International Journal of Obesity (1997) 21, 1167±1175
ß 1997 Stockton Press All rights reserved 0307±0565/97 $12.00
Age- and maturity-related changes in body
composition during adolescence into
adulthood: The Fels Longitudinal Study
SS Guo, WC Chumlea, AF Roche and RM Siervogel
Division of Human Biology, Departments of Community Health and of Pediatrics, Wright State University School of Medicine,
Yellow Springs, OH, USA
OBJECTIVES: To examine patterns of change in total body fat (TBF), percent body fat (%BF), and fat-free mass (FFM),
from 8±20 y of age and the effect of rate of skeletal maturation. To determine the degree of tracking of body
composition for individuals from childhood into adulthood.
RESEARCH DESIGN: Annual serial data for TBF, %BF and FFM from underwater weighing using a multicomponent
body composition model were collected from 130 Caucasian males and 114 Caucasian females between 1976 and
1996. Rate of maturation was de®ned as FELS skeletal age (SA) less chronological age (CA). Random effects models
were used to evaluate general patterns of change and tracking of individual serial data over the 12 y age range.
RESULTS: Changes in TBF followed a quadratic model for males and for females with declining rates of change.
Changes for %BF followed a cubic model for males and females. General patterns of change for FFM followed a cubic
model for males and a quadratic model for females. TBF for males and females increased with age, but the rates of
change declined with age. %BF for females increased from age 8±20 y. For males, %BF increased with age, but the
positive rates of change declined and became a negative when aged about 13 y and reached a minimum at about
the age of 15 y. The rate of change for %BF increased thereafter. FFM for males and females increased with age,
but the rates of change decreased with age. The extent of tracking is inversely related to the length of the time
interval. At the same age, rapidly-maturing children have signi®cantly larger amounts of TBF, %BF and FFM than slowmaturing children. Tracking in body composition for individuals persisted from childhood to adulthood.
CONCLUSIONS: (1) There are gender-associated differences in these patterns of change for %BF and FFM but not for
TBF; (2) TBF, %BF and FFM increased with increased rates of maturation; (3) signi®cant tracking in body composition
for individuals persists from childhood to adulthood.
Keywords: body composition; maturation; patterns of change
Introduction
Changes in body composition occur in conjunction
with growth in body size and shape during adolescence.1 Growth produced changes in total body fat
(TBF), percent body fat (%BF) and fat-free mass
(FFM) during childhood affect adult body composition and fat distribution, all of which, in turn, affect
risk factors for cardiovascular and related diseases.1±4
However, there is a paucity of knowledge regarding
the long-term patterns of change in body composition
from childhood into adulthood. The tracking of
indices of body composition such as the body mass
index (BMI) from childhood into adulthood has been
reported.5,6 BMI values at an age are highly correlated
with concurrent measures of TBF and %BF7 and
indicate potential risk for obesity later in adulthood.6
However, the tracking of measured values for TBF,
Correspondence: Shumei S Guo, PhD, Department of
Community Health, Wright State University School of Medicine,
1005 Xenia Avenue, Yellow Springs, OH 45387-1695, USA.
E-mail: [email protected]
Received 19 February 1997; revised 14 July 1997; accepted 6
August 1997
%BF and FFM for individuals from childhood into
adulthood has not been reported. Understanding and
quantifying the patterns of change in body composition and the tracking of levels of body composition
from childhood into adulthood, would allow the early
recognition of children and adolescents with aberrant
changes and=or unusual levels of body composition.
This knowledge would facilitate their subsequent
management.
Our knowledge of the changes in body composition, from childhood into adulthood, depends on the
availability of long-term serial studies of children, but
available serial data are scant. Most reports of changes
in body composition during childhood and adolescence have used cross-sectional data,1,8,9 or were
based upon changes in estimates of body composition
from indirect methods.10,11 Analyses of limited serial
data, have been reported by Chumlea and colleagues.12,13 These investigators reported annual increments in body composition variables for 69 children
aged 10±18 y, using data from the Fels Longitudinal
Study. These short term serial data, permitted simple
descriptions of changes, based upon two or three
consecutive annual examinations only. These data
were insuf®cient for modelling individual patterns of
Age- and maturity-related changes in body composition
SS Guo et al
1168
change over time or for relating changes to subsequent
adult outcomes.
It is well recognized that there is a relationship
between the maturation of a child and his or her
growth and development, especially during adolescence. Using data from the Fels Longitudinal Study,
Reynolds14 demonstrated that early sexual maturity in
girls and boys, aged 6±16 y, based upon secondary
gender characteristics, was related to increased rates
of growth in radiographic measures of muscle and
subcutaneous adipose tissue. Similar ®ndings have
been reported by others.15,16 However, these earlier
studies were not able to incorporate the rate or level of
maturation, with concurrent changes in body composition over long time periods, into multivariate serial
analytical models.
The present study examined the pattern of agerelated changes in TBF, %BF and FFM during a 12 y
age range, from childhood and adolescence to young
adulthood. We addressed three questions. First, what
were the age- and gender-speci®c patterns of change
in TBF, %BF and FFM from 8±20 y of age? The
patterns of change in TBF, %BF and FFM were
modelled, and the annual rates of change for body
composition were derived. Secondly, what was the
effect of the rate of biological maturation on the
patterns of change in body composition values?
Indices of rates of maturation were incorporated into
the analysis to evaluate their effects on changes in
body composition during adolescence. Thirdly, what
was the extent of the tracking of body composition
from childhood into young adulthood? The likelihood
of individuals remaining in the same percentile channel in body composition over time was evaluated. The
sensitivity and speci®city of childhood and adolescent
body composition measures, as predictors of adult
values, were computed.
Subjects and Methods
The data for the present study were from 130 Caucasian males and 114 Caucasian females, all of whom
are participants in the Fels Longitudinal Study in
southwestern Ohio. The Fels Longitudinal Study
started in 1929 and has been described in detail.17
Participants enter the Fels Longitudinal Study at or
before birth and are followed at regularly scheduled
visits. Body composition measures start at the age of
8 y. For the present analysis, we included participants
who had at least 6 serial body composition measures
taken between the ages of 8 and 23 y. All procedures
were approved by the Institutional Review Board of
Wright State University.
To study changes in body composition from 8±20 y
of age, individual, annual serial measures of TBF,
%BF and FFM from 8±23 y of age were analyzed.
These measurements were taken near each partici-
pant's birthdays from 1976 to 1988. Because of the
timing of scheduled visits and the ages of the participants at the beginning and the end of this period of data
collection, some participants had incomplete data sets
within the age range of the analysis. Some participants
were older than 8 y when the measurements began in
1976, and some were younger than 23 y when the most
recent data for the present analysis were recorded. The
absence of these values did not alter the descriptions of
the changes that occurred in body composition. The
inclusion of serial data up to the age of 23 y in the
analysis improved the accurate description of the patterns of change near the age of 20 y.
Body composition methods
Measures of TBF, %BF and FFM were obtained from
a multicomponent body composition model. The ageand gender-speci®c values for the concentrations of
the major constituents of FFM were used to calculate
values for its density in different age groups for each
gender.18 We smoothed the Lohman18 values for the
density of FFM by using a ®xed-knot cubic spline
technique and calculated FFM from body density
using these smoothed values.19 Let d1 be the density
and d2 be the density of fat (0.9 g=cc). The value for
%BF is:
1
d1 d2
d2
ÿ
100;
%BF ˆ
BD d1 ÿ d2 d1 ÿ d2
and TBF and FFM can be obtained from %BF as:
TBF ˆ W %BF and FFM ˆ W…1 ÿ %BF†:
Body density was determined from underwater weighing and residual volume, which was measured on a
Gould 2100 computerized spirometer.19 In the multicomponent model, the density of the fat-free component varied due to changes in its water and bone
mineral content.18,19 Stature was measured to 0.1 cm
on a Holtain stadiometer, (Seritex, Carlstadt, NJ,
USA) and weight was measured to 0.1 kg on a beam
balance scale.
Maturational data
Skeletal ages were assessed using the FELS method20
from left hand-wrist radiographs obtained on the same
day as the body composition measures. Rate of
maturation was de®ned as skeletal age (SA) of the
left hand-wrist less chronological age (CA), that is,
SA 7 CA. For each participant, the average of
SA 7 CA over the period from 8 y to maturity was
obtained. A participant with an average value > 1
indicated a tendency for rapid maturation; a participant with an average value < 7 1 had a tendency for
slow maturation and `intermediate' otherwise.
Preliminary analysis
Each individual's serial data were divided into seven,
2 y segments: 8±10, 10±12, 12±14, 14±16, 16±18, 18±
Age- and maturity-related changes in body composition
SS Guo et al
20 and 21±23 y. If a participant had more than one
measurement within a two year age range, the average
was computed and used in the analysis. Means and
s.d. were calculated for weight, stature, TBF, %BF
and FFM for males and females separately within
each two year age segment.
General patterns of change
This analysis focused on the overall behavior of body
composition values for individuals across the age
range. If body composition measurements at different
ages for individuals follow a multivariate normal
distribution, then the mean and covariance structure
of these data are suf®cient to describe the process of
changes in body composition over time.
To describe the patterns of change for TBF, %BF
and FFM from childhood into adulthood, the corresponding age-speci®c values for each individual were
plotted by age. Each individual's serial data, were
®tted by a family of low-degree polynomials (including linear quadratic and cubic) with four possible
covariance models, to describe the variance-covariance structure between pairs of serial data within
each body composition variable. The likelihood ratio
tests and the Akaike Information Criterion21 were
used to select the appropriate model. The likelihood
ratio method, tests various nested models using
asymptotic chi-square statistics. The Akaike Information Criterion (AIC) was used as a overall measure of
adequacy of ®t of the speci®c models. The computation was performed using `SAS PROC MIXED'
program.22
Tracking
Tracking for body composition was determined from
the predicted values of the individual ®tted model as
the extent to which individuals remained in the same
percentile channel over time. The percentile channel
was de®ned as the upper tertile of the study sample for
each age. Sensitivity (Se) refers to the percentage of
participants who remained in the upper tertile group
from one age to another. Speci®city (Sp) refers to the
percentage of participants who remained in the combined middle and lower tertile group from one age to
another. Tracking of TBF, %BF and FFM was evaluated between 5 y age intervals, 8±13 and 13±18 y,
and a 10 y age interval, 8±18 y. These selected age
intervals approximately represent a prepubertal to
pubertal period, a pubertal to postpubertal period
and a prepubertal to postpubertal period.
Results
General patterns of change
Gender-speci®c means and standard deviations for
TBF, %BF and FFM, are presented in Table 1 for
each 2 y age segment. The means for TBF and %BF
were signi®cantly larger (P < 0.05) in females than in
males from 8±20 y of age and the gender difference in
these mean values increased with age. The means for
FFM, stature and weight were similar for males and
females from the age of 8±14 y, but after 14 y of age,
the means for FFM for males were signi®cantly larger
than those for females. Similarly, the males were
signi®cantly taller and heavier than females > 14 y.
The selection of the appropriate model for each
body composition variable, was based on the greatest
AIC value and the likelihood ratio tests (Table 2).
Two of the covariance models did not converge and
were not considered further. A quadratic model with
random intercepts and slopes was chosen for TBF for
both males and females. The best ®t for %BF for both
males and females was a cubic model with random
intercepts and slopes. For FFM in males, a cubic
model with random intercepts and slopes was
selected, but in the females, a quadratic model with
random intercepts and slopes was selected. The
regression parameter estimates with their standard
Table 1 Means s.d. for body composition from underwater weighing and for anthropometry by age segments and gender*
8^10 y
10^12 y
12^14 y
14^16 y
16^18 y
18^20 y
Males
TBF (kg)
BF(%)
FFM (kg)
Stature (cm)
Weight (kg)
BMI (kg=m2)
4.73 3.97*
14.89 10.01*
24.47 3.44
133.10 5.57
29.19 5.90
16.35 2.26
6.35 5.06*
16.45 8.87*
28.91 3.87
143.81 5.92
35.26 7.85
16.92 2.83*
8.18 5.50*
17.62 7.93*
35.19 6.00
154.50 7.71
43.36 10.10*
17.99 3.06*
9.08 7.23*
14.95 8.01*
48.03 8.59*
169.45 7.96*
57.10 12.78
19.76 3.63
8.00 7.14*
11.40 6.98*
57.43 8.06*
176.72 7.04*
65.42 12.2*
20.91 3.53
9.72 7.27*
12.99 7.00*
60.20 7.52*
179.17 7.00*
69.90 12.5*
21.74 3.49
Females
TBF (kg)
BF(%)
FFM (kg)
Stature (cm)
Weight (kg)
BMI (kg=m2)
6.39 3.80
20.12 9.42
24.06 3.97
133.75 6.10
30.45 5.82
16.91 2.24
8.87 4.36
22.54 7.53
29.90 5.11
144.53 7.61
37.86 8.09
17.95 2.64
11.21 4.88
23.15 6.99
35.96 6.24
156.17 7.73
47.16 9.28
19.20 2.75
13.56 6.10
23.65 6.83
42.20 5.90
163.63 7.07
55.75 9.71
20.77 3.10
15.16 6.46
25.10 6.75
43.67 5.39
164.91 7.01
58.82 9.66
21.61 3.12
16.25 6.41
26.33 7.01
43.96 5.51
165.42 6.82
60.19 9.71
21.96 2.95
*
Sign®cant sex-associated differences at P < 0.05.
TBF ˆ total body fat; %BF ˆ percent body fat; FFM ˆ fat-free mass; BMI ˆ body mass index.
1169
Age- and maturity-related changes in body composition
SS Guo et al
1170
Table 2 Goodness of ®t of models and estimated regression coef®cients for the selected models for total body fat (TBF), percent body
fat (%BF) and fat-free mass (FFM)
Models
AIC
TBF (kg) for males
Random intercept
Random intercept, slope*
TBF (kg) for females
Random intercept
Random intercept, slope*
72 log likelihoods
Chi-square statistics
P value
2
TBF* ˆ 2.4 ‡ 0.65 age 70.003 age
2985.45
±
2851.76
133.69
TBF* ˆ 76.28 ‡ 1.84 age 70.03 age2
2562.06
±
2448.30
113.76
71494.72
71429.88
71281.03
71228.15
±
< 0.001
±
< 0.001
BF (%) for males
Random intercept
Random intercept, slope*
71631.41
71612.96
BF* ˆ 766.75 ‡ 18.63 age 71.28 age2 ‡ 0.028 age3
3258.82
±
3217.92
40.90
±
< 0.001
BF (%) for females
Random intercept
Random intercept, slope*
71419.54
71401.17
BF* ˆ 712.6 ‡ 7.01 age 70.44 age2 ‡ 0.01 age3
2835.07
±
2794.35
40.72
±
< 0.001
FFM (kg) for males
Random intercept
Random intercept, slope*
71584.84
71540.34
FFM* ˆ 131.93 727.67 age ‡ 2.28 age2 70.05 age3
3165.67
±
3072.67
93.00
±
< 0.001
FFM (kg) for females
Random intercept
Random intercept, slope*
7120.95
71193.03
FFM* ˆ 728.26 ‡ 7.54 age 70.19 age2
2422.90
±
2378.05
33.85
±
< 0.001
*Selected models based upon chi-square tests for P < 0.05. For all the selected models, while there are general patterns of change over
time for each variable, there are also differences among individuals. These differences are re¯ected in the baselines (intercept) and the
rates of change (slope).
errors for each selected model are summarized in
Table 3.
The regression parameter estimates of these models
were used to predict corresponding values for TBF,
%BF and FFM, for each individual. Gender-speci®c
plots of the means of these predicted values for TBF,
%BF and FFM are shown in Figures 1±3, respectively.
The predicted values were close to the corresponding
observed values. This concordance indicated that the
chosen models adequately described the general patterns of change in the body composition of these
children into young adulthood.
From these tabular and graphic results, it appears
that on average from childhood into adulthood, TBF
and %BF increased in both the females and males, but
more in the females than the males. The average
increase in TBF for females was from 6.4 kg at age
8 y to 16.3 kg at age 20 y; the corresponding increase
in males was from 4.7 kg at age 8 y to 9.7 kg at age
20 y. Similarly, the average %BF for the females
increased from 20±26% BF. In the males, the means
for %BF increased from age 8±14 y, then decreased
from age 14±18 y, but increased again from age 18±
20 y. The overall increase in FFM from age 8±14 y
was similar for both males and females, but afterwards, the males demonstrated a much greater
increase in FFM than the females. Average amounts
of FFM increased about 10 kg from age 8±14 y in
both males and females. After the age of 14 y, the
increase in mean FFM was 9.0 kg for females, but
for males, the average amount of FFM increased
about 25.0 kg. These changes in FFM are re¯ected
by corresponding changes in stature and weight
(Table 1).
Table 3 The regression parameter estimates and effects of maturity on total body fat (TBF), percent body fat (%BF) and fat-free mass
(FFM) for males and females
Intercept
Males
TBF(kg)
BF(%)
FFM(kg)
Females
TBF(kg)
BF(%)
FFM(kg)
Age
2.40
(2.72)
766.74
(12.46)
131.92
(10.89)
0.65
(0.32)
18.63
(2.51)
727.67
(2.23)
76.28
(2.82)
712.60
(14.20)
728.26
(2.55)
1.84
(0.37)
7.01
(2.95)
7.54
(0.32)
Age2
Age3
70.003
(0.01)
71.28
(0.17)
70.05
(0.003)
±
±
0.027
(0.004)
±
±
75.24(1.75)*
±
76.71(2.60)*
±
77.05(1.7)*
±
72.76(1.36)*
±
73.16(1.98)
±
74.95(1.4)
±
70.03
(0.01)
70.44
(0.20)
70.19
(0.01)
±
±
0.01
(0.004)
±
±
73.40(1.77)
±
72.98(2.55)
±
76.09(2.12)*
±
72.73(1.01)*
±
73.38(1.01)*
±
71.81(1.20)
±
Slow-maturing
Intermediate
*The values are expressed relative to rapid-maturity ˆ zero, signi®cant at P < 0.05. For example, on the average, a rapid-maturer has
larger TBF than the intermediate-maturer by 2.76 kg.
age2 ˆ age6age.
age3 ˆ age6age6age.
Age- and maturity-related changes in body composition
SS Guo et al
General patterns of rates of change
Figure 1 Average status values in total body fat (TBF) from age
8±20 y for males and females.
By taking the ®rst derivative of the selected models
for TBF, %BF and FFM with respect to time, we
generated gender-speci®c velocity curves for each
body composition variable (Figures 4±6). Males and
females had positive annual velocities for TBF, indicating that TBF increased continuously from the age
of 8±20 y. The velocity in TBF for females was larger
than that for males, particularly at young ages, and
declined with age. The maximum velocity for TBF
was 0.6 kg=y for males and 1.4 kg=y for females at
about age 8 y for both.
Percent body fat increased continuously from the
age of 8±20 y for females, but the velocity for %BF
decreased for much of this age range. The maximum
velocity for girls was 1.9%=y at about the age of 8 y,
and the minimum velocity was 0.6%=y at about the
age of 15 y, then the velocity started to increase again.
For males, %BF increased from age 8 y to about 12 y,
but %BF velocity during this age range decreased
from 3.5%=y to zero. After the age of 12, %BF in
males decreased because of the negative velocity in
%BF which reached a minimum of 7 0.9%=y at
about age 15 y. After the age of 15 y, the velocity in
%BF for males remained negative but increased
toward zero at age 18 y. By age 20 y, the velocity in
%BF in males was 1%=y.
The velocity for FFM, for males increased from the
age of 8 y to reach a maximum velocity of 7.0 kg=y at
about age 15 y. The velocity for FFM in males then
decreased to 3.5 kg=y at age 20 y. The velocity for
FFM in females decreased continuously over the same
period. The maximum velocity for FFM for females
was about 4.5 kg=y at around the age of 8 y. Females
only had larger annual velocities in FFM than the
males from the age of 8±11 y and the velocity for FFM
for females decreased to zero by the age of 20 y.
Figure 2 Average status values in percent body fat (%BF) from
age 8±20 y for males and females.
Rates of Maturation
The associations of patterns of change in TBF, %BF
and FFM with rates of maturation are presented in
Table 3. Independent of age, rapidly maturing males
had larger values for TBF, %BF and FFM than slowly
maturing males and females. At the same age, rapidlymaturing females had signi®cantly larger mean FFM
values than slowly-maturing females. Rapidly-maturing males had signi®cantly larger mean TBF values
than the intermediate maturing males. Rapidly-maturing females had signi®cantly larger mean TBF BF
values than intermediate-maturing females.
Tracking
Figure 3 Average status values in fat-free mass (FFM) from age
8±20 y for males and females.
The sensitivity for children to remain in the upper
tertile from one age to another are presented in Table
4. The tracking analyses for TBF showed that, of the
males whose TBF values were in upper tertile at age
8 y, about 74% remained in the upper tertile at age
13 y (sensitivity). Of the females with TBF values in
the upper tertile at age 8 y, 68% remained in the upper
1171
Age- and maturity-related changes in body composition
SS Guo et al
1172
Figure 4 Velocity in total body fat (TBF) from age 8±20 y for
males and females.
Figure 5 Velocity in percent body fat (%BF) from age 8±20 y for
males and females.
tertile group at 13 y. This proportion decreased as the
length of the time interval increased. For example,
only 60% of males with values in the upper tertile at
age 8 y were also in the upper tertile at age 18 y, and
the proportion remaining in the same percentile zone
decreased to 46% for females. However, of the males
in the upper tertile for TBF at age 13 y, 86% remained
in the upper tertile at age 18 y as did 77% of the
females. This pattern of a stronger degree of tracking
from age 13±18 y than from 8±13 y or 8±18 y was
fairly consistent for both males and females for %BF
and FFM also. Tracking for 5 y intervals was slightly
greater for those age 13 y than for those aged 8 y,
indicating better tracking from the pubertal to the
postpubertal period (13±18 y) than from the prepubertal to the pubertal period (8±13 y). These same patterns persisted for the individuals in the middle and
lower tertiles (speci®city) across the same age ranges
for TBF, %BF and FFM.
The 5 y tracking for %BF was 91% for males aged
8 y and 81% at age 13 y. Tracking in %BF in males
decreased from the prepubertal to the pubertal period
and from the pubertal to postpubertal period. The
extent of tracking continued to decrease as the timeintervals increased. Tracking in %BF (both sensitivity
and speci®city) was less marked in females than males
for all age intervals. Furthermore, for females, 5 y
tracking from the prepubertal to the pubertal period
was similar to that from the pubertal to postpubertal
period.
Of the males whose FFM values were in upper
tertile at age 8 y, about 79% remained in the upper
tertile at age 13 y (sensitivity). Of the female participants with FFM values in the upper tertile at age 8 y,
89% remained in the upper tertile group at age 13 y.
For the 13±18 y intervals, sensitivity was greater for
males and only slightly greater for females aged 8 y
than for males of the same age, indicating better
tracking from the pubertal to the postpubertal period
than from the prepubertal to the pubertal period. The
degree of tracking decreased slightly as the length of
the time intervals increased, but more so in males than
females. For example, of those males with FFM
values in the upper tertile at age 8 y, only 70%
remained in the upper tertile at age 18 y, while 83%
of females remained in the upper tertile for FFM at
age 18 y. This greater degree of tracking in FFM in the
females is probably because most girls have completed their growth in FFM during this period.
Discussion and Conclusions
Figure 6 Velocity in fat-free mass (FFM) from age 8±20 y for
males and females.
The present study con®rms previous reports of
changes in body composition during adolescence. In
this study, long-term serial data, over a 12-year
period, were analyzed to examine age- and maturity-
Age- and maturity-related changes in body composition
SS Guo et al
1173
Table 4 Tracking in body composition from childhood to adulthood at 5 y and 10 y intervals
TBF
Age Intervals (y)
Males
8±13
13±18
8±18
Females
8±13
13±18
8±18
%BF
FFM
Sensitivity
Specificity
Sensitivity
Specificity
Sensitivity
Specificity
0.74
0.86
0.60
0.87
0.93
0.80
0.91
0.81
0.72
0.79
0.90
0.86
0.79
0.91
0.70
0.89
0.95
0.84
0.68
0.77
0.46
0.84
0.88
0.72
0.71
0.71
0.43
0.86
0.86
0.71
0.89
0.91
0.83
0.94
0.96
0.91
TBF ˆ total body fat: %BF ˆ percent body fat: FFM ˆ fat-free mass:
related changes in measured body composition for
groups and for individuals from childhood into young
adulthood. The statures and weights of the present
sample were within the corresponding 5th and 95th
percentiles for stature and weight, for persons at the
same ages from the National Center for Health Statistics.23 Body composition measures were obtained
from underwater weighing using a multi-component
body composition model and the effects of rate of
maturation on long-term patterns of change in body
composition were examined.
Numerous studies have reported changes in body
composition during growth from the prepubertal,
pubertal, to postpubertal periods,1,8,24 but these studies
were based predominantly upon cross-sectional data.
Only a few studies have used serial data to analyze
changes in body composition, but these previous serial
studies had small sample sizes.1,13 Furthermore, previous studies calculated changes in body composition
as differences between two successive measurements
or applied a simple linear regression model. Differences between pairs of successive measurements provide inadequate descriptions of the long-term patterns
of change and individual data can be excluded when
there are missing values. A linear regression assumes
changes over time are constant, which may not
accurately re¯ect the actual patterns of change. The
present study took advantage of the availability of
long-term serial data and improved statistical methods
to investigate long-term patterns of change in body
composition.
A random effects model was used to determine the
patterns of change over time in body composition and
the parameters were used to characterize individual
differences. This model analyzed the complete set of
serial data. The random effects model handled the
occurrence of missing values and included measurements made at various time intervals. Missing values
were estimated, using maximum likelihood procedures, with the assumption that the patterns of
change for individuals followed paths similar to
those of their peers. The random effects model also
allowed for the inclusion of covariates such as gender
and rate of maturation.
The general pattern of change in TBF values was
one of increase throughout the study period (Figure 1)
but the velocity in TBF declined (Figure 3). The
females consistently had larger values for TBF than
males, and the extent of these gender differences
increased after 13 y. For females, %BF also increased
from age 8±20 y (Figure 2), but the velocity in %BF
declined and reached a minimum at about age 15 y
(Figure 5). The signi®cant increase in body fat in
females during pubescence and their greater degree of
fatness, compared with males, is largely due to their
greater production of estrogen.1 The %BF values for
males increased from age 8±12 y, then decreased until
about age 18 y, but started to increase thereafter.
Haschke24 reported an increase in %BF from age 7±
12 y followed by a decrease in values from age 12±
17 y in males, whereas Chumlea and co-workers13
reported an annual decrease of 1.15% in %BF from
age 10±18 y.13
The results from the application of the random
effects model, indicated that the general patterns of
change for FFM followed a cubic model in males and
a quadratic model in females (Figure 3). FFM values
were similar between males and females before age
12±14 y, but males had substantially larger FFM
values than females thereafter. Values for FFM
increased with age in each gender, but the increase
tended to decline continuously for females as indicated by the decreasing velocity in FFM for females
(Figure 6). For males, the increase in FFM was just
beginning to decline in the postpubertal period. The
velocity in FFM which reached a peak of 7 kg=y was
still 4 kg=y at age 20 in males. The gender differences
in the present ®ndings were slightly smaller than those
reported by Forbes1 who considered the differences
to be largely due to the greater production of testosterone by the males. The increase in FFM of 36 kg for
males from age 10±20 y in the present study, was
larger than the increase of 33 kg reported by Forbes1.
The increase in FFM for females from age 10±20 y
was 19 kg which was also larger than that of 16 kg
reported by Forbes.1 These study differences, in the
amounts of change in FFM may be partly due to
whether serial data for individuals or cross-sectional
data were analyzed or a multicomponent model was
used.
The major changes in the components of FFM from
the prepubertal to young adulthood are due to altera-
Age- and maturity-related changes in body composition
SS Guo et al
1174
tions in the water and bone mineral content of the
FFM8,24 leading to increases in the density of FFM
with growth and maturation.18 There are also gender,
race and maturational differences.25,26 It is necessary
to use a multicomponent model to estimate body
composition in children and adolescents, from underwater weighing, because the multicomponent model
assumes that the total body consists of fat, water, bone
and protein, and that each of these components has a
®xed density. Several studies have documented the
advantages of multi-compartment model over the
traditional 2-compartment model.18,27,28
Within the study age period, the velocity curves for
TBF and FFM have their maximum values at age 8 y
for females (Figure 4 and Figure 6). The study period
began at age 8 y, so that there are no body composition measures for younger participants. If data for
younger participants were available, it is possible that
the maximum velocity in TBF and FFM could have
been earlier. We are uncertain as to the age of
maximum velocity for TBF and FFM because of the
age constraint of performing underwater weighing for
young children. With the availability of pediatric
software for dual energy X-ray absorptiometry
machines, future investigators should be able to
remove this uncertainty for the age at maximum
velocity for TBF and FFM in females.
The selected random effects models ®t the individual serial data well. As expected, the variations
between the predicted and the observed values in
the individual curves were greater than those in the
mean curves. There were differences among individuals in the intercepts (baseline values at 8 y) and the
slopes (rates of change). The individual parameter
estimates from the present results allow future studies
of covariate in¯uences on patterns of change in body
composition. These in¯uences include genetic factors,
physical activity, diet, hormone and maturational
levels.
In the present study, relative skeletal age expressed
relative to chronological age (CA), that is SA±CA,
was used as an index of maturity. This measure of
maturity de®ned whether an individual was rapidlymaturing, slowly-maturing or near-average. A mean
relative skeletal age for each child was obtained over
the study period and incorporated into the random
effects model, to study the effects of maturity levels
on body composition. At the same chronological age,
rapidly-maturing children have signi®cantly larger
values for TBF, %BF and FFM than slowly-maturing
children. At the same chronological age, rapidlymaturing males and females have signi®cantly larger
values for TBF and %BF, than those maturing at near
average rates and signi®cantly larger values for FFM
than those who are maturing slowly. These results
using relative skeletal age, support the similar ®ndings
by other investigators1,14,16,31 who have used this and
other measures of maturity (age at menarche, peak
height velocity and secondary gender characteristics);
earlier maturing children have greater rates of growth
in muscle and fat than later maturing children at the
same chronological age.
The participants in the present study were born
between 1958 and 1988. Potentially, there could be
secular trends in the body composition measures,
affecting these results for the older children. To
examine this possibility, we separated the participants
into four groups by their birth years: 1953±1963,
1963±1973, 1973±1983 and 1983 ‡ . Values for
TBF, %BF and FFM were compared among the four
groups at age 8, 13, and 18 y. There were consistent
increasing trends from older to younger groups by
birth years, but these trends were not signi®cant.
However, these trends in increased fatness among
the younger birth groups corresponds with the
reported increased prevalence of obesity in US children.29
We have demonstrated, for the ®rst time, that the
tracking of measured values of TBF, %BF and FFM
persists over 5 and 10 y time periods from childhood
to adulthood. The occurrence of tracking at these
levels has signi®cant implications for the clinical
management, public health programs and epidemiological studies of US children. Clinical and public
health programs are aimed at altering the percentile
levels of children with high values for BMI that are
associated with increased health risk. Values for BMI
are highly correlated with TBF and %BF.7 Clearly,
when a high value is found for TBF or %BF (or a low
value for FFM) it should be regarded as a signi®cant
guide to future body composition values and potentially indicative of increased health risk in adulthood,30 because childhood body composition values
are closely related to corresponding values in young
adulthood. The present ®ndings indicate that high
values of TBF and %BF for individuals in childhood
strongly persist into adulthood and this persistence is
even stronger for adolescence into adulthood. Therefore, the presence of obesity in a child early in
adolescence, has a high probability of continuing, as
that child develops into an adult. Body composition
should be measured or accurately estimated in clinical
and epidemiological studies of children, even perhaps
at younger ages than those covered in the present
investigation.
Conclusions
From the present results, we are able to describe the
patterns of change in body composition from childhood into adulthood, the effect of the rate of maturation on body composition and the degree of tracking
in body composition for an individual. These ®ndings
con®rm reports from earlier cross-sectional and shortterm serial studies. For TBF, there is a continual
increase, but a declining rate of change. For %BF
the same pattern exists as for TBF for females, but for
Age- and maturity-related changes in body composition
SS Guo et al
males the pattern of change in %BF re¯ects the
concurrent changes in FFM. For FFM, there is an
increase in girls that slows down in adolescence, but
in males, there is a continual rapid increase into young
adulthood. For the same age, rapidly-maturing children have signi®cantly larger TBF, %BF and FFM
than the slowly-maturing children. For the same age,
rapidly-maturing females have signi®cantly larger
TBF and %BF than the intermediate-maturing females
and signi®cantly larger FFM than the slow-maturing
females. Signi®cant tracking in body composition
exists. The extent of tracking is inversely related to
the length of time intervals between two measurements.
Many aspects of these ®ndings have been reported
previously but for shorter time periods and smaller
(and sometimes cross-sectional) samples.7,12±14,31
With the use of a large serial data set, we have
con®rmed the pattern of change in body composition
among children and that these patterns are affected by
a child's level of maturity. More importantly, we have
demonstrated that the pattern of these changes in body
fatness persist into young adulthood, demonstrating
the signi®cant tracking of body composition and
obesity.
Acknowledgement
This work was supported by the grants HD 27063 and
HD 12252 from the National Institutes of Health,
Bethesda, MD, USA.
11
12
13
14
15
16
17
18
19
20
21
References
1 Forbes GB. Human Body Composition. Growth, Aging, Nutrition, and Activity. Springer-Verlag: New York, 1987. pp 350.
2 Burke GL, Savage PJ, Manolio TA, Sprafka JM, Wagenknecht
LE, Sidney S, Perkins LI, Liu K, Jacobs DR. Correlates of
obesity in young black and white women ± The Cardia Study.
Am J Pub Health 1992; 82: 1621±1625.
3 Wilson PW, McGee DL, Kannel WB. Obesity, very low
density lipoproteins, and glucose intolerance over fourteen
years. Am J Epid 1981; 114: 697±704.
4 Freedman DS, Burke GL, Harsha DW, Srinivasan SR, Cresanta JL, Webber LS, Berenson GS. Relationship of changes in
obesity to serum lipid and lipoprotein changes in childhood
and adolescence. JAMA 1985; 254: 515±520.
5 Cronk CE, Roche AF, Chumlea WC, Kent R, Berkey C.
Longitudinal trends in weight=stature2 in relation to adulthood
body fat measures. Hum Biol 1982; 54: 751±764.
6 Guo S, Chumlea WC, Roche AF, Gardner JD, Siervogel RM.
The predictive value of childhood body mass index values for
overweight at age 35 years. Am J Clin Nutr 1994; 59: 810±819.
7 Roche AF, Siervogel RM, Chumlea WC, Webb P. Grading of
body fatness from limited anthropometric data. Am J Clin Nutr
1981; 34: 2831±2838.
8 Boileau RA, Lohman TG, Slaughter MHJ, Ball TE, Going SB,
Hendrix MK. Hydration of the fat-free body in children during
maturation. Hum Biol 1984; 56: 651±666.
9 Van Loan M. Total body composition: Birth to old age. In:
Ellis KJ, Eastman JD (eds). Human Body Composition.
Plenum Press: New York, 1996, pp 205±215.
10 Deurenberg P, van der Kooy K, Hautvast JGAJ. The assessment of the body composition in the elderly by densitometry,
22
23
24
25
26
27
28
29
anthropometry and bioelectrical impedance. In: Yasumura S,
Harrison JE, McNeill KG, Woodhead AD, Dilmanian FA
(eds). In Vivo Body Composition Studies. Recent Advances.
Basic Life Sciences, Volume 55. Plenum Press: New York,
1990. pp 391±393.
Slaughter MH, Lohman TG, Boileau RA, Horswill CA, Stillman RJ, Van Loan MD, Bemben DA. Skinfold equations for
estimation of body fatness in children and youth. Hum Biol
1988; 60: 709±723.
Chumlea WC, Knittle JL, Roche AF, Siervogel RM, Webb P.
Size and number of adipocytes and measures of body fat in
boys and girls 10 to 18 years of age. Am J Clin Nutr 1981; 34:
1791±1797.
Chumlea WC, Siervogel RM, Roche AF, Webb P, Rogers E.
Increments across age in body composition for children 10 to
18 years of age. Hum Biol 1983; 55: 845±852.
Reynolds EL. Sexual maturation and the growth of fat,
muscle and bone in girls. Child Development 1946; 17: 121±
144.
Roche AF, Wainer H, Thissen D. The RWT method for the
prediction of adult stature. Pediatrics 1975; 56: 1026±1033.
Tanner JM. Growth at Adolescence. Blackwell Scienti®c
Publications: Oxford, 1962, pp 121±130.
Roche AF. Growth, Maturation and Body Composition: The
Fels Longitudinal Study 1929±1991. Cambridge University
Press: Cambridge, UK, 1992. pp 282.
Lohman T. Applicability of body composition techniques and
constants for children and youths. Exerc Sports Sci Rev 1986;
14: 325±357.
Guo S, Roche AF, Houtkooper L. Fat-free mass in children
and young adults predicted from bioelectric impedance and
anthropometric variables. Am J Clin Nutr 1989; 50: 435±443.
Roche AF, Chumlea WC, Thissen D. Assessing the Skeletal
Maturity of the Hand-Wrist: FELS Method. Charles C
Thomas: Spring®eld, IL, 1988. pp 339.
Akaike H. Information theory and an extension of the maximum likelihood principle. In: Petrov BN, Caski F (eds).
Second International Symposium on Information Theory.
Akademai Kiado: Budapest, 1973, pp 267±281.
SAS. SAS Technical Report P-229. SAS/STAT Software:
Changes and Enhancements. Release 6.07. SAS Institute,
Inc., North Carolina, 1992. pp 620.
Najjar MF, Rowland M. Anthropometric Reference Data and
Prevalence of Overweight. United States, 1976±80. Vital and
Health Statistics, Series 11, No 238, DHEW Publication (PHS)
87±168. 1987. National Center for Health Statistics: Rockville,
pp 1±73.
Haschke F. Body composition of adolescent males. Part I:
Total body water in normal adolescent males. Part II: Body
composition of the male reference adolescent. Acta Paediat
Scand 1983; Suppl. 307: 1±23.
Cote KD, Adams WC. Effect of bone density on body
composition estimates in young adult black and white
women. Med Sci Sports & Exerc 1993; 25: 290±296.
Schutte JE, Townsend EJ, Hugg J, Shoup RF, Malina RM,
Blomqvist CG. Density of lean body mass is greater in blacks
than in whites. J Appl Physiol 1984; 56: 1647±1649.
Heyms®eld S, Lichtman S, Baumgartner R, Wang J, Kamen
Y, Aliprantis A, Pierson R. Body composition of humans±
comparison of 2 technical complexity, and radiation exposure.
Am J Clin Nutr 1990; 51: 52±58.
Guo S, Chumlea WC, Wu X, Wellens R, Roche AF, Siervogel
RM. A comparison of body composition models. In: Ellis KJ,
Eastman JD (eds). Human Body Composition. Plenum Press:
New York, 1993, pp 27±30.
Malina R, Bouchard C. Growth, Maturation and Physical
Activity. Human Kinetics Books: Champaign, IL, 1991. pp 501.
1175