点击显示 收起
1 From the Division of Gastroenterology and Nutrition, Children's Hospital of Philadelphia, Philadelphia, PA (HLB and WCH); the Department of Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, PA (HLB); Mathematica Policy Research, Inc, Princeton, NJ (RCW); the Division of Cardiology, Cincinnati Children's Hospital Medical Center, Cincinnati, OH (SRD); and the Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, OH (SRD)
2 Supported by grant no. R01HL/DK64022 from the National Institutes of Health. 3 Address reprint requests and correspondence to HL Burdette, Division of Gastroenterology and Nutrition, Children's Hospital of Philadelphia, 34th Street and Civic Center Boulevard, Philadelphia, PA 19104. E-mail: burdette{at}email.chop.edu.
ABSTRACT
Background: Although dual-energy X-ray absorptiometry (DXA) is considered the most accurate measure of adiposity in children, it has rarely been used to examine the relation between infant feeding and adiposity during childhood.
Objective: The objective was to ascertain whether adiposity at age 5 y was related to breastfeeding, to the timing of the introduction of complementary foods during infancy, or to both.
Design: Body composition was measured in 313 children at age 5 y by using DXA. Data on breastfeeding, formula feeding, and the timing of the introduction of complementary foods were obtained from the mothers when the children were 3 y old. Regression analysis was used to examine the relation between infant feeding and fat mass after adjustment for lean body mass, sex, birth weight, maternal obesity, race, and other sociodemographic variables.
Results: Fifty-three percent of the children were boys, 80% were white, and 20% were black. There was no significant difference in adjusted fat mass between those ever breastfed and those never breastfed (
Key Words: Breastfeeding • complementary feeding • children • preschool-age children • obesity • dual-energy X-ray absorptiometry
INTRODUCTION
With the rising prevalence of childhood obesity, interest has increased in ascertaining whether breastfeeding or the delayed introduction of complementary foods (or both) can reduce the risk of later obesity. However, the many studies on this subject have conflicting results (1–8). Debate about the lack of agreement among these studies has focused primarily on the methods of measuring infant feeding and on the appropriate adjustment for confounding factors in nonexperimental study designs (4, 9). Less emphasis has been placed on the selection of the measure of adiposity or obesity.
Body mass index (BMI; in kg/m2) provides information about body weight, but it is only a surrogate measure of adiposity. Differences in body weight are only partly due to differences in body fatness. For example, black children, at any given level of BMI, have a significantly lower percentage of fat mass (FM) and a significantly higher percentage of lean mass than do white children (10, 11). It is adiposity, rather than weight, that is thought to explain the major health comorbidities associated with obesity. In contrast to BMI, dual-energy X-ray absorptiometry (DXA) is a technique for directly measuring adiposity in children (12, 13). Despite the superiority of DXA over BMI as a measure of adiposity, to our knowledge, only one study has used DXA to examine the association between breastfeeding and adiposity after age 2 y (14). We know of no studies that used DXA to examine the association between complementary feeding practices in infancy and adiposity beyond infancy.
Breastfeeding and the timing of the introduction of complementary foods were examined as independent determinants of later obesity. However, in practice, breastfeeding and complementary feeding are not independent decisions, and they may jointly explain variances in later obesity (15). It has been suggested that children who are exclusively or predominantly breastfed (ie, no formula or energy-containing complementary foods) for the first 4 mo of life have a different pattern of growth than do children who receive energy from nonbreastmilk sources in the first 4 mo of life (16, 17). Thus, in addition to examining the independent effects of breastfeeding and complementary feeding, it is important to examine their combined association with later adiposity.
The main purpose of the current study was to ascertain whether adiposity at age 5 y, as assessed by DXA, was related to breastfeeding, to the timing of the introduction of complementary foods during infancy, or to both. The secondary purpose was to examine these same relations by using BMI as a surrogate measure of adiposity.
SUBJECTS AND METHODS
Setting and participants
This analysis involved data collected on 313 preschool-aged children who were participating in a prospective cohort study. The principal aim of this study was to describe normal changes in body fatness during early childhood. At the age of 3 y, 372 children were enrolled in this study. To be eligible, children had to be born after full-term gestation (37 wk), to be free of chronic health problems affecting growth and development, and to have parents who were either both black or both white.
Written informed consent was obtained from all of the mothers. The Institutional Review Board at Cincinnati Children's Hospital Medical Center approved the study.
Study measures
Breastfeeding and formula feeding
As part of the first study visit, which occurred when the children had a mean (± SD) age of 3.3 ± 0.3 y, mothers reported information about feeding practices during the child's first year of life. Each mother was asked if she breastfed her child and, if so, what age the child was (<1 mo, 1–2 mo, 3–5 mo, 6 mo–1 y, or >1 y) when breastfeeding stopped.
The breastfeeding mothers were also asked if they regularly gave their children formula in addition to breastmilk and, if so, what age the child was when the mothers began that practice (0–3 mo, 4–6 mo, or 7–12 mo). Using the formula and breastfeeding data, we defined the following mutually exclusive breastfeeding and formula-feeding groups: never breastfed, breastfed but also fed formula before 4 mo, breastfed 4 mo and fed formula after 4 mo, and breastfed 12 mo and never fed formula.
Introduction of complementary foods
For the purposes of this study, complementary foods included foods fed to the infant other than breastmilk, formula, or water. Mothers were asked when their children started eating infant cereal (from a bottle or spoon), baby food, and mashed or regular table food. Mothers were also asked when the children started drinking fruit juice and other sweetened beverages, such as Kool-Aid (Kraft Foods, Inc, Northfield, IL), Gatorade (PepsiCo, Inc, Chicago, IL), or carbonated beverages. Using these data, we placed children into 1 of 2 categories according to whether or not their mothers followed the American Academy of Pediatrics (AAP) recommendations for introduction of complementary foods and juice (18, 19). Following the AAP recommendation was defined as no introduction of solid foods (including infant cereal, baby food, or table food) before age 4 mo (18) and no introduction of sweetened beverages before age 6 mo (19).
Combination of breastfeeding, formula feeding, and use of complementary foods
In addition to analyzing separately the data on breastfeeding, formula feeding, and the timing of the introduction of complementary foods, we constructed feeding categories that combined these data. In accordance with prior analysis using such data (17), we divided the children into 3 mutually exclusive groups: never breastfed, partially breastfed, and exclusively breastfed. The exclusively breastfed children were breastfed and received no formula, complementary foods, or juice during the first 4 mo of life. Partially breastfed children were breastfed but received formula, complementary foods, or juice before age 4 mo.
Dual-energy X-ray absorptiometry
The primary outcome measure for our analyses was adiposity, which was assessed with the use of DXA when the children were an average of 5.0 y old (range: 4.8–5.2 y). The child's entire body was scanned in the array-fan beam mode with a Hologic 4500 instrument (Hologic, Bedford, MA) and HOLOGIC PEDIATRIC software (version 12.3; Hologic). For the DXA scans, the child was lightly dressed in a tee-shirt and shorts or pants with an elastic waist. No sedation was used. Body-composition measures derived from DXA included lean body mass (LBM), FM, and percentage body fat [(%BF) total FM divided by total body mass]. DXA data at age 5 y were available for 313 children (84% of the original cohort).
Anthropometric measurements
Height and weight were measured at the initial visit and at the time of the DXA scan. Height was measured by using a wall-mounted stadiometer (602VR; Holtain, Crymych, United Kingdom), and weight was measured with a digital scale (model 770; Seca, Hamburg, Germany) while the children wore light clothing such as that worn for the DXA scan. Height and weight were obtained twice, and the average value for each measure was calculated. These height and weight values were then used to calculate BMI (kg/m2), and the BMI percentile and z score for each BMI value were established by using the 2000 Centers for Disease Control and Prevention growth charts (20).
Covariates
All information on the covariates was obtained from the mother at the first study visit. Child covariates included race (white or black), sex, birth weight (< 4 kg or 4 kg), and participation in the Special Supplemental Nutrition Program for Women, Infants, and Children as an infant. Other sociodemographic variables included maternal age, marital status, education, and household income. Each mother was asked to report on health indicators that might influence either her feeding practices or her child's adiposity. These factors included the mother's current height, weight, and smoking status and whether she had diabetes (gestational or preexisting) during pregnancy. Maternal BMI was calculated from self-reported height and weight (or prepregnancy weight if the mother was pregnant at the first study visit). Maternal obesity was defined as a BMI 30. Data on BMI were missing for 2 mothers.
Statistical analysis
FM, adjusted for LBM and sex, was used as the primary dependent variable. These adjustments of FM were made to account for normal variation in fat mass due to body size and sex (21, 22), and they allowed us to pool data across sexes. Adjusted FM, rather than %BF, was selected as the primary measure of adiposity in the analyses involving a continuous dependent variable. This decision was made for 2 reasons. First, of all measures of adiposity, adjusted FM is the one most highly correlated with cardiovascular disease risk factors (23). Second, the use of a ratio (or percentage) may introduce bias in the results if the regression relating the ratio denominator (total body mass) and the ratio numerator (total FM) has a non-zero intercept, as was the case for our data (24). Regression-adjusted FM has been used by others to assess the relation between maternal feeding practices and child adiposity (25).
To facilitate clinical interpretation of our results from the DXA scans, we also wished to conduct our analysis by using a binary dependent variable that classified children as having either "high" or "normal" adiposity. We examined this outcome because breastfeeding may reduce the proportion of children at both ends of the adiposity range without affecting the mean (26). However, there are no well-established reference data from which to identify a cutoff for overweight or obesity in children on the basis of body-composition data from DXA scans. We classified children as having "high adiposity" if they had a %BF in the sex-specific highest quartile for the cohort. In our cohort, this was >29% for females and >24% for males. Prior studies suggested similar cutoffs for %BF by comparing the association of %BF with cardiovascular disease risk factors (27, 28). Despite the aforementioned limitation of using %BF, we based our binary outcome variable on that measure because it is the measure that has been most widely used for this purpose (29, 30).
In all our analyses, the feeding variables were handled as categorical measures, and comparisons were made to a referent group (either never breastfed or did not follow the AAP recommendations on complementary feeding). Chi-square tests were used to evaluate the bivariate relations between the exposures (breastfeeding and introduction of complementary foods) and the covariates. Multivariate linear regression was used to calculate the mean FM, adjusted for covariates, at each level of the given feeding variable. This allowed us to compare the difference (and 95% CI of the difference) between the mean FM of children in the referent group and the mean FM of children in another level of the feeding variable (for example, comparing adjusted mean FM for those never breastfed with that for those breastfed 6 mo–1 y). In the first regression model, we adjusted FM only for LBM and sex. In the second model, we further adjusted for other covariates.
Similarly, we used multivariate regression to evaluate the covariate-adjusted relation between breastfeeding (and introduction of complementary foods) and other secondary dependent variables, such as the percentage of children with high adiposity, BMI z score, and percentage of children who were overweight (BMI 85th percentile). All analyses were conducted by using SPSS for WINDOWS software (version 12.0; SPSS Inc, Chicago, IL).
RESULTS
Description of sample
The sex and racial composition of the study sample (n = 313) was 142 white males, 24 black males, 107 white females, and 40 black females. At the time of the children's birth, the mean age of the mothers was 30 ± 5.0 y. Half of the mothers had completed college, and almost 90% were married (Table 1). The mean maternal BMI at the time of the first study visit was 26.7 ± 6.6.
View this table:
TABLE 1. Characteristics of the study sample1
The 313 children who had DXA scans at age 5 y were significantly more likely to have been ever breastfed than were the 59 children in the original cohort who did not have a DXA scan at age 5 y (74% and 56%, respectively; P = 0.007). However, at age 3 y, there was no significant difference between the 2 groups in BMI z score (0.34 and 0.37, respectively; P = 0.86) or the percentage with BMI 85th percentile (22% and 25%, respectively; P = 0.67) or 95th percentile (6.7% and 7.1%, respectively; P = 0.91).
Feeding practices
Half of the mothers (n = 157) followed the AAP recommendations for introducing complementary foods, and only 8% (n = 26) of the mothers did not start feeding their infants complementary foods until age 6 mo (31). One-quarter of the mothers never breastfed (Table 2). Of the 231 mothers who did breastfeed, 88 breastfed for < 6 mo, 77 did so for 6 mo–1 y, and 66 did so for >1 y. Fifty-two women (17%) breastfed for 1 y without using formula, and, of these, 16 did not start complementary foods until the infant was 6 mo old. Compared with women who never breastfed, those who ever breastfed—and particularly those who did so for a longer time—were significantly more likely to follow the AAP recommendations for the introduction of complementary foods (Table 2). Ninety-five children (30%) were exclusively breastfed in the first 4 mo of life insofar as they received no formula, complementary foods, or sweetened beverages before 4 mo of age.
View this table:
TABLE 2. Prevalence of breastfeeding and the relation of breastfeeding status to the introduction of complementary foods according to the recommendations of the American Academy of Pediatrics (AAP)1
Body-composition and anthropometric measurements
The mean FM of the children at age 5 y was 4.55 ± 1.64 kg, and the mean LBM and %BF were 14.05 ± 1.93 kg and 23.12 ± 5.43%, respectively (Table 3). Girls had significantly greater FM and %BF than did boys, and boys had greater LBM than did girls. After adjustment for LBM, the FM remained higher for girls than for boys (5.18 ± 0.12 and 3.99 ± 0.11 kg, respectively; P < 0.001). Because of this sex difference, mean FM is hereafter reported after adjustment for both LBM and sex.
View this table:
TABLE 3. Descriptive statistics for dual-energy X-ray absorptiometry (DXA) and anthropometric measurements at age 5 y
Body weight did not differ significantly between black and white children (Table 3). After adjustment for LBM and sex, black children had significantly lower FM than did white children (4.19 ± 0.18 and 4.64 ± 0.09 kg, respectively; P = 0.03).
The proportion of children with BMI 85th and 95th percentiles was 26% and 10%, respectively. The proportion of children with BMI 85th percentile was significantly higher in blacks than in whites (P = 0.03), but there was no significant difference in mean BMI z scores between the races. There was no significant sex x race interaction for any of the DXA or anthropometric measures in Table 3. FM and BMI z score were significantly correlated (r = 0.74, P < 0.001), as were %BF and BMI z score (r = 0.55, P < 0.001).
Relation of breastfeeding and introduction of complementary foods to adiposity
Mothers who were white, more educated, married, older, and nonsmokers and who had higher household income and did not enroll their children in the Special Supplemental Nutrition Program for Women, Infants, and Children were significantly more likely to breastfeed and to follow the AAP recommendations for the introduction of complementary foods than were mothers without those characteristics (Table 4). However, none of these covariates except race was significantly related to the FM of the child. Obese mothers were significantly less likely than were nonobese mothers to breastfeed, and the children of the obese mothers had significantly higher FM than did the children of nonobese mothers. The interaction between maternal obesity and infant feeding method was not significant.
View this table:
TABLE 4. Relation between characteristics of the child and the mother and breastfeeding, introduction of complementary foods, and fat mass1
The difference in FM (after adjustment for LBM and sex only) between children who were never breastfed and those who were ever breastfed was not significant. Compared with children who were never breastfed, those who were ever breastfed had an FM that was 0.33 kg lower (95% CI: 0.68 kg lower, 0.03 kg higher) at 5 y of age (Table 5).
View this table:
TABLE 5. Mean fat mass and percentage of subjects with high adiposity (highest quartile of percentage body fat) by breastfeeding status and introduction of complementary foods
Regardless of the duration of breastfeeding, breastfed children did not have significantly less FM than did children who were never breastfed. When FM was compared between groups of children who had different combinations of breastfeeding and formula feeding and children who were never breastfed, the differences were not significant. FM did not differ significantly between children whose mothers followed the AAP complementary feeding recommendations and children whose mothers did not. When we repeated these analyses using %BF as a continuous dependent variable, we had results similar to those obtained when we used FM as the dependent variable (data not shown). The results were also similar when percentage "high adiposity" was used as a dichotomous dependent variable (Table 5). There was no significant difference in either FM or the proportion with high adiposity between those never breastfed and those exclusively breastfed for the first 4 mo.
In unadjusted analysis, the BMI z score tended to be lower in children who were breastfed (Table 6). However, after adjustment for multiple covariates, the BMI z score for breastfed children did not differ significantly from that in children who were never breastfed. In unadjusted analysis, the percentage of children who were overweight (BMI 85th percentile) was significantly lower in the children ever breastfed than in those never breastfed. When obesity (BMI 95th percentile) was used as the dependent variable, the significant relation also disappeared after adjustment for covariates (unadjusted never breastfed versus ever breastfed: 18% versus 7%; P = 0.005; adjusted never breastfed versus ever breastfed: 14% versus 9%; P = 0.20). There was no significant difference in either overweight or BMI z score between those never breastfed and those exclusively breastfed in the first 4 mo. The results shown in Tables 5 and 6 were unchanged when maternal BMI, and not maternal obesity, was used as a covariate.
View this table:
TABLE 6. Mean BMI z score and prevalence of overweight (BMI 85th percentile) by breastfeeding status and introduction of complementary foods
DISCUSSION
We found no statistically significant relation between adiposity at age 5 y, as measured by DXA, and either breastfeeding or the timing of the introduction of complementary foods during infancy. Even children who were breastfed for > 1 y and who were never fed formula had a total FM that did not differ significantly from that of children who were never breastfed. When BMI was used as a surrogate measure of adiposity, an apparent protective effect of breastfeeding on later adiposity was no longer present after adjustment for covariates.
Our findings are consistent with 2 previous studies that examined the relation between breastfeeding and direct measures of adiposity by using DXA. In a prospective cohort study, Butte et al (32) found no significant difference in body composition at age 24 mo between infants who were primarily breastfed and those who were formula fed. Tulldahl et al (14) measured body fat in Swedish adolescents and found that the group that was exclusively breastfed for >3 mo did not have significantly lower %BF than did a combined group of those who were not breastfed at all or who were exclusively breastfed for <3 mo. Unlike many smaller studies (< 500 subjects) of the relation between breastfeeding and later BMI, such as some included in the meta-analysis of Owen et al (4), the current study found no significant association between breastfeeding and either BMI or overweight. This difference in findings may be best explained by the fact that we adjusted our analysis for potentially confounding covariates such as maternal obesity, socioeconomic status, and maternal smoking.
To our knowledge, only one randomized trial has examined whether the timing of the introduction of solid foods affects later adiposity. In that study, there was no difference at age 1 y in fatness, as assessed by DXA, between those who began solids early, at age 3–4 mo, or later, at age 6 mo (33). However, that study was restricted to formula-fed infants, and the effect of the timing of the introduction of solid foods on later adiposity may differ between breastfed and formula-fed infants. None of the observational studies examining the relation between the timing of the introduction of solid foods and childhood adiposity have used DXA (5–8). One of these observational studies found significantly greater fatness in children who were introduced to solid foods early than in those who were introduced to solid foods later (5). Two of the other studies, as did the current study, found no difference (7, 8), and one found that delayed introduction of solids was associated with greater fatness in early childhood (6).
Our sample size was small in comparison to many studies that used BMI to examine the relation between breastfeeding and obesity (4), so we may have lacked the power to detect a statistically significant difference in adiposity by using DXA. However, even if our estimate of the difference in adiposity between the ever-breastfed and the never-breastfed children was statistically significant, it is not clear that this difference (0.17 SD units) was clinically significant. When we used a binary outcome measure based on DXA ("high" versus "normal" adiposity), there was no difference in the percentage with "high" adiposity between those never breastfed and those who were breastfed >1 y without formula.
The application of our findings to other populations may be limited. The mothers in our sample had relatively high education levels, and none of our subjects were Hispanic. Furthermore, because we had a low percentage of blacks in our sample, we were unable to examine whether the relation of breastfeeding to fatness differed between blacks and whites, as was shown in earlier studies (26, 34). The fact that the information on infant feeding was collected when the children were 3 y old could have led to some bias if some mothers of heavier 3-y-olds overreported the duration of breastfeeding.
Large sample sizes are required to detect the relatively small effects that breastfeeding may have on later adiposity and to also adequately control for confounding. Investigators have appropriately used BMI as the measure of adiposity in large epidemiologic studies, because height and weight measures are less costly and easier to obtain than is DXA. However, because BMI may not be the most valid measure of adiposity, there is some inherent trade-off between study methods. The current study was not large, but we used DXA as a measure of adiposity and also controlled for confounding factors.
It is possible that some of the protective effect of breastfeeding on later obesity is overestimated by using BMI as an outcome measure. In our study, for example, there were suggestive differences in the effect of breastfeeding on adiposity, depending on the measure of adiposity used. In our unadjusted analysis, BMI levels and the prevalence of overweight (BMI 85th percentile) were both significantly lower in breastfed children than in formula-fed children, whereas FM and "high adiposity" (highest quartile of %BF) were not. Nonetheless, breastfeeding had no significant association with either BMI level or the prevalence of overweight after control for confounding factors.
To the extent that the association between breastfeeding and later adiposity may differ when different measures of adiposity are used, researchers and clinicians should be cautious about concluding that breastfeeding protects against later obesity solely on the basis of studies using BMI. Families should be given balanced information about the likelihood that breastfeeding will protect their children from obesity. Our study and the 2 others examining breastfeeding and later adiposity by using DXA (14, 32) show no protective effect of breastfeeding on later adiposity. This in no way, however, diminishes the importance of recommending breastfeeding for its multiple other benefits to mother and child.
ACKNOWLEDGMENTS
We extend our gratitude to Karen Munson for her role in managing the data collection and to Nicolas Stettler for his suggestions on an early draft of the manuscript.
All authors were involved in the design of the study. HLB had the primary responsibility for conducting the analysis and writing the manuscript. RCW helped analyze and interpret the data, helped with drafting and reviewing the manuscript, and provided supervision. WCH was involved in interpreting the data and helped with writing the manuscript. SRD was involved in acquisition and interpretation of the data and in critical review of the manuscript. RCW and SRD obtained funding for the study. None of the authors had any personal or financial conflict of interest.
REFERENCES