Association Between Infant Breastfeeding and Overweight in Young Children

Context It has been suggested that breastfeeding is protective against children becoming overweight, and that there is a dose-dependent effect of its duration.

Objective To determine whether breastfeeding and its duration are associated with a reduced risk of being overweight among young children in the United States.

Design and Setting Data on infant feeding and child overweight status were taken from the third National Health and Nutrition Examination Survey (NHANES III), a cross-sectional health examination survey conducted from 1988-1994.

Subjects Sample of 2685 US-born children between the ages of 3 and 5 years, with birth certificates, height and weight measures, and information on infant feeding.

Main Outcome Measures A body mass index (BMI) between the 85th and 94th percentile was considered at risk of overweight and a BMI in the 95th percentile or higher was considered being overweight.

Results After adjusting for potential confounders, there was a reduced risk of being at risk of overweight for ever breastfed children (adjusted odds ratio [AOR], 0.63; 95% confidence interval [CI], 0.41-0.96) compared with those never breastfed. There was no reduced risk of being overweight (AOR, 0.84; 95% CI, 0.62-1.13). There was no clear dose-dependent effect of the duration of full breastfeeding on being at risk of overweight or overweight and no threshold effect. The strongest predictor of child overweight status was the mother's concurrent weight. The rate of children being overweight nearly tripled with maternal overweight status (BMI, 25.0-29.9 kg/m²; AOR, 2.95; 95% CI, 1.35-6.42) and more than quadrupled with maternal obesity status (BMI ≥30.0 kg/m²; AOR, 4.34; 95% CI, 2.50-7.54).

Conclusions There are inconsistent associations among breastfeeding, its duration, and the risk of being overweight in young children. Breastfeeding continues to be strongly recommended, but may not be as effective as moderating familial factors, such as dietary habits and physical activity, in preventing children from becoming overweight.

With rare exceptions, breast milk is the preferred feeding for infants and confers unique immunologic, growth, and developmental benefits.^1,2 Immunologic benefit and reduction in risk for or severity of many acute and chronic diseases, including diarrhea, lower respiratory tract infections, urinary tract infections, otitis media, and asthma, have been associated with exclusive breastfeeding for at least 4 months.^1-3 Shorter and less intensive periods of breastfeeding may be less protective. Cognitive developmental scores for infants who were fully breastfed for 6 months or more are increased in comparison with those scores of infants who were never breastfed.⁴

There is some suggestion that breastfeeding is also protective against the development of overweight in young children and adolescents.^5,6 If breastfeeding is protective against overweight in early childhood, the explanation may lie less in inherent obesity-preventing properties of breast milk than in the fact that breastfeeding displaces potentially more energy-dense formula-feeding. In addition, the higher protein/nitrogen content of infant formula compared with breast milk may cause a metabolic response of increased insulin and insulin-like growth factor-1 secretion in formula-fed infants leading to excessive weight gain, or there may be differences in the regulation of intake of formula-fed infants compared with breastfed infants.

Any protective mechanism of breastfeeding is difficult to identify because many of the same factors related to child growth and the development of overweight, such as race/ethnicity, maternal education, maternal size, and birth weight, are also related to the initiation and duration of breastfeeding or the decision to formula-feed.^7-11 Our objective was to determine if breastfeeding and its duration are associated with a reduced risk of overweight in a nationally representative sample of young US-born children.

Information on infant feeding and child overweight status were taken from the National Health and Nutrition Examination Survey III (NHANES III), a cross-sectional survey conducted by the National Center for Health Statistics/Centers for Disease Control and Prevention (NCHS/CDC) from 1988-1994.¹² As part of the stratified, multistage probability design, infants and children from ages 2 to 71 months at interview, blacks, and Mexican-Americans were oversampled. These analyses focus on young children, 3 to 5 years (36-71 months) at interview, for whom retrospective information on infant feeding was obtained. NHANES III included 3461 children, ages 3 to 5 years, and 94.1% (n = 3257) were measured for height and weight.

Birth certificates were sought for US-born children ages 6 years or younger.^13-15 For these analyses, 2879 birth certificates (94% of US-born children ages 3-5 years) were linked to the NHANES III file. Variables taken from the birth certificates included sex, birth plurality (singleton or twin), birth order, birth weight (grams), and length of gestation (weeks). As in previous studies using the linked file,^10,13-15 gestational length from the mother's last menstrual period was examined for completeness and validity^16-18; 235 subjects (8.2%) were found to have missing or invalid length of gestation. For those with valid gestational ages, preterm delivery was defined as less than 37 weeks and term as 37 weeks or more.

Birth-weight categories were defined as low birth weight (<2500 g), normal (2500-3999 g), and high birth weight (≥4000 g). Children with valid gestational ages were categorized as small (<10th percentile of birth weight for gestation), appropriate (10th-89th percentile), and large for gestational age (≥90th percentile), using reference percentiles derived separately for non-Hispanic white, non-Hispanic black, and Mexican-American singleton infants.¹⁸ Infants of "other" races/ethnicities were categorized using the reference for non-Hispanic whites.

Race/ethnicity and region of residence were based on US Bureau of the Census categories and definitions.¹² The education of the family reference person in NHANES III was used to represent household education level. The following were taken from the questionnaire administered to the child's parent or other adult caretaker in the home: mother's age at the time of the child's birth, smoking during pregnancy, reported current weight and height, and whether the infant received special neonatal care and duration of that care. Sex was taken from NHANES III and checked for discrepancies with the birth certificate. Mother's weight and height were converted into a body mass index (BMI, kg/m²), except for 145 subjects who had missing information. Maternal BMI was classified as underweight (< 18.5 kg/m²), normal weight (18.5-24.9 kg/m²), overweight (25.0-29.9 kg/m²), and obese (≥30.0 kg/m²).¹⁹ In multivariate analyses, the underweight and normal weight women were combined as the reference because only 88 women were classified as underweight.

Questions on infant feeding practices included whether the child was ever breastfed and the age (converted into months) when the child completely stopped breastfeeding, was first fed formula daily, completely stopped drinking formula, was first fed milk daily, and started eating solid foods daily.¹² Infant feeding was defined based on the duration (in months) that the children were fully breastfed (no liquids daily other than breast milk or water). Infants who were partially breastfed (supplemented daily with formula or milk) were no longer considered fully breastfed once daily supplementation with formula or milk began. Feeding groups were classified as never breastfed and fully breastfed for 2 months or less, 3 to 5 months, 6 to 8 months, and 9 months or more. Timing (in months) of the introduction of solid foods was considered separately because timing of the introduction of solids does not appear to alter growth and body composition in the first year of life.^20,21

Body weight and height were measured using standard anthropometric techniques.^22,23 Body mass index was calculated, and weight status was defined using BMI-for-age percentiles from the revised NCHS/CDC growth charts.²⁴ Body mass index and weight status based on BMI-for-age were chosen as the primary dependent variables because BMI is the worldwide standard for screening for overweight in children, it is based on highly reliable measurements (height and weight), and even in early childhood BMI is highly correlated with fatness.^24-26 Twenty-nine children had missing heights or weights. Following current guidelines for children and adolescents, a BMI between the 85th through 94th percentile was considered "at risk of overweight" and at or above the 95th percentile as overweight.^24-26 Young children generally are not designated "obese," even at or above the 95th percentile of BMI-for-age, in the absence of clinical evaluation. The 95th percentile corresponds to a BMI of 30.0 or grade 2 obesity in adults.¹⁹

From the base sample of 2879 subjects with interview, examination, and birth certificate data, 76 twins and triplets were excluded, as were cases missing birth weight (n = 5), sex (n = 4), and infant feeding (n = 5). Twenty-nine children born very low birth weight (<1500 g) were excluded because they were not sufficiently mature at birth to initiate breastfeeding, and they remained small through childhood. For the same reason, 75 children with extended periods (≥2 weeks) of special neonatal care were excluded. The final analytic sample was 2685 subjects.

Following NCHS/CDC analytic guidelines, statistical sample weights for examined children were used (weighted) to account both for the oversampling and unit nonresponse.²⁷ Both nonresponse and poststratification adjustments impose a great deal of heterogeneity into the weights, and thus ignoring the sample weights in this case would have imposed a bias in both the prevalence estimates and in the estimates of sampling error.

SUDAAN software, which incorporates the heterogeneity in survey weight along with the clustering effects in survey design, was used for statistical analysis and to estimate standard errors (SEs).²⁸ The effect of duration of full breastfeeding (in months) on child BMI was examined by multiple regression. The likelihood of being at risk of overweight and overweight among fully breastfed as compared with never breastfed children was estimated using the SUDAAN MULTILOG procedure. MULTILOG is an extension of traditional logistic regression with the capability to model multiple level outcomes. Using the MULTILOG procedure, odds and adjusted odds ratios for at risk of overweight and overweight were estimated in the same model with the normal BMI group (< 85th percentile) as the reference. Possible confounding variables were selected for inclusion based on both forward selection and backward deletion, with the decision to include or delete based on changes in the significance and β values of the exposure effects (duration of breastfeeding).²⁹ In parallel, most analyses were performed unweighted to confirm that the significance of the findings was not affected by the variance estimates used to account for the sample design. The unweighted results were not different and are not shown.

Sample sizes and characteristics of the sample are given in Table 1. In terms of infant feeding, 46% were never breastfed, nearly 25% were fully breastfed for 2 months or less, and then approximately 10% were fully breastfed for 3 to 5 months, for 6 to 8 months, and for 9 months or more. The timing of introduction of solid foods was consistent with the American Academy of Pediatrics recommendations for infant feeding¹ for 4 to 6 months for only about half the sample.

Using weighted estimates, 11% of the US-born children were classified as being at risk of overweight and 8.2% as overweight (Table 2). For females, but not for males, there was a trend toward increasing overweight from ages 3 to 5 years.

There were differences in full breastfeeding and its duration by key demographic, maternal, and perinatal characteristics that were likely to be associated with child BMI, and these were tested for inclusion in the multivariable models (Table 3). Non-Hispanic black infants were less likely to have been ever breastfed, while non-Hispanic white and infants of "other" race/ethnicity were more likely to have been breastfed longer. Only about 10% of non-Hispanic blacks continued full breastfeeding for 3 months or longer, compared with 37% of non-Hispanic whites.

There were no differences in the duration of full breastfeeding by infant sex. Moderately low birth weight infants (1500-2499 g) were less likely and high birth weight infants more likely to have been breastfed. Infants who were introduced to solids at the recommended age (4-6 months) were more likely to have been breastfed, but infants for whom the introduction of solids was delayed beyond 6 months were breastfed longer. Mother's concurrent BMI was strongly related to infant feeding. Underweight mothers breastfed for a shorter duration compared with normal weight mothers. Over 50% of overweight and obese mothers did not breastfeed at all (Table 3).

There was a reduction in the prevalence of being at risk of overweight if ever breastfed, although the reduction did not appear dose-dependent (Table 4). The reduction in risk was not significant in unadjusted analyses (odds ratio [OR], 0.66; 95% confidence interval [CI], 0.44-1.00), but there was a significant reduction in being at risk of overweight associated with having ever been breastfed (AOR, 0.63; 95% CI, 0.41-0.96), adjusting for birth weight status, race/ethnicity, sex, age group, mother's BMI status, and the timing of introduction of solid foods. However, there was no clear dose-dependent effect of duration of full breastfeeding (Table 4) nor was there a recognizable threshold effect of breastfeeding. The AOR for being at risk of overweight with full breastfeeding for 3 months or more was 0.67 (95% CI, 0.39-1.14), indistinguishable from the AOR of 0.66 (95% CI, 0.34-1.29) for being at risk of overweight with full breastfeeding for 6 months or more.

For overweight, there was no significant reduction in being overweight with being ever breastfed either in unadjusted (OR, 0.81; 95% CI, 0.61-1.09) or in adjusted analyses (AOR, 0.84; 95% CI, 0.62-1.13). As in the case of being at risk of overweight, for overweight status there was neither a clear dose-dependent effect with duration (Table 4) nor a threshold effect. The AOR of being overweight with full breastfeeding for 3 months or more was 0.70 (95% CI, 0.39-1.25), again indistinguishable from the AOR of 0.70 (95% CI, 0.37-1.32) for full breastfeeding for 6 months or more.

The lack of strong association between duration of full breastfeeding and child BMI was confirmed in a linear multiple regression model using duration of full breastfeeding (in months) to predict child BMI, adjusting for the race/ethnicity, sex, age group, birth weight status, mother's BMI, and timing (in months) of introduction of solid foods. There was a slight reduction in child BMI with duration of full breastfeeding (−0.02 ± 0.01 kg/m² per month full breastfeeding), but it was not statistically significant (P = .20).

However, there were factors that were significantly associated with child overweight in the multiple logistic regression models. Compared with non-Hispanic white children, Mexican-American children were at significantly increased risk of overweight (AOR, 1.76; 95% CI, 1.05-2.94). The timing of introduction of solid foods was associated with a 0.1% reduction in risk of overweight (AOR, 0.9994; 95% CI, 0.9990-0.9997) for each month that the introduction of solids was delayed (P<.001).

By far the strongest predictor of child BMI status was mother's concurrent BMI (Table 5). Children were at moderately increased risk for being at risk of overweight with an overweight mother (AOR, 1.54; 95% CI, 0.93-2.57) but nearly 3 times more likely to be at risk of overweight with maternal obesity (AOR, 2.97; 95% CI, 1.88-4.69). Furthermore, overweight in early childhood was nearly 3-fold greater (AOR, 2.95; 95% CI, 1.35-6.42) with maternal overweight and more than 4-fold greater (AOR, 4.34; 95% CI, 2.50-7.54) with maternal obesity.

We found a significant 37% reduction in being at risk of overweight (BMI 85th-94th percentile) among children ever breastfed compared with those who were never breastfed but only a 16% reduction in overweight (≥95th percentile). There was no strong dose-dependent effect of the duration of full breastfeeding on overweight status in early childhood, nor was there an apparent threshold effect, that is, a reduction in risk with full breastfeeding for 3 months or more or for 6 months or more. Children fully breastfed for 3 months or longer were at about a 30% decreased risk for overweight, but the CIs for the estimates were wide and included unity. Duration of full breastfeeding as an independent variable showed only a weak linear association with child BMI.

Our results are in contrast to the findings of von Kries et al,⁵ who reported a clear dose-dependent effect for duration of exclusive breastfeeding (defined as the child having been fed "no food" other than breast milk) on the prevalence of "obesity" (>97th percentile for BMI) in Bavarian children at ages 5 to 6 years. However, the current recommendation is to allow supplementation with solids by 4 to 6 months, while continuing full breastfeeding for 12 months or more. The findings of von Kries et al⁵ may be attributed to reverse causality, that is, the Bavarian children who were exclusively breastfed for 6 months or longer were smaller as infants, able to be nutritionally supported on breast milk alone longer, and were therefore less likely to be at risk of overweight or to be overweight in early childhood.

Our findings are less likely to be attributable to such reverse causality, and the study has additional strengths. The nationally representative sample of US-born young children is ethnically diverse, and the findings for breastfeeding are consistent both with prevalence data and with demographic associations reported in other national studies.^7-9 Our conclusions therefore are less likely due to chance or sample bias, although neither explanation can be totally ruled out. Having excluded infants born weighing less than 1500 g or with extended periods (≥ 2 weeks) of special neonatal care, we were able to adjust for the effects of both moderately low (1500-2499 g) and high birth weight, since both have been shown to be associated with early childhood growth in NHANES III.^10,13-15 By not limiting our definition of breastfeeding to those exclusively breastfed and allowing the introduction of solids at the recommended ages, we were better able to determine if there were inherent overweight-preventing properties of breastfeeding, as compared with lesser weight gain from a lower energy intake.

Basing the duration of breastfeeding on mother's recall is a limitation of this study. However, studies of the reliability of recalling infant feeding mode and duration have shown that, if anything, more highly educated mothers tend to overestimate the duration of full breastfeeding.^2,30 This would have the effect of exaggerating the beneficial effects of breastfeeding on child overweight since children born to more highly educated parents tend to be healthier, better nourished, and leaner. It is unlikely, therefore, that recall bias is masking an association between breastfeeding and early childhood overweight.

We did find that factors other than infant feeding were associated with child overweight, implying that the association between breastfeeding and being at risk of overweight may be confounded by unmeasured sociodemographic or intervening familial factors. Compared with non-Hispanic white children, young Mexican-American children were at significantly increased risk of overweight, a finding that has been reported previously for NHANES III.^11,15 The timing of the introduction of solid foods was associated significantly with a 0.1% reduction in risk of overweight for each month that introduction of solids was delayed. For half of the sample, timing of introduction of solid foods was consistent with American Academy of Pediatrics recommendations of 4 to 6 months.¹ A further delay in introduction of solid foods in infancy would appear to provide a statistically significant but relatively minor reduction in risk for overweight at ages 3 to 5 years. Mehta et al²⁰ have reported that early (3-4 months), as compared with late (≥6 months), introduction of solid foods did not affect infant growth or body composition in the first year of life and concluded that early introduction of solid foods in infancy simply displaced energy intake from liquid sources (breast milk or formula), rather than supplying additional energy that would contribute to overweight.

By far the strongest predictor of BMI status in young children was mother's concurrent BMI. Risk of overweight among young children was nearly tripled with maternal overweight and more than quadrupled with mothers' obesity. The overwhelming significance of the association between parental obesity and overweight status in children younger than 5 years has also been demonstrated.³¹ It was shown that risk among children ages 3 to 5 years of subsequently becoming obese as adults was nearly tripled for those with obese mothers compared with risk among children whose mothers were not obese, although overweight in early childhood is not by itself a strong predictor of adult obesity. The extent to which familial tendencies to obesity are genetic or associated with shared dietary and activity habits has yet to be determined, but it is unlikely that breastfeeding alone would be sufficient to prevent the development of overweight in such a situation.

Our findings therefore agree with previous studies that have failed to find an association between mode of infant feeding and body composition in early childhood^10,21 or overweight, but have implicated maternal or parental obesity.^32-35 For example, Baranowski et al³² reported for 246 children from 3 ethnic groups that no measure of infant feeding (breastfeeding or duration) was associated with adiposity at 3 to 4 years of age. Likewise, Zive et al³³ found no relationship between duration of breastfeeding and adiposity, as measured by BMI, for 331 children at 4 years of age. However, as we found in this study, Zive et al³³ found that mother's BMI explained the largest portion of the variance in child fatness.

Our study is best compared with that of O'Callaghan et al³⁴ who reported in a large study of 4062 children at age 5 years that birth weight, paternal BMI, and sleeplessness at 6 months predicted risk of being at risk of overweight (BMI 85th-94th percentile), while birth weight, being female, and maternal and paternal BMI predicted increased risk of overweight (≥95th percentile). On the other hand, whereas birth weight and parental BMI were strongly associated with risk of overweight, breastfeeding and its duration were not.³⁴ We also found that a high birth weight (≥4000 g) was associated with an increased risk of having a high BMI in early childhood (>85th percentile), but the effect was diminished when maternal BMI was considered.

Between infancy and ages 3 to 5 years, patterns of food consumption and physical activity probably have more impact on weight status in early childhood than infant feeding. Because the NHANES III is a cross-sectional survey, no data are available to describe relevant behaviors during this intervening period or to characterize the timing or relative weight level at which the adiposity rebound may be occurring. Children who reach the nadir of their BMI at ages younger than 6 years before it once again increases tend to be at increased risk for overweight.^36,37 A limitation of cross-sectional survey data is that the contribution of these time-dependent patterns to the development of overweight cannot be evaluated.

Overweight in early childhood is a growing problem in the United States. Recent findings from NHANES III (1988-1994), compared with earlier national surveys,^11,15 demonstrated that children as young as 4 or 5 years, and especially females and Mexican-Americans, show an increased prevalence of overweight. Mei et al³⁸ have documented an increasing trend in overweight among low-income children aged 5 years or younger who were included in the CDC Pediatric Nutrition Surveillance System from 1983 to 1995. Consistent with our findings, Mei et al³⁸ also observed that overweight prevalence was higher for young females than for males, a trend that emerges as early as the preschool years and persists into adulthood. In this context, the suggestion that breastfeeding may be protective against the development of overweight in children^5,6 has raised expectations that promoting breastfeeding might help reverse this trend.

The findings from our analyses of US national survey data do not support the contention that breastfeeding and its duration are necessarily protective against early childhood overweight, although there is some conferred protection from being at risk of overweight. However, maternal obesity (suggesting shared familial dietary habits and activity patterns) far supersedes infant feeding as a risk factor for being at risk of overweight or overweight in early childhood, and the dose-dependent effect for duration of breastfeeding and risk for overweight is inconsistent compared with that for maternal obesity. The finding that overweight and obese mothers were less likely to have breastfed at all may suggest a synergistic effect of maternal overweight status, combined with the absence of breastfeeding, that results in increased risk for overweight in the children of these mothers.

It cannot be too strongly emphasized that breastfeeding has numerous attributes that render it the preferred feeding choice for almost all infants. However, duration of full breastfeeding does not appear to be predictive of or necessarily have preventive properties for overweight in early childhood, and encouraging breastfeeding for overweight prevention would not be as effective as moderating familial factors in preventing early childhood overweight.