| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Departments of Exercise Science and Foods and Nutrition, University of Georgia, Athens, Georgia 30602-3654
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
FOOTNOTES
REFERENCES
ABSTRACT 
Prior, Barry M., Kirk J. Cureton, Christopher M. Modlesky,
Ellen M. Evans, Mark A. Sloniger, Michael Saunders, and Richard D. Lewis. In vivo validation of whole body composition estimates from
dual-energy X-ray absorptiometry. J. Appl.
Physiol. 83(2): 623-630, 1997.
We validated whole
body composition estimates from dual-energy X-ray absorptiometry (DEXA)
against estimates from a four-component model to determine whether
accuracy is affected by gender, race, athletic status, or
musculoskeletal development in young adults. Measurements of body
density by hydrostatic weighing, body water by deuterium dilution, and
bone mineral by whole body DEXA were obtained in 172 young men
(n = 91) and women
(n = 81). Estimates of body fat
(%Fat) from DEXA (%FatDEXA)
were highly correlated with estimates of body fat from the
four-component model [body density, total body water, and total
body mineral (%Fatd,w,m);
r = 0.94, standard error of the
estimante (SEE) = 2.8% body mass (BM)] with no significant
difference between methods [mean of the difference ± SD of
the difference = 
0.4 ± 2.9 (SD) % BM,
P = 0.10] in women and men. On
the basis of the comparison with
%Fatd,w,m, estimates of
%FatDEXA were slightly more
accurate than those from body density
(r = 0.91, SEE = 3.4%; mean of the difference ± SD of the difference = 1.2 ± 3.4% BM).
Differences between %FatDEXA and
%Fatd,w,m were weakly related to
body thickness, as reflected by BMI (r = 0.34), and to the percentage of water in the fat-free mass
(r = 0.51), but were not
affected by race, athletic status, or musculoskeletal development. We
conclude that body composition estimates from DEXA are accurate
compared with those from a four-component model in young adults who
vary in gender, race, athletic status, body size, musculoskeletal
development, and body fatness.
body composition; body water; bone mineral; densitometry; women; gender; men; multicomponent models; musculoskeletal development; race
INTRODUCTION BODY COMPOSITION is important in the assessment of
nutritional status, disease risk, physical fitness, and effectiveness
of interventions (11). Traditional methods of estimating body
composition such as densitometry, hydrometry, and
40K counting are limited in their
accuracy, availability, and suitability for all populations. Estimates
of body composition from multicomponent models are more accurate than
those from traditional methods but are time consuming and not practical
for large groups. Dual-energy X-ray absorptiometry (DEXA) is widely
used for measuring bone mineral, and technological advancements have
enabled DEXA to quickly become a preferred method for estimating whole
body composition. Body composition measurements from DEXA are
relatively quick, easy, and have good precision, but their validity has
not been adequately established (21, 25).
Although the accuracy of DEXA has been evaluated by chemical analysis
(5, 37) and in vitro validation (38, 39), in vivo validation is
required to assess potential errors in whole body composition estimates
caused by regional variation in tissue thickness, the water content of
lean tissue (nonbone, nonfat) mass, and the difficulty in estimating
the composition of nonbone tissue above and below bone. These potential
errors are not adequately addressed by in vitro or chemical analysis
studies (21). Several studies have validated body composition estimates
from DEXA against estimates from body density on the basis of a
two-component model (8, 17, 36, 41). However, body composition
estimated from body density is affected by errors resulting from
variation in the density of the fat-free mass (FFM) and is no longer
considered acceptable as a criterion measure (13). Body composition
estimated from a four-component model (fat, water, mineral, protein)
accounts for variation in the water and mineral fractions and the
density of the FFM and, therefore, provides a better criterion measure for validating newer methods of estimating body composition (23). Only
a few studies have used body composition estimates from a four-component model to validate body composition estimates from DEXA
(3, 12, 18). These studies have not included athletes who represent
extremes in body size, musculoskeletal development, and body
composition. Thus the purpose of this study was to validate whole body
estimates of percent body fat from DEXA
(%FatDEXA) with
estimates from a four-component model [percentage of body mass
(BM) that was fat estimated from body density, total body water, and
total body mineral;
%Fatd,w,m] in a large group
of young, black and white, athletic and nonathletic, men and women who
varied markedly in body size, musculoskeletal development, and
proportion of fat. Estimates of body fat from body density (%Fatd) were also compared with
those from the four-component model and from DEXA to assess the
relative accuracy of estimates from DEXA and body density. We
hypothesized that the agreement between
%FatDEXA and
%Fatd,w,m would be better than
the agreement between %FatDEXA
and %Fatd and between
%Fatd and
%Fatd,w,m.
METHODS
Table 1.
Subject characteristics
Measurement
Women (n = 81)
Men
(n = 91)
Age, yr
20.7 ± 2.6
21.2 ± 2.1
Height, cm
165.4 ± 6.9
182.9 ± 7.9
Mass, kg
61.5 ± 11.0
91.0 ± 20.8
Mesomorphy
4.1 ± 1.3
6.2 ± 1.7
BMI, kg/m2
22.5 ± 3.8
27.0 ± 4.6
Db, g/cm3
1.047 ± 0.017
1.066 ± 0.012
Body water, kg
34.6 ± 5.1
58.4 ± 12.7
%Fatd
22.7 ± 7.7
14.4 ± 5.4
%FatDEXA
22.5 ± 8.4
13.1 ± 5.7
%Fatd,w,m
22.3 ± 7.6
12.5 ± 5.9
Fat mass, kg*
14.2 ± 7.6
11.9 ± 7.4
Fat-free mass, kg*
47.3 ± 6.5
79.2 ± 16.3
Values are means ± SD; n, no. of subjects. BMI, body
mass index; Db, body density measured by hydrostatic
weighing; %Fatd, %FatDEXA, and
%Fatd,w,m: fat estimated from body density; dual-energy x-ray absorptrometry (DEXA); and body density, total body water, and
total body mineral, respectively.
*
Estimated from
%Fatd,w,m.
Significantly different from
%Fatd,w,m.
3 (27).
Body water.
Total body water was determined by deuterium
(2H2O) dilution according to the
technique of Davis et al. (9). After a baseline blood sample was taken,
subjects ingested a known quantity of 2H2O [0.31 ± 0.01 (SD)
g 2H2O/kg BM] in 100 ml of
distilled water and rinsed with an additional 100 ml of distilled
water. After a 3-h equilibration period during which all urine was
collected, another blood sample was taken. Blood samples were
centrifuged, and the plasma was stored at 80°C. Plasma
samples were purified by diffusion (10). In 12 of the subjects, blood
samples could not be obtained and urine was purified instead. The
purified sample was analyzed by using an infrared spectrophotometer
(Miran 1FF, Foxboro, Foxboro, MA). Total body water was corrected for
2H2O lost in urine and decreased
4% to account for hydrogen exchange with protein and carbohydrate
during the 3-h equilibration period (33). The within-subjects SD of
duplicate measurements of body water on 2 days ~1 wk apart was 0.75 liter. The within-subjects SD of 2-4 measurements of body water on
the basis of the same blood sample on a given day was 0.71 liter (27).
We considered this a measure of the technical error associated with the
measurement of total body water.
Body mineral.
Total body mineral was estimated from bone mineral ash measured by
using DEXA (Hologic QDR-1000W, enhanced whole body analysis software
version 5.71, Waltham, MA). Bone mineral ash was multiplied by 1.2741 to estimate total body mineral content (4). Thirty-one subjects were
taller than the scanning region of the bone densitometer. Body
composition and bone ash for these subjects were estimated from the
summed values from two separate scans (upper and lower body divided at
the neck) by using regression. This method has been shown to yield
accurate estimates of BM
[r = 0.99, standard error of the estimate (SEE) = 338 g], %Fat
(r = 0.99, SEE = 0.2% BM),
and bone ash (r = 0.99, SEE = 29 g)
(B. M. Prior, C. M. Modlesky, R. D. Lewis, and K. J. Cureton,
unpublished observations). The within-subjects SD of
duplicate measurements of body mineral on 2 days ~1 wk apart was 7.2 g (27).
Musculoskeletal development.
Musculoskeletal development per unit height was characterized by using
the mesomorphy rating from the Heath-Carter anthropometric somatotype.
This rating was determined from upper arm and calf circumferences
corrected for skinfold thickness, knee and elbow diameters, and height,
following procedures described by Carter (6). Mesomorphy was determined
from the measurements by using a regression equation (7).
Body composition calculations.
The percentage of BM that was fat was estimated from body density
(i.e., %Fatd) on the basis of a
two-component model by using the Siri equation (34)
![%Fat<SUB>d</SUB> = [(4.95/D<SUB>b</SUB>) − 4.50]100](/content/vol83/issue2/fulltext/623/img001.gif)
|
![%Fat<SUB>d,w,m</SUB> = [(2.747/D<SUB>b</SUB>) − (0.714w) + (1.146m) − 2.0503]100](/content/vol83/issue2/fulltext/623/img002.gif)
|
[
(Y
Y2)/n],
where Y is %FatDEXA or is
estimated from densitometry, and Y is
%Fatd,w,m} (8). An
experiment-wise 
level of 0.05 was used for all significance tests.
Significance values were corrected for multiple comparisons by using
the Bonferroni method.
RESULTS
Individual differences between scale BM and BM estimated by DEXA ranged
from 0.9 to 3.1 kg (Fig. 1). The
largest deviations were in two subjects with the greatest BM. Mean
total BM estimated by DEXA was significantly less than scale BM by 0.6 ± 0.5 kg. Total BM estimated by DEXA was highly correlated to scale
BM (r = 1.0, SEE = 0.5 kg). The
relationship of total BM estimated by DEXA to scale BM was similar in
men (y = 1.01x 0.2, r = 1.0, SEE = 0.5 kg)
and women (y = 1.01x + 0.03, r = 1.0, SEE = 0.4 kg). Because of the
close agreement between scale BM and that estimated by DEXA, only
results for %Fat estimates are presented.
, Women
[y = 1.01x + 0.03, r = 1.0, standard error of estimate
(SEE) = 0.37 kg]; 
, men (y = 1.01x 0.24, r = 1.0, SEE = 0.52 kg). Solid line,
line of identity.
Individual differences between
%Fatd,w,m and
%FatDEXA ranged from
9.9 to 7.5% BM (Fig.
2A).
Mean %FatDEXA was not
significantly different from
(
diff ± SDdiff = 0.4 ± 2.9% BM;
P = 0.10) and highly related
(r = 0.94, SEE = 2.8% BM)
to %Fatd,w,m. TE predicting
%Fatd,w,m from
%FatDEXA (2.9% BM) was only
slightly different from the SEE, indicating little systematic
difference in the regression equation from the line of identity. The
relationship of %FatDEXA to
%Fatd,w,m was slightly different
in men (y = 0.90x + 0.75, r = 0.87, SEE = 2.9% BM) and women
(y = 0.85x + 3.30, r = 0.94, SEE = 2.6% BM, Fig.
3). There was no significant difference in
the slopes of the regression lines, but the intercept for women was
significantly greater than that for men. The gender difference appeared
to result primarily from overestimation of
%Fatd,w,m by
%FatDEXA in the fattest women.
Race, athletic status (athlete/nonathlete), and musculoskeletal
development did not significantly affect the relationship of
%FatDEXA to
%Fatd,w,m.
, Men; , women. Solid
line, mean difference; dashed lines, ±2 SD.
Mean %FatDEXA was 0.8 ± 3.0%
BM less than %Fatd, with
individual differences ranging from 8.4 to 9.3% BM (Fig.
2B).
%FatDEXA was highly related to
%Fatd
(r = 0.94, SEE = 2.8% BM),
with no difference in the relationship between men
(y = 0.83x + 3.54, r = 0.88, SEE = 2.6% BM) and
women (y = 0.84x + 3.92, r = 0.92, SEE = 3.0% BM). Across all
subjects, TE predicting %Fatd
from %FatDEXA (3.1%) was only
slightly higher than the SEE. Gender, race, and athletic status, in
addition to %FatDEXA, were
significant predictors of %Fatd
and probably accounted for the slight deviation from the line of
identity.
To assess the relative accuracy of body composition estimates from DEXA
and those from densitometry, estimates of
%Fatd were compared with
%Fatd,w,m. Mean
%Fatd was significantly greater than %Fatd,w,m
diff ± SDdiff = 1.2 ± 3.4% BM), with differences ranging from 8.1 to 8.5% BM (Fig.
2C).
%Fatd was highly correlated with
%Fatd,w,m
(r = 0.91, SEE = 3.4% BM), but there
was a slight difference in the relationship of
%Fatd to %Fatd,w,m between men
(y = 0.96x 1.42, r = 0.89, SEE = 2.7% BM) and women (y = 0.87x + 2.62, r = 0.87, SEE = 3.7% BM).
The intercept for women was significantly greater than that for men. For all subjects, TE predicting
%Fatd,w,m from
%Fatd (3.6%) was slightly
greater than the SEE.
Body mass index (BMI), used as an indicator of body thickness, was
moderately correlated with differences between
%Fatd,w,m and
%FatDEXA
(y = 0.21x + 4.82, r = 0.34, SEE = 2.75% BM, Fig. 4A).
Water content of the fat-free mass (W/FFM), used to represent hydration
of lean tissue mass, was more strongly correlated with differences
between %Fatd,w,m and
%FatDEXA
(y = 0.55x + 39.77, r = 0.51, SEE = 2.5% BM, Fig.
4B).
%FatDEXA to body mass index
(BMI). Regression equation: y = 0.21x + 4.82, r = 0.34, SEE = 2.75% BM.
B: relationship of
%Fatd,w,m %FatDEXA to water content of
fat-free mass (W/FFM), expressed as %FFM. Regression equation: y = 0.55x + 39.77, r = 0.51, SEE = 2.51% BM. Symbols are
defined as in Fig. 2. Solid line, line of identity.
DISCUSSION
We used estimates of body composition from a four-component model as a criterion measure to validate body composition estimates from DEXA in a large, diverse group of men and women that included blacks and whites, athletes and nonathletes, and marked variability in body size, musculoskeletal development, and body fatness. We found that %FatDEXA agreed well with %Fatd,w,m in men and women. The agreement of %Fatd,w,m with %FatDEXA was better than with %Fatd, suggesting that estimates with %FatDEXA were slightly more accurate than those using %Fatd. Differences between %FatDEXA and %Fatd,w,m were weakly related to body thickness, as reflected by BMI, and to W/FFM, but they were not affected by race, athletic status, or musculoskeletal development. The results indicate that body composition estimates from DEXA are accurate in young adults, who vary in gender, race, athletic status, body size, musculoskeletal development, and body fatness.
The major finding of this study was that %FatDEXA (17.5 ± 8.5% BM) agreed well with %Fatd,w,m (17.1 ± 8.3% BM) in young, black and white, men and women over a wide range of body size (BM 42.9 to 133.5 kg, BMI 17.1 to 41.2 kg/m2), musculoskeletal development (mesomorphy 1.6 to 9.6), and %Fat (3 to 50% BM). Lohman (25) has suggested that evaluation of the accuracy of a new method of estimating %Fat should be on the basis of the SEE of predicting a criterion estimate of %Fat, with SEEs <3% BM indicating good accuracy of the new method. On the basis of Lohman's criteria, body composition estimates from DEXA were accurate (SEE = 2.8% BM). Another indicator of predictive accuracy is the TE (22). A regression equation may have a small SEE but a large TE, indicating a systematic error in the equation (8). Lohman (22) has suggested that regression equations with TE >3.3% BM are inaccurate. In this study, TE for predicting %Fatd,w,m from %FatDEXA was only slightly greater than the SEE (2.9% BM), indicating little systematic error in the prediction and good accuracy.
Several studies have validated
%FatDEXA against
%Fatd on the basis of a
two-component model. Snead et al. (36) found
%FatDEXA to be quite inaccurate
when validated against %Fatd, as
reflected by a relatively large SEE (3.8% BM) in a large group of men
and women aged 21-81 yr. On the basis of mean differences, it
appeared that DEXA was more accurate in younger adults 21-39 yr
(diff = ~1% BM) than in older adults >60 yr
(diff = ~5% BM). These results suggested that the lack of accuracy may have
been due to an underestimation of
%FatDEXA, especially in the
truncal region of older individuals, a problem that may have been at
least partially corrected in later software revisions (25). Wellens et
al. (41) found somewhat better accuracy in 78 women (SEE = 3.2% BM)
and very good accuracy in 50 men (SEE = 2.3% BM) aged 18-67 yr.
These authors suggested that differences between
%Fatd and
%FatDEXA may have been due
to variation in the density and composition of the FFM, affecting
%Fatd estimates, beam-hardening effects, and error in soft tissue analysis by DEXA. Error in soft tissue analysis was hypothesized to have affected
%FatDEXA in men to a greater
extent than in women. This is surprising because, on the basis of
regression analysis, %FatDEXA was
more accurate in men than women. Clark et al. (8) found poorer accuracy
in 35 men aged 22-75 yr. The SEE of predicting
%Fatd from
%FatDEXA was 3.0% BM, but the
mean difference between %Fatd and
%FatDEXA was large (3.9 ± 1.2% BM). The TE of predicting
%Fatd from
%FatDEXA was 5.2% BM, indicating
that there was a systematic difference between
%Fatd and
%FatDEXA and that DEXA was not
highly accurate. Differences between
%Fatd and
%FatDEXA were attributed to
errors inherent in the DEXA software and variation in the density of the FFM. Hansen et al. (17) found
%FatDEXA to be very accurate when
predicting %Fatd (SEE = 2.4% BM)
in 100 women aged 28-39 yr. Unlike Clark et al. (8), they found a
small mean difference between
%Fatd and
%FatDEXA (0.2% BM), emphasizing
the good agreement between methods. However, the regression equation
predicting %Fatd from
%FatDEXA
(y = 0.77x + 6.96) had a large intercept and a slope substantially less than one,
indicating that there were differences as values deviated from the
means. TE of the prediction was not reported.
In recent years, the use of %Fatd
as a criterion method for validating other body composition methods has
been challenged because it assumes a constant density and composition
of the FFM (13, 18, 24). In theory,
%Fatd,w,m should be a better
criterion measure for validating body composition estimates from DEXA
because it accounts for the effects of variation in water and mineral and their effects on the density of the FFM. Several studies have compared %FatDEXA estimates with
those from multicomponent models. Boileau et al. (3) found reasonably
good agreement between %FatDEXA and
%Fatd,w,m (SEE = 3.1% BM) in a
large group of men and women (n = 308), ranging in age from 8 to 75 yr; however, it is not known whether
there was greater disagreement at younger ages compared with older
ages, as some investigators have found (16). In a smaller study
focusing on the reliability of %Fat estimates from a four-component
model, Friedl et al. (12) presented body composition estimates from 5 different methods in 10 subjects. Using their data, we compared
%FatDEXA and
%Fatd,w,m estimates and found
that %FatDEXA agreed
very well with %Fatd,w,m
(diff ± SDdiff = 0.4 ± 1.9% BM; SEE 1.9% BM). Modlesky et al. (35) found good
agreement between %FatDEXA and
%Fatd,w,m
(diff ± SDdiff = 0.4 ± 2.5% BM)
across two groups of men who differed in musculoskeletal development.
In a more sophisticated approach, Heymsfield et al. (18) and Wang et
al. (40) compared body composition estimates from dual-photon
absorptiometry (similar to DEXA) and DEXA to a multicomponent model on
the basis of in vivo neutron-activation analysis (IVNA). In men and
women aged 24-93 yr (n = 31),
Heymsfield et al. (18) found that fat mass estimated by IVNA agreed
well with that measured by dual-photon absorptiometry (SEE = 1.5 kg).
Similarly, Wang et al. (40) found no difference between
%FatDEXA and %Fat estimated from
IVNA in a larger (n = 65) group of
black and white men and women. In three of the four subgroups, SEEs
ranged from 2.6 to 3.2% BM, indicating good agreement between methods.
However, in white women there was substantially less agreement between
methods (SEE = 5.18% BM). The close agreement between IVNA and DEXA or
dual-photon absorptiometry is important because body composition
estimates by IVNA are independent of those by DEXA, whereas
%Fatd,w,m is dependent on bone
mineral measured by DEXA.
We hypothesized that the agreement between %Fatd,w,m and %FatDEXA would be better than the agreement between %Fatd,w,m and %Fatd. Our hypothesis was based on the fact that variation in the water (i.e., W/FFM) and mineral (M/FFM) contents of the FFM introduces potentially large errors in %Fatd compared with methods that account for W/FFM and M/FFM variation, e.g., %Fatd,w,m (B. Prior, K. J. Cureton, C. M. Modelsky, E. M. Evans, M. A. Sloniger, M. J. Saundes, and R. D. Lewis, unpublished observations). Body composition estimates by DEXA are unique in that they should not be affected by variation in M/FFM and are less affected by variation in W/FFM (21). We found that %FatDEXA agreed well with %Fatd,w,m (SEE = 2.8% BM) but that there was substantially less agreement between %Fatd and %Fatd,w,m (SEE = 3.4% BM). The smaller SEE between %FatDEXA and %Fatd,w,m indicated that body composition estimates from DEXA were slightly more accurate than those from body density.
We also hypothesized that the agreement between %FatDEXA and %Fatd,w,m would be better than the agreement between %FatDEXA and %Fatd because there is less biological error associated with the density and composition of the FFM for %Fatd,w,m than for %Fatd. A surprising finding was that this was not the case. The SEEs associated with prediction of %Fatd,w,m and %Fatd from %FatDEXA were the same (2.8% BM). Similarly, the SDdiff between %FatDEXA and %Fatd,w,m (2.9% BM) was nearly the same as that between %FatDEXA and %Fatd (3.0% BM). While surprising, this finding is not inconsistent with the literature. Studies using body composition estimates from body density to validate estimates from DEXA have found SEEs ranging from 2.4 to 3.9% BM (8, 17, 36, 41), whereas studies using body composition estimates from multicomponent models as criterion measures have found SEEs ranging from 1.9 to 3.2 (3, 12, 40).
Errors associated with %FatDEXA
may contribute to differences between this estimate and
%Fatd,w,m and
%Fatd. In estimating body
composition from DEXA, it is assumed that the nonbone, nonfat, soft
lean tissue mass has a constant hydration of 73 g/ml. Any deviation
from the assumed value will cause systematic error in %FatDEXA (32). The magnitude of
the error, however, has not been established. Horber et al. (19) found
that of 0.9-4.4 kg of BM lost during hemodialysis, nearly all
(94%) was correctly identified as lean tissue and only a small
(0.1-kg) portion was incorrectly identified as fat tissue. In that
study, ingestion of 0.8-2.4 liters of water did not appreciably
affect estimates of fat or lean tissue mass measured by DEXA, although
there was considerable variation. Similarly, in a dehydration study,
Going et al. (14) found that DEXA correctly identified 98% of the total mass lost as lean tissue mass, whereas there were no changes in
fat mass or bone mass. Theoretical calculations by Kohrt (21) support
these findings; a 5% change in W/FFM would, according to Kohrt, result
in an error in fat mass by DEXA of <0.5 kg. In this study, W/FFM was
moderately related to the difference between %FatDEXA and
%Fatd,w,m
(r = 0.51; Fig.
4B). A difference in W/FFM should
reflect differences in lean tissue mass hydration. Regression analysis
indicated that a 5% change in W/FFM resulted in a 2.7% BM (0.4 kg
fat) change in the difference between
%FatDEXA and %Fatd,w,m. Although our findings
indicate a relatively small effect of lean tissue mass hydration on
%FatDEXA, variation in lean
tissue mass hydration may have accounted for part of the difference
among %FatDEXA,
%Fatd,w,m, and
%Fatd.
Error in %FatDEXA estimates could result from inaccurate detection of fat mass in the trunk. Snead et al. (36) found that when packets of lard were placed on the thighs of subjects, DEXA measured all of the added mass and correctly identified 96% of it as fat. However, when the lard was placed over the truncal region of the body, 93% of the added mass was detected, but only 55% was correctly identified as fat, with the remainder being identified as lean tissue mass. This finding suggests that DEXA does not accurately measure fat in the trunk. Similar results were reported by Milliken et al. (26), who found that DEXA correctly estimated the %Fat of a lard packet placed on the thighs of 28 men and women but significantly underestimated %Fat of the lard packet (52 vs. 90% actual) when it was placed on the abdomen. Furthermore, when two lard packets were placed on top of each other, DEXA significantly underestimated %Fat in both the thigh (77%) and abdomen (47%). These findings suggest that DEXA may be in error to a greater extent in fatter subjects. Error in the detection of fat mass in the trunk appears to result from the method used by DEXA to determine soft tissue composition in tissue above and below bone. DEXA uses the average soft tissue composition surrounding bone to estimate the soft tissue composition directly above and below bone. Because of the quantity of bone in the trunk, DEXA may not accurately estimate soft tissue composition and fat mass in the trunk region. In the present study, however, an inability to accurately measure fat mass in the trunk would cause %FatDEXA to be underestimated relative to %Fatd,w,m, particularly in the fattest subjects. This did not appear to occur; rather, %FatDEXA appeared to slightly overestimate %Fatd,w,m. Thus inaccurate estimates of truncal fat mass by DEXA did not appear to contribute to the differences between %FatDEXA and the other methods.
Another source of error in
%FatDEXA estimates results from
changes in body (tissue) thickness. Increasing object thickness causes
preferential loss of lowerenergy photons relative to high-energy photons, an effect called beam hardening. The resulting altered attenuation constants for the various tissues can cause error in body
composition estimates by DEXA (1). In the present study, we used BMI as
an indicator of body thickness, assuming that higher values of BMI
would correspond to greater body thickness. The negative relationship
between BMI and the difference between
%Fatd,w,m and
%FatDEXA
(r = 0.34; Fig.
4A) suggests that as body thickness increased, DEXA increasingly overestimated
%Fatd,w,m. Our findings are in
contrast to those of Tothill et al. (38) and Jebb et al. (20), who
found that increasing phantom thickness had little effect on %Fat
estimates from DEXA. Furthermore, the effect of beam hardening, e.g.,
increasing object thickness, is that fat mass is underestimated
relative to true fat mass or content (38). Thus, if beam hardening were
contributing to error in
%FatDEXA, the expected effect
would be opposite to what we observed.
Estimates of %Fat from DEXA are also affected by the type of bone densitometer used. For example, Tothill et al. (38) found that measurements of %Fat in a phantom by using a Hologic densitometer were, on average, 2% BM lower than those from a Lunar. Similarly, Modlesky et al. (28) found that a Hologic densitometer underestimated %Fat measured by a Lunar densitometer by 1.5% BM in 13 young men. These findings indicate that although some of the difference between %FatDEXA and %Fatd,w,m in the present study may be a function of the type of densitometer used, the close agreement with %Fatd,w,m suggests the error is not large.
We have based the preceding discussion on the assumption that any difference between %FatDEXA and %Fatd,w,m resulted from error in the estimate of %FatDEXA. However, estimates of %Fatd,w,m are not error free. Day-to-day biological variation and technical errors in each of the measurements used to estimate %Fatd,w,m (body density, bone mineral content, and total body water) contribute to day-to-day differences in %Fatd,w,m. Our laboratory (27) and others (12) have reported that these differences are small, with the within-subjects SD for replicate determinations of %Fatd,w,m being ~1% BM.
Other systematic errors, not reflected in repeated measurements in an individual, may also contribute to errors in %Fatd,w,m. For example, it could be that increasing body thickness resulted in error in %Fatd,w,m estimates. Such an error could occur if bone mineral estimates systematically varied with body thickness. However, Tothill et al. (39), Jebb et al. (20), and Milliken et al. (26) found that bone mineral estimates varied little as phantom thickness increased. Thus it is not likely that %Fatd,w,m estimates were affected by a systematic error due to body thickness.
Errors in the total body water measurement in selected individuals could have contributed to error in %Fatd,w,m. Replicate measurements of total body water in a small group of subjects indicated that day-to-day differences within individuals were quite small (SD = 0.75 liter). Most of this variation was due to the technical error of measuring total body water (0.71 liter). This technical error corresponds to an error of 0.7% BM in the estimate of %Fatd,w,m and an error of 0.9% in W/FFM. Assuming that the total variance in these measurements is the sum of biological variability and technical error [(SDtot)2 = (SDbiol)2 + (SDtech)2], these values amount to 6% of the variance of %Fatd,w,m (SD = 8.8% BM) and 30% of the variance of W/FFM (SD = 2.7% FFM) in all subjects. Thus technical error in measuring total body water appears to explain some, but not a predominate proportion, of the differences between %FatDEXA and %Fatd,w,m.
The similar agreement of %FatDEXA with %Fatd,w,m and %Fatd, despite the poorer agreement of %Fatd,w,m with %Fatd than with %FatDEXA, is difficult to understand. It suggests that there is some shared error between %FatDEXA and %Fatd such that differences between estimates of %FatDEXA and %Fatd are less than those between %Fatd and %Fatd,w,m. Indeed, the differences from the criterion %Fatd,w,m of %Fatd and %FatDEXA were positively correlated (r = 0.55, P < 0.001). We reported that variation in the difference between %Fatd and %Fatd,w,m for the present sample was completely explained by variation in the density of the FFM, which in turn was mostly explained by variation in W/FFM and M/FFM (B. Prior, K. J. Cureton, C. M. Modelsky, E. M. Evans, M. A. Sloniger, M. J. Saundes, and R. D. Lewis, unpublished observations). %FatDEXA estimates should not affected by variability in M/FFM. Thus the shared error is probably associated with variability in W/FFM. This is supported by significant correlations of W/FFM with deviations of %FatDEXA (r = 0.51) and %Fatd (r = 0.95) from %Fatd,w,m. On the basis of the discussion above, about one-third of this shared variation is due to technical error associated with the measurement of body water.
In summary, we have found that body fatness estimates by DEXA are accurate on the basis of close agreement with criterion estimates from a four-component model in a large sample of young men and women, including blacks and whites, athletes and nonathletes, and marked variability in body size, musculoskeletal development, and body fatness. Across all subjects, there was no systematic error in the prediction of %Fatd,w,m from %FatDEXA and a reasonably small SEE (2.8% BM). Small effects of variation in lean tissue mass hydration and body thickness may explain part of the difference between body composition estimates by DEXA and those from the four-component model. Body composition estimates from DEXA were slightly more accurate than those from body density compared with those from the four-component model. We conclude that body composition estimates from DEXA are accurate compared with those from a four-component model in young adults. The accuracy is slightly better in women than in men but is not affected by race, athletic status, body size, musculoskeletal development, and body fatness.
FOOTNOTES
Address for reprint requests: B. Prior, Dept. of Physiology, Giltner Hall, Michigan State Univ., East Lansing, MI 48824 (E-mail: bprior{at}pslsun.psl.msu.edu).
Received 21 November 1996; accepted in final form 15 April 1997.
REFERENCES
| 1. | Blake, G. M., D. B. McKeeney, S. C. Chhaya, P. J. Ryan, and I. Fogelman. Dual energy X-ray absortiometry: the effects of beam hardening on bone density measurements. Med. Phys. 19: 459-465, 1992[Medline]. |
| 2. | Bland, J. M., and D. G. Altman. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1: 307-310, 1986[Medline]. |
| 3. | Boileau, R. A., M. H. Slaughter, R. J. Stillman, C. B. Christ, J. Clasey, G. Dykstra, K. Farrell-Lee, D. Giddens, S. Sunvold, and R. Whalen. Precision of fat and fat-free body (FFB) composition estimates: comparison of four methods (Abstract). FASEB J. 8: A179, 1994. |
| 4. | Brozek, J., F. Grande, J. T. Anderson, and A. Keys. Densitometric analysis of body composition: revision of some quantitative assumptions. Ann. NY Acad. Sci. 110: 113-140, 1963. |
| 5. |
Brunton, J. A.,
H. S. Bayley,
and
S. A. Atkinson.
Validation and application of dual-energy X-ray absorptiometry to measure bone mass and body composition in small infants.
Am. J. Clin. Nutr.
58:
839-845,
1993 |
| 6. | Carter, J. E. L. The Heath-Carter anthropometric somatotype. In: The Heath-Carter Somatotype Method, edited by J. E. L. Carter. San Diego, CA: San Diego State University, 1972, p. A1-A8. |
| 7. |
Carter, J. E. L.,
and
B. H. Heath.
SomatotypingDevelopment and Applications. Cambridge, UK: Cambridge Univ. Press, 1990, p. 374.
|
| 8. | Clark, R. R., J. M. Kuta, and J. C. Sullivan. Prediction of percent body fat in adult males using dual energy X-ray absorptiometry, skinfolds, and hydrostatic weighing. Med. Sci. Sports Exerc. 25: 528-535, 1993[Medline]. |
| 9. |
Davis, J. M.,
D. R. Lamb,
W. A. Burgess,
and
W. P. Bartoli.
Accumulation of deuterium oxide in body fluids after ingestion of D2O-labeled beverages.
J. Appl. Physiol.
63:
2060-2066,
1987 |
| 10. | DeLuca, J. P., and K. E. Friedl. Validation of a simplified method of total body water measurement using deuterium oxide and diffusion dish purification (Abstract). Med. Sci. Sports Exerc. 24, Suppl: S8, 1992. |
| 11. | Forbes, G. B. Human Body Composition: Growth, Aging, Nutrition, and Activity. London: Springer-Verlag, 1987, p. 153-195. |
| 12. |
Friedl, K. E.,
J. P. DeLuca,
L. J. Marchitelli,
and
J. A. Vogel.
Reliability of body-fat estimations from a four-component model by using density, body water, and bone mineral measurements.
Am. J. Clin. Nutr.
55:
764-770,
1992 |
| 13. | Going, S. B. Densitometry. In: Human Body Composition, edited by A. F. Roche, S. B. Heymsfield, and T. G. Lohman. Champaign, IL: Human Kinetics, 1996, p. 3-24. |
| 14. |
Going, S. B.,
M. P. Massett,
M. C. Hall,
L. A. Bare,
P. A. Root,
D. P. Williams,
and
T. G. Lohman.
Detection of small changes in body composition by dual-energy x-ray absorptiometry.
Am. J. Clin. Nutr.
57:
845-850,
1993 |
| 15. | Goldman, R. F., and E. R. Buskirk. Body volume measurement by underwater weighing: description of a method. In: Techniques for Measuring Body Composition, edited by J. Brozek. Washington, DC: National Academy of Sciences, 1959, p. 78-89. |
| 16. | Guo, S. S., W. C. Chumlea, X. Wu, R. Wellens, A. F. Roche, and R. M. Siervogel. A comparison of body composition models. In: Human Body Composition: In Vivo Methods, Models, and Assessment, edited by K. J. Ellis, and J. D. Eastman. New York: Plenum, 1993, p. 27-30. |
| 17. |
Hansen, N. J.,
T. G. Lohman,
S. B. Going,
M. C. Hall,
R. W. Pamenter,
L. A. Bare,
T. W. Boyden,
and
L. B. Houtkooper.
Prediction of body composition in premenopausal females from dual-energy X-ray absorptiometry.
J. Appl. Physiol.
75:
1637-1641,
1993 |
| 18. |
Heymsfield, S. B.,
S. Lichtman,
R. N. Baumgartner,
J. Wang,
Y. Kamen,
A. Aliprantis,
and
R. N. Pierson, Jr.
Body composition of humans: comparison of two improved four-compartment models that differ in expense, technical complexity, and radiation exposure.
Am. J. Clin. Nutr.
52:
52-58,
1990 |
| 19. | Horber, F. F., F. Thomi, J. P. Casez, J. Fonteille, and P. Jaeger. Impact of hydration status on body composition as measured by dual energy X-ray absorptiometry in normal volunteers and patients on haemodialysis. Br. J. Radiol. 65: 895-900, 1992[Abstract]. |
| 20. | Jebb, S. A., G. R. Goldberg, G. Jennings, and M. Elia. Dual-energy X-ray absorptiometry measurements of body composition: effects of depth and tissue thickness, including comparisons with direct analysis. Clin. Sci. (Lond.) 88: 319-324, 1995[Medline]. |
| 21. | Kohrt, W. Body composition by DEXA: tried and true? Med. Sci. Sports Exerc. 27: 1349-1353, 1995[Medline]. |
| 22. | Lohman, T. G. Skinfolds and body density and their relation to body fatness: a review. Hum. Biol. 53: 181-225, 1981[Medline]. |
| 23. | Lohman, T. G. Applicability of body composition techniques and constants for children and youths. In: Exercise and Sport Sciences Reviews, edited by K. Pandolf. New York: Macmillan, 1986, p. 325-357. |
| 24. | Lohman, T. G. Advances in Body Composition Assessment. Champaign, IL: Human Kinetics, 1992, p. 1-150. |
| 25. | Lohman, T. G. Dual energy X-ray absorptiometry. In: Human Body Composition, edited by A. F. Roche, S. B. Heymsfield, and T. G. Lohman. Champaign, IL: Human Kinetics, 1996, p. 63-78. |
| 26. | Milliken, L. A., S. B. Going, and T. G. Lohman. Effects of variations in regional composition on soft tissue measurements by dual-energy X-ray absorptiometry. Int. J. Obes. 20: 677-682, 1996. |
| 27. |
Modlesky, C.,
K. Cureton,
R. Lewis,
B. Prior,
M. Sloniger,
and
D. Rowe.
Density of the fat-free mass and estimates of body composition in weight trainers.
J. Appl. Physiol.
80:
2085-2096,
1996 |
| 28. |
Modlesky, C. M.,
R. D. Lewis,
K. A. Yetman,
B. Rose,
L. B. Rosskopf,
T. K. Snow,
and
P. B. Sparling.
Comparison of body composition and bone mineral measurements from two DEXA instruments in young men.
Am. J. Clin. Nutr.
64:
669-676,
1996 |
| 29. | Pedhazur, E. J. Multiple Regression in Behavioral Research. Fort Worth, TX: Holt, Rinehart and Winston, 1982, p. 436-450. |
| 32. |
Roubenoff, R.,
J. J. Kehayias,
B. Dawson-Hughes,
and
S. B. Heymsfield.
Use of dual-energy X-ray absorptiometry in body composition studies: not yet a gold standard.
Am. J. Clin. Nutr.
58:
589-591,
1993 |
| 33. | Schoeller, D. A. Hydrometry. In: Human Body Composition, edited by A. F. Roche, S. B. Heymsfield, and T. G. Lohman. Champaign, IL: Human Kinetics, 1996, p. 25-44. |
| 34. | Siri, W. E. Body composition from fluid spaces and density: analysis of methods. In: Techniques for Measuring Body Composition, edited by J. Brozek, and A. Henschel. Washington, DC: National Academy of Sciences, 1961, p. 223-244. |
| 35. |
Slosman, D. O.,
J. P. Casez,
C. Pichard,
T. Rochat,
F. Fery,
R. Rizzoli,
J. P. Bonjour,
A. Morabia,
and
A. Donath.
Assessment of whole-body composition with dual-energy X-ray absorptiometry.
Radiology
185:
593-598,
1992 |
| 36. |
Snead, D. B.,
S. J. Birge,
and
W. M. Kohrt.
Age-related differences in body composition by hydrodensitometry and dual-energy X-ray absorptiometry.
J. Appl. Physiol.
74:
770-775,
1993 |
| 37. |
Svendsen, O. L.,
J. Haarbo,
C. Hassager,
and
C. Christiansen.
Accuracy of measurements of body composition by dual-energy X-ray absorptiometry in vivo.
Am. J. Clin. Nutr.
57:
605-608,
1993 |
| 38. | Tothill, P., A. Avenell, and D. M. Reid. Comparisons between Hologic, Lunar and Norland dual-energy X-ray absorptiometers and other techniques used for whole-body soft tissue. Eur. J. Clin. Nutr. 48: 781-794, 1994[Medline]. |
| 39. | Tothill, P., A. Avenell, and D. M. Reid. Precision and accuracy of measurements of whole-body bone mineral: comparisons between Hologic, Lunar and Norland dual-energy X-ray absorptiometers. Br. J. Radiol. 67: 1210-1217, 1994[Abstract]. |
| 40. | Wang, J., A. Dilmanian, J. Thornton, M. Russel, S. Burastero, M. Mazariegos, S. B. Heymsfield, and R. N. Pierson, Jr. In vivo neutron activation analysis for body fat: comparisons by seven methods. In: Human Body Compositon: In Vivo Methods, Models, and Assessment., edited by K. J. Ellis, and J. D. Eastman. New York: Plenum, 1993, p. 31-34. |
| 41. |
Wellens, R.,
W. C. Chumlea,
S. Guo,
A. F. Roche,
N. V. Reo,
and
R. M. Siervogel.
Body composition in white adults by dual-energy X-ray absorptiometry, densitometry, and total body water.
Am. J. Clin. Nutr.
59:
547-555,
1994 |
This article has been cited by other articles:
|
|
 |
|
  C. A Clay, S. Perera, J. M Wagner, M. E Miller, J. B Nelson, and S. L Greenspan Physical Function in Men With Prostate Cancer on Androgen Deprivation Therapy Physical Therapy, October 1, 2007; 87(10): 1325 - 1333. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. G. Esposito, S. G. Thomas, L. Kingdon, and S. Ezzat Comparison of Body Composition Assessment Methods in Patients with Human Immunodeficiency Virus-Associated Wasting Receiving Growth Hormone J. Clin. Endocrinol. Metab., August 1, 2006; 91(8): 2952 - 2959. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. F. Piepoli, A. Kaczmarek, D. P. Francis, L. C. Davies, M. Rauchhaus, E. A. Jankowska, S. D. Anker, A. Capucci, W. Banasiak, and P. Ponikowski Reduced Peripheral Skeletal Muscle Mass and Abnormal Reflex Physiology in Chronic Heart Failure Circulation, July 11, 2006; 114(2): 126 - 134. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. E Williams, J. C. Wells, C. M Wilson, D. Haroun, A. Lucas, and M. S Fewtrell Evaluation of Lunar Prodigy dual-energy X-ray absorptiometry for assessing body composition in healthy persons and patients by comparison with the criterion 4-component model Am. J. Clinical Nutrition, May 1, 2006; 83(5): 1047 - 1054. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. J.M. Serlie, J. H. de Haan, C. J. Tack, H. J. Verberne, M. T. Ackermans, A. Heerschap, and H. P. Sauerwein Glycogen Synthesis in Human Gastrocnemius Muscle Is Not Representative of Whole-Body Muscle Glycogen Synthesis Diabetes, May 1, 2005; 54(5): 1277 - 1282. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Sun, C. R French, G. R Martin, B. Younghusband, R. C Green, Y.-g. Xie, M. Mathews, J. R Barron, D. G Fitzpatrick, W. Gulliver, et al. Comparison of multifrequency bioelectrical impedance analysis with dual-energy X-ray absorptiometry for assessment of percentage body fat in a large, healthy population Am. J. Clinical Nutrition, January 1, 2005; 81(1): 74 - 78. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Casaburi, S. Bhasin, L. Cosentino, J. Porszasz, A. Somfay, M. I. Lewis, M. Fournier, and T. W. Storer Effects of Testosterone and Resistance Training in Men with Chronic Obstructive Pulmonary Disease Am. J. Respir. Crit. Care Med., October 15, 2004; 170(8): 870 - 878. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. H. H. Goh, C. F. Tain, T. Y. Y. Tong, H. P. P. Mok, and M. T. Wong Are BMI and other anthropometric measures appropriate as indices for obesity? A study in an Asian population J. Lipid Res., October 1, 2004; 45(10): 1892 - 1898. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. J. Gately, D. Radley, C. B. Cooke, S. Carroll, B. Oldroyd, J. G. Truscott, W. A. Coward, and A. Wright Comparison of body composition methods in overweight and obese children J Appl Physiol, November 1, 2003; 95(5): 2039 - 2046. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. E. van der Ploeg, R. T. Withers, and J. Laforgia Percent body fat via DEXA: comparison with a four-compartment model J Appl Physiol, February 1, 2003; 94(2): 499 - 506. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. Smith, B. Engel, A. M Diskin, P. Spanel, and S. J Davies Comparative measurements of total body water in healthy volunteers by online breath deuterium measurement and other near-subject methods Am. J. Clinical Nutrition, December 1, 2002; 76(6): 1295 - 1301. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Y. McDermott, A. Shevitz, T. Knox, R. Roubenoff, J. Kehayias, and S. Gorbach Effect of highly active antiretroviral therapy on fat, lean, and bone mass in HIV-seropositive men and women Am. J. Clinical Nutrition, November 1, 2001; 74(5): 679 - 686. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Cicoira, L. Zanolla, L. Franceschini, A. Rossi, G. Golia, M. Zamboni, P. Tosoni, and P. Zardini Skeletal muscle mass independently predicts peak oxygen consumption and ventilatory response during exercise in noncachectic patients with chronic heart failure J. Am. Coll. Cardiol., June 15, 2001; 37(8): 2080 - 2085. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. R. Wagner and V. H. Heyward Validity of two-component models for estimating body fat of black men J Appl Physiol, February 1, 2001; 90(2): 649 - 656. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P Tothill, W J Hannan, and S Wilkinson Comparisons between a pencil beam and two fan beam dual energy X-ray absorptiometers used for measuring total body bone and soft tissue Br. J. Radiol., February 1, 2001; 74(878): 166 - 176. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. B Houtkooper, S. B Going, J. Sproul, R. M Blew, and T. G Lohman Comparison of methods for assessing body-composition changes over 1 y in postmenopausal women Am. J. Clinical Nutrition, August 1, 2000; 72(2): 401 - 406. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. G. LOHMAN, M. HARRIS, P. J. TEIXEIRA, and L. WEISS Assessing Body Composition and Changes in Body Composition: Another Look at Dual-Energy X-ray Absorptiometry Ann. N.Y. Acad. Sci., May 1, 2000; 904(1): 45 - 54. [Abstract] [Full Text] [PDF]
|
|
|
|
