International Journal of Obesity (1998) 22, 980±987
ß 1998 Stockton Press All rights reserved 0307±0565/98 $12.00
http://www.stockton-press.co.uk/ijo
The effect of covertly manipulating the energy
density of mixed diets on ad libitum food intake
in `pseudo free-living' humans
RJ Stubbs1, AM Johnstone1, LM O'Reilly1, K Barton and C Reid2
1
The Rowett Research Institute and 2Biomathematics and Statistics Scotland, The Rowett Research Institute, Bucksburn, Aberdeen, UK
OBJECTIVE: This study examined the effects of covert alterations in the energy density (ED) of mixed, medium fat
(MF) diets on ad libitum food and energy intake (EI), subjective hunger and body weight in humans.
DESIGN: Randomised cross-over design. Subjects were each studied three times (factorial design), during 14 d,
throughout which they had ad libitum access to one of three covertly-manipulated MF diets.
SUBJECTS: Six healthy men, mean age (s.e.m.) ˆ 30.0 y (12.76 y), mean weight ˆ 71.67 kg (19.80 kg); mean
height ˆ 1.79 m (0.22 m), body mass index (BMI) ˆ 22.36 (2.60) kg/m2, were studied. The fat, carbohydrate (CHO) and
protein in each diet (as a proportion of the total energy) and energy density (ED) were, low-ED (LED), 38:49:13%;
373 kJ/100 g; medium-ED (MED), 40:47:13%; 549 kJ/100 g; high-ED (HED), 39:48:13%; 737 kJ/100 g. Subjects could alter
the amount but not the composition of foods eaten. They were resident in (but not con®ned to) a metabolic suite
throughout the study.
RESULTS: Solid food intake decreased as ED increased, giving mean values of 2.84, 2.51 and 2.31 kg/d, respectively.
This was insuf®cient to defend energy balance, since energy intake increased with increasing ED (F (2,10) 16.08;
P < 0.001) giving mean intakes of 10.12, 12.80 and 16.17 MJ/d, respectively. Rated pleasantness of food (measured on
visual analogue scales) was not signi®cantly different between diets nor was subjective hunger different between the
LED, MED and HED diets, respectively. Diet signi®cantly affected body weight (F (2,10) ˆ 4.62; P ˆ 0.038), producing
changes of 71.20, 0.02 and 0.95 kg, respectively, by day 14.
CONCLUSION: Dietary ED can in¯uence EI and body weight, since changes in amount eaten alone are insuf®cient to
defend energy balance, when subjects feed on unfamiliar diets and diet selection is precluded. Comparison with our
previous studies suggest that there was compensation in solid food intake when ED was altered using mixed diets (as
in this study) compared to previous studies which primarily used fat or CHO to alter dietary ED.
Keywords: carbohydrate; fat; energy; macronutrients; food intake; appetite; energy density; humans
Introduction
Increased fat intake is considered to be a major factor
in Western society, predisposing individuals to gain
weight. Indeed, government reports now suggest that
in order to limit weight gain, people should speci®cally attempt to reduce dietary fat intake (for example,
see Ref. 1). The evidence that fat somehow bypasses
appetite control and that changes in fat intake are a
major factor underlying secular trends in body weight
appears to be compelling.2 ± 6
The food industry has responded to dietary recommendations regarding fat, by producing a large
number of food products which are `low' or `lowerfat' foods. These foods are not always foods which are
low in energy density (ED) or in the case of some
`lower fat' foods, low in absolute fat content. Despite
this fact, some consumers appear to have developed
the perception that in order to lose weight, a person
simply needs to reduce their total dietary fat intake.
Correspondence: RJ Stubbs, The Rowett Research Institute,
Greenburn Road, Bucksburn, Aberdeen, AB21 9SB, UK.
Received 22 August 1997; revised 12 March 1998; accepted
22 May 1998
This may not be the case since two recent studies
found that the increased consumption of modi®ed low
fat (LF) foods decreases fat but not energy intake (EI),
in free-living consumers.7,8 These results suggest that
while dietary fat intake may be a risk factor for weight
gain, substantial weight loss is unlikely to be achieved
simply by increasing the consumption of LF food
products.
The majority of studies that compare the effects of
high fat (HF) and LF diets on appetite and indices of
energy balance, confound diet composition with ED,
since dietary fat content and ED tend to co-vary.
These studies have not therefore assessed whether
the excess EI induced by consumption of HF diets is
a macronutrient-speci®c effect. Two studies have
compared the effects of isoenergetically-dense HF
and LF diets on EI. When the ED of HF and LF
diets was very similar, the effect of fat in promoting
excess EI was not apparent.9,10 These ®ndings have
led to the suggestion that the effects of fat in promoting excess EI is simply a non-nutrient speci®c effect
of dietary ED.11 However, many experts would agree
that the effects of diet composition on feeding behaviour cannot be simply reduced to a question of
whether dietary fat or ED promotes excess EI.
Energy density of mixed diets and energy intake
RJ Stubbs et al
Other studies which have compared the effects of
isoenergetic (1.0 ±1.5 MJ) macronutrient loads on
subjective hunger, have found protein to be the most
satiating macronutrient.12 ± 14 Differences between isoenergetically-dense HF and high carbohydrate (HC)
diets are detectable, but tend to be more subtle than
when fat contributes disproportionately to dietary ED
(for example, see Ref. 15). A consideration of the
literature suggests, that under ecological conditions,
there is a hierarchy in the satiating ef®ciencies of the
macronutrients protein (most satiating), carbohydrate
(CHO) and fat (least satiating).16 This hierarchy may
be over-ridden by changes in dietary ED, since we
have recently shown that subjects can spontaneously
overeat on high-CHO (HC) diets which are high in
ED.17 A much more complex picture of the way
human subjects respond to different controlled (and
often covert) manipulations of the nutrient content
and=or ED of the diet is thus beginning to emerge.
Walls and Koopmans18 have found that in rodents,
compensatory responses were greater when rats were
given intravenous infusions of mixed nutrients, than
when either fats or glucose were given alone. Compensatory responses were greater for glucose than fat.
This effect has been found in humans.19 Campbell et
al20 observed marked (albeit incomplete) compensatory responses for covert alterations in the ED of
mixed, liquid diets in lean but not obese subjects.
These observations, together with the above discussion, suggest that subjects may compensate to a
greater extent for variations in the ED of mixed
diets than when the ED of the diet is altered primarily
using fat or CHO.
This study therefore examined subjects' responses
to variations in the ED (but not the composition) of
mixed diets, under the same experimental conditions
that we have used to examine responses to ED
manipulations of the diet by primarily using fat,5,6
CHO17 or isoenergetically dense HF and HC diets.10
Materials and methods
Subjects
Six healthy men (mean (s.e.m.) age 30.00 (12.76) y;
weight 71.67 (19.80) kg; height 1.79 (0.22) m; body
mass index (BMI) 22.36 (2.60) kg=m2) were recruited
by advertisement. They were normal weight, nonsmoking, non-trained, healthy men, who were not
taking any medication. Subjects were resident in, but
not con®ned to, the metabolic suite, since they could
come and go as they pleased. They also had ad libitum
access to a diet that was of a ®xed composition per run
and so were not strictly free-living in the true sense of
the phrase.
Study design
Each subject was studied three times in 16 d during
treatments which began with 2 d equilibration on a
diet designed to standardise energy and macronutrient
intakes at 1.6 basal metabolic rate (BMR), as
described previously.17 The ED of these meals were,
on average 600 kJ=100 g. During the subsequent 14 d,
the subjects had continuous ad libitum access to one
of three covertly manipulated diets, which were low
(LED), medium (MED) or high-energy density (HED)
(Table 1). The proportion of energy derived from
protein, CHO and fat was the same on each diet.
Each dish was made in three versions, corresponding
to the LED, MED and HED regimes. The ED of each
food type increased on going from the LED to HED
version of the same food. The order of diets was
randomised in a counterbalanced design across subjects. The energy and macronutrient composition of
each time of the diet, the diet recipes, three-day
rotating menu and instructions for preparation of the
foods, can be obtained by contacting the authors. The
means of anthropometric measurement, food preparation, presentation and food intake measurements have
been described in detail previously.17
Subjects completed subjective hunger, appetite and
palatability ratings 15 min after each meal as
described by Hill and Blundell.21 There was a minimum period of one week between treatments, during
which subjects had access to free food. The subjects
entirely determined their own daily activity regimes.
We did not therefore ensure that physical activity was
strictly controlled across treatments, although subjects
reported that they did not change their normal routines
during the course of the study. The study was
approved by the Grampian Health Board and University of Aberdeen Joint Ethical Committee.
Table 1 Average composition of dietary treatments expressed as water, dry matter, energy and nutrient composition for the high
(HED), medium (MED) and low-energy density (LED) diets
HED
(s.e.m.)
MED
(s.e.m.)
LED
(s.e.m.)
Water
(g)
Dry matter
(g)
Energy
(MJ)
Protein
(MJ)
Fat
(MJ)
CHO
(MJ)
Sugar
(g)
Starch
(g)
NSP
(g)
%
Protein
%
CHO
%
Fat
583
(26)
671
(27)
762
(23)
417
(26)
329
(27)
238
(23)
7.38
(0.14)
5.49
(0.05)
3.56
(0.03)
0.98
(0.01)
0.72
(0.01)
0.48
(0.01)
2.89
(0.12)
2.19
(0.03)
1.38
(0.02)
3.51
(0.02)
2.57
(0.02)
1.70
(0.01)
110
(14)
88
(11)
57
(7)
100
(14)
63
(11)
40
(7)
5
(1)
6
(1)
6
(1)
13
(0)
13
(0)
13
(0)
48
(1)
47
(0)
48
(0)
39
(1)
40
(0)
39
(0)
CHO ˆ carbohydrate; NSP ˆ non-starch polysaccharides.
981
Energy density of mixed diets and energy intake
RJ Stubbs et al
982
Statistical analysis
ANOVA was conducted on the intakes of energy, fat,
CHO and protein for the individual 24 h results from
each subject, using the Genstat 5 statistical program
(Genstat 5 Rothampstead Experimental Station, Harpenden, UK). The analysis treated subject, run and
week (week 1 or 2 of each run) as blocking factors,
and diet and week as factors. Menu days were
included as a covariate. Post-hoc Scheffe tests were
conducted to assess signi®cant differences between
speci®c mean values. The analysis of the visual
analogue ratings was complicated by the large
number of missing observations for some subjects.
Therefore it was necessary to check that there were no
systematic differences in the times of non-response
between diets, as this may bias the results. A logistic
regression was performed with binary response, observation or missing value. Terms for subject, diet, time
and the interactions between subject and diet, and
subject and time, were included in the model. The
interactions between time and diet, and subject, time
and diet were then added to the model. After ®nding
no evidence for systematic differences in response,
average ratings were calculated for each day and these
were analysed by ANOVA, with diet as factor and
subject as blocking factor.
Hunger ratings were analysed by ANOVA with
subject, run, week and weekday as blocking factors
and diet, week and weekday as factors. Pleasantness
and satisfaction were analysed by ANOVA with
subject, run, week and meal as blocking factors
and diet, week and meal as factors. Menu day
was included as a covariate. Change in body
weight was analysed by ANOVA, with subject,
run and week as blocking factors and day, week
and diet as factors.
Results
Food, energy and nutrient intakes
Table 2 summarises the mean intakes of weight,
energy, fat, CHO, protein, sugar and starches, and
gives the standard errors of the difference between the
means (s.e.d.), F ratios and probability values, associated with the main effects for dietary manipulation.
There were no signi®cant differences in the weight of
food and non-caloric drinks (summed) consumed
between diets (Figure 1). There was no signi®cant
increase or decrease in the rate of food and drink
intake over the 14 d of the study (Figure 1). When the
wet weight of food without non-caloric drinks was
examined, there was a signi®cant compensatory
decrease in food intake as dietary ED increased,
which was insuf®cient to defend energy balance,
since the effect of dietary ED on EI was highly
signi®cant (Table 2). Menu-day effects were also
signi®cant and will not be discussed further. Since
the composition of the diets was constant, macronutrient intakes (Figure 2) paralleled EI.
Figure 3 shows that subjects lost weight over the
14 d on the LED diet, weight did not signi®cantly
change on the MED diet and increased on the HED
diet, giving cumulative weight changes of 71.20 kg,
0.02 kg and 0.91 kg on the LED, MED and HED diets,
respectively, [F (2,10) ˆ 4.62; P ˆ 0.038] by the
morning of day 15.
Subjective hunger, appetite and perceived pleasantness
of the diets
Subjects differed from one another in their expression
of subjective hunger (F (5,8) ˆ 294; P < 0.001). Mean
Table 2 Average daily food, energy and nutrient intakes (s.e.m.) for the low energy density (LED), medium energy
density (MED) and high energy density (HED) dietary treatments, for the six subjects. Variance ratios, standard
errors of the difference between the means (s.e.d.) and statistical probability for the main treatment effects has also
been summarised
Food and drink (g)
s.e.m.
Water (g)
s.e.m.
Solid food (g)
s.e.m.
Energy (MJ)
s.e.m.
Protein (MJ)
s.e.m.
Fat (MJ)
s.e.m.
CHO (MJ)
s.e.m.
Sugar (g)
s.e.m.
Starch (g)
s.e.m.
NSP (g)
s.e.m.
a
LED
MED
HED
Variance ratio
Probability
s.e.d.
4075
(215)
3513
(194)
2839a
(637)
10.19a,c
(0.41)
1.35a,c
(0.05)
3.80a,b
(0.15)
4.97a,c
(0.20)
135a,b
(5)
154a,c
(8)
20a,b
(1)
3410
(166)
2718
(140)
2506
(550)
12.80
(0.59)
1.68
(0.08)
5.03
(0.23)
6.00
(0.28)
173
(8)
167
(10)
13
(1)
3663
(199)
2819
(173)
2304
(540)
16.17
(0.60)
2.11
(0.08)
6.23
(0.23)
7.64
(0.28)
192
(8)
265
(12)
14
(1)
1.77
0.219
356.3
3.93
0.055
308.8
3.93
< 0.01
192.3
13.23
< 0.002
1165.2
13.63
< 0.001
145.9
13.46
< 0.001
466.8
7.64
< 0.002
546.4
8.15
0.008
14.74
20.34
< 0.001
19.06
14.85
0.001
1.546
Indicates a signi®cant difference between LED and HED diets; b indicates a signi®cant difference between LED and
MED diets; cindicates a signi®cant difference between MED and HED diets; CHO ˆ carbohydrate; NSP ˆ non-starch
polysaccharides.
Energy density of mixed diets and energy intake
RJ Stubbs et al
983
Figure 1 Mean (s.e.m.) daily intake of food and energy for the
six subjects on the three dietary treatments. ANOVA con®rmed
the treatment effect was signi®cant at P < 0.002.
daily hunger was 26 mm, 24 mm and 24 mm on the
LED, MED and HED diets, respectively. Results for
`fullness', `urge to eat', `desire to eat' and `thoughts
of food', exhibited similar non-signi®cant trends.
There was a signi®cant diet effect for prospective
consumption (F (2,8); 6.41; P ˆ 0.02). Mean values
were 36 mm, 37 mm and 31 mm on the LED, MED
and HED diets, respectively. Clearly this effect was
accounted for by a contrast between the HED diet and
the other two diets. There were no signi®cant differences in rated pleasantness of the food between days
or diets (F (2,8) 1.18; P ˆ 0.380). The mean values for
pleasantness were 80 mm, 84 mm and 84 mm for the
LED, MED and HED diets, respectively.
Figure 2 Mean (s.e.m.) daily protein, carbohydrate (CHO) and
fat intakes for the six subjects on the three dietary treatments.
ANOVA con®rmed the treatment effect was signi®cant at
P < 0.001.
Discussion
The effect of increasing the ED of mixed diets on food
and EI
Covertly increasing the ED of mixed diets, led to a
graded increase in the EI of these subjects. The MED
diet was chosen to have an average ED of approximately 550 kJ=100 g, because subjects have been
shown to eat to approximate energy balance on similar
diets, of this composition, in our laboratory. Indeed, in
the present study, it appears from the body weight
data, that subjects were in approximate energy
Figure 3 Mean (s.e.m.) change in body weight, relative to day 1
(set to zero), for ®ve of the six subjects on the three dietary
treatments. ANOVA con®rmed the treatment effect was signi®cant at P ˆ 0.038.
balance on the MED diet. The LED (373 kJ=100 g)
and HED (737 kJ=100 g) diets were chosen to be
approximately 36% less or more in ED than the
MED diet, respectively. At ®rst sight, it appeared
Energy density of mixed diets and energy intake
RJ Stubbs et al
984
Figure 4 Mean (s.e.m.) daily food intake in response to changes
in the energy density (ED) of the diet using mixed nutrients
(present study), primarily fat6 and primarily carbohydrate
(CHO).21 There were six men in each experiment, which lasted
14 d per dietary treatment. Signi®cance levels are indicated for
treatments effects within each study. NS ˆ not statistically
signi®cant.
that the subjects in the present study were consuming
an amount of food and drink that was similar on each
diet. We have previously expressed EI on MED and
HED diets, in relation to that on the LED diet, to
assess whether subjects ate in direct proportion to the
ED of the diets. Indeed, this was the case when the ED
of diets was previously increased covertly primarily
using fat.5,6 In our previous study, where the ED of
the diet was manipulated primarily using CHO, subjects showed an incipient but non-signi®cant tendency
to eat more food on the LED diet17 (Figure 4). EI was
thus largely in proportion to the ED of the diet.17 In
the present study, expressing the average ED of each
diet relative to the LED diet, gave ratios of 1.00, 1.47
and 1.98. Expressing EI on each diet, relative to the
LED diet, gave ratios of 1.00, 1.26 and 1.60, respectively, indicating that the difference in EI between
diets was less than it would have been had subjects not
compensated to some degree by altering food intake.
The wet weight of solid food also decreased as the ED
of the diet increased giving values of 2.84, 2.51 and
2.31 kg=d on the LED, MED and HED diets, respectively. Thus subjects consumed 12% and 19% less in
solid food, on the MED and HED diets, in comparison
to the LED diet. Clearly this degree of caloric compensation was insuf®cient to defend energy balance. It
is important to question how quantitatively signi®cant
these effects actually are. In order to appreciate this,
the nature of the experimental design must be taken
into account.
Limitations of the current experimental design
In studies where subjects have access to a range of
familiar foods, after ingesting macronutrient-manipulated foods,7,8,22 ± 25 caloric compensation is far more
complete than when they can alter the amount but not
type, composition or ED of the foods eaten. Studies,
such as the present study, tend to be relatively
insensitive, in that they only allow the subject to
respond to a dietary manipulation by eating more or
less of a diet in a ®xed composition and ED. Furthermore, dietary manipulations are often covert, the order
of runs is randomised and the diets are formulated to
appear as similar as possible in taste, texture and
appearance. As in most studies of feeding behaviour,
there were limitations to the design of this experiment.
The study design (and hence conclusions arising from
it) was subject to the following constraints. The food
items are not common, familiar or `real' foods.
Furthermore the ED of the diets range from 356 ±
738 kJ=100 g, which in absolute terms is a narrow
range relative to foods available in the free-living
environment. It is likely that subjects would learn to
compensate more rapidly for changes in the ED and
composition of manipulated diets, if the manipulations
were overtly paired to appropriate learning cues.
Ideally, the results of overt and covert manipulations
should be compared. This study was conducted in a
very small number of subjects and these results should
perhaps be replicated before attempting to extrapolate
them to other groups or the general population. This
design precludes learning about the physiological
consequences of ingesting manipulated foods and
the ability of subjects to respond to those manipulations, is blunted by these constraints of the experimental design. The result of this is to arti®cially raise
the causal effect of dietary ED on EI. In other words,
if a diet is manipulated in a way that subjects respond
weakly, signi®cant differences in energy (but not
food) intake will occur. Subjects fail to compensate
to any great degree under these experimental conditions and EI tends to re¯ect the initial dietary manipulation. Thus, face-value interpretations of
quantitative-response studies, such as these, may
tend to overestimate the `regulatory' signi®cance of
passive overconsumption, as the ED of diets is
increased, and underestimate the importance of
active changes in food intake or subjective hunger.
This argument is supported by studies conducted
under more ecological conditions7,8 and studies
using diets more similar to those encountered in
everyday life,22 ± 25 in which subjects tend to compensate more accurately for alterations in dietary ED.
These arguments also imply that it is important to
design experiments which examine both quantitative
(amount eaten) and qualitative (selection of food type,
composition and ED) responses to dietary manipulations.
Comparison of a series of quantitative studies
examining variations in the ED of the diet primarily
using fat, CHO or mixed nutrients
If the above arguments are accepted, interesting differences in group responses occurred in a series of
studies which manipulated dietary ED primarily using
fat5,6 or CHO,17 under the same experimental conditions. When the ED of the diet was altered primarily
using fat, subjects did not compensate at all for
alterations in dietary ED.5,6 Neither were there
Energy density of mixed diets and energy intake
RJ Stubbs et al
signi®cant differences in subjective hunger between
the diets.5 When ED was altered, primarily using
CHO, there was no compensation for the ED of the
diets, but subjects were signi®cantly (and anecdotally)
more hungry on the LED diet, compared to the HED
diet.17 In the present study, there were no signi®cant
differences in subjective hunger between the diets, but
there was a signi®cant diet effect for prospective
consumption. Subjects also showed a highly signi®cant (albeit quantitatively small) compensation of
food and EI on going from the LED to the HED
diet. Thus, subjects responded more effectively to
changes in the ED of the mixed nutrient manipulations
(by active changes in food intake) than the primarily
CHO-based manipulations (detectable changes in
hunger),17 which in turn showed a greater response
than the primarily fat-based manipulations.5,6 These
patterns are illustrated in Figure 4. These comparisons
are consistent with infusion studies in rodents18 and
humans,19 which have found greater caloric compensation in response to mixed infusions than glucoseinfusions, which in turn produced greater caloric
compensation than lipid infusions. These ®ndings in
turn, are consistent with the notion that dietary macronutrients exert hierarchical effects on satiety (protein
having the greatest effect and fat the least).16
Body weight changes
By day 14 of each dietary treatment, mean body
weights had changed by 71.20, 0.02 and 0.95 kg on
the LED, MED and HED diets, respectively. Change
in body weight can be used to estimate change in
energy balance, using estimates of the energy cost of
weight gain and weight loss. The energy content of
adipose tissue and lean body mass can be combined
with the estimated ratio of fat (61%) to lean tissue
(39%) contributing to body weight changes in a group
of men with the physical characteristics of these
subjects, to estimate the approximate energy content
of weight change.26,27 It should be noted that the
above ratio of fat to lean tissue corresponds to a
ratio of 75% adipose tissue to 25% lean body mass,
since adipose tissue comprises 80% lipid and 20%
lean tissue. The adipose tissue of Reference Man
comprises 80% fat, 15% water and 5% protein
as wet tissue.28 The gross energy of fat of
39.9 MJ=kg and of protein is 23.6 MJ=kg. The accurate estimate of the energy content of adipose tissue
should therefore be (0.8 39.9) ‡ (0.15 0) ‡
(0.05 23.6) ˆ 33.1 MJ=kg. The gross energy of
average meat protein is 23.6 MJ=kg. The estimated
protein content of lean tissue is 20% wet weight,
giving an estimated energy content of lean tissue
amounting to 23.6 MJ=kg 0.2 ˆ 4.72 MJ=kg. The
estimated energy content of body weight change can
therefore be calculated as (0.75 33.1) ‡ (0.25 4.72) ˆ 26.2 MJ=kg, assuming that 39% of tissue
gain is lean tissue and 61% is fat (or 25% lean body
mass and 75% adipose tissue).
Forbes et al26 made estimates of the energy cost of
weight gain in adults. Using empirical data from their
own laboratory and a number of rigorous studies from
the literature, they have estimated the relationship
between total excess energy ingested and total
weight gain. From this they estimate the energy cost
of weight gain in non-obese subjects to be 33.7 MJ=kg
weight gain. Subtracting the energy content of weight
change from the energy cost of weight gain gives an
estimate of the cost of synthesis and maintenance of
tissue deposition, 33.7 7 26.2 ˆ 7.5 MJ=kg.
The energy cost of weight loss is likely to be less
than that for weight gain. This is because the energy
cost of tissue mobilisation is far less than the energy
cost of tissue synthesis, and is likely to be cancelled
out by the decrease in the energy required to maintain the tissues that have been lost. We have
assumed this to be so and have therefore used the
energy content of weight change (26.2 MJ=kg), estimated above as an estimate of the energy cost of
weight loss for this group of subjects. This is
necessarily an approximation, and in early life the
energy cost of weight gain and loss may be in¯uenced by a number of factors, including body composition and physical activity of the subjects.26,27
Using these values and the changes in body weight
gives an estimated change in energy balance
of 71.2 26.2 ˆ 731.4 MJ, 0.02 33.7 ˆ 0.7 MJ
and 0.95 33.7 ˆ 32.0 MJ on the LED, MED and
HED diets, respectively. Assuming that subjects were
in zero energy balance on the MED diet, the differences in EI would have cumulated to produce energy
balances of 736.4 and 47.2 MJ for the LED and
HED diets, respectively. Thus the estimated change
in energy balance from increase in weight was 15 MJ
lower, and from decrease in weight 5 MJ lower, than
the estimates using differences in EI, assuming zero
energy balance on the MED diet.
Body weight ¯uctuated, on a day-to-day basis, by
an average 220 g=d (range ˆ 20±670 g) on the MED
diet. Using an average of the energy cost of weight
gain (33.7 MJ=kg) and loss (26.2 MJ=kg), gives an
approximate mean energy cost of weight change of
30.0 MJ=kg. This would give an average daily uncertainty of (0.22 30.0) ˆ 6.6 MJ (range ˆ 0.6±
20.1). The above discrepancy is therefore at the
extreme end of the error-range, arising from an
assumed energy cost of weight change of
30.0 MJ=kg. These differences between energy balance estimated from change in body weight and
change in EI are also within the errors arising from
dietary intake measurements and the assumptions
used in calculating the energy cost of weight gain
and loss. Subjects may have been more physically
active or less compliant on the HED diet. However,
if subjects were non-compliant on this diet, it would
not have been due to their eating foods not on the
menu, since if that were the case they should have
gained more and not less weight than expected. It is
unlikely that subjects pretended to eat more food,
985
Energy density of mixed diets and energy intake
RJ Stubbs et al
986
since they were feeding ad libitum and could eat as
much or as little as they wanted and could simply
leave the food in their fridge.
ED, ¯uid and food intake
In the present study, ¯uid intake estimates were not as
precise as the food intake estimates, since subjects
would occasionally drink water from other sources ±
the most common being the kettle, for tea and coffee.
However, it was of interest to note that subjects
tended to drink more of the drinks provided as the
ED and hence dry matter of the diets increased.
Furthermore, failing to differentiate between ¯uid
and food intake in the present study would have led
to the erroneous conclusion that the subjects did not
compensate at all for the increase in dietary ED. The
exact relationship between ¯uid and food intake in
human feeding behaviour is unclear. It is also unclear
how ¯uids may interact with the food matrix undergoing digestion in the gut (Mela, personal communication). It may be hypothesised that when ¯uids and
food matter interact to form a complex in the gut, that
this may enhance short-term satiety. When solid food
matter and ¯uids interact less in the gut, it is likely
that ¯uid intake will exert less in¯uence over solid
food intake.
Conclusions
This study concludes a series of studies where we
have examined the effects of covertly altering the ED
of the diet, primarily in the form of fat,5,6 or CHO,17
and in this study using mixed diets of ®xed composition. These studies suggest that under the precise, but
behaviourally insensitive, experimental conditions of
quantitative feeding studies, where subjects can only
alter the amount, but not the composition of covertly
manipulated foods eaten, changes in dietary ED as fat,
as CHO and as mixed diets, can exert profound effects
on EI and body weight. The evidence from more
ecological studies7,8,22 ± 25 suggests that in real-life,
where subjects can alter the amount, type, composition and ED of foods they ingest, compensatory
responses to changes in diet composition and=or ED
are likely to be more accurate, though often incomplete. Comparing the results of this study to our
previous work, suggests that subjects may respond
to a greater extent to covert manipulations of the ED
of mixed diets than changes in the ED of CHO-rich
diets, which produce greater responses than manipulations of the ED of the diet by increasing its fat content.
The signi®cance of these effects under ecological
conditions has yet to be determined. It is recommended where possible, that quantitative studies
such as the present study be compared to studies
where subjects can respond to prior dietary manipulation by altering both the amount and composition of
foods they subsequently select. Food and nutrient
selection may well be an important component of
caloric compensation.
Acknowledgements
This work was supported by funding from the Scottish
Of®ce Department of Agriculture, Environment and
Fisheries. We are grateful to Scienti®c Hospital Supplies Ltd, Liverpool, UK, for the donation of the
MaxiJoule.
References
1 Department of Health. The health of the nation: A strategy for
health in England. HMSO: London, 1992.
2 Lissner L, Heitmann BL. Dietary fat and obesity: evidence
from epidemiology. Eur J Clin Nutr 1995; 49: 79 ± 90.
3 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 ± 942.
4 Kendall A, Levitsky DA, Strupp BJ, Lissner L. Weight-loss
on a low fat diet: consequence of the impression of the control
of food intake in humans. Am J Clin Nutr 1991; 53:
1124 ± 1129.
5 Stubbs RJ, Harbron CG, Murgatroyd PR, Prentice AM. Covert
manipulation of dietary fat and energy density: effect on
substrate ¯ux and food intake in men feeding ad libitum. Am
J Clin Nutr 1995; 62: 316 ± 330.
6 Stubbs RJ, Ritz P, Coward WA, Prentice AM. Covert manipulation of the ratio of dietary fat to energy density: effect on
food intake and energy balance in free-living men feeding ad
libitum. Am J Clin Nutr 1995; 62: 330 ± 338.
7 Gatenby SJ, Aaron G, Morton G, Mela DJ. Nutritional
implications of reduced-fat food use by free-living consumers.
Appetite 1995; 25: 241 ± 252.
8 Aaron JI, Gatenby SJ, Jack V, Mela DJ. Effect of reduced fat
and reduced sugar food use on diet and weight control status of
free-living consumers (abstract). Proc Nutr Soc 1996; 55:
215A.
9 Van Stratum P, Lussenburg RN, van Wezel LA, Vergroesen
AJ, Cremer HD. The effect of dietary carbohydrate:fat ratio on
energy intake by adult women. Am J Clin Nutr 1978; 31:
206 ± 212.
10 Stubbs RJ, Harbron CG, Prentice AM. The effect of covertly
manipulating the dietary fat to carbohydrate ratio of isoenergetically dense diets on ad libitum food intake in free-living
humans. Int J Obes 1996; 20: 651 ± 660.
11 Prentice AM, Poppitt SD. Importance of energy density and
macronutrients in the regulation of energy intake. Int J Obes
1996; 20: S18 ± S23.
12 Stubbs RJ, Van Wyk MCW, Johnstone AM, Harbron C.
Breakfasts high in protein, fat or carbohydrate: effect on
within-day appetite and energy balance. Eur J Clin Nutr
1996; 50: 409 ± 417.
13 Johnstone AM, Harbron CG, Stubbs RJ. Macronutrients,
appetite and day-to-day food intake in humans. Eur J Clin
Nutr 1996; 50: 418 ± 430.
14 Johnstone AM, Stubbs RJ, Reid C. Macronutrient composition
of weight-reducing diets: effects on compliance and weight
loss [Abstract]. Int J Obes 1997; 21 (Suppl 2): S212.
15 Rolls BJ, Kim-Harris S, Fischmann MW, Foltin RW, Moran
TH, Stoner SA. Satiety after preloads with different amounts
of fat and carbohydrate: implications for obesity. Am J Clin
Nutr 1994; 60: 476 ± 487.
16 Stubbs RJ. Macronutrient effects on appetite. Int J Obes 1995;
19 (Suppl 5): S11 ± S19.
Energy density of mixed diets and energy intake
RJ Stubbs et al
17 Stubbs RJ, Harbron C, Johnstone AM. The effect of covertly
manipulating the energy density of high-carbohydrate diets on
ad libitum food intake in `pseudo free-living' humans. Int J
Obes (this volume).
18 Walls EK, Koopmans HS. Differential effects of intravenous
glucose, amino acids and lipid on daily food intake in rats. Am
J Physiol 1992; 262: R225 ± R234.
19 Gil K, Skeie B, Kvetan V, Askanazi J, Friedman MI. Parenteral nutrition and oral intake: Effect of glucose and fat
infusion. J Parenter Ent Nutr 1991; 15: 426 ± 432.
20 Campbell RG, Hashim SA, Van Itallie TB. Studies of food
intake regulation in man. Responses to variations in nutritive
density in lean and obese subjects. New Engl J Med 1971; 285:
1402 ± 1407.
21 Hill AJ, Blundell JE. Nutrients and behaviour: research
strategies for the investigation of taste characteristics food
preferences, hunger sensations and eating patterns in man.
J Psychol Res 1982; 17: 203 ± 212.
22 Foltin RW, Fischman MW, Moran TH, Rolls BJ, Kelly TH.
Caloric compensation for lunches varying in fat and carbohydrate contents by humans in a residential laboratory. Am J Clin
Nutr 1990; 52: 969 ± 980.
23 Foltin RW, Rolls BJ, Moran TH, Kelly TH, McNelis AL,
Fischman MW. Caloric, but not macronutrient, compensation
by humans for required eating occasions with meals and
snacks varying in fat and carbohydrate. Am J Clin Nutr
1992; 55: 331 ± 342.
24 Caputo FA, Mattes RD. Human dietary responses to covert
manipulations of energy, fat and carbohydrate in a midday
meal. Am J Clin Nutr 1992; 56: 36 ± 43.
25 Mattes RD, Pierce CB, Friedman MI. Daily caloric intake of
normal weight adults: responses to changes in dietary energy
density of a luncheon meal. Am J Clin Nutr 1988; 48:
214 ± 219.
26 Forbes GB, Brown MR, Welle SL, Lipinski BA. Deliberate
overfeeding in women and men: energy cost and composition
of the weight gain. Br J Nutr 1986; 56: 1 ± 9.
27 Forbes GB. Do obese individuals gain weight more easily
than nonobese individuals? Am J Clin Nutr 1990; 52:
224 ± 227.
28 Snyder WS, Cook MJ, Nasset ES, Karkhausen LR, Parry
Howells G, Tipton IH. Report of the Task Group on Reference
Man. International commission on radiological protection
No. 23. Pergamon Press: Oxford, 1974.
987