Daily food intake in relation to dietary energy density in the freeliving environment: a prospective analysis of children born at

different risk of obesity13

Tanja VE Kral, Albert J Stunkard, Robert I Berkowitz, Virginia A Stallings, Danielle D Brown, and Myles S Faith

KEY WORDS
Energy density, daily energy intake, daily food
intake, obesity risk status, children, caloric compensation
INTRODUCTION

The prevalence of childhood obesity is on the rise worldwide

(110), and the dietary patterns contributing to this increase are
poorly understood. One dietary factor that has received less attention in the study of pediatric obesity is dietary energy density
(ED), which is defined as calories per weight of food (kcal/g).
Controlled laboratory-based feeding studies found that experimentally increasing the ED of foods significantly increases
short-term energy intake in adults (1114). Specifically, across
conditions of ED, subjects ate a similar weight of food, thereby

increasing their energy intake during a single meal. This suggests

that under laboratory conditions, as well as in the free-living
environment (15), adults regulate the amount of food they consume (by weight or volume) to a greater extent than the calories
they consume.
Studies of children that experimentally altered the ED of foods
tested childrens ability to compensate for energy (16 23). Energy compensation refers to the adjustment of food intake during
an ad libitum meal in response to variations in the ED from a
preload (ie, a fixed amount of food or liquid consumed before the
meal). As reviewed by Birch and Fisher (24), most laboratory
studies have shown that young children compensate for energy
from a preload, although compensation is usually partial and
shows considerable interindividual variability. In addition, the
effect of compensation ability on longer-term intake regulation
(ie, beyond a single laboratory meal) remains poorly understood.
Laboratory protocols have the advantage of assessing childrens food intake objectively. However, they generally limit the
food choices, which, in turn, may create conditions that alter the
subjects habitual eating behavior. Children may compensate
reasonably well for energy during a single meal under controlled
conditions but not necessarily outside of the laboratory. It is
important to assess childrens eating behavior in relation to dietary ED under free-living conditions when they have access to
an ample array of foods that vary in ED. It is of particular interest
to further elucidate the nature of childrens intake regulation.
Specifically, do children regulate the number of calories they
consume on a daily basis (25), or do they regulate the amount or
the volume of food they ingest, as the data in adults suggest
(26 28)? Also, are there differences in intake regulation between
children who are born with a different predisposition to obesity
and, if so, how does this affect their ability to compensate? It is
possible that obesity-promoting genes or family environments
influence compensation ability in children. It has also been
1

From the University of Pennsylvania, Philadelphia, PA (TVEK, AJS,

RIB, VAS, DDB, and MSF), and The Childrens Hospital of Philadelphia,
Philadelphia, PA (RIB, VAS, and MSF).
2
Supported by NIH grant DK068899 and the General Clinical Research
Center (grant RR00240) of the Childrens Hospital of Philadelphia.
3
Address reprint requests to TVE Kral, Center for Weight and Eating Disorders, University of Pennsylvania School of Medicine, 3535 Market Street, 3rd
Floor, Philadelphia, PA 19104. E-mail: tkral@mail.med.upenn.edu.
Received January 18, 2007.
Accepted for publication March 16, 2007.

Am J Clin Nutr 2007;86:417. Printed in USA. 2007 American Society for Nutrition

41

Downloaded from ajcn.nutrition.org by guest on May 26, 2016

ABSTRACT
Background: Young children adjust their short-term intake in response to variations in energy density (ED; kcal/g) from preloads in
laboratory studies. It remains unknown whether this compensation
also occurs under free-living conditions.
Objective: The aims of the study were to test whether children aged
3 6 y regulate their habitual daily food (g) and energy (kcal) intakes
in relation to ED and whether compensation differs for children born
at different risk of obesity.
Design: Participants were children born at high risk (n 22) or low
risk (n 27) of obesity on the basis of maternal prepregnancy body
mass index (BMI; in kg/m2). Daily ED, food intake, and energy
intake were assessed from 3-d food records that either included or
excluded beverages. Intake regulation was explored by relating childrens daily food and energy intakes to ED and, more importantly, by
examining residual scores derived by regressing daily food intake
on ED.
Results: For both risk groups, daily food intake was inversely correlated with ED (P 0.05), whereas daily energy intake was not
significantly correlated with ED at most ages (P 0.05). In analyses
that excluded beverages, mean residual scores significantly increased from 3 to 6 y of age in high-risk children, which indicates
relative overconsumption, but decreased in low-risk children, which
indicates relative underconsumption (risk group time interaction,
P 0.005).
Conclusions: Children adjusted their daily food intake in relation to
ED, which suggests caloric compensation under free-living conditions. Compensation ability may deteriorate with age in a manner
that favors relative food overconsumption among obesity-prone
children.
Am J Clin Nutr 2007;86:417.

42

KRAL ET AL

suggested that calories consumed from beverages are less well

regulated than are those from solid foods (29, 30). Thus, it is
possible that the results for childrens intake regulation may
differ when liquids (ie, beverages) are included or excluded
from the analyses.
A first aim of this study was to examine the interrelation
between daily ED (kcal/g), daily food intake (g), and daily energy
intake (kcal) among children born at high or low risk of obesity.
We predicted that children would reduce their daily food intake
as a function of the ED of their diet, thereby regulating daily
energy intake. A second aim was to examine childrens daily
food intake in relation to their predicted daily food intake for a
given level of daily energy density and to assess potential
changes in childrens compensation ability over time (from ages
3 to 6 y). We predicted that childrens daily food intakes, relative
to their predicted ones, would differ as a function of their obesity
risk status. We also predicted that compensation ability would
deteriorate over time in both risk groups. All analyses were
completed including or excluding all beverages.

Subjects
The subjects in this report were part of an ongoing longitudinal
study of growth and development in early life that has been
conducted at the University of Pennsylvania and the Childrens
Hospital of Philadelphia. The children in this study were born at
either low or high risk of obesity (31). The childrens obesity risk
status was based on maternal prepregnancy body mass index
(BMI; in kg/m2). Infants born to mothers with a prepregnancy
BMI less than the 33rd percentile (mean BMI of 19.5 1.1) were
classified as being at low risk of obesity (n 37). Children born
to mothers with a prepregnancy BMI greater than the 66th percentile (mean BMI of 30.3 4.2) were classified at being at high
risk of obesity (n 35). Children were enrolled in the study at the
age of 3 mo, and their growth and development were followed up
to year 12. All children in the study were white. Further details of
parental and subject characteristics and the study design were
reported previously (3135). The present report is based on a
subsample of this cohort (low-risk: n 27; high-risk: n 22) for
whom 3-d weighed-food records were available at 3, 4, 5, and 6 y
of age. Written informed consent was obtained from the parents.
The protocol was approved by the institutional review boards of
the University of Pennsylvania and the Childrens Hospital of
Philadelphia.
Dietary assessment
Each year, within 2 wk of each childs birthday, the primary
caretakers of children in both risk groups were asked to complete
3-d weighed-food records (2 weekdays, 1 weekend day). Caretakers were provided electronic food scales to preweigh all foods
and beverages (except water) consumed by the child and to weigh
all leftovers. Food records were analyzed by research nutritionists at the General Clinical Research Center at the Childrens
Hospital of Philadelphia by using the Food Processor Nutrition
Analysis software (ESHA Research, Salem, OR). Only those
food records that were completed for 2 d were included in
the present analyses. Two- and three-day food records were

Dietary outcome measures

The following 3 dietary outcome measures were computed
from the 3-d weighed-food records: daily ED (kcal/g), daily food
intake (g), and daily energy intake (kcal). All 3 outcome variables
were computed either to include or exclude all beverages. The
daily dietary ED was computed by dividing the total daily calories by the total amount of food and beverages (ie, weight in
grams) consumed (36). Daily food and energy intakes were computed as the sums of the individual weights and calories consumed from foods and beverages on a given day. All dietary
outcome measures were averaged across the total number of days
of completed records included in the analyses. A detailed description of the childrens daily energy density and daily energy
intake is provided elsewhere (37).
Statistical analysis
Descriptive statistics are presented as means SEMs. To test
aim one, a Pearson correlation analysis examined relations between daily ED and daily energy and food intakes. Dietary ED
was correlated first with daily food intake and second with daily
energy intake within each risk group for each year. These correlation analyses were completed by using intakes that included or
excluded all beverages.
Because the correlations tested in aim one could have been due
in part to the functional relation between these variables (38), we
proceeded by using an analysis of residuals to further examine
childrens intake regulation. Thus, to test the second aim of the
study, an analysis of residuals was conducted in which, for each
age, daily food intake was regressed onto daily ED and the
residual scores were saved. The residual is that part of daily food
intake that is uncorrelated with daily ED. Residuals are the differences between an observed value of the response variable and
the value predicted by the model. In our model, the residuals
represent the vertical difference between the regression line (or
the predicted value) and the actual (or observed) data. The residual represents the difference between an individuals actual
food intake and what their food intake would be expected to be
for a given level of ED based on this sample. Thus, the residuals
allow us to break down the sample into children who eat more or
less than would be expected. Positive residuals represent greater
food consumption than would be expected (ie, overconsumption), whereas negative residuals represent less food consumption than expected (ie, underconsumption). A residual of zero
represents what the model would predict food intake to be for a
given ED value on the basis of this sample; thus, the deviation
from the perfect prediction is zero.
A 2 (risk group) 4 (age) mixed linear model analysis of
variance (ANOVA) tested for (linear) differences in the mean
residual food scores as a function of risk group and child age.

Downloaded from ajcn.nutrition.org by guest on May 26, 2016

SUBJECTS AND METHODS

available for the following numbers of children at ages 3, 4, 5,

and 6 y: 45 (40 three-day records and 5 two-day records); 48 (45
three-day records and 3 two-day records); 42 (38 three-day
records and 4 two-day records); and 42 (39 three-day records
and 3 two-day records), respectively. One three-day food record
was not included in the analyses because the child was fed infant
formula almost exclusively. Likewise, 2 one-day records were
not included because the children were either sick or fell asleep
before dinner.

43

DAILY INTAKE IN RELATION TO ENERGY DENSITY

TABLE 1
Daily energy intake, food intake, and energy density stratified by age and risk group1
Daily intake and risk group
Energy intake, food only (kcal)2
Low risk
High risk
Energy intake, food and beverages (kcal)4
Low risk
High risk
Food intake, food only (g)5
Low risk
High risk
Food intake, food and beverages (g)6
Low risk
High risk
Energy density, food only (kcal/g)7
Low risk
High risk
Energy density, food and beverages (kcal/g)8
Low risk
High risk

Age 3 y

Age 4 y

Age 5 y

Age 6 y

898 47
847 51

939 55
965 57

1064 52
1177 70

1159 45
1393 673

1145 49
1189 59

1219 62
1253 61

1311 58
1493 84

1435 43
1687 69

443 25
395 24

451 30
447 24

466 25
526 34

525 31
620 36

948 45
1062 74

1018 58
1097 58

980 54
1210 90

1104 42
1262 51

2.14 0.10
2.23 0.09

2.23 0.08
2.24 0.11

2.37 0.09
2.36 0.11

2.31 0.08
2.34 0.10

1.26 0.06
1.21 0.07

1.27 0.05
1.19 0.05

1.39 0.06
1.30 0.07

1.35 0.06
1.39 0.09

Risk group was a between-subjects variable with 2 levels (lowrisk or high-risk), and child age was a within-subjects variable
with 4 levels (3, 4, 5, or 6 y). A main effect of risk group would
imply that the tendency to overconsume food differs for high-risk
and low-risk children. Because of the repeated-measures structure of the data, the model used restricted maximum likelihood
estimates and a compound symmetry error structure. Significant
main effects or interactions were followed up by pair-wise comparisons, as well as contrasts, to test for potential linear effects of
time across all 4 ages and linear time by risk group interactions.
Again, these analyses were completed by using intakes that included or excluded all beverages.
The ANOVA model to test aim 2 was run with and without
child BMI z score as a covariate to ensure that any risk group
differences in residual scores were not merely due to differences
in childrens weight status. We identified 2 children in the lowrisk group who, at age 4 y, qualified as statistical outliers per
Tukeys criteria (39) with regard to BMI z score within their
respective risk group. We comment in the Results section that the
outcomes did not change when the outliers were removed.
We adjusted the level by using the Bonferroni correction
(ie, 0.05/n, where n equals the number of comparisons in
a given analysis). We note associations that were significant
both at the conventional 0.05 level as well as when using
the Bonferroni correction because this correction can be
overly stringent (40). We used the same approach in previous
analyses of this cohort (33).
The data were analyzed by using SPSS software (version 12.0;
SPSS Inc, Chicago, IL) and SAS software (version 9.1; SAS

Institute, Inc, Cary, NC). For all analyses, P values 0.05 were
considered statistically significant.
RESULTS

Child characteristics
The number of low-risk children included in the present study
at ages 3, 4, 5, and 6 y were 23 (13 boys, 10 girls), 27 (14 boys,
13 girls), 23 (11 boys, 12 girls), and 22 (8 boys, 14 girls); the
number of high-risk children were 22 (12 boys, 10 girls), 21 (9
boys, 12 girls), 19 (10 boys, 9 girls), and 20 (11 boys, 9 girls),
respectively. The mean (SEM) BMI z scores for high-risk
children were 0.4 0.3, 0.4 0.3, 0.5 0.4, and 0.2 0.3,
and those for low-risk children were 0.4 0.2, 0.1 0.1,
0.1 0.2, and 0.3 0.2 at ages 3, 4, 5, and 6 y, respectively.
Student t tests showed that, at each of the years, none of these
BMI z scores were significantly different between risk
groups. More detailed analyses of anthropometric measures
for this subsample (37) as well as the full cohort (34) are
provided elsewhere.
Daily food intake, energy intake, and energy density
The mean daily food intake (g), daily energy intake (kcal), and
daily energy density (kcal/g) for children in both risk groups at
ages 3, 4, 5, and 6 y are shown in Table 1. These data are
displayed for descriptive purposes only. A detailed discussion of
these results can be found in a previously published paper by Kral
et al (37) that investigated the changes in dietary energy density
in children in this cohort over time.

Downloaded from ajcn.nutrition.org by guest on May 26, 2016

All values are x SEM. These data were part of a previous publication by Kral et al (37).
2 4 ANOVA indicated a significant (linear) risk group time interaction (P 0.02).
3
Significantly different from the low-risk group, P 0.05.
4
2 4 ANOVA indicated a significant main and linear effect of time (P 0.0001) but no significant risk group time interaction (P 0.19).
5
2 4 ANOVA indicated a significant risk group time interaction (P 0.02).
6
2 4 ANOVA indicated a significant linear effect of time (P 0.0001) but no risk group time interaction (P 0.40).
7
2 4 ANOVA indicated a significant linear effect of time (P 0.03) but no risk group time interaction (P 0.98).
8
2 4 ANOVA indicated a significant linear effect of time (P 0.002) but no risk group time interaction (P 0.58).
1

44

KRAL ET AL

TABLE 2
Pearson correlation coefficients summarizing the relation between dietary energy density and daily food and energy intakes (excluding beverages) by risk
group and child age
Pearsons r with daily energy density (kcal/g)1
Risk group
Low risk
High risk
Low risk
High risk

Daily intake

Age 3 y

Age 4 y

Age 5 y

Age 6 y

Food intake, food only (g)

Food intake, food only (g)
Energy intake, food only (kcal)
Energy intake, food only (kcal)

0.55
0.405
0.31
0.21

0.49
0.23
0.10
0.503

0.50
0.652,4
0.30
0.13

0.702,4
0.513
0.01
0.29

With the use of Fisher r-to-z transformations and a 2-tailed significance test, none of the correlation coefficients was significantly different within each
age group or across age groups.
2
Correlation significant at 0.01 level.
3
Correlation significant at 0.05 level.
4
P 0.006 (Bonferroni correction, 0.05/8).
5
P 0.06.

Relation between daily food intake and energy density

and between daily energy intake and energy density

Residual analysis

All beverages excluded

Mean residual food intake scores for each risk group by child
age are displayed in Figure 1. The 2 (risk group) 4 (age)
ANOVA indicated a significant risk group age category interaction (P 0.005) and a linear risk group age (as a continuous variable) interaction (P 0.0005). This interaction was still
significant when applying a Bonferroni-adjusted alpha in the
respective ANOVA models (0.05/3 0.02). Specifically, mean
residual scores increased between ages 3 and 6 y among high-risk
children, reflecting a greater tendency to overconsume food relative to predicted intake, but decreased among low-risk children,
reflecting a greater tendency to underconsume food relative to
predicted intake. A pairwise follow-up comparison indicated that
the mean residual food intake score was significantly greater in
the high-risk children than in the low-risk children at age 6 y
(residual 53.4 137.0 versus 48.5 105.1; P 0.01).
The risk group age interaction remained significant (P
0.02) when BMI z score was added to the model as a covariate.
The main effect of BMI z score was significant in this model
(P 0.04), indicating that heavier children tended to overconsume food (relative to their predicted intake) compared with
thinner children. After removal of the 2 statistical outliers for
BMI z score from the model, the main effect of BMI z score was
not significant (P 0.06); however, the risk group age
interaction remained significant (P 0.02). We note that this

All beverages included

Daily food intake was significantly inversely correlated with
daily ED for both the low-risk and the high-risk group at all ages
(P 0.05; Table 3). Six of these 8 associations remained significant when applying a Bonferroni-adjusted (0.05/8
0.006). Daily energy intake, however, was not significantly correlated with daily ED for either group at any age (Table 3).

TABLE 3
Pearson correlation coefficients summarizing the relations between dietary energy density and daily food and energy intakes (including beverages) by risk
group and child age
Pearsons r with daily energy density (kcal/g)1
Risk group
Low risk
High risk
Low risk
High risk
1

Daily intake

Age 3 y

Age 4 y

Age 5 y

Age 6 y

Food intake, food and beverages (g)

Food intake, food and beverages (g)
Energy intake, food and beverages (kcal)
Energy intake, food and beverages (kcal)

0.58
0.692,3
0.28
0.002

0.45
0.494
0.13
0.40

0.64
0.622,3
0.05
0.13

0.782,3
0.612,3
0.26
0.42

2,3

2,3

With the use of Fisher r-to-z transformations and a 2-tailed significance test, none of the correlation coefficients was significantly different within each
age group or across age groups.
2
Correlation significant at 0.01 level.
3
P 0.006 (Bonferroni correction, 0.05/8).
4
Correlation significant at 0.05 level.

Downloaded from ajcn.nutrition.org by guest on May 26, 2016

Daily food intake was significantly inversely correlated with

daily ED for the low-risk group at all ages and for the high-risk
group at ages 5 and 6 y (P 0.05; Table 2). The inverse relation
was borderline significant in the high-risk group at 3 y of age
(P 0.06). This indicates that children who consumed a diet that
was relatively higher in ED tended to eat less food on a daily
basis than did children who consumed a diet that was relatively
lower in ED. Two of these 8 associations remained significant,
and 2 were borderline significant, when applying a Bonferroniadjusted (0.05/8 0.006).
Daily energy intake was not significantly correlated with daily
ED for the low-risk group at all ages and for the high-risk group
at ages 3, 5, and 6 y (Table 2). This indicates that across a range
of dietary ED, daily caloric intake did not vary significantly as a
function of ED during these years.

All beverages excluded

DAILY INTAKE IN RELATION TO ENERGY DENSITY

100

There was a main effect of BMI z score when added to the

model as a covariate (P 0.0001), whereas the main effect of
risk group (P 0.21) and the risk group age interaction (P
0.76) were not significant. After removal of the 2 outliers, the
main effect of BMI z score remained significant (P 0.0001);
the main effect of risk group remained nonsignificant (P 0.19),
as did the risk group age interaction (P 0.72).

UC <---------------> OC

80
60
40
20
0
-20
-40

DISCUSSION

-60
-80

Age (y)

interaction effect approached but did not reach significance

when using the Bonferroni-adjusted (0.05/4 0.01).
All beverages included
Mean residual food intake scores for each risk group by child
age are displayed in Figure 2. The 2 (risk group) 4 (age)
ANOVA indicated a trend for a main effect of risk group (P
0.07), which suggests a tendency for high-risk children to overconsume, relative to their predicted intake, compared with
low-risk children across all 4 y. The risk group age interaction was not significant (P 0.55), nor was the linear risk
group age interaction (P 0.38). None of the main effects
or the interaction was significant when using a Bonferroniadjusted (0.05/3 0.02).
160
110
60
10
-40
-90
-140

Age (y)

FIGURE 2. Mean residual food intake (food and beverages) scores by risk
group (, low risk; f, high risk) and child age. Residual food intake scores
were generated by regressing the daily food intake (including all beverages)
onto daily energy density (including all beverages). These residuals represent
the difference between childrens actual food and beverage intakes and their
predicted food and beverage intakes for a given energy density based on this
sample. UC, relative underconsumption; OC, relative overconsumption. The
risk group age interaction for residual scores was not significant (P
0.55), nor was the linear risk group age interaction (P 0.38).

The main finding of the present study was that across a range
of ED, children born at high or low risk of obesity adjusted the
daily amount of food they consumed on the basis of the ED of
their diet. Children who consumed a diet that was higher in ED
consumed less food daily than did children whose diet was lower
in ED. These findings held true regardless of whether beverages
were included or excluded. This is the first longitudinal study to
show that children aged 3 6 y also adjust their daily intake under
free-living conditions. These findings thus confirm and extend
previous research conducted under laboratory conditions.
The prospective nature of this design also allowed us to test for
changes in energy regulation ability over time. It has been suggested that developmental differences in energy compensation
ability may exist in children (24). Preschool children, on average,
show compensation (24); however, Anderson et al (41) found no
effect on food intake when varying the ED of preloads in 9 10y-old children. These findings indirectly suggest developmental
changes in compensation ability; however, this phenomenon has
never been documented in long-term prospective studies that
track children during growth.
With the use of a novel approach of residual score analysis, this
present study assessed changes in compensation ability over the
course of 4 y. The results indicated different patterns of residuals
between high- and low-risk children, which suggests that familial predisposition to obesity may partially operate through errors
in compensation. This supports theories and experimental studies regarding energy compensation ability and the development
of pediatric obesity (2224, 42). However, it should be pointed
out that the results from the residual analysis are sampledependent and therefore may not apply to other cohorts of children. The main finding of the residual analysis was that high-risk
children tended to overconsume, whereas low-risk children
tended to underconsume, relative to their predicted intake, although the results differed depending on whether beverages were
included in the analyses. For analyses that excluded beverages,
the risk group differences gradually emerged over time, with
high-risk children not showing a tendency to overconsume until
after 4 y of age. This is consistent with the notion that compensation ability gradually deteriorates during early childhood, possibly starting during the years corresponding to adiposity rebound (43), rather then being compromised consistently and as
early as 3 y of age. This gradual decline in compensation ability
was not seen when beverages were included in the analyses (see
below). Another key finding was that the risk group differences
remained significant even when the analyses were controlled for
BMI z score, which suggests that the errors in compensation
cannot be fully explained by differences in weight status. Again,
this was not the case when beverages were included in the analyses (see below).
When beverages were included in the residual score analyses,
the results indicated a trend (P 0.07) for high-risk children to

Downloaded from ajcn.nutrition.org by guest on May 26, 2016

FIGURE 1. Mean residual food intake (food only) scores by risk group
(, low risk; f, high risk) and child age. Residual food intake scores were
generated by regressing daily food intake (excluding all beverages) onto
daily energy density (excluding all beverages). These residuals represent the
difference between childrens actual food intake and their predicted food
intake for a given energy density based on this sample. UC, relative underconsumption; OC, relative overconsumption. There was a significant (linear)
risk group age interaction (P 0.005) for residual scores. *Significantly
different from the low-risk group, P 0.01.

UC <---------------> OC

45

46

KRAL ET AL

are ethnically diverse and who show greater variability in their

weight status and age than did the children in the current cohort.
Thus, the results from this study are limited to the relatively
narrow study sample, namely, healthy, full-term, white children,
and may not be generalizable to children as a whole. Second, it
would be desirable to obtain intake data over a longer period of
time at each age to assess childrens habitual diet. Third, the
validity of the data may be increased by complementing selfreported intakes, which may be subject to measurement or reporting errors by mothers, with measured intake data in the feeding laboratory. Fourth, future research should also collect data of
energy expenditure to be able to relate childrens intake data to
their energy needs. Last, the present design, unlike a classic twin
design or other more genetically sensitive designs, could not
formally test the respective influence of genetic and environmental influences on compensation ability.
In conclusion, the results of the present study suggest that
children between the ages of 3 and 6 y have the ability to adjust
their daily food intake on the basis of the ED of their diet to
maintain a certain daily energy intake in the free-living environment. Energy compensation ability, however, may differentiate
children who were born at high-risk or at low-risk of obesity.
Future research is needed that separates the effects of food and
beverages on childrens compensation ability.
We thank the General Clinical Research Center of the Childrens Hospital
of Philadelphia and the staff of the Infant Growth Study for their support. We
also thank Michael J Rovine for statistical consulting on the study.
The contributions of the authors were as followsTVEK: contributed to
the study design, data acquisition and analysis, and writing of the manuscript;
AJS, RIB, VAS, and MSF: were responsible for the study concept, acquisition of the data, writing of the manuscript, and obtaining funding; DDB:
contributed to the data analysis. None of the authors had any conflicts of
interest.

REFERENCES
1. Livingstone MB. Childhood obesity in Europe: a growing concern. Public Health Nutr 2001;4:109 16.
2. James PT. Obesity: the worldwide epidemic. Clin Dermatol 2004;22:
276 80.
3. Kelishadi R, Pour MH, Sarraf-Zadegan N, et al. Obesity and associated
modifiable environmental factors in Iranian adolescents: Isfahan
Healthy Heart Program - Heart Health Promotion from Childhood. Pediatr Int 2003;45:435 42.
4. al-Nuaim AR, Bamgboye EA, al-Herbish A. The pattern of growth and
obesity in Saudi Arabian male school children. Int J Obes Relat Metab
Disord 1996;20:1000 5.
5. Freedman DS, Srinivasan SR, Valdez RA, Williamson DF, Berenson
GS. Secular increases in relative weight and adiposity among children
over two decades: the Bogalusa Heart Study. Pediatrics 1997;99:420 6.
6. Zametkin AJ, Zoon CK, Klein HW, Munson S. Psychiatric aspects of
child and adolescent obesity: a review of the past 10 years. J Am Acad
Child Adolesc Psychiatry 2004;43:134 50.
7. Kotani K, Nishida M, Yamashita S, et al. Two decades of annual medical
examinations in Japanese obese children: do obese children grow into
obese adults? Int J Obes Relat Metab Disord 1997;21:91221.
8. Lobstein TJ, James WP, Cole TJ. Increasing levels of excess weight
among children in England. Int J Obes Relat Metab Disord 2003;27:
1136 8.
9. Moreno LA, Sarria A, Popkin BM. The nutrition transition in Spain: a
European Mediterranean country. Eur J Clin Nutr 2002;56:9921003.
10. Rolland-Cachera MF, Deheeger M, Thibault H. [Epidemiologic bases of
obesity. ] Arch Pediatr 2001;8(suppl 2):287s9s (in French).
11. Bell EA, Castellanos VH, Pelkman CL, Thorwart ML, Rolls BJ. Energy
density of foods affects energy intake in normal-weight women. Am J
Clin Nutr 1998;67:41220.
12. Bell EA, Rolls BJ. Energy density of foods affects energy intake across

Downloaded from ajcn.nutrition.org by guest on May 26, 2016

overconsume and low-risk children to underconsume across all

4 y. That is, errors in compensation did not appear to gradually
emerge over time but may have been present as early as 3 y of age.
This interpretation, if true, would suggest that errors in compensation ability for (food and) beverages are established by 3 y of
age and, therefore, that any deterioration in compensation ability
may have occurred before the age of 3 y. This finding raises the
possibility that infancy and the first 2 y of life may be critical
periods for the development of compensation ability for food and
beverages (20, 21, 44, 45) and therefore may be important periods for early-life interventions. To more convincingly identify
critical periods of time when childrens ability to compensate
may start to deteriorate, however, studies are needed that examine (or track) the stability of compensation indexes. Findings
from the current study are also consistent with studies reporting
that excess beverage consumption (especially from fruit juice) is
associated with excess weight gain in preschool-age children
(46, 47). Hunger and thirst mechanisms are believed to be distinct
entities, because the consumption of beverages primarily acts
on thirst and not hunger mechanisms (48). It is possible that
liquids, which are thought of as having a generally lower
satiating efficiency (49), may thus further weaken childrens
compensation ability.
Another unique finding of the residual score analyses that
included beverages was that the risk group differences no longer
approached statistical significance when BMI z score was added
as a covariate. BMI z score was consistently associated with
greater residual scores across analyses, which suggests that errors in compensation ability for beverages (and food) may be
affected by child weight status and the determinants thereof.
Heavier children may be more likely to exceed their energy needs
through consumption of beverages (46). Despite these findings,
it should be noted that beverages, because of their low energy
density, add weight but fewer calories to the daily total, and, thus,
by computation, disproportionately influence dietary ED.
Hence, residual analyses that include beverages may need to be
interpreted with more caution.
The result of this investigation agrees with findings from 2
recent experimental studies (22, 23) that showed that energy
compensation in children 1) is often incomplete and shows great
interindividual variability and 2) tends to decline with age. It is
possible that children, especially those who are genetically predisposed to obesity, may become more responsive to environmental influences (such as the portion size of foods) as they
become older. A study conducted by Rolls et al (50) showed that
when presented with larger portions of food, 5-y-old children,
but not 3-y-old children, significantly increased their intakes
relative to when they were presented with smaller portions of
food. The authors suggested that as children age they become
increasingly responsive to environmental, social, and cultural
factors that all are likely to affect their intake. To what extent
these factors increase childrens vulnerability to overeat may
depend on parental obesity, a line of research that has yet to be
further explored.
The strengths of the present study include 1) the unique sample
of children born at different risk of obesity and who were prospectively studied and 2) the use of extensive training and electronic food scales to measure the weights of the foods that the
children consumed. The study might well be extended in several
ways. First, it would be desirable to examine ability to adjust food
intake in relation to dietary ED in a larger cohort of children who

DAILY INTAKE IN RELATION TO ENERGY DENSITY

13.
14.
15.
16.
17.
18.
19.
20.

22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.

33.

34.

35.

36.

37.

38.

39.
40.
41.

42.
43.

44.

45.

46.

47.

48.
49.

50.

Eating in the absence of hunger: a genetic marker for childhood obesity

in prepubertal boys? Obes Res 2006;14:131 8.
Faith MS, Berkowitz RI, Stallings VA, Kerns J, Storey M, Stunkard AJ.
Parental feeding attitudes and styles and child body mass index: prospective analysis of a gene-environment interaction. Pediatrics 2004;
114:e429 36.
Berkowitz RI, Stallings VA, Maislin G, Stunkard AJ. Growth of children
at high risk of obesity during the first 6 y of life: implications for prevention. Am J Clin Nutr 2005;81:140 6.
Stunkard AJ, Berkowitz RI, Schoeller D, Maislin G, Stallings VA. Predictors of body size in the first 2 y of life: a high-risk study of human
obesity. Int J Obes Relat Metab Disord 2004;28:50313.
Ledikwe JH, Blanck HM, Khan LK, et al. Dietary energy density determined by eight calculation methods in a nationally representative United
States population. J Nutr 2005;135:273 8.
Kral TVE, Berkowitz RI, Stunkard AJ, Stallings VA, Brown DD, Faith
MS. Dietary energy density increases during early childhood irrespective of familial predisposition to obesity: results from a prospective
cohort study. Int J Obes (in press).
Cohen J, Cohen P, West SG, Aiken LS. Applied multiple regression/
correlation analysis for the behavioral sciences. 3rd ed. Mahwah, NJ:
Lawrence Erlbaum Associates, 2003.
Tukey J. Exploratory data analysis. Reading, MA: Addison-Wesley
Publishing Company, 1977.
Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing. Stat Med 1990;9:811 8.
Anderson GH, Saravis S, Schacher R, Zlotkin S, Leiter LA. Aspartame:
effect on lunch-time food intake, appetite and hedonic response in children. Appetite 1989;13:93103.
Johnson SL, Birch LL. Parents and childrens adiposity and eating style.
Pediatrics 1994;94:653 61.
Rolland-Cachera MF, Deheeger M, Maillot M, Bellisle F. Early adiposity rebound: causes and consequences for obesity in children and adults.
Int J Obes 2006;30:S117.
Fomon SJ, Filer LJ, Thomas LN, Anderson TA, Nelson SE. Influence of
formula concentration on caloric intake and growth of normal infants.
Acta Pediatr Scand 1975;64:172 81.
Birch LL, McPhee L, Sullivan S. Childrens food intake following drinks
sweetened with sucrose and aspartame: time course effects. Physiol
Behav 1989;45:38796.
Faith MS, Dennison BA, Edmunds LS, Stratton HH. Fruit juice intake
predicts increased adiposity gain in children from low-income families:
weight status-by-environment interaction. Pediatrics 2006;118:2066
75.
Welsh JA, Cogswell ME, Rogers S, Rockett H, Mei Z, Grummer-Strawn
LM. Overweight among low-income preschool children associated with
the consumption of sweet drinks: Missouri, 1999 2002. Pediatrics
2005;115:e2239.
Fitzsimons JT. The physiology of thirst and sodium appetite. Cambridge,
United Kingdom: Cambridge University Press, 1979.
DellaValle DM, Roe LS, Rolls BJ. Does the consumption of caloric and
non-caloric beverages with a meal affect energy intake? Appetite 2005;
44:18793.
Rolls BJ, Engell D, Birch LL. Serving portion size influences 5-year-old
but not 3-year-old childrens food intakes. J Am Diet Assoc 2000;100:
232 4.

Downloaded from ajcn.nutrition.org by guest on May 26, 2016

21.

multiple levels of fat content in lean and obese women. Am J Clin Nutr
2001;73:1010 8.
Kral TVE, Roe LS, Rolls BJ. Does nutrition information about the
energy density of meals affect food intake in normal-weight women?
Appetite 2002;39:137 45.
Rolls BJ, Bell EA, Castellanos VH, Chow M, Pelkman CL, Thorwart
ML. Energy density but not fat content of foods affected energy intake
in lean and obese women. Am J Clin Nutr 1999;69:86371.
Poppitt SD, Prentice AM. Energy density and its role in the control of
food intake: evidence from metabolic and community studies. Appetite
1996;26:15374.
Birch LL, Deysher M. Caloric compensation and sensory specific satiety: evidence for self regulation of food intake in young children. Appetite 1986;7:32331.
Birch LL, Deysher M. Conditioned and unconditioned caloric compensation: evidence for self regulation of food intake by young children.
Learn Motiv 1985;16:34155.
Birch LL, McPhee L, Shoba C, Steinberg L, Krehbiel R. Clean up your
plate: effects of child feeding practices on the conditioning of meal size.
Learn Motiv 1987;18:30117.
Johnson SL, McPhee L, Birch LL. Conditioned preferences: young
children prefer flavors associated with high dietary fat. Physiol Behav
1991;50:124551.
Birch LL, Johnson SL, Jones MB, Peters JC. Effects of a non-energy fat
substitute on childrens energy and macronutrient intake. Am J Clin Nutr
1993;58:326 33.
Fomon SJ, Filer LJ, Thomas LN, Rogers RR, Proksch AM. Relationship
between formula concentration and rate of growth in normal infants. J
Nutr 1969;98:24154.
Cecil JE, Palmer CN, Wrieden W, et al. Energy intakes of children after
preloads: adjustment, not compensation. Am J Clin Nutr
2005;82:302 8.
Johnson SL, Taylor-Holloway LA. Non-Hispanic elementary school
childrens self-regulation of energy intake. Am J Clin Nutr 2006;83:
1276 82.
Birch LL, Fisher JO. Food intake regulation in children. Fat and sugar
substitutes and intake. Ann N Y Acad Sci 1997;819:194 220.
Birch LL, Johnson SL, Andresen G, Peters JC, Schulte MC. The variability of young childrens energy intake. N Engl J Med 1991;324:
2325.
Duncan KH, Bacon JA, Weinsier RL. The effects of high and low energy
density diets on satiety, energy intake, and eating time of obese and
nonobese subjects. Am J Clin Nutr 1983;37:7637.
Lissner L, Levitsky DA, Strupp BJ, Kalkwarf HJ, Roe DA. Dietary fat
and the regulation of energy intake in human subjects. Am J Clin Nutr
1987;46:886 92.
Kendall A, Levitsky DA, Strupp BJ, Lissner L. Weight loss on a low-fat
diet: consequence of the imprecision of the control of food intake in
humans. Am J Clin Nutr 1991;53:1124 9.
Flood JE, Roe LS, Rolls BJ. The effect of increased beverage portion size
on energy intake at a meal. J Am Diet Assoc 2006;106:1984 90.
Mattes R. Fluid calories and energy balance: the good, the bad, and the
uncertain. Physiol Behav 2006;89:66 70.
Stunkard AJ, Berkowitz RI, Stallings VA, Schoeller DA. Energy intake,
not energy output, is a determinant of body size in infants. Am J Clin Nutr
1999;69:524 30.
Faith MS, Berkowitz RI, Stallings VA, Kerns J, Storey M, Stunkard AJ.

47