Papers

BMJ: first published as 10.1136/bmj.327.7420.901 on 16 October 2003. Downloaded from http://www.bmj.com/ on 23 August 2021 by guest. Protected by copyright.
Anthropometry and body composition of 18 year old men
according to duration of breast feeding: birth cohort
study from Brazil
Cesar G Victora, Fernando C Barros, Rosângela C Lima, Bernardo L Horta, Jonathan Wells

Abstract risk of overweight (85th to 94th centiles of body mass Post-Graduate

Programme in
index), but not with actual overweight (95th centile or Epidemiology,
Objective To assess the association between duration above), for children aged 3-5 years.3 The second paper Universidade
of breast feeding and measures of adiposity in showed that children aged 9-14 years who had been Federal de Pelotas,
adolescence. mostly breast fed for the first six months of life had a
CP 464, 96001-970,
Pelotas, RS, Brazil
Design Population based birth cohort study. 20% lower risk of overweight than those who were only Cesar G Victora
Setting Pelotas, a city of 320 000 inhabitants in a or mostly fed infant formula.4 In 2003 Dewey professor
relatively developed area in southern Brazil. suggested that most large studies show a protective Rosângela C Lima
Participants All newborn infants in the city’s hospitals effect of breast feeding against overweight in children
post-doctoral fellow
were enrolled in 1982; 78.8% (2250) of all male and adolescents.5 PAHO/WHO
participants were located at age 18 years when Latin-American
We report on a birth cohort of over 2000 male Center for
enrolling in the national army. Brazilians followed for 18 years, for whom detailed Perinatology and
Main outcome measures Weight, height, sitting breastfeeding information was collected in early child- Human
height, subscapular and triceps skinfolds, and body Development,
hood. In addition to several anthropometric measures Montevideo,
composition (body fat, lean mass). of adiposity, we ascertained body composition through Uruguay
Results Neither the duration of total breast feeding bio-impedance. Fernando Barros
nor that of predominant breast feeding (breast milk consultant
plus non-nutritive fluids) showed consistent Faculty of Medicine,
associations with anthropometric or body Methods Catholic University
of Pelotas, Brazil
composition indices. After adjustment for Pelotas (population 320 000) is a relatively affluent city Bernardo L Horta
confounding factors, the only significant associations in southern Brazil. The population is mostly of associate professor
were a greater than 50% reduction in obesity among European background. The mean annual gross domes- Institute of Child
participants breast fed for three to five months tic product per capita is about US$2700 (£1690; €2398), Health, University
compared with all other breastfeeding categories of London, London
and infant mortality in 1982 was 38 per thousand. Jonathan Wells
(P = 0.007) and a linear decreasing trend in obesity The study recruited all 5914 infants born alive dur- lecturer
with increasing duration of predominant breast ing 1982 in the three maternity hospitals in Pelotas Correspondence to:
feeding (P = 0.03). Similar significant effects were not (over 99% of all city births).6 7 The infants were C G Victora
observed for other measures of adiposity. Borderline examined, and their mothers were interviewed. In early cvictora@
direct associations also occurred between total terra.com.br
1983 children born in January-April 1982 were sought
duration of breast feeding and adult height (P = 0.06). at home. In 1984 and 1986 all households in the city bmj.com 2003;327:901
Conclusions The significant reduction in obesity were visited in search of cohort children; 87% and 84%
among children breast fed for three to five months is of the original cohort were located. Duration of breast
difficult to interpret, as no a priori hypothesis existed feeding and age at introduction of fluids and foods
regarding a protective effect of intermediate duration were recorded. In 2000 we invited men in the cohort
of breast feeding. The findings indicate that, in this who were registering at the local army base to join the
population, breast feeding has no marked protective study. Informed consent was obtained in all phases of
effect against adolescent adiposity. the study, and confidentiality was ensured. A single
anthropometrist carried out each measurement on all
participants.
Introduction We used standard definitions of breast feeding.8
Whether breast feeding protects against obesity is a Total duration in months and days was collected at the
hotly debated issue in the literature. In a review in 1983, 1984, and 1986 follow ups. Exclusive breast feed-
2001, Butte concluded that “the protective effect of ing was of short duration, because virtually all children
breastfeeding on later obesity remains controversial.”1 also received herbal teas or water from the first week of
Later in 2001 Dietz reviewed two recent papers from life and were thus classified as predominantly breast
the United States.2 The first reported a significant fed. Duration of predominant breast feeding recorded
association between any breast feeding and being at the age when foods other than breast milk or teas or

BMJ VOLUME 327 18 OCTOBER 2003 bmj.com page 1 of 5

Papers

water were introduced. We used the earliest available

Table 1 Percentage of cohort located according to
information in order to minimise recall bias.

BMJ: first published as 10.1136/bmj.327.7420.901 on 16 October 2003. Downloaded from http://www.bmj.com/ on 23 August 2021 by guest. Protected by copyright.
socioeconomic characteristics, birth weight, and breast feeding
Confounding variables were measured in 1982:
monthly family income, maternal education, pre- Original cohort No (%) traced
Baseline characteristic (No (%)) in 2000
pregnancy body mass index (in kg/m2; height was Monthly family income (US$) in 1982:
measured soon after delivery, and weight was obtained ≤50 666 (22.0) 483 (72.5)
from antenatal records or by recall), smoking during 51-150 1463 (48.4) 1170 (80.0)
pregnancy (non-smokers, 0-14, or ≥ 15 cigarettes a 151-300 544 (18.0) 456 (83.8)
day), birth weight, and gestational age (estimated by the 301-500 184 (6.1) 146 (79.3)
date of the last menstrual period and recorded as less >500 167 (5.5) 128 (76.6)
than 37, 37-38, 39-41, or ≥ 42 weeks or unknown). We Total 3024* P<0.001†
also collected information on confounders in 2000: Maternal schooling (years) in 1982:
skin colour (self reported, classified as white or other), 0-4 1008 (33.2) 768 (76.2)
physical activity (minutes per week spent on physical 5-8 1288 (42.5) 1048 (81.4)

exercise), type of diet (Block classification9: best low fat 9-11 330 (10.9) 249 (75.5)
≥12 406 (13.4) 323 (79.6)
food, low fat food, eating American diet, diet is quite
Total 3032* P=0.008†
high in fat, diet is high in fat), daily smoking, and alco-
Birth weight (g) in 1982:
hol intake in the preceding week.
<2500 g 244 (8.0) 189 (77.5)
We measured several outcomes in 2000. We ≥2500 g 2791 (92.0) 2344 (78.9)
measured standing and sitting height with a CMS Total 3035* P=0.30†
stadiometer (London, UK) and obtained leg length by Duration of total breast feeding (months), 1983-6:
subtraction. We weighed participants in their under- <1 636 (23.3) 511 (80.3)
pants with a Tanita body fat analyser scale (model 1-2.9 708 (25.9) 577 (81.5)
TBF-305, Tokyo, Japan). We measured subscapular and 3-5.9 611 (22.3) 519 (84.9)
triceps skinfolds by using a Holtain skinfold caliper 6-8.9 261 (9.5) 210 (80.5)
(Dyfed, UK). We used standard World Health 9-11.9 114 (4.2) 90 (78.9)
Organization definitions and reference curves.10 We ≥12 405 (14.8) 328 (81.0)

defined overweight as body mass index at or above the Total 2735* P=0.29†
Duration of predominant breast feeding (months), 1983-6:
85th centile of the sex and age specific value and obes-
<1 757 (28.9) 619 (81.8)
ity as body mass index at or above the 85th centile plus
1-1.9 405 (15.4) 328 (81.0)
subscapular and triceps skinfolds at or above the 90th
2-2.9 489 (18.6) 407 (83.2)
centile. We repeated all analyses with body mass index
3-3.9 662 (25.2) 561 (84.7)
cut-offs of 25 kg/m2 and 30 kg/m2. ≥4 309 (11.8) 243 (78.6)
We estimated fat mass and fat-free mass in kg by Total 2622* P=0.17†
using the Tanita bio-impedance scale. A validation *Total numbers do not add to 3037 owing to missing information.
substudy in a mirror sample of 48 participants in the age †
2 test for heterogeneity (
2 test for proportions; analysis of variance for means).
range of the study cohort assessed total body water by
deuterium dilution.11 We used the resulting validation highest for the 5-8 years and ≥ 12 years groups. No
equation (total body water = 4.437+(0.378 × weight)+ significant differences in follow up occurred according
(0.189 × height2)/impedance standard) to calculate fat to birth weight or duration of breast feeding. Only
mass (total body water/0.732) and fat-free mass (weight 8.4% of all mothers reported not having attempted to
minus fat mass). We calculated percentage fat mass by breast feed, and 14.9% reported breast feeding for 1-29
dividing fat mass by total weight. We also calculated fat days. Misclassification between these two groups
mass/height2 and fat-free mass/height2. We selected the seemed to be frequent,13 and we merged them in the
power of two because log-log regressions of fat and fat- analyses.
free mass in kg over height showed regression Both short (less than one month) and long (12
coefficients of 2.12 and 2.05.12 months or more) durations of breast feeding were
We analysed dichotomous outcomes by using
2 more prevalent in poor families with less educated
tests for heterogeneity and for linear trends in propor- mothers (data available on request). Non-white infants
tions. We used logistic regression to adjust for tended to be breast fed for longer than white ones, and
confounding. We compared means by using analysis of birth weight was inversely associated with duration of
variance (crude analyses) and general linear models breast feeding. Similar patterns occurred for predomi-
(adjusted analyses). Except for birth weight (a continu- nant breast feeding. Income, maternal education,
ous variable), we categorised confounders as shown in smoking, skin colour, maternal body mass index,
table 1. gestational age, and birth weight were treated as

Results Table 2 Distribution of sample according to key indicators

Of the 3037 liveborn boys in the cohort, we examined (n=2250)
2250 (74.1%) in 2000, and 143 (4.7%) are known to Sample characteristic No %
have died. We therefore traced 78.8% of the original Overweight 392 17.4
cohort. Another 37 cohort members were identified Obese 182 8.1
but refused to participate. Daily smokers 356 15.8
Table 1 shows that middle class participants were Physical activity <90 min/week 403 17.9
more often traced than either poor or wealthy ones. High fat diet 1085 48.2

Regarding maternal education, follow up rates were Alcohol intake in previous week 1352 60.1

page 2 of 5 BMJ VOLUME 327 18 OCTOBER 2003 bmj.com

 Papers

Table 3 Anthropometry and body composition of 18 year old men according to total duration of breast feeding. Values are means

BMJ: first published as 10.1136/bmj.327.7420.901 on 16 October 2003. Downloaded from http://www.bmj.com/ on 23 August 2021 by guest. Protected by copyright.
(SDs) unless stated otherwise
Duration of total breast feeding (months)
Outcomes <1 1-2 3-5 6-8 9-11 ≥12 P value* All participants
Overweight (No (%)) 69 (14.1) 75 (13.5) 55 (11.0) 33 (16.2) 17 (19.3) 44 (13.7) 0.3 291 (13.5)
Obese (No (%)) 47 (9.6) 50 (9.0) 19 (3.8) 22 (10.8) 9 (10.2) 27 (8.4) 0.004 125 (8.1)
Body mass index 22.5 (3.8) 22.3 (3.6) 22.1 (3.2) 22.6 (4.4) 22.9 (3.7) 22.5 (4.2) 0.25 22.4 (3.8)
Fat mass percentage 17.6 (2.9) 17.8 (2.8) 17.7 (2.8) 17.9 (3.0) 18.3 (2.6) 17.9 (2.7) 0.42 17.8 (2.8)
Fat mass/height2 4.0 (1.2) 4.0 (1.1) 3.9 (1.0) 4.1 (1.3) 4.2 (1.0) 4.1 (1.3) 0.19 4.0 (1.1)
Lean mass/height2 18.5 (2.9) 18.3 (2.7) 18.2 (2.5) 18.6 (3.2) 18.7 (2.9) 18.4 (3.1) 0.27 18.3 (2.8)
Height in cm 172.7 (6.9) 173.1 (6.8) 173.5 (6.5) 174.3 (7.0) 174.7 (6.8) 173.9 (6.7) 0.001† 173.4 (6.8)
Percentage leg/total 44.5 (1.4) 44.6 (1.3) 44.7 (1.3) 44.8 (1.3) 44.7 (1.2) 44.7 (1.5) 0.02† 44.6 (1.3)
height
No of participants‡ 489 555 498 204 88 321 2155
*Test for heterogeneity (
2 test for proportions; analysis of variance for means).
†Test for linear trend.
‡The number of participants ranged from 2127 to 2155, depending on the outcome.

potential confounders in the multivariable analyses. longer (P = 0.006). Table 5 shows adjusted analyses of
No associations occurred between breast feeding and continuous outcomes. No association occurred
current behavioural variables (smoking, alcohol drink- between total duration of breast feeding and mean
ing, type of diet, or physical exercise), but we retained body mass index, fat, or fat-free mass. The association
these variables in the model as covariates. Table 2 with height was almost significant (P = 0.06), but differ-
shows characteristics of the sample in 2000. ences were small. No association with leg length
Table 3 shows the unadjusted effects of total dura- occurred.
tion of breast feeding. No significant associations We repeated all analyses for predominant breast
occurred with the prevalence of overweight or with feeding (table 6). In the crude analyses, we found an
mean body mass index, fat, or fat-free mass, but obesity inverse association with obesity (P = 0.03) and direct
was about three times less prevalent in participants associations with height (P = 0.03) and possibly with
who were breast fed for three to five months than in percentage leg length (P = 0.06). Adjustment for
the other categories (P = 0.004). Height increased confounding confirmed the lack of association with
steadily with duration of breast feeding until 11 overweight (table 7). However, we found evidence of a
months, with a significant overall linear trend. A protective effect against obesity (P = 0.03), although
significant linear association also occurred with children who were predominantly breast fed for at least
percentage leg length, but the differences were small. four months showed a slight increase compared with
Table 4 shows adjusted analyses of overweight and those breast fed for three months.
obesity. Total duration of breast feeding remained Table 8 shows that after adjustment for confound-
unassociated with overweight, but the lower risk of ers no significant associations occurred between dura-
obesity for participants breast fed for three to five tion of predominant breast feeding and body mass
months persisted, with an adjusted odds ratio of 0.41 index, fat, fat-free mass, height, or percentage leg
compared with those breast fed for 12 months or length. We repeated all analyses with the adult cut-offs

Table 4 Logistic regression analyses for prevalence of overweight and obesity among 18 year old men according to duration of total
breast feeding (n=2082)
Odds ratio (95% CI) according to duration of total breast feeding (months)
Outcomes Model <1 1-2 3-5 6-8 9-11 ≥12 P value*
Overweight (%) Crude 1.04 (0.69 to 1.56) 0.98 (0.66 to 1.47) 0.78 (0.51 to 1.19) 1.21 (0.74 to 1.98) 1.51 (0.81 to 2.78) 1.0 0.27
Adjusted† 1.14 (0.77 to 1.68) 0.91 (0.61 to 1.35) 0.85 (0.57 to 1.27) 0.86 (0.53 to 1.40) 1.34 (0.74 to 2.44) 1.0 0.41
Obese (%) Crude 1.16 (0.71 to 1.91) 1.08 (0.66 to 1.76) 0.43 (0.24 to 0.79) 1.32 (0.73 to 2.38) 1.24 (0.56 to 2.74) 1.0 0.007
Adjusted† 1.08 (0.63 to 1.84) 1.09 (0.64 to 1.84) 0.38 (0.20 to 0.72) 1.20 (0.64 to 2.24) 1.05 (0.46 to 2.42) 1.0 0.006
*Likelihood ratio test.
†Adjusted for family income and maternal education at birth, maternal body mass index, skin colour, birth weight, gestational age, maternal smoking during
pregnancy, and current behavioural variables (smoking, alcohol drinking, type of diet, and physical exercise).

Table 5 Anthropometry and body composition of 18 year old men according to duration of total breast feeding: adjusted analyses*.
Values are means (SEs)
Duration of total breast feeding (months)
Outcomes <1 1-2 3-5 6-8 9-11 ≥12 P value
Body mass index 23.0 (0.2) 22.9 (0.2) 22.6 (0.2) 22.9 (0.3) 23.3 (0.4) 23.0 (0.2) 0.35
Fat mass percentage 17.5 (0.2) 17.7 (0.2) 17.6 (0.2) 17.8 (0.2) 18.2 (0.3) 17.8 (0.2) 0.24
Fat mass/height2 4.1 (0.1) 4.1 (0.1) 4.0 (0.1) 4.2 (0.1) 4.3 (0.1) 4.2 (0.1) 0.28
Lean mass/height2 18.9 (0.2) 18.8 (0.2) 18.6 (0.2) 18.8 (0.2) 19.1 (0.3) 18.9 (0.2) 0.45
Height in cm 173.9 (0.4) 174.1 (0.4) 174.4 (0.4) 174.5 (0.5) 174.8 (0.7) 174.8 (0.4) 0.49
Percentage leg/total height 44.6 (0.1) 44.6 (0.1) 44.7 (0.1) 44.8 (0.1) 44.7 (0.2) 44.6 (0.1) 0.52
*General linear model (general factorial); estimates adjusted for family income and maternal education at birth, maternal body mass index, skin colour, birth weight
and gestational age, maternal smoking during pregnancy, and current behavioural variables (smoking, alcohol drinking, type of diet, and physical exercise). The
number of participants ranged from 2106 to 2133, depending on the outcome.

BMJ VOLUME 327 18 OCTOBER 2003 bmj.com page 3 of 5

Papers

Table 6 Anthropometry and body composition of 18 year old men according to the duration of predominant breast feeding. Values

BMJ: first published as 10.1136/bmj.327.7420.901 on 16 October 2003. Downloaded from http://www.bmj.com/ on 23 August 2021 by guest. Protected by copyright.
are means (SDs) unless stated otherwise
Duration of predominant breast feeding (months)
Outcomes <1 1-1.9 2-2.9 3-3.9 ≥4 P value* All participants
Overweight (%) 14.9 14.0 11.4 13.1 14.5 0.63 13.5
Obese (%) 10.3 8.3 7.0 6.5 7.4 0.03† 8.1
Body mass index 22.5 (3.7) 22.2 (3.4) 22.2 (3.7) 22.5 (4.0) 22.4 (3.6) 0.70 22.3 (3.7)
Fat mass/height2 4.0 (1.2) 4.0 (1.1) 4.0 (1.1) 4.0 (1.1) 4.0 (1.1) 0.95 4.0 (1.1)
Lean mass/height2 18.4 (2.9) 18.2 (2.8) 18.2 (2.8) 18.3 (2.8) 18.6 (2.9) 0.42 18.4 (2.8)
Height in cm 173.0 (6.9) 173.2 (6.9) 173.3 (6.4) 173.9 (6.6) 173.6 (7.0) 0.03† 173.4 (6.7)
Percentage leg/total height 44.6 (1.4) 44.5 (1.3) 44.7 (1.3) 44.7 (1.3) 44.7 (1.4) 0.06† 44.6 (1.3)
No of participants‡ 593 312 393 546 237 — 2081
*Test for heterogeneity (
2 test for proportions; analysis of variance for means).
†Test for linear trend.
‡Number of participants ranged from 2051 to 2081, depending on the outcome.

Table 7 Logistic regression analyses for prevalence of overweight and obesity among 18 year old men according to duration of
predominant breast feeding (n=1993)
Odds ratio (95% CI) according to duration of predominant breast feeding (months)
Outcomes Model <1 1-1.9 2-2.9 3-3.9 ≥4 P value*
Overweight (%) Crude 1.03 (0.67 to 1.58) 0.96 (0.60 to 1.56) 0.76 (0.47 to 1.23) 0.89 (0.58 to 1.38) 1.0 0.62
Adjusted† 1.05 (0.69 to 1.58) 0.82 (0.51 to 1.32) 0.86 (0.55 to 1.34) 0.82 (0.54 to 1.25) 1.0 0.56
Obese (%) Crude 1.43 (0.82 to 2.49) 1.12 (0.60 to 2.10) 0.94 (0.50 to 1.74) 0.87 (0.49 to 1.56) 1.0 0.03‡
Adjusted† 1.42 (0.79 to 2.56) 1.14 (0.59 to 2.22) 1.02 (0.53 to 1.94) 0.80 (0.43 to 1.48) 1.0 0.03‡
*Likelihood ratio test.
†Adjusted for family income and maternal education at birth, maternal body mass index, skin colour, birth weight, gestational age, maternal smoking during
pregnancy, and current behavioural variables (smoking, alcohol drinking, type of diet, and physical exercise).
‡Test for linear trend.

for body mass index of 25 kg/m2 and 30 kg/m2; results milk, predominantly breastfed babies received water,
were virtually unchanged. herbal teas, or both, usually starting in the first week of
life, and most of them received fruit juices from the
Discussion age of 2-3 months. Exclusive breast feeding was rare in
this population, so its effect cannot be assessed. One
Main findings and interpretation might speculate whether truly exclusive breast
Our results were mostly negative. Neither the duration feeding, if present, would have resulted in a clearer
of total breast feeding nor that of predominant breast pattern of association with obesity. Nevertheless, we
feeding had marked effects on the anthropometric or found no associations between predominant breast
body composition indices. Allowance for confounding
feeding and any of the other measures of adiposity
factors eliminated some of the associations observed in
studied. We also observed a possible positive effect of
the crude analyses. Two significant associations
total duration of breast feeding on adult height
remained after adjustment. The first was a greater than
(P = 0.06), a finding that was not reported in the
50% reduction in obesity in participants breast fed for
three to five months. Although it is tempting to relate available literature.
this finding to the “critical window” theory of develop- This study has some limitations. About 21% of the
ment, it has to be interpreted with caution, because we participants were lost to follow up over the 18 year
did not find significant reductions in this group for any period; losses were higher at both extremes of the
of the other measures of adiposity, and because we had socioeconomic scale and were not associated with
no a priori hypothesis regarding a protective effect of duration of breast feeding. Another limitation is that
intermediate duration of breast feeding. results are not available for women. About 4.5% of the
The second significant finding was a linear trend cohort died by 18 years of age, but 69% of these deaths
for a reduction in obesity with increasing duration of occurred before 6 months of age. Survivor bias is
predominant breast feeding. In addition to breast therefore unlikely to have affected our analyses.

Table 8 Anthropometry and body composition of 18 year old men according to duration of predominant breast feeding: adjusted
analyses*. Values are means (SEs)
Duration of predominant breast feeding (months)
Outcomes <1 1-1.9 2-2.9 3-3.9 ≥4 P value
Body mass index 23.1 (0.2) 22.9 (0.2) 22.9 (0.2) 22.8 (0.2) 23.1 (0.3) 0.61
Fat mass percentage 17.7 (0.2) 17.6 (0.2) 17.8 (0.2) 17.8 (0.2) 17.5 (0.2) 0.68
Fat mass/height2 4.1 (0.1) 4.1 (0.1) 4.1 (0.1) 4.1 (0.1) 4.1 (0.1) 0.96
Lean mass/height2 19.0 (0.1) 18.8 (0.2) 18.8 (0.2) 18.8 (0.1) 19.1 (0.2) 0.38
Height in cm 174.1 (0.4) 174.5 (0.4) 174.3 (0.4) 174.1 (0.4) 174.5 (0.5) 0.79
Percentage leg/total height 44.7 (0.1) 44.6 (0.1) 44.7 (0.1) 44.6 (0.1) 44.6 (0.1) 0.52
*General linear model (general factorial); estimates adjusted for family income and maternal education at birth, maternal body mass index, skin colour, birth weight
and gestational age, maternal smoking during pregnancy, and current behavioural variables (smoking, alcohol drinking, type of diet, and physical exercise). The
number of participants ranged from 2034 to 2060, depending on the outcome.

page 4 of 5 BMJ VOLUME 327 18 OCTOBER 2003 bmj.com

 Papers

Comparison with literature

A review of the English language literature on breast
What is already known on this topic

BMJ: first published as 10.1136/bmj.327.7420.901 on 16 October 2003. Downloaded from http://www.bmj.com/ on 23 August 2021 by guest. Protected by copyright.
feeding and adolescent adiposity found, in addition to Although no consensus exists, some recent papers
the study by Gillman et al mentioned in the introduc- suggest that breast feeding may protect against
tion,4 three other studies. A Canadian case-control overweight and obesity in adolescence
study compared obese, overweight, and non-obese
12-18 year olds.14 Feeding history was ascertained What this study adds
retrospectively by telephone. A significant increasing This is the largest birth cohort study reporting on
trend in duration of breast feeding occurred among the effect of breast feeding on several measures of
the three groups, but no additional benefit from adiposity, including body composition, in
delayed introduction of solids was seen, in either the adolescence
crude or the adjusted analyses. Tulldahl et al studied
the effect of duration of breast feeding on height, skin- Among 18 year old male Brazilians, duration of
folds, and body composition in 18 year olds.15 Those neither total nor predominant breast feeding was
who were breast fed for longer periods tended to be consistently protective against adolescent adiposity
shorter and leaner, but not all associations were signifi-
cant. In a study of 136 rural American teenagers, breast
feeding for two months or less was positively associated by the Ministry of Health of Brazil. Earlier phases of the cohort
with obesity.16 This association was only present in the study were financed by the International Development
lower socioeconomic status group. However, the Research Center of Canada and by the Overseas Development
association disappeared in the adjusted analysis. Administration of the United Kingdom.
Our study differed from the four studies cited Competing interests: None declared.
above, as it was a prospective cohort study assessing the Ethical approval: The Brazilian Medical Research Council
approved the study protocol.
association of duration and exclusivity of breast
feeding with several measures of adolescent adiposity. 1 Butte NF. The role of breastfeeding in obesity. Pediatr Clin North Am
The only other cohort study on the subject, the 2001;48:189-98.
2 Dietz WH. Breastfeeding may help prevent childhood overweight. JAMA
growing up today study, was a partly retrospective 2001;285:2506-7.
cohort in which information on breast feeding was 3 Hediger ML, Overpeck MD, Kuczmarski RJ, Ruan WJ. Association
obtained through maternal recall when participants between infant breastfeeding and overweight in young children. JAMA
2001;285:2453-60.
were aged 9-14 years, with an overall response rate of 4 Gillman MW, Rifas-Shiman SL, Camargo CA Jr, Berkey CS, Frazier AL,
less than 50%.4 Rockett HR, et al. Risk of overweight among adolescents who were
breastfed as infants. JAMA 2001;285:2461-7.
5 Dewey K. Is breastfeeding protective against child obesity? J Hum Lact
Conclusions 2003;19:9-18.
We cannot at this stage make a strong statement about 6 Barros FC, Victora CG, Vaughan JP. The Pelotas birth cohort study, 1982-
1987: strategies for following up 6,000 children in a developing country.
the protective role of breast feeding against obesity in Perinat Pediatr Epidemiol 1990;4:267-82.
adolescence. Although the subject of breast feeding 7 Victora CG, Barros FC, Lima RC, Behague DP, Gonçalves H, Horta BL,
and obesity in childhood is also still controversial,1 at et al. The Pelotas (Brazil) birth cohort study, 1982-2001. Cad Saude Pub-
lica 2003 (in press).
least one study suggests that there is an inverse 8 World Health Organization, Division of Diarrhoeal and Acute
dose-response relation between the duration of breast Respiratory Disease Control. Indicators for assessing breast-feeding practices.
Geneva: WHO, 1991.
feeding and obesity at 5-6 years of age.17 Our results do 9 Block G, Clifford C, Naughton MD, Henderson M, McAdams M. A brief
not support this finding. dietary screen for high fat intake. J Nutr Educ 1989;21:199-207.
10 World Health Organization, Expert Committee on Nutrition. Physical sta-
Regardless of the role breast feeding may have in tus: uses and interpretation of anthropometry. Geneva: WHO, 1995. (WHO
preventing obesity, it has been consistently associated Technical Report Series, No 854.)
11 Wells JCK, Gigante DP, Wright A, Hallal PC, Victora CG. Validation of
with many advantages for the mother and child, rang- leg-to-leg impedance for body composition assessment among Brazilian
ing from decreased childhood mortality to a likely pro- males aged 16-19 years. Int J Body Composition Res 2003;1(2):1-6.
12 VanItallie TB, Yang M, Heymsfield SB, Funk RC, Boileau RA.
tection against breast cancer.18 19 The continued Height-normalized indices of the body fat-free mass and fat mass: poten-
protection, promotion, and support of breast feeding tially useful indicators of nutritional status. Am J Clin Nutr 1990;52:953-9.
13 Huttly SRA, Victora CG, Barros FC, Beria JU, Vaughan JP. Do mothers
remains a major public health priority. overestimate breastfeeding duration? An example of recall bias from a
study in Southern Brazil. Am J Epidemiol 1990;132:572-5.
We acknowledge the logistical support of the Brazilian Army, in 14 Kramer MS. Do breast-feeding and delayed introduction of solid foods
particular Colonel J C Poppe, Major L M Coutinho, Captain J L protect against subsequent obesity? J Pediatr 1981;98:883-7.
Barros, and Mr O Petiz. 15 Tulldahl J, Pettersson K, Andersson SW, Hulthen L. Mode of infant feed-
ing and achieved growth in adolescence: early feeding patterns in
Contributors: CGV and FCB have coordinated the cohort study relation to growth and body composition in adolescence. Obes Res
from its inception, supervising data collection and analyses and 1999;7:431-7.
writing up the manuscript. RCL coordinated the fieldwork for 16 Elliott KG, Kjolhede CL, Gournis E, Rasmussen KM. Duration of breast-
the current phase of the study and constructed the database. feeding associated with obesity during adolescence. Obes Res 1997;5:538-
41.
BLH carried out data analyses and contributed to the writing of
17 Von Kries R, Koletzko B, Sauerwald T, von Mutius E. Does breast-feeding
the manuscript. JW advised on body composition measurement, protect against childhood obesity? Adv Exp Med Biol 2000;478:29-39.
carried out the validation study for body composition, and pro- 18 WHO Collaborative Study Team on the Role of Breastfeeding on the
vided critical inputs to the analysis and writing up of the manu- Prevention of Infant Mortality. How much does breastfeeding protect
script. All authors approved the final version of the paper. CGV against infant and child mortality due to infectious diseases? A pooled
analysis of six studies from less developed countries. Lancet
is the guarantor for the paper.
2000;355:451-5.
Funding: The study was financed by the Division of Child and 19 Beral V, Bull D, Doll R, Peto R, Reeves G. Breast cancer and breastfeeding.
Adolescent Health of the World Health Organization, by the Lancet 2003;361:e177.
Programa Nacional de Núcleos de Excelência (PRONEX), and (Accepted 24 July 2003)

BMJ VOLUME 327 18 OCTOBER 2003 bmj.com page 5 of 5