Вы находитесь на странице: 1из 7

Clinical Science (1992) 82, 687-693 (Printed in Great Britain)

Four-component model for the assessment of body


composition in humans: comparison with alternative
methods, and evaluation of the density and hydration of
fat-free mass
N. J. FULLER, S. A. jEBB, M. A. LASKEY*, W. A. COWARD and M. EllA
MRC Dunn Cfinical Nutrition Centre, Cambridge, u'K., and * Department of Nuclear Medicine,
Addenbrooke's Hospital, Cambridge, U.K.

(Received 30 September 1991/4 February 1992; accepted 11 February 1992)

appropriate, and the mean hydration fraction \Vasclose to


1. Body comllOsition was assessed in 28 healthy subjects
values which are generaIly applied to the fat-free mass
(body mass index 20-28 kg(m2) by dual-enerl,'Y X-ray (0.72-0.73). Despite concealing considerable inter-
absorptiometry, deuterium dilutioll, densitometry, 4üK individual variation, dIese mcan values llIay be applied to
counting and fom prediction methods (skinfold thick-
groul's with characteristics similar to tbosc in tbis stud)'.
ness, bioe\ectrical impedance, near-Í.r. intcractance and Finally, \Vith the notable cxception of skinfold thickness,
body mass index). Three- and four-component Illodels of bedsidc I'rcdiction metbods sllow poor agreement wilh
body composition were constrncted from combinations of both the tluee- and the four-component 1lI0dels.
the reference methods. The resnlts of all methods were
compared. Precisioll was evaluated by analysis of propa-
gation of errors. The density and hydration fraction of the
fat-frec mass wcre detcrmincd.
INTRODUCTION
2. From the precision of the basic mcasurcments, the
propagation of crrors for the estimation of fat (± SD) by Dual-energy X-ray absorptiometry (DEXA) is being
the four-component model was found to be ± 0.54 kg, by increasingly used for the predictioIl of gross body compo-
the three-component mudel, ± 0.49 kg, by deuterium sitian because o[ its ability to differentiate between bone
dilution, ± 0.62 kg, and by dcnsitometry, ± 0.78 kg. mineral, and fat ami fat -free soft tissues [1]. Despite its
Precisioll for the mcasuremcnt of thc density and relatively high cost, DEXA has certain advantages over
hydration fractioll of fat-free mass was ± 0.0020 kg(1 and ather reference methods (densitometry aud deuterium
± 0.0066, respcctively. dilution) in view of its speed and ease of applicalion lo a
3. The agreement behveen reference methods was variety of subjects.
gencraIly better than between reference and aIternative Classical two-component madels (fat and fat-free mass)
methods. Dual-energy X-ray absoptiometry predicted far the measurement-i body carnpositio[J are lirnited by
three- and four-component model body comllOsition assulJ1ptions regarding 'the constancy of compositjon o f
slightly less weIl than densitometry or deuterium dilution fat-free mass. A three-component model (fat, water and
(both of which greatly influence these Illulti-component fat-free dry matter [2]), which is based on measurclllcnts
models). obtained from both densitometry and deuteriulll dilution,
4. The hydration fraction of fat-free mass was calculated overcomes some unccrtainlies concerning lhe hydration
to be 0.7382±0.0213 (range 0.6941-0.7837) amI the [ruction o[ fat-free mass, but it does assume a conslant
density of fat-free mass was 1.1015 ± 0.0073 kg(1 (range ratio of protein lo mineral (the [at-free dry matter in this
1.0795-1.1110 kg(I), with no significant difference model). On theOIetical grounds, a further improvement
bctween men and women for either. may be achievcd by the use of a four-component l1lodel
5. The results sllggest that the three- and fOllr-colllpo- ([at, protein, water and mineral [3, 4]), which segregates
nent Illodels are not comprolllised by error s arising frolll protein from mineral. Thismodel removes Ihe need fOI
individual techniques. Dual-energy X-ray absorptiometry somc o[ the assumptions, inherent in lhe two- and
would appear to be a suitable alternative method for the three-component models, regarding proportions ancl
assessment of body composition in these healthy aduIts, average den sities of different components of tlle fat-free
The traditionalmean vallle assllmed for density of the fat- mass. Thc four-component moclel incorporates direct
free mass in c1assic densitometry (1.1 kg(l) appears to be mcasurcmcnt of total-body bone mineral content, as wcll

Key words: bady fatodensitómetry, dual-energy X-ray absarptiametry. fat-free mass. tatal-bady water.
Abbreviatlons; BMI,bady mass index; DEXA. dual-energy X-ray absarptiametry.
Correspondence: Mr N. J. Fuller, MRC Dunn Clinical Nutritian Centre. 100 Tennis Court Raad, Cambridge CB2 IQL. U.K.
688 N. J. Fuller el al.

as the established measurements of body density and alld were analysed using the Lunar DPX software package
totaI-body water. However, the densities of fat and 3.1.
protein are assumed to be eonstant, as are both the
density of totaI-body mineral and the ratio of osscous to
non-osseous mineral. Densitometl'Y
The aims of this study were four-foId; first, to establish Body volume alld density were assessed by the
the extent of agreement between the various methods of technique of Akers & Buskirk [6], modified by the use of
body composition anaIysis, with emphasis on DEXA and a helium-dilution technique for the measuremcnt of lung
the three- and four-component modeIs; secondly, to volume (based on that of Gnaedinger et al. [7]). In the
quantify the extent of the variation in density of the fat- calculation of body composition using this technique, the
free mass in tbis sample populatioll, and to test whether density of fat was assumed to be 0.9007 kg/l and the
differences in mean values exist between men and women; density of the faHree tissue was assumed to be 1.100 kg/I
thirclly, to cstablish the hydration fra~tion of the faHree [2].
mass and, once again, to evaluate any putative gender
differences; and finaliy, to assess the extent to which
measurement errors of the individual reference tech- Isotope.dilution technique
niques are propagated in the determination of body com- Total-body water volume (litres) was dctermincd from
position by multi-component modcis. measurement of isotope dilution, using saliva samples
obtained at 4, 5 and 6 h after dosing [8J. Phantom dilu-
tions were made at 20°C, and the eOlTesponding volumes
METHODS at 36°C were calculated from the densities of water at
200C and 36°C (0.99823 kg/l and 0.99371 kg/I, respce-
Subjects
tively, Smithsonian Tables). Sillce the density of water at
Twenty-eight healthy adults (12 women and 16 men) 20"C (0.99823 kgj1) is virtually identical with that foul1d
volunteered to participate in the study. The charac- for deionized Cambridge tap water (0.9985 kg/l [8]), the
tcristics of these subjects are shown in Table 1. Approval difference was considered small enough to be ignored. In
for the study was granted by the Ethical Committee of the calculatillg total-body water, it was assumed that
Dunn ClínicaI Nutrition Centre, and all subjects gave deuterium-dilution space was a factor of 1.04 times
their informed written consent. greater than total-body water, owing to praton exchange
[9], and for calculating body composition it was assumed
that the hydration fraction of the fat-free mass was 0.7194
Weight and height
(ca!culated fram the study by Siri [2]).
Body weight, corrected to nude weight, was measured
using elcctronic scates with a digital read-out (Sauter Type
Whole.body potassium measul'ement
E121O; Todd Scalcs, Newmarket, Suffolk, U.K.). Height
was mcasured with a wall-mounted stadiometer (Holtain Estimates of body fat and fat-free mass were obtained
Ltd, Crosswell, Crymyeh, Dyfed, U.K.). from whole-body 4°K measurements [9, 10].

DEXA Prediction 01' bedside methods

Total-body fat, fat-free soft tissu' and bone mineral ash Skinfold thickness was measured at four sites with
measurements were obtained for allsubjects using a standard skinfold calipers (Holtain Ltd), ane! estimates of
Lunar DPX whole-body scanner (Lunar Radiation Cor- body [al were obtailled by the mcthod o[ Durnin &
poration, Madison, WI, U.S.A.) as dcscribcd elsewhere [5] Womersley [11] using agc and sex appropriate equations.

73.71
23.40±2.19
1.56-1.78
19.7614.26
Mean±sD
26.13-57.62
66.50
59.39-90.06
48.25-90.06
17.10-24.65
22.28
1.56-1.88
12.01-25.04
1.64-1.88
12.01-37.95
33.8
58.741
51.61
49.28-77.83
Mean±SD
20-59
43.05 ±9.07
26.13-34.76
38.061732
48.25-67.80
1.72±
32.91
22.31 ±
18-55
± 10.6 11.55
5.38
10.7
7.39
±0.09
5.40
36.60-77.83 1.65 ±0.06
56.8916.25
10.24 20.86
17.10-28.10-"42.10±3.18
19.84-28.10
18-59
±2.46
36.94-57.62
19.56-37.95
1.77±0.06 31.8± 11.0(n = 28)
31.38±2.57
Range ±Men(n=16)
36.60-46.6325.6314.962.08
Combined Table 1. Characteristics of subjects

Women (n= 12)


/ Whole-body bioelectrical resistance/impedance
Body composition

was
measured using the technique described previously [12].
These measurements were used to estimate total-body fat
methodology 689

Clcarly, although only the equation for fat is shown, fat-


free mas s is also obtainablc frol1l this model.

mass (as a percentage of body weight) and fat-free mass


Four-component model
(kg) using equatiolls (dctails unknown) supplied in the
analyser software (Valhalla Scientific, San Díego, CA, The four-component model was based on principIes
U.SA.). Estimates of body fat, fat-free ma~s and water similar to those adopted for the three-component model,
were obtained from near-i.r. interactance measurements but involved the additional measurement of bone mineral
by the method described by Elia el al. [13]' Body mass by DEXA, allowing the body to be segregated into fat,
index (BMI) was calculated from weight and height (kgl water, mineral and protein. ,.
m2), enabling estimation of body fat from previously The density of total-body mineral (using data fram
puhlished prediction equations [14 l. Brozek el al. [15]) was calculated to be 3.0375 kg/l,
assuming a constant ratio (0.8191:0.1809) of osseous
(bone) mineral (density = 2.982 kg/l) to non-osseous
Three-component rnodel
mineral (density= 3.317 kg/l).
For the purposes of this study, a three-component Total-body mincral mas s was obtailled from the DEXA
mode! of body composition was constructed on the basis estimate of bone ash:
of three major and distinct chemical cOl11ponents within
Total-body mineralmass = ash x 1.2741 kg
the body: fat, water and the remaining fat-free dry l11ass
(consisting of protein and mineral in constant ratio). Body where the fraction of osseous mineral comprising ash was
composition was calculated by using combined assumed t'o be 0.9582, and, fram the ratio of osseous to
measurements of body weight, total-body water (fram nOll-osseous mineral, the mass of non-osseous mineral
deuterium dilution) and body volume (from densito- was assumed to be ash x 0.23048 kg (calculated from
metry). Brozek el al. [15]).
This three-component model was based on the follow- Total-body mineral volume was obtained fram total-
ing assumptions [15]: that the density of body fat was a body mineral mass and mean dcnsity of osseous and non- , v" ,"
constant 0.9007 kg/l, that the density of water at 36°C osseous lnineral, thus: ?f,.l v,''' /',,5";JL ~"'.,
was 0.99371 kg/l filld that the density of protein plus Total-body niineral volume = ash x (1.2741/3.0375)
lnineral was 1.5157 kg/l [obtained when protein (density or ash x OA1946litres
1.34 kg/l) and mineral (dcnsity 3.0375 kg/l) were
eombined in the assumed whole-body ratio of The fraetion of body fat· (f) was then ealculated in a
0.7926:0.2074]. manner analogous to that used in the three-colllponent
The fraction of body fat (f) was calculated according to model [in this case, d2 of [he equalion of Keys & Brozek
the following general equation, a rearrangel11ent of that of [16] (see above) represents protcin alone, and D
Keys & Brazek [16]: represents [he combined density of fat and protein, Dr+p]'
Hence .
f= d¡ (d2 - D)/D(d2 - di)
whcre, for the purposes of the three-component model, d[ f= [2.747 4/(D¡.,)] - 2.0503
represents the density of fat, d2 represents the density of where the tenn llf+p was ealculated from the difference
protein combined with mineral (constant ratio in the between whole-body mass and total-body water plus
whole body, as above), and D répresents the density offat mineral mass divided by the difference between whole-
eombined with this same ratio of protein and mineral body volume and total-body water plus mineral volume,
(D¡+pm)' in a malmer analogous to the three-component model:
With substitution of the assumed density values (see
above) into the equation of Keys & Brozek [16]: D f+ p = [body mass - (water mass + mineral mass )]1
[body volume - (water volume + mineral volume)]
f=[2.21989/(Df+pm)]-1.46462
In terlllS of the basic measurelllents:
The only unknown in this equation, D!+pm' was calculated
from the general equation: density = mass/volume, where Fat (kg)=2.747BV-0.7JOTBW+ L460A -2.0S0Wt. 4--
mass and volume of fat plus protein/mineral were where the abbreviatiol1s BV, TBW and Wt. are the same
obtained by the differcnee between the whole-b6dy and
as in the three-component model and A = ash in kg (from
the total-body water eomponent: DEXA).
D¡ +pm = (body mass - water mass)/(body volume - water Protein (kg) may be ealculated in a similar manller to
volume) fat, and fat-free mass is also available using this model.
In terms of the basie measurements:
Fat (kg) = 2.220BV - 0.764 TBW - 1.465 Wt. Calculation of density and hydration of fat·free mass
where BV=body volume in litres (from densitometry), The density of fat-free mass (Dffm) was calculated fram
TBW=total-body water volul11e in litres (fram deuterium the mass and volume of each individual component
dilution) and Wt. = body weight in kg. (water, protein and mineral) using the four-component
'\
~
690

model, thus; volume,


N. J. Fuller et al. 0.45 litres; total-body bone mineral ash, 0.03 kg
~ '\
Dffm = (mass ofwater+ protein + mineral)/ [5]. The precision for estimates of total-body water was "
based on sequential measurements of the isotopic enrich- .
(volume of water + protein + mineral)
ment of water in saliva samples taken at 4, 5 and 6 h after
In terms of the basic mcasuremcnts: oral adrninistration of the isotope. Thc results (or all
suhjects were comparcd with the enrichmcnt of a single
Drrm=(0.710TBW+ 3.050Wt.-2.747BV-1.4GOA)/
phantom dilution o( thc stock do se. Precision for the
(0.78STBW+2.276WL-2.050BV-1.621A)
measurement of water calculated from this study was 0.45
where the abbreviations TBW, Wt., Bv and A are as litres (about 1%), although a precision of 0.86 litres (2%),
defined for the four-eomponent modeL which may be more appropriate under ,certain circuITl-
The hydration fraetion was ealculated from the mass of stances [lS], was also considered.
total-body water (from deuterium dilution) and total-body
fat-free mass, obtained from the.four-component mode!. RESULTS
In terms of the basic measurements: .
The bias and 95% limits of agreement (shown as
Hydration fraetion = TBW/(0.715TBW+ 3.070Wt.
bias ± 2 SD) between methods are given in Table 2 for
- 2.765BV-1.469A)
percentage of body weight as fat, and in Table 3 for fat-
where the abbreviations TBW, Wt., BV ami A are as free mass(kg). In many of the eomparisolls, such as the
defined for the four-eomponent mode!. three- versus the four-component model, there is no
material differenee in the bias and 95% Jimits of agree-
$tatistics meut between men and women. However, in otl1ers, for
instance deusitometry versus the BMI formula, the bias is
The bias and 95% (± 2 SD) limits of agreemcnt [17] for seen to be of opposite sign and the 95% limits of agree-
eomparison of the estimates of body eomposition by a ment differ in magnitude betweeu men and women.
particular method eompared with other methods were Therefore, results obtained for each gender separately
established as previously in our laboratory (e.g. FuIler & have been deposited as Clinical Science Tables A and B
Elia [12]). Student's !-test was used to test for the signifi- 92/2 with the Librarian, Royal Society of Medicine,
cance of any differences between populatiol1 samples 1 Wimpole Street, London WIM 8AE, U.K., from whom
studied here (male versus female). copies are available on request. Table 4 shows the resuits
Estimates of the overaIl measurement precision in of this same statistical analysis performed on the alterna-
terms of fat or fat-free mass were calculated for eaeh tive methods aloue (for all subjects combined). Clearly,
model of body eomposition from analysis of the propaga- there is less overall bias ami better agreement when
tion of errors. These estimates were based on a 70 kg man reference methods are compared with eaeh other than
consisting of 15% fat, a hydration fraetion of fat-free mas s when the alternative methods are compared (either with
of 0.7194 (total-body water voJume is therefore 42.8 reference methods Ol" with eaeh other).
litres) a\1(1a total-body bone mineral ash eontent of 3.2 Tl1e density of the fat-free mass (mean ± SD) was found
kg. The values adopted for errors in the basic to be 1.l003±0.0066 kg(l and 1.l024±O,0078 kg/I
measurements were as foIlows: body dcnsity, 0.0025 kg/I for women and men respcctively, with no significant
[2]; body weight, 0.01 kg [18]; body volume (derived from difference betwcen the sexes. Thc ovcrallmean was thus
body weight and density), 0.157 litres; total-body water 1.1015 ±0.0073 kg/I with a range of 1.0795-1.1110 kg/1.

Table 2. Comparison of varíous body composition techniques with some relerence methods: bias and 95% (±2so)
Iimits 01 agreement lor estímates 01 body fat (as percentage of body weight). The Table shows the reference method!
model (top row 01 the Table) minus the alternative method (Ieft·hand column of the Table). Values lor the bias for fat-free
mass as a percentage of body weight are equal and opposite to those lor percentage lat, and the 95% limits of agreement
(bias ± 2 so) are equal far both (see [ i7]).

DEXA
-1.83±
-0.47
0.I3±S.60
-0.60±4.46
model
-0.81
2.16
2.03
0048 ±4.13
-0.94±6.12
-1.96±8.66
-0,51
-1.14±5.S7
-0.65 ±8.60
-1.27±5A2
034±7.50
±±
±8.4610.78
3.15
7.87 0.7S±3.78
Three-component1.70±3.91
1.23± 1043
OAI±7.79
±8.680.09
0.71 ±8.21
1.36±4.9S
7,07-0.60±
339 7.30
10.42
± 5042
Four-component Densitometry Deuterium
model DEXA dilution
ry
mponent model

0,95 ±6.83
-0.04±9.09 -0.99±8.71
-0.66±S.00 -1.61±7.13
2,63± 7.87 1.69±8,62
- 1.35± 10.23 - 230 ± 12040
-034±7A9 -1.29±10.15
Body composition methodology

Table 3. Comparison ofvarious body compositlon technlques with so me reference methods: bias and 95% (±2 SD)
Iimits of agreement for estimates of fat-free mass (in kg). The Table shows the reference method/model (top row 01 the
Table) minus the alternative method (Ieft·hand column 01 the Table). Values lor the bias lor kg 01 fat are equal and opposite to
thOse for fat·free mass. and the 95% limits 01 agreement (bias ± 2 SD) are equal in both (see [17]).

model
-0.6312.57
-0.42±4.48
0.631
-0.48±2.98
0.5414.04
0:06
0.2016.54
1.27±4.75
DEXA
0.80
0.45
0.03 ±
5.91
3.57
-0.74
± ±6.39
8,48
3.87-1.37
-0.69±3,19
3,43- 0.71 ±2.97
±I.02
±
-0.20±5.31-0.54±6.34
0.96
1.6916,43
1.22± 5.84
-1.1614.07
6.04- 2,43
Three·component 16,04
0.05±8.17
Four-component
-0)0±3,88 Densitometry Deuterium
model DEXA dilution
metry
omponent model

-0.68 ±4.65
0.1516.05 0.83±6,46
0.48 ± 3.21 1.16± 5.03
- 1.75± 5.29 -I,06±6.30
0.74±7.16 1.42± 8.90
-0.29±4.88 0.65 ±6.91

Table 4. Comparison of varlous alternative body composition techniques: bias and 95% (±2SD) Iimlts of agreement
for estimates of fat and fat·free mass. The Table shows the reference method/model (top row of the Table) minus the
alternative method (Ieft·hand eolumn of the Table). Values lor the bias lor lat-free mass as per eent body weight and those lor
lat mass (kg) are equal and opposite to those for per cent fat and fat-lree mass (kg). respeetively, and the 95% limits 01
agreement (bias±2\o) are equal for both (see [17]).

Skinfold
-0.77
0.26±7.63
-0.51
--4.01
2.49
2.2314.86
±7.25
± 8.7
- Whole-body
-0.69
3.30±7.08
2.97±7.98
0.3318.61
1.72±
1,01±
± 1 1-0.62±7,n
±5.61 0.3415.14
-0.18±6.68
--1.89
0.59±
1.31±
±1282-0.30±9.70
potassium
impedanee5.3
11.38
11.07 14.65
10.08
268±5.11
BMI± 3.85
hiekne\\ Skinfold
tassium Whole·body
pOtaSsium Whole·bodyimpedanee
impedance Skinlold thiekness

The hydration fraction of the fat -free tissue (mean ± su) equations for calculating densíty and hydration fractíoll of
was calculated to be 0.7449±0.0192 for women and fat-free mass). The largest cOlltributíon to the ímprecísion
0.7332±0.0219 for men, withno significant differcncc associated with del1sítometry origínated from the estimate
between them. Thus the mean hydration fraction of all of body volume (± 0.77 kg). Precisíon in Úle three- and
subjects combíned was calculated to be 0.7382 ± 0.0213 four-component models, and in density and hydralíon
with a mnge of 0.6941-0.7837; fractÍon of fat-free mass, was found to be mostly attríbut-
Precísíon (su) assocíated with estímated amounts o[ fat ablc to the eslÍmatíon of total-body water and body
(kg) or fat-free mass (kg) for a 70 kg man using the three- volume, with thc contríbutíon of body weíght and mineral
and four-component models was found to be ± 0.49 kg ash being negligible.
and ± 0.54 kg, respectively, when a 1% precísion for SubstitulÍon of thc assumed and commonly accepted
wate!' estimation was assumed. Alternatívely, assumíng a values for either densíly (1.1 kg/l) or hydration fraction
2% precision for water estímation (other measuremelll (0.72-0.73) of the fat-free mass by values representing the
precisíons beíng held constant), the correspondillg values extremes o[ the range found in thís study demonstrated
were ± 0.74 kg alld ± 0.75 kg, respectívely. The precísíon potential errors ín the estímatíon of fat (and fat-free mass)
found for densitometry was ± 0.78 kg and that for of up to 2.4 kg and 2.6 kg, respcctívely. These rcpresent
c!euteríum dilutíoll was ± 0.62 kg, assumíng 1% precísion, about a 25% error in the fat cstímatíon (and about 5% in
amI 1.20 kg when 2% precisíon was consídered. The fat-free mass) of a 70 kg reference mano
precisíon associated with the calculatíon of hydratioll
fractíon was calculated to be ± 0.0066, and that for the DISCUSSION
densíty of fat-free mass was ± 0.0020 kg/l (only the 1%
prccísion in water estimate ís shown here because the This study has híghlighted the polentíal use of DEXA
errors from this source largely cancelled out, see the for assessíng body composition. It is simple to perform
./692 N. J. Fulleret al.
and exposes the voluntccr al' patient to only a low levcl of composition. The calculations in this study suggest that
radiatíon [5]. Determination of fat (percentage of body lhere is considerable inter-individual variation in the
weight) 01' fat-free mass (kg) by DEXA agreed slightly less density of fat-free mass, despite the close proximity of the
well with the three- or four-component models than did mean value to 1.1 kg/l for both men and women. Since
dcnsitomelry or dcutcrium dilulioll. This is not very these differences do exist, both within and between
surprising, since densitomctry and deulerium dilutioll populations, a single value for the density of the fat-free
pro vide the muy measured delerminants of body fat in the mass should not be universally applied. tor example,
three-component model, and are the major determinants either osteoporosis or overhydration (oedema) would
of fat in the four-component model. Moreover, DEXA tend to decrease the mean density of fat-free mass,
was found to be the best predietor of mean body compo- whereas greater proportions of bone minqal (as found in
sition obtained by four separate two-component some Negroid populations [201) would tend to increase it.
reference methods (densitometry, deuterium dilution, The four-component model of body composition also
DEXA and total-body potassium, results not shown). enables the hydration fraction of fat-free mass to be
Although, of these two-component methods, total-body assessed. In water-dilution techniques, the value for
potassium predicted body composition with leasl hydration fraction of the fat-free mass applied to bocly
accuracy and agreemcnt (compared with either the composition assessment ranges from 0.7194 to 0.7320.
reference methods/models or the mean of the two-com- The results of this study suggest that it is the latter value
ponent methods), this does not necessarily imply that a11 which is the more appropriate for groups of individuals
models based on whole-body potassium have this predis- with similar characterístics to those studied here (mean
position. The whole-body counter used in this study had hydration fraction calculated to be 0.7382, with no sig-
only two sodium iodide erystals, which yield relatively nifícant difference betwecn men amI women). Further-
poor counting preeision (coefficient of variation greater more, applying the formcr value (0.7194, as suggested by
than 5% [19]). calculation from the data of Siri [2J) to the deuteríum-
Thc application of bedside predietion melhods to dilution technique (in isolation) of body eomposition
estimate body composition has certain advantages over assessment is not appropriate in our group of subjeets
the more sophisticated primary lIlethods in terms of because ir overestimates fat-free mass relative to the four-
simplicity, eonvenience and relative eost. lrrespeetive of component model (for both men and women). However,
which rcference method is selected for eomparison, the despite lhe use of the most appropriate mean hydration
use of the sum of four skinfold thicknesses [11], measured fraetion for the water-dilution teehniques (two-compo-
by a single observer, has pro ved to be a better predictor of nent model), there may sti11be substantial error in body
body composition than other bedside prediction methods composition analysis because of inter-individual varia-
(equations based on BMI, bioelectrical impedance/resist- tion. Perhaps surprisingly, the hydration fractions of the
anee and near-i.r. interactanee) in this group of subjeets. fat-free mass assumed for reference man and rcferenee
Certain components of the body were not taken inlo woman [21] differ substantialJy (0.74 and 0.69, respee-
consideration in the three- and four-eomponent models. tively). To establish these values, the Task Group on
Of these, nucleic acids and glycogen probably represent Reference Mi.l11 [21] considered measurements of
the greatest fraetions. However, their mass in the whole water obtained from one group of individuals and
body is relatively low (approximately, 0.7 kg and 0.5 kg, measurements of fat-free mass from another. lt is
respectively, corresponding to about 1% and less than 1% sllggested that this approaeh is inappropríate, and may
body weight) and their densities (DNA, 1.42 kg/l; RNA, have been responsible for the apparent diserepancy. By
1.49 kg/l; glycogen, 1.52 kg/J) are not so very different the use of the more consistent experimental approach
from that of protein (1.34 kg/l), Hence, the errors adopted in this study, an appropriate value for the
involved as a result of neglecting nucleic aeids and hydration fraction of both men and women has been
glycogen from these models are eonsidered to be small, identified.
and, for the most part, these eomponents are likely to be Although many of the above difficulties may be over-
included with the protein fraction [4]. Similarly, amino come by the use of the four-component model, there is
acids (a 70 kg reference man contains about 250 g of free some concern that measurement errors arising from the
amino acids. the densities of which are both above and individual techniques may be propagated to produce a
belo\v 1.34 kg/l depending on the particular amino acid), large error in the final estimate of fat and fat-free mass.
urea (10-15 g in a 70 kg man, density 1.32 kg/l) and other However, the present analysis suggests that the individual
minor cornponents, sueh as uric acid and creatinille, are measurement errors are not substantia11y additive, ami
partitioned mostly wíth the protein fraction; this is that the overall measurement errors for both fat and fat-
because the densities o[ these eomponents are closer tu free mass amount to only about 1% 01' lcss of body weight
the density of protein (1.34 kg/l) than to that of fat (0.9 (whieh is largely clue to the measurement of body water
kg/l). and body volume). This error is as low 01' lower than
The four-component model also provides important measurement errors associated with the determinatíon of
information cOlleerning the likely variation in density of body [at using either densitometry or deuterium-dilution
the fat-free mass. In c1assical densitometry this value is in isolation.
generally assumed to be 1.1 kg/l. Deviations from thís The precision (SD) for the estima te of the hydration
value will result in errors in the predictioIl of body fractíon (± 0.0066) ami the density (± 0.0020 kg/l) of fat-
693

/ Body composition methodology


/ free mass was found to eorrespond to approximatcly one- REFERENCES
third of the inter-individual variation (± 0.02 and
1. Sartoris. D.). & Resnick, D. Dual-energy radiographic absorptiometry:
± 0.0073 kg/l, respeetively). This variability, whieh cannot current status and perspective. A.J.R.-Am. J. Roen,genol. 1989; 152, 241-6.
be aeeounted for by propagation oí measuremenl error, is 2. Siri, W.E. Body composition from lluid spaces and density: analysis of
eonsidered to be largely lIue lo biological variation in the methods. In: Brozek, j. & Henschel, A .• eds. T echniques for measuring body
relative proportions of eomponents of the fat-free mass, composition. Washington, D.c.: National Academy 01 Sdences N.R.e.,
1961: 223-44.
bul it may also be clue to the use of incorreel assumptions
J. Lohman, T.G. Research progre ss in validation 01 laboratory methods 01
in the four-eomponent model. These assumptions relate assessing body composition. Med. Sci. Sports Exerdse 1984; 16, 596-603.
to components whieh remain unaccounted for by the 4. Heymslield. S.B., Lichtman, S., Baumgartner, R.N. et al. Body composition of
Il1mlel (diseussed above), the densities of various eompo- humans: comparison 01 two improved four-compartment models that difler

nenls whieh are eonsidered in the model, the use oí a in expense, technical complexity. and radiation exposul'e. Am. J. Clin. Nutr.
1990; 52,52-8
factor for total-body water whieh is 1.04 times greater 5. Fulier, N.) .. Laskey, MA & E1ia. M. Assessment of the composition 01 major
than lhe measured deuterium-dilution space, and the body regions by dual-energy X-ray absorptiometry (DEXA), with special
\\hole-body ratio of osseous 'relativ~ to non-osseous reference to 11mb musde mass. Clin. Physiol. 1992; 12, 1-15.
mineral. Of these assumptions, the grealest level of 6. Akers, R. & Buskirk. E.R. An underwater weighing system utilizing "force

eonfidenee is assoeiated with the assumed densities of cube" transducers. ). Appl. Physiol. 1969; 26, 649-52.
7. Gnaedinger, R.H .• Reineke, E.P., Pearson, A.M .• Van Huss, W.D., Wessel, JA
water and fal. In eontrast, there is relatively less & Montoye, H.J. Detennination 01 body density by air displacement, helium
J eonfidenee in the ubiquitous applieation oí the remaining dilutlon, and underwater weighing. Ann. N.Y. Acad. Sci. 1963; 110,91-5.
assumptions. For example, sinee body proteins eomprise a 8. Puliicino, E.. Coward, W.A., Stubbs, R.J. & Elia. M. Bedside and field me,hods
for assessing body composition: comparison with the deuterium dilution
heterogeneous group of a large number of different
technique. Eur. J. Clin. Nutr. 1990: 44,753-62.
proteins with different densities, the propoI1ions in whieh 9. Coward, W.A., Parkinson, S.A. & Murgatroyd, P.R. Body composition
they are lo be founcl in the whole bocly may vary between measurements for nutrition research. Nutr. Res. Rev. 1988; 1,115-24.
c1ifferent individuals or in c1ifferent eireumstanees. 10. Burkinshaw, L Measurement 01 human body composition in vivo. Prog. Med.

However, mammalian muscle proteills, whieh are the Radiat. Phys. 1985; 2, 113-37.
11. Dumin, j.V.GA & Womersley,). Body fat assessed lrom total body density
most abundant proteins in the bocly, have a density and its estimation from skinfold thickness: measurements on 481 men and
estimatecl lo be between 1.33 amI 1.35 kg/I (calculatcd women aged 16-72 years. Br. J. Nutr. 1974; 32,77-97.
froll1 data taken from Mendez & Keys [22]). Collagen, 12. Fuller. N.J. & Elia, M. Potential use 01 bioelectrical impedance of the 'whole
whieh is estimated to aecounl for 25- 30% of lotal protein body' and 01 body segments for the assessment 01 body composition: com-
parison with densitometry and anthropometry. Eur. J. Clin. Nutr. 1989; 43,
in the human body (ll1ainly in bone ami skin), is eon- 779-91.
sidered lo have a densily of 1.36 kg/I [23J. These values 13. Elia, M., Parkinson, S.A. & Diaz, E. Evaluation 01 near infra-red interaetance
are c10se to the value for the density 0[" prolein applied lo as a method for predicting body composition. Eur. J. Clin. Nutr. 1990; 44,
lhe rnodels used in this study. 113-21.

It is eoncluded that DEXA provides a suitable aiterna- 14. Black, D" James, W.P.T., Besser, G.M. et al. Obesity. A report 01 the Royal
College 01 Physidans. J. R. ColI. Physicians London 1983; 17,5-65.
tÍve melhod lo densilomelry or deulerium dilution for lhe
15. Brozek, J., Grande, F.. Anderson, J,T. & Keys, A. Densitometrit analysis 01
assessll1ent of body composition. Furthermore, skinfold body eomposition: revision of so me quantitative assumptions. AnCl. N.Y.
thickness measuremenls predieted body eOll1position (as Acad. Sci. 1963;110, 113-40.

assessed by the referenee methods) better than any of the 16. Keys, A. & Brozek, J. Body ¡at in adult mano Physiol. Rev. 1953; 33,
245-325.
other bedside ll1ethocls. In additioll, values are proffered
17. Bland, j.M. & Altman, D.G. Statistical methods lor assessing agreement
(derived from lhe four-eomponent model) whieh repre- between two methods of clinical measurement. Lancet 19B6; 1, 307 -10.
scnt the mean density and hyclration fraetion of the fat- 18. 11urgatroyd, P.R. & Coward, W.A. An improved method for estimating
free mas s in healthy non-obese adults with similar changes in whole.body lat and protein mass in mano Br. J. Nutr. 1989; 62,
311-14.
eharacteristies to those in this study.
19. Parkinson, SA /n vivo measurement of changes in body composition [Ph.D.
Thesisj. Cambridge: University of Cambridge, 1990.
20. Schutte, j.E., Townsend, E.J.• Hugg. J. et al. Density 01 lean body mass is
greater in blaeks than whites. J. Appl. Physiol. 19B4; 56, 1647-9.
21. Snyder, W.S., Cook, M.J., Nasset, ES et al. Repon 01 the Task Group on
ACKNOWLEDGMENTS
Relerence Man. Oxlord: Pergamon Press, 1974: 27-32.
22. Mendez, J. & Keys. A. Density and composition of mamrnalian musde.
\Ve are decply indebled lo Mr K. F. Szaz, Mr P. R. Metab. Clin. Exp. 1960; 9, 184-8.
Murgatroyd and Dr T. J. Cole for hclp aml adviee 23. Hulmes. D.J.S. & Miliar, A. Quasi-hexagonal molec"lar packing in collagen
lhroughoul the eourse or this study. fibrils. Nature (London) 1979; 282, 878-BO.

Похожие интересы