Randomized trial on protein vs carbohydrate in
ad libitum fat reduced diet for the treatment of
, S Toubro
, B Rùnn
, L Holm
and A Astrup
Research Department of Human Nutrition, The Royal Veterinary and Agricultural University, Copenhagen, Denmark and
of Mathematics, The Royal Veterinary and Agricultural University, Copenhagen, Denmark
OBJECTIVE: To study the effect on weight loss in obese subjects by replacement of carbohydrate by protein in ad
libitum consumed fat-reduced diets.
DESIGN: Randomized dietary intervention study over six months comparing two ad libitum fat reduced diets (30% of
total energy) strictly controlled in composition: High-carbohydrate (HC, protein 12% of total energy) or high-protein
(HP, protein 25% of total energy).
SETTING AND PARTICIPANTS: Subjects were 65 healthy, overweight and obese subjects (50 women, 15 men, aged
18 ± 55 y) randomly assigned to HC (n 25), HP (n 25) or a control group (C, n 15). All food was provided by self-
selection in a shop at the department, and compliance to the diet composition was evaluated by urinary nitrogen
MAIN OUTCOME MEASURE: Change in body weight, body composition and blood lipids.
RESULTS: More than 90% completed the trial. Weight loss after six months was 5.1 kg in the HC group and 8.9 kg in
the HP group (difference 3.7 kg, 95% con®dence interval (CI)(1.3 ± 6.2 kg) P < 0.001), and fat loss was 4.3 kg and 7.6 kg,
respectively (difference 3.3 kg (1.1 ± 5.5 kg) P < 0.0001), whereas no changes occurred in the control group. More
subjects lost >10 kg in the HP group (35 %) than in the HC group (9 %). The HP diet only decreased fasting plasma
triglycerides and free fatty acids signi®cantly.
CONCLUSIONS: Replacement of some dietary carbohydrate by protein in an ad l ib itu m fat-reduced diet, improves
weight loss and increases the proportion of subjects achieving a clinically relevant weight loss. More freedom to
choose between protein-rich and complex carbohydrate-rich foods may allow obese subjects to choose more lean
meat and dairy products, and hence improve adherence to low-fat diets in weight reduction programs.
Keywords: low-fat diets; ad libitum; high-protein; high-carbohydrate; cardiovascular risk factors; blood lipids; body
The prevalence of obesity is increasing rapidly in the
Western world, and its comorbidities are of major
concern. To prevent obesity, it is recommended that
fat should be no more than 30% of the energy intake.
The background for this advice is that overconsump-
tion of high-fat foods plays a role in weight gain and
obesity in susceptible individuals.
This concept has
been used clinically to induce and maintain weight
loss in obese subjects by administration of low-fat
diets consumed ad libitum. However, there is some
debate about the ef®ciency of the low-fat ad libitum
principle, as compared to calorie counting.
reduction in energy intake can be achieved by a
reduction in dietary fat content, which can induce a
modest weight loss, but the optimal relative propor-
tion of dietary carbohydrate and protein, both in terms
of potential weight loss and of potential adverse
effects, has never been addressed in long-term inter-
A number of short-term studies suggest that protein
per kJ exerts a more powerful effect on satiety than
both carbohydrate and fat.
1,3,7 ± 15
If this is also true in
the long-term, replacing some of the dietary carbohy-
drate by protein should improve the weight loss
obtained by using low-fat diets under ad libitum
conditions. In contrast, observational studies have
found that dietary protein content is positively asso-
ciated with body fatness.
We therefore undertook
the present study to compare two ad libitum, strictly
controlled, low-fat diets, with respect to changes in
body weight, body composition and blood lipids in
obese subjects over a period of six months.
Subjects and methods
Included in the study were 65 overweight and obese
subjects (25 < body mass index (BMI) < 34 kg=m
*Correspondence: Arne Astrup, Research Department of Human
Nutrition, The Royal Veterinary and Agricultural University,
Rolighedsvej 30, 1958 Frederiksberg C, Copenhagen, Denmark.
Received 26 August 1998; revised 30 November 1998; accepted
8 January 1999
International Journal of Obesity (1999) 23, 528±536
ß 1999 Stockton Press All rights reserved 0307±0565/99 $12.00
of both genders, aged 18 ± 56 y, (Table 1). All subjects
were volunteers recruited through advertisement or a
waiting list. They underwent a brief medical screening
examination, including a medical history, a routine
physical examination and blood tests (haemoglobin,
leucocytes, sodium, potassium, glucose, alkaline
phosphates and electrocardiogram (ECG)) before
enrollment. In addition to normal screening results,
the subjects all met the criterion of being weight
stable for 2 months before entry. This was con-
®rmed by weighing at the department.
The subjects of the intervention group were ran-
domly assigned to either high-carbohydrate (HC: 25
subjects) or high-protein (HP: 25 subjects) diet, both
low in fat (30% of total energy) or to a control group
(C: 15 subjects) (Table 2). To ensure group matching
with respect to BMI, gender, age and smoking habits,
a third party, who did not know the subjects or their
identity, exchanged group membership of six subjects.
Alcohol intake, assessed by 7 d dietary records, was
equal in the three groups. The subjects in the three
groups had similar histories with respect to the course
of their obesity; of the subjects in the HC, HP and C
groups, 20%, 28% and 20%, respectively, reported
that they were overweight at school start (not statis-
tically signi®cant (NS)), and 16%, 28% and 28%
reported that they were overweight at age 18 y (NS).
Self help, public health services and alternative thera-
pies for weight reduction had all been used to a similar
degree in the three groups.
Approval was obtained from the Municipal Ethical
Committee of Copenhagen and Frederiksberg. The
study was performed in accordance with the Helsinki
II declaration, and each subject signed an informed
consent document before the study commenced.
The study was conducted as a dietary intervention for
six months, strictly controlled in terms of macronu-
trient composition. The outcome measures were
changes in body weight and composition, proportion
of subjects achieving a certain weight loss ( >5kg
and >10 kg), total and intra-abdominal fat mass and
changes in the plasma values of total and high density
lipoprotein (HDL) cholesterol, triglycerides and free
All the food for the intervention groups was provided
by a shop at the Department and could be consumed
ad libitum. The C group was instructed not to change
anything in their dietary habits whilst shopping in
ordinary food shops. The targeted composition of the
two diets was: HP: 25% of energy (E%) as protein and
45 E% as carbohydrate; HC: 12 E% as protein and 58
E% as carbohydrate (See Table 2). A variety of
different food items made up an all-round assortment
offered by the shop and this covered the most
common foods. The selection varied seasonally.
Table 2 Macronutrient composition and energy content in intervention diets
Hig h- carbohydrate Hig h- protein
0 ^ 3 months
4 ^ 6 months
Ta r g e t e d
0 ^ 3 months
4 ^ 6 months
Energy from protein (%) 12 12.1 0.1 12.2 0.1 25 24.7 0.1 24.1 0.2
Energy from carbohydrate (%) 58 59.4 0.2 59.0 0.2 45 45.9 0.2 46.8 0.2
Energy from fat (%) 30 28.5 0.2 28.8 0.2 30 29.4 0.2 29.1 0.2
Total energy (MJ/d) - 10.6 0.3** 11.2 0.5** - 8.6 0.4 9.3 0.4
Fiber content (g)
- 22.8 1.6** - 18.6 1.4
- 14.5 3.2 - 14.4 2.7
Energy density (kJ/g)
- 4.9 0.1 5.0 0.1 - 4.7 0.1 5.0 0.2
Plus-minus values are means s.e.m.
The targeted macronutrient composition according to the protocol.
Dietary composition, intake and energy density as registered by the shop computer system during the initial three months of dietary
intervention, calculated as mean daily values.
Dietary composition, intake and energy density as registered by the shop computer system during the last three months of dietary
intervention, calculated as mean daily values.
Data from 7 d dietary records.
Calculated without drinks, on the basis of computer registrations.
**P < 0.001, *P < 0.01 as compared to the corresponding value in the high-protein group.
Table 1 Physical characteristics of subjects in the two intervention groups and one control group
A ge Gender Smoking
B ody weig ht Heig ht BMI B ody f at
(y) (M/F) (kg) (cm) (kg/m
High-carbohydrate (n 25) 39.4 2.0 6/19 9 (5.7 1.5) 88.6 1.9 169.5 0.0 30.8 0.4 30.5 1.5
High-protein (n 25) 39.8 1.9 6/19 9 (8.0 2.0) 87.0 1.9 170.0 0.0 30.0 0.4 28.5 1.4
Controls (n 15) 37.6 2.2 3/12 5 (9.1 2.4) 88.1 1.8 171.0 0.0 30.3 0.7 29.6 1.8
Values are means s.e.m. There were no differences between groups by ANOVA.
Number of smokers in group (number of cigarettes smoked per day).
BMI body mass index.
Low-fat diets: High-protein
A Rosenvinge Skov
Protein sources were primarily dairy products and
meat (beef, pork, poultry, lamb, ®sh and offal).
Carbohydrate sources were primarily vegetables,
fruits, breads, rice and pasta, but chocolate and
simple sugars, in the form of sweets, were also
The subjects collected their food from the shop
twice a week. Food items could be chosen freely
within the dietary design and individual `shoppings'
were registered in a computer system designed speci-
®cally for the purpose, described in detail pre-
At each shop visit, all food items were
selected by the subject and bar code scanned by a
dietician. This made it possible to monitor achieve-
ment of the scheduled macronutrient distribution and,
if necessary, to modify the selected provisions. Unea-
ten food and left-overs, weighed to the nearest 1 g,
were taken into account in the calculation of the
energy content of the actual selection. We used bar
codes, unique for each food item, and uncoded infor-
mation about energy and macronutrient composition
of the food item. The information used was provided
by the database, Dankost
dietary assessment soft-
ware (National Food agency of Denmark, Sùborg,
Denmark) or by the food manufacturers. The calcu-
lated energy content of the food was not known by the
Subjects were instructed thoroughly in how to
prepare the foods, but they could also choose ready-
prepared dishes. The aim was to fully control the
dietary composition of the LP and HP subjects' food,
and they were encouraged to collect all their foods,
including `empty calories' and caloric beverages
(except for alcohol) from the grocery store. Any
deviation from this principle should be recorded
analogous to recording of food waste and leftovers.
For validation of the food registration method in the
shop the protein intake was monitored by an objective
biological markers, 24 h urinary nitrogen excretion
(24-h UN) each month and the completeness of the
urine sample was controlled.
Urine was collected at
baseline and at three months and six months of the
study. A questionnaire investigation was performed
after six months of dietary intervention, to asses the
impact of the dietary intervention on the quality of
The questionnaires were structured with mostly
precoded response categories and a few open ques-
tions. The subjects were asked to rate numerous life
quality variables relating to the dietary altera-
tion=intervention. These included physical, social
and psychological well-being, acceptability and
palaltability, and discomforts from different organ
systems such as nausea, constipation, frequent bowel
movements, abdominal pain, musculo-skeletal dis-
comforts and tiredness.
The subjects were instructed not to change their
physical activity pattern or smoking habits during the
study. The subjects were also allowed to leave their
alcohol habits unchanged, given that the intake was no
more than 20 g=d. This was controlled by self-report-
ing of alcohol intake at each visit in the shop.
Anthropometric measurements and body composition
Body weight was measured weekly, with subjects
wearing light clothing, on a decimal scale (Seca
model 707, Copenhagen, Denmark) in both interven-
tion groups. Subjects in the control group were only
weighed at baseline and after three and six months.
Sagittal diameter and waist and hip circumferences
were measured in all groups at baseline and at the end
of the study. Body composition was determined by a
dual energy X-ray absorptiometry (DEXA) scanning
(Hologic 1000=W, Hologic, Inc., Waltham, MA, soft-
ware version 5.61). Subjects wore only underwear and
a cotton T-shirt during the scan. For quality control,
spine phantoms were scanned daily.
Intra-abdominal adipose tissue was estimated from
DEXA-scans and anthropometry by the equation
given by Treuth et al:
Intra-abdominal fat area (cm
) 7 208.2
4.62(sagittal diameter, cm) 0.75 (age, y) 1.73
(waist, cm) 0.78 (trunk fat, %).
Venous blood samples were drawn from an antecubi-
tal vein after an overnight fast. After centrifugation,
aliquots were stored at 7 20
C, prior to analysis.
Plasma cholesterol, HDL-cholesterol and triglycerides
were determined enzymatically with a Cobas Mira-
Analyzer (Boehringer Mannheim Gmbh, Mannheim,
Germany) and plasma nonesteri®ed fatty acids
(NEFA) were determined by an enzymatic colori-
metric method using a Wako NEFA C test kit
(Wako Chemicals GmbH, Neuss, Germany).
Differences between groups in proportion of subjects
achieving a certain weight loss, that is, 5kg or
10 kg after three months and six months, respec-
tively, were tested by a chi-squared test and the
difference between intervention groups is expressed
as odds-ratio (OR). Group differences in changes in
body weight and blood lipids after 0, three months and
six months of intervention, were analyzed by a mixed
model for analysis of variance, with interaction
between `group' and `time' included as ®xed effects
and `subjects' included as random effect. One chi-
squared test was performed to test for equal baseline
levels of blood lipids in the three groups and another
chi-squared test was made to test for the effect of time
in the control group. Changes in body weight, com-
position and blood lipids are given as expected mean-
s s.e.m. (or 95% con®dence intervals (CI)), with
corresponding P-values estimated under the statistical
Low-fat diets: High-protein
A Rosenvinge Skov
model. Differences between groups in intra-abdom-
inal fat area were tested by one-way ANOVA.
P < 0.05 was considered signi®cant.
Life quality variables were tested with nonpara-
metric statistics: Chi-squared tests for differences
between groups in yes=no questions and Kruskal-
Wallis one-way analysis of variance by ranks for
group differences with respect to multiple choice
questions (four choices). To account for the multiple
comparisons the signi®cance level was set as
P < 0.01. Statistics Analysis Package, SAS
(SAS Institute, Cary, NC, USA) and SigmaStat
(Jandel Scienti®c GmbH, Erkrath, Germany) were
used in the statistical analysis.
Compliance and acceptability
Two subjects dropped out of each intervention group,
due to change of address or non-compliance, and one
subject was excluded from the control group, due to
elective surgery. A total of 60 subjects completed the
trial (92%), 23 in each intervention group and 14 in
the control group.
Table 2 shows the average daily macronutrient
intake, energy intake and energy density in the six
intervention months, separated into two three month
periods. The achievement of the targeted differences
in protein intakes in the intervention groups was
supported by the use of 24 h UN as an objective
marker of protein intake. At baseline, dietary protein
intake calculated from 24 h UN was similar in the
three groups and did not change in the control group
over the period. However, in the HP group, protein
intake increased from a baseline value of 91.4 g=d
(81.0 ± 101.82) to a six months intervention average of
107.8 g=d (102.2 ± 112.1 g=d) (P < 0.05), while corre-
spondingly, a decrease from 91.1 g=d (82.5 ± 99.7 g=d)
to 70.4 g=d (64.8 ± 76.0 g=d) (P < 0.05) was observed
in the HC group (Group difference: P < 0.0002).
Dietary ®ber intake changed from 17.8 g=d at baseline
to 18.6 g=d in the intervention period in the HP group,
whereas dietary ®ber intake correspondingly changed
in the (HC) group from 16.1 g=d to 22.8 g=d. Hence,
the increase in daily dietary ®ber content was 7 g
lower in the HP group than in the HC group
(P < 0.05). Alcohol intake at baseline was
17.7 3.3 g=d in the (HC) group, 15.0 2.2 g=din
the HP group and 11.1 3.0 g=d in the C group (NS)
and did not change during dietary intervention.
There were no signi®cant group differences in the
questionnaire responses in any of the measures of
appetite or palatability. None of the subjects in either
group responded that they most of the time felt hungry
soon after a meal or felt a bit hungry during the whole
day. Both in the HP and in the (HC) group, only 4% of
the subjects did not agree that low-fat food is as
attractive and tastes as good as `normal' food. To
assess if differences in weight loss may have been
in¯uenced by palatability and acceptability, we
analyzed the weight loss in sub-groups of both inter-
vention groups and found no evidence to support the
contention that differences in acceptability affected
weight loss. In addition, no differences were found
with respect to discomforts from different organ
systems, such as tiredness=sleeping problems,
shortness of breath, abdominal symptoms (rum-
bling=distended stomach, constipation=frequent
bowel movements, abdominal pain after meal), gen-
eral edema or discomforts in muscles or joints. The
subjects generally considered the dietary alteration to
be easier to comply with than they had expected.
Body weight and composition
Pre-treatment body weights were similar in all three
groups and no signi®cant change occurred in the
control group (Figure 1) Weight loss after three
months was greater in the HP group than in the HC
group: 7.5 kg vs 5.0 kg (difference 2.5 kg (0.6 ± 4.2 kg)
P < 0.02). After six months, weight loss was 5.0 kg
(3.6 ± 6.4 kg) in the HC group and 8.7 kg (7.3 ±
11.9 kg) in the HP group (difference 3.7 kg (1.3 ±
6.2 kg) P 0.0002). After three months of dietary
intervention, more subjects had lost 5 kg body
weight in the HP group (19=24 (79%)) than in the
HC group (12=23 (52%)) (P < 0.05) (Figure 2). After
six months, more subjects had lost 10 kg body
weight in the HP group (8=23 35%) than in the
HC group (2=23 9%) (OR 5.6 (1.1 ± 30.2)
P < 0.001). At three months, fat loss was 3.8 kg
(2.6 ± 5.0 kg) in the HC group and 5.8 kg (4.6 ±
7.0 kg) in the HP group (difference: 2.0 kg (0.4 ± 3.7)
P < 0.02). After six months, fat loss was 4.3 kg (3.1 ±
Figure 1 Changes in body weight in overweight and obese
subjects randomized to ad libitum fat-reduced diets: high-carbo-
hydrate (protein 12% of total energy; n 25), high-protein (pro-
tein 25% of total energy; n 25) or to a control group (no
intervention; n 15). Values are means s.e.m. There were no
differences in baseline values of body weight. Grouptime
interaction: P < 0.0001.
Low-fat diets: High-protein
A Rosenvinge Skov
5.5 kg) in the HC group and 7.6 kg (6.2 ± 9.0 kg) in the
HP group (difference: 3.3 kg (1.1 ± 5.7) P < 0.0001)
(Figure 3). Intra-abdominal adipose tissue decreased
by 33.0 cm
in the HP group and by 16.8 cm
HC group (P < 0.0001), whereas it increased in the
control group by 15.2 cm
, differing from both inter-
vention groups (P < 0.0001) (Figure 4).
No group differences in baseline values of blood
lipids were found, and no signi®cant changes were
seen in the control group during the six months of
intervention (Figure 5). Total cholesterol and HDL-
cholesterol decreased in both the HC and HP groups,
with no group differences (Figure 5). Plasma free fatty
acids decreased by approx 30% after six months in the
HP group, while they were unchanged in the HC
group (P < 0.05). In contrast to the increase in
plasma triglycerides after three months in the HC
group, a decrease by 0.37 mmol=l (0.15 ± 0.59
Figure 2 Proportion of subjects having lost >5 kg or 10 kg body
weight after three months and six months of dietary interven-
tion. Comparisons between groups were made by a chi-squared
test. *P < 0.05 for the comparison of difference with the high-
Figure 3 Changes from baseline in body fat mass in overweight
and obese subjects randomized to two ad libitum fat-reduced
diets, either a high-carbohydrate (protein 12% of total energy;
n 25), or high-protein (protein 25% of total energy; n 25) or a
control group (no intervention; n 15). *P < 0.02 for the compar-
ison between the two intervention groups. **P < 0.0001 for the
comparison between the two intervention groups. Values are
Figure 4 Changes in intra-abdominal adipose tissue (IAAT)
estimated from dual energy X-ray absorptiometry (DEXA)
scans and anthropometry by the equation given by Treuth et
: IAAT (cm
) 7 208.2 4.62(sagittal diameter, cm) 0.75
(age, y) 1.73 (waist, cm) 0.78 (trunk fat, %). * P < 0.0001 for
the comparison of changes between high-protein group and the
two others. Values are means s.e.m.
Figure 5 Changes in blood lipids from base-line values in
overweight and obese subjects randomized to ad libitum fat-
reduced diets: high-carbohydrate (HC: protein 12% of total
energy; n 25), high-protein (HP: protein 25% of total energy;
n 25) or to a control group (C: no intervention; n 15). *P < 0.05
and **P < 0.01 for the comparison of change from baseline and
for difference between intervention groups and control group. P
HDL high density lipoproteins.
Low-fat diets: High-protein
A Rosenvinge Skov
mmol=l) was found in the HP group (P 0.001). After
six months, no signi®cant group differences remained
The present study shows that two diets with a dietary
fat content reduced to slightly below 30 E%, cause
clinically relevant weight losses during ad libitum
consumption, compared to a control diet with a fat
content of about 40 E%. This study further shows that
the HP diet induces a larger weight loss than the HC
diet. After six months intervention, the HP diet
induced a 3.7 kg (1.3 ± 6.2 kg) larger weight loss,
which was mainly due to a reduction in body fat
mass. Moreover, in the HP group, 35% of the subjects
lost >10 kg, whereas only 9% in the HC group
achieved this goal (OR 5.6 (1.1 ± 30.2)). Both fat-
reduced diets decreased the intra-abdominal fat stores,
but the decrease in the HP group was two-fold greater
than in the HC group.
The weight loss on the two ad libitum fat-reduced
diets was markedly higher than those previously
reported in normal weight subjects
2,19 ± 21
above the weight losses reported in overweight and
There are two likely reasons for this.
Firstly, the small weight losses observed in some of
the low-fat intervention trials can partly be attributed
to low adherence to the low-fat diet composition.
Most trials using the ad libitum low-fat principle
reported so far have instructed the subjects how to
make the dietary changes, but have not ensured that
the subjects actually consumed a diet with the pre-
scribed composition. Adherence to the diet, as
assessed by recovered label in expired air in subjects
consuming meals enriched with
free-living conditions, has been shown to be positively
related to weight loss.
In contrast, our shop system,
where all foods during six months were free of charge,
allowed a more strict control of macronutrient com-
position, while allowing the subjects freedom to select
appropriate food items in the shop. Thus the compli-
ance to the two diets as assessed by UN excretion was
high and 92% of the subjects completed the 6 months
We ®nd it very likely that the provision of free
food during the intervention trial played a role and
that the high compliance was also economically
motivated. However, in a recent study, obese sub-
jects participated in four different behavioural weight
control programs that differed only with respect to
the way the food was provided to the subjects: no
food provision, meal plans, provision of food (paid
for by the subjects) or food provided free.
losses were similar in the three latter groups, but
signi®cantly different from that of the group that
received the behavioural program alone. Therefore, it
is unlikely that the provision of free food enhanced
Dietary composition was not monitored as closely
in the C-group as in the intervention groups, since
food was not provided to the control group from the
study shop. However, as we chose experimental
conditions that were very natural, we do not consider
this will invalidate the status of the C-group as a
The adherence to the dietary compositions of the
two intervention groups was high, as assessed by the
excretion of 24 h UN, which was used as a marker of
The agreement between the dietary
protein intake, as estimated by the shop computer and
the UN excretion was very high (r 0.84,
P < 0.0001), and the achievement of a two-fold dif-
ference in UN excretion between the HP and HC
groups supports that statement that the targeted
macronutrient compositions of the intervention diets
were actually reached.
We ®nd it very likely that some unintended volun-
tary energy restriction occurred in both intervention
groups, due to the subjects being highly motivated to
lose weight. This may have enhanced the weight loss
in the two intervention groups, but is unlikely to have
in¯uenced the weight loss difference between the HP
and HC groups.
The mechanisms responsible for the larger weight
loss caused by the HP diet than by the HC diet might
be due to both a reduced energy intake and a greater
thermogenic effect of protein.
We found the reported energy intake during the
intervention was lower in the HP group than in the
HC group by 2 MJ (0.94 ± 3.05 MJ, P < 0.001), which is
more than suf®cient to explain the larger weight loss in
the HP group. Rolls et al
found that high protein and
high starch foods produced greater satiety than high fat,
high sucrose or mixed content foods. The lower energy
intake in the HP group is in accordance with most meal
test studies, showing a higher satiating effect of protein
than carbohydrate, when compared joule for
A high protein intake also seems to be
able suppress the following day's energy intake more
than an isoenergetic amount of carbohydrate. Stubbs et
studied the relationship between carbohydrate and
protein balances and the next day's spontaneous energy
intake during a seven-day stay in a respiration chamber,
and found that every megajoule of increased protein
stores on day 1 produced a reduction in energy intake on
the subsequent day amounting to 2.1 MJ. For a similar
increase in carbohydrate stores, the reduction in energy
intake was only 0.4 MJ. Thus the more pronounced
effect of protein than of carbohydrate in inhibition of
energy intake found in short-term studies is con®rmed
by the present study and shown to be maintained for at
least six months.
Palatability of the diet has been shown to be an
important determinant of energy intake,
lower energy intake in the HP group than in the HC
group could therefore have been due to a lower
Low-fat diets: High-protein
A Rosenvinge Skov
palatability of the HP diet. However, we ®nd this
explanation unlikely since no differences were found
in palatability between the intervention groups after
six months. Moreover we found that in the question-
naire on appetite the subjects' responses to questions
were independent of their achieved weight loss.
Moreover, no differences with respect to physical
well-being were found. The subjects generally consid-
ered the dietary alteration to be easier to comply with
than they had expected (Holm L, SKov AR, Astrup A.
unpublished results). The results therefore suggest that
the lower energy intake in the HP diet was due to a
higher satiating effect of protein than of carbohydrate.
In addition to the effect on energy intake, the HP
diet may increase energy expenditure more than the
HC diet, as the post-prandial thermogenesis of protein
amounts to 30% of its energy content, whereas that of
carbohydrate is only 4 ± 8%.
On a daily basis, the
difference in protein and carbohydrate intakes
between the HP group and the HC group can be
estimated to produce a difference of about 300 kJ=d,
which is only about 15% of the observed difference in
energy balance. The greater weight loss caused by the
HP diet than the HC diet can therefore mainly be
attributed to a reduction in energy intake.
The mechanisms responsible for the high satiating
effect of protein are not known. The energy density of
foods is an important determinant of spontaneous
energy intake and seems to be responsible for the
higher energy intakes observed on high-fat than on
However, differences in energy den-
sity are unlikely to be involved because the HP diet
and the HC diet had similar energy densities of 4.7 ±
5.0 kJ=g. This is in agreement with the ®nding that a
high-protein meal suppressed hunger to a greater
extent than two isoenergetic high-fat and high-carbo-
hydrate meals with the same energy density.
change in daily dietary ®ber content was expectedly
lower in the HP group than in HC group, by 7 g. This
difference cannot explain the larger weight loss in the
HP group, as the higher ®ber intake would rather have
contributed to a larger weight loss in the HC group.
Possible differences in fat quality may have played a
role, but there are no published human data to support
that differences in fat types in¯uence the satiating
effect of the diet. Consequently, the inhibition of
energy intake caused by the HP diet may be due to
mechanisms other than the energy density, for exam-
ple, release of gut peptides, liver metabolism and a
direct central effect of certain amino acids.
Obesity is an important risk factor of cardiovascular
disease (CVD) and abdominal obesity in particular is
strongly associated with an adverse lipid pro®le,
ischaemic heart disease, stroke and premature
Overwhelming epidemiological data have
demonstrated a close association between obesity
and coronary heart disease (CHD) mortality,
which is attributed partly to its effects on plasma
lipid metabolism. The dyslipidaemic pro®le asso-
ciated with fatness and especially with excessive
intra-abdominal fat deposition is characterised by
increased total cholesterol, low density lipoprotein
(LDL)-cholesterol, triglyceride and free fatty acid
levels, and decreased HDL levels.
larger reduction in the intra-abdominal fat depots in
the HP group, may be expected to reduce the risk of
these comorbidities. Although a bene®cial effect of
weight loss on plasma lipids was found in both
intervention groups, more favourable improvements
were seen in the HP group (Figure 5). There was a
slight transient increase in plasma triglycerides after
three months in the HC group, whereas a reduction
was seen in the HP group. The increase in plasma
triglycerides has been reported to occur on isoener-
getic low-fat, high carbohydrate diets,
, but not
under ad libitum conditions where weight loss is
allowed to occur.
Moreover, plasma NEFA were
reduced only in the HP group. The greater improve-
ment in that cardiovascular risk pro®le after six
months on the HP diet may be due to a combination
of the greater fat loss, reduction in intra-abdominal fat
and to the diet composition per se. Intervention
studies comparing isoenergetic low-fat diets with
either high or low ratios of protein to carbohydrate
have demonstrated that, without changes in body
weight, the exchange of protein for carbohydrate
reduced LDL-cholesterol and triglycerides, and
increased HDL-cholesterol in hypercholesterolaemic
The more favourable effects of the HP
diet may be only partially attributable to the larger
reduction in body fat.
We did not measure blood pressure in the present
study, but it is unlikely that dietary protein increases
Furthermore, weight loss has con-
sistently been associated with clinically relevant
reduction in both systolic and diastolic blood pres-
A recent intervention study on moderately
hypertensive patients demonstrated that a fat-reduced
diet, rich in fruits and vegetables and low-fat dairy
products, providing 18% of energy from protein,
reduced systolic and diastolic blood pressure by
5.5 mm Hg and 3.0 mm Hg more than a control
This intervention resulted in a weight loss of
< 0.5 kg, so there is no reason to believe that an
increase in dietary protein can offset the bene®cial
effect of the weight loss on blood pressure in obese
A protein-rich diet may have other health implica-
tions and its effects on osteoporosis, kidney function
and colonic cancers are still a matter for debate. We
failed to detect any detrimental effect of the HP diet
on bone mineral density and kidney size and glomer-
ular function (data not shown), but more studies are
needed to elucidate the contribution of high-protein
diets to the development of these disorders. The use of
fat-reduced, high-protein diets in the treatment of
obesity seems justi®ed because the health bene®ts of
a weight loss of the magnitude observed in the present
study is associated with a marked improvement in risk
factors for non-insulin dependent diabetes and CVD,
Low-fat diets: High-protein
A Rosenvinge Skov
and possibly with a reduction in mortality.
uncertainty about possible adverse effects means that
the bene®cial effects observed in this treatment pro-
gram for obesity cannot yet be extrapolated to the
recommendation of a high-protein diet to the general
The study shows that replacement of some dietary
carbohydrate by protein in ad libitum fat-reduced
diets, for treatment of obesity, improves mean
weight loss and increases the proportion of subjects
achieving a clinically relevant weight loss. Slight
improvements in blood lipids were also observed.
More freedom to choose between protein-rich and
complex carbohydrate-rich foods may allow obese
subjects to eat more lean meat and dairy products
and hence improve adherence to low-fat diets during
weight reduction programs.
The study was supported by The Danish Research and
Development Programme for Food Technology, The
Federation of Danish Pig Producers and Slaughter-
houses, Danish Dairy Research Foundation and The
Danish Livestock and Meat Board. We also thank the
staff of Energy Metabolism and Obesity Group.
Finally, we thank several food producers for kindly
contributing to the food selection.
1 Rolls BJ. Carbohydrates, fats, and satiety. Am J Clin Nutr
1995; 61: 960S ± 967S.
2 Lissner L, Levitsky DA, Strupp BJ, Kalkwarf HJ, Roe DA.
Dietary fat and the regulation of energy intake in human
subjects. Am J Clin Nutr 1987; 47: 886 ± 892.
3 Rolls BJ, Kim Harris S, Fischman 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.
4 Schlundt DG, Hill JO, Pope-Cordle J, Arnold D, Virts KL,
Katahn M. Randomized evaluation of a low fat ad libitum
carbohydrate diet for weight reduction. Int J Obes 1993; 17:
623 ± 629.
5 Shah M, McGovern P, French S, Baxter J. Comparison of a
low-fat, ad libitum complex-carbohydrate diet with a low-
energy diet in moderately obese women. Am J Clin Nutr 1994;
59: 980 ± 984.
6 Toubro S, Astrup A. Randomised comparison of diets for
maintaining obese subjects' weight after major weight loss: ad
lib, low fat, high carbohydrate v ®xed energy intake. BMJ
1997; 314: 17 ± 22.
7 Hill AJ, Blundell JE. Macronutrients and satiety: the effects of
a high-protein or high carbohydrate meal on subjective
motivation to eat and food preferences. Nutr Behav 1986; 3:
133 ± 144.
8 Hill AJ, Blundell JE. Comparison of the action of macronu-
trients on the expression of appetite in lean and obese human
subjects. Ann NY Acad Sci 1990; 580: 529 ± 531.
9 Barkeling B, Rossner S, Bjorvell H. Effects of a high-protein
meal (meat) and a high-carbohydrate meal (vegetarian) on
satiety measured by automated computerized monitoring of
subsequent food intake, motivation to eat and food prefer-
ences. Int J Obes 1990; 14: 743 ± 751.
10 Booth DA, Chase A, Campbell AT. Relative effectiveness of
protein in the late stages of appetite suppression in man.
Physiol Behav 1970; 5: 1299 ± 1302.
11 Astrup A, Raben A. Glucostatic control of intake and obesity.
Proc Nutr Soc 1996; 55, 485 ± 495.
12 Rolls BJ, Hetherington M, Burley VJ. The speci®city of
satiety: the in¯uence of different macronutrient contents on
the development of satiety. Physiol Behav 1988; 43: 145 ± 153.
13 Stubbs RJ. Macronutrient effects on appetite. Int J Obes 1995;
19 (Suppl 5): S11 ± S19.
14 Hannah JS, Dubey AK, Hansen BC. Postingestional effects of
a high-protein diet on the regulation of food intake in mon-
keys. Am J Clin Nutr 1990; 52: 320 ± 325.
15 Stubbs RJ, van Wyk MCW, Johnstone AM, Harbron CG.
Breakfasts high in protein, fat or carbohydrate: effect on
within-day appetite and energy balance. Eur J Clin Nutr
1996; 50: 409 ± 417.
16 Buemann B, Bouchard C, Tremblay A. Social class interacts
with the association between macronutrient intake and sub-
cutaneous fat. Int J Obes 1995; 19: 770 ± 775.
17 Skov AR, Toubro S, Raben A, Astrup A. A method to achieve
control of dietary macronutrient composition in ad libitum
diets consumed by free-living subjects. Eur J Clin Nutr 1997;
51: 667 ± 672.
18 Treuth MS, Hunter GR, Kekes-Szabo T. Estimating intraab-
dominal adipose tissue in women by dual-energy X-ray
absorptiometry. Am J Clin Nutr 1995; 62: 527 ± 532.
19 Kendall A, Levitsky DA, Strupp BJ, Lissner L. Weight loss on
a low-fat diet: consequence of the imprecision of the control
of food intake in humans. Am J Clin Nutr 1991; 53: 1124 ±
20 Raben A, Jensen NJ, Marckmann P, Sandstro
m B, Astrup A.
Spontaneous weight loss during 11 weeks' ad libitum intake of
a low fat=high ®ber diet in young, normal weight subjects. Int
J Obes 1995; 19: 916 ± 923.
21 Siggaard R, Raben A, Astrup A. Weight loss during 12 weeks'
ad libitum carbohydrate-rich diet in overweight and normal-
weight subjects at a Danish working site. Obes Res 1996; 4:
347 ± 356.
22 Schaefer EJ, Lichtenstein AH, Lamon Fava S, McNamara JR,
Schaefer MM, Rasmussen H, Ordovas JM. Body weight
and low-density lipoprotein cholesterol changes after
consumption of a low-fat ad libitum diet. JAMA 1995; 274:
1450 ± 1455.
23 Lyon X, Di Vetta V, Milon H, Je
quier E, Schutz Y. Com-
pliance to dietary advice directed towards increasing the
carbohydrate to fat ratio of the everyday diet. Int J Obes
1995; 19: 260 ± 269.
24 Wing RR, Jeffery RW, Burton LR, Thorson C, Sperber
Nissinoff K, Baxter JE. Food provision vs structured meal
plans in the behavioral treatment of obesity. Int J Obes 1996;
20: 56 ± 62.
25 Isaksson B. Urinary nitrogen output as a validity test in dietary
surveys. Am J Clin Nutr 1980; 33:4±5.
26 Teff KL, Young SN, Blundell JE. The effect of protein or
carbohydrate breakfasts on subsequent plasma amino acid
levels, satiety and nutrient selection in normal males. Phar-
macol Biochem Behav 1989; 34: 829 ± 837.
27 Stubbs RJ, Ritz P, Coward WA, Prentice AM. Covert
manipulation of the ratio of dietary fat to carbohydrate and
energy density: effect on food intake and energy balance in
free-living men eating ad libitum. Am J Clin Nutr 1995; 62:
330 ± 337.
28 Saltzman E, Dallal GE, Roberts SB. Effect of high-fat and
low-fat diets on voluntary energy intake and substrate oxida-
tion: studies in identical twins consuming diets matched for
energy density, ®ber and palatability. Am J Clin Nutr 1997; 66:
Low-fat diets: High-protein
A Rosenvinge Skov
1332 ± 1339.
29 Robinson SM, Jaccard C, Persaud C, Jackson AA, Jequier E,
Schutz Y. Protein turnover and thermogenesis in response to
high-protein and high-carbohydrate feeding in men. Am J Clin
Nutr 1990; 52: 72 ± 80.
30 Karst H, Steiniger J, Noack R, Steglich HD. Diet-induced
thermogenesis in man: thermic effects of single proteins,
carbohydrates and fats depending on their energy amount.
Ann Nutr Metab 1984; 28: 245 ± 252.
31 Peters JC, Harper AE. Acute effects of dietary protein on food
intake, tissue amino acids and brain serotonin. Am J Physiol
1987; 252: R902 ± R914.
32 Bray GA. Obesity and the heart. Mod Concepts Cardiovasc
Dis 1987; 56: 67 ± 71.
33 Must A, Jacques PF, Dallal GE, Bajema CJ, Dietz WH. Long-
term morbidity and mortality of overweight adolescents. A
follow-up of the Harvard Growth Study of 1922 to 1935. New
Engl J Med 1992; 327: 1350 ± 1355.
34 Manson JE, Colditz GA, Stampfer MJ, Willett WC, Rosner B,
Monson RR, Speizer FE, Hennekens CH. A prospective study
of obesity and risk of coronary heart disease in women. New
Engl J Med 1990; 322: 882 ± 889.
35 Lamon-Fava S, Wilson PW, Schaefer EJ. Impact of body mass
index on coronary heart disease risk factors in men and
women. The Framingham Offspring Study. Arterioscler
Thromb Vasc Biol 1996; 16: 1509 ± 1515.
36 Jeppesen J, Schaaf P, Jones C, Zhou M, Chen Y, Reaven GM.
Effects of low-fat, high-carbohydrate diets on risk factors for
ischemic heart disease in postmenopausal women. Am J Clin
Nutr 1997; 65: 1027 ± 1033.
37 Wolfe BM. Potential role of raising dietary protein intake for
reducing risk of atherosclerosis. Can J Cardiol 1995; 11
(Suppl G): 127G ± 131G.
38 Wolfe BM, Giovanetti PM. Short-term effects of sub-
stituting protein for carbohydrate in the diets of moderately
hypercholesterolemic human subjects. Metabolism 1991; 40:
338 ± 343.
39 Obarzanek E, Velletri PA, Cutler JA. Dietary protein and
blood pressure. JAMA 1996; 275: 1598 ± 1603.
40 Goldstein DJ. Bene®cial health effects of modest weight loss.
Int J Obes 1992; 16: 397 ± 415.
41 Appel LJ, Moore TJ, Obarzanek E, et al. A clinical trial of the
effects of dietary patterns on blood pressure. New Engl J Med
1997; 336: 1117 ± 1124.
42 Williamson DF, Pamuk ER, Flanders TM, Byers TE, Heath C.
Prospective study of intentional weight loss and mortality in
never-smoking overweight US white women aged 40 ± 64
years. Am J Epidemiol 1995; 141: 1128 ± 1141.
Low-fat diets: High-protein
A Rosenvinge Skov