CarbohydrateFat Interactions and Obesity

Examined by a Two-Compartment Computer

Model
Jean-Pierre Flatt

Abstract
FLATT, JEAN-PIERRE. Carbohydratefat interactions and
obesity examined by a two-compartment model. Obes Res.
2004;12:20132022.
Objective: A systems dynamics computer model was developed to examine how the interactions between carbohydrate
and fat metabolism influence body weight regulation. It
reflects the operation of a two reservoir-system: one representing the bodys limited glycogen, and the other, its large
fat reserves. The outflows from the reservoirs correspond to
the oxidation of glucose and fat, whose relative contributions are affected by the size of the prevailing glycogen and
fat reserves. Together, they meet the bodys energy expenditure. Replenishments occur three times per day, in portions restoring total glycogen content to specific levels. A
parameter mimicking the action of insulin is necessary to
create realistic responses.
Research Methods and Procedures: The model was run for
125-day periods to establish the degree of adiposity for
which rates of fat oxidation become commensurate with fat
intake and the influence thereon of various dietary, environmental, lifestyle, and inherited variables.
Results: Equivalent degrees of adiposity can be sustained
under a variety of conditions. For instance, the impact on
steady-state body fat contents of a 10% increase or decrease
in the energy provided by dietary fat is offset by a 26-gram
decrease or increase in mean glycogen levels.
Discussion: Environmental factors such as food diversity,

Received for review January 5, 2004.

Accepted in final form October 12, 2004.
The costs of publication of this article were defrayed, in part, by the payment of page
charges. This article must, therefore, be hereby marked advertisement in accordance with
18 U.S.C. Section 1734 solely to indicate this fact.
Department of Biochemistry and Molecular Pharmacology, University of Massachusetts
Medical Center, Worcester, Massachusetts.
Address correspondence to J. P. Flatt, Department of Biochemistry and Molecular Pharmacology, University of Massachusetts Medical Center, 364 Plantation Street, Room 813,
Worcester, MA 01605.
E-mail: j.p.flatt@umassmed.edu
Copyright 2004 NAASO

palatability, and availability can be expected to raise the

range within which glycogen levels are habitually maintained. This restrains fat oxidation, until expansion of the fat
mass is sufficient to promote fat oxidation to a rate commensurate with dietary fat intake. This metabolic leverage
can explain why increased food offerings tend to raise the
prevalence of obesity.
Key words: system dynamics, body weight regulation,
glycogen levels, dietary fat, exercise

Introduction
Most adults maintain relatively stable body weights and
body compositions during long periods of their lives. This
implies not only that their average energy intake matches
their average energy expenditure (EE),1 but also that the
composition of the fuel mix that their body oxidizes is
equivalent, on average, to the nutrient mixture that they
consume. As long as this is not the case, body composition
keeps changing, until the particular configuration is reached
for which both conditions are satisfied (1). A steady state of
approximate weight maintenance then becomes established.
If disturbed by periods of dietary restriction or overconsumption (2), the particular body composition for which a
given individual had previously reached weight stability
tends to reestablish itself, as long as the environmental and
lifestyle parameters have not changed. The problem is that,
for many individuals, this steady state is reached only after
an excessive amount of body fat has been accumulated (1).
Protein contributes a minor and fairly constant proportion
of dietary energy, and biological evolution has led to regulatory features that spontaneously maintain protein balance.
It is, therefore, the interactions between carbohydrate

1
Nonstandard abbreviations: EE, energy expenditure; CHO, carbohydrate; RQ, respiratory
quotient; GLY, glycogen; TG, triglyceride; BMR, basal metabolic rate; %BoFat, percent
body fat; CoxB, CHO oxidation by the brain; CoxA, CHO oxidation by the large turbine;
FCR, fat-CHO ratio; JNS, Jnsulin; PAL, physical activity levels.

OBESITY RESEARCH Vol. 12 No. 12 December 2004

2013

Carbohydrate/Fat Metabolism and Obesity, Flatt

(CHO) and fat intakes and oxidation that primarily determine body weight maintenance and the level of adiposity
for which this is achieved. Whereas glucose and fatty acids
can both provide fuel to sustain ATP regeneration, the
conditions controlling the storage and use of glucose and
fatty acids differ greatly because of the vast difference in the
size of the bodys CHO and fat reserves. Glycogen (GLY)
reserves are limited to a few hundred grams, not much more
than the quantity of CHOs usually consumed in 1 day (3). In
contrast, the bodys capacity for storing fat is considerable.
Fat reserves typically amount to 100,000 150,000 kcal in
normal weight adults, which is 100 times more energy
than present as glycogen. In overweight and obese individuals, these reserves can reach much higher levels. Because
the fat stores are so large, daily gains or losses of fat are not
only easily tolerated, but also essentially unnoticeable. In
effect, it is unlikely that short-term gains or losses of fat can
be physiologically detected with sufficient precision to elicit
appropriate adjustments in food intake (1).
The bodys metabolic fuel economy is thus handled
through the operation of a system comprising two reservoirs
of very different dimensions. We have previously proposed
an analogy with a hydraulic model, featuring a reservoir
with a small diameter and one with a 10 times wider
diameter, to represent the bodys CHO and fat reserves,
respectively (Figure 1) (4). The small reservoir is the sole
source of fuel for a small turbine representing the brain,
which uses exclusively glucose under usual conditions. The
flow through the large turbine reflects the EE of the rest of
the body. It is supplied by both reservoirs. The contributions
made by the small and the large reservoirs are considered to
be determined by the relative heights of the levels prevailing
in the reservoirs at a given time. The flow through the two
turbines thus depicts the use of glucose and fat by the body
to regenerate the ATP used, at an overall rate determined by
the bodys EE. The reservoirs are replenished from time to
time, in proportions reflecting the CHO-to-fat ratio of the
diet. These additions are the equivalent of meals. They
cause a substantial change in the level of the small reservoir,
while leaving the level in the large reservoir essentially
unchanged. The contribution made by the small reservoir to
the large turbine flow is thereby temporarily increased. As
time elapses after a meal and the level in the small reservoir
declines, this contribution decreases, while that made by the
large reservoir increases. The conduit from the small to the
large reservoir describes the conversion of glucose into fat
by de novo lipogenesis. Under usual living condition, glycogen reserves are spontaneously kept far below the levels
for which this process becomes quantitatively important (5,6).
The rates of CHO and fat oxidation in humans can be
assessed by continuous measurements of CO2 production
and oxygen consumption (7). The ratio of CO2 produced to
O2 consumed is known as the respiratory quotient (RQ). It
2014

OBESITY RESEARCH Vol. 12 No. 12 December 2004

Figure 1: Two-compartment model showing the impact of circumstantial, lifestyle, and genetic factors on glycogen levels and fat
stores in the body, described in terms of kilocalories. The flow
through the small turbine reflects the exclusive use of glucose by
the brain; the flow through the large turbine represents the EE in
the rest of the body, to which the two reservoirs contribute in
proportion to the levels of fuel that they contain at a given time.
The conduit from the small to the large reservoir depicts lipogenesis, a process that assumes quantitative significance only when
glycogen levels reach levels far higher than those maintained
under usual ad libitum food intake conditions. The small containers at the top of the diagrams represent the portions of CHO and fat
provided intermittently by the consumption of meals. The top and
bottom panels show inherited differences affecting the competition
between fat and glucose for oxidation, such that different amounts
of fat need to be present to achieve comparable levels in the two
reservoirs and, hence, comparable proportions of glucose and fat
use (adapted from ref. 1).

varies between 0.7, when fat only is oxidized, and 1.0, when
glucose is the sole substrate oxidized. After consumption of
a CHO-containing meal, the RQ rises promptly and later
declines as the GLY reserves are drawn down. The postprandial changes in the proportions of glucose and fat

Carbohydrate/Fat Metabolism and Obesity, Flatt

Figure 2: Main components of a two-compartment systems dynamics computer model constructed using Stella v.8 software. The
rectangular boxes are containers, the double arrows represent
fluxes, and the single arrows show the influences of the contents of
container contents and of converters (circles) on fluxes. Explanations are provided in the text.

oxidized in vivo are much greater than the alteration in the

outflows from the small and the large reservoir that one
could expect to occur in the simple model considered above.
This is because, in the body, CHO intake stimulates the
secretion of insulin, which inhibits fatty acid mobilization,
reduces fatty acid oxidation, and promotes glucose oxidation (8).
To examine the interactions between CHO and fats and
their impact on body weight maintenance and body fat
content, we developed a two-compartment computer model,
which includes a parameter mimicking the action of insulin
in promoting CHO oxidation when GLY levels are high. In
addition, various parameters were built in to reflect variations in environmental and lifestyle conditions. The model
was used to examine how differences in dietary fat content,
physical activity, food availability, diversity, and palatability could be expected to affect the degree of adiposity for
which steady states tend to become established and how this
can be modified by inherited traits affecting the use of
glucose relative to fat.

Research Methods and Procedures

Description of the Model: Structure and Assumptions
The systems dynamics modeling approach was used to
create a model corresponding to the situation described in
Figure 1. In this type of model (Figure 2), fluxes (shown by
double arrows) into and out of containers are determined by
the contents of the containers and by appropriate auxiliary

functions (shown by single arrows). Fluxes drawing from

sources outside the model into the gut containers, or representing losses to the outside (i.e., CO2 release), originate, or
feed, into clouds (shaped like propellers in the diagram).
The fluxes that take place during a specified time interval
are calculated, and the contents of the containers are recomputed at the end of this interval. These computation loops
are repeated for the number of cycles specified to complete
a model-run. The changes can be displayed as a function
of time in graphs and tables, which is well suited to examine
biological phenomena. The Stella version 8 software was
used; information can be obtained at http://www.iseesystems.
com/. Substrate contents and fluxes are expressed in terms
of kilocalories and kilocalories per hour, respectively; a
computing time interval of one-tenth of an hour was used.
One container represents the bodys total CHO reserve,
i.e., the sum of its GLY contents (primarily muscle and liver
GLY). Another container represents the bodys total fat, i.e.,
triglyceride reserves (TG). No provision is made for the
transfer of calories from one of these containers to the other,
because fatty acids cannot be transformed into CHO and
because conversion of glucose into fat through the process
of de novo lipogenesis proceeds at negligible rates under
usual living conditions (5,6).
The outflows from these two containers are calculated to
be equal to the individuals EE at a given time. The contributions made by GLY and TG to this prevailing rate of EE
are determined as follows. The energy need of the brain (the
small turbine in Figure 1) is equal to 20% of the basal
metabolic rate (BMR). Thus, CHO oxidation by the brain
(CoxB in Figure 2) is 0.2 BMR, and it is derived exclusively by outflow from the GLY compartment. The bodys
fat stores typically contain 100 times more energy than its
GLY reserves, which corresponds to the diameter of the
large reservoir being 10 times that of the small reservoir, as
shown in Figure 1. Despite the two orders of magnitude
difference in these reserves, glucose and fat contributions to
substrate oxidation adjust themselves to CHO and fat intakes. In Figure 1, the large reservoir has to contain 100
times more fuel than the small reservoir to generate equivalent pressures at the reservoirs outlets and, therefore, to
elicit comparable contributions to the flow through the large
turbine. The contributions made by GLY and TG in meeting
EE other than by the brain are, therefore, computed in the
model to occur in the same proportions as the bodys GLY
content and its TG content divided by 100. Thus, CHO
oxidation by the large turbine (CoxA) is computed as (EE
CoxB) GLY/(GLY TG/100). Therefore, fat oxidation
through the large turbine is equal to (EE CoxB CoxA).
Inherited variability in enzymatic and/or hormonal regulation may alter the influence of the bodys fat stores in
promoting fat oxidation relative to glucose oxidation. In
Figure 1, this is shown by changing the ratio of the diameters and surfaces of the two reservoirs, because this will
OBESITY RESEARCH Vol. 12 No. 12 December 2004

2015

Carbohydrate/Fat Metabolism and Obesity, Flatt

affect how the fat stores compete with the GLY reserves in
supplying the substrate needed by the large turbine. A
parameter allowing alteration of the use of fat relative to
CHO, or a fat-CHO ratio modifier (FCR modifier), was
therefore introduced, such that the denominator in the equation determining CHO and fat flow through the large turbine
becomes [GLY TG/(100 FCR)].
A 24-hour period is considered to start at 6:00 AM. Meals
occur according to a 24-hour cycle, in portions designed to
restore specified GLY contents. Meals are simulated by
influxes into the gut and from there into the GLY and TG
containers. Because the palatability and diversity of the food
available (9), as well as social facilitation (10), increase
meal size, a palatability parameter (Palat) was built in,
which can be applied to increase or decrease the target
levels set for GLY replenishment. Finally, the rate of absorption from the gut can be modified to mimic the effect of
differences in the glycemic index of the CHOs consumed.
As mentioned, the in vivo changes in the proportion of
CHO and fat oxidized after meals are greatly enhanced by
the postprandial increase in insulin secretion and this hormones major effects on metabolism. To mimic the influence of insulin, a function called Jnsulin (JNS) is used. The
effect of JNS is implemented by using (GLY JNS)
instead of GLY in computing the contributions made by the
outflows from the GLY and TG compartments to the large
turbine. Thus, CoxA (EE CoxB) (GLY JNS)/
[GLY TG/(100 FRC)]. JNS is equal to [GLY/(1312)2],
where 1312 is the value found to lead to the same steadystate body composition for the default parameter setting as
in the absence of JNS (i.e., when JNS is set permanently to
1.0). As one would expect, this brings about much greater
variations in the composition of the fuel mix oxidized
during the day, and the time-course of some key daily
phenomena reflects much more closely what happens in
vivo when JNS is operating (Figure 3).
To keep track of daily events, a number of additional
components are needed in the model, which are not shown
in Figure 2 [e.g., to compute cumulative daily (denoted by
the prefix cd) CHO and fat intakes (cdCint and cdFint),
oxidation (cdCox and cdFox), and EE (cdEE)]. Cumulative
values reached during the previous 24-hour period are denoted by a p prefix, as in pminGLYg, pmeanGLYg, and
pmaxGLYg, which describe the minimum, mean, and maximal GLY values during the preceding 24-hour period,
respectively. The suffix g indicates that a particular value
is expressed in terms of grams (although kilocalories are the
units used in the computations). The model calculates the
RQ and the food quotient, which describes the ratio of CO2
produced to O2 consumed during the biological oxidation of
a representative sample of the diet (1). Several other values
of interest are derived, including the CHO and fat content of
the diet (c%Diet and f%Diet), the contribution of glucose to
prevailing and daily EE (C%ox and cdC%ox), body fat
2016

OBESITY RESEARCH Vol. 12 No. 12 December 2004

Figure 3: Daily patterns (from 6:00 AM to 6:00 AM) of key

descriptors of energy metabolism in the two-compartment model.
Top: (1) EE (scale, 0 to 4 kcal/min), (2) cumulative daily energy
intake (cdEint; scale, 0 to 4000 kcal/d), and (3) cumulative daily
EE (cdEE; scale, 0 to 4000 kcal/d). Middle: (1) GLY-reserve
without JNS (GLY; scale, 0 to 4000 kcal), (2) GLY reserve with
JNS (jGLY; scale, 0 to 4000 kcal), and (3) JNS (jGLY/1312)2
(scale, 0 to 2; no dimension). Bottom: RQ CO2 production/O2
consumption (1) without JNS (scale, 0.7 to 1.0) or (2) with JNS
(jRQ).

content in kilograms (TG/9000), and percentage body fat

(%BoFat) of body weight (BWt). The values that prevail for
these various components can be listed in tables or displayed in graphs. Variations in the bodys fat content cause
body weight to change, to 1 kg/6000 kcal gained or lost,
reflecting the fact that approximately two-thirds of body
weight gains or losses incurred by adults are fat, and onethird is lean body mass, which is primarily water. The
corresponding changes in BMR are automatically computed.
Default Conditions
The initial, or default, conditions are set to correspond to
an individual with a body weight of 65 kg, whose basal EE
(BMR) is 1 kcal/min and whose fat reserve amounts to 12
kg, or 12,000 grams 9 kcal/g 108,000 kcal. Fat is
stored in adipose tissue, which is 90% fat. It was assumed
that, under resting conditions, adipose tissue accounts for a

10
0
8
0
290
322
65
46
Experimental vs. model. PRO, protein.

Experiment 1
Model Run 1
Experiment 2
Model Run 2

1.33
1.33
1.32
1.32

0.82
0.83
0.82
0.83

1.47
1.48
1.5
1.49

0.89
0.85
0.91
0.85

300
356
301
356

53
126
427
502

129
0
130
0

22
7
22
8

PRO
balance
(kcal)
FAT
balance
(kcal)
CHO
balance
(kcal)
PRO
intake
(kcal)
FAT
intake
(kcal)
CHO
intake
(kcal)
Postmeal
Max RQ
Postmeal EE
(kcal/min)
Premeal
RQ
Premeal EE
(kcal/min)

Table 1. Comparison of substrate balances after test meals

Testing the Models Responses

The model was evaluated by examining its response to
influxes into the gut and from there into the GLY and TG
containers of portions of CHO and fat with the same energy
contents as two test meals given to human volunteers to
study the difference in the metabolic responses to CHO and
fat consumption (11). One of the test meals provided 75
grams of CHOs, 6 grams of fat, and 32 grams of protein; in
the other meal, the fat content was increased to 47 grams.
To accommodate the fact that the test meals contained
protein, a nutrient not replicated in the model, the CHO-tofat ratio in the model meals was set to be oxidized with the
same RQ as those of the experimental meals, i.e., 0.923 and
0.828. The models initial conditions were modified from
the default condition to raise the BMR to the rate measured
in the human subjects, and the thermic effect associated
with CHO absorption was raised from 6% to 17% to imitate
the predictable thermic effect of the protein-containing
meal. The data obtained by continuous indirect calorimetry
in the human study during the 9 hours after the meal are
shown in Table 1, along with the GLY, TG, and energy
balances computed by the model over a 9-hour postintake
period. The similarity of the energy balances is the result of
the adjustments just discussed, and thus has no particular
significance. In the experiment and in the model, CHO or
GLY balances were similar after both test meals, reflecting

Energy
balance
(kcal)

rate of EE of 10 kcal/d per kilogram. The BMR is, therefore,

computed as 54.4 10/24 body fat/(9000 90%) 60
kcal/h. The rate of EE is determined by the BMR, multiplied
by an activity pattern (ActivPattern) mimicking changes
induced by daily activities plus the thermic effect of food
(TEF), set at 0.06 CHO absorption and 0.03 fat
absorption. In the default conditions, the activity pattern is
set to elicit a total daily EE (TEE) of 2500 kcal/d, corresponding to a physical activity level (PAL) of 1.74 (PAL
TEE/BEE, where BEE BMR extrapolated to 24 hours).
The PAL can be modified by applying an appropriate multiplier (ActivLevel) to the daily activity pattern.
CHO and fat intakes at breakfast (7:30 AM), lunch (12:30
PM), and supper (7:00 PM) simulate three daily meals. Their
sizes are such that the GLY reserve is raised to 1500, 1600,
and 1800 kcal, respectively. Fat intake at these meals is
calculated to provide a fraction of total intake equal to the
diets fat content (dietF), which is set to 0.4 (i.e., 40%) in
the default condition, The rates of CHO and fat absorption
from the gut correspond to a half-life of 0.69 hours. This
manner of computing meal size inherently serves to maintain CHO balance, although the range within which GLY
levels oscillate can vary depending on the extent of PALinduced depletions between meals or changes in the palatability parameter. It will also bring about energy balance,
once the contribution to EE made by fat oxidation matches
the percentage of fat in the diet.

322
317
51
54

Carbohydrate/Fat Metabolism and Obesity, Flatt

OBESITY RESEARCH Vol. 12 No. 12 December 2004

2017

Carbohydrate/Fat Metabolism and Obesity, Flatt

the fact that these provided the same amounts of CHO. The
difference between the observed CHO balances (22 kcal)
and the computed change in GLY (7 and 8 kcal) was
10% of the amounts of CHO provided, or 4% of the
9-hour EE. The observed fat balances of 290 and 65
kcal were similar to the TG balances of 322 and 46 kcal
computed by the model. Thus, the model replicates the
biological behavior, namely that CHO ingestion promotes
CHO oxidation, whereas fat ingestion does not promote fat
oxidation (11).
A digital file of the model can be obtained by e-mailing
a request to the author in which reference to this publication
is made and the requesters name, address, and affiliation
are stated.
Use of the Model
Unless stated otherwise, the model was run for periods of
125 days. The first 10-day period proceeded under the
default conditions described above, at the end of which
daily energy imbalances were 1 kcal. At that point, one or
more parameters were altered by using the STEP function
provided by the software. The resulting changes and the
subsequent progressive approach toward a new steady state
were recorded in tabular form and displayed in various
graphs. Because the approach to new steady-state conditions is asymptotic and slow, appropriate additions or withdrawals from the TG container (TG) were programmed to
occur on day 105, allowing 20 more days to verify that a
new steady state had indeed become established. This was
judged to be the case when daily energy imbalances were
2 kcal/d. Figure 4 shows a run in which palatability was
increased by 100 kcal on day 11, so that the target GLY
contents after the meal became 1600, 1700, and 1900 kcal,
respectively. This caused an increase in mean GLY content
from 330 to 352 grams. A steady state reestablished itself
after an expansion of the TG reserve from 12 to 16.8 kg had
occurred, which corresponds to a body weight gain from 65
to 72.1 kg and an increase in body fat content from 18.5%
to 23.3%. An increase in mean glycogen content of 88 kcal
thus required a 490 times greater increment of 43,000 kcal
in fat content to raise the overall rate of fat oxidation back
to 40% of total EE. When palatability was instead reduced
by 100 kcal, mean GLY levels (pdmeanGLY) established
themselves at 308 grams, and a steady state reestablished
itself after a TG loss of 4.1 kg, for a body fat content of
13.4%.
Various computations were carried out in which the values of percent dietary fat, palatability, PAL, and FCR modifier parameters were altered on day 11. The steady-state
values reached for TG, %BoFat, minimal, mean, and maximal daily GLY values, and PAL (expressed as the ratio of
total to basal daily EE) were recorded and entered into a
Statview spreadsheet to create figures and for statistical
analysis.
2018

OBESITY RESEARCH Vol. 12 No. 12 December 2004

Figure 4: 125-day (3000-hour) run showing the impact of enhanced palatability and/or food availability and diversity on body
weight maintenance. The change was set to cause a 25-gram
increase in the targeted postmeal CHO content. The figure shows
that this led (1) to an increase from 330 to 352 grams in the mean
GLY reserve, which represents the bodys total glycogen content
(jpmeanGLYg; scale, 0 to 400 grams), (2) to an increase in
cumulative daily energy intake (jpcdEint; vertical scale, 0 to 4000
kcal), but essentially no change (3) in cumulative daily EE
(jpcdEE; vertical scale, 0 to 4000 kcal). Also shown are (4) the
effect on daily TG (fat) oxidation, shown as percentage of daily fat
intake oxidized (jpcd%Fox; vertical scale, 0% to 200%), and (5)
body weight (jBWt; vertical scale, 60 to 100 kg). To reach the
ultimate steady-state condition more rapidly than asymptotically,
the TG reserve (fat mass) was increased by 25,000 kcal (2.8 kg) on
the 105th day (2520 hours). After this addition, energy balance
became established, and TG oxidation matched fat intake.

Results
It was considered that the most informative form to
present the results was in terms of the %BoFat prevailing
when the steady state pertaining to a particular set of parameters was reached. These values are displayed as a
function of dietary fat content, with the prevailing mean
GLY levels (in grams) shown by the numbers appearing
next to each data-point. Figure 5 shows the outcome of the
runs performed with the JNS function made inactive by
setting it to a constant value of 1.0. In this condition, only
unmodified mass effects operate, and differences in physical
activity and in mean GLY levels exert much less influence
than differences in dietary fat content. When the JNS function is active, however, palatability (which alters the targeted GLY levels determining meal size) and PAL exert
considerable influence on steady-state body fat contents
(Figure 6). The effectiveness with which body fat competes
with glucose for oxidation or, in the model, the effectiveness with which TG content competes with GLY content,
can also very powerfully affect steady-state levels of adiposity (Figure 7), suggesting that inherited metabolic differences may affect the propensity for developing obesity
by altering the competition between glucose and fat oxidation.

Carbohydrate/Fat Metabolism and Obesity, Flatt

Figure 5: Steady-state body fat content as a function of dietary

energy consumed as fat and mean GLY levels, without a functioning JNS parameter. Results of 125-day computer runs for various
settings of the palatability (100, 0, 100, or 200 kcal) and
PAL parameters (PAL 1.4, 1.74, 2.1, or 2.4). The numbers
shown near the data-points describe the mean GLY levels (in
grams) maintained under a particular condition.

To evaluate the relative impacts of various parameters on

the %BoFat for which steady states would become established, the results of 33 such model runs were examined by
multiple regression analysis, with % BoFat as the dependent
variable. This led to the following equation (SE in parenthesis):
%BoFat 3.13 8.00 0.95 FCR 0.64 0.08
%diFat 0.10 0.02 Palat
10.3 2.2 PAL
N 33 and R2 0.80 (p 0.0001 for all parameters).
The prediction was improved when mean GLY was used
instead of palatability and PAL:
%BoFat 93.8 9.57 0.55 FCR 0.62 0.04
diFAT 0.237 0.016 meanGLY

Figure 6: Steady-state body fat content as a function of dietary

energy consumed as fat and mean GLY levels on steady-state body
fat content (as implied from the models TG levels), as a function
of dietary energy consumed as fat, and mean GLY levels, with the
JNS parameter operating. Results of 125-day computer runs for
various settings of the palatability (100, 0, 100, or 200 kcal)
and PAL parameters (PAL 1.4, 1.74, 2.1, or 2.4). The numbers
shown near the data-points describe the mean GLY levels (in
grams) maintained under a particular condition. The diagonal grid
lines show how mean GLY levels influence steady-state TG levels
and the degree of adiposity that they imply.

N 33 and R2 0.94 (p 0.0001 for all parameters).

This reflects the fact that, with the use of the model
presented here, where replenishment restores GLY to specified levels, the leverage of physical activity is mainly in
causing greater depletion of GLY between meals. This can
be conceptualized as lengthening the meal-to meal interval, if
calories expended rather than time elapsed is considered (1).

Discussion
The metabolism of living organisms and its regulation are
exceedingly complex, reflecting the outcome of a biological
evolution that enables humans and animals to function and
to sustain themselves under a multitude of conditions by
ingestion of variable proportions of CHOs, fats, and protein
OBESITY RESEARCH Vol. 12 No. 12 December 2004

2019

Carbohydrate/Fat Metabolism and Obesity, Flatt

Figure 7: Impact of permanent alterations in the effectiveness with

which fat competes against GLY for oxidation on steady-state
body fat content as a function of dietary energy consumed as fat
and mean GLY levels (see also Figure 1). Results of 125-day
computer runs with the JNS parameter operating. The FCR parameter, which has the effect of modifying the 1:100 ratio of the
surface areas in the two reservoirs shown in Figure 1, was set at
values of 0.5, 0.75, 1.0, 1.5, 2.0, or 3.0. The numbers shown near
the data-points describe the mean GLY levels maintained under a
particular condition. The diagonal grid lines show the impact of
difference FCR settings on body fat contents, as implied by the
steady-state TG levels that become established in the model.

and continuous extraction of oxygen from the air. The sum

of all these processes can be assessed by measuring the
consumption of oxygen, the production of CO2, and the
urinary excretion of nitrogen, an experimental approach
known as indirect calorimetry (7). All theories and predictions that may be made about metabolic regulation must
be compatible with the assessment of overall metabolic
substrate transformations made in this manner.
Indirect calorimetry has provided extensive evidence
about the relative contributions that glucose and fatty acids
make to substrate oxidation at a given time. It shows that
ingestion of CHO and variations in GLY content exert a
much greater influence on the composition of the fuel mix
oxidized than consumption of fat or variations in the bodys
fat content (11). It is also evident that regulatory mechanisms operate to keep GLY stores within boundaries that are
2020

OBESITY RESEARCH Vol. 12 No. 12 December 2004

narrower than the limits set by the need to avoid hypoglycemia (12) and well below the level at which substantial
conversion of glucose to fat is initiated (5,13). Quite obviously, these behaviors are dictated by the importance of
maintaining an adequate supply of GLY to ensure the supply of glucose in the blood, but without unnecessarily inducing the transformation of glucose into fat. The intake of
CHO and fat and their oxidation, which is stoichiometrically linked to the respiratory exchange, constitute the
bodys main exchanges with the environment. To explore
the interactions between CHO and fat intake and metabolism and their impact on body composition, a computer
model must be able to realistically reproduce these phenomena but not necessarily the complex network of underlying
mechanisms. As shown by Figure 3 and in Table 1, this is
possible by a model built simply on the operations of a
small and a large reservoir, provided that variations in the
content of the small reservoir are allowed to play a predominant role. This is enabled by the use of a parameter mimicking the in vivo influence of insulin. The good imitation of
actual metabolism that this very simple model achieves
suggests that the bodys metabolic responses are also dictated primarily by the conditions imposed for the operation
of any system comprising a small and a large reservoir with
a shared purpose.
The computations performed with this two-compartment
model indicate that several variables affecting daily life can
be expected to exert substantial impacts on the body fat
content that must prevail to bring about a rate of fat oxidation commensurate with fat intake. Quite obviously, one of
these variables is the fat content of the diet itself (Figures 4,
5, and 6). However, factors influencing GLY levels, against
which fat oxidation has to compete, are important as well
(Figure 6). These include physical activity by causing
greater GLY depletion between meals, as well as food
availability, palatability, diversity (9), and social facilitation of eating (10), which can all be expected to influence
the range within which GLY levels are habitually maintained (1). The influence of these food-related effects is
taken into account in the model by altering the palatability
variable.
There has been an alarmingly rapid rise in the prevalence
of obesity during the last two decades (14), even in populations whose gene pool has not changed (15). The reasons
likely to explain this phenomenon in affluent societies have
remained elusive. While a high dietary fat content is widely
considered to be an important contributor to the development of obesity (16), recent increments in the preponderance of obesity cannot be attributed to an increase in dietary
fat content, whose contribution to the average diet may even
have slightly declined during this period (17). A further
decline in average physical activity may have taken place as
well (14), perhaps best revealed by surveys of the number of
hours spent watching television (18), but it is not yet pos-

Carbohydrate/Fat Metabolism and Obesity, Flatt

sible to evaluate the contribution made by this decline. Let

us apply our model to this issue, assuming for instance, that
the diets fat content has declined from 40% to 37%, that
PAL has decreased from 1.74 to 1.60, and that the increase
in food offerings has raised the range within which habitual
GLY levels are maintained by 40 grams. The multiple
regression equation derived from the model runs predicts
that these shifts would cause changes in body fat content of
1.9%, 1.4%, and 4.0%, respectively. Together this
would result in an overall rise in steady-state body fat
content from 18.7% to 22.3%. Based on the analysis of the
NHANES III data presented in a recently published report
on daily reference intakes (19), this would correspond to an
increase in the average BMI of men from 24.7 to 26.7
kg/m2. Using a coefficient of variation of 29% for the
mean BMI in men (19), this would cause increases from
48% to 59% in the preponderance of overweight (BMI 25
kg/m2) and from 24% to 33% in the preponderance of
obesity (BMI 30 kg/m2) among men. Such changes
would seem to be reasonably consistent with recent trends
in the progression of obesity in the United States (15).
Whereas gains or losses in body fat content are recognized by shifts in body weight, changes in GLY levels are
neither noticeable nor measurable. The importance of altered GLY levels for body weight regulation has, therefore,
been generally overlooked. The outcome of the model runs
(Figure 6) and the example given above suggest that environmental factors can, and indeed should, be expected to
manifest metabolic leverage by influencing the range within
which GLY levels are habitually maintained (1). This mechanism offers a simple and plausible explanation for recent
increases in the prevalence of obesity. This view, which is
compatible with the continuous operation of regulatory factors by which energy intake adjusts itself to EE, seems to
provide a better explanation than merely to attribute obesity
to a positive energy balance (20). Given the large daily
variations in daily food consumption (21), phenomena serving to adjust energy intake to expenditure over time must be
operating to account for the maintenance of fairly accurate
long-term energy balance by lean and obese subjects, although in the case of the latter, only after excessive fat
stores have been built.
The influence of some other factors considered to affect
body weight was also examined with the model, e.g., that
the availability of food at all times can induce snacking,
thereby reducing glycogen depletion between meals. The
consumption of a 300-kcal snack at 4:00 PM is compensated
by a decrease in energy intake at supper (from 1184 to 850
kcal) and minor changes at breakfast and lunch. It requires
an increase in body fat content from18.5% to 20% before a
steady state is reestablished at which energy turnover is
2532 instead of 2502 kcal/d. This corresponds to a weight
gain of 2.3 kg and to a 0.9 unit increment in BMI.

It is often considered that switching from a high to a low

glycemic index diet may help weight control (22). We,
therefore, tested the models response to changes in the rate
of nutrient absorption from the gut. Accordingly, the halflife for CHO and fat removal from the gut compartments
was changed from 0.7 to 1.4, or to 2.8 hours, on day 11.
This led to decreases in the TG reserve, corresponding to
decreases of 2.8 and 5.7 kg in steady-state body weights,
with concomitant reductions in body fat content from 18.5%
to 16.2% and 13.8%, respectively.
Inherited traits are well known to affect the development
of obesity. Some genetic differences can be envisioned as
altering the effectiveness with which body fat competes for
oxidation relative to GLY. Differences of a permanent nature can be implemented in the model by altering the FCR
parameter, which is equivalent to changing the ratio of the
diameters and surfaces of the two reservoirs (Figure 1).
Such structural changes can cause substantial shifts in the
%BoFat maintained under given sets of circumstances (Figure 7). This response suggests that antiobesity drugs seeking
to alter metabolism should aim at enhancing fat oxidation
relative to CHO oxidation rather than merely to increase
overall EE, because the latter tends to be compensated by
adjustments in food intake.
In conclusion, the outcome of the model studies shows
that particular degrees of adiposity can be maintained under
a variety of conditions. According to the multiple regression
equation derived from a series of model runs, increases in
the proportion of energy provided by dietary fat from 40%
to 50%, or decreases to 30%, would be offset by 26-gram
decreases or increases in mean GLY levels. Thus, the leverage exerted by GLY levels on body weight maintenance
provides a potential mechanism to account for the impact of
increasing food offerings on the prevalence of obesity. The
model also shows that inherited traits influencing the use of
fat relative to glucose can greatly affect the degrees of
adiposity for which weight maintenance tends to be
achieved.

Acknowledgments
This publication was made possible by National Institute
of Health Grant DK56817. Its contents are solely the responsibility of the author and do not necessarily represent
the official views of the National Institute of Health.
References
1. Flatt JP. McCollum Award Lecture, 1995: diet, lifestyle and
weight maintenance. Am J Clin Nutr. 1995;62:820 36.
2. Roberts SB, Young VR, Fuss P, et al. Energy expenditure
and subsequent nutrient intakes in overfed young men. Am J
Physiol. 1990;259:R4619.
3. Bjorntorp P, Sjostrom L. Carbohydrate storage in man:
speculations and some quantitative considerations. Metabolism. 1978;27:1853 65.
OBESITY RESEARCH Vol. 12 No. 12 December 2004

2021

Carbohydrate/Fat Metabolism and Obesity, Flatt

4. Flatt JP. The difference in the storage capacity for carbohydrate and for fat, and its implications in the regulation of body
weight. Ann N Y Acad Sci. 1987;499:104 23.
5. Acheson KJ, Flatt JP, Jequier E. Glycogen synthesis versus
lipogenesis after a 500 gram carbohydrate meal in man. Metabolism. 1982;31:1234 40.
6. Hellerstein MK. De novo lipogenesis in humans: metabolic
and regulatory aspects. Eur J Clin Nutr. 1999;53:S53 65.
7. Jequier E, Acheson K, Schutz Y. Assessment of energy
expenditure and fuel utilization in man. Ann Rev Nutr. 1987;
7:187208.
8. Randle PJ, Hales CN, Garland PB, Newsholme EA. The
glucose fatty-acid cycle: its role in insulin sensitivity and the
metabolic disturbances of diabetes mellitus. Lancet. 1963;1:
7859.
9. Rolls BJ, Rowe EA, Rolls ET, Kingston B, Megson A,
Gunary R. Variety in a meal enhances food intake in man.
Physiol Behav. 1981;26:21521.
10. de Castro JM. Social facilitation of the spontaneous meal size
of humans occurs on both weekdays and weekends. Physiol
Behav. 1991;49:1289 91.
11. 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.
12. Russek M. Current status of the hepatostatic theory of food
intake control. Appetite. 1981;2:137 43.
13. Acheson KJ, Schutz Y, Bessard T, Anantharaman K, Flatt
JP, Jequier E. Glycogen storage capacity and de novo lipo-

2022

OBESITY RESEARCH Vol. 12 No. 12 December 2004

14.

15.

16.
17.
18.

19.

20.
21.

22.

genesis during massive carbohydrate overfeeding in man.

Am J Clin Nutr. 1988;48:240 7.
ObesityPreventing and Managing the Global Epidemic. Report of a WHO Consultation Technical Report Series 2000.
Geneva, Switzerland, 2000.
Mokdad AH, Bowman BA, Ford ES, Vinicor F, Marks JS,
Kaplan PJ. The continuing epidemics of obesity and diabetes
in the United States. J Am Med Assoc. 2001;286:1195200.
Bray GA, Popkin BM. Dietary fat intake does affect obesity!
Am J Clin Nutr. 1998;68:115773.
Willett WC. Dietary fat and obesity: an unconvincing relation. Am J Clin Nutr. 1998;68:1149 50.
Gortmaker SL, Must A, Sobel AM, Peterson K, Colditz
GA, Dietz WH. Television viewing as a cause of increasing
obesity among children in the United States, 1986 1990. Arch
Pediatr Adolesc Med. 1996;150:356 62.
Dietary Reference Intakes: Energy, Carbohydrate, Fiber, Fat,
Fatty Acids, Cholesterol, Protein and Amino Acids. Report of
the Institute of Medicine. Washington, DC: National Academies Press, 2000.
Flatt JP. How NOT to approach the obesity problem. Obes
Res. 1997;5:6323.
Bingham SA, Gill C, Welch A, et al. Comparison of dietary
assessment methods in nutritional epidemiology: weighed
records v. 24h recalls, food-frequency questionnaires and estimated-diet records. Br J Nutr. 1994;72:619 43.
Ludwig DH. Dietary glycemic index and obesity. J Nutr.
2000;130:280S3S.