Anthropometry and body composition of 18 year old men according to duration of breast feeding: birth cohort study from Brazil

Introduction

Whether breast feeding protects against obesity is a hotly debated issue in the literature. In a review in 2001, Butte concluded that “the protective effect of breastfeeding on later obesity remains controversial.”1 Later in 2001 Dietz reviewed two recent papers from the United States.2 The first reported a significant association between any breast feeding and being at risk of overweight (85th to 94th centiles of body mass index), but not with actual overweight (95th centile or above), for children aged 3-5 years.3 The second paper showed that children aged 9-14 years who had been mostly breast fed for the first six months of life had a 20% lower risk of overweight than those who were only or mostly fed infant formula.4 In 2003 Dewey suggested that most large studies show a protective effect of breast feeding against overweight in children and adolescents.5

We report on a birth cohort of over 2000 male Brazilians followed for 18 years, for whom detailed breastfeeding information was collected in early childhood. In addition to several anthropometric measures of adiposity, we ascertained body composition through bio-impedance.

Methods

Pelotas (population 320 000) is a relatively affluent city in southern Brazil. The population is mostly of European background. The mean annual gross domestic product per capita is about US$2700 (£1690; €2398), and infant mortality in 1982 was 38 per thousand.

The study recruited all 5914 infants born alive during 1982 in the three maternity hospitals in Pelotas (over 99% of all city births).6 7 The infants were examined, and their mothers were interviewed. In early 1983 children born in January-April 1982 were sought at home. In 1984 and 1986 all households in the city were visited in search of cohort children; 87% and 84% of the original cohort were located. Duration of breast feeding and age at introduction of fluids and foods were recorded. In 2000 we invited men in the cohort who were registering at the local army base to join the study. Informed consent was obtained in all phases of the study, and confidentiality was ensured. A single anthropometrist carried out each measurement on all participants.

We used standard definitions of breast feeding.8 Total duration in months and days was collected at the 1983, 1984, and 1986 follow ups. Exclusive breast feeding was of short duration, because virtually all children also received herbal teas or water from the first week of life and were thus classified as predominantly breast fed. Duration of predominant breast feeding recorded the age when foods other than breast milk or teas or water were introduced. We used the earliest available information in order to minimise recall bias.

Confounding variables were measured in 1982: monthly family income, maternal education, pre-pregnancy body mass index (in kg/m²; height was measured soon after delivery, and weight was obtained from antenatal records or by recall), smoking during pregnancy (non-smokers, 0-14, or ≥ 15 cigarettes a day), birth weight, and gestational age (estimated by the date of the last menstrual period and recorded as less than 37, 37-38, 39-41, or ≥ 42 weeks or unknown). We also collected information on confounders in 2000: skin colour (self reported, classified as white or other), physical activity (minutes per week spent on physical exercise), type of diet (Block classification9: best low fat food, low fat food, eating American diet, diet is quite high in fat, diet is high in fat), daily smoking, and alcohol intake in the preceding week.

We measured several outcomes in 2000. We measured standing and sitting height with a CMS stadiometer (London, UK) and obtained leg length by subtraction. We weighed participants in their underpants with a Tanita body fat analyser scale (model TBF-305, Tokyo, Japan). We measured subscapular and triceps skinfolds by using a Holtain skinfold caliper (Dyfed, UK). We used standard World Health Organization definitions and reference curves.10 We defined overweight as body mass index at or above the 85th centile of the sex and age specific value and obesity as body mass index at or above the 85th centile plus subscapular and triceps skinfolds at or above the 90th centile. We repeated all analyses with body mass index cut-offs of 25 kg/m² and 30 kg/m².

We estimated fat mass and fat-free mass in kg by using the Tanita bio-impedance scale. A validation substudy in a mirror sample of 48 participants in the age range of the study cohort assessed total body water by deuterium dilution.11 We used the resulting validation equation (total body water = 4.437+(0.378 x weight)+(0.189 x height²)/impedance standard) to calculate fat mass (total body water/0.732) and fat-free mass (weight minus fat mass). We calculated percentage fat mass by dividing fat mass by total weight. We also calculated fat mass/height² and fat-free mass/height². We selected the power of two because log-log regressions of fat and fat-free mass in kg over height showed regression coefficients of 2.12 and 2.05.12

We analysed dichotomous outcomes by using χ² tests for heterogeneity and for linear trends in proportions. We used logistic regression to adjust for confounding. We compared means by using analysis of variance (crude analyses) and general linear models (adjusted analyses). Except for birth weight (a continuous variable), we categorised confounders as shown in table 1.

Results

Of the 3037 liveborn boys in the cohort, we examined 2250 (74.1%) in 2000, and 143 (4.7%) are known to have died. We therefore traced 78.8% of the original cohort. Another 37 cohort members were identified but refused to participate.

Table 1 shows that middle class participants were more often traced than either poor or wealthy ones. Regarding maternal education, follow up rates were highest for the 5-8 years and ≥ 12 years groups. No significant differences in follow up occurred according to birth weight or duration of breast feeding. Only 8.4% of all mothers reported not having attempted to breast feed, and 14.9% reported breast feeding for 1-29 days. Misclassification between these two groups seemed to be frequent,13 and we merged them in the analyses.

Both short (less than one month) and long (12 months or more) durations of breast feeding were more prevalent in poor families with less educated mothers (data available on request). Non-white infants tended to be breast fed for longer than white ones, and birth weight was inversely associated with duration of breast feeding. Similar patterns occurred for predominant breast feeding. Income, maternal education, smoking, skin colour, maternal body mass index, gestational age, and birth weight were treated as potential confounders in the multivariable analyses. No associations occurred between breast feeding and current behavioural variables (smoking, alcohol drinking, type of diet, or physical exercise), but we retained these variables in the model as covariates. Table 2 shows characteristics of the sample in 2000.

Table 3 shows the unadjusted effects of total duration of breast feeding. No significant associations occurred with the prevalence of overweight or with mean body mass index, fat, or fat-free mass, but obesity was about three times less prevalent in participants who were breast fed for three to five months than in the other categories (P = 0.004). Height increased steadily with duration of breast feeding until 11 months, with a significant overall linear trend. A significant linear association also occurred with percentage leg length, but the differences were small.

Table 4 shows adjusted analyses of overweight and obesity. Total duration of breast feeding remained unassociated with overweight, but the lower risk of obesity for participants breast fed for three to five months persisted, with an adjusted odds ratio of 0.41 compared with those breast fed for 12 months or longer (P = 0.006). Table 5 shows adjusted analyses of continuous outcomes. No association occurred between total duration of breast feeding and mean body mass index, fat, or fat-free mass. The association with height was almost significant (P = 0.06), but differences were small. No association with leg length occurred.

We repeated all analyses for predominant breast feeding (table 6). In the crude analyses, we found an inverse association with obesity (P = 0.03) and direct associations with height (P = 0.03) and possibly with percentage leg length (P = 0.06). Adjustment for confounding confirmed the lack of association with overweight (table 7). However, we found evidence of a protective effect against obesity (P = 0.03), although children who were predominantly breast fed for at least four months showed a slight increase compared with those breast fed for three months.

Table 8 shows that after adjustment for confounders no significant associations occurred between duration of predominant breast feeding and body mass index, fat, fat-free mass, height, or percentage leg length. We repeated all analyses with the adult cut-offs for body mass index of 25 kg/m² and 30 kg/m²; results were virtually unchanged.

Discussion