Carbohydrate balance predicts weight and fat gain in adults13

Robert H Eckel, Teri L Hernandez, Melanie L Bell, Kathleen M Weil, Trudy Y Shepard, Gary K Grunwald,
Teresa A Sharp, Coni C Francis, and James O Hill

KEY WORDS
Carbohydrate balance, dietary fat, dietary carbohydrate, metabolic predictor, weight gain, fat mass gain, body
composition, energy balance, insulin sensitivity, obesity, indirect
calorimetry
INTRODUCTION

Because obesity now affects 44 million adults in the United

States alone (1), its control creates a public health challenge.
Although most of the population is overweight (2), some persons
seem more resistant to weight gain than others. Persons can be
susceptible to weight gain because of either behavioral or metabolic factors (3). Consequently, finding effective prevention
strategies has become crucial.
Although genetics plays a role in weight gain (4), it cannot
independently explain the dramatic increase in the development
of obesity over the past several decades (5). During the past 20 y,

several metabolic predictors of obesity, such as lower basal metabolic rates (6, 7), reductions in nonexercise activity thermogenesis (NEAT) (8), increases in carbohydrate (CHO) oxidation (9,
10), insulin sensitivity (SI) (1113), low concentrations of leptin
(14), and reduced levels of sympathetic nervous system activity
(15, 16), were reported. A better understanding of these metabolic factors could be useful in identifying obesity prevention
interventions for those at greater risk.
Exposing persons to an interval of positive energy balance
could elucidate metabolic differences in the propensity for
weight gain. The standard energy balance equation, by definition, indicates equality between energy intake (EI) and energy
expenditure (EE; EI EE). Conversely, positive energy balance
indicates that EI exceeds EE (EI EE), which sets the stage for
weight gain. Energy balance can also be quantified for each of the
macronutrients (protein, fat, and CHO), such that CHO balance
is the rate of CHO intake minus the rate of CHO oxidation,
reflecting a change in CHO stores. Thus, the term positive CHO
balance reflects storage of CHO in the body (17).
The purpose of the present study was to investigate the effect
of diets on the degree to which the components of energy balance
and SI predict long-term changes in weight and body composition. Here, we address the question of whether prospective differences in energy balance on a high-CHO diet compared with a
high-fat diet during 1 day of physical inactivity could predict
weight or adipose tissue gain over 4 y.
SUBJECTS AND METHODS

Subjects
The Colorado Multiple Institutional Review Board at the University of Colorado at Denver and Health Sciences Center
1

From the Division of Endocrinology, Metabolism, and Diabetes (RHE,

TLH, KMW, and TYS), the Department of Preventive Medicine and Biometrics
(MLB and GKG), the Department of Pediatrics, Center for Human Nutrition
(TAS and JOH), and the Adult General Clinical Research Center (CCF), University of Colorado at Denver and Health Sciences Center, Denver, CO.
2
Supported by grant R01DK-46881 from the National Institute of Diabetes and Digestive and Kidney Diseases, grant P30 DK-48520-01 from the
Colorado Clinical Nutrition Research Unit (Metabolic and Energy Balance
Laboratories), and grant M01-RR00051 from the Adult General Clinical
Research Center by the NIH Division of Research Resources.
3
Address reprint requests to RH Eckel, Division of Endocrinology, Metabolism, and Diabetes, Division of Cardiology, University of Colorado at
Denver and Health Sciences Center, PO Box 6511, MS 8106, Aurora, CO
80045. E-mail: robert.eckel@uchsc.edu.
Received July 26, 2005.
Accepted for publication January 4, 2006.

Am J Clin Nutr 2006;83:803 8. Printed in USA. 2006 American Society for Nutrition

803

Downloaded from ajcn.nutrition.org by guest on March 28, 2016

ABSTRACT
Background: The prevention and treatment of obesity is a public
health challenge.
Objective: We investigated the effects of dietary composition, insulin sensitivity (SI), and energy balance on predicted changes in
body composition.
Design: In a randomized crossover design study, 39 normal-weight
(n 23), overweight (n 8), and obese (n 8) men and women
(aged 2536 y) each followed a 15-d isocaloric high-fat (HF; 50%
fat) and high-carbohydrate [HC; 55% carbohydrate (CHO)] diet
with a 4 6-wk washout period during the first year. During each
treatment, energy balance was measured while the subjects were
inactive by using indirect calorimetry on day 15, and SI was measured by using a euglycemic clamp study (40 mU m2 min1) on
day 16. Weight and body composition were then measured annually
for 4 y. The outcomes for fat mass, percentage body fat, and weight
were measured by using a linear 2-stage mixed model.
Results: CHO balance (day 15) and SI (day 16) on the HC diet were
highly and significantly correlated (r 0.55, P 0.001). On the HC
diet, the subjects who had a higher positive CHO balance (day 15)
gained less fat mass (P 0.001), percentage body fat (P 0.006),
and weight (P 0.024) over time. When adjusted for SI, CHO
balance remained a significant predictor of changes in fat mass (P
0.021) and percentage body fat (P 0.025).
Conclusions: On a HC diet, the subjects who had a higher positive
CHO balance on day 15 while they were inactive gained less fat mass
during 4 y, a predictive effect independent of SI. As suggested in
rodents, the capacity to expand the glycogen pool might reduce
energy intake and protect against fat and weight gain.
Am J Clin
Nutr 2006;83:803 8.

804

ECKEL ET AL

TABLE 1
Baseline characteristics of subjects1

Age (y)
Sex (M/F)
Weight (kg)
BMI (kg/m2)
RMR (kcal/d)
RQ, fasting
Glucose (mg/dL)
Insulin (U/mL)
SI (GIR; mg m2 min1)

Normal-weight subjects
(n 23)

Overweight subjects
(n 8)

Obese subjects
(n 8)

28 32
14/9
68 12
21.5 2.0
1629 315
0.82 0.06
88.5 8
5.7 2.2
355 126

30 4
6/2
85 114
26.9 1.33
1850 226
0.77 0.05
87.6 9
8.0 3.3
323 184

33 13
2/6
108 235
37.1 6.64
2138 5075
0.79 0.07
93.4 11
20.3 13.3
145 453

n 39. RMR, resting metabolic rate; RQ, respiratory quotient; SI, insulin sensitivity; GIR, glucose infusion rate.
x SD (all such values).
35
Significantly different from normal-weight subjects (t tests with Bonferroni adjustment): 3P 0.0001, 4P 0.001, 5P 0.01.

period; subjects were asked to abstain from rigorous exercise for

24 h before metabolic measurements. SI was measured on day
16 of each phase by using the hyperinsulinemic euglycemic
clamp technique (40 mU m2 min1) (19). The subjects were
discharged for a 4 6-wk washout phase under free-living conditions but returned weekly for weight measurements to ensure
weight stability. Results of the response to the 14-d diet periods
and short-term results of the calorimeter studies were previously
reported (20).

Procedures

Whole-room indirect calorimeter

On day 0, fasting glucose and insulin were measured. At

breakfast on day 0, each subject began 1 of 2 randomized diets,
either a 15-d isocaloric high-fat (HF) diet or a 15-d isocaloric
high-CHO (HC) diet. Initial estimates of individual daily energy
requirements were made based on the Diet Habit Survey (18) and
an ensuing measurement of resting metabolic rate (RMR). Actual energy consumption was adjusted daily during the first 7 d to
achieve weight stability and maintenance of admission weight;
minimal adjustments were made during the second 7 d. The
subjects were weight-stable (within 1 kg) during the 14-d feeding
period.
Each diet phase lasted 14 d and was followed by a 23-h stay in
the whole-room indirect calorimeter, during which the subjects
were fed the same diet as the preceding 14 d. EI was not reduced
for the stay in the calorimeter. Because confinement to the calorimeter limited physical activity, the subjects had a positive
energy balance on day 14. Although positive energy balance can
be achieved through overfeeding, it is likely that a substantial
amount of weight gain in the population results from reductions
in physical activity. Thus, the positive energy balance observed
in the calorimeter could mimic such a population phenomenon.
Physical activity was not controlled per se during each feeding

The whole-room calorimeter was described previously (20,

21). This is a small room (2.6 3.4 m). All oxygen consumption
and carbon dioxide production is continuously measured while
the subjects are inside. EE and oxidation of protein, CHO, and fat
are determined from these measurements (22). This information,
combined with measured nutrient intakes, can be used to determine daily balances for total energy and for each macronutrient.
The accuracy and precision of this whole-room indirect calorimeter system was previously established (21).

Diets
The HC diet provided a macronutrient content of 55% of
energy as CHO, 25% as fat, and 20% as protein. The HF diet
provided 30% of energy as CHO, 50% as fat, and 20% as protein.
During the calorimeter stay, the subjects consumed 3 meals and
2 snacks. All foods were weighed, and the diets were analyzed by
using a computer program (FOOD PROCESSOR PLUS; ESHA
Research, Salem, OR).

Resting metabolic rate

For the purposes of estimating EI and monitoring the effect of
dietary composition on respiratory quotient (RQ), it was necessary to obtain RMR measurements before the subjects were admitted to the whole-room calorimeter. In these instances, the
RMR was measured by using indirect calorimetry (Sensormedics Metabolic Cart, Model 2900; Sensormedics, Yorba Linda,
CA). Measurements were made in the morning after a 12-h fast
and 24-h abstention from exercise. After 30 min of rest, RMR
was measured for 1520 min with the use of a ventilated hood.
Oxygen consumption and carbon dioxide production were used
to calculate RMR according to the formula of Weir (23). Criteria

Downloaded from ajcn.nutrition.org by guest on March 28, 2016

approved the protocol. Thirty-nine normal-weight [body mass

index (BMI; in kg/m2): 25; n 23), overweight (BMI: 25 to
30; n 8), and obese (BMI: 30; n 8) white men and women
(aged 2536 y; BMI range: 18.750.2) (Table 1) were recruited
and granted their written informed consent. The subjects were at
their maximum BMI and were weight stable for 3 mo. The
subjects usual dietary patterns were assessed by using The Diet
Habit Survey (18). Eight subjects dropped out of the study: 2
completed only the initial diet phases, 1 completed through year
1, 2 completed through year 2, and 3 completed through year 3.
Two subjects missed the year 4 time point but did complete a
measurement at year 5. Three subjects missed the year 3 time
point but completed their participation. Because of the statistical
method used, it was possible to use all of the data collected.
Screening laboratory tests included urinalysis; complete blood
count; serum glucose, electrolytes, liver function, and renal function; and thyroid-stimulating hormone. All values were within
normal limits for each subject. None of the subjects were taking
medications that would affect lipid or CHO metabolism.

CARBOHYDRATE BALANCE PREDICTS WEIGHT AND FAT GAIN

for valid RMR was a minimum of 15 min of steady state, determined as 10% fluctuation in minute ventilation and oxygen
consumption and 5% fluctuation in RQ.
Insulin sensitivity
On day 16 of each diet phase, a 3-h 40 mU m2 min1
hyperinsulinemic euglycemic clamp study was performed to
measure SI (19). The euglycemic goal was measured as the fasting blood glucose concentration measured on the morning of day
16 of diet phase 1. The individual SI with each diet was then
measured as the mean glucose infusion rate (in mg m2 min1)
over the last hour of the infusion study (120 180 min). Therefore, the glucose infusion rate is an expression of SI. Fasting
serum insulin concentrations were also considered to be an indicator of relative SI (24). The female subjects were each studied
in the early follicular phase of their menstrual cycle.
Physical activity index

PAI 14-d average EI (in kcal)/RMR at baseline

(1)

Data analysis
The goal of the analysis was to estimate the degree to which
baseline responses to the HC and HF diets predicted changes in
outcomes of body weight, fat mass, and percentage fat during the
next several years. Candidate predictor variables were 24-h energy balance, CHO and fat oxidation and balance, and SI on each
diet. Associations of candidate variables individually with
changes in outcomes separately were estimated by using a
2-stage mixed model (32). All mixed models included a fixedeffect intercept, a fixed effect for time (which represented the
population mean change in outcome), and random effects for
subject-specific intercepts and slopes. Each model also included
a fixed effect for one candidate predictor as well as an interaction
between the predictor and time. The interaction estimates and
tests the association of the predictor with change in outcome.
This analysis is similar to first regressing each subjects outcome
on time to obtain a slope (rate of change of outcome), then
regressing this slope on the predictor, but it is preferred in the
presence of missing data. Two of the 39 subjects had only initial
measurements of weight and fat mass, so these measurements
were omitted from all analyses of changes in weight and fat mass.
Results were similar with the regression approach. Fixed effects
for sex and baseline BMI and their interactions with time were
included in all models to adjust for these characteristics. Analyses were also repeated with adjustment for estimated baseline
physical activity outside the calorimeter. Two forms were considered: PAI (25) and NEAT (8).

Nonexercise activity thermogenesis

NEAT, the thermogenesis that accompanies nonvolitional exercise (8), was calculated by using the following formula:

NEAT EE RMR thermic effect of food

(2)

The thermic effect of food was estimated as 10% RMR (26). As

with PAI, EE outside the chamber (14 d) was estimated as the EI
fed to the subjects to achieve weight stability.
Body weight and body fat mass
To power the present study, the expected weight change over
4 y was 1.2 kg, on the basis of the following formula obtained
from data from the second National Health and Nutrition Examination Survey (27):

Expected weight change 0.3 kg/y x y

(3)

Body weight and body composition were measured annually for

4 y after first beginning the study. However, some variability was
observed in the timing of these measures, as previously described. Body composition was measured by hydrodensitometry,
with residual volume measured simultaneously by using the
open-circuit nitrogen-dilution technique (28). Nitrogen was
measured by using a Med-Science 505-D Nitralizer (St Louis,
MO). Percentage body fat was estimated from body density (average of 710 repeat measurements) by using the revised equation of Brozek et al (29).
Laboratory procedures
Serum insulin concentrations were measured by radioimmunoassay (30). Measurements of total urinary nitrogen were done
by pyrochemiluminescence (31) with the use of the Antek (Houston, TX) nitrogen analyzer system (21).

RESULTS

Weight remained stable during the 14 d of isocaloric HC and

HF feeding (data previously reported) (20). During the calorimeter stay of both diets, the subjects had a positive energy balance
(P 0.001), with mean (SD) energy storage of 859 138
kcal/d on the HC diet and 944 176 kcal/d on the HF diet. No
significant difference in energy balance was observed between
the HC and HF diets (P 0.50). However, total energy balance
was significantly related to PAI during the HC calorimeter stay
(r 0.85, P 0.0001), as was CHO balance (r 0.54, P
0.0007). During the HF calorimeter stay, CHO balance (r 0.38,
P 0.02) and total energy balance (r 0.57, P 0.003) were
also correlated with PAI.
By design, CHO balance was higher on the HC diet (P
0.005), and fat balance was higher on the HF diet (P 0.005).
The differences in macronutrient expenditures were consistent
with observed differences in the 24-h RQ. For all subjects, the
mean (SEM) 24-h RQ from the HC diet was 0.815 0.010,
whereas that from the HF diet was 0.775 0.009 (P 0.01). No
significant differences in protein balance or expenditure were
noted between the diet groups. Additionally, no significant difference in SI was observed between the HC and HF diets (P
0.17) (Table 2).
During the 4 y of follow-up, the mean (SEM) increase in
body weight was 0.29 0.15 kg/y (P 0.0628). Fat mass
increased by 0.31 0.15 kg/y (P 0.0508). Candidate predictor
variables were examined individually as predictors of changes
over time in the 3 outcomes, with adjustment for sex and baseline
BMI. No significant relation was observed between CHO or fat
oxidation and any of the 3 outcomes on either diet (P 0.10).
Borderline inverse relations were observed between total energy

Downloaded from ajcn.nutrition.org by guest on March 28, 2016

Physical activity level was estimated by calculating the physical activity index (PAI) (25). This index is the ratio of total EE
to basal EE. In the present study design, actual EE was not
measured during each 14-d feeding phase before the calorimeter
stay. However, each feeding period was highly controlled and
monitored; the subjects were weight-stable within 1 kg and were
therefore in energy balance. PAI was calculated as follows:

805

806

ECKEL ET AL

TABLE 2
Results from the indirect calorimeter stay on both the high-carbohydrate
(HC) and high-fat (HF) diets1
Outcome
Carbohydrate balance (kcal)
Fat balance (kcal)
24-h RQ
Energy expenditure (kcal)
SI (mg m2 min1)

HC diet

HF diet

P2

642 85
4 112
0.815 0.010
2516 98
305 24

350 80
415 127
0.775 0.009
2391 117
278 23

0.0033
0.0017
0.0076
0.1618
0.1712

1
All values are x SEM. n 36. RQ, respiratory quotient; SI, insulin
sensitivity.
2
A paired t test was used for the difference between diets.

FIGURE 1. Relation between physical activity index (year 1) and the

change in fat mass over 4 y (r 0.2, P 0.2; 2-stage mixed model,
described in Methods).

additional 100 mg m2 min1 in SI, the subjects gained, on

average, 0.30 0.10 kg/y less weight (P 0.008) and 0.22
0.11 kg/y less fat (P 0.051). R2 values indicate that CHO
balance and SI explained 9% and 19% of the variation in weight
change, and 21% and 19% of the variation in fat mass change,
respectively.
The significant inverse relation between CHO balance after
the HC diet had been consumed for 15-d in year 1 and the change
in fat mass during 4 y is shown in Figure 2 (n 37, r 0.46,
P 0.007). The pattern of the relation (slope of each regression
line) between CHO balance and change in fat mass was similar
in all initial BMI categories (P 0.79 for interaction between
CHO balance and BMI category).
The effects of CHO balance and SI after the HC diet when the
other variable was added to the model (still adjusting for sex and
baseline BMI) are also shown in Table 3. CHO balance remained
inversely related to changes in fat mass and percentage fat (P
0.037 and P 0.029, respectively) and SI was inversely related
to weight change (P 0.027), but the significance of each and the
effects were attenuated because of the strong relation between the
2 variables (r 0.55, P 0.001). Thus, these effects do not
completely substitute for each other, but they are difficult to
completely separate.

TABLE 3
Effects of carbohydrate (CHO) balance and insulin sensitivity (SI) with a high-CHO diet at baseline measurement on changes in fat and weight over 4 y

Outcome
Effect of CHO balance
Weight change (kg y1 100 kcal1)
Fat mass change (kg y1 100 kcal1)
Percentage fat change (% y1 100 kcal1)
Effect of SI
Weight change [(kg/y)/100 mg m2 min1)]
Fat mass change [(kg/y)/100 mg m2 min1)]
Percentage fat change [(%/y)/100 mg m2 min1)]
1
2

P1

Adjusted for
sex, baseline
BMI, and SI

0.081 0.0372
0.097 0.033
0.097 0.036

0.038
0.007
0.013

0.30 0.10
0.22 0.11
0.15 0.12

0.008
0.051
0.23

Adjusted for sex and

baseline BMI

P1

Adjusted for sex,

baseline BMI, and
CHO balance

P1

0.038 0.038
0.082 0.037
0.095 0.041

0.32
0.037
0.029

0.26 0.11
0.13 0.11
0.051 0.12

0.027
0.25
0.68

A 2-stage mixed model was used to determine predictors of future changes in body composition.
x SEM (all such values).

Downloaded from ajcn.nutrition.org by guest on March 28, 2016

balance and weight gain for both diets (P 0.053 for HC diet;
P 0.058 for HF diet). Fat balance on the HF diet was also
inversely related to weight gain (P 0.038) but was otherwise
not a significant predictor for any of the outcomes on either
diet (P 0.25). CHO balance and SI on the HF diet were not
predictive of future gains in any of the 3 outcomes (P 0.30
for all). Adjustment for usual physical activity as quantified
by PAI or NEAT did not substantially change any of these
results; additionally, no relation was observed between either
measure of physical activity and weight or fat gain over 4 y for
either diet (P 0.1 for all). Specifically, no relation was
observed between the PAI at the beginning of year 1 and the
change in fat mass (r 0.2, P 0.2) (Figure 1), weight
(r 0.2, P 0.3), or percentage body fat (r 0.12,
P 0.4) over 4 y.
The strongest and most consistent predictors of future weight
and fat change were CHO balance and SI after the HC diet. On the
HC diet, the subjects who had a higher positive CHO balance
(day 15) gained less fat mass (P 0.001), percentage body fat
(P 0.006), and weight (P 0.024) over time. When adjusted
for SI, CHO balance remained a significant predictor of changes
in fat mass (P 0.021) and percentage body fat (P 0.025). The
effects of CHO balance and SI on each of the 3 outcomes, adjusted for sex and baseline BMI, are shown in Table 3. Both
predictors were consistently and inversely related to the outcomes (with the exception of SI for fat mass change, P 0.051),
and all were significant except SI for percentage body fat. For
example, for each additional 100 kcal/d in CHO balance, the
subjects gained a mean (SEM) 0.081 0.037 kg/y less weight
(P 0.038) and 0.097 0.033 kg/y less fat (P 0.007). For each

CARBOHYDRATE BALANCE PREDICTS WEIGHT AND FAT GAIN

FIGURE 2. Carbohydrate (CHO) balance after 15 d of a high-CHO diet

in year 1 at baseline as a predictor of change in fat mass over 4 y (r 0.46,
P 0.007; 2-stage mixed model, described in Methods).

DISCUSSION

substantial differences in the study designs, that is, sample size

[n 6, all men; x (SD) BMI: 23.0 2.0], short-term dietary
manipulation (total of 5 d), and no longitudinal follow-up. Of
interest, evidence shows that short-term glycogen storage could be
increased more after exercise-induced glycogen depletion than by
high CHO feeding (40). If a long-term relation does exist between
glycogen stores and EI, this may partly explain why continuance of
physical activity promotes weight maintenance. It is also possible
that some of the subjects changed their lifestyle during the 4 y of
follow-up; that is, they decreased or increased their habitual physical
activity. However, the relations between CHO balance, glycogen
stores, and the regulation in long-term EI remain to be tested.
The strong relations between energy and CHO balance in the
chamber and PAI in the setting of 1 d of physical inactivity on the HC
diet suggests that the protection from increases in gain of body
weight and fat mass over 4 y could be explained by the level of
physical activity, not the response to HC feeding. However, the lack
of any relation between the PAI and NEAT and the gain in weight
and fat mass indicates a separate mechanism is operational.
In the present study, the RQ remained 1.0, and mean
(SEM) fat balance on the HC diet was 4 112 kcal/24 h.
Thus, de novo lipogenesis from glucose was not the predominant
fate of excess dietary CHO during the period of physical inactivity. However, because it often takes days for fat balance to
adjust to changes in diet composition (33, 41), and the effect of
sudden changes in physical activity may also be important in the
long-term regulation of energy balance, the ability of changes in
fat balance over 1 d to predict changes in weight or body composition over 4 y would less likely be observed. Although fat
oxidation during HC feeding was not a significant predictor of
gain in weight and fat mass, it could still be an important part of
the explanation for our results. Persons who continue to rely to a
greater extent on fat oxidation when eating a HC diet would be
more likely to show more positive CHO balance when they
become physically inactive. Moreover, because the same could
be said about athletes or even persons who are highly physically
active, it is intriguing that, even when PAI and NEAT were
accounted for in the model, CHO balance remained a strong
predictor of changes in body composition over time.
Although the present study was highly controlled, a limitation
of these data is that total EE was not measured in the subjects
while they were outside of the indirect calorimeter. Yet, because
their food intakes were adjusted to sustain their weights, the
method of calculation proved highly accurate. Moreover, during
the 4-y follow-up, the subjects were in their free-living environment in which neither their diet nor their physical activity levels
were controlled. This is perhaps a strength, rather than a limitation, of the study, in that the baseline data predicted the response
in a free-living environment.
These results do not necessarily implicate a HC diet in the
development of obesity. The role of diet composition in the
development of obesity is controversial (42, 43), and our results
do not directly address the question. They do, however, suggest
that differences in handling of excess CHO during inactive periods could be important for gain of weight and fat mass. The
results are significant because they provide both a model for
studying differences in substrate utilization and a predictor of
future weight and fat gain. Many factors could have contributed
to the differences in weight gain, making it even more remarkable
that a single metabolic predictor could be identified. Overall, the
long-term results of the present highly controlled, short-term diet

Downloaded from ajcn.nutrition.org by guest on March 28, 2016

We identified a strong prospective metabolic predictor of

long-term gains in body weight and body fatness. This marker
may be useful in identifying persons who are susceptible to
weight gain, so that interventions for weight gain prevention can
be more efficiently tested. As a model, it may be useful to study
subjects on a day of relative inactivity, when their EI remains
constant for the normal level of activity. Thus, they would be
studied on a day of positive energy and CHO balance. The duration of HC feeding before, as well as the period of inactivity
during, testing would require additional study.
These results also provide useful information about the mechanisms by which differences in substrate metabolism affect susceptibility to a gain in weight and fat mass. Evidence has implicated differences in fat oxidation and fat balance in weight gain
(9, 33), and we were surprised to find that CHO balance was so
strongly predictive of long-term gain in weight and fat mass
compared with fat balance. Although fat balance on the HF diet
was predictive of change in weight only, and energy balance was
predictive of weight gain on both diets, we were less enthusiastic
about these seemingly isolated and only mildly significant values
in comparison to the strength of the relations between CHO
balance on the HC diet and changes in body weight and composition. Clearly, those persons who had the greatest degree of
positive CHO balance gained the least weight and fat mass over
time, even when SI was accounted for in the model.
Possible explanations exist as to why CHO balance on the HC
diet most predicted gain in weight and fat mass. First, those
persons who showed more positive CHO balance on the HC diet
could be ones with an enhanced ability to store glycogen or have
a lesser tendency to deplete glycogen stores. Flatt (34) suggested
the critical role of CHO balance in the regulation of body weight.
Specifically, glycogen stores may be related to EI such that
persons with a greater tendency to deplete glycogen stores may
experience more hunger, thereby causing ingestion of more total
energy. Persons with a lower tendency to deplete glycogen
stores, then, do not consume as much energy. However, the
studies by Flatt (35, 36) were mostly conducted in rodents, and
short-term studies conducted in humans, in which glycogen
stores were manipulated by dietary means, have produced inconsistent effects on EI (3739). In particular, the findings of Shetty
et al (39) may not be relevant to our observations because of

807

808

ECKEL ET AL

and energy balance experiment require confirmation, and mechanisms need to be pursued.

REFERENCES
1. Mokdad AH, Ford ES, Bowman BA, et al. Prevalence of obesity, diabetes, and obesity-related health risk factors, 2001. JAMA 2003;289:
76 9.
2. Flegal KM, Carroll MD, Ogden CL, Johnson CL. Prevalence and trends
in obesity among US adults, 1999 2000. JAMA 2002;288:17237.
3. Hill JO, Peters JC. Environmental contributions to the obesity epidemic.
Science 1998;280:1371 4.
4. Chagnon YC, Rankinen T, Snyder EE, Weisnagel SJ, Perusse L, Bouchard C. The human obesity gene map: the 2002 update. Obes Res
2003;11:313 67.
5. Flegal KM, Carroll MD, Kuczmarski RJ, Johnson CL. Overweight and
obesity in the United States: prevalence and trends, 1960 1994. Int J
Obes Relat Metab Disord 1998;22:39 47.
6. Ravussin E, Lillioja S, Knowler WC, et al. Reduced rate of energy
expenditure as a risk factor for body-weight gain. N Engl J Med 1988;
318:46772.
7. Roberts SB, Savage J, Coward WA, Chew B, Lucas A. Energy expenditure and intake in infants born to lean and overweight mothers. N Engl
J Med 1988;318:461 6.
8. Levine JA, Eberhardt NL, Jensen MD. Role of nonexercise activity thermogenesis in resistance to fat gain in humans. Science 1999;283:212 4.
9. Zurlo F, Lillioja S, Esposito-Del Puente A, et al. Low ratio of fat to
carbohydrate oxidation as predictor of weight gain: study of 24-h RQ.
Am J Physiol 1990;259:E650 7.
10. Seidell JC, Muller DC, Sorkin JD, Andres R. Fasting respiratory exchange ratio and resting metabolic rate as predictors of weight gain: the
Baltimore Longitudinal Study on Aging. Int J Obes Relat Metab Disord
1992;16:66774.
11. Swinburn BA, Nyomba BL, Saad MF, et al. Insulin resistance associated
with lower rates of weight gain in Pima Indians. J Clin Invest 1991;88:
168 73.
12. Travers SH, Jeffers BW, Eckel RH. Insulin resistance during puberty and
future fat accumulation. J Clin Endocrinol Metab 2002;87:3814 8.
13. Yost TJ, Jensen DR, Eckel RH. Weight regain following sustained
weight reduction is predicted by relative insulin sensitivity. Obes Res
1995;3:5837.
14. Filozof CM, Murua C, Sanchez MP, et al. Low plasma leptin concentration and low rates of fat oxidation in weight-stable post-obese subjects. Obes Res 2000;8:20510.
15. Spraul M, Ravussin E, Fontvieille AM, Rising R, Larson DE, Anderson
EA. Reduced sympathetic nervous activity. A potential mechanism predisposing to body weight gain. J Clin Invest 1993;92:1730 5.
16. Peterson HR, Rothschild M, Weinberg CR, Fell RD, McLeish KR,
Pfeifer MA. Body fat and the activity of the autonomic nervous system.
N Engl J Med 1988;318:1077 83.

Downloaded from ajcn.nutrition.org by guest on March 28, 2016

We thank the nurses, CORE laboratory technicians, and kitchen personnel

of the General Clinical Research Center (GCRC) for their professional assistance and the personnel of the Energy Balance Laboratory of the Colorado
Clinical Nutrition Research Unit and of the GCRC, who skillfully managed
the whole-room indirect calorimeter.
RHE was the principal investigator and was responsible for the original design
of study, data collection, analysis and interpretation of the data, and manuscript
construction. TLH was responsible for monitoring of study conduct, data compilation, analysis and interpretation of the data, and manuscript construction.
MLB was responsible for the data analysis design, analysis and interpretation of
the data, and manuscript construction. KMW was responsible for monitoring of
study conduct, data collection, analysis and interpretation of the data, and manuscript construction. TYS was responsible for the original design of the study,
monitoring of study conduct, data collection, and manuscript construction. GKG
was responsible for the data analysis design, analysis and interpretation of the
data, and manuscript construction. TAS was responsible for the original design
of the study, data collection and compilation, analysis and interpretation of the
data, and manuscript construction. CCF was responsible for data collection and
compilation, data interpretation, and manuscript construction. JOH was responsible for the original study design, data interpretation, and manuscript construction. None of the authors had a conflict of interest.

17. Swinburn B, Ravussin E. Energy balance or fat balance? Am J Clin Nutr

1993;57(suppl):766S70S.
18. Connor SL, Gustafson JR, Sexton G, Becker N, Artaud-Wild S, Connor
WE. The Diet Habit Survey: a new method of dietary assessment that
relates to plasma cholesterol changes. J Am Diet Assoc 1992;92:417.
19. Insel PA, Liljenquist JE, Tobin JD. Insulin control of glucose metabolism in man. J Clin Invest 1975;55:1057 66.
20. Shepard TY, Weil KM, Sharp TA, et al. Occasional physical inactivity
combined with a high-fat diet may be important in the development and
maintenance of obesity in human subjects. Am J Clin Nutr 2001;73:703 8.
21. Hill JO, Peters JC, Reed GW, Schlundt DG, Sharp T, Greene HL. Nutrient balance in humans: effects of diet composition. Am J Clin Nutr
1991;54:10 7.
22. Jequier E, Acheson K, Schutz Y. Assessment of energy expenditure and
fuel utilization in man. Annu Rev Nutr 1987;7:187208.
23. Weir J. New methods for calculating metabolic rate with special reference to protein metabolism. Nutrition 1949;6:21321.
24. Laakso M. How good a marker is insulin level for insulin resistance?
Am J Epidemiol 1993;137:959 65.
25. Institute of Medicine, Food and Nutrition Board. Dietary Reference
Intakes for energy, carbohydrates, fiber, fat, fatty acids, cholesterol,
protein and amino acids (macronutrients). Washington, DC: National
Academy Press, 2002.
26. DAlessio DA, Kavle EC, Mozzoli MA, et al. Thermic effect of food in
lean and obese men. J Clin Invest 1988;81:17819.
27. Kuczmarski RJ. Prevalence of overweight and weight gain in the United
States. Am J Clin Nutr 1992;55(suppl):495S502S.
28. Goldman R, Buskrik E. Body volume measurement by underwater weighing: description and methods. Techniques for measuring body composition.
Washington, DC: National Academy of Science, 1962:78 89.
29. Brozek J, Grande J, Keys A. Densitometric analysis of body composition: revision of some qualitative assumptions. Ann N Y Acad Sci 1963;
110:113 40.
30. Desbuquois B, Aurbach GD. Use of polyethylene glycol to separate free
and antibody-bound peptide hormones in radioimmunoassays. J Clin
Endocrinol Metab 1971;33:732 8.
31. Skogerboe KJ, Labbe RF, Rettmer RL, Sundquist JP, Gargett AM.
Chemiluminescent measurement of total urinary nitrogen for accurate
calculation of nitrogen balance. Clin Chem 1990;36:7525.
32. Laird NM, Ware JH. Random effects models for longitudinal data. Biometrics 1982;38:96374.
33. Thomas CD, Peters JC, Reed GW, Abumrad NN, Sun M, Hill JO.
Nutrient balance and energy expenditure during ad libitum feeding of
high-fat and high-carbohydrate diets in humans. Am J Clin Nutr 1992;
55:934 42.
34. Flatt JP. Carbohydrate balance and body-weight regulation. Proc Nutr
Soc 1996;55:449 65.
35. Flatt JP, Ravussin E, Acheson KJ, Jequier E. Effects of dietary fat on
postprandial substrate oxidation and on carbohydrate and fat balances.
J Clin Invest 1985;76:1019 24.
36. Flatt JP. The difference in the storage capacities for carbohydrate and for
fat, and its implications in the regulation of body weight. Ann N Y Acad
Sci 1987;499:104 23.
37. Snitker S, Larson DE, Tataranni PA, Ravussin E. Ad libitum food intake
in humans after manipulation of glycogen stores. Am J Clin Nutr 1997;
65:941 6.
38. Sparti A, Milon H, Di Vetta V, et al. Effect of diets high or low in
unavailable and slowly digestible carbohydrates on the pattern of 24-h
substrate oxidation and feelings of hunger in humans. Am J Clin Nutr
2000;72:1461 8.
39. Shetty PS, Prentice AM, Goldberg GR, et al. Alterations in fuel selection
and voluntary food intake in response to isoenergetic manipulation of
glycogen stores in humans. Am J Clin Nutr 1994;60:534 43.
40. Goforth HW Jr, Laurent D, Prusaczyk WK, Schneider KE, Petersen KF,
Shulman GI. Effects of depletion exercise and light training on muscle
glycogen supercompensation in men. Am J Physiol Endocrinol Metab
2003;285:E1304 11.
41. Costill DL, Fink WJ, Hargreaves M, King DS, Thomas R, Fielding R.
Metabolic characteristics of skeletal muscle during detraining from competitive swimming. Med Sci Sports Exerc 1985;17:339 43.
42. Bray GA, Popkin BM. Dietary fat intake does affect obesity! Am J Clin
Nutr 1998;68:115773.
43. Willett WC. Dietary fat plays a major role in obesity: no. Obes Rev
2002;3:59 68.