Is Parental Control over Children's Eating Associated with Childhood Obesity? Results from a Population-Based Sample of Third Graders
Abstract
Objective: Identifying parental behaviors that influence childhood obesity is critical for the development of effective prevention and treatment programs. Findings from a prior laboratory study suggest that parents who impose control over their children's eating may interfere with their children's ability to regulate intake, potentially resulting in overweight. These findings have been widely endorsed; however, the direct relationship between parental control of children's intake and their children's degree of overweight has not been shown in a generalized sample.
Research Methods and Procedures: This study surveyed 792 third-grade children with diverse ethnic and socioeconomic backgrounds from 13 public elementary schools. Parental control over children's intake was assessed through telephone interviews using a state-of-the-art instrument, and children were measured for height, weight, and triceps skinfold thickness.
Results: Counter to the hypothesis, parental control over children's intake was inversely associated with overweight in girls, as measured by body mass index, r = −0.12, p < 0.05, and triceps skinfolds, r = −0.11, p < 0.05. This weak relationship became only marginally significant when controlling for parents’ perceptions of their own weight, level of household education, and children's age. No relationship between parental control of children's intake and their children's degree of overweight was found in boys.
Discussion: Previous observations of the influence of parental control over children's intake in middle-class white families did not generalize to 8- to 9-year-olds in families with diverse socioeconomic and ethnic backgrounds. The present findings reveal a more complex relationship between parental behaviors and children's weight status.
Introduction
The prevalence of obesity in children and adolescents in the United States is rising, with the highest prevalence rates among Latino/Hispanic and African American children (1). Also alarming are data demonstrating that childhood and adolescent obesity are often resistant to most available treatments (2, 3) and are predictive of subsequent adult obesity (4). Thus, identifying factors that influence childhood and adolescent obesity are critical for the development of effective prevention and treatment programs.
Parental behaviors are thought to influence children's weight (5). For instance, it has been hypothesized that parents who wish to spare their children the negative social consequences of being overweight may impose limits on their children's access to food (6). In contrast, it has also been hypothesized that parents who control their children's eating habits may interfere with their children's ability to regulate their own intake, ironically resulting in excess weight gain (6, 7, 8).
Preliminary evidence supports the second hypothesis. In a small cross-sectional study, parents’ restraint over their children's eating behavior (as measured by a 4-item scale, e.g., child was not free to eat whenever hungry) was significantly associated with overweight in girls (n = 26), although not in boys (n = 16), perhaps due to the small sample of boys (7). In a laboratory study, Johnson and Birch (8) examined whether parents’ reported control of children's intake undermined their children's ability to regulate their intake. In a sample of 77 preschoolers (3 to 5 years old), parents who reported greater control over their preschoolers’ intake (as measured by a 6-item scale) had preschoolers who were less able to self-regulate their intake after a high-calorie preload (8). Furthermore, girls who were less able to self-regulate their intake were heavier, as measured by skinfolds, than girls who were better able to regulate. Birch and Fisher (9) reported similar findings from a subsequent laboratory study of 156 white, 5-year-old daughter and mother dyad pairs from two-parent families (9). Although these preliminary findings have been widely endorsed (10), the direct relationship between parental control of children's intake and their children's degree of overweight has not been shown in a generalized sample. In fact, in the two laboratory studies, the correlations between parental control over intake and children's degree of overweight were not reported (8, 9). A recent discordant sibling analysis in 18 middle- to upper-middle-class white families found no differences between maternal control over feeding among 7- to 12-year-old obese and non-obese siblings (11).
The potential impact of parental control over children's eating on children's overweight has provocative implications for the design of future obesity prevention programs, especially in light of the possible gender differences. However, because the past studies have had small and/or selected samples (e.g., teachers selected which overweight and normal weight children would participate in the Costanzo and Woody study) (7), confidence in these findings is extremely limited. Therefore, we examined the strength of the relationship between parental control over intake and children's degree of overweight in a large population-based sample of third-grade children with diverse ethnic and socioeconomic backgrounds. Identifying factors that predict the development of obesity in preadolescence is especially critical given the need to preempt the adolescent weight gain predictive of adult obesity (4).
Research Methods and Procedures
The study was conducted during the fall of 1996 in 13 northern California public elementary schools. All third-grade students and their parents or guardians were eligible to participate in the study. Parents were informed of the study in writing and given an opportunity to refuse participation for their children. Trained study personnel collected children's data, including physical measurements and questionnaires, during regular school hours. Parents or guardians were interviewed over the telephone and interviews were preferably conducted with the mothers (biological, adoptive, or stepmothers) or female guardians. However, if the mothers were unavailable, interviews were conducted with the fathers or male guardians. In the few families where parents or guardians were not available, grandparents were interviewed. The Human Subjects Panel at Stanford University approved the study.
Measures
Household Educational Level.
Parents reported the education level of all parents or guardians in the household. The highest reported education level was used as the household education level.
Parents' Perceptions of Their Own Weight.
Parents were asked to judge their own weight on a three-point scale: underweight, about the right weight, or overweight.
Parental Control over Their Children's Food Intake.
This construct was measured using the state-of-the-art instrument developed by Birch and colleagues (8). Parents rated each of six items on a five-point Likert scale ranging from disagree (1) to agree (5) (see Table 1 for items). The Cronbach α in this sample of third-graders was 0.61, and the internal consistency could not be increased substantially by excluding items.
| 1. When my child does not finish dinner, he/she should not get dessert. |
| 2. My child should always eat all of the food on his/her plate. |
| 3. Generally, my child should only be permitted to eat at set mealtimes. |
| 4. My child often has to be strongly encouraged to eat things he/she doesn't like because those foods are often good for him/her. |
| 5. My child should be strongly reprimanded for playing or fiddling with food. |
| 6. I have to be especially careful to make sure my child eats enough. |
Parents' Perceptions of Their Children's Weight.
Parents were asked to evaluate their children's weight on a five-point Likert scale ranging from very underweight (1) to very overweight (5).
Children's Age.
Children circled the month and year in which they were born, and their age was calculated in months.
Children's Ethnicity.
Parents identified the ethnicity of their children. Children were classified into the following ethnic groups: Asian, African American, Latino/Hispanic, Native American, Pacific Islander, white, and multiethnic.
Children's Body Mass Index and Triceps Skinfold Thickness.
Height, weight, and triceps skinfold thickness were obtained using standard protocols (12). Test-retest reliabilities were >0.99. Body mass index (BMI; kg/m2) is a preferred measure of childhood obesity in clinical and epidemiological studies (13). Triceps skinfold thickness is a measure of subcutaneous body fat.
Data Analysis
To examine the sample for possible bias due to nonparticipation and/or missing data, we used t tests and χ2 analyses to compare children with complete data (the analysis sample) and children without complete data, on age, gender, and BMI, and to test for differences between girls and boys in the sample. To test the hypothesis that parental control over children's eating is positively associated with children's body fatness, we calculated Spearman correlation coefficients between parental control over children's food intake and BMI and triceps skinfold thickness. Because previous research has suggested that the relationship between parental control over children's intake and children's body fatness may differ for girls and boys, these and all subsequent analyses were performed separately for girls and boys. Spearman correlation coefficients were also calculated between parental control over children's intake, BMI, triceps skinfold thickness, and other conceptually related measures, i.e., household educational level, parents’ perceptions of their own weight (also a proxy for parent overweight), parents’ perceptions of their child's weight, and the child's age. These variables were chosen because of previous theory or research linking them to eating behavior, child adiposity, and/or parenting behaviors (10). To investigate the independent relationship of parental control over children's eating and BMI, controlling for these conceptually important variables, we performed a multiple linear regression, including household education level, parents’ perception of their own weight, and children's age as covariates. A two-tailed α level of 0.05 was used for all statistical tests. For the analysis sample of girls (n = 395) and boys (n = 397), we had 80% power to detect correlation coefficients as small as 0.14 (14).
Results
There were 998 eligible children, including 13 sets of siblings. One child of each sibling pair was randomly selected for use in the analyses. Of the 985 remaining eligible third-grade girls and boys, only 28 (2.8%) of the parents refused to let their children participate. Of the 957 children who participated, 792 (82.8%) of the children had complete data for the two variables of primary interest in this study: they had parents who participated in the telephone interviews and BMI measurements. Of the parents who participated in the telephone interviews, 656 (82.8%) were mothers or female guardians, 126 (15.9%) were fathers or male guardians, and 10 (1.3%) were grandparents living in the home.
Of the 165 children without complete data, 153 (92.7%) had parents who did not participate in the interviews, 7 (4.2%) did not have BMI data, and 5 (3.0%) had neither. Parents did not participate in the interviews for the following reasons: 13 had problem phone numbers; 18 were called at least 10 times without an answer; 41 were contacted but the parents were unavailable; 6 participated in the interviews but were relatives other than parents, guardians, or grandparents; 29 did not speak English or Spanish; 42 refused to participate in the parent interviews; and 9 reported other reasons.
There were no significant differences between children with complete data (N = 792) and those without (n = 165) for age, gender, or BMI.
Descriptive Statistics
Because the research to date has indicated that the relationship between parental control over children's intake and children's degree of overweight may differ for girls and boys, statistical comparisons were conducted by gender (see Table 2). Although girls were younger than boys, the difference was not clinically significant. As expected for this age group (15), girls had larger triceps skinfolds than boys.
| Girls | Boys | |
|---|---|---|
| n | 395 | 397 |
| Parents who view themselves as overweight (%) | 53.7 | 52.6 |
| Households with a college degree (%) | 43.0 | 43.1 |
| Parental control over intake (range 1–5) | 2.8 ± 1.0 | 2.7 ± 1.0 |
| Children's age (years) | 8.4 ± 0.4 | 8.5 ± 0.4* |
| Children's ethnicity (%) | ||
| Asian | 21.3 | 20.0 |
| African American | 4.1 | 3.5 |
| Latino/Hispanic | 20.0 | 18.5 |
| Multiethnic | 9.0 | 10.4 |
| Native American | 0.3 | 1.0 |
| Pacific Islander | 0.8 | 1.8 |
| White | 44.6 | 44.8 |
| Children's BMI (kg/m2) | 17.7 ± 3.3 | 18.0 ± 3.3 |
| Children's triceps skinfold thickness (mm) | 15.1 ± 5.3 | 13.2 ± 5.6† |
- * p < 0.05.
- † p < 0.001.
Bivariate Statistics
Spearman correlations between the variables were examined separately by gender (see Table 3). For girls, the relationship between parental control over children's intake was inversely and significantly correlated with BMI. Similar results were seen for triceps skinfold thickness. For boys, the relationship between parental control over children's intake and both measures of overweight were close to zero, and not statistically significant. The strength of the relationships between parental control over children's intake and BMI were not statistically significantly different between girls and boys when directly comparing the two correlation coefficients after Fisher's Z-transformations.
| Highest household educational level | Parents' perception of their weight | Parental control over intake | Children's age | Children's BMI | |
|---|---|---|---|---|---|
| Girls | |||||
| Highest household educational level | — | ||||
| Parents' perception of their own weight | −0.12* | — | |||
| Parental control over children's intake | 0.01 | −0.13† | — | ||
| Children's age | −0.08 | −0.02 | −0.14† | — | |
| Children's BMI | −0.19‡ | 0.21‡ | −0.12* | 0.11* | — |
| Children's triceps skinfolds | −0.07 | 0.15† | −0.11* | 0.12* | 0.82‡ |
| Boys | |||||
| Highest household educational level | — | ||||
| Parents' perception of their own weight | 0.03 | — | |||
| Parental control over children's intake | −0.02 | −0.15† | — | ||
| Children's age | −0.07 | 0.02 | 0.00 | — | |
| Children's BMI | −0.09 | 0.21‡ | −0.02 | 0.09 | — |
| Children's triceps skinfolds | −0.02 | 0.15† | 0.01 | 0.06 | 0.77‡ |
- * p < 0.05.
- † p < 0.01.
- ‡ p < 0.001.
Although parent and child adiposity and weight-related attitudes and behaviors have been reported to vary by ethnicity (16, 17), previous research on parental control over children's intake has been limited almost exclusively to whites (10). We performed an exploratory analysis of the relationships between parental control over intake and children's BMI, separately by gender for the three largest ethnic groups in this sample. For Asian and white girls, the correlations between parental control and BMI were similar to all girls, r = −0.15 and r = −0.11, respectively. For Hispanic girls, the correlation was close to zero, r = −0.03. For Asian, Latino/Hispanic, and white boys, similar to all boys, the correlations were fairly close to zero, r = −0.08, r = 0.09, and r = 0.02, respectively.
To investigate whether the weak associations in girls were due to the mix of mothers and fathers interviewed, the relationship between parental control and children's overweight was examined for mothers and female guardians only. The correlation was weaker and not significant, r = −0.08.
Multivariate Statistics
To investigate the independent relationship of parental control over intake and children's BMI controlling for other conceptually important variables, a multiple linear regression was conducted for all girls. The covariates included parents’ perception of their own weight, household educational level, and children's age. After controlling for the covariates, parental control over intake was only marginally significantly related to girls’ BMI (B = −0.31, SE = 0.16, β = −0.10, t[388] = −2.0, p = 0.05). However, two covariates remained significant, parents’ perception of their own weight, (B = 1.04, SE = .27, β = 0.19, t[388] = 3.8, p < 0.001) and household educational level (B = −0.38, SE = 0.11, β = −0.17, t[388] = −3.4, p < 0.001).
Exploratory Analyses
As mentioned previously, parents who wish to spare their children the negative consequences of overweight may impose limits on their access to food. In an exploratory analysis, the relationship between parental control over their children's intake and their perceptions of their children's weight was examined. However, parents who controlled their daughters’ intake were more likely to perceive their daughters as underweight (r = −0.14, p < 0.01). Parents who controlled their sons’ intake were no more likely to perceive their sons as underweight than parents who did not control intake (r = −0.01).
Parents who are worried about their daughters being underweight may control their children's intake to ensure they are eating enough. Indeed, examining the content of individual scale items post hoc revealed that the endorsement of some items may be associated with encouraging children to eat rather than restricting children from eating. For instance, parents may have insisted their children eat all the food on their plate, eat foods they didn't like because those foods were good for them, and not fiddle with their food (i.e., actually eat the food), because they wanted their children to gain weight. Items measuring different types of parental control may account for the moderate internal consistency for the scale in this sample (Cronbach α = 0.61) and the weak relationship of the scale with BMI. Therefore, we conducted a principal components analysis of the scale items, which yielded two significant factors accounting for 35% and 18% of the total variance in this sample, respectively. The first factor comprised items 2 through 6 (Cronbach α = 0.64) and the second factor comprised item 1. However, the relationship between the subscale comprising items 2 through 6 and children's BMI was similar to the original scale for girls (r = −0.11, p < 0.05) and boys (r = −0.03). Again, similar to the original scale, the subscale did not remain significantly related to girls’ BMI (B = −0.26, SE = 0.14, β = −0.09, t[388] = −1.8, p = 0.07) after controlling for the covariates in a multiple linear regression. Finally, item 1 was not related to BMI in girls (r = −0.05) or boys (r = 0.02).
Discussion
We examined the hypothesized relationship between parental control over children's intake and overweight in children. The large multiethnic population-based sample of preadolescent girls and boys enabled a statistically powerful test of this hypothesis outside of a laboratory setting. Counter to the hypothesis, we found only a weak and, in fact, inverse relationship among girls; parents who reported greater control over their children's intake had daughters who were less overweight. In addition, this weak relationship became only borderline significant when controlling for other conceptually important variables in multivariate analyses. Consistent with past research (7), parental control over intake was not associated with overweight among boys. Moreover, it is also interesting to note that parents who reported that they were overweight themselves, a risk factor found to predict future overweight in their children in previous research (18), reported exerting significantly less control over their children's intake.
The lack of or weak inverse association between parental control over intake and overweight in girls may result from several reasons. First, parents of 3- to 5-year-olds, as studied by Johnson and Birch (8) and Birch and Fisher (9), may have more influence over their children's eating behaviors than parents of 8- to 9-year-olds, as studied in this sample. As girls reach preadolescence, the impact of parental control over food intake may be diluted by influences from their peers, mass media, and the school environment (17, 19). Alternatively, parents may relax their control strategies as their daughters get older. In our sample, the parents of older girls reported less control over their daughters’ intakes. Despite previous suggestions that parental behaviors regarding children's intake may differ by socioeconomic status (10), we found no association between household educational level and parental control over intake.
Second, the lack of an association may be due to the fact that some of the parents who were interviewed may not have been the adult primarily responsible for their children's eating environment. In this present study, more than 80% of the parents interviewed were mothers and female guardians. Although it may be safe to conclude that mothers are typically in control of their children's eating environment, future research needs to more clearly identify the role of fathers in dual- and single-parent households on their children's eating habits.
Third, the lack of an association between parental control over intake and girls’ overweight may stem from the fact that the scale items were measuring different types of parental control such as encouragement or restriction over their children's intake. The exploratory analysis revealed that parents who controlled their daughters’ intake were more likely to view them as underweight, suggesting that such parents may control their children's intake to ensure their daughters are eating enough rather than eating too much. However, based on post hoc analyses, subscales that examined parents’ encouragement versus restriction over their children's intake were also neither strongly nor significantly related to children's overweight.
It is always possible that our results could be influenced by biased sampling. However, data were available for 80% of the eligible population and there were no significant differences in children's age, gender, or BMI between participants with and without complete data. Therefore, this is unlikely.
A limitation of this study is that the proposed mechanism of the hypothesis linking parental control over children's intake and overweight in children, i.e., children's ability to self-regulate their own intake, was not measured. However, the hypothesis was undermined more generally because the direct relationship between parental control and overweight in children was not seen. Given that the covariates were significant in multivariate analyses (i.e., parents’ perception of their own weight and household educational level), these variables may be more conceptually important than parental control of children's intake per se in influencing children's overweight. Conceivably, parents who are overweight and have less education may not be as likely to provide their children with access to healthy foods (20) or to reinforce low-fat eating habits (21).
In summary, previous observations of the influence of parental control over intake among 3- to 5-year-old preschoolers in middle-class white families did not generalize to 8- to 9-year-olds in families of diverse socioeconomic and ethnic backgrounds. Our findings reveal the complexity of the relationships between parental behaviors and children's weight status (10, 22). There is a critical need for future research to explore the full range of parental influences, both positive and negative, systematically. In particular, longitudinal data are essential to clarify how parental influences may contribute to subsequent childhood obesity (23).
Acknowledgments
This work was done during the tenure of a Clinician-Scientist Award from the American Heart Association (T. N. R.), and funded in part by Public Health Service, National Institutes of Health, National Heart, Lung, and Blood Institute Grants R01 HL54102 (T. N. R.) and R29 HL60154 (M. K.). We thank Joel D. Killen, Ph.D., Sarah J. Erickson, Ph.D., Christina Russell, Kathy Valenzuela, Mireya Samaniego, Dina L.G. Borzekowski, Ed.D., Sally McCarthy, Connie Watanabe, M.S., Ann Varady, M.S., and Helena C. Kraemer, Ph.D. for their contributions to this study, and the students, parents, teachers, and administrators who participated in this project.
