Content uploaded by Marie-Pierre St-Onge
Author content
All content in this area was uploaded by Marie-Pierre St-Onge on Jan 18, 2016
Content may be subject to copyright.
The effectiveness of breakfast recommendations on weight loss:
a randomized controlled trial
1–3
Emily J Dhurandhar, John Dawson, Amy Alcorn, Lesli H Larsen, Elizabeth A Thomas, Michelle Cardel, Ashley C Bourland,
Arne Astrup, Marie-Pierre St-Onge, James O Hill, Caroline M Apovian, James M Shikany, and David B Allison
ABSTRACT
Background: Breakfast is associated with lower body weight in
observational studies. Public health authorities commonly recom-
mend breakfast consumption to reduce obesity, but the effectiveness
of adopting these recommendations for reducing body weight is
unknown.
Objective: We tested the relative effectiveness of a recommendation
to eat or skip breakfast on weight loss in adults trying to lose weight
in a free-living setting.
Design: We conducted a multisite, 16-wk, 3-parallel-arm randomized
controlled trial in otherwise healthy overweight and obese adults
[body mass index (in kg/m
2
) between 25 and 40] aged 20–65 y.
Our primary outcome was weight change. We compared weight
change in a control group with weight loss in experimental groups
told to eat breakfast or to skip breakfast [no breakfast (NB)]. Ran-
domization was stratified by prerandomization breakfast eating
habits. A total of 309 participants were randomly assigned.
Results: A total of 283 of the 309 participants who were randomly
assigned completed the intervention. Treatment assignment did not
have a significant effect on weight loss, and there was no interaction
between initial breakfast eating status and treatment. Among skippers,
mean (6SD) baseline weight-, age-, sex-, site-, and race-adjusted
weight changes were 20.71 6 1.16, 20.76 6 1.26, and 20.61 6
1.18 kg for the control, breakfast, and NB groups, respectively. Among
breakfast consumers, mean (6SD) baseline weight-, age-, sex-, site-,
and race-adjusted weight changes were 20.53 6 1.16, 20.59 6 1.06,
and 20.71 6 1.17 kg for the control, breakfast, and NB groups, re-
specti vely. Self-reported compliance with the recommendation was
93.6% for the breakfast group and 92.4% for the NB group.
Conclusions: A recommendation to eat or skip breakfast for weight
loss was effective at changing self-reported breakfast eating habits,
but contrary to widely espoused views this had no discernable effect
on weight loss in free-living adults who were attempting to lose weight.
This trial was registered at clinicaltrails.gov as NCT01781780.
Am J Clin Nutr doi: 10.3945/ajcn.114.089573.
INTRODUCTION
A common public health message from reputable sources is that
eating breakfast is important to achieve and maintain a healthy weight
(1). Observ ational e vidence suggests that there is an association of
breakfast with body weight and weight loss. Howev er , there is little
causal evidence to support this conjecture. The need for evidence to
determine whether there is a causal effect of regular breakfast con-
sumption on weight loss was recently emphasized (2, 3).
The association of breakfast consumption with lo wer body
weight is well established (4–8); howe v er, observational evidence
does not preclude the possibility that breakfast eaters tend to
weigh less because of other weight-related factors associated with
breakfast eating. Short-term studies highlight potential physiologic
mechanisms by which breakfast may influence appetite, energy
expenditure, fat oxidation, and body weight (9–12). Nevertheless,
whether the proposed physiologic mechanisms translate to long-
term ef fects on energy intake and body weight is unclear.
Some hypotheses with regard to breakfast consumption and
lower body weight are based on the conjecture that breakfast
consumption is important for the regulation of energy intake.
Some studies suggest that skipping breakfast results in higher
energy intake at lunch compared with when breakfast is con-
sumed (9, 13–15). Others suggest that skipping breakfast may
not be compensated for through increasing energy intake later in
the day, resulting in net negative energy balance relative to when
breakfast is consumed (16, 17). Energy balance may be main-
tained through successive compensations over several days (18),
but long-term studies of the influence of breakfast consumption
1
From the Department of Health Behavior, School of Public Health
(EJD); the Office of Energetics, Nutrition Obesity Research Center, Depart-
ment of Nutrition Sciences, School of Public Health and School of Health
Professions (JD, A Alcorn, and DBA); and the Division of Preventative
Medicine, School of Medicine (JMS), University of Alabama at Birming-
ham, Birmingham, AL; The OPUS Centre, Department of Nutrition, Exer-
cise and Sports, University of Copenhagen, Copenhagen, Denmark (LHL
and A Astrup); the Department of Endocrinology, Metabolism, and Diabetes
(EAT) and the Department of Pediatrics, Anshutz Medical Campus (MC and
JOH), University of Colorado Denver, Denver, CO; the Nutrition and Weight
Management Center, Boston Medical Center, Boston, MA (ACB and CMA);
and the New York Obesity Research Center, Columbia University, New
York, NY (M-PS-O).
2
Supported by awards P30DK056336, P30DK046200, DK26687,
T32DK007446, and T32DK062710 from the National Institute of Diabetes
and Digestive and Kidney Diseases; award T32HL072757 from the National
Heart, Lung, and Blood Institute; Clinical and Translational Research Award
grant UL1TR000040; and The Nordea Foundation, a foundation in Denmark
giving health-related grants to promote public well-being. No funder was
involved in the design, analysis, or interpretation of this study.
3
Address correspondence to DB Allison, 1665 University Boulevard,
RPHB 140J, Birmingham, AL 35294-0022. E-mail: dallison@uab.edu; or
EJ Dhurandhar, 1665 University Boulevard, RPHB 227J, Birmingham, AL
35294-0022. E-mail: edhurand@uab.edu.
Received April 4, 2014. Accepted for publication May 13, 2014.
doi: 10.3945/ajcn.114.089573.
Am J Clin Nutr doi: 10.3945/ajcn.114.089573. Printed in USA. Ó 2014 American Society for Nutrition 1of7
AJCN. First published ahead of print June 4, 2014 as doi: 10.3945/ajcn.114.089573.
Copyright (C) 2014 by the American Society for Nutrition
compared with prolonged fasting on energy intake under free-
living, ad libitum conditions have not been conducted. There-
fore, the long-term impact of breakfast on energy intake and
weight loss is not clear.
An extensive review (3) showed only one randomized con-
trolled trial (RCT) pertinent to the effects of breakfast eating
compared with breakfast skipping on weight in a nonmalnourished
industrialized population. This RCT tested the impact of eating or
skipping breakfast on weight loss (19). Their findings suggested
that the effect of breakfast on weight loss might depend on
breakfast eating habits before the study, such that switching
breakfast eating habits as a result of study assignment increased
weight loss. However, the study was not well powered to detect
this interaction, and the result was only suggestive of significance
(P = 0.06).
Herein, we conducted an RCT to determine whether advising
good nutrition habits is more effective at producing weight loss if
paired with advice to skip or eat breakfast. We tested the effect of
breakfast recommendations on weight loss in free-living adults
who were attempting to lose weight, because this population is
likely affected by public health breakfast recommendations. On
the basis of previous findings (19), we stratified randomization by
baseline breakfast eating habits and hypothesized that individuals
who were advised to switch their usual breakfast eating habits due
to their group assignment would lose more weight than the
control group.
SUBJECTS AND METHODS
Study design
We conducted a multisite, 3-parallel-arm RCT. Study sites
included the University of Alabama at Birmingham, University of
Copenhagen (Denmark), Boston Medical Center, Columbia
University, and the University of Colorado, Denver. All sites
conducted the study in a clinical research setting. After screening,
participants were randomly assigned to 1 of 3 groups [control,
breakfast, or no breakfast (NB)], and random ization was strat-
ified by typical bre akfast eating habits at baseline. The study
duration was 16 wk, and the primary outcome measure was
weight chang e from baseline. Weight was measured at baseline
and after 16 wk, and additional contacts included phone calls at
weeks 4, 8, and 12. A ll study visits occurred between January
2013 and January 2014. The tr ial was stopped when recruitment
goals were met and all participants had completed the 16-wk
duration. The study was registered at clinicaltrials.gov (registry
no. NCT01781780).
Study sample
Inclusion criteria were as follo ws: aged 20–65 y, BMI (in kg/m
2
)
.25 but ,45, interested in weight loss, and beginning the day
by 0900 h at least 5 d/wk. Exclusion criteria were participation
in a weight-reduction program i n the past 3 mo, weight loss or
gain $5% of body weight in the past 6 mo, t aking medication
that affects appet ite (weight-loss drugs o r antidepressant, ste-
roid, or thyroid medication, unless d osage has been sta ble for
at least 6 mo), taking medication that re quires eating food in
the morning, a history of prior surgical proce dure for weight
control, current smoker or had smoked ,6 mo before the
start of the study, any major disease, a score on the Brief
Symptom Inventory that exceeded the 90th percentile (20) or
a T-score $63 on the global severity index or on 2 of the 9
symptom scales on the Symptom Checklist 90-R (21), current
eating disorder [a 26-item Eating Attitudes Test (22) score .20],
a recent or ongoing problem with drug a buse or addiction,
excessive alcohol intake, night shift work, pregnancy, or
breastfeeding.
Participants were recruited from cities of participating sites
through e-mail, campus newspaper advertisements, flyers, and
word of mouth. Participants were screened over the phone for
initial eligibility criteria and asked to come into the clinic for
a second level of screening. After informed consent, weight and
height, a pregnancy test for all women, the Brief Symptom
Inventory–18 (20), or Symptom Checklist 90-R21 (Copenhagen
site only), and the Eating Attitudes Test (22) were all adminis-
tered at a second screening to determine final eligibility. If
participants passed the second level of screening and were in-
terested, they underwent random assignment. The institutional
review board, or ethical committee, of each participating in-
stitution approved the protocol, and all participants provided
written informed consent.
Intervention
Participants enrolled in the study were randomly assigned to 1
of 3 groups at the baseline visit. The control group received
a USDA pamphlet “Let’s Eat for the Health of It,” describing
general good nutrition habits (with no mention of breakfast)
(23), along with a handout emphasizing the main points of the
pamphlet. The breakfast group received the USDA pamphlet
with a handout instructing participants to consume breakfast
before 1000 h every day. The breakfast handout also provided
suggestions of food items that might constitute a healthy
breakfast; however, no specific restrictions were given on types
of foods that could be consumed for the breakfast meal. The NB
group received the USDA pamphlet with a handout instructing
participants not to consume any calories before 1100 h every
day, and that only water or zero-calorie beverages could be
consumed from the time of waking until 1100 h.
The study coordinators reviewed the USDA pamphlet and the
appropriate handout documents with participants in detail by
reading them aloud and answering any questions. It was rec-
ommended that participants incorporate the suggestions as best
they could into their daily life. Follow-up phone calls occurred at
weeks 4, 8, and 12 to enhance participant compliance and at-
trition. During these calls, participants were asked if they were
following the recommendations, keeping their compliance diary
(breakfast and NB groups), and if they were experiencing any
barriers to incorporating the recommendations into their diet.
Weight and height were measured at baseline, and body weight
was measured again at the end of the trial period. Participants
were not blinded to their treatment condition; however, all study
staff who measured final weights were blinded to participants’
group assignment.
Randomization
Participants who ate breakfast regularly ($4 times/wk) were
randomly assigned separately from those who did not frequently
2of7 DHURANDHAR ET AL
eat breakfast (#3 times/wk); hereafte r, this dis tinction wil l
be referred to as “ prerandomization status,” with participants
classified as either breakfast “eaters” or “skippers.” Participants
were randomly assigned to control, breakfast, or NB in a manner
enforcing equal (1:1:1) allocation, specifically a single block
size of n = 27, with 9 control, 9 breakfast, and 9 NB assign-
ments. This was done within each site and prerandomization
status classification. Because of implementation errors or sites
not meeting recruitment goals, perfect balance was achieved at
only one site; however, these issues did not cause treatment
assignments to be unduly unbalanced across sites or pre-
randomization statuses (see Figure 1). The assigned treatment
groups were specified on cards contained in sequentially num-
bered, opaque, sealed envelopes that were prepared by the Office
of Energetics at the University of Alabama at Birmingham.
Sealed envelopes were sent to partic ipating sites, and study
coordinators enrolled participants and opened the next consec-
utively numbered envelope in the appropriate site-by-initial-
status series in the presence of the participant. Randomization
was performed with R, version 2.15.3 (24), by using a pseudo-
random number generator with an arbitrary but fixed seed.
Measures
Participants’ height was measured at baseline without shoes
by using a wall-mounted stadiometer. Weight was measured at
baseline and at the final study visit after 16 wk by using elec-
tronic scales to the nearest 0.1 kg with participants wearing light
clothing without shoes. Weight change from baseline to follow-
up was the primary outcome of interest. BMI was calculated as
kilograms divided by meters squared by using measured weight
and height for the purpose of screening. Compliance with the
recommendations in the breakfast and NB groups was tracked
by using a self-report diary that asked participants to circle
“yes” or “no” for each day they were enrolled in the study to
indicate if they had complied with the breakfast recommenda-
tion. Compliance with the intervention was calculated as the
percentage of days participants complied with the breakfast
recommendation, and this was a secondary outcome.
Statistical analyses
All analyses were performed by using R (version 2.15.3) (24).
An intent-to-treat analysis of change in body weight over the
16-wk intervention period (n = 309) was used to determine the
primary outcome of this RCT. Multiple imputation was used to
account for missing data. A completers-only analysis was also
conducted (n = 283). Ordinary least-squares linear regression
models, adjusted for baseline measurements and including ini-
tial weight as a covariate, were used to compare participants
across assignment groups. Means and SDs were calculated on
the basis of data at the baseline and week 16 time points, as well
as for changes in the outcome measures between these 2 time
points.
As a secondary analysis, data from the diaries of completers
were used to assess the impact of compliance on the breakfast and
NB arms, to perform an as-treated analysis (n = 185). Participants
in the control group did not complete diaries, because they were
not given specific instructions with respect to breakfast con-
sumption. For the breakfast and NB participants, their group
indicator was multiplied by the proportion of days in compliance
and treated as a continuous covariate. During an exit interview
with participants, it was determined that circled diary days were
sometimes used by participants to indicate breakfast had been
eaten on that day, rather than compliance with their group in-
structions, as was instructed (see Supplemental Figure 1, A and
B, under “Supplemental data” in the online issue). Therefore,
because it was recorded for most participants during their exit
interview if they filled out the diary according to breakfast
consumption, or compliance with instructions, compliance pro-
portions were discerned through participant intent when filling
out the diary as best as possible (see Supplemental Figure 1, C
and D, under “Supplemental data” in the online issue). Because
this information was not available for every participant and our
FIGURE 1. Flow of participants through enrollment, allocation, and follow-up: CONSORT diagram. “Eat” indicates participants who were breakfast eaters
before the study ($4 times/wk); “Skip” indicates participants who were breakfast skippers before the study (#3 times/wk). B, breakfast group; C, control
group; CONSORT, Consolidated Standards of Reporting Trials; Incl/Excl, inclusion/exclusion; LTFU, lost to follow-up; NB, no-breakfast group.
THE EFFECT OF BREAKFAST RECOMMENDATIONS ON WEIGHT 3of7
efforts did not completely ameliorate the compliance question
(note the tails bounded by the dashed boxes in panels C and D),
we performed the secondary analysis twice, once with compli-
ance as given in panels C and D and once with the compliance
“yes” and “no” counts swapped for boxed individuals. In cases
in which compliance proportion could not be ascertained, either
due to missing diaries or an inability to discern intent, this
“missingness” was handled by multiple imputation.
As a tertiary analysis, we examined whether or not participants
randomly assigned to the control group changed their self-
reported breakfast eating habits, with specific interest in pre-
randomization status. Logistic regression was used to perform
this analysis of completers (n = 105).
To accommodate missing values in the participants who did
not complete the week 16 follow-up or who were missing
compliance information, multiple imputation was performed. In
the imputation process, a total of 10 imputed data sets were
created and analyzed. For this trial, given the very small amount
of missing data (8.4% of weight-change data and 12% of com-
pliance data), 10 imputations were sufficient for imputation (25).
The software package used to implement the multiple imputation
and pooling procedure was Multivariate Imputation by Chained
Equations library, version 2.18 (26), found in R, version 2.15.3
(24). As a sensitivity analysis, we also conducted the same overall
modeling with the use of completers only. Tests of significance
were conducted at the 0.05 2-tailed significance level.
The study sample was calculated such that we had at least 90%
power at a 2-tailed a of 0.05 for pairwise comparisons between
groups to detect an interaction between treatment and stratifi-
cation group. We did this by first using the Schlundt et al (19)
study interaction effect to determine sample size with the use of
the calculation in Tiwari et al (27). Because the Schlundt et al
study was a 2-group study, and we had 3 treatment groups, the
sample size of 177 was multiplied by 1.5 to account for the extra
group, making 266 participants necessary for our study. To ac-
count for a dropout rate of 13.5% as reported in the Schlundt
et al article, a total of 308 participants were expected to be re-
cruited. This would leave an adequate sample size of 266 for 90%
power to detect a small to moderate effect size of r = 0.22–0.34
reported in Schlundt et al.
RESULTS
Demographic characteristics
Of the 309 participants who were randomly assigned, 283
completed the study. Fourteen participants were lost to follow-up,
and 12 participants discontinued the study for various reasons
(pregnancy, time or work constraints, or did not wish to participate
in the control group). Comparisons of baseline covariates across
treatment groups are given in Table 1. There were no significant
differences in baseline covariates between the treatment groups at
baseline.
Effect of treatment on weight loss
For the completers-only analysis (n = 283), there was no effect
of treatment assignment on weight loss (P = 0.77; see Table 2
and Figure 2). Among breakfast skippers, baseline mean (6SD)
weight-, age-, sex-, site-, and race-adjusted weight changes
were 20.71 6 1.16, 20.76 6 1.26, and 20.61 6 1.18 kg for the
control, breakfast, and NB groups, respectively. Among break-
fast eaters, baseline mean (6SD) weight-, age-, sex-, site-, and
race-adjusted weight changes were 20.53 6 1.16, 20.59 6
1.06, and 20.71 6 1.17 kg for the control, breakfast, and NB
groups, respectively.
After adjustment for age, race, site, sex, prerandomization
status, and baseline weight, there was no evidence of an interaction
between treatment assignment and site (P = 0.32) nor between
treatment assignment and prerandomization status (P = 0.70) on
TABLE 1
Baseline demographic characteristics
1
Group
Control (n = 105) Breakfast (n = 101) No Breakfast (n = 103) P
Height (cm) 168.2 6 8.1
2
167.0 6 7.3 166.7 6 10.0 0.42
Weight (kg) 90.8 6 16.8 89.6 6 13.9 91.9 6 15.9 0.57
Women (n) 79 74 81 0.66
Age (y) 42.1 6 11.2 40.6 6 12.0 42.0 6 12.4 0.58
Race (n) 0.39
White non-Hispanic 54 45 48
Black non-Hispanic 32 40 34
White Hispanic 6 5 12
Black Hispanic 2 5 3
Other 11 6 6
Site 1.00
UAB 24 25 23
Columbia University 24 20 23
Colorado 15 14 15
Copenhagen 18 18 18
Boston 24 24 24
Breakfast skippers (n) 52 45 50 0.73
1
n = 309. There were no missing data for these variables. P values were calculated by using either an F test or Fisher’s
exact test, as appropriate. UAB, University of Alabama at Birmingham.
2
Mean 6 SD (all such values).
4of7 DHURANDHAR ET AL
weight change. With restriction to a strictly additive model (see
Table 3), weight change was not significantly affected by pre-
randomization status (P = 0.71). Subject sex was nominally
significant, with men losing, on average, 0.99 kg more than
women (P = 0.040). The age of the subject was suggestive of
a linear trend, with older individuals losing more weight, on
average, than their younger counterparts (a difference of 0.34 kg
for every 10 y of age; P = 0.050). Significant differences in
weight change by race (P = 0.034) were primarily driven by
evidence that black Hispanics tended to lose less weight than
other ethnicities (weight change shifted by +3.4 kg), although
the sample size for this subgroup was quite small (n = 10).
Significant differences in weight change by site (P = 0.035) were
primarily driven by the University of Copenhagen site, where
participants lost more weight, on average, than those at other
sites (weight change shifted by 21.75 kg). All of the afore-
mentioned effects in the additive model were calculated simul-
taneously (type II ANOVA). The multiple imputation produced
comparable results, with no differences in weight change by
treatment assignment or interactions of treatment assignment
with site or with prerandomization status. Findings with regard
to adjusted covariates were also similar, with no change in
nominal significance.
Compliance-adjusted treatment effects
On average, self-reported compliance (calculated as percentage
of days that participants followed breakfast recommendation after
determining intent when filling out the diary) was 93.6% for the
breakfast group and 92.4% for the NB group in completers.
Adjustment of treatment assignment by compliance did not result
in a significant effect of treatment on weight loss (P = 0.82) when
using the compliance information “as is” in a completers-only
analysis (Table 4). Results were not different under multiple
imputation, “fixing” the compliance information, or both.
Effect of control condition on breakfast eating habits
Among those assigned to the control group, there were 44
a priori breakfast skippers and 52 breakfast eaters. Over the
course of the experiment, 15 of 44 breakfast skippers began
eating breakfast regularly, whereas only 4 of 52 regular breakfast
eaters began skipping breakfast. The difference in odds of changing
behaviors cannot be explained by initial weight, age, race, study
site, or sex (Table 5); thus, the proclivity of changing breakfast
eating habits differed significantly by prerandomization status,
with skippers more likely to become breakfast eaters during the
intervention than vice versa (covariate-adjusted OR: 6.1; 95%
CI: 1.72, 27.65; P = 0.01) (Table 5). Results were not different
under multiple imputation.
DISCUSSION
Although breakfast consumption can help ensure adequate
nutrient intake and may have several health benefits (28–30), this
TABLE 2
Completer interaction model: weight-loss type II ANOVA
1
Sum of squares df FP
Initial weight 10.47 1 1.05 0.31
Age 31.16 1 3.12 0.079
Race 112.25 4 2.81 0.026
Site 107.45 4 2.69 0.032
Male sex 42.57 1 4.26 0.040
Prerandomization status 0.73 1 0.07 0.79
Assignment 5.13 2 0.26 0.77
Site 3 assignment 7.12 2 0.36 0.70
Prerandomization status 3 assignment 92.98 8 1.16 0.32
Residuals 2578.91 258
1
n = 283. Values were determined by type II ANOVA for the analysis of
weight loss as a function of initial weight, age, race, site, sex, prerandom-
ization status, and experimental assignment under a “completers only” anal-
ysis. In addition, interactions of assignment by site and by prerandomization
status were considered. No effect involving assignment was significant.
FIGURE 2. Mean (6SD) covariate-adjusted weight changes by treatment
and stratification. The gray dashed line represents “0” or no weight change
from baseline. ANOVA found no significant effects of treatment or stratifi-
cation on weight change. Among breakfast skippers, baseline mean (6SD)
weight-, age-, sex-, site-, and race-adjusted weight changes were 20.71 6
1.16, 20.76 6 1.26, and 20.61 6 1.18 kg for the C, B, and NB groups,
respectively. Among breakfast eaters, baseline mean (6SD) weight-, age-,
sex-, site-, and race-adjusted weight changes were 20.53 6 1.16, 20.59 6
1.06, and 20.71 6 1.17 kg for the C, B, and NB groups, respectively. B,
breakfast group; C, control group; NB, no-breakfast group.
TABLE 3
Completer additive model: weight-loss type II ANOVA
1
Sum of squares df FP
Initial weight 6.43 1 0.64 0.42
Age 38.69 1 3.87 0.050
Race 105.99 4 2.65 0.034
Site 104.74 4 2.62 0.035
Male sex 42.70 1 4.27 0.040
Prerandomization status 1.38 1 0.13 0.71
Assignment 5.13 2 0.26 0.77
Residuals 2679.77 268
1
n = 283. Values were determined by type II ANOVA for the analysis of
weight loss as a function of initial weight, age, race, site, sex, prerandom-
ization status, and experimental assignment in a strictly additive model (no
interactions allowed) under a “completers only” analysis. Assignment was
not significant.
THE EFFECT OF BREAKFAST RECOMMENDATIONS ON WEIGHT 5of7
RCT showed no effect of a recommendation to eat or skip
breakfast on weight loss. This experiment was not designed to
test the efficacy of a particular breakfast [eg, protein quality
(31)] or of the precise timing of isocaloric amounts of food on
weight loss (32), and conclusions with regard to the influence of
breakfast type or meal timing on weight loss cannot be drawn
from this study. Rather, we intended to test the effectiveness of
a public health recommendation to skip or eat breakfast in
causing weight loss in free-living individuals attempting to lose
weight over a 16-wk period.
Compliance with recommendations to eat or skip breakfast, at
least as judged by self-report, was high (93.6% for the breakfast
group and 92.4% for the NB group), suggesting that the rec-
ommendation was effective at producing the intended effects on
breakfast consumption. Therefore, we tested the effect of public
health breakfast recommendations among individuals who
wanted to lose weight and who took the recommendation seri-
ously and made an attempt to follow that recommendation. It
should also be noted that some individuals interpreted the control
general nutrition recommendations to mean that they should
begin eating breakfast, particularly if they were skippers before
the study, even though breakfast was not mentioned in the control
group instructions. This may have weakened our power to detect
differences between the control and breakfast groups but not
between the breakfast-skipping group and the other groups.
Contrary to the near-significant interaction found by Schlundt
et al (19) between initial breakfast eating habits and treatment
assignment that suggested that individuals who switch their
breakfast eating habits may lose more weight, we saw no in-
teraction of prerandomization status and treatment assignment.
They tested the impact of breakfast on weight loss as part of
a clinical weight-loss program that involved specific caloric
intake reduction targets and counseling to help participants meet
that target. Our findings suggest that the observed interaction in
the Schlundt et al study may have been a type 1 error.
Alternatively, intake was considerably more ad libitum in our
study than in the Schlundt et al study and reflective of what
individuals would do in a free-living condition if exposed to
a public health recommendation to eat more healthfully and
either skip or eat breakfast. It is likely that participants were freer
to compensate for any changes in energy intake that would have
resulted from changes in breakfast eating habits in our study
compared with the Schlundt et al study, which may account for
a lack of treatment by prerandomization status interaction in our
study. Our findings do not preclude the possibility that in a more
controlled, clinical weight-loss setting, switching breakfast
eating habits would produce more weight loss and only suggest
that in a free-living less controlled setting, breakfast eating habits
do not influence weight loss.
Several strengths of our study design should be considered.
This is the largest RCT designed to determine the effectiveness of
breakfast recommendations on weight loss. The large sample size
and multiple sites ensure both adequate power and reasonable
generalizability. In addition, our randomized design and rigorous
statistical analysis strengthen inferences about causation.
Several limitations of our study should also be considered. We
did not measure body composition or other metabolic variables.
Previous findings suggest that breakfast consumption may affect
endocrine regulation of appetite and body fatness (9–12), In
addition, our study was 16 wk in duration, which may not have
been long enough to detect an effect on body weight if the
overall change in energy intake resulting from the breakfast
habit changes is very small.
Because a general recommendation to eat or skip breakfast did
not influence weight loss in this study, future research might
assess whether more specific recommendations with regard to the
timing and quantity of meals or meal compositions might im-
prove weight-loss outcomes. In addition, our findings can only be
applied to free-living adults who are attempting to lose weight,
and therefore trials to determine the effectiveness of breakfast
recommendations in a clinical weight-loss setting, or in children
and adolescents, merit consideration. This is an important area in
need of investigation to ensure that public health recommen-
dations are effective and not unnecessarily using public resources
or discouraging public trust of authorities in the efforts to reduce
the population prevalence of obesity.
We thank registered dieticians Bettina Belmann Mirasola and Maria Roed
Andersen who helped collect data at the University of Copenhagen.
The authors’ responsibilities were as follows—DBA and EJD: designed
the research; EJD, A Alcorn, MC, ACB, EAT, and LHL: conducted the
research; JMS, CMA, JOH, DBA, M-PS-O, and A Astrup: provided essential
materials and facilities; JD: performed statistical analysis; EJD and JD:
wrote the manuscript (only authors who made a major contribution); and
EJD: had primary responsibility for final content. The authors disclose that,
although not for the support of this study, the authors or their institutions
TABLE 4
Completer additive model: compliance-adjusted weight loss
1
Sum of squares df FP
Initial weight 11.72 1 1.14 0.28
Age 27.36 1 2.65 0.10
Race 93.69 4 2.27 0.062
Site 76.34 4 1.85 0.12
Male sex 41.51 1 4.03 0.046
Prerandomization status 0.87 1 0.08 0.77
Compliance-adjusted assignment 4.05 2 0.20 0.82
Residuals 2574.32 250
1
n = 183. Values were determined by type II A NOVA for the analysis
of weight loss as a function of initial weight, age, race, site, sex, preran-
domization status, and compliance-adjusted assignment in a strictly addi-
tive model (no interactions allowed) under a “compl eters only” analysis.
Compliance-adjusted assignment was not signi ficant.
TABLE 5
Control group breakfast habit changes: completers
1
Likelihood ratio x
2
df P
Initial weight 1.71 1 0.19
Age 0.030 1 0.86
Race 8.94 4 0.063
Site 9.11 4 0.058
Male sex 0.15 1 0.70
Prerandomization status 6.65 1 0.010
1
n = 105. Values were determined by type II analysis of deviance for
the analysis of spontaneous changing of breakfast eating habits among the
control group. The binary outcome of habit change was considered in a lo-
gistic regression, regressing on prerandomization status while adjusting for
initial weight, age, race, site, and sex under a “completers only” analysis.
There was a significant effect for prerandomization status (15 of 44 breakfast
skippers became breakfast eaters, whereas only 4 of 52 breakfast eaters
became breakfast skippers).
6of7 DHURANDHAR ET AL
have received gifts, grants, or consulting fees from multiple organizations
that market products commonly consumed at breakfast, including Kraft
Foods (DBA), Kellogg Company (DBA), Cooking Light magazine (DBA),
Quaker (Pepsi; DBA), the Dairy Research Institute (DBA), the Egg Board
(DBA), Global Dairy Platform (AA), Arla (AA), Cargill (LHL), McDonalds
(JOH), General Mills (JOH), Walt Disney Company (JOH), Retrofit (JOH),
Nutrisystem (CMA), MetaProteomics (CMA), Dunkin’ Donuts (CMA),
Au Bon Pain (CMA), Bay State Milling Company (CMA), and Post Cereals
(M-PS-O). None of these organizations were involved in the support, design,
analysis, or interpretation of this study.
REFERENCES
1. Zeratsky K. Why does eating a healthy breakfast help control weight?
Available from: http://www.mayoclinic.org/food-and-nutrition/expert-
answers/FAQ-20058449 (cited 12 February 2014).
2. Casazza K, Fontaine KR, Astrup A, Birch LL, Brown AW, Bohan
Brown MM, Durant N, Dutton G, Foster EM, Heymsfield SB, et al.
Myths, presumptions, and facts about obesity. N Engl J Med 2013;368:
446–54.
3. Brown AW, Bohan Brown MM, Allison DB. Belief beyond the evi-
dence: using the proposed effect of breakfast on obesity to show 2
practices that distort scientific evidence. Am J Clin Nutr 2013;98:
1298–308.
4. Horikawa C, Kodama S, Yachi Y, Heianza Y, Hirasawa R, Ibe Y, Saito
K, Shimano H, Yamada N, Sone H. Skipping breakfast and preva-
lence of overweight and obesity in Asian and Pacific regions:
a meta-analysis. Prev Med 2011;53:260–7.
5. Mesas AE, Munoz-Pareja M, Lopez-Garcia E, Rodriguez-Artalejo F.
Selected eating behaviours and excess body weight: a systematic re-
view. Obesity Rev 2012;13(2):106-35.
6. Rampersaud GC, Pereira MA, Girard BL, Adams J, Metzl JD.
Breakfast habits, nutritional status, body weight, and academic per-
formance in children and adolescents. J Am Diet Assoc 2005;105:743–
60; quiz 61–2.
7. Szajewska H, Ruszczynski M. Systematic review demonstrating that
breakfast consumption influences body weight outcomes in children
and adolescents in Europe. Crit Rev Food Sci Nutr 2010;50:113–9.
8. Deshmukh-Taskar PR, Nicklas TA, O’Neil CE, Keast DR, Radcliffe
JD, Cho S. The relationship of breakfast skipping and type of breakfast
consumption with nutrient intake and weight status in children and
adolescents: the National Health and Nutrition Examination Survey
1999-2006. J Am Diet Assoc 2010;110:869–78.
9. Astbury NM, Taylor MA, Macdonald IA. Breakfast consumption af-
fects appetite, energy intake, and the metabolic and endocrine re-
sponses to foods consumed later in the day in male habitual breakfast
eaters. J Nutr 2011;141:1381–9.
10. Smeets AJ, Westerterp-Plantenga MS. Acute effects on metabolism
and appetite profile of one meal difference in the lower range of meal
frequency. Br J Nutr 2008;99:1316–21.
11. Leidy HJ, Racki EM. The addition of a protein-rich breakfast and its
effects on acute appetite control and food intake in ’breakfast-skipping’
adolescents. Int J Obes (Lond) 2010;34:1125–33.
12. Farshchi HR, Taylor MA, Macdonald IA. Deleterious effects of
omitting breakfast on insulin sensitivity and fasting lipid profiles in
healthy lean women. Am J Clin Nutr 2005;81:388–96.
13. Geliebter A, Yahav E, Haq S, Hashim SA. Cholesterol and weight
change following daily high or low fiber breakfast cereals. Obes Res
2000;8:25S-S.
14. Hamedani A, Akhavan T, Samra RA, Anderson GH. Reduced energy
intake at breakfast is not compensated for at lunch if a high-insoluble-
fiber cereal replaces a low-fiber cereal. Am J Clin Nutr 2009;89:1343–
9.
15. Geliebter A, Yahav E, Forbes G, Hashim SA. Lunch meal intake fol-
lowing high and low glycemic breakfast cereals. FASEB J 1999;13(5):
A871-A.
16. Kral TV, Whiteford LM, Heo M, Faith MS. Effects of eating breakfast
compared with skipping breakfast on ratings of appetite and intake at
subsequent meals in 8- to 10-y-old children. Am J Clin Nutr 2011;93:
284–91.
17. Levitsky DA, Pacanowski CR. Effect of skipping breakfast on sub-
sequent energy intake. Physiol Behav 2013;119:9–16.
18. Champagne CM, Han H, Bajpeyi S, Rood J, Johnson WD, Lammi-
Keefe CJ, Flatt JP, Bray GA. Day-to-day variation in food intake and
energy expenditure in healthy women: the Dietitian II Study. J Acad
Nutr Diet 2013;113:1532–8.
19. Schlundt DG, Hill JO, Sbrocco T, Pope-Cordle J, Sharp T. The role of
breakfast in the treatment of obesity: a randomized clinical trial. Am J
Clin Nutr 1992;55:645–51.
20. Derogatis LR, Melisaratos N. The Brief Symptom Inventory: an in-
troductory report. Psychol Med 1983;13:595–605.
21. Derogatis LR, Unger R. Corsini Encyclopedia of Psychology. Hoboken,
NJ; John Wiley & Sons, Inc., 2010.
22. Garner DM, Olmsted MP, Bohr Y, Garfinkel PE. The Eating Attitudes
Test: psychometric features and clinical correlates. Psychol Med 1982;
12:871–8.
23. USDA. Let’s eat for the health of it. Fayetteville, AR; University of
Arkansas Division of Agriculture Research & Extension, 2011. (De-
partment of Health and Human Services publication HHS-ODPHP-
2010-01-DGA-B.)
24. Team RCR. A language and environment for statistical computing.
Vienna, Austria: R Foundation for Statistical Computing, 2013.
25. Schafer JL. Analysis of incomplete multivariate data. London, United
Kingdom: Chapman & Hall, 1997.
26. Stef van Buuren KG-O. MICE: multivariate imputation by chained
equations in R. J Stat Softw 2011;45:1–67.
27. Tiwari HK, Birkner T, Moondan A, Zhang S, Page GP, Patki A, Allison
DB. Accurate and flexible power calculations on the spot: applications
to genomic research. Stat Interface 2011;4:353–8.
28. Nicklas TA, O’Neil C, Myers L. The importance of breakfast con-
sumption to nutrition of children, adolescents, and young adults. Nutr
Today 2004;39:30–9.
29. Kleinman RE, Hall S, Green H, Korzec-Ramirez D, Patton K, Pagano
ME, Murphy JM. Diet, breakfast, and academic performance in chil-
dren. Ann Nutr Metab 2002;46(suppl 1):24-30.
30. Berner LA, Keast DR, Bailey RL, Dwyer JT. Fortified foods are major
contributors to nutrient intakes in diets of US children and adolescents.
J Acad Nutr Diet 2014 (in press).
31. Vander Wal JS, Gupta A, Khosla P, Dhurandhar NV. Egg breakfast
enhances weight loss. Int J Obes (Lond) 2008;32:1545–51.
32. Stote KS, Baer DJ, Spears K, Paul DR, Harris GK, Rumpler WV,
Strycula P, Najjar SS, Ferrucci L, Ingram DK, et al. A controlled trial
of reduced meal frequency without caloric restriction in healthy, normal-
weight, middle-aged adults. Am J Clin Nutr 2007;85:981–8.
THE EFFECT OF BREAKFAST RECOMMENDATIONS ON WEIGHT 7of7
