.1 mg) into tared extraction thimbles with infu-
sorial earth. Loadedthimbles were placedin a convectionovenat 60øCovernightandreweighedbefore
extraction.I used a $oxtec SystemHT6 (Tecator, Inc.,
Hemdon, Virginia) solventextractionapparatusand
5:1 (v:v) petroleumether and chloroformto extract
total lipids from the three aliquots.Extractedlipids
wereweighedto determinelipid content(%dry mass)
for eachaliquot.The lipid contentof eachdried carcass was determined
from the mean of the three ali-
quots.If the standarddeviation of the mean exceeded
1%, additional aliquots were extracteduntil the SD
was less than 1% (averaged SD = 0.255%, n = 62).
valuesat 940 and 950 nm are usedto predictpercent Totalbodylipid (g) wascalculatedfrom the product
body fat (Futrex, Inc., 1988). Consequently,I calcu- of lipid content(% dry mass)and dry carcass
mass.
lated the difference between the corrected OD values
Lean body masswas calculated as the difference be-
July1991]
Noninvasive
Body
FatMeasurement
513
tween total bodymassand total bodylipid mass.Lipid
content (% live mass)was calculatedusing the following formula:
Lipid content (%) = TBL/(FMB - CC) x 100, (1)
where TBLis total body lipid, FMBis freshbody mass
immediately after death, and CC is the massof crop
contents.
Dataanalysis.--Toexplorethe relationshipbetween
IRI at the three body sitesand percent body lipid as
determinedby solventextraction,I usedleastsquares
linear and multiple regressionanalysis.I calculated
regressionsof each set of IRI measurementsagainst
total body lipid (% fresh body mass)both separately
and together to determine the best predictor(s).
The SA-1 was validated by regressingEM-SCAN
number againstlean body massas determined by solvent extraction.To investigatethe effectsof other
variables(body temperature,total body water, total
body lipid, total feather mass,date of analysis,sex)
on EM-SCAN number,I regressedthe residualsabout
the regressionof EM-SCAN number on lean body
massagainsteachvariable.I usedthe EM-SCAN number for each subjectto calculateTOBEC-estimatedtotal body lipid, and the least squaresregressionequation to estimate lean body mass.Estimatedlean body
masswas subtractedfrom live body massto provide
an estimateof totalbody lipid. TOBEC-estimatedtotal
bodylipid wasthen regressedagainsttotalbodylipid
as determined by solvent extraction. Inverse regression (Sokal and Rohlf 1981) was used to calculate
confidencelimits for TOBEC-estimatedtotal body lipid at differentlevelsof totalbodylipid. Statistics
were
calculated using StatWorks 1.0.
RESULTS
460
,'60
2'oo
,',o
2'40
2'6o
260
Live Body Mess (g)
Fig. 1. Percent body lipid as a function of total
live bodymassin gamefarm-rearedand wild-caught
Northern Bobwhites.Dotted squaresrepresentgame
farm-reared individuals,and solid diamondsrepresentwild-caughtindividuals.
variationin percentbodylipid, respectively(Fig.
1).
Percentbody lipid was not significantly correlated with dietary regime among game farmreared bobwhites (R = 0.202, P = 0.151, n = 62).
Birds presumably compensatedfor lower energy diets by consuminggreater quantitiesof
food. Percentbody lipid was significantlycorrelated with date of analysis (R = 0.418, F =
12.72, df = 1/60, P < 0.001), which indicates
that averagebody lipid levels slowly declined
during the validationperiod (January2 to March
14).
Live massof male game farm-reared bobwhites was not significantly greater than females (t = 1.435, P > 0.05). However, game
farm-rearedmaleshadsignificantlyhigher percent body lipid than females (t = 3.323, P <
0.05, Table 1). The sametrend was apparent in
Carcasscomposition.--A
total of 62 bobwhites
(51 gamefarm-reared,! ! wild-caught)were analyzed by the TOBEC technique (Table 1). Of wild-caught bobwhites (Table 1), but small samthe 51 gamefarm-rearedquail, 42 quail were ple sizesprecludedstatisticalsignificance(t =
analyzedby IRI spectroscopy
(20 males,22 fe- 0.698, P > 0.05).
males).Averagebody massof gamefarm-reared
Near infraredinteractance.--Thecorrectedvalbobwhites (237.7 g) was significantly greater uesfor optical density (OD) at either 940 nm or
than that of wild-caught bobwhites(185.2 g, t 950 nm were not significantly correlated with
= 15.681, P < 0.001); there was no overlap in percent body lipid for any of the three body
total body massbetween the two groups.How- sites(P > 0.10). Optical density at 940 nm was
ever,total body lipid asa percentof fresh mass highly correlatedwith OD at 950 nm for both
was not significantly greater in game farm- the subscapular(R = 0.959, P < 0.0005, n = 42)
reared bobwhitescomparedwith wild-caught and pectoralissites(R = 0.931, P < 0.0005, n =
bobwhites (t = 1.1902, P > 0.05, Table 1). Per-
42). The coefficient of variation in OD associated
centbodylipid wassignificantlycorrelatedwith
total body massin both gamefarm-reared (R =
0.371,P = 0.007,n = 51) and wild-caught (R =
with variation in position of the light wand on
eachsubjectaveraged1.00%for subscapularOD
0.675, P = 0.022, n = 11) bobwhites, but total
at 940 nm and 0.99% at 950 nm (n = 42 individuals); 0.80% at 940 nm and 0.82% at 950 nm
body massexplained only 14%and 46% of the
for pectoralisOD (n = 42); and 1.95%at 940 nm
514
DANIELD. ROSY
0.02
[Auk, Vol. 108
0.02 -
y = - 0.011 + 6.73e-4x R = 0.32
o
O.Ol
B
o
Q.
y=-0.016+0.001x
R=0.42
-0.02
Total Body Lipid (% live mass)
Fig.2. Differencein opticaldensityat 940nm and
Total Body Lipid (% live mass)
Fig.3. Differencein opticaldensityat 940nm and
950 nm at the subscapular
siteasa functionof percent 950 nm at the pectoralissite as a functionof percent
bodylipid in gamefarm-rearedNorthernBobwhites. bodylipid in gamefarm-rearedNorthern Bobwhites.
and 1.81% at 950 nm for subfemoral OD (n =
29).
Fig. 4). The coefficientof variation in EM-SCAN
number associatedwith variation in position of
The difference between OD at 940 nm and
the subject in the sample chamber averaged
950 nm (OD difference)was correlatedsignifi- 1.243%(range = 0.209-4.216%,n = 53 individcantly with percentbody lipid at the subscap- uals). Estimatedtotal body lipid from TOBEC
ular (R = 0.42, F = 8.452, df = 1/40, P = 0.006, measurements
wassignificantlycorrelatedwith
n = 42; Fig. 2) and pectoralissites(R = 0.32, F total body lipid, explaining 59%of the variation
= 4.469, df = 1/40, P = 0.038, n = 42; Fig. 3). in total body lipid (R = 0.768, F = 86.06, df =
However, subscapularOD differenceexplained 1/60, P < 0.0005, n = 62; Fig. 5).
only 18%and pectoralisOD differenceonly 10%
There was no significant difference in EMof the variation in percent body lipid. Optical SCAN number for subjectswith and without
density difference at the subscapularsite was aluminum leg bands(paired t = 0.470, P = 0.647,
not significantlycorrelatedwith OD difference n = 29 pairs).There was alsono significantdifat the pectoralissite (R = 0.090, P = 0.579, n = ference in EM-SCAN number from the two dif42).Opticaldensitydifferenceat the subfemoral ferent instruments(paired t = - 0.193,P = 0.844,
site was not significantly correlated with per- n = 12 pairs).
centbodylipid (F = 0.038,df = 1/27, P = 0.841).
When subscapularOD difference and pectoralis OD difference were included in a multiple
regressionmodel,thesetwo independentvariables explained only 25% of the variation in
percent body lipid (F = 6.612, df = 2/39, P =
0.004, n = 42). The multiple regressionmodel
that explained the greatestportion of the variance in percent body lipid included the variablessubscapularOD difference,sex,and analysisdate (r2 = 0.571, F = 16.838,df = 3/38, P <
0.0005,n = 42). No other independentvariables
explained a significant portion of the residual
variation in percent body lipid when entered
into the multiple regressionmodel (P > 0.05).
Total bodyelectricalconductivity.--EM-SCAN
number was significantlycorrelatedwith lean
body mass(R = 0.959) and EM-SCAN number
explained 92% of the variation in lean body
mass (F = 695.1, df = 1/60, P < 0.0005, n = 62;
The absolute values of the residuals about the
regressionof EM-SCAN number on lean body
masswere negatively correlated with date of
analysis(R = 0.446, F = 14.92, df = 1/60, P =
0.001, n = 62). A plot of the residualsvs. date
of analysisindicated that the precisionof TOBEC measurements
increased
over the valida-
tion period (Fig. 6). Specifically,for subjectsanalyzed before 23 Januarywe cut a hole in the
nylon stocking restraint so that the subject's
headprotrudedfrom the stocking.After 23 Januarythe subject's
entirebody,includingits head,
was enclosed in the stocking. When the data
collectedbefore23 Januarywere removedfrom
the data set, there was no longer a significant
correlationbetween the residualsand analysis
date (R = 0.127,P = 0.576, n = 43). Consequent-
ly, datafrombobwhitesanalyzedbefore23 January were removed from the data set.
In the new data set, EM-SCAN
number ex-
July1991]
Noninvasive
Body
FatMeasurement
515
30'
1000'
y•-427.648+5.851X
R--a0.96
=E
Z
800'
m
m
n-
,60
,;o
2;0
2;0
260
-2O
2O
40
60
80
Lean Body Mass (g)
Fiõ. 4. EM-$C^N number (TOBEC)as a function
of lean bodymassasdeterminedby solventextraction
in Northern
Bobwhites.
Julian
Date
Fig. 6. Residualsaboutthe regressionof EM-SCAN
numbervs. lean body massas a function of analysis
date.
plained 98%of the variation in lean body mass
(r • = 0.980, F = 1976.2, df = 1/41, P < 0.0005,
numbervs. lean body massand included the
n = 43), and estimatedtotal body lipid from
four outliers
EM-SCAN number explained 83% of the variation in total body lipid (r• = 0.832, F = 203.5,
were removed, EM-$CAN number explained
99% of the variation in lean body mass (r= =
df = 1/41, P < 0.0005, n = 43). The residuals
0.991, F = 3965.2, df = 1/36, P < 0.0005, n =
aboutthe new regressionequationof EM-SCAN
number vs. lean body masswere not significantly correlatedwith body temperature(R =
38). EM-SCAN number was related to lean mass
0.044, F = 0.081, df = 1/41, P = 0.774, n = 43).
in the model.
When
these data
asdeterminedby solventextractionby
EM-$CAN
number = -413.93
+ 5.831(LM),
(2)
Averagebodytemperatureof subjects
was41.8øC
(n = 61) and varied over a range of nearly 4øC
(SD = 0.87, range = 39.7-43.6). The residuals where LM is leanbodymass.The SEof the slope
were significantly correlatedwith total body of the regressionequation was +0.093. Lean
water (% of lean mass, R = 0.502, F = 13.14, df
mass can be estimated
= 1/39, P = 0.001, n = 41; Fig. 7). Five of six
ments of bobwhitesby the equation,
subjectswhosebody water was <66% of lean
massfell below the regressionline of EM-SCAN
30-
O.
20-
(3)
body lipid (r• = 0.917,F = 396.3,df = 1/36, P
40.
< 0.0005, n = 38; Fig. 8). I used the inverse
regressionprocedureto determinethe 95%confidencelimits for the estimateof totalbodylipid
from TOBEC measurements:+4.8 g at 25 g es-
-I
•
LM = (EM-SCAN number + 413.93)/5.831.
Estimatedtotal body lipid from EM-SCAN
number explained 92% of the variation in total
5O
_•
from TOBEC measure-
timated body lipid, _+4.9g at 10 g estimated
body lipid, and _+5.6g at 50 g estimatedbody
lipid.
[]
y= -0.476+1.021x R=0.77
DISCUSSION
20
30
40
50
6O
Total Body Lipid (g)
Fig. 5. Estimatedtotal body lipid from EM-SCAN
numberasa functionof total bodylipid asdetermined
by solventextractionin Northern Bobwhites.
The near infrared interactance (IRI) results
indicate serious limitations
in the usefulness of
the Futrex-5000for estimation of body compositionin bobwhites.The low correlationbe-
516
DANIEL
D. ROBY
10'
[Auk,Vol. 108
5O
40-
3O
2O
10
y = 0.254 + 0.988x R = 0.96
y = 103.338- 1.545x R =0.51
0
64
65
66
67
68
10
69
water (percentlean mass)in Northern Bobwhites.
3o
4o
5o
6o
Total Lipid Mass (g)
Total Body Water (% lean mass)
Fig.7. Residuals
abouttheregression
of EM-SCAN
numbervs.leanbodymassasa functionof totalbody
2o
Fig. 8. Estimatedtotal body lipid from EM-SCAN
numberasa functionof totalbodylipid asdetermined
by solventextraction.Data collectedfrom subjects
analyzedbefore23 Januaryor from dehydratedsubjects are not included.
tween IRI measurementsand percentbody lipid
is unexplained. Except for subfemoral optical
densitymeasurements,
the coefficientsof variationof OD measurements
for eachsubjectwere
not higher than thosefor total body electrical
conductivity measurements. Consequently,
variationbetweenreplicatemeasuresof OD due
to variation in light-wand position,subjectposition, or pressureexerted on the light wand
were apparently not responsiblefor the low
correlationbetweenOD and percentbody lipid.
Optical density measurementsfrom different
siteson the samesubjectwere not significantly
correlated. Consequently,either OD measure-
gression of EM-SCAN number on lean mass
subjectposture within the sample chamber af-
with
fected
The resultsof the present TOBEC validation
were comparedwith thoseof the only two pre-
when the subject'shead was not included in
the nylonstockingrestraint.Second,careshould
be taken to avoid subjectdehydration. In the
presentstudy, EM-SCAN number from dehydratedsubjectsunderestimatedlean mass.Walsberg (1988) obtained similar resultswith Gambel'sQuail (Callipepla
gambelii)deprived of water
for 3 days. Alterations in the electrolyte concentration and compositionof artificial "phantom" subjectsare known to alter EM-SCAN
number (Klish et al. 1984). Subjectswith abments failed to reflect differences in composi- normally high extracellularfluid volume,howtion under the light wand, or these differences ever, do not significantlyaffectthe accuracyof
number
were weakly correlatedwith differencesin total lean mass estimates from EM-SCAN
body composition. Differences between indi- (Cochranet al. 1989).This is apparentlydue to
viduals, sexes, or both, in the distribution of fat
the predominanceof water in determining the
depots may be related to preferential mobili- conductivity of biological systems.In dehyzation of lipid from certain depots.
drated subjectswhere body water constitutesa
The SA-1 Small Animal Body Composition relatively small proportion of lean body mass,
Analyzer (TOBEC)is a promising method for TOBEC measurements underestimate the true
tracking changesin body composition of ex- lean body mass. The lost precision associated
perimental subjects.Normalized TOBEC mea- with subjectdehydration can be mitigated by
efforts to minimize subjectstressand the time
surementsare apparently consistentamong instrumentsand,asreportedby Castroet aL (1990), interval between capture and measurement.
aluminum leg bandsdo not significantlyaffect However, TOBEC measurements from wildEM-SCAN
number.
Two factors affected the
caught birds whose hydration state is suspect
precision of lean massestimatesfrom EM-SCAN
(e.g. immediately following a long flight or unnumber. First, seeminglyminor differencesin der conditions of heat stress) should be treated
TOBEC
measurements.
This
was dem-
onstratedby the higher residualsaboutthe re-
caution.
July1991]
Noninvasive
BodyFatMeasurement
viously published validations using live birds
(Walsberg1988,Castroet al. 1990).Thesestudies used 11 and 8 speciesß
respectively,to generate prediction equationsfor lean massas a
function of EM-SCAN number.Walsberg(1988)
obtained
an r 2 of 0.988 for a fitted second-order
polynomial model, and Castroet al. (1990) obtained
an r 2 of 0.95 for a linear
model.
These
values are higher than the r2 for the complete
bobwhite data set reported in this study (r• =
0.920).Once datacollectedearly in the present
validation and those from dehydrated individuals were removed from the data set, however,
517
ACKNOWLEDGMENTS
Futrex, Inc., provided a Futrex-5000for this validation and EM-SCAN, Inc., provided a back-upSA-1
Small Animal Body CompositionAnalyzer to facilitate the study.G. Castroprovidedvaluableadviceon
the use of the SA-1 Analyzer. M. Eichholz and A.
Asch assistedin collecting TOBEC and IRI measurements, and J. Byrd and J. B. Tanner assistedwith
carcass
analyses.C. R.Blem,R.D. Drobney,R.J.Gates,
A. WooIf, and an anonymousreviewer madehelpful
commentsthat improved earlier drafts. Support for
this researchwas provided by the Cooperative Wildlife ResearchLaboratoryand the Office of Research
Developmentand AdministrationßSouthernIllinois
University, Carbondale.
the r2was 0.991.Castroet al. (1990) reportedan
average coefficientof variation associatedwith
LITERATURE CITED
position in the chamberof 6.8%,considerably
higher than the value reported here (1.24%).
G. A., R. J. WHYTE,& E.G. BOLEN. 1980.
These resultsemphasizethe importanceof the B^LD•SSARRE,
Use of ultrasonicsound to estimatebody fat deconsistencyof subject position and restraint
pots in the Mallard. Prairie Nat. 12: 79-86.
technique.In the presentstudy,the prediction BLEM,C. R. 1976. Patternsof lipid storageand utiequation for lean mass as a function of EMlization in birds. Am. ZooL 16: 671-684.
SCAN number had a considerablylower interß 1980. The energeticsof migration. Pp. 175cept (-413.93) and higher slope (5.831) than
213 in Animal migration, orientation, and navigation (S. A. Gauthreaux,Ed.). New York, Acathat of Castro et al. (-42.66 and 2.71, respecdemic Press.
tively). This suggeststhat a linear regression
1990. Avian energy storage.Pp. 59-113 in
model is only appropriateover a fairly narrow
Current ornithology, vol. 7 (D. M. Power, Ed.).
range of lean body massand that prediction
New York, Plenum Press.
equations need to be derived for each species
BRACCO,E. F., M. YANG, K. SEGAL,S. A. HASIM, & T.
to be measured.
B. VAN ITALLIE.
Total body electrical conductivity measurementscanyield estimatesof total body lipid in
bobwhites
that are within
+ 10% of actual total
1983.
A new method
for deter-
mining body compositionin the live rat. Proc.
Soc.Exp. Biol. Med. 174: 143-146.
CAS?RO,G., B. A. WUNDER, & F. L. FRrrz. 1990. Total
body lipid (mean= 10.42%,range = 0.20-64.1%,
body electricalconductivity(TOBEC)to estimate
total body fat of free-living birdsßCondor 92:
n = 38). In all respects(precision,subjectstress,
496-499.
ease,time required), the TOBECtechniquewas
COCHP•N,
W. J., M. L. FIOROTTO,
H.-P. SHENG,& W. J.
superiorto the IRI techniquefor nondestructive
KLISH.
1989.
Reliability
of
fat-free mass estimeasurementof total body lipid in bobwhites.
The
results
of the TOBEC
validation
on bob-
matesderived from total-bodyelectricalconductivity measurementsasinfluencedby changesin
whites indicatethat this techniquewill provide
extracellular fluid volume. Am. J. Clin. Nutr. 49:
new opportunities for measuring body com29-32.
position over time in individual birds. While CONWAY,J. M., & K. H. NORRIS. 1987. Noninvasive
the high accuracyof TOBECbody composition
body compositionin humans by near infrared
analysismakes it an attractive nondestructive
interactance. Pp. 163-170 in In vivo body composition studies(K. J. Ellis et al., Eds.). Londonß
method,commerciallyavailableTOBECanalyzInst. Phys. Sci. Med.
erscurrentlylimit the sizerangeof subjects
that
ß--,
& C. E. BODW•. 1984. A new apcan be analyzed. The SA-1 and SA-2 Small Anproach for the estimationof body composition:
imal Body CompositionAnalyzer (EM-SCAN,
Inc., Springfield,Illinois) can only accommodate birds up to approximately 300 g, and the
instrument is apparently most accuratein the
range of 50-250 g live mass(Castroet al. 1990,
D. D. Roby unpubl. data). Models designedfor
adult human or pediatric applicationsare too
largefor usein the field andareveryexpensive.
infrared interactance.Am. J. Clin. Nutr. 40:11231130.
DosusH, G. R., C. D. ANKNE¾,& D. G. KR•M•NTZ.
1985.
The effect of apparatusß
extractiontime, and solvent type on lipid extractionsof Snow Geese.
Can. J. Zool. 63: 1917-1920.
DOMERMUTH,W., T. L. VEUM, M. A. ALEXANDER,H. B.
HEDRICK,J. CLARK,& D. EKLUND. 1976. Predic-
518
DANIEL
D. ROB¾
tion of lean body compositionof live market
weight swine by indirect methods.J. Anim. Sci.
43: 966-976.
DRENT,
R. H., & S.DAAN. 1980. The prudentparent:
energeticadjustmentsin avian breeding.Ardea
68: 225-252.
--.
DROBNEY,
R.D. 1980. Reproductive
bioenergetics
of
Wood Ducks. Auk 97: 480-490.
FUTREX, INC.
1988.
Futrex-5000
1983b. Extractinginformationfrom spectrophotometric curves. Predicting chemical composition from visible and near infrared spectra.
Pp. 95~113in Proc.I.U.F.O.S.T.Symp.FoodRes.
Data Anal. Oslo, Norway, Appl. Sci.Publ.
1985. Reflectancespectroscopy.
Pp. 167-186
in Modern methodsof food analysis(J. R. Whitaker and K. Stewart, Eds.). New Haven, Connecticut, Avi Publ. Co.
Research Manual.
Gaithersburg,Maryland.
[Auk,Vol.108
HARKER,
W.H. 1973. Methodandapparatus
for mea-
PETHIG,R. 1979. Dielectric and electronicproperties
of biological materials.New York, J. Wiley and
suring fat content in animal tissueeither in vivo
or in slaughteredand preparedform. U.S. Patent
PRESTA,E., K. R. SEGAL,B. GUTIN, C. G. HARRISON,&
3,735,247.
HARRISON, G. G.
Sons.
t. B. VAN ITALLIE.1983a. Comparisonin man
of totalbodyelectricalconductivityandleanbody
massderived from body density:validation of a
new body compositionmethod. Metabolism 32:
1987. The measurement of total
bodyelectricalconductivity.
HumanBiol.59:311317.
HARRISON, G. G., & t. B. VAN ITALLIE. 1982. Esti-
534-537.
mation of body composition:a new approach PRESTA,E., J. WANG, G. C. HARRISON,P. BJORNTORP,
basedon electromagnetic
principles.Am. J.Clin.
W. H. HARTER, & T. B. VAN ITALLIE. 1983b. MeaNutr.
35: 1176-1179.
surementof total body electricalconductivity:a
new method for estimationof body composition.
JOHNSON,
D. H., G. L. KRAPU,K. J. REINECKE,
& D. G.
JORDE. 1985. An evaluation of condition indices
Am. J. Clin. Nutr. 37: 735-739.
for birds.J. Wildl. Manage.49: 569-575.
RICKLEFS,
R. E. 1974. Energeticsof reproduction in
KEIM,N. L., P. L. MAYCLIN,S.J.TAYLOR,
& D. L. BROWN.
birds.Pp. 152-297in Avian energetics(R. A. Payn1988. Total-bodyelectricalconductivitymethod
ter Jr., Ed.). Cambridge, Massachusetts,Nuttall
for estimatingbody composition:
validationby
direct carcassanalysisof pigs. Am. J. Clin. Nutr.
47: 180-185.
Ornithol.
Club Publ. No. 15.
ROBEL,
R.J. 1972. Body fat content of bobwhitesin
relation to food plantings in Kansas.Proc. Natl.
BobwhiteQuail Symp. 1: 139-149.
KING,J.R. 1973. Energeticsof reproductionin birds.
Pp. 78-107 in Breeding biology of birds (D. S. ROSENTHAL,R. D. 1977. An introduction to near inFarher, Ed.). Washington, D.C., Natl. Acad. Sci.
fraredquantitativeanalysis.Proc.Annu. Mtg. Am.
KLISH,W. J.,G. B.FORBES,
A. GORDON,
& W. J.COCHRAN.
Assoc. Cereal Chemists 1977: 1-15.
1984.
New
method
for the estimation
of lean
--.
body massin infants (EMME-Instrument): validation in nonhuman models. J.Pediatr. Gastroenterol. Nutr.
3: 119-204.
KREMENTZ,D. G., & G. W. PENDLETON. 1990. Fat scor-
ing: sourcesof variability. Condor 92: 500-507.
LANZA,E. 1983. Determinationof moisture,protein,
fat, and caloriesin raw pork and beef by nearinfrared spectroscopy.
J. FoodSci.48: 471-474.
1986.
Characteristics
of non-destructive
near-
infrared instrumentsfor grain and foodproducts.
Proc. Mtg. Japan Food Science Inst. 1986: 1-23.
SCOTT,M.L. 1973. Nutrition in reproduction--directeffectsand predictivefunctions.Pp. 46-68 in
Breeding biology of birds (D. S. Farner, Ed.).
Washington, D.C., Natl. Acad. Sci.
SOKAL,R. R., & F. J. ROHLF. 1981. Biometry, 2nd ed.
San Francisco, W. H. Freeman.
LEMAHO, Y., H. V. VAN KHA, H. KOUBI,G. DEWASMES, VAN LOAN, M., & P. MA¾CLIN. 1987. A new TOBEC
J. GIRARD,P. FERRf•,
& M. CAGNARD.1981. Body
composition, energy expenditure, and plasma
instrument and procedure for the assessmentof
body composition:use of Fourier coefficentsto
predict lean body massand total body water. Am.
metabolitesin long term fastinggeese.Am. J.
Physiol. 241: E342-E354.
LUKASKI,H. C. 1987. Methods for the assessment of
human body composition:traditional and new.
Am. J. Clin. Nutr. 46: 537-556.
NAGY, K. A., & D. P. COSTA. 1980. Water flux in
animals: analysisof potential errors in the tritiated water method. Am. J. Physiol. 238: R454R465.
NORRIS,
K. H. 1983a. Instrumentaltechniquesfor
measuringqualityof agriculturalcrops.Pp. 471484 in Post-harvest
physiologyand croppreservation (M. Lieberman, Ed.). New York, Plenum
Press.
J. Clin. Nutr. 45: 131-137.
--,
K. R. $EGAL,E. F. BRACCO,P. MAYCLIN, & T.
B. VAN ITALLIE.1987. TOBEC methodologyfor
body compositionassessment:
a cross-validation
study. Am. J. Clin. Nutr. 46: 9-12.
WALSBERG,
G.E. 1983. Avian ecologicalenergetics.
Pp. 161-220 in Avian biology,vol. 7 (D. S. Farner,
J. R. King, and K. C. Parkes, Eds.). New York,
Academic
1988.
Press.
Evaluation
of a nondestructive
meth-
od for determining fat storesin small birds and
mammals.Physiol. Zool. 61: 153-159.