Lower protein in infant formula is associated with lower weight up to
age 2 y: a randomized clinical trial1–4
Berthold Koletzko, Ru ¨diger von Kries, Ricardo Closa Monasterolo, Joaquı´n Escribano Subı´as, Silvia Scaglioni,
Marcello Giovannini, Jeannette Beyer, Hans Demmelmair, Dariusz Gruszfeld, Anna Dobrzanska, Anne Sengier,
Jean-Paul Langhendries, Marie-Francoise Rolland Cachera, and Veit Grote for the European Childhood Obesity
Trial Study Group
Background: Protein intake during infancy was associated with
rapid early weight gain and later obesity in observational studies.
Objective: The objective was to test the hypothesis that higher
protein intake in infancy leads to more rapid length and weight gain
in the first 2 y of life.
Design: In a multicenter European study, 1138 healthy, formula-fed
infants were randomly assigned to receive cow milk–based infant
and follow-on formula with lower (1.77 and 2.2 g protein/100 kcal)
or higher (2.9 and 4.4 g protein/100 kcal) protein contents for the
first year. For comparison, 619 exclusively breastfed children were
also followed. Weight, length, weight-for-length, and BMI were
determined at inclusion and at 3, 6, 12, and 24 mo of age. The pri-
mary endpoints were length and weight at 24 mo of age, expressed
as length and weight-for-length z scores based on World Health
Organization growth standards 2006.
Results: Six hundred thirty-six children in the lower (n ¼ 313) and
higher (n ¼ 323) protein formula groups and 298 children in the
breastfed group were followed until 24 mo. Length was not different
between randomized groups at any time. At 24 mo, the weight-for-
length z score of infants in the lower protein formula group was 0.20
(0.06, 0.34) lower than that of the higher protein group and did not
differ from that of the breastfed reference group.
Conclusions: A higher protein content of infant formula is associ-
ated with higher weight in the first 2 y of life but has no effect on
length. Lower protein intake in infancy might diminish the later risk
of overweight and obesity. This trial was registered at clinicaltrials.gov
as NCT00338689.Am J Clin Nutr 2009;89:1–10.
of later obesity in a large number of observational studies sum-
marized in 3 recent systemic reviews (1–3). Compared with
breastfed infants, formula-fed term infants have greater body
weight gains in infancy (4–6). The greater weight gain in formula-
fed infants than in infants fed breast milk might be explained by
different metabolizable substrate intakes (7), particularly protein:
protein intake per kilogram body weight is 55–80% higher in
formula-fed than in breastfed infants (8). It was proposed that
a higher protein intake stimulates secretion of insulin-like growth
factor I (IGF-I) and consecutively cell proliferation, which leads
to accelerated growth and increased adipose tissue (7, 9). A
positive association of protein intake with early growth was seen
in 2 observational studies (10, 11); whereas no effect on growth in
the first months of life was seen in other studies (12–14). Some
observational studies found a higher protein intake in the first 2 y
of life that was predictive of overweight in later childhood,
whereas energy, carbohydrate, or fat intake was not predictive (9,
To testthe hypothesis that a higher early protein intake leads to
more rapid growth in the first 2 y of life, we performed a mul-
ticenter, double-blind intervention trial in infants fed formula
randomly assigned to receive infant and follow-on formulas with
a lower or higher content of cow milk protein during the first year
of life. The growth pattern offormula-fed children was compared
with that of breastfed children recruited as an additional ob-
1From Dr von Hauner Children’s Hospital, University of Munich Medical
Centre, Munich, Germany (BK, JB, HD and VG); the Institute of Social
Paediatrics and Adolescent Medicine, University of Munich, Munich, Ger-
many (RvK and VG); the Universitad Rovira i Virgili, Reus, Spain (RCM
and JES); the Department of Paediatrics, University of Milano, Milano, Italy
(SS and MG); the Children’s Memorial Health Institute, Warsaw, Poland
(DG); the Department of Paediatrics, Universite ´ Libre de Bruxelles, Brussels,
Belgium (AS); CHC St Vincent, Lie `ge-Rocourt, Belgium (J-PL); and
INSERM, U557, Bobigny, France (M-FRC).
2This manuscript does not necessarily reflect the views of the Commis-
sion and in no way anticipates the future policy in this area.
3Supported in part by the Commission of the European Communities,
specific RTD Programme ‘‘Quality of Life and Management of Living
Resources,’’ within the 5th Framework Programme (research grant nos.
QLRT–2001–00389 and QLK1-CT-2002-30582) and the 6th Framework
Programme (contract no. 007036); the Child Health Foundation, Munich,
Germany; LMU innovative research priority project MC-Health (sub-project
I); the Kompetenznetzwerk Adipositas (Competence Network Obesity)
funded by the German Federal Ministry of Education and Research (FKZ
01GI0828); and the International Danone Institutes. BK is the recipient of
a Freedom to Discover Award of the Bristol-Myers-Squibb Foundation,
New York, NY.
4Address correspondence to B Koletzko, Dr von Hauner Children’s Hos-
pital, University of Munich, Lindwurmstraße 4, D-80337 Mu ¨nchen, Ger-
many. E-mail: firstname.lastname@example.org.
Received October 13, 2008. Accepted for publication March 3, 2009.
Am J Clin Nutr 2009;89:1–10. Printed in USA. ? 2009 American Society for Nutrition
AJCN. First published ahead of print April 22, 2009 as doi: 10.3945/ajcn.2008.27091.
Copyright (C) 2009 by the American Society for Nutrition
SUBJECTS AND METHODS
The study was a double-blind, randomized controlled trial
comparing 2 groups of children each fed 2 types (standard and
follow-on) of cow milk–based formula with either a lower or
higher protein content for the first year of life. Additionally, an
observational group of exclusively breastfed children was in-
cluded, as recommended by the European Society of Pediatric
Gastroenterology, Hepatology and Nutrition (18), the European
Commission (19), and the US Food and Nutrition Board of the
Institute of Medicine of the National Academies (20).
Eligible for study participation were apparently healthy, sin-
gleton, term infants who were born between 1 October 2002 and
31 July 2004. Children of mothers with a hormonal or metabolic
disease or illicit drug addiction during pregnancy were not in-
Italy, Poland, and Spain). Anthropometric measurements were
made at 11 sites: 2 in Germany (Munich and Nuremberg), 2 in
Belgium (Liege and Brussels), 4 in Italy (Milano), 1 in Poland
(Warsaw), and 2 in Spain (Reus and Tarragona). Except for Italy
and Poland, where one team was responsible for all measure-
ments, different teams were responsible per site in Germany,
Spain, and Belgium. A prerequisite for recruitment was that the
distance from the place of residence to the local study center was
compatible with later study visits.
Infants were enrolled during the first 8 wk of life. Formula-fed
infants had to be exclusively formula-fed at the end of the eighth
week of life. Breastfed children had to be breastfed since birth.
Noncompliance, prompting exclusion and no further follow-up,
was determined by maternal interviews at child ages 2, 3, 6, and
9 mo. For the intervention groups, noncompliance was defined as
feeding of nonstudy formula or breastfeeding for .10% of
feedings (or .3 bottles/wk) over ?1 wk in the first 9 mo of life.
Breastfed children had to be exclusively breastfed (,10% of
feedings or ,3 bottles of formula/wk) for the first 3 mo of life.
Additional reasons for exclusion were medical conditions that
might restrict growth or relocation too far away from the study
center to attend visits. The study was approved by the ethics
committeesofall studycenters.Written informedparental consent
was obtained for each infant.
The lower- and higher-protein infant formulas (manufactured
and provided free of charge to families by Bledina, Steenvoorde,
France) differed in the amount of cow milk protein (7.1% and
11.7% of energy) but had identical energy densities by compen-
satoryadaptation ofthefatcontent(Table1).The sourceoflipids
was a mix of vegetable oils: palm, rapeseed, coconut, and sun-
flower oils emulsified with soy lecithin. For details of the formula
composition, see the ‘‘Supplemental data’’ in the online issue.
After introduction of complementary feeding, but not before
the start of the fifth month of life, families of randomized infants
were provided with follow-on formulas with protein contents of
8.8% and 17.6% of energy, respectively, until the infants reached
the age of 12 mo. The composition of protein and fat and the
content of all other ingredients were identical (Table 1). The
whey-to-casein ratio in all study formulas was 1:4.
The composition of all study formulas complied with the 1991
EU Directive on Infant and Follow-on Formulae (22), and the
protein contents represented approximately the lowest and highest
levels, respectively, of the range accepted in this Directive.
Recruitment procedures in all centers were designed to pro-
mote and support breastfeeding. Introduction of any food other
than study formula or breast milk before the age of 4 completed
months was discouraged, but no other attempts were made to
influence the local and family traditions of introduction of solids
into the infants’ diet.
Shortly after birth, the parents were approached and invited to
participate in a study on obesity prevention, and the influence of
If the parents had opted for exclusive formula feeding, partici-
pation in the intervention group with cost-free provision of
formula was offered. In case of consent, children were randomly
assigned to treatment, and the first supply of the randomized
study formula was distributed with uniform instructions on how
to prepare the formula. Inclusion into the study was possible until
8 wk after birth. If a child was not exclusively formula-fed or
parents were indecisive, breastfeeding was encouraged and
children were included in an observational group of breastfed
children if the mother planned to exclusively breastfeed for ?4
mo. Once included in the study, children in the intervention and
the breastfed group were followed up identically.
Randomization lists for each country, stratified by sex, were
prepared by using random permuted blocks of 8. Two colors each
were used to label the lower and higher protein formulas (ie, 4
colors in total). Randomization numbers with respective colors
were drawn through an Internet-based platform. The formulas
were packaged in otherwise identical cans by the manufacturer
and labeled as infant or follow-on formula. The color codes were
only disclosed to the statisticians performing the final analysis.
All investigators and participating families were kept blinded to
allow for blinded follow up until later ages.
Birth weight and length were obtained from hospital data. All
other anthropometric measures were obtained at visits to the
The study entry visit with baseline anthropometric measurements
was scheduled at the time of randomization or as shortly as
measuring weight (Seca 336 scales at ?24 mo and Seca 702
scales at 24mo; Seca,Hamburg, Germany) and recumbent length
(Seca 232 until age 6 mo and PED LB 35–107 X scales after
6 mo; Ellard Instruments, Monroe, WA). Standing height was
measured in some children at 24 mo (Seca Stadiometer 242). All
measurements were performed twice and recorded with an ac-
curacy of 10 g for weight and 0.1 cm for length. The mean of
both measurements was used for further analysis. Standing
heights were only available for 12 children at 24 mo of age. To
make the measurement comparable with recumbent length, 0.7
cm was added in accordance with the procedures established in
the World Health Organization (WHO) Growth Multicentre
Reference Study (23). Written standard operating procedures
were also based on the procedures used in the WHO growth
study. Repeated training sessions were performed for all study
personnel by experts in anthropometry, with input from the
WHO Growth Multicentre Reference Study. Repeated site visits
KOLETZKO ET AL
were made by the coordinating center (Munich, Germany) for
monitoring and to ensure compliance with the protocol.
Information on the course of pregnancy, medical history,
lifestyle and behavior choices, socioeconomic background, and
mother’s prepregnancy weight was obtained from standardized
parent interviews at the baseline visit. Weight and height of the
fathersand mothersweremeasured atstudy entryandatthe12-mo
the first available height measurement of the mother and the first
available weight and height measurement of the father.
Infant food intakes were recorded by prospective 3-d weighed
food records at the infant ages of 3, 6, 12, and 24 mo obtained
with food scales (Unica 66006; Soehnle, Murrhardt, Germany)
provided to all participating families. Parents were advised
on how to record all food and beverages consumed during
2 weekdays and 1 weekend day. Intakes of energy and macro-
nutrients were calculated by using a database that was based on
the German BLS II.3 (24) (10652 original food items 1 2241
ones added before the study). Food items and recipes not
identified in the database were added at each study center ac-
cording to information from the manufacturers, other databases,
or ingredients. Energy intake was not calculated for food records
with any breastfeeding, because breastfeeding as intake of breast
milk was only measured in a subgroup of infants. Food records
with energy intakes .3 SDs of the mean by month and those
deemed incomplete and inaccurate or with reported concurrent
illness were excluded.
To compare the macronutrient contents of our study formulas
with those of other formulas commercially available during the
study period in Europe, we collected compositional data from all
infants (recommended before age 4 mo) and follow-on (from
4 mo of age) formulas used by participating families as recorded
in the 3-d weighed food records.
were expressed as SD scores (z scores) for length-for-age and
weight-for-length. Weight-for-length shows less variation than
weight-for-age and is a better descriptor of body composition in
children than weight. We expressed outcomes as z scores because
this makes the data more comparable with other studies, stand-
ardizes for sex, and takes into account the true age at the mea-
surement. Growth up to 24 mo of age was analyzed by adjusting
the respective anthropometric measurement for its baseline mea-
surement, as recommended (25). The study followed the recom-
mendations made in the CONSORT guidelines (26).
Data management and statistical analyses were carried out
with the software packages SAS version 9.2 (SAS Institute Inc,
Cary, NC) and Stata version 9.2 (StataCorp LP, College Station,
TX). Anthropometric results were expressed as z scores relative
to the growth standards of the WHO for breastfed children (27),
which were calculated by using WHO programs (http://
www.who.int/childgrowth/software/en/). Means (6SD) or me-
dians with interquartile ranges (IQR: 25th and 75th percentiles)
were used as appropriate. Pearson chi-square and Fisher’s exact
tests were used for statistical comparison of categorical data,
and a t test was used for normally distributed continuous data.
Composition of the study formulas, human milk, and commercial formula1
Study formulas (cow milk–based) Formulas used during the study in all
Infant formulasFollow-on formulas
(n ¼ 58)
(n ¼ 45)
(n ¼ 94)Lower proteinHigher proteinLower proteinHigher protein
Energy (g/100 mL)
Proteins (g/100 mL)
Proteins (g/100 kcal)
Proteins (% of energy)
Lipids (g/100 mL)
Carbohydrates (g/100 mL)
Sodium (mg/100 mL)
Calcium (mg/100 mL)
Iron (mg/100 mL)
Zinc (mg/100 mL)
A (lg/100 mL)
D (lg/100 mL)
E (lg/100 mL)
B-6 (mg/100 mL)
B-12 (lg/100 mL)
Folic acid (lg/100 mL)
70 6 6.7
1.2 6 0.2
3.6 6 0.7
7.4 6 0.2
1The quality of protein, carbohydrate, and fat of the study formulas was identical; the source of lipids was a mix of vegetable oils (palm, rapeseed,
coconut, and sunflower oils emulsified with soy lecithin). A mix of synthetic vitamins and vitamin B-12 from a fermentation source was added.
2All values are means 6 SDs and are from the Darling study (21).
3All values are medians (ranges) and are based on all formulas noted by the study participants in 3-d weighted food protocols.
LOWER PROTEIN INTAKE LOWERS WEIGHT AT 24 MO
Linear regression analysis was applied to adjust the effects of
type of feeding on z scores at 24 mo for weight, length, weight-
for-length, and BMI for the respective baseline values. For
pairwise comparison of the effect of each type of formula feeding
with the breastfeeding group, potential confounders (sex, moth-
er’s educational status, smoking in pregnancy, and country) were
also considered. To assess the effect of clustering by 8 different
study teams, we used robust cluster variance estimators.
Description of baseline characteristics is based on all children
allocated to study formula or to the breastfed group. To test
whether therewere different effects of the type offeeding between
countries, an interaction term between formula and country was
added to the linear regression model. Questionnaire data were
evaluatedfor all participatingsubjects, evenifanthropometric data
were missing. An intention-to-treat analysis was not performed
because children that switched to nonstudy formula (or breast-
by study protocol.
coefficient models as described by Singer and Willet (28) and
Fitzmaurice et al (29) were applied to model growth differences
between the lower- and higher-protein formula groups by using
all available measurements from baseline to 24 mo. Both models
account for the correlated data structure because of the repeated
measurements and use the exact age of measurement. The
piecewise-linear-random-coefficient model was chosen to ana-
lyze the age-dependent effect of study formula on the anthro-
pometric outcome. The idea of the model is to split the time in
fixed segments with different slopes in each segment, in contrast
with the usual multilevel linear growth model, which uses one
slope over the whole analysis time. The choice of the time
segments (0–3 mo, 3–6 mo, 6–12 mo, and 12–24 mo) for this
model is based on the measurement points as planned per pro-
tocol. Statistical significance of differences between trajectories
of the study groups in the piecewise-linear-random-coefficient
model was assessed by 95% prediction bands. If the 95% pre-
diction bands of one group (eg, lower protein) does not overlap
with the average trajectory of the other study group (eg, higher
protein), there is a significant difference between these trajec-
tories with a 5% probability error.
Study size and power calculation
The study was designed to have sufficient power (90%) to
2 formula groups, with a level of significance of 0.05 and an
expected population mean (WHO standard population) of 87.8 6
3 cm. The necessary number of 296 children in each intervention
group was inflated to a target of ?500 randomized children in
each group to meet an expected dropout rate of 30%.
A total of 1138 formula-fed infants were randomly assigned
for treatment, and 1090 were allocated to study formula (Figure
1). The median age at randomization was 14 d (IQR: 3–30 d)
and at the baseline visit was 16 d (IQR: 2–29 d); 249 (23%)
children were exclusively formula-fed since birth, and all others
gradually switched from breastfeeding to formula feeding within
the first 8 wk of life.
After allocation to the study formula, 229 children were lost to
and 169 were excluded for lack of compliance (Figure 1). One-
hundred eighteen (70%) of the latter children were excluded
from further study participation during the first 3 mo of life, 29
by 6 mo of age, and 22 children by 9 mo of age (Figure 1). Eight
of these 169 children completed the study up to 24 mo and were
excluded according to the study protocol after study termination
because questionnaire data or food protocols indicated lack of
The observational group of breast-fed infants included 619
infants, of whom 349 participated in the follow-up visits at 6 mo,
327 at 12 mo, and 304 at 24 mo. About 7% of all children in the
intervention group completed questionnaires at 24 mo but had no
anthropometric measurement taken. Thus, the final analysis was
confined to 313 (follow-up rate: 58%) children in the lower-
protein group, 323 (59%) children in the higher-protein group,
and 298 (51%) children in the breastfed group.
Randomization of the infants to the lower- and higher-protein
groups was successful: there were no significant differences in
descriptors of socioeconomic status, smoking habits, parental
anthropometric measures, gestational age, and weight and length
at birth and at baseline (Table 2). Parents in the observational
group of breastfed children had a higher educational level and
fewer mothers smoked than in the formula-fed groups.
Parents of children lost to follow-up had a lower educational
level, and the children’s mothers were more likely to be smokers
than were those still in the study at 24 mo, with no difference
between the lower- and higher-protein formula groups. In con-
trast, the 169 children that were excluded for lack of compliance
were not significantly different in any of the characteristics
considered. They also had no differences in weight or length at
baseline or later visits compared with the children staying in the
study (data not shown).
Mean weight, length, weight-for-length, and BMI of all par-
ticipating infants (both formula fed and the breastfed children)
Standard: average weight of children in the lower-protein group,
for example, was 0.42 SDs lower (Table 3).
Energy intake in the lower- and higher-protein formula groups
wasidentical at3, 12,and 24mo. At6moofage,energy intakein
the lower-protein formula group was slightly higher (715 6 127
kcal) than in the higher-protein formula group (690 6 121 kcal)
(P ¼ 0.010) (Figure 2). Energy intake in the intervention groups
increased from 584 6 99 kcal/d at 3 mo (n ¼ 809) to 1108 6
236 kcal/d at 24 mo (n ¼ 484). The protein intake was signifi-
cantly different between the 2 formula groups (P , 0.001) at all
time points until the end of the intervention at 12 mo, but not
thereafter (Figure 2). Correspondingly, the fat intake was sig-
nificantly lower in the higher-protein group at all time points up
to age 12 mo, whereas there was no difference in carbohydrate
intake (data not shown).
Differences in weight, weight-for-length, and BMI between
the formula groups emerged at 6 mo of age and remained rel-
atively stable thereafter with a decreasing tendency toward the
end of the study. Length, weight, weight-for-length, and BMI
increaseswere lower for breastfedthanfor higher-proteinformula-
fed children between baseline and 12 mo of age (Figure 3).
At 24 mo of age, length was not different between the in-
tervention groups. The mean weight attained at 24 mo was 12.42
KOLETZKO ET AL
and 12.60 kg for the lower- and higher-protein groups, re-
spectively (Table 3), without a significant difference in adjusted
z scores. At 24 mo the adjusted z score for weight-for-length was
found to be 0.20 (95% CI: 0.06, 0.34; P ¼ 0.005) greater in the
higher- than in the lower-protein formula group (Table 3); the CI
was unaffected by clustering by study teams. The differences
found in anthropometric z scores at 24 mo of age would trans-
late, eg, for girls, into a difference in length, weight, and BMI
(in kg/m2) of 20.1 (95% CI: 20.6, 0.4) cm, 150 (95% CI: 213,
325) g, and 0.3 (95% CI: 0.1, 0.4), respectively.
Although the anthropometric measures at 24 mo of age dif-
fered significantly between countries, the effect of the in-
analyzed anthropometric measures was the interaction between
country and type of formula significant. For instance, the base-
line-adjusted z scores for weight-for-length at 24 mo for the
higher- and lower-protein groups were 0.46 and 0.18 in Spain
and 0.23 and 20.05 in Belgium, respectively.
in the higher-protein group than in the lower-protein group,
whereas the z scores for weight-for-length and BMI were signif-
icantly higher only at 6 and 12 mo of age. In general, the differ-
at 12 mo of age for weight, weight-for-length, and BMI, indicat-
ing the strongest effect of the intervention at this time point
When modeling the whole growth trajectories of each child
over the first 2 y of life with multilevel linear growth models, we
saw significant interactions between type of study formula and
age, which reflected the above differences in effect of the study
formula over age (ie, the strongest effect at 12 mo of age). To
depict this age-dependent effect of study formula, we used
a piecewise-linear-random-coefficient model. A statistically sig-
nificant difference between the lower- and higher-protein groups
originated at 3–6 mo of age and slightly decreased after 12 mo of
age for weight-for-length, BMI, and weight. (See Supplemental
Figure 4 under ‘‘Supplemental data’’ in the online issue.)
To compare the observational group of breastfed children with
children from the intervention group, we adjusted for baseline
measurements and potential confounders. Compared with breast-
fed children, children fed higher-protein formula had significantly
higher z scores for weight, length, weight-for-length, and BMI at
24 mo, with a difference of 0.30 (0.15, 0.45), 0.27 (0.12, 0.43),
0.18 (0.02, 0.33), and 0.20 (0.05, 0.36) SDs, respectively. In
contrast, there was no significant difference in z scores of weight-
for-length and BMI between the lower-protein group and the
breastfed group at 24 mo, but the z scores for weight and length
were 0.16 (0.01, 0.31) and 0.29 (0.13, 0.45) SDs higher in the
lower-protein group than in the breastfed group.
This randomized, controlled trial showed that a higher protein
FIGURE 1. Randomization, allocation, and follow-up of study participants in the intervention group.
LOWER PROTEIN INTAKE LOWERS WEIGHT AT 24 MO
there was no significant difference with respect to length, a higher
formula feeding. Because formula groups at 24 mo showed no
difference in weight-for-length and in BMI is probably due to
a difference in body fat or a difference in adiposity.
which consistentlyreportedlowergrowth inthe first 2 y of life(10,
15, 16) and lower body size (9, 15–17) in children and adoles-
cents fed a lower-protein diet in infancy. Three small randomized
trials that compared the growth of children fed formula of dif-
ferent protein contents showed either no (1–3) or some (31)
positive effect of higher protein intake on growth in infancy; the
different results were potentially due to limited sample sizes.
Interestingly, no effect of the intervention on length growth
was observed at any time point over the first 24 mo of life. A
potential reduction in length growth had been chosen as an
experimental primary endpoint because of suggestions that early
lengthgrowthmightbea novel predictor oflater overweight(32).
However, many studies (1–3) showed that early weight gain is
the best predictor of later childhood overweight, a question
explicitly addressed by Toschke et al (33).
The effects observed are likely to be conservative estimates.
Many of the study children were partially breastfed during the
first 8 wk of life. If all formula-fed children had been exclusively
formula-fed from birth on, the effects might have been greater.
The protein contents of the lower- and higher-protein formulas
in our study werewithin the range of commercial infant formulas
available in Europe during the study period (Table 1). Tradi-
tionally, the protein content in infant formulas has been far
higher than in human milk. The protein content of formula was
as high as 4 g/100 kcal in the 1970s, decreased to ’3 g/100 kcal
in the 1980s, and tended to decrease further thereafter. Whereas
adverse effects of higher protein intakes were not of major
concern, worries about the deleterious effects of too low an in-
take of protein (34) prevailed. Considering that the protein content
of human milk varies and tends to have a higher biological value
than cow milk protein (19), the protein composition of infant
formula was designed to always meet the assumed minimum
requirements of protein and indispensable amino acids of infants.
The protein intake provided by the lower-protein formula was
still higher than that of the breastfed children or than that rec-
ommended for infants (19, 35, 36). For instance, formula-fed
children had protein intakes of ’14 and 20 g/d at 3 and 6 mo,
respectively, in our study, whereas breastfed children in the
Baseline characteristics for children in the intervention and the observational study groups
Intervention group Observational group
Lower protein Higher proteinBreastfed
n (%) Mean 6 SDn (%)Mean 6 SD P value1
n (%)Mean 6 SDP value2
Mother’s educational level
Mother smoked in past 3 mo
before or during pregnancy
Mother smoked beyond
12th wk of gestation
Age of mother (y)
BMI of mother (kg/m2)
BMI of father (kg/m2)
Birth length (cm)
Birth weight (kg)
257 (47.6)—238 (43.3)—0.495145 (24.7)——
29.8 6 5.2
23.7 6 4.6
25.9 6 3.8
50.6 6 2.7
3.3 6 0.3
29.8 6 5.4
23.7 6 4.6
26.1 6 3.5
50.6 6 2.6
3.3 6 0.4
31.2 6 4.5
22.4 6 3.6
25.6 6 3.4
50.5 6 2.4
3.3 6 0.3
1Chi-square test for categorical data and t test for comparison of means.
2For comparison between the intervention and observational group.
KOLETZKO ET AL
DARLING study (37) had protein intakes of 7 and 8 g/d. Con-
sumption of the lower-protein formula supported normal length
growth, and parental reports did not indicate any untoward ef-
fects. Given that the supply of total protein and essential amino
acids with the lower-protein formula is clearly higher than ref-
erence intakes for infants that are regarded as safe (19, 35, 36),
one would not expect any untoward effects on growth or func-
tional outcomes, such as neurologic development or immune
response. However, further follow-up of the study cohort up to
school age is planned to document neurologic and other out-
One of the strengths of our study was the high external validity
due to its multinational design. Although the anthropometric
measures at 24 mo of age differed in each country, the effect of
Number of children, age at measurement, and anthropometric measures with z scores for the intervention (formula) and observational group (breastfed) at
baseline and 24 mo
Intervention groupObservational group
Lower protein Higher protein
at 24 mo1
Baseline24 mo Baseline24 mo Baseline 24 mo
No. of children2
Age at measurement (d)
539 313550322 630585298
16 (2, 30)3
733 (730, 737)15 (2, 28)733 (729, 737)12 (3, 21) 734 (730, 738)
3.72 6 0.764
20.42 6 0.78
12.42 6 1.32
0.34 6 0.85
3.67 6 0.75
20.43 6 0.75
12.60 6 1.46
0.44 6 0.95
3.55 6 0.61
20.31 6 0.78
12.26 6 1.41
0.24 6 0.93
0.12 (20.011, 0.25)5
52.1 6 2.9
20.25 6 1.02
88.3 6 3.1
0.31 6 0.94
52.0 6 3.1
20.23 6 1.00
88.1 6 3.2
0.24 6 1.02
51.7 6 2.5
20.02 6 1.04
87.5 6 3.4
0.09 6 1.06
20.039 (20.18, 0.10)6
13.6 6 1.6
20.47 6 0.86
20.48 6 1.07
16.1 6 1.2
0.19 6 0.89
0.18 6 0.86
13.5 6 1.5
20.50 6 0.82
20.52 6 1.05
16.4 6 1.3
0.40 6 0.95
0.37 6 0.93
13.2 6 1.4
20.46 6 0.91
20.64 6 1.13
16.2 6 1.3
0.25 6 0.95
0.21 6 0.93
0.23 (0.089, 0.36)7
0.20 (0.060, 0.34)8
1Derived from linear regression adjusted for the respective anthropometric baseline value; 95% CIs in parentheses.
2Numbers vary slightly between anthropometric measures.
3Median; 25th and 75th percentile in parentheses (all such values).
4Mean 6 SD (all such values).
5P ¼ 0.072.
6P ¼ 0.589.
7P ¼ 0.001.
8P ¼ 0.005.
FIGURE 2. Mean (6SD) daily energy and protein intakes at 3, 6, 12, and 24 mo of age in 456 children in the lower-protein formula group and in 454
children in the higher-protein formula group. **,***Significantly different from the lower-protein group (t test): **P , 0.01, ***P , 0.001.
LOWER PROTEIN INTAKE LOWERS WEIGHT AT 24 MO
the intervention was not different between the countries. A
further strength of the study was the documented difference in
protein intakes between intervention groups that matches the
Attrition was higher than expected. This may have been
explained by the fact that the questionnaires and visit schedules
were demanding and study participation of families with healthy
infants was voluntary without any perceived benefit other than
access to free formula. However, because attrition was equal in
both arms and randomization was initially successful, attrition
likely did not bias our results.
A potentially more serious issue was the exclusion of 169
noncompliant children that should have been included for a proper
the study protocol, because compliance was not expected to be
related to type of study formula or growth. All of the study
FIGURE 3. Mean z scores (with 95% CIs) for length, weight, weight-for-length, and BMI in the lower-protein (n ¼ 540) and higher-protein (n ¼ 550)
groups and in the breastfed (n ¼ 588) children at baseline (0–8 wk of age) and at 3, 6, 12, and 24 mo of age. *,**,***Significantly different from the lower-
protein group (ANOVA adjusted for baseline value): *P , 0.05, **P , 0.01, ***P , 0.001.
KOLETZKO ET AL
formulae because they complied with European regulatory
standards in all aspects, including protein contents. Furthermore,
formula whenever parents perceive a potential problem (eg, the
in most cases, this is not related to the formula used. Importantly,
there was no difference in frequency of noncompliance between
shortly after randomization within the first 3 mo of life, which
rendered it unlikely that noncompliance was related to differences
in weight gain between the 2 formula groups.
The proportion of smokers was high: 46% of the mothers in
both intervention groups were smokers, and ’30% smoked after
the 12th week of pregnancy. However, the population, for which
the data of this intervention trial applies, was the population of
nonbreastfeeding mothers only, who are generally more likely to
smoke (38, 39). Thus, the high proportion of smoking mothers is
expected and unlikely to limit the external validity of the study.
Potential effect of a lower protein diet on obesity
Monteiro et al (40) found an odds ratio of 1.87 (95% CI: 1.10,
3.18) for obesity in 14–16-y-old Brazilian adolescents for every
1-SD change in weight-for-length gain during the first 2 y of life.
According to Ong and Loos (3), this is the average effect seen in
studies of weight gain in infancy on later obesity. Thus, the
observed increase in z score for weight-for-length at the age of
24 mo by 0.19 SDs in the higher-protein formula group com-
pared with the lower protein group would yield an extrapolated
odds ratio of 1.13 (95% CI: 1.02, 1.25) [OR ¼ exp(ln(1.87) 3
0.19) ¼ 1.13] for being obese in adolescence.
This large randomized controlled trial showed significant
effects of a lower protein intake from infant formula on weight,
weight-for-length, and BMI in the first 2 y of life. Limiting the
protein content of infant and follow-on formula and, more
generally, the dietary protein intake during infancy, might con-
stitute a potentially important approach to reducing the risk of
childhood overweight and obesity.
We thank the participating families and all project partners for their enthu-
siasticsupportofthe projectandWCameronChumlea(Departments ofCom-
munity Health and Pediatrics, Lifespan Health Research Center, Wright State
University Boonshoft School of Medicine, Dayton, OH) for his help in setting
up standardized anthropometric measures and in training the study personnel.
Members of The European Childhood Obesity Trial Study Group: Annick
Xhonneux, Jean-Noel Van Hees, and Franc xoise Martin (CHC St Vincent,
Janas, and Ewa Pietraszek (Children’s Memorial Health Institute, Warsaw,
Poland); Sabine Verwied-Jorky, Sonia Schiess, Ingrid Pawellek, Uschi Handel,
Maximilians UniversityofMunich, Munich, Germany);Helfried Groebe,Anna
Reith, and Renate Hofmann (Klinikum Nurnberg Sued, Nurnberg, Germany);
Joana Hoyos, Philippe Goyens, Clotilde Carlier, and Elena Dain (Universite ´
Me ´ndez Riera (UniversitatRovira i Virgili,Spain);and SabrinaTedeschi,Carlo
The authors’ responsibilities were as follows—BK (initiator and principal
RvK: concept of statistical evaluation and writing of manuscript; RCM (study
center coordinator): critical reading of manuscript; JES, SS, and DG: coor-
dination of study and critical reading of manuscript; MG (study center coor-
dinator): critical reading of manuscript; JB and AD: recruitment, conduct of
study, data entry, and critical reading of manuscript; HD: administrative co-
ordination, conduct of study, and critical reading of manuscript; AS and J-PL:
recruitment, conduct of study, and critical reading of manuscript; M-FRC:
counseling on standardized anthropometric measures, training of study per-
sonnel, and critical reading of manuscript; and VG: data management and
analysis and writing of manuscript. None of the authors reported a conflict
ofinterest followingtheguidelinesof the International Committeeof Medical
1. Baird J, Fisher D, Lucas P, Kleijnen J, Roberts H, Law C. Being big or
growing fast: systematic review of size and growth in infancy and later
obesity. BMJ 2005;331:929–31.
2. Monteiro POA, Victora CG. Rapid growth in infancy and childhood
and obesity in later life - a systematic review. Obes Rev 2005;6:143–54.
3. Ong KK, Loos RJF. Rapid infancy weight gain and subsequent obesity:
systematic reviews and hopeful suggestions. Acta Paediatr 2006;95:
4. Dewey KG. Growth characteristics of breast-fed compared to formula-fed
infants. Biol Neonate 1998;74:94–105.
5. Kramer MS, Guo T, Platt RW, et al. Feeding effects on growth during
infancy. J Pediatr 2004;145:600–5.
6. Victora CG, Morris SS, Barros FC, Horta BL, Weiderpass E, Tomasi E.
Breast-feeding and growth in Brazilian infants. Am J Clin Nutr 1998;67:
7. Koletzko B, Broekaert I, Demmelmair H, et al. Protein intake in the first
year of life: a risk factor for later obesity? The EU Childhood Obesity
project. In: Koletzko B, Dodds PF, Akerblom H, Ashwell M, eds. Early
nutrition and its later consequences: new opportunities. Dordrecht,
Netherlands: Kluwer Academic, 2005:69–79.
8. Alexy U, Kersting M, Sichert-Hellert W, Manz F, Schoch G. Macro-
nutrient intake of 3- to 36-month-old German infants and children:
results of the DONALD Study. Dortmund Nutritional and Anthropo-
metric Longitudinally Designed Study. Ann Nutr Metab 1999;43:
9. Rolland Cachera MF, Deheeger M, Akrout M, Bellisle F. Influence of
macronutrients on adiposity development: a follow up study of nutrition
and growth from 10 months to 8 years of age. Int J Obes Relat Metab
10. Stunkard AJ, Berkowitz RI, Schoeller D, Maislin G, Stallings VA.
Predictors of body size in the first 2 y of life: a high-risk study of human
obesity. Int J Obes Relat Metab Disord 2004;28:503–13.
11. Nielsen GA, Thomsen BL, Michaelsen KF. Influence of breastfeeding
and complementary food on growth between 5 and 10 months. Acta
12. Turck D, Grillon C, Lachambre E, et al. Adequacy and safety of an
infant formula with a protein/energy ratio of 1.8 g/100 kcal and en-
hanced protein efficiency for term infants during the first 4 months of
life. J Pediatr Gastroenterol Nutr 2006;43:364–71.
13. Heinig MJ, Nommsen LA, Peerson JM, Lonnerdal B, Dewey KG. Intake
and growth of breast-fed and formula-fed infants in relation to the
timing of introduction of complementary foods: the DARLING study.
Davis Area Research on Lactation, Infant Nutrition and Growth. Acta
14. Lonnerdal B, Hernell O. Effects of feeding ultrahigh-temperature
(UHT)-treated infant formula with different protein concentrations or
powdered formula, as compared with breast-feeding, on plasma amino
acids, hematology, and trace element status. Am J Clin Nutr 1998;68:
15. Gunnarsdottir I, Thorsdottir I. Relationship between growth and feeding
in infancy and body mass index at the age of 6 years. Int J Obes Relat
Metab Disord 2003;27:1523–7.
16. Gunther AL, Buyken AE, Kroke A. Protein intake during the period of
complementary feeding and early childhood and the association with
body mass index and percentage body fat at 7 y of age. Am J Clin Nutr
17. Scaglioni S, Agostoni C, Notaris RD, et al. Early macronutrient intake
and overweight at five years of age. Int J Obes Relat Metab Disord 2000;
LOWER PROTEIN INTAKE LOWERS WEIGHT AT 24 MO
18. Aggett PJ, Agostini C, Goulet O, et al. The nutritional and safety
assessment of breast milk substitutes and other dietary products for
infants: a commentary by the ESPGHAN Committee on Nutrition.
J Pediatr Gastroenterol Nutr 2001;32:256–8.
19. Scientific Committee on Food. Report of the Scientific Committee on
Food on the revision of essential requirements of infant formulae and
follow-on formulae. Brussels, Belgium: European Commission, 2003.
20. Committee on the Evaluation of the Addition of Ingredients New to
Infant Formula. Comparing infant formulas with human milk. infant
formula: evaluation of the safety of new ingredients. Washington, DC:
The National Academies Press, 2004:49–66.
21. Nommsen LA, Lovelady CA, Heinig MJ, Lonnerdal B, Dewey KG.
Determinants of energy, protein, lipid, and lactose concentrations in
human milk during the first 12 mo of lactation: the DARLING Study.
Am J Clin Nutr 1991;53:457–65.
22. European Commission. Commission Directive 91/321/EEC of 14 May 1991
oninfantformulae andfollow-onformulae.Off JEur Union1991:0035–0049.
23. de Onis M, Onyango AW, Van den Broeck J, Chumlea WC, Martorell R.
Measurement and standardization protocols for anthropometry used in
the construction of a new international growth reference. Food Nutr Bull
24. Federal Institute for Health Protection of Consumers and Veterinary
Medicine. The German Food Code and Nutrient Data Base: conception,
structure and documentation of the data base blsdat. BLS II, 3rd ed.
Berlin, Germany: BgVV Publications, 1999.
25. Twisk J, Proper K. Evaluation of the results of a randomized controlled
trial: how to define changes between baseline and follow-up. J Clin
26. Moher D, Schulz KF, Altman D. The CONSORT statement: revised
recommendations for improving the quality of reports of parallel-group
randomized trials. JAMA 2001;285:1987–91.
27. WHO Multicentre Growth Reference Study Group. WHO child growth
standards: length/height-for-age, weight-for-age,weight-for-length, weight-
for-height and body mass index-for-age: methods and development.
Geneva, Switzerland: World Health Organization, 2006.
28. Singer J, Willett J. Applied longitudinal data analysis. Modelling
change and event occurence. New York, NY: Oxford University Press,
29. Fitzmaurice G, Laird N, Ware J. Applied longitudinal analysis. Hoboken,
NJ: John Wiley & Sons, Inc, 2004.
30. Forbes GB. Lean body mass-body fat interrelationships in humans. Nutr
31. Axelsson IE, Jakobsson I, Raiha NC. Formula with reduced protein
content: effects on growth and protein metabolism during weaning.
Pediatr Res 1988;24:297–301.
32. Rolland Cachera MF, Meance S, Deheeger M. Length gain in infancy is
associated with body fat and fat pattern at age 14 years. Int J Obes 2000;
33. Toschke AM, Grote V, Koletzko B, von Kries R. Identifying children at
high risk for overweight at school entry by weight gain during the first
2 years. Arch Pediatr Adolesc Med 2004;158:449–52.
34. Dewey KG, Beaton G, Fjeld C, et al. Protein requirements of infants and
children. Eur J Clin Nutr 1996;50:S119–50.
35. Assessment of nutrient requirements for infant formulas. J Nutr 1998;
36. Energy and protein requirements. Report of a Joint FAO/WHO/
UNU Expert Consultation. World Health Organ Tech Rep Ser 1985;
37. Heinig MJ, Nommsen LA, Peerson JM, Lonnerdal B, Dewey KG. Energy
and protein intakes of breast-fed and formula-fed infants during the first
year of life and their association with growth velocity: the DARLING
Study. Am J Clin Nutr 1993;58:152–61.
38. Riva E, Banderali G, Agostoni C, Silano M, Radaelli G, Giovannini M.
Factors associated with initiation and duration of breastfeeding in Italy.
Acta Paediatr 1999;88:411–5.
39. Noble L, Hand I, Haynes D, McVeigh T, Kim M, Yoon JJ. Factors influ-
encing initiation of breast-feeding among urban women. Am J Perinatol
40. Monteiro PO, Victora CG, Barros FC, Monteiro LM. Birth size, early
childhood growth, and adolescent obesity in a Brazilian birth cohort. Int
J Obes Relat Metab Disord 2003;27:1274–82.
KOLETZKO ET AL