at SciVerse ScienceDirectSocial Science & Medicine 75 (201.docx

at SciVerse ScienceDirect
Social Science & Medicine 75 (2012) 323e330
Contents lists available
Social Science & Medicine
journal homepage: www.elsevier.com/locate/socscimed
Breastfeeding and risk of overweight and obesity at nine-years
of age
Cathal McCrory*, Richard Layte 1
The Economic and Social Research Institute, Whitaker Square,
Sir John Rogerson’s Quay, Dublin 2, Ireland
a r t i c l e i n f o
Article history:
Available online 17 April 2012
Keywords:
Ireland
Breastfeeding
Children
Overweight
Obesity
Body mass index (BMI)
Cohort study
* Corresponding author. Tel.: þ353 1 8632027; fax:
E-mail address: [email protected] (C. McCror
1 Tel.: þ353 1 8632027; fax: þ353 1 8632100.

0277-9536/$ e see front matter � 2012 Elsevier Ltd.
doi:10.1016/j.socscimed.2012.02.048
a b s t r a c t
Whether breastfeeding is protective against the development of
childhood overweight and obesity
remains the subject of considerable debate. Although a number
of meta-analyses and syntheses of the
literature have concluded that the greater preponderance of
evidence indicates that breastfeeding
reduces the risk of obesity, these findings are by no means
conclusive. The present study used data from
the Growing Up in Ireland study to examine the relationship
between retrospectively recalled breast-
feeding data and contemporaneously measured weight status for
7798 children at nine-years of age
controlling for a wide range of variables including; socio-
demographic factors, the child’s own lifestyle-
related behaviours, and parental BMI. The results of the
multivariable analysis indicated that being
breastfed for between 13 and 25 weeks was associated with a 38
percent (p < 0.05) reduction in the risk
of obesity at nine-years of age, while being breastfed for 26
weeks or more was associated with a 51
percent (p < 0.01) reduction in the risk of obesity at nine-years
of age. Moreover, results pointed towards
a doseeresponse patterning in the data for those breastfed in
excess of 4 weeks. Possible mechanisms
conveying this health benefit include slower patterns of growth
among breastfed children, which it is
believed, are largely attributable to differences in the
composition of human breast milk compared with
synthesised formula. The suggestion that the choice of infant
feeding method has important implications
for health and development is tantalising as it identifies a
modifiable health behaviour that is amenable

to intervention in primary health care settings and has the
potential to improve the health of the
population.
� 2012 Elsevier Ltd. All rights reserved.
Introduction
The belief that breastfeeding during infancy affords protection
against a number of diseases features prominently in the epide-
miological literature; there is considerable evidence to support
this
assertion. Breastfeeding is associated with reduced risk for
a number of neonatal infections including gastro-intestinal
infec-
tions, diarrhoeal infections, and types of extra-intestinal
infections
(Jackson & Nazar, 2006).
The claim that breastfeeding may protect against obesity in
childhood and later life is less well established. Although two
separate reviews of the literature (Arenz, Ruckerl, Koletzko, &
von
Kries, 2004; Owen, Martin, Whincup, Davey Smith, & Cook,
2005)
have concluded that having been breastfed as an infant is
associated
with significantly reduced odds of childhood obesity, these
meta-
analyses disguise considerable heterogeneity in findings across
studies. While Arenz and colleagues calculated an OR of 0.78
þ353 1 8632100.
y).
All rights reserved.
(95% CI: 0.71e0.85) across the nine studies which met their
criteria

for inclusion, careful scrutiny of the pattern of results reveals
that of
the seven studies that included a measure of parental weight
status,
three reported a statistically significant protective effect of
breast-
feeding (Bergmann et al., 2003; Gillman et al., 2001; Toschke
et al.,
2002) and four found that there was no statistically significant
effect
(Hediger, Overpeck, Kuczmarski, & Ruan, 2001; Li, Parsons, &
Power,
2003; O’Callaghan, Williams, Andersen, Bor, & Najmans, 1997;
Poulton & Williams, 2001); although the point estimates for all
but
the study by Li et al. suggested a protective effect.
A subsequent review by Owen et al. (2005) showed that the
pooled OR across six studies was markedly reduced when
adjusted
for socio-economic status, parental BMI and maternal smoking
e
decreasing from 0.86 (95% CI: 0.81e0.91) to 0.93 (95% CI:
0.88e0.99) e but remaining significant. The most heavily
weighted
of these was the study by Grummer-Strawn and Mei (2004),
which
involved 177,304 children up to 5 years of age. However, this
study
only had important covariate information (mother’s age, educa-
tional attainment, mother’s self-reported pre-pregnancy weight,
measured height, weight gain during pregnancy, and post-
partum
smoking) for a subset of the sample (n ¼ 12,587), and crucially,
residual confounding cannot not be ruled out.

mailto:[email protected]
www.sciencedirect.com/science/journal/02779536
http://www.elsevier.com/locate/socscimed
http://dx.doi.org/10.1016/j.socscimed.2012.02.048
C. McCrory, R. Layte / Social Science & Medicine 75 (2012)
323e330324
A further review by Harder, Bergmann, Kallischnigg, &
Plagemann (2005), which included only those studies where the
odds ratio, 95% confidence interval and duration of
breastfeeding
were reported and which used exclusively formula fed infants as
the reference group, also concluded that breastfeeding was
protective against obesity with the results of their meta-
regression
indicating a clear doseeresponse effect in the data. Each month
of
breastfeeding was associated with a 4% reduction in risk of
over-
weight averaged across the 17 studies which met their criteria
for
inclusion. Again though, these studies varied widely in the list
of
confounders adjusted for, with only five of the studies including
a control for parental BMI. If we consider only those studies
which
included adjustment for parental BMI, we find that four of these
(i.e. Hediger et al., 2001; Parsons, Power, & Manor, 2003;
Poulton &
Williams, 2001; Wadsworth, Marshall, Hardy, & Paul, 1999)
did not
find any statistically significant effect of breastfeeding when
adjusted for confounding factors.

Failure to adjust for parental weight status may be an important
shortcoming since parental BMI has been shown to be amongst
the
strongest determinants of childhood overweight (Danielzik,
Langnase, Mast, Spethmann, & Muller, 2002; Li, Law, Lo
Conte, &
Power, 2009), reflecting the contribution of shared genes and
shared environment. What is more, studies have shown that
women who are overweight or obese are less likely to breastfeed
(Amir & Donath, 2007; Li, Jewell, & Grummer-Strawn, 2003).
Parental weight status is correlated with a range of familial (e.g.
shared diet) and environmental variables (e.g. lifestyle factors)
that
may mediate the association with childhood overweight. Parents
directly influence the types and varieties of foodstuffs to which
children are exposed. Research shows that children and parents’
dietary intakes are correlated for most nutrients (Oliveria et al.,
1992: cited in Taylor, Evers, & McKenna, 2005); mothers with
higher BMI are more likely to give their children low nutrient
snacks and to consume more fat as a proportion of food intake
(Davison & Birch, 2001). A U.S. study of 2149 children aged
9e19
years participating in the National Health and Growth Study
found that the percentage of kilocalories from fat was inversely
related to parental education and family income (Crawford,
Obarzanek, Schreiber et al., 1995). Studies of household food
purchases also generally report a positive association between
household SES and the quality and variety of purchased foods
(Darmon & Drewnowski, 2008). Similarly, studies have docu-
mented an inverse association between parental BMI and rates
of
physical activity in adolescents (Kahn et al., 2008; Williams &
Mummery, 2011), which suggests that parental BMI may serve
as
a proxy for other lifestyle-related behaviours that are associated

with rates of obesity.
The present study used data from the first wave of the Growing
Up in Ireland study, a large nationally representative study of
Irish
school-children to explore the relationship between
breastfeeding
exposure and levels of overweight and obesity at nine-years of
age
controlling for a wide range of potentially confounding
variables.
Method
Sample
The sample comprised 8568 nine-year-old school-children
participating in the Growing Up in Ireland (GUI) study, a
nationally
representative cohort study of children living in the Republic of
Ireland. The sample was selected through a two-stage sampling
method within the national school system. Eligible children
were those who were born between 1st November 1997 and
31st October 1998. In the first stage, 1105 primary schools from
the national total of 3177 were selected using a probability
proportionate to size (PPS) sampling method. In the second
stage,
a random sample of eligible children was selected within each
school. At the school level, a response rate of 82.3% was
achieved,
while at the level of the household (i.e. eligible child selected
within the school) a total of 57% of children and their families
participated in the study. Interviews were carried out with the
teacher and parents of the study child. Fieldwork for the school-
based component was carried out between MarcheNovember
2007, while fieldwork for the home-based phase of data collec-

tion ran from July 2007eJuly 2008. The data were weighted
prior to
analysis to account for the complex sampling design, which
involves the structural adjustment of the sample to the
population
using Census of Population statistics while maintaining the case
base of 8568 children. More detailed information about the
sample
selection process and derivation of weights is contained in the
sampling document that accompanies the anonymised microdata
file (ISSDA, 2010). All stages of the Growing Up in Ireland
project
were approved by the Health Research Board’s standing
Research
Ethics Committee based in Dublin.
Measures
Breastfeeding measure
Information relating to breastfeeding initiation and duration
was obtained retrospectively when the child was nine-years of
age
via parental recall. Parents were asked about whether the child
was
ever breastfed, even if only for a short time, as well as the total
number of weeks for which the child was breastfed. Duration of
breastfeeding in weeks was grouped into a 6 level ordered cate-
gorical variable: never breastfed, breastfed for 4 weeks or less,
breastfed for 5e8 weeks, breastfed for 9e12 weeks, breastfed for
13e25 weeks, and breastfed for 26 weeks or more. Although
individual validation of breastfeeding information to an outside
source was not possible, analysis of hospital records on the
proportion of mothers breastfeeding at discharge following birth
for the period during which the study children were born shows
strong concurrence by maternal characteristics. Li, Scanlon, and

Serdula (2005) examined the validity and reliability of maternal
recall of breastfeeding practice across 11 studies with variable
recall
periods. They found that retrospective report could yield
accurate
estimates of breastfeeding initiation and duration, particularly
when the recall period was within the first three years. Very few
studies have examined the validity of maternal recall over more
extended periods, though one study found strong concurrence
for
initiation (85% correctly identified) when infant clinic records
were
compared with retrospective report 15 years after the event, but
that recall of breastfeeding duration was lower with 37%
accurately
recalling to within one month and 59% accurately recalling to
within two months (Tienboon, Rutishauser, & Wahlqvist, 1994).
Nevertheless, Li et al. (2005) estimated that the mean difference
in
breastfeeding duration between recall and the validation
standard
with a recall period of 6 months was less than a week and
increased
to 5 weeks with a recall period of 14e15 years.
Measurement of BMI
Height and weight measurements were obtained from the
primary and secondary caregiver as well as the study child as
part of
the household interview by trained interviewers using
scientifically
calibrated measuring instruments. Weight measurements were
recorded to the nearest 0.5 kg using a SECA 761 medically
approved
(Class IIII) flat mechanical scale that graduated in 1 kg

increments
and had an upper capacity of 150 kg. Height was recorded to the
nearest millimetre using a Leicester portable height stick.
Respon-
dents were asked to remove footwear, headwear and any heavy
clothing prior to being measured. The data were screened by the
GUI
323e330 325
data management team for biologically implausible data prior to
deposit in the archive and extreme outliers were set to missing.
Valid height and weight measurements were obtained in respect
of
94.5 percent of the sample of children.
Definition of overweight and obesity in early childhood
Body mass index (BMI) is the most widely used method for
assessing the degree of adiposity in the general population. It is
calculated by dividing weight in kilograms by height in metres
squared and has been shown to correlate strongly with measures
of
body fat obtained using direct physiological assessment
(Lindsay
et al., 2001). There are no universally agreed thresholds for
defining overweight and obesity in child and adolescent pop-
ulations as BMI cut-offs have to be standardised for age,
ethnicity
and gender. Cole, Bellizzi, Flegal, and Dietz (2000) pooled data
from
six international studies and employed a smoothing procedure to
develop age and gender specific cut-offs that dissected the 25
and

30 kg/m2 at 18 years of age. As children could be interviewed
at any
stage between their ninth and tenth year of age, the IOTF cut-
offs
for children aged 9.5 years were used in the present analysis.
This
definition of overweight and obesity has the obvious and
desirable
benefit of providing internationally comparable estimates of
prevalence.
Covariates
Child variables. A wide range of child, family, cultural and
social
variables have been found to influence both the propensity to
breastfeed and children’s BMI status. We chose control
variables on
the basis of their association with obesity or breastfeeding in
the
literature, as informed by the most recent reviews of the subject
(e.g. Kleiser, Rosario, Mensink, Prinz-Langenohl, & Kurth,
2009;
Reilly et al., 2005). In addition to the gender of the study child,
parent-reported child variables included birth weight in
kilograms,
which is represented as a dichotomous variable (<2500 g/
�2500 g), gestational period, which is represented as a four
level
variable (late (42 weeks or more), on-time (37e41 weeks), early
(33e36 weeks), very early (32 weeks or less)), the study child’s
screen time, represented as a four level variable (none/less than
an
hour/1 h to less than 3 h/more than 3 h). Childhood physical
activity
level was indexed using a question which asked on how many
occasions in the past 14 days the child had done exercise hard

enough to make him/her breathe heavily and make his/her heart
beat faster, which is represented as a 5 level variable: none/1e2
days/3e5 days/6e8 days/9 or more days. Finally, dietary intake
was indexed using a semi-quantitative food frequency question-
naire that asked the primary carer to recall whether the study
child
consumed each type of food, once, more than once or not at all
in
the 24 hour period preceding the interview. We summarised the
overall difference in dietary quality by combining the different
types of food consumption into a single index of dietary quality
with lower scores indicating worse dietary quality. We did this
by
assigning positive values (1 ¼ eaten once, 2 ¼ more than once)
to
foods seen as beneficial (such as fresh fruit, cooked vegetables,
raw
vegetables/salad) and a negative value to those generally seen
as
less beneficial (burger, sausage, chips, crisps etc). The range of
scores varied from �0.55 to þ0.70 with a mean of 0.11 and a
stan-
dard deviation of 0.17. We then developed a categorical
variable by
partitioning our measure of dietary quality into tertiles with
lower
scores indicating worse dietary quality.
Parental variables. Adult weight status is indexed using the
stan-
dard GarroweWebster cut-offs with BMI� 25.0 and less than
30.0
defining overweight and BMI in excess of 30.0 defining obesity.
Some 4.4 percent of primary carers’ and 5.4 percent of
secondary
carers’ (in instances in which there was a resident secondary

carer)
did not have their weight measured during the course of the
household interview. Tests showed that these were not missing
at
random and thus a missing code was used in analysis for this
group.
Maternal prenatal smoking status was captured via parental
recall
at nine-years of age and is represented as a three level variable
(never smoked, smoked occasionally during pregnancy, smoked
daily during pregnancy). Given the small numbers involved in
some
of the ethnic categories, we use Irish/non-Irish background as
a proxy for this.
Socio-economic characteristics of the household. Three
different
measures of socio-economic status of the child’s household are
used in the analysis: primary and secondary carer’s social class,
primary carer’s education and household income. Household
Social
Class was measured using the Irish Central Statistics Office’s
social
class schema and coded using the International Standard
Classifi-
cation of Occupations 1988 (ISCO88). Household social class is
established using a dominance procedure. This meant that in
two-
parent families where both members of the household were
economically active, the family’s social class group was
assigned as
the higher of the two. Primary Carer’s Level of Education was
rep-
resented as a four category variable: lower secondary education
or
less, higher secondary education, post-secondary education, and

third level education. Self-reported household net income was
adjusted for household size and composition using the modified
Organisation for Economic Co-Operation and Development
(OECD)
equivalence scale and is represented as income quintiles. The
primary caregiver’s employment status was indexed using
a dichotomous variable (not in FT work/in FT work).
Missing cases analysis
Non-biological parents and fathers completing the question-
naire were not asked the questions relating to whether the child
was breastfed during infancy (n ¼ 211). Overall, the degree of
missing data was small for most covariates. The exception was
household income, which was missing for 626 cases.
Consequently,
missing values on household income were imputed using the
Multiple Imputation UVIS programs implemented in STATA by
Royston (2004). Thus, the effective case base for the analyses
that
follow was 7798. Inferential statistics reported in the tables
have
been weighted to take account of the complex survey design.
Results
Mean BMI for the sample of 7798 nine-year-old children was
17.97 (S.D. ¼ 3.13). The estimated proportion of children in the
non-
overweight, overweight, and obese categories according to the
IOTF
cut-offs was 74.3%, 19.0% and 6.6% respectively. Table 1
shows the
odds of being classified as overweight or obese for children
who
were breastfed for variable durations during infancy relative to

those who were never breastfed. In unadjusted analysis, with
breastfeeding treated as an ordered categorical variable repre-
senting varying durations of breastfeeding exposure,
breastfeeding
for 5 weeks or more was associated with significantly reduced
odds
of being obese at nine-years of age and a clear doseeresponse
relationship was evident in the data. Those who were breastfed
for 5e8 weeks were 47 percent less likely to be obese compared
with those who were never breastfed (OR ¼ 0.53 CI.95 ¼
0.32e0.89),
increasing through 58 percent for those breastfed for between 9
and 12 weeks (OR ¼ 0.42 CI.95 ¼ 0.24e0.73), and 13e25 weeks
(OR ¼ 0.42 CI.95 ¼ 0.27e0.64), and 62 percent for those
breastfeed
in excess of 26 weeks (OR ¼ 0.38 CI.95 ¼ 0.24e0.62). There
was no
statistically significant protective effect of breastfeeding
against
risk of overweight in the crude model.
Table 1
Mean BMI and the probability of being overweight or obese at
nine-years of age by breastfeeding status in the crude
multinomial model.
Duration Unweighted n BMI (s.d.) Overweight Obese
Weighted % OR (95% CI) Weighted % OR (95% CI)
Never breastfed 3788 18.18 (3.33) 20.1 1.00 8.1 1.00
Breastfed 4 weeks or less 964 18.05 (3.06) 17.3 0.83
(0.66e1.03) 7.9 0.94 (0.64e1.37)
Breastfed 5e8 weeks 568 17.62 (2.86) 18.1 0.84 (0.63e1.12) 4.6

0.53 (0.32e0.89)*
Breastfed 9e12 weeks 623 17.75 (2.65) 17.8 0.81 (0.62e1.07)
3.7 0.42 (0.24e0.73)**
Breastfed 13e25 weeks 926 17.65 (2.73) 18.4 0.85 (0.67e1.07)
3.7 0.42 (0.27e0.64)***
Breastfed 26 wksþ 929 17.41 (2.77) 17.3 0.78 (0.60e1.01) 3.4
0.38 (0.24e0.62)***
Reference category on the dependent variable: non-overweight.
* significant at the 0.05 level ** significant at the 0.01 level
*** significant at the 0.001 level.
323e330326
Table 2 shows the probability of being overweight/obese at
nine-years of age treated as a binary variable (non-overweight
vs
overweight/obese) and the probability of having been breastfed
as
an infant classified as a binary variable (ever vs never)
according to
important characteristics of the sample. It is evident that rates
of
overweight/obesity and the probability that a child will have
been
breastfed as an infant are strongly associated with socio-
economic
characteristics of the household, and with parental weight
status.
Other significant predictors included gestational age, prenatal
smoking, and child level variables such as the frequency of hard
exercise and children’s screen time.
To determine whether breastfeeding remained protective
against childhood overweight and obesity in a multivariate
model

when considered alongside other putative confounding
variables,
we performed a multinomial logistic regression analysis using
forced entry and robust standard errors to estimate the effect of
variable durations of breastfeeding on the probability of being
overweight or obese controlling for all other variables in the
anal-
ysis. The choice of variables to be used in the multivariate
model
was dependent on their association with breastfeeding in Table
2.
The derived estimates are expressed as Adjusted Odds Ratios
(AOR)
relative to the baseline category (i.e. non-overweight). Table 3
shows the relationship between breastfeeding exposure and risk
of overweight and obesity in the full multivariate model,
control-
ling for gestational age, nationality, prenatal smoking, maternal
education, household social class, household income, frequency
of
hard exercise, screen time, the study child’s dietary quality, and
parental weight status. The final model revealed that
breastfeeding
for 13 weeks or more was associated with significantly reduced
odds of being obese controlling for other factors. Although
there
was a trend towards a doseeresponse effect in the data, with
breastfeeding in excess of one month associated with decreasing
odds of being obese, the relationship was statistically
significant
only for those who breastfed for 13 weeks or more.
Breastfeeding
for between 13 and 25 weeks was associated with a 38 percent
reduction in the risk of obesity (AOR ¼ 0.62 CI.95 ¼
0.39e0.99;
p < 0.05) in the full multivariable adjusted model while breast-

feeding for 26 weeks or more was associated with a 51 percent
reduction in the risk of obesity (AOR ¼ 0.49 CI.95 ¼
0.29e0.82;
p < 0.01) at nine-years of age.
Discussion
This study sought to examine whether being breastfed during
infancy was protective against overweight and obesity at nine-
years of age using data from a large, nationally representative
cohort study in the Republic of Ireland. In agreement with the
results of other epidemiologic studies, our analyses indicate that
being breastfed for a period in excess of 13 weeks during
infancy
was associated with a significantly reduced risk of being obese
at
nine-years of age after controlling for a wide range of potential
confounding variables including parental overweight status.
Breastfeeding for between 13 and 25 weeks was associated with
a 38 percent reduction in the risk of obesity at nine-years of age
in the full multivariable adjusted model, while being breastfed
in
excess of 26 weeks was associated with a 51 percent reduction
in
risk of obesity. While being breastfed for less than this amount
of
time was not associated with any statistically significant protec-
tive effect, the results pointed towards a doseeresponse rela-
tionship for children who were breastfed for more than four
weeks. It could be argued that the finding of a doseeresponse
gradient in the data adds to our confidence in a causal relation-
ship as it becomes increasingly difficult to envisage how some
unobserved variable could explain away the protective effect of
breastfeeding at different levels of exposure. Being breastfed
was
not associated with any statistically significant reduction in the

risk of being classified as ‘overweight’ for any duration of
expo-
sure. Why breastfeeding should be protective only at the higher
end of the BMI distribution is clearly a topic that requires
further
examination in subsequent studies (see Beyerlein & Von Kries,
2011).
While the mechanism conferring this protective effect is not
well understood, a number of tentative theories have been
advanced to account for this phenomenon, which can be broadly
characterised as (1) nutritional and (2) behavioural
explanations.
The first of these suggests that differences in the composition of
human breast milk are protective against the development of
obesity. The growth acceleration hypothesis (see Singhal &
Lanigan,
2007) holds that the protective effect of breastfeeding is a result
of
a slower pattern of growth among breastfed children relative to
those who were bottle fed. Consistent with this proposition,
anthropometric studies of early infant growth patterns have
established that children who are breastfed gain height and
weight
more slowly than those who were bottle fed (Ong et al., 2002;
Ziegler, 2006), and that the extent of the divergence is such that
it
can amount to a difference of 600e650 g by one year of age
(Dewey,
1998). It has been suggested that rapid weight gain in early life
defined by early centile crossing may predispose to later
metabolic
risk by bringing forward the timing of the adiposity rebound
(Taylor, Grant, Goulding & Williams, 2005) and a number of
longitudinal studies have found that the velocity with which
infants cross weight-for-age reference centiles is related to later

cardiovascular and metabolic risk (Ong et al., 2000; Stettler,
Zemel,
Kumanyika, & Stallings, 2002).
Given that the energy density of infant formula can be anything
from 10 to 18 percent higher compared with breast milk
(Heinig,
Nommsen, Peersen, Lonnerdal, & Dewey, 1993), this represents
a plausible etiologic pathway. An alternative hypothesis is that
it is
the protein density, as opposed to the energy density of infant
formula that is causal to increased rates of adiposity in children.
Again, research has shown that the concentration of protein is
much higher in infant formula compared with breast milk
(Darragh
& Moughan, 1998; Feng et al., 2009). Some studies have
indicated
that it is high intakes of protein, rather than high intakes of
energy,
Table 2
Independent association of each of the potential confounding
variables with the probability of being overweight/obese at
nine-years of age and having been breastfed as an
infant using logistic regression analysis
Variable Overweight/Obese Breastfed Unweighted n
Prevalence (%) O.R. (95% CI) Prevalence (%) O.R. (95% CI)
Breastfeeding status Never Breastfed 28.1 1.00 - - 3788
Breastfed 4 wks or less 25.2 0.86 (0.70-1.06) - - 964
Breastfed 5-8 wks 22.7 0.75 (0.58-0.97)* - - 568
Breastfed 9-12 wks 21.5 0.70 (0.54-0.90)** - - 623

Breastfed 13-25 wks 22.1 0.73 (0.59-0.90)** - - 926
Breastfed 26 wks + 20.8 0.67 (0.53-0.85)*** - - 929
Child’s Gender Male 22.0 1.00 45.5 1.00 3761
Female 29.5 1.48 (1.30-1.68)*** 44.0 0.94 (0.84-1.06) 4037
Birth-weight BW >¼2500 grams 28.1 1.00 45.0 1.22 (0.93-1.60)
367
BW <2500 grams 25.2 1.12 (0.81-1.55) 40.2 1.00 7431
22.7
Gestation Late (>¼42 wks) 21.5 1.26 (1.09-1.47)** 43.4 0.88
(0.77-1.01) 1895
On-timey (37-41 wks) 22.1 1.00 46.6 1.00 4912
Early (33-36 wks) 20.8 0.93 (0.76-1.13) 38.0 0.70 (0.58-
0.85)*** 873
V. early (<¼32 wks) 1.13 (0.65-1.96) 47.4 1.03 (0.67-1.61) 118
22.0
Nationality Irish 29.5 1.00 40.9 1.00 6555
Non-Irish 0.94 (0.79-1.13) 66.1 2.82 (2.35-3.38)*** 1243
28.1
Prenatal Smoking Never smoked 25.2 1.00 51.5 1.00 6064
Occasionally smoked 22.7 1.30 (1.06-1.59)* 36.0 0.53 (0.44-
0.65)*** 736
Daily smoked 21.5 1.37 (1.15-1.64)*** 22.2 0.27 (0.22-
0.33)*** 998
22.1
Maternal Education Lower secondary 20.8 1.82 (1.49-2.21)***
24.1 0.11 (0.09-0.13)*** 1344
Higher secondary 1.41 (1.18-1.69)*** 42.3 0.25 (0.21-0.29)***

2479
Post-secondary 22.0 1.36 (1.13-1.65)*** 57.1 0.45 (0.38-
0.53)*** 1960
Third level 29.5 1.00 74.8 1.00 2015
Household Social Class Unclassified 28.1 1.62 (1.14-2.29)**
37.1 0.28 (0.21-0.37)*** 366
Unskilled 25.2 2.19 (1.37-3.49)*** 25.9 0.17 (0.10-0.27)***
124
Semi-skilled 22.7 2.55 (1.90-3.43)*** 30.0 0.20 (0.16-0.26)***
551
Skilled manual 21.5 1.96 (1.50-2.55)*** 34.9 0.25 (0.20-
0.31)*** 1092
Non-manual 22.1 1.85 (1.44-2.37)*** 38.8 0.30 (0.25-0.36)***
1545
Managerial & Technical 20.8 1.47 (1.16-1.88)** 54.6 0.57
(0.47-0.68)*** 3052
Professional Managers 1.00 68.0 1.00 1068
22.0
Income Quintile Lowest 29.5 1.39 (1.14-1.70)*** 35.6 0.39
(0.32-0.47)*** 1550
2nd 1.22 (0.99-1.49) 41.1 0.49 (0.41-0.59)*** 1573
3rd 28.1 1.34 (1.10-1.63)** 45.7 0.59 (0.50-0.71)*** 1575
4th 25.2 1.30 (1.06-1.59)* 51.5 0.75 (0.62-0.90)*** 1565
Highest 22.7 1.00 58.7 1.00 1535
21.5
Employment Status Not working FT 22.1 1.00 42.1 1.00 3317
Working FT 20.8 1.09 (0.96-1.24) 47.1 1.23 (1.09-1.38)***
4481
Frequency of Hard Exercise none 22.0 2.26 (1.53-3.33)*** 35.7
0.67 (0.45-0.97)* 167

1-2 days 29.5 1.68 (1.26-2.26)*** 41.1 0.84 (0.64-1.08) 397
3-5 days 1.67 (1.42-1.97)*** 43.8 0.93 (0.80-1.08) 1336
6-8 days 28.1 1.26 (1.07-1.48)*** 45.9 1.01 (0.88-1.17) 1594
9 or more days 25.2 1.00 45.5 1.00 4304
22.7
TV Viewing Hrs Noney 21.5 1.00 71.8 1.00 193
<1 hr 22.1 3.65 (1.90-7.01)*** 51.1 0.41 (0.27-0.63)*** 1849
1 to <3 hrs 20.8 4.36 (2.31-8.25)*** 43.9 0.31 (0.20-0.47)***
5047
3 hrs or more 5.91 (3.06-11.43)*** 32.7 0.19 (0.12-0.30)***
709
22.0
Dietary Quality Low 29.5 0.89 (0.76-1.04) 33.4 0.36 (0.31-
0.42)*** 2360
Medium 0.91 (0.79-1.06) 44.0 0.56 (0.49-0.64)*** 2783
High 28.1 1.00 58.4 1.00 2655
25.2
Parental Weight Status Neither parent overweight/obese 22.7
1.00 55.1 1.00 798
One parent overweight/obese 21.5 2.03 (1.48-2.79)*** 47.3
0.73 (0.60-0.90)** 2877
Both parents overweight/obese 22.1 4.45 (3.26-6.07)*** 43.2
0.62 (0.51-0.77)*** 2508
Mum not measured 20.8 3.76 (2.48-5.70)*** 45.8 0.69 (0.51-
0.93)* 345
Dad not measured 3.73 (2.52-5.54)*** 40.8 0.56 (0.41-0.76)***
424
No resident partner 22.0 3.50 (2.47-4.97)*** 38.0 0.50 (0.39-
0.65)*** 846

* significant at the 0.05 level
** significant at the 0.01 level
*** significant at the 0.001 level
323e330 327
Table 3
Results of the multinomial logistic regression analysis
expressing the odds of being overweight or obese by various
risk factors in the full multivariable model.
Variable Overweight Obese
Adjusted odds ratio Sig. Adjusted odds ratio Sig.
Breastfeeding status Never breastfed 1.00 e 1.00 e
Breastfed 4 wks or less 0.87 (0.69e1.09) ns 1.05 (0.71e1.55) ns
Breastfed 5e8 wks 0.90 (0.67e1.21) ns 0.68 (0.40e1.15) ns
Breastfed 9e12 wks 0.95 (0.71e1.26) ns 0.61 (0.34e1.09) ns
Breastfed 13e25 wks 1.01 (0.78e1.30) ns 0.62 (0.39e0.99) p <
0.05
Breastfed 26 wksþ 0.88 (0.67e1.15) ns 0.49 (0.29e0.82) p < 0.01
Gestation Late (�42 wks) 1.14 (0.96e1.35) ns 1.32 (1.00e1.74)
ns
On-timea (37e41 wks) 1.00 e 1.00 e
Early (33e36 wks) 0.86 (0.69e1.07) ns 0.85 (0.55e1.29) ns
V. early (�32 wks) 0.64 (0.30e1.36) ns 1.91 (0.84e4.35) ns
Nationality Non-Irisha 1.01 (0.82e1.25) ns 1.25 (0.88e1.78) ns
Prenatal smoking Never smoked 1.00 e 1.00 e

Occasionally smoked 1.34 (1.05e1.72) p < 0.05 1.18 (0.80e1.74)
ns
Daily smoked 1.26 (1.01e1.57) p < 0.05 1.33 (0.96e1.84) ns
Maternal Education Lower secondary 1.20 (0.90e1.59) ns 1.92
(1.13e3.28) p < 0.01
Higher secondary 1.10 (0.87e1.38) ns 1.51 (0.92e2.48) ns
Post-secondary 1.12 (0.89e1.41) ns 1.62 (0.99e2.65) ns
Third level 1.00 e 1.00 e
Household social class Unclassified 1.01 (0.61e1.66) ns 2.67
(1.20e5.93) p < 0.05
Unskilled 1.73 (0.99e3.02) ns 2.64 (0.97e7.17) ns
Semi-skilled 1.69 (1.17e2.43) p < 0.01 5.13 (2.68e9.79) p <
0.001
Skilled manual 1.47 (1.05e2.05) p < 0.05 4.01 (2.19e7.36) p <
0.001
Non-manual 1.24 (0.91e1.68) ns 3.06 (1.70e5.52) p < 0.001
Managerial & Technical 1.19 (0.91e1.56) ns 2.76 (1.60e4.75) p
< 0.001
Professional Managers 1.00 e 1.00 e
Income Quintile Lowest 0.86 (0.66e1.11) ns 1.17 (0.74e1.85) ns
2nd 0.87 (0.68e1.11) ns 0.81 (0.51e1.29) ns
3rd 0.99 (0.78e1.26) ns 1.02 (0.65e1.60) ns
4th 0.98 (0.79e1.23) ns 1.34 (0.84e2.14) ns
Highest 1.00 e 1.00 e
Employment status Working FTa 1.20 (1.00e1.43) p < 0.05 1.21
(0.91e1.61) ns
Frequency of hard exercise None 1.57 (0.97e2.55) ns 4.53
(2.60e7.90) p < 0.001
1e2 days 1.25 (0.89e1.76) ns 2.71 (1.71e4.28) p < 0.001
3e5 days 1.45 (1.20e1.75) p < 0.001 2.52 (1.85e3.41) p < 0.001
6e8 days 1.18 (0.98e1.43) ns 1.63 (1.20e2.21) p < 0.01

9 or more days 1.00 e 1.00 e
TV viewing hrs None 1.00 e 1.00 e
<1 h 3.21 (1.51e6.81) p < 0.01 3.55 (0.79e15.92) ns
1 to <3 h 3.25 (1.55e6.82) p < 0.01 5.10 (1.17e22.20) p < 0.05
3 h or more 4.18 (1.93e9.04) p < 0.001 5.48 (1.18e25.49) p <
0.05
Dietary quality Low 0.66 (0.55e0.80) p < 0.001 0.68
(0.50e0.92) p < 0.05
Medium 0.80 (0.68e0.96) p < 0.05 0.82 (0.62e1.08) ns
High 1.00 e 1.00 e
Parental weight status Neither parent overweight/obese 1.00 e
1.00 e
One parent overweight/obese 1.90 (1.35e2.68) p < 0.001 3.20
(1.43e7.15) p < 0.01
Both parents overweight/obese 3.60 (2.57e5.05) p < 0.001 9.50
(4.40e20.51) p < 0.001
Mum not measured 3.05 (1.93e4.81) p < 0.001 10.12
(4.08e25.14) p < 0.001
Dad not measured 2.78 (1.80e4.30) p < 0.001 7.79 (3.30e18.38)
p < 0.001
No resident partner 3.09 (2.07e4.61) p < 0.001 6.84
(2.97e15.75) p < 0.001
Pseudo R2 0.061
Reference category on the Dependent Variable: Non-
overweight.
a Reference categories on the dichotomous Independent
Variables: Irish background, not in FT employment.
323e330328
fat or carbohydrates that predict early adipose rebound and
higher

BMI in childhood (Scaglioni et al., 2000). Most of this research
is
summarised in the excellent paper by Koletzko et al. (2009).
Other
investigators have suggested that it is not breastfeeding per se,
but
rather, the delayed introduction of complementary foods that
may
be protective against the development of obesity in later life
(Ong,
Emmett, Noble, Ness, & Dunger, 2006; Schack-Nielson,
Mortensen,
& Michaelsen, 2010; Wilson et al., 1998). Alternatively, it
could be
that bioactive compounds such as leptin or ghrelin which have
a role in satiety and regulation of hunger, occur naturally in
human
breast milk and are absent in infant formula that underlies the
association (see Lawrence, 2010).
Behavioural explanations, by contrast, postulate that the
method of infant feeding may lead to different behavioural
patterns
among breastfed and bottle fed infants, resulting in a
predisposi-
tion towards obesity risk in later life. Much of this evidence is
summarised in Bartok and Ventura (2009). For example, one
study
showed how dietary intake patterns varied across groups:
breast-
fed children consumed a large feed in the morning followed by
smaller feeds over the course of the day, while bottle fed infants
consumed the same quantity at regular intervals, suggesting that
parental control rather than hunger cues might be driving infant
feeding behaviour (Wright, Fawcett, & Crow, 1980).
Breastfeeding

mothers, by contrast, may be more responsive to children’s cues
indicating satiety. Consistent with such a hypothesis, several
studies have shown that children are able to moderate their
consumption of formula feed or breast milk when energy
density is
increased (Fomon, Filer, Thomas, Anderson, & Nelson, 1975).
Furthermore, it has been hypothesised that breastfed children
may
also regulate the milk production of their mother (e.g.
Bergmann
et al., 2003). A recent retrospective study, comparing children
fed
human breast milk directly via the breast (as opposed to
indirectly
323e330 329
with the bottle), found that the method of feeding could have
lasting effects on appetite regulation (Di Santis et al., 2011).
Nevertheless, the possibility cannot be ruled out that some
other unmeasured factor accounts for the association and that
breastfeeding simply serves as a marker, albeit a powerful
marker,
of other nutritional or lifestyle-related exposures. In trying to
locate
this study within the broader framework of research examining
the
benefits of breastfeeding on childhood BMI, it should be
acknowl-
edged that this study has a number of limitations. Although we
have demonstrated support for the idea that breastfeeding
during
infancy for a period in excess of 13 weeks is protective against

obesity in middle childhood, an obvious limitation is that the
study
is cross-sectional in nature, examining BMI at only one time
point.
This is an important qualification because some investigators
have
speculated that the protective effect of breastfeeding against
obesity is weak in early childhood and may not manifest until
later
in childhood (e.g. Dewey, 2003; Dietz, 2001). Other studies
provide
tentative support for this proposition (Gillman et al., 2001;
Poulton
& Williams, 2001), although a recent study by O’Tierney,
Barker,
Osmond, Kajantie, and Eriksson (2009), which followed a birth
cohort until 60 years of age has complicated the issue further.
O’Tierney’s group analysed BMI and obesity within sibling
pairs
discordant for breastfeeding duration and found that a longer
period of breastfeeding was associated with lower BMI at one
year
of age, but this effect had disappeared by 7 years of age. At 60
years
of age, being breastfeed for 8 months or longer or for less than
2
months was associated with increased BMI. The reason for these
age related variations in obesity risk is an interesting avenue for
empirical investigation, ideally in a longitudinal context
employing
an exclusively breastfed reference group.
Although we used breastfeeding duration as a proxy for dose,
this obscures considerable heterogeneity in breastfeeding
exposure
across individuals. It would be anticipated for example that a

child
breastfed exclusively for 6 months would receive a higher dose
of
breast milk on average compared with those who used comple-
mentary feeding methods or transitioned to other milks or solid
foods earlier. We also lacked a measure of the child’s age at
time of
transition to solid foods and were thus unable to examine the
claim
that it is the delayed introduction of solid foods rather than the
nutritional or bioactive properties of breast milk that is
protective
against obesity.
A real strength of the current study is the large and represen-
tative nature of the sample, which accounts for approximately
1/7
of all children born in Ireland between 1997 and 1998. In
addition,
the reasonably proportionate split between those who were
breastfed and those who were not allows for the estimation of
robust main effects. We were able to control for multiple
possible
confounding factors including the child’s low dietary quality,
screen
time and frequency of hard exercise. Perhaps most importantly,
we
could control for parental BMI, which reflects the confounding
influence of genetic inheritance and environment.
The suggestion that the choice of infant feeding method has
important implications for child health and development is
tanta-
lising as it identifies a modifiable health behaviour that is
amenable
to intervention and has the potential to improve the health of

the
population (Lawrence, 2010). However, further experimental
research is required to elucidate the causal mechanism and to
establish why the effect is apparent at certain ages and not
others.
Acknowledgements
The Growing Up in Ireland data have been funded by the
Government of Ireland through the Department of Children and
Youth Affairs; have been collected under the Statistics Act,
1993, of
the Central Statistics Office. The project has been designed and
implemented by the joint ESRI-TCD Growing Up in Ireland
Study
Team.
In addition to the funders, the authors would like to thank the
entire Growing Up in Ireland Project and Study teams, and the
children and families who participated in the study.
References
Amir, L. H., & Donath, S. (2007). A systematic review of
maternal obesity and
breastfeeding intention, initiation and duration. BMC Pregnancy
and Childbirth,
7, 9.
Arenz, S., Ruckerl, R., Koletzko, B., & von Kries, R. (2004).
Breastfeeding and child-
hood obesity: a systematic review. International Journal of
Obesity and Related
Metabolic Disorders, 28, 1247e1256.
Bartok, C., & Ventura, A. K. (2009). Mechanisms underlying
the association between

breastfeeding and obesity. International Journal of Pediatric
Obesity, 4, 196e204.
Bergmann, K. E., Bergmann, R. L., von Kries, R., Bohm, O.,
Richter, R.,
Dudenhausen, J. W., et al. (2003). Early determinants of
childhood overweight
and adiposity in a birth cohort study: role of breastfeeding.
International Journal
of Obesity and Related Metabolic Disorders, 27, 162e172.
Beyerlein, A., & Von Kries, R. (2011). Breastfeeding and body
composition in chil-
dren: will there ever be conclusive empirical evidence for a
protective effect
against overweight? The American Journal of Clinical Nutrition,
94(6 Suppl),
1772Se1775S.
Cole, T. J., Bellizzi, M. C., Flegal, K. M., & Dietz, W. H.
(2000). Establishing a standard
definition for child overweight and obesity worldwide:
international survey.
British Medical Journal, 320, 1240e1243.
Crawford, P. B., Obarzanek, E., Schreiber, G. B., et al. (1995).
The effects of race,
household income, and parental education on nutrient intakes of
9- and 10-year-
old girls: NHLBI Growth and Health Study. Annals of
Epidemiology, 5, 360e368.
Danielzik, S., Langnase, K., Mast, M., Spethmann, C., &
Muller, M. J. (2002). Impact of
parental BMI on the manifestation of overweight 5e7 year old
children. Euro-

pean Journal of Nutrition, 41, 132e138.
Darmon, N., & Drewnowski, A. (2008). Does social class
predict diet quality?
American Journal of Clinical Nutrition, 87(5), 1107e1117.
Darragh, A. J., & Moughan, P. J. (1998). The amino acid
composition of human milk
corrected for amino acid digestibility. British Journal of
Nutrition, 80, 25e34.
Davison, K. K., & Birch, L. L. (2001). Childhood overweight: a
contextual model and
recommendations for future research. Obesity Reviews, 2(3),
159e171.
Dewey, K. G. (1998). Growth characteristics of breastfed
compared to formula fed
infants. Biology of the Neonate, 74, 94e105.
Dewey, K. G. (2003). Is breastfeeding protective against child
obesity? Journal of
Human Lactation, 19, 9e18.
Dietz, W. H. (2001). Breastfeeding may help prevent childhood
overweight. Journal
of the American Medical Association, 285, 2506e2507.
DiSantis, K., Collins, B. N., Fisher, J. O., & Davey, A. (2011).
Do infants fed directly
from the breast have improved appetite regulation and slower
growth during
early childhood compared with infants fed from a bottle?
International Journal
of Behavioural Nutrition and Physical Activity, 8, 89.

Feng, P., Gao, M., Holley, T., Zhou, T., Burgher, A., Trabulsi,
J., et al. (2009). Amino acid
composition and protein content of mature human milk from
nine countries.
The Journal of the Federation of American Societies of
Experimental Biology, J, 23,
LB448.
Fomon, S. J., Filer, L. J., Jr., Thomas, L. N., Anderson, T. A.,
& Nelson, S. E. (1975).
Influence of formula concentration on caloric intake and growth
of normal
infants. Acta Pediatrica Scandinavia, 64, 172e181.
Gillman, M. W., Rifas-Shiman, S. L., Camargo, C. A., Berkey,
C. S., Frazier, A. L.,
Rockett, H. R. H., et al. (2001). Risk of overweight among
adolescents who were
breastfed as infants. Journal of the American Medical
Association, 285,
2461e2467.
Grummer-Strawn, L., & Mei, Z. (2004). Does breastfeeding
protect against Pediatric
overweight? Analysis of longitudinal data from the Centers for
Disease Control
and Prevention Pediatric Nutrition Surveillance System.
Pediatrics, 113, e81ee86.
Harder, T., Bergmann, R., Kallischnigg, G., & Plagemann, A.
(2005). Duration of
breastfeeding and risk of overweight: a meta-analysis. American
Journal of
Epidemiology, 162, 397e403.
Hediger, M. L., Overpeck, M. D., Kuczmarski, R. J., & Ruan,

W. J. (2001). Association
between infant breastfeeding and overweight in young children.
Journal of the
American Medical Association, 285, 2453e2460.
Heinig, M. J., Nommsen, L. A., Peersen, J. M., Lonnerdal, B., &
Dewey, K. G. (1993).
Energy and protein intakes of breastfed and formula-fed infants
during the first
year of life and their association with growth velocity: the
DARLING study.
American Journal of Clinical Nutrition, 58, 152e161.
Irish Social Sciences Data Archive (ISSDA). (2010). Sample
design and response in
wave 1 of the nine-year cohort of Growing Up in Ireland.
http://issda.ucd.ie/
documentation/GUI/GUI-SampleDesign9YearCohort.pdf.
Jackson, K. M., & Nazar, A. M. (2006). Breastfeeding, the
immune response, and
long-term health. The Journal of the American Osteopathic
Association, 106,
203e207.
Kahn, J. A., Huang, B., Gillman, M. W., Field, A. E., Austin, S.
B., Colditz, G. A., et al.
(2008). Patterns and determinants of physical activity in U.S.
adolescents.
Journal of Adolescent Health, 42, 369e377.
http://issda.ucd.ie/documentation/GUI/GUI-
SampleDesign9YearCohort.pdf
http://issda.ucd.ie/documentation/GUI/GUI-
SampleDesign9YearCohort.pdf

323e330330
Kleiser, C., Rosario, A. S., Mensink, G. B. M., Prinz-
Langenohl, R., & Kurth, B. M.
(2009). Potential determinants of obesity among children and
adolescents in
Germany: results from the cross-sectional KiGGS study. BMC
Public Health, 9, 46.
Koletzko, B., von Kries, R., Monasterola, R. C., Subias, J. S.,
Scaglioni, S.,
Giovannini, M., et al.for the European Childhood Obesity Trial
Study Group.
(2009). Can infant feeding choices modulate later obesity risk?
American Journal
of Clinical Nutrition, 89(Suppl. 1), 1502Se1508S.
Lawrence, R. A. (2010). Does breastfeeding protect against
overweight and obesity
in children? A review. Childhood Obesity, 6(4), 193e197.
Li, R., Jewell, S., & Grummer-Strawn, L. (2003). Maternal
obesity and breast-feeding
practices. American Journal of Clinical Nutrition, 77(4),
931e936.
Li, L., Law, C., Lo Conte, R., & Power, C. (2009).
Intergenerational influences on
childhood body mass index: the effect of parental body mass
index trajectories.
The American Journal of Clinical Nutrition, 89, 551e557.
Li, L., Parsons, T. J., & Power, C. (2003). Breastfeeding and
obesity in childhood:
cross-sectional study. British Medical Journal, 327, 904e905.

Li, R., Scanlon, K. S., & Serdula, M. (2005). The validity and
reliability of maternal
recall breastfeeding practice. Nutrition Reviews, 63, 103e110.
Lindsay, R. S., Hanson, R. L., Roumain, J., Ravussin, E.,
Knowler, W. C., &
Tataranni, P. A. (2001). Body mass index as a measure of
adiposity in children
and adolescents: relationship to adiposity by dual energy X-ray
absorptiometry
and to cardiovascular risk factors. Journal of Clinical
Endocrinology and Metab-
olism, 86, 4061e4067.
O’Callaghan, M. J., Williams, G. M., Andersen, M. J., Bor, W.,
& Najmans, J. M. (1997).
Prediction of obesity in children at 5 years: a cohort study.
Journal of Paediatrics
and Child Health, 33, 311e316.
O’Tierney, P. F., Barker, D. J. P., Osmond, C., Kajantie, E., &
Eriksson, J. G. (2009).
Duration of breast-feeding and adiposity in adult life. The
Journal of Nutrition,
139(2), 422Se425S.
Ong, K. K. L., Ahmed, M. L., Emmett, P. M., Preece, M. A.,
Dunger, D. B., & the Avon
Longitudinal Study of Pregnancy and Childhood Study Team.
(2000). Associa-
tion between post-natal catch-up growth and obesity in
childhood: prospective
cohort study. British Medical Journal, 320, 967e971.
Ong, K. K. L., Emmett, P. M., Noble, S., Ness, A., & Dunger,

D. B. (2006). Dietary energy
intake at the age of 4 months predicts postnatal weight gain and
childhood
body mass index. Pediatrics, 117, 503e508.
Ong, K. K. L., Preece, M. A., Emmett, P. M., Ahmed, M. L.,
Dunger, D. B., & for the
ALSPAC study team. (2002). Size at birth and early childhood
growth in relation
to maternal smoking, parity and infant breast-feeding:
longitudinal birth cohort
study and analysis. Pediatric Research, 52(6), 863e867.
Owen, C. G., Martin, R. M., Whincup, P. H., Davey Smith, G.,
& Cook, D. G. (2005).
Effect of infant feeding on the risk of obesity across the life
course: a quanti-
tative review of Published evidence. Pediatrics, 115,
1367e1377.
Parsons, T. J., Power, C., & Manor, O. (2003). Infant feeding
and obesity through the
lifecourse. Archives of Disease in Childhood, 88, 793e794.
Poulton, R., & Williams, S. (2001). Breastfeeding and risk of
overweight. Journal of
the American Medical Association, 286, 1449e1450.
Reilly, J. J., Armstrong, J., Dorosty, A. R., Emmett, P. M.,
Ness, A., Rogers, I., et al.for the
Avon Longitudinal Study of Parents and Children Study Team.
(2005). Early life
risk factors for obesity in childhood: cohort study. British
Medical Journal, 330,
1357.
Royston, P. (2004). Multiple imputation of missing values.

Stata Journal, 4, 227e241.
Scaglioni, S., Agostini, C., De Notaris, R., Radaelli, G., Radice,
N., Valenti, M., et al.
(2000). Early macronutrient intake and overweight at five years
of age. Inter-
national Journal of Obesity, 24, 777e781.
Schack-Nielson, L., Sorensen, T. I. A., Mortensen, E. K., &
Michaelsen, K. F. (2010). Late
introduction of complementary feeding, rather than duration of
breastfeeding,
may protect against adult overweight. American Journal of
Clinical Nutrition,
91(3), 619e627.
Singhal, A., & Lanigan, J. (2007). Breastfeeding, early growth
and later obesity.
Obesity Reviews, 8(Suppl. 1), 51e54.
Stettler, N., Zemel, B. S., Kumanyika, S., & Stallings, V. A.
(2002). Infant weight gain
and childhood overweight status in a multicenter cohort study.
Pediatrics, 109,
194e199.
Taylor, J. P., Evers, S., & McKenna, M. (2005). Determinants
of healthy eating in
children and youth. Canadian Journal of Public Health,
96(Suppl. 3), S20eS26.
Taylor, R. W., Grant, A. M., Goulding, A., & Williams, S. M.
(2005). Early adiposity
rebound: review of papers linking this to subsequent obesity in
children and
adults. Current Opinion in Clinical Nutrition and Metabolic

Care, 8, 607e612.
Tienboon, P., Rutishauser, I. H. E., & Wahlqvist, M. L. (1994).
Maternal recall of infant
feeding practices after an interval of 14 to 15 years. Australian
Journal of
Nutrition & Dietetics, 51(1), e25ee27.
Toschke, A. M., Vignerova, J., Lhotska, L., Osancova, K.,
Koletzko, B., & von Kries, R.
(2002). Overweight and obesity in 6- to 14-year-old Czech
children in 1991:
protective effect of breastfeeding. Journal of Pediatrics, 141,
764e769.
Wadsworth, M., Marshall, S., Hardy, R., & Paul, A. (1999).
Breast feeding and obesity.
Relation may be accounted for by social factors. British
Medical Journal, 319,
1576.
Williams, S. L., & Mummery, W. K. (2011). Links between
adolescent physical
activity, body mass index, and adolescent and parent
characteristics. Health
Education and Behaviour, 38(5), 510e520.
Wilson,A. C., Forsyth,J.S.,Greene,S.A.,Irvine, L., Hau,C.,&
Howie,P. W. (1998).Relation
of infant diet to childhood health: seven year follow up of
cohort of children in
Dundee infant feeding study. British Medical Journal,
316(7124), 21e25.
Wright, P., Fawcett, J., & Crow, R. (1980). The development of
differences in the

feeding behaviour of bottle and breast fed human infants from
birth to two
months. Behavioural Processes, 5, 1e20.
Ziegler, E. E. (2006). Growth of breast-fed and formula-fed
infants. In J. Rigo, &
E. E. Ziegler (Eds.), Protein and energy requirements in infancy
and childhood (pp.
51e59). Basel: Karger, AG.
Breastfeeding and risk of overweight and obesity at nine-years
of ageIntroductionMethodSampleMeasuresBreastfeeding
measureMeasurement of BMIDefinition of overweight and
obesity in early childhoodCovariatesChild variablesParental
variablesSocio-economic characteristics of the
householdMissing cases
analysisResultsDiscussionAcknowledgementsReferences
American Journal of Epidemiology
Copyright © 2005 by the Johns Hopkins Bloomberg School of
Public Health
All rights reserved
Vol. 161, No. 1
Printed in U.S.A.
DOI: 10.1093/aje/kwh338
at K
ingston U
niversity L
ibrary on M
arch 11, 2013

http://aje.oxfordjournals.org/
D
ow
nloaded from
PRACTICE OF EPIDEMIOLOGYMETA-ANALYSIS
Breastfeeding in Infancy and Blood Pressure in Later Life:
Systematic Review and
Meta-Analysis
Richard M. Martin, David Gunnell, and George Davey Smith
From the Department of Social Medicine, University of Bristol,
Bristol, United Kingdom.
Received for publication January 29, 2004; accepted for
publication June 25, 2004.
The influence of breastfeeding on blood pressure in later life is
uncertain. The authors conducted a systematic
review of published studies from which estimates of a mean
difference (standard error) in blood pressure between
breastfed and bottle-fed subjects could be derived. They
searched MEDLINE and Excerpta Medica (EMBASE)
bibliographic databases, which was supplemented by manual
searches of reference lists. Fifteen studies (17
observations) including 17,503 subjects were summarized.
Systolic blood pressure was lower in breastfed
compared with bottle-fed infants (pooled difference: –1.4
mmHg, 95% confidence interval (CI): –2.2, –0.6), but
evidence of heterogeneity between study estimates was evident
(χ216 = 42.0, p < 0.001). A lesser effect of
breastfeeding on systolic blood pressure was observed in larger

(n ≥ 1,000) studies (–0.6 mmHg, 95% CI: –1.2,
0.02) compared with smaller (n < 1,000) studies (–2.3 mmHg,
95% CI: –3.7, –0.9) (p for difference in pooled
estimates = 0.02). A small reduction in diastolic blood pressure
was associated with breastfeeding (pooled
difference: –0.5 mmHg, 95% CI: –0.9, –0.04), which was
independent of study size. If causal, the small reduction
in blood pressure associated with breastfeeding could confer
important benefits on cardiovascular health at a
population level. Understanding the mechanism underlying this
association may provide insights into pathways
linking early life exposures with health in adulthood.
blood pressure; bottle feeding; breast feeding; cardiovascular
system; hypertension; infant nutrition; milk,
human; review literature
Abbreviation: CI, confidence interval.
Evidence is growing that blood pressure levels in both
childhood and young adulthood are influenced by factors
operating early in life (1–4) and are associated with later
cardiovascular disease (5). Specifically, several cohort
studies suggest that blood pressure may be determined by
early nutritional exposures, including sodium intake in
infancy (6), consumption of formula feed (7), and breast-
feeding (8). Detection, treatment, and control of hyperten-
sion in adulthood does not reduce cardiovascular disease risk
to normotensive levels (9), supporting efforts to identify
primary prevention interventions that could be started in
early life. Any long-term effect of breastfeeding on blood
pressure levels may have implications for policies promoting
breastfeeding, particularly among the least affluent families
with the lowest breastfeeding rates (10) and the highest risks
of premature cardiovascular disease (11), and it may

increase understanding of cardiovascular disease mecha-
nisms operating through early life exposures.
Interpreting individual studies of the association between
breastfeeding and blood pressure in isolation is complicated.
Firstly, cohort studies include infants born in different
decades during the 20th century (8, 12, 13). The composition
of bottle (artificial) feeds has changed during this time, and
associations with particular components of these feeds may
explain differences in results. Secondly, different definitions
of breastfeeding have been used (13, 14). Thirdly, the
strength of the relation may depend on the age at outcome
measurement (15, 16). Finally, control for confounding
Correspondence to Dr. Richard M. Martin, Department of Social
Medicine, University of Bristol, Canynge Hall, Whiteladies
Road, Bristol,
United Kingdom, BS8 2PR (e-mail: [email protected]).
15 Am J Epidemiol 2005;161:15–26
16 Martin et al.
at K
ingston U
niversity L
ibrary on M
arch 11, 2013
D
ow

nloaded from
factors may have been inadequate (17). We conducted a
systematic review and meta-analysis of studies reporting on
blood pressure levels in breast- and bottle-fed subjects and
explored possible sources of heterogeneity using meta-
regression (18).
MATERIALS AND METHODS
Included studies
Articles were included if they fulfilled the following
criteria: 1) having been breastfed in infancy was compared
with bottle (artificial) feeding, 2) systolic or diastolic blood
pressure had been measured as an outcome, and 3) an esti-
mate of the mean difference in blood pressure between
breast- and bottle-fed groups could be extracted from the
article. Our review was restricted to human subjects.
Data sources
We systematically searched all published papers, letters,
abstracts, and review articles on infant feeding and cardio-
vascular disease, cardiovascular disease risk factors, and
growth by using the MEDLINE and Excerpta Medica
(EMBASE) bibliographic databases from their inception to
April 2003. We used a combined text word and MESH
heading search strategy (refer to the Appendix), and we
manually searched reference lists of all studies that fulfilled
our eligibility criteria. Using the “saved searches” and “auto
alerts” automated facilities incorporated within the
MEDLINE and EMBASE databases, we reran the search
every week until May 2004. No restriction was made
regarding language of publication. Two papers then in press

but not yet published (19, 20) were also considered for inclu-
sion. When clarifications were required, we corresponded
with the authors, but no additional data were supplied. One
of the authors (R. M. M.) assessed study eligibility and
extracted data by using a prepiloted, standardized form.
We did not use a simple quality score, which might be
arbitrary. Instead, we conducted meta-regression analyses to
assess specific aspects of quality, including control of
confounding, loss to follow-up, recall bias, definition of
breastfeeding, and sample size (refer to the information
below).
Statistical analysis
A meta-analysis of the mean differences, and their stan-
dard errors, in systolic and diastolic blood pressures between
breastfed and bottle-fed infants was conducted. The fully
adjusted estimates from individual studies were used in the
meta-analysis where available; otherwise, the crude esti-
mates were used. Heterogeneity was assessed by using the Q
test (18). Because heterogeneity was marked, random-
effects models were computed. One paper followed up
subjects at ages 13–16 years (15), some of whom were
included in an analysis based on follow-up at ages 7.5–8
years (16). Because the two studies cannot be considered
independent in a meta-analysis, we performed a meta-
analysis with and without including this later follow-up
study to determine its impact on the overall pooled mean
difference.
Selected study characteristics, chosen a priori, were
entered as indicator variables in separate meta-regression
analyses (18) to assess their impact on between-study varia-
tion (heterogeneity), as follows: study size (<1,000/ ≥1,000);

reliance on maternal recall of breastfeeding beyond infancy
(yes/no); whether breastfeeding occurred for at least 2
months (yes/no); whether breastfeeding was exclusive for at
least 2 months (yes/no); age at measurement of blood pres-
sure (<10 years/11–45 years/>45 years); decade of birth
(before 1980/after 1980); proportion of target population
included in the main analysis (<30 percent/31–60 percent,
>61 percent); method of blood pressure measurement (auto-
mated/manual); and whether effect estimates in the final
models controlled for social factors in childhood or adult-
hood (yes/no), maternal factors in pregnancy (yes/no), or
current weight (yes/no). Papers that assessed blood pressure
in infancy only (age <1 year) were investigated separately
because the focus of our inquiry was on the long-term, rather
than acute, effects of breastfeeding. Funnel plots, the Egger
(weighted regression) test, and the Begg and Mazumdar
(rank correlation) tests for funnel plot asymmetry were
conducted to examine the relation between sample size and
observed mean differences in blood pressure by infant
feeding group (21).
Sensitivity analysis
We examined the likely impact on the overall pooled
relation between breastfeeding and blood pressure of also
including the five potentially eligible studies that did not
provide quantitative estimates (table 1). In all five studies,
null results were reported, and a mean difference in systolic
blood pressure of 0.0 mmHg between breast- and bottle-fed
subjects was assigned. The meta-analysis was then
repeated to estimate the pooled mean difference when all
studies were included (i.e., both those with published esti-
mates and the five studies without published estimates).
For the five studies without quantitative data, an estimate
of the standard error was based on the sample size and
assumed a standard deviation of 10 mmHg where this

parameter was not reported (22–24).
RESULTS
Description of studies
The electronic search yielded 3,403 references. Abstract
review suggested that 17 were potentially relevant to the
analysis relating breastfeeding with blood pressure beyond
12 months (8, 12–16, 23–33). Ten other papers were identi-
fied from a manual search of reference lists (22, 34–42). Of
the 27 studies, 12 published studies were included in the
meta-analysis (8, 12–16, 25–27, 34–36) (Web table 1; this
information is described in the supplementary table referred
to as “Web table 1” in the text, which is posted on the
Journal’s website (http://aje.oupjournals.org/)). Reasons for
exclusion (n = 15) are given in figure 1. Together with the
three additional studies identified after April 2003 (which
Am J Epidemiol 2005;161:15–26
Breastfeeding and Blood Pressure: A Systematic Review 17
at K
ingston U
niversity L
ibrary on M
arch 11, 2013
D
ow

nloaded from
involved 10,062 subjects) (19, 20, 43), 15 studies with
17,503 participants were included in the meta-analysis
relating breastfeeding with blood pressure beyond 12
months (Web table 1).
Two of these 15 studies were based on a follow-up of a
randomized controlled trial in preterm infants (15, 16), eight
were prospective cohorts (8, 14, 20, 25–27, 36, 43), and one
was a historical cohort (13); in four cross-sectional surveys
of blood pressure, infant feeding history was based on retro-
spective recall by the mother (12, 19, 34, 35). These studies
included populations from the United Kingdom, Finland,
Holland, Belgium, Italy, Czech Republic, Croatia, South
Africa, and Australia. Individual studies were relatively
homogeneous with respect to ethnicity. The year of birth of
the subjects ranged from 1918 to 1994. The proportion of the
target population included in the main analysis was unstated
in one paper (35), less than 30 percent in four studies (12, 13,
15, 36), 30–60 percent in four studies (8, 20, 27, 43), and
more than 60 percent in six studies (14, 16, 19, 25, 26, 34).
From these 15 studies, 17 estimates of systolic blood pres-
sure differences were derived, of which 12 included males
and females combined and five were sex specific. Eleven
systolic blood pressure observations (nine studies) were of
children (aged 1–16 years), and six observations (five
studies) occurred in later adulthood (age ≥17 years). One
study reported results for diastolic blood pressure only (25).
From the 15 studies, 13 estimates of diastolic blood pressure
differences were derived, 12 of which included males and
females combined and one of which was for males only.
Nine diastolic blood pressure observations (eight studies)
were of children aged 1–16 years, and four observations

(four studies) occurred in adulthood (age ≥17 years).
Definitions of breastfeeding
The 15 studies used different definitions of breastfeeding.
In a randomized controlled trial with follow-up at ages 7.5–
8 years (16) and ages 13–16 years (15), preterm infants were
randomly assigned to donated, banked breast milk or
TABLE 1. Studies reporting on associations between method
of infant feeding and blood pressure beyond 12 months of age
that
were not included in the current meta-analysis
* Includes partially breastfed.
First author, source
(year of publication)
(reference no.)
No. breastfed*;
no. bottle fed
(sex)
Infant feeding
comparison
Infant year of
birth
Age at which infant
feeding was assessed
Age at which

outcome
measurement
occurred
Description
of results
Baranowski, families
from an ethnically
diverse population in
Texas (1992) (22)
245 total (M† + F†) Duration of any
breastfeeding
Not stated Interviewer
administered
questions to mother
3–4 years after
infant’s birth
3–4 years No significant correlations
between duration of
breastfeeding and SBP†
or DBP† observed;
quantitative estimates not
reported
Cobaleda Rodrigo,
Madrid, Spain (1989)
(23)
1,893 total (M + F) Ever vs. never
breastfed

1965–1983 0–18 years; method
unclear
0–18 years No significant differences
between duration of
breastfeeding and SBP
or DBP observed; no
quantitative estimates
given
Simpson, births in
Dunedin maternity
hospital, New Zealand
(1981) (37)
692 total (M + F) Ever vs. never
breastfed
1972–1973 3 years; method
unclear
7 years No significant difference in
breastfeeding rates or
duration of breastfeeding
when comparing children
with high, medium, and
low blood pressure; no
given
Marmot, subsample of
238 eligible subjects
living in London and
Bristol, United
Kingdom who were
part of the 1946

national birth cohort
(n = 5,362), England
(1980) (24)
95; 47 (M + F) Exclusively breastfed
for 5 months vs.
exclusively bottle
fed
1946 First and third year of
life; methods not
stated
31–32 years “There were no consistent
differences [in blood
pressure levels] between
those who had been
breastfed and those who
had been bottle fed”; no
given
Fall, 297 women born
and still living in East
Hertfordshire (total
births = 5,585),
England (1995) (41)
279; 11 (F) Breastfed, bottle fed,
breast- and bottle
fed
1923–1930 During infancy; infant
feeding mode
recorded by health
visitors

60–71 years “No differences occurred
between the three
feeding groups in any of
the risk factors
measured” (included
systolic and diastolic
blood pressures); no
given
18 Martin et al.
at K
ingston U
niversity L
ibrary on M
arch 11, 2013
D
ow
nloaded from
preterm formula (either as the sole diet or a supplement to
mother’s milk) until they weighed 2,000 g or were
discharged to home. In the other studies, the exposure was
defined as 1) any breastfeeding in five studies (12, 19, 25,
26, 35); 2) exclusive breastfeeding in five studies (exclusive

for the first 10 days only (13), for at least 3 months (27, 34),
for at least 15 weeks (8), or for at least 12 months (36)); 3)
both any breastfeeding and exclusive breastfeeding for at
least 2 months in one study (43); and 4) any breastfeeding for
at least 3 months in one study (14) and at least 6 months in
another (20). In all studies except the randomized controlled
trial (15, 16), the comparator group was exclusive bottle
feeding. Five of the studies (providing six observations)
relied on maternal recall beyond infancy, ranging from 3–18
years (14), to 3 years (27), to 5–7 years (34), to 20–28 years
(12), and to 44–60 years (19).
Breastfeeding and systolic blood pressure
The results for systolic blood pressure, shown in figure 2,
are based on 14 studies with 17 observations. Mean systolic
blood pressure was lower in breastfed infants compared with
bottle-fed infants according to 10 observations from eight
studies (8, 14, 15, 20, 26, 35, 36, 43). Seven observations
(from six studies) showed no or little difference in systolic
blood pressure among breastfed versus formula-fed infants
(12, 13, 16, 19, 27, 34). Two of these seven observations
were from the randomized controlled trial in preterm infants
with follow-up at ages 7–8 years (16). When the original
study was followed up into adolescence (ages 13–16 years),
having received breast milk was associated with a 2.7-
mmHg reduction in blood pressure (15).
In a random-effects model, mean systolic blood pressure
was lower among breastfed infants (mean difference: –1.4
FIGURE 1. Summary of outcomes of studies retrieved for
analysis, 1966–2004.

at K
ingston U
niversity L
ibrary on M
arch 11, 2013
D
ow
nloaded from
mmHg, 95 percent confidence interval (CI): –2.2, –0.6; p =
0.001) (figure 2). There was also evidence of marked hetero-
geneity between studies (χ216 = 42.0, p < 0.001). Exclusion
of the study by Singhal et al. (15) (because of lack of inde-
pendence from Lucas et al.’s study (16)) had little impact on
the pooled difference (–1.3, 95 percent CI: –2.2, –0.5).
Controlling for study size in a meta-regression analysis
lowered the τ2 estimate of between-study variation from 1.69
when study size was not included in the model to 0.47 when
study size was included, suggesting that some of the
observed heterogeneity was explained by study size. In a
stratified meta-analysis, a smaller effect of breastfeeding on
later systolic blood pressure was observed in the larger
studies (n ≥ 1,000) (difference: –0.6 mmHg, 95 percent CI:
–1.2, 0.02; p = 0.06) compared with the smaller studies (n <
1,000) (difference: –2.3 mmHg, 95 percent CI: –3.7, –0.9;
p = 0.001). This difference was unlikely to be due to chance

(p = 0.02). There was evidence of heterogeneity in models
restricted to small studies (χ212 = 27.1, p = 0.007) but less
evidence among the four larger studies (χ23 = 6.1, p = 0.1).
FIGURE 2. Mean difference (95% confidence interval) in
systolic blood pressure (mmHg) for infants who were breastfed
minus infants who
were bottle fed: studies reporting on the association between
breastfeeding and systolic blood pressure, 1966–2004. The first
author, the year of
publication, and the reference number (in parentheses) are
indicated on the y-axis. These studies are arranged in
descending order of mean age
at which blood pressure was measured. The box corresponding
to each study is proportional to the inverse of the variance, with
horizontal lines
showing the 95% confidence intervals of the mean difference in
systolic blood pressure (mmHg). The combined estimate is
based on a random-
effects model shown by the dashed vertical line and diamond
(95% confidence interval). The solid vertical line represents the
null result, that is,
zero mean difference in blood pressure. Lucas 1 or 2 denotes
estimates using different comparator groups (Web table 1). *
Female-specific esti-
mates; ** male-specific estimates.
20 Martin et al.
at K
ingston U

niversity L
ibrary on M
arch 11, 2013
D
ow
nloaded from
In studies where the duration of breastfeeding was at least
2 months, the pooled blood pressure difference between
breast- and bottle-fed groups (–2.0 mmHg) was on average
1.6 mmHg larger (95 percent CI: –0.4, 3.5; p = 0.1) than in
studies with a shorter duration of breastfeeding (pooled
difference: –0.6 mmHg). Similarly, the difference in blood
pressure between breast- and bottle-fed groups was 1.4
mmHg greater (95 percent CI: –0.4, 3.2; p = 0.1) in those
born up to 1980 (pooled difference: –2.7 mmHg) compared
with those born after 1980 (pooled difference: –0.8 mmHg).
Only four of the 17 observations on systolic blood pressure
controlled for potential socioeconomic (19, 20, 43) or
maternal antenatal factors (such as body mass index,
smoking in pregnancy, education, parity, marital status) (8,
20, 43) or current body size (8, 20, 43). Controlling for
confounding produced a greater than 30 percent reduction in
crude effect estimates in two (19, 43) of three studies in
which comparison with crude estimates was possible. In
meta-regression analysis, there was weak evidence that
studies not controlling for socioeconomic factors (pooled
difference: –2.0 mmHg) had mean differences in blood pres-
sure 1.4 mmHg higher (95 percent CI: –0.6, 3.3; p = 0.17)
than in studies controlling for socioeconomic factors (pooled
difference: –0.9 mmHg). In one study, a large reduction in

blood pressure associated with having been breastfed for at
least 3 months (Web table 1) was reported to have been
somewhat attenuated after controlling for current weight,
age, birth weight, time of birth, birth order, mother’s age,
and history of high antenatal maternal blood pressure (14),
but quantitative estimates suitable for inclusion in the meta-
analyses were not available. Several studies controlled for
current weight (14) or body mass index (8, 15) or ponderal
index (20) in their final model, which may have had the
effect of overcontrolling for a factor on the causal pathway if
breastfeeding lowers blood pressure by reducing later
adiposity (44).
In meta-regression analyses, there was little evidence that
heterogeneity was explained by reliance on maternal recall
of breastfeeding (p = 0.9), age at measurement of blood pres-
sure (p = 0.8), whether breastfeeding was exclusive for at
least 2 months (p = 0.6), method of blood pressure measure-
ment (p = 0.2), or proportion of the target population
included in the main analysis (p = 0.9).
Breastfeeding and diastolic blood pressure
The results for 13 observations (12 studies) relating to
diastolic blood pressure are shown in figure 3. Mean dia-
stolic blood pressure was lower among breastfed infants
according to nine observations from eight studies (8, 12, 15,
16, 19, 20, 25, 43). In a random-effects model, the pooled
mean diastolic blood pressure was lower among breastfed
infants (difference: –0.5 mmHg, 95 percent CI: –0.9, –0.04;
p = 0.03). There was less evidence of heterogeneity between
estimates (χ212 = 20.2; p = 0.06) than in the analysis of
breastfeeding and systolic blood pressure. Exclusion of the
study by Singhal et al. (15) had little impact on the pooled
difference (–0.4, 95 percent CI: –0.8, –0.01). The effect of

breastfeeding on later diastolic blood pressure was similar in
the four larger studies (n ≥ 1,000) (difference: –0.4 mmHg,
95 percent CI: –0.9, 0.1; p = 0.10) compared with the seven
smaller studies (n < 1,000) (difference: –0.6 mmHg, 95
percent CI: –1.5, 0.2; p = 0.15). Studies that relied on
maternal recall of breastfeeding beyond infancy showed
pooled differences in mean diastolic blood pressure (0.0
mmHg) that were 0.6 mmHg smaller (95 percent CI: 0.2,
1.1; p = 0.004) than in studies that did not rely on recall
(pooled difference: –0.7 mmHg).
We found little evidence that between-study heterogeneity
in estimates was explained by age at measurement of blood
pressure (p = 0.5), decade of birth (p = 0.2), stipulation of a
minimum duration of breastfeeding (p = 0.5), proportion of
the target population in the main analysis (p = 0.2), whether
breastfeeding was exclusive for at least 2 months (p = 0.2),
method of blood pressure measurement (p = 0.4), or whether
effect estimates controlled for socioeconomic factors (p =
0.9), maternal factors in pregnancy (p = 0.9), or current
weight (p = 0.9).
Studies that formally tested for interaction found little
evidence of sex differences in the association between
breastfeeding and systolic or diastolic blood pressure (20,
43). Repeating analyses after excluding the first published
(in 1981) of the included studies (35, 36), which could be
regarded as hypothesis-generating reports, made little differ-
ence to the pooled-effect estimates for systolic (mean differ-
ence: –1.1 mmHg, 95 percent CI: –1.8, –0.4; p = 0.003) or
diastolic (mean difference: –0.5 mmHg, 95 percent CI: –1.0,
–0.06; p = 0.03) blood pressure.
Small study effects
For systolic blood pressure, there was evidence of differ-

ential small study effects on inspection of funnel plots
(figure 4) and the Begg (p = 0.09) test for funnel plot asym-
metry, but there was no such evidence for diastolic blood
pressure (Begg test: p = 0.3). That is, we found some
evidence that small studies (i.e., those with higher standard
errors, located to the right of the figure), compared with
larger studies, reported larger mean differences in systolic
blood pressure between infant feeding groups.
Excluded studies
Table 1 summarizes the results from the five studies not
included in the meta-analysis because a mean difference in
blood pressure could not be obtained (22–24, 37, 41). All
reported no “statistically significant” association between
breastfeeding and either systolic or diastolic blood pressure.
These studies were relatively small—only 3,262 subjects in
total compared with 17,503 included in the meta-analysis. In
a sensitivity analysis, inclusion in the meta-analysis of the
assumed zero estimates from the five studies (table 1) with
no published mean differences attenuated the overall
summary estimate for systolic blood pressure (mean differ-
ence: –1.0 mmHg, 95 percent CI: –1.6; –0.4; p = 0.002), but
there was still strong evidence of an inverse association.
at K
ingston U
niversity L
ibrary on M

arch 11, 2013
D
ow
nloaded from
Blood pressure in infancy
Overall, six studies were identified that examined the rela-
tion between infant feeding mode and blood pressure
measured before 12 months of age (32, 39, 40, 45–47) (table
2). The mean difference in blood pressure by feeding mode,
and the associated standard error, could be estimated from
four of these studies (six observations) (32, 40, 45, 46). In
random-effects models, the pooled systolic blood pressure
difference in infancy associated with breastfeeding was –1.7
mmHg (95 percent CI: –4.0, 0.6; p = 0.15), although there
was some evidence of heterogeneity (χ25 = 11.8; p = 0.04).
The pooled diastolic blood pressure difference in infancy
associated with breastfeeding was –1.1 (95 percent CI: –4.0,
1.8; p = 0.4; χ23= 8.2, p = 0.04).
DISCUSSION
Breastfeeding was associated with a 1.4- and 0.5-mmHg
reduction in systolic and diastolic blood pressure, respec-
FIGURE 3. Mean difference (95% confidence interval) in
diastolic blood pressure (mmHg) for infants who were breastfed
minus infants who
were bottle fed: studies reporting on the association between
breastfeeding and diastolic blood pressure, 1966–2004. The first

author, the year
of publication, and the reference number (in parentheses) are
indicated on the y-axis. These studies are arranged in
descending order of mean
age at which blood pressure was measured. The box
corresponding to each study is proportional to the inverse of the
variance, with horizontal
lines showing the 95% confidence intervals of the mean
difference in diastolic blood pressure (mmHg). The combined
estimate is based on a
random-effects model shown by the dashed vertical line and
diamond (95% confidence interval). The solid vertical line
represents the null result,
that is, zero mean difference in blood pressure. Lucas 1 or 2
denotes estimates using different comparator groups (Web table
1). * Male-specific
estimate.
22 Martin et al.
at K
ingston U
niversity L
ibrary on M
arch 11, 2013
D
ow

nloaded from
tively, although differences in systolic blood pressure
between feeding groups were reduced in large (difference:
–0.6 mmHg) compared with smaller (difference: –2.3 mmHg)
studies. These pooled estimates are similar to those found by
Owen et al. (48) in a recent review, even though the current
report includes recently published data on an extra 10,062
subjects from three studies that included more than 1,500
participants each.
Chance, bias, and confounding
A number of studies reported inverse associations between
breastfeeding and blood pressure, including two (of three)
with more than 3,500 subjects each (20, 43), suggesting that
these findings are unlikely to be due to type 1 error alone.
Selection bias would arise if excluded subjects had a
different breastfeeding–blood pressure association
compared with those who were included. In one study, a
protective effect of breast milk on blood pressure was
observed when 26 percent of the original cohort were
followed up at ages 13–16 years (15), but not when 81
percent were examined at ages 7.5–8 years (16), suggesting
either the possibility of selection bias in the later follow-up
or an amplification of the breastfeeding–blood pressure
association (49). When all the studies were considered, we
found similar effect estimates in studies with more than 60
percent follow-up and in those with less than 30 percent
follow-up, suggesting that the association between breast-
feeding and blood pressure did not systematically vary
between studies according to follow-up rates.
Although reporting of ever having been breastfed after up
to 20 years is highly correlated with obstetric records (50),

breastfeeding duration may be remembered less accurately
(51). Three cross-sectional studies relied on retrospective
reporting of exclusive (34) or any breastfeeding 7 years (34),
28 years (12), and 60 years (19) after birth, and these studies
showed little evidence of an association between breast-
feeding and blood pressure. In meta-regression analysis, reli-
ance on maternal recall was associated with an attenuation of
the difference in diastolic (but not systolic) blood pressure
between breast- and bottle-fed groups. Publication bias is a
concern because most studies in this review were small, and
mean blood pressure differences were greater in the smaller
compared with the larger studies.
Relatively few studies controlled for potential
confounding factors, although adjusted effect estimates were
attenuated by at least 30 percent in two studies (19, 43). In
the meta-regression analyses, studies controlling for socio-
FIGURE 4. Begg’s funnel plot (with pseudo 95% confidence
limits) for studies reporting on the association between
breastfeeding and systolic
blood pressure, 1966–2004.
at K
ingston U
niversity L
ibrary on M
arch 11, 2013

D
ow
nloaded from
economic factors showed smaller systolic blood pressure
differences between breast- and bottle-fed subjects. The
distribution of breastfeeding was less socially determined
before World War II (52) compared with now (10), and
results from prewar cohorts may be free from confounding
by social class (53). The two prewar studies reviewed
showed little evidence of any association between breast-
feeding and blood pressure (13, 19), although nondifferential
misclassification is a possibility in the Caerphilly cohort that
relied on recall of breastfeeding status 44–60 years after
infancy (19), and the Dutch famine cohort may not be gener-
alizable (13). Accelerated postnatal weight gain is a potential
confounding factor because it is associated with raised blood
pressure (54) and may influence infant feeding practices
(55). In the only study to examine this issue (43), the associ-
ation of breastfeeding with blood pressure was not altered by
postnatal growth.
Relevance to contemporary cohorts
Modern formula feeds, which more closely resemble the
nutrient content of breast milk, were not developed until the
mid-1970s (56). Previously, bottle-fed infants were given
unmodified cow’s milk preparations and other alternatives
such as condensed milk (52, 57). Several studies of infants
born since 1980, however, show a blood-pressure-lowering
TABLE 2. Studies relating breastfeeding to blood pressure

levels in infancy (before 12 months of age), by year of
publication
* M, male; F, female; SBP, systolic blood pressure; DBP,
diastolic blood pressure.
First author, source
(year of publication)
(reference no.)
No. breastfed;
no. bottle fed (sex)
Infant feeding
comparison
Infant year
of birth
Age at which
infant
feeding was
assessed
Age at which
outcome
measurement
occurred
Mean difference (breast-bottle) in
mmHg (standard error) Covariates in
multiply-adjusted

modelsUnadjusted or
simple model
Fully
adjusted
model
Studies included in the meta-analysis
Pomeranz, infants born
in a single hospital,
Israel (2002) (45)
7; 31 (M* + F*) Ever breastfed vs.
milk formula
made with
either mineral
water (low
sodium) or tap
water (high
sodium)
Not stated Birth 6 months SBP*: –6.1 (2.0);
DBP*: –7.3 (3.1)
Not given None
Bernstein, term infants
born in
Johannesburg
Hospital, South
Africa (1990) (46)
43; 81 (M + F) Exclusively
breastfed (n =

43) vs. low-
sodium formula
(n = 42) or
high-sodium
formula (n = 39)
1988 6 weeks 6 weeks Breastfed vs. low-
sodium formula:
–1.6 (2.2);
breastfed vs.
high-sodium
formula: –4.1
(2.0)
Not given None
Zinner, about 4% of
infants born in
hospitals in Boston,
Massachusetts, and
Rhode Island (1980)
(32)
154; 264 (M + F) Breastfed vs.
bottle fed
Not stated Infancy 1–6 days SBP: 0.0 (0.95);
DBP: –0.7 (0.92)
Not given None
Schachter, hospital
births, Pittsburgh,
Pennsylvania (1979)
(40)

30; 141 (M + F) Breastfed vs.
bottle fed
Not stated Infancy 6 months White ethnicity:
SBP: –0.5 (1.8);
DBP: –1.5 (1.3).
Black ethnicity:
SBP: 3.3 (3.3);
DBP: 5.8 (3.5)
Not given Results
stratified by
ethnicity
Studies excluded from the meta-analysis
Cohen, neonates born
at two hospitals,
United States (1992)
(47)
7; 11 (M + F) Breastfed, bottle
fed
Not stated Infancy 24–94
hours
(mean:
55
hours)
During a feed; blood
pressure of
breastfed babies
approximately 15
mmHg higher
than those bottle

fed but about 2
mmHg higher
(derived from
figure 2) before
and 30–60
minutes after a
feed
de Swiet, 500 infants
born in a hospital in
Kent, England
(1977) (39)
Not stated
(M + F)
Breastfed, bottle
fed
1975 Infancy 4 days and
6 weeks
No differences in
blood pressure
levels between
infants breastfed
vs. bottle fed
24 Martin et al.
at K
ingston U

niversity L
ibrary on M
arch 11, 2013
D
ow
nloaded from
effect of breastfeeding (8, 15, 20, 25, 26, 43), suggesting that
if the results are causal, they are relevant to modern cohorts.
Population health implications
Reductions in population mean blood pressure levels of as
little as 2 mmHg could reduce the prevalence of hyperten-
sion by up to 17 percent, the number of coronary heart
disease events by 6 percent, and strokes and transient
ischemic attacks by 15 percent (9, 58). This reduction
equates to preventing 3,000 coronary heart disease events
and 2,000 strokes annually among those under age 75 years
in the United Kingdom (59). The effect estimates from our
meta-analysis could therefore translate into the prevention of
a substantial number of deaths annually.
Mechanisms
Breastfeeding could influence blood pressure via a variety
of mechanisms, including 1) reducing sodium intake in
infancy (60); 2) increasing intake of long-chain polyunsatu-
rated fatty acids, important structural components of tissue
membrane systems, including the vascular endothelium
(25); and 3) protecting against hyperinsulinemia in infancy

(61–63) and insulin resistance in early life (64), adolescence
(65), and adulthood (13), processes that may in turn raise
blood pressure via a number of mechanisms (66).
The concomitant association of breastfeeding with both
taller stature (particularly leg length) (67, 68) and lower
blood pressure is in line with previously reported inverse
relations between stature (particularly leg length) and blood
pressure in adulthood (64, 69). Height and leg length may
reflect the dynamic properties of the arterial tree, with short
height increasing the systolic peak because of the early
return of reflected arterial pulse waves (64). Two studies that
controlled for current height found that this made very little
difference to effect estimates (34, 43), suggesting that height
may not be on the causal pathway between breastfeeding and
blood pressure. Alternatively, breastfeeding may program
both growth rate and the formation of blood pressure control
mechanisms (70).
Conclusions
Breastfeeding is inversely associated with blood pressure,
but the possibility of publication bias and residual
confounding cannot be excluded. If causal, the observed
reduction in blood pressure associated with breastfeeding
may have a small, but important effect on public health,
especially in populations where early bottle feeding is
common.
ACKNOWLEDGMENTS
R. M. M. is a Wellcome Trust research training fellow in
clinical epidemiology.
All three authors developed the hypothesis. R. M. M.
acquired the data, performed the analysis, wrote the first

draft of the paper, and coordinated its completion under the
supervision of G. D. S. and D. G. The first draft was signifi-
cantly revised after comments from these two authors. All
authors contributed to and approved the final version.
Help in developing the electronic search of the MEDLINE
and EMBASE databases was provided by Margaret Burke,
Cochrane Heart Group Trials Search Coordinator.
REFERENCES
1. Stary HC. Lipid and macrophage accumulations in arteries of
children and the development of atherosclerosis. Am J Clin
Nutr 2000;72:1297S–306S.
2. Berenson GS, Srinivasan SR, Bao W, et al. Association
between multiple cardiovascular risk factors and atherosclero-
sis in children and young adults. N Engl J Med 1998;338:1650–
6.
3. Zinner SH, Martin LF, Sacks F, et al. A longitudinal study of
blood pressure in childhood. Am J Epidemiol 1974;100:437–
42.
4. Barker DJP. Mothers, babies and health in later life. London,
United Kingdom: Churchill Livingstone, 1998.
5. McCarron P, Davey Smith G, Okasha M, et al. Blood pressure
in young adulthood and mortality from cardiovascular disease.
Lancet 2000;355:1430–1.
6. Geleijnse JM, Hofman A, Witteman JCM, et al. Long-term
effects of neonatal sodium restriction on blood pressure.
Hypertension 1997;29:913–17.

7. Martin RM, McCarthy A, Davies DP, et al. Association
between infant nutrition and blood pressure in early adulthood:
the Barry Caerphilly Growth cohort study. Am J Clin Nutr
2003;77:1489–97.
8. Wilson AC, Forsyth JS, Greene SA, et al. Relation of infant
diet
to childhood health: seven year follow up of cohort of children
in Dundee infant feeding study. BMJ 1998;316:21–5.
9. Cook NR, Cohen J, Hebert PR, et al. Implications of small
reductions in diastolic blood pressure for primary prevention.
Arch Intern Med 1995;155:701–9.
10. Foster K, Lader D, Cheesborough S. Infant feeding 1995: a
sur-
vey of infant feeding practices in the United Kingdom carried
out by the Social Survey Division of ONS on behalf of the
Department of Health, the Scottish Office Department of
Health, the Welsh Office and the Department of Health and
Social Services in Northern Ireland. London, United Kingdom:
The Stationary Office, 1997.
11. Frankel S, Davey Smith G, Gunnell D. Childhood socioeco-
nomic position and adult cardiovascular mortality: the Boyd
Orr Cohort. Am J Epidemiol 1999;150:1081–4.
12. Leeson CP, Kattenhorn M, Deanfield JE, et al. Duration of
breast feeding and arterial distensibility in early adult life: pop-
ulation based study. BMJ 2001;322:643–7.
13. Ravelli ACJ, van der Meulen JH, Osmond C, et al. Infant
feed-
ing and adult glucose tolerance, lipid profile, blood pressure,
and obesity. Arch Dis Child 2000;82:248–52.

14. Taittonen L, Nuutinen M, Turtinen J, et al. Prenatal and
postna-
tal factors in predicting later blood pressure among children:
cardiovascular risk in young Finns. Pediatr Res 1996;40:627–
32.
15. Singhal A, Cole TJ, Lucas A. Early nutrition in preterm
infants
and later blood pressure: two cohorts after randomised trials.
Lancet 2001;357:413–19.
16. Lucas A, Morley R. Does early nutrition in infants born
before
term programme later blood pressure? BMJ 1994;309:304–8.
at K
ingston U
niversity L
ibrary on M
arch 11, 2013
D
ow
nloaded from
17. Simmons D. NIDDM and breastfeeding. Lancet

1997;350:157–
8.
18. Deeks JJ, Altman DG, Bradburn MJ. Statistical methods for
examining heterogeneity and combining results from several
studies in meta-analysis. In: Egger M, Davey Smith G, Altman
DG, eds. Systematic reviews in health care: meta-analysis in
context. London, United Kingdom: BMJ Publishing Group,
2003:285–312.
19. Martin RM, Ben-Shlomo Y, Gunnell D, et al. Breastfeeding
and cardiovascular disease risk factors, incidence and mortal-
ity: the Caerphilly Study. J Epidemiol Community Health (in
press).
20. Lawlor DA, Najman JM, Sterne J, et al. Associations of
paren-
tal, birth, and early life characteristics with systolic blood pres-
sure at 5 years of age: findings from the Mater-University study
of pregnancy and its outcomes. Circulation 2004;110:2417–23.
21. Sterne JAC, Egger M, Davey Smith G. Investigating and
deal-
ing with publication and other biases. In: Egger M, Davey
Smith G, Altman DG, eds. Systematic reviews in health care:
meta-analysis in context. London, United Kingdom: BMJ Pub-
lishing Group, 2001:189–208.
22. Baranowski T, Bryan GT, Harrison JA, et al. Height, infant-
feeding practices and cardiovascular functioning among 3 or 4
year old children in three ethnic groups. J Clin Epidemiol 1992;
45:513–18.
23. Cobaleda Rodrigo A, Hidalgo Vicario MI, Plaza Perez I, et
al.
Prevalence of breast feeding and its relation to cardiovascular

risk factors in the pediatric population of Fuenlabrada. (In
Spanish). An Esp Pediatr 1989;31:350–5.
24. Marmot MG, Page CM, Atkins E, et al. Effect of breast
feeding
on plasma cholesterol and weight in young adults. J Epidemiol
Community Health 1980;34:164–71.
25. Forsyth JS, Willatts P, Agostoni C, et al. Long chain
polyunsat-
urated fatty acid supplementation in infant formula and blood
pressure in later childhood. BMJ 2003;326:953–5.
26. Smith RE, Kok A, Rothberg AD, et al. Determinants of
blood
pressure in Sowetan infants. S Afr Med J 1995;85:1339–42.
27. Kolacek S, Kapetanovic T, Luzar V. Early determinants of
car-
diovascular risk in adults. B. Blood pressure. Acta Paediatr
1993;82:377–82.
28. Kolacek S, Kapetanovic T, Zimolo A, et al. Influence of
early
infant nutrition on cardiovascular risk factors in adults. (In
Croatian). Arhiv Za Zastitu Majke i Djeteta 1990;34:215–26.
29. Strbak V, Hromadova M, Kostalova L, et al. Search for
optimal
age for weaning. Ten year prospective study. Endocr Regul
1993;27:215–21.
30. Strbak V, Skultetyova M, Hromadova M, et al. Late effects
of
breast-feeding and early weaning: seven year prospective study
in children. Endocr Regul 1991;25:53–7.

31. Roberts SB. Prevention of hypertension in adulthood by
breast-
feeding? Lancet 2001;357:406–7.
32. Zinner SH, Lee YH, Rosner B, et al. Factors affecting blood
pressures in newborn infants. Hypertension 1980;2:99–101.
33. Fall C. Nutrition in early life and later outcome. Eur J Clin
Nutr
1992;46:S57–S63.
34. Whincup PH, Cook DG, Shaper AG. Early influences on
blood
pressure: a study of children aged 5–7 years. BMJ 1989;299:
587–91.
35. Zeman J, Simkova M. Blood pressure values in infants and
young children in relation to the duration of breast feeding. (In
Czech). Ceska Slov Pediatr 1981;36:593–4.
36. Boulton J. Nutrition in childhood and its relationships to
early
somatic growth, body fat, blood pressure, and physical fitness.
Acta Paediatr Scand Suppl 1981;284:1–85.
37. Simpson A, Mortimer JG, Silva PA, et al. Correlates of
blood
pressure in a cohort of Dunedin seven year old children. In:
Onesti G, Kim KE, eds. Hypertension in the young and the old.
New York, NY: Grune and Stratton, 1981:155–63.
38. Lucas A, Morley R, Hudson GJ, et al. Early sodium intake
and
later blood pressure in preterm infants. Arch Dis Child 1988;63:

at SciVerse ScienceDirectSocial Science & Medicine 75 (201.docx

at SciVerse ScienceDirectSocial Science & Medicine 75 (201.docx

Recommended

Recommended

More Related Content

Similar to at SciVerse ScienceDirectSocial Science & Medicine 75 (201.docx

Similar to at SciVerse ScienceDirectSocial Science & Medicine 75 (201.docx (20)

More from ikirkton

More from ikirkton (20)

Recently uploaded

Recently uploaded (20)

at SciVerse ScienceDirectSocial Science & Medicine 75 (201.docx