Copyright 2006 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2006
To examine the relationship between maternal infant-feeding style and adiposity in childhood and to determine whether feeding style explains any of the association between maternal obesity and childhood adiposity.
Prospective cohort study.
Cincinnati metropolitan area.
A total of 313 preschool children; 80% were white and 20% were black.
Main Outcome Measures
Seven factors describing maternal infant-feeding style derived from the Infant Feeding Questionnaire administered at age 3 years; maternal obesity, defined as a body mass index of 30 or higher before pregnancy; and adiposity at 5 years of age as assessed by dual-energy x-ray absorptiometry.
The mean ± SD fat mass was 4.55 ± 1.64 kg. Seventeen percent of the mothers were obese before pregnancy. Children whose mothers had high concern about the infant overeating or becoming overweight (the highest tertile of the “overeating” factor) had 0.67 kg (95% confidence interval, 0.31-1.03 kg) higher fat mass than children whose mothers did not have high concern (the other 2 tertiles). None of the other 6 feeding factors were related to childhood adiposity. Children of obese mothers had 0.54 kg (95% confidence interval, 0.10-0.98 kg) higher fat mass than children of nonobese mothers. High concern about the infant overeating, which was more common in obese mothers, accounted for 15% of this 0.54-kg difference.
High maternal concern about an infant overeating or becoming overweight was associated with higher adiposity at 5 years of age and explained some of the association between maternal obesity and child adiposity.
Amidst the current epidemic of childhood obesity, more consideration is being given to starting obesity prevention efforts early in life, even during gestation or at birth.1 One potential target for these prevention efforts is infant feeding. Although there has been much research on the association between later obesity and what infants eat, such as breast milk, infant formula, and other complementary foods,2,3 there has been far less research on contextual factors in infant feeding. Many of these contextual factors are related to the parental feeding style, particularly the feeding style of the mother, who is often the primary infant feeder.
The term maternal infant-feeding style is used herein to describe a mother's practices and beliefs about feeding her infant, such as whether the mother controls feeding with a regular schedule, uses food to soothe her infant, is aware of her infant's hunger cues, or worries about her infant eating too much or too little. To the extent that the mother's infant-feeding style has an enduring impact on the child's developing system of appetite regulation and is associated with the child's later adiposity, this maternal feeding style, and not just what the infant eats, may be an appropriate focus in obesity prevention.
Although maternal obesity has been identified as the strongest risk factor for the early development of obesity in children,4 this risk factor can operate through many potential mechanisms.5 One of the modifiable mechanisms may be the mother's style of feeding her infant. If obese and nonobese mothers have different feeding styles and these styles are also associated with childhood obesity, it is possible that altering these feeding styles might be a way to help obese parents prevent obesity in their children.
Some maternal feeding practices, such as pushing children to eat or restricting children from certain snack foods, are thought to be related to the development of later obesity, but those studies have focused on feeding preschool children and not on feeding infants.6,7 To our knowledge, there are only 2 validated instruments that ask mothers about aspects of their infant feeding style hypothesized to affect later adiposity.8,9 The developers of these instruments examined the association of maternal infant-feeding style with children's later height and weight, but there were no direct measures of children's adiposity in these studies, and all anthropometric measurements were made at younger than 24 months.9,10 To our knowledge, there have been no other studies relating maternal infant-feeding style to children's later weight or adiposity.
In this study, we attempted to address 2 research questions. (1) Are there any aspects of the mother's feeding style during infancy that are associated with the child's adiposity at 5 years of age? (2) How much, if any, of the association between maternal obesity before pregnancy and childhood adiposity is explained by maternal infant-feeding style?
This study involves data collected on preschool children participating in a prospective cohort study describing normal changes in body fatness during early childhood. In 2001 and 2002 we recruited 372 children aged 3 years for this study from the Cincinnati metropolitan area by placing brochures in pediatric offices and day care centers and by placing advertisements in community newspapers. All the children met the following eligibility criteria: full-term gestation (≥37 weeks), no chronic health conditions affecting growth and development, and parents who were either both black or both white. Parental race was reported by the child's mother. The institutional review board at Cincinnati Children's Hospital Medical Center approved the study. Informed written consent was obtained from the parents.
Dual-energy x-ray absorptiometry (DXA) is considered the most accurate measure of adiposity in children,11 and it was used to measure children's body composition at a mean age of 5.0 years (range, 4.8-5.2 years). The entire body was scanned in the array-fan beam mode using a Hologic 4500 instrument (Hologic Inc, Bedford, Mass) and pediatric software (version 12.3, Hologic Inc). Body composition measures derived from DXA included lean body mass, total fat mass, and percentage of body fat (total fat mass divided by total body mass). The DXA measurements at 5 years of age were available for 313 children (84% of the original cohort).
Height and weight were measured at the initial study visit and at the time of DXA. Heights were measured using a wall-mounted stadiometer, and weights were measured using a digital scale. Height and weight were measured twice, and, for each measure, the average value was calculated. These height and weight values were then used to calculate body mass index (BMI) (calculated as weight in kilograms divided by the square of height in meters), and the BMI percentile and z score for each BMI value were established using the 2000 Centers for Disease Control and Prevention growth reference.12 For the DXA and anthropometric measurements, the child was lightly dressed in a T-shirt and in shorts or pants with an elastic waist.
At the time of the initial study visit (mean ± SD child age, 3.3 ± 0.3 years), each mother was asked to complete the Infant Feeding Questionnaire (IFQ),9 a self-administered instrument containing 20 questions about different aspects of maternal infant-feeding style during the child's first year of life that are hypothesized to be related to the risk of childhood obesity. The IFQ has 12 items about the context of a mother's feeding practices, such as how, when, and why she fed her infant, and 8 questions about a mother's beliefs concerning infant feeding (Table 1). Each item has 5 response options ranging from “never” to “always” for questions about practices and from “disagree a lot” to “agree a lot” for questions about beliefs.
The IFQ measures the following 7 factors: (1) concern about infant undereating or becoming underweight (undereating), (2) concern about infant's hunger (hunger), (3) awareness of infant's hunger and satiety cues (infant cues), (4) concern about infant overeating or becoming overweight (overeating), (5) feeding infant on schedule (schedule), (6) using food to calm infant's fussiness (food to calm), and (7) social interaction with the infant during feeding (social) (Table 1). The 5 response options for each question were assigned the values 0 (never/disagree a lot) through 4 (always/agree a lot). Answers were averaged across the items for each factor to give a possible score for each factor ranging from 0 to 4. This is the same scoring method used in a previous study.9
The mother's BMI was calculated from maternal report of height and weight before pregnancy (data collected at the first study visit). Maternal obesity was defined as having a BMI of 30 or higher.
At the first study visit, the mothers provided information about the children's sex, birth weight, and breastfeeding history. The duration of breastfeeding was categorized as follows: never and less than 3, 3 to 5, 6 to 12, and more than 12 months. Mothers also reported information to allow construction of the following maternal variables: age (years), race (white vs black), marital status (single vs married), education (high school or less vs some college vs college or more), current smoking status (smoker vs nonsmoker), household income (<$50 000 vs $50 000-$74 999 vs ≥$75 000), and whether the mother enrolled the child in WIC (Special Supplemental Nutrition Program for Women, Infants, and Children) during infancy (yes vs no).
We created continuous and binary outcome measures from the DXA data and from the BMI data (4 total outcome variables). From DXA, the continuous outcome measure was fat mass, which was adjusted for lean body mass and sex, as has been previously described.13,14 These adjustments in fat mass were made to account for normal variation in fat mass related to body size and sex, and they allowed us to pool data across the sexes. Fat mass was selected a priori as the primary outcome variable because it is a direct measure of adiposity and, as a continuous measure, was likely to be more sensitive than a binary measure to any differences in maternal infant-feeding style.
To facilitate clinical interpretation of the DXA results, we also classified children as having either “high” or “normal” adiposity, creating a secondary (binary) outcome measure. Because there are no established reference data to determine a cutoff point for overweight/obesity in children based on body composition data from DXA, children were categorized as having high adiposity if they had a percentage of body fat in the sex-specific upper quartile for the cohort. In this cohort, this value was greater than 29% for females and greater than 24% for males, cutoff points that are consistent with previous studies15,16 investigating cutoff points for obesity in children based on the relationship between percentage of body fat and cardiovascular risk factors.
Because most other studies of childhood obesity risk factors are based on BMI, we also analyzed 2 other secondary outcome measures—BMI z score (continuous) and BMI at the 85th percentile or higher (binary). These 2 measures are analogous to our DXA measures of fat mass and high adiposity, respectively.
The 7 feeding factor scores from the IFQ were not normally distributed (the median value was close to either 0 or 4) (Table 1), and not all the scores ranged from 0 to 4, so we converted each factor into a binary variable using the following method. For each factor, we divided the sample into 3 groups of approximately equal size, breaking the group as close as possible to the tertile cutoff points for the factor score. Given the skew of the data, 1 tertile contained the values in the “outlier” tail of the distribution. We further divided the children into 2 groups—a group with values in this outlier tertile and a group with values in the other 2 tertiles (Table 1). In all the analyses using the 7 IFQ factors, we made comparisons between these 2 groups. For 6 of the 7 factors, the children in the outlier tertile were those we hypothesized would have increased adiposity at 5 years of age. We hypothesized that the outlier tertile of the factor “high concern about the child undereating or being underweight” would be associated with lower fat mass.
Multivariate linear regression was used to compare the mean fat mass, adjusted for lean body mass and sex, between the 2 levels of each feeding factor (Table 1) and between obese and nonobese mothers. χ2 Tests and t tests were used to evaluate the bivariate relationship between the 3 other outcome measures (high adiposity, BMI z score, and overweight) and (1) the 7 feeding factors and (2) maternal obesity.
After examining these bivariate relationships among child adiposity, maternal infant-feeding style (7 feeding factors), and maternal obesity, we used regression models to evaluate the extent to which the relationship between maternal obesity before pregnancy and child adiposity at age 5 years was mediated by maternal infant-feeding style. Using fat mass as the dependent variable, linear regression models were run with the base model containing lean body mass and sex and with 5 subsequent models, each adding variables to the previous model as follows: model 1, maternal obesity; model 2, birth weight; model 3, maternal covariates (race, age, marital status, smoking status, education, household income, and enrolled child in WIC); model 4, breastfeeding duration; and model 5, the maternal infant-feeding factors that were significantly associated, in bivariate analysis, with child fat mass and maternal obesity. We then assessed the change in the regression coefficient for maternal obesity between models 4 and 5 to determine the extent to which these feeding factors mediated the relationship between maternal obesity and child adiposity.17 Similar linear regression models were used with BMI z score as the dependent variable, and similar logistic models were used for the binary dependent variables of high adiposity and overweight. None of these models, however, contained lean body mass or sex.
Of the 313 children in this analysis, 166 (53%) were boys; 80% of the mothers were white and 20% were black. The mean ± SD birth weight of the children was 3388 ± 554 g, with 12% of the newborns weighing 4000 g or more. Sixty-eight mothers (22%) enrolled their infants in WIC, and 231 children (74%) were breastfed. At the time of the child's birth, the mean ± SD age of the mothers was 30 ± 5.0 years. Half of the mothers had completed college, 86% were married, and 7% were smokers.
The mean ± SD maternal BMI before pregnancy was 24.9 ± 5.6, and 17% of the mothers (n = 52) were obese before becoming pregnant. At 5 years of age, the percentages of children with BMI at the 85th percentile or higher and at the 95th percentile or higher were 26% and 10%, respectively. Total fat mass and BMI z score were significantly correlated (r = 0.735; P<.001), as were percentage of body fat and BMI z score (r = 0.552; P<.001). The mean ± SD fat mass (unadjusted), lean body mass, and percentage body fat were 4.55 ± 1.64 kg, 14.05 ± 1.93 kg, and 23% ± 5.43%, respectively. After adjusting for lean body mass, girls had significantly higher fat mass than boys (5.18 vs 3.99 kg; P<.001). Mean fat mass is, hereafter, reported after adjustment for lean body mass and sex.
Comparing the 313 children who underwent DXA at 5 years of age with the 59 children in the original cohort who did not undergo DXA, there was no significant difference between the 2 groups in BMI z score at 3 years of age (0.34 vs 0.37; P = .86) or in the percentage with BMI at the 85th percentile or higher (22% vs 25%; P = .67) or at the 95th percentile or higher (6.7% vs 7.1%; P = .91).
The only maternal infant-feeding factor that was significantly related to fat mass at 5 years of age was high concern about the infant overeating or becoming overweight (Table 2). Compared with children whose mothers had lower levels of concern (lowest 2 tertiles), the children whose mothers had high concern had a 0.67 kg (95% confidence interval, 0.31-1.03 kg) greater fat mass at 5 years of age. Because the standard deviation in fat mass was 1.64 kg, this 0.67-kg difference represents an effect size of 0.41 SD unit (0.67/1.64). These children were also more likely to have high adiposity. Mothers who reported high control over their infant feeding schedules tended to have children with lower rather than higher fat mass at age 5 years.
Mothers who were concerned about their infant overeating or becoming overweight also had children with higher BMI z scores and prevalence of overweight (Table 2). In the opposite direction, mothers with high concern about their infant undereating or becoming underweight had children with lower BMI z scores and prevalence of overweight.
At 5 years of age, children born to mothers who were obese before pregnancy were more likely to have higher fat mass and BMI z scores and to be overweight (Table 3). Mothers who were obese before pregnancy were significantly more concerned about their infant overeating or becoming overweight. Although none of the other differences in maternal infant-feeding style between obese and nonobese mothers were statistically significant, obese mothers also tended to exert less control over the infant feeding schedule, to have lower social contact during feeding, and to have less concern about the infant being underweight.
Because “overeating” was the only factor significantly related to adiposity at age 5 years and maternal obesity before pregnancy, this was the only feeding factor included in the regression analyses. Based on the regression models, children born to obese mothers had a 0.54-kg greater fat mass (95% confidence interval, 0.10-0.98 kg) than children born to nonobese mothers, even after controlling for the child's lean body mass, sex, birth weight, a variety of maternal covariates (race, age, marital status, smoking status, education, household income, and WIC enrollment), and breastfeeding status (model 4 under “fat mass” in Table 4). This is a difference in fat mass of one third of a standard deviation.
This difference was reduced to 0.46 kg (a 15% reduction) by the addition of the “overeating” feeding factor in the final model (model 5 under “fat mass” in Table 4). In this final model, the unstandardized regression coefficient on the “overeating” factor was 0.60 (95% confidence interval, 0.23-0.98), which means that after controlling for all the covariates, including birth weight, maternal obesity, and breastfeeding duration, children whose mothers had high concern about their overeating or becoming overweight during infancy had a 0.60-kg higher fat mass at 5 years of age. Maternal obesity, birth weight, maternal covariates, breastfeeding duration, and the “overeating” feeding factor together explained 36% of the variance in child fat mass.
The regression models using the 2 BMI-derived dependent variables produced results similar to those using fat mass as the dependent variable. When BMI z score was used, there was a 13% reduction in the regression coefficient on maternal obesity (from 0.53 SD of BMI to 0.46 SD of BMI) between model 4 and model 5 (Table 4). The addition of the “overeating” infant feeding factor to the logistic model for overweight reduced the odds ratio for maternal obesity by 8% (Table 5).
The IFQ described maternal infant-feeding style in terms of 7 factors, but only a single factor—concern about the infant overeating or becoming overweight—was significantly associated with fat mass at 5 years of age, as measured directly using DXA. This factor was also associated with high adiposity (percentage of body fat in the upper quartile of study participants), BMI z score, and overweight (BMI ≥85th percentile). Low awareness of infant satiety cues, high use of food to calm infant fussiness, and low social contact with the infant during feeding were not significantly related to childhood adiposity. High control over the infant feeding schedule tended to be associated with lower, rather than higher, childhood adiposity.
Mothers who were obese (BMI ≥30) before pregnancy had children with significantly higher fat mass, BMI z scores, and prevalence of overweight at 5 years of age. This relationship between maternal obesity and the child's later adiposity/BMI was explained, in part, by maternal concern that the infant was overeating or would become overweight. This finding was independent of birth weight, breastfeeding, and maternal covariates, such as race, education, and income.
This study was limited by the fact that the mothers completed the IFQ when the child was 3 years of age rather than at the end of infancy. This could have led to poorer recall of the infant feeding style or could have biased that recall because of the child's weight at the time the IFQ was completed. Controlling for BMI at age 3 years would not necessarily overcome this potential limitation, and doing so would inappropriately control for a factor on the causal pathway between infant-feeding style and adiposity at age 5 years.18
In addition to this limitation, IFQ scores were not normally distributed and could have failed to detect meaningful clinical variation in some of the 7 factors. The present study may have also lacked power to detect some differences in maternal infant-feeding style that are of potential clinical significance. For example, there was a tendency for children to have lower fat mass if their mothers exerted high control over the infant feeding schedule, and there was also a tendency for obese mothers to report being less likely to exert high control over the infant feeding schedule. These findings warrant further study because they have implications for advising parents about infant feeding.
Kramer and colleagues8 administered the 10-item Maternal Feeding Attitudes Questionnaire to mothers at delivery. This instrument asked mothers about how they planned to feed their children. It was designed to assess a mother's tendency to “push” food into her child (eg, by trying to feed a fussy infant even if the infant had just been fed), with higher scores indicating a “pushier” feeding style. Scores were not significantly associated with BMI at 12 (n = 382) or 24 (n = 347) months of age.10 The IFQ was previously administered to 453 mothers of varying sociodemographic backgrounds with children 11 to 23 months of age.9 Consistent with the present study, maternal concern about the infant overeating or becoming overweight was the only infant-feeding factor that was associated with an increased relative weight (weight-for-height ≥90th percentile) in the children.
A strong and graded relationship between maternal BMI before pregnancy and childhood obesity (BMI ≥95th percentile) has previously been demonstrated,4 suggesting that maternal obesity is the primary clinical risk factor, available at birth, for predicting preschooler obesity. To our knowledge, only 1 other study19 has demonstrated the relationship between maternal obesity before pregnancy and childhood adiposity using DXA, a direct measure of adiposity.
Maternal concern about her infant overeating or becoming overweight explained some of the relationship between maternal obesity and childhood adiposity. However, this feeding factor was associated with later fatness, independent of maternal obesity, suggesting that this feeding factor represents more than a mother's concern about transmitting to the child any of her own genetic susceptibility to obesity. Maternal concern about the infant overeating may be a marker for a behavioral eating pattern in the child that promotes obesity. Others have shown that an infant's eating style, as measured by the vigor of nipple sucking, is associated with later body weight.20- 22
We do not have data on weight during infancy, but the present study suggests that mothers have some concern about their infant's appetite and weight, independent of birth weight, that predicts adiposity at age 5 years. Mothers may begin to alter feeding strategies in response to their infant's appetite and weight. Several studies indicate that higher levels of maternal control, manifested as restriction over a child's intake of snack foods, are associated with higher BMI levels in the children,23,24 especially in children of obese mothers.25 However, it is not clear whether this restriction is a response to the child's weight or the cause of it. Our findings suggest that mothers are concerned about their child's weight before the preschool years.
Our finding that maternal concern about the infant becoming overweight is associated with later adiposity seems to contradict other studies26,27 suggesting that parents do not accurately perceive when their preschool children are overweight. One possible explanation for these different findings is that many parents may consider it unreasonable to label infants or young children as “overweight” or that such a label may be a reflection of poor parenting. However, asking parents whether they are concerned that their child “will become overweight” may reveal parents' more latent concerns about their infant's weight.
Interactions during feeding are an important part of establishing attachment between mother and child.28 Therefore, absent any clearer evidence that specific maternal infant-feeding styles are related to later adiposity, practitioners should avoid overzealous recommendations that might interfere with feeding interactions.
Because obesity is partly a heritable disorder and because feeding and infant growth contribute to parents' sense of competence in their ability to nurture their children, the topic of overeating or overweight can be sensitive. Clinicians may find it useful, however, to ask parents during their child's infancy whether they are “concerned about their infant eating too much or becoming overweight.” These questions could also be paired with questions about the infant “not eating enough or becoming underweight.”
Such questions may serve to open an early and constructive dialogue with parents about weight-related issues. For all parents, certain household behaviors can be encouraged that may prevent obesity29 while providing health benefits unrelated to obesity.30 For parents who express concern about obesity in their children, there are constructive ways to expand the dialogue.31 Useful areas for open-ended inquiry include asking about the family history of obesity, which members of the family share a concern about the child's obesity risk, what specifically is causing the concern, what household interventions have already been tried, and what are the perceived barriers and facilitators to those efforts. Responses to these questions can help form an action plan that is based on the family's input.
Correspondence: Hillary L. Burdette, MD, MS, Division of Gastroenterology and Nutrition, Children's Hospital of Philadelphia, 34th Street and Civic Center Boulevard, Philadelphia, PA 19104 (email@example.com).
Accepted for Publication: November 15, 2005.
Author Contributions: Dr Burdette had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Funding/Support: This work was supported by grant R01HL/DK64022 from the National Institutes of Health (Drs Whitaker and Daniels).
Role of the Sponsor: The sponsor had no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, and approval of the manuscript.
Acknowledgment: We thank Karen Munson, RN, for her role in managing the data collection.
Burdette HL, Whitaker RC, Hall WC, Daniels SR. Maternal Infant-Feeding Style and Children's Adiposity at 5 Years of Age. Arch Pediatr Adolesc Med. 2006;160(5):513-520. doi:10.1001/archpedi.160.5.513