|Home | About | Journals | Submit | Contact Us | Français|
To assess whether parental overweight status and disinhibited overeating are predictive of daughters’ accelerated weight gain and disinhibited overeating.
Participants were part of a longitudinal study of girls (N = 197) and their parents. Measured height and weight were used to calculate BMI [weight (kilograms)/height (meters)2]. Parents’ disinhibited eating behavior was assessed using the Eating Inventory. Girls’ disinhibited eating was assessed using a behavioral protocol to measure eating in the absence of hunger. Girls were classified based on parental overweight at study entry into four groups: neither, mother only, father only, or both parents overweight.
Girls with both parents overweight had the most rapid increases in BMI from 5 to 13 years of age; BMI increased most slowly among the neither parent overweight group, with intermediate increases in BMI among mother only and father only overweight groups. Daughters with both parents overweight at study entry were eight times more likely to be overweight at age 13, controlling for daughters’ weight at age 5. Girls with both parents overweight had higher levels of disinhibited eating across all ages than all other groups. Although girls in all parental weight status groups showed increases in disinhibited eating over time, girls with both parents overweight had larger increases in disinhibited eating over time compared with all other groups.
Girls growing up in families differing in parental overweight had divergent developmental trajectories for BMI and disinhibited overeating. Findings reveal the need to focus prevention efforts on overweight parents of young children.
Parents provide both genes and environments that may promote behaviors associated with excessive weight gain in children (1-8). Additionally, obesity runs in families, and having obese parents increases obesity risk in children (9-15). Parental weight status also predicts tracking of childhood overweight (11,16-18); an overweight child living in a family where one or more parent is overweight is likely to remain overweight throughout his or her childhood and into adolescence and adulthood (16). Additionally, Treuth et al. (19) reported that number of overweight parents predicted fat gain among normal weight girls.
Accelerated weight gain in infancy and early childhood predicts later risk for overweight and obesity during childhood, adolescence, and adulthood (20-24). Stettler et al. (23) found that accelerated weight gain in infancy had the strength to predict obesity at age 20. Similar results were found by Law et al. (25), providing evidence for links between accelerated weight gain from 1 to 5 years of age and higher weight status at age 22. Furthermore, accelerated weight gain in childhood was linked to several negative health outcomes in adulthood, including coronary heart disease, insulin resistance, hypertension, and type 2 diabetes (25,26).
Several studies have addressed the ways in which parental overweight status may also influence the development of children’s eating behavior, dietary intake, and physical activity patterns (2,3,5,8,27,28). This study describes daughters’ developing patterns of disinhibited overeating, an eating style that has been consistently associated with weight gain, overweight, and obesity, among both children (29-31) and adults (32-36). Eating in the absence of hunger (EAH)1 is a behavioral measure of disinhibited overeating and is characterized as the tendency to consume large amounts of palatable foods in a short period of time in a fashion that is not a response to hunger. Given its associations with weight gain and overweight, disinhibited eating has been proposed as a behavioral phenotype for obesity (37). Both genetic and environmental differences among families may shape the emergence of differences in eating styles such as disinhibited overeating. Additionally, overweight parents may be exhibiting distinctly different eating styles than normal weight parents and, in particular, may serve as models for their children’s disinhibited eating (38-40).
In this research, we address whether having one or two overweight parents is predictive of accelerated weight gain and disinhibited eating during middle childhood and into adolescence and whether disinhibited overeating aggregates within families. We hypothesized that overweight parents would report higher levels of disinhibited overeating than parents who are not overweight, and that, across middle childhood, daughters growing up in families with both parents overweight would exhibit greater increases in BMI, be more at risk for overweight, and have higher levels of disinhibited overeating than girls from families with one or neither parent overweight.
Participants included 197 non-Hispanic, white families living in central Pennsylvania recruited as part of a longitudinal study of the health and development of young girls; the sample was not recruited based on weight status. Eligibility criteria for girls’ participation at the time of recruitment included living with both biological parents, the absence of severe food allergies or chronic medical problems affecting food intake, and the absence of dietary restrictions involving animal products. Families were recruited for participation in the study using flyers and newspaper advertisements. In addition, families with age-eligible female children within a five-county radius received mailings and follow-up phone calls (Metromail Limited, County Durham, UK).
At study entry, participants included 197 girls who averaged 5.4 ± 0.4 (standard deviation) years of age and their parents, of whom 192 families were reassessed 2 years later when girls averaged 7.3 ± 0.3 years. A third assessment with 183 families was conducted 2 years later when girls averaged 9.3 ± 0.3 years, followed by a fourth assessment with 177 families when girls averaged 11.3 ± 0.3 years and a fifth assessment with 168 families when girls averaged 13.3 ± 0.3 years. Attrition was primarily caused by family relocation outside of the study area. No significant differences were found between the initial weight statuses of participants lost to follow-up (n = 29) and of participants remaining in the study through age 13 (n = 168). This finding was also true for girls’, mothers’, and fathers’ BMI at daughter age 5. The University’s Institutional Review Board approved all study procedures, and parents provided consent for their family’s participation before the study began.
Parental measures used in this study were collected at study entry, when daughters were age 5; girls’ measures were collected when daughters were ages 5, 7, 9, 11, and 13 years.
Mothers completed a Background Questionnaire, developed in our laboratory, that assessed family background characteristics including combined family income and mothers’ and fathers’ years of education. Parents chose from four categories for income: less than $20,000/yr, $20,000 to $35,000/yr, $35,000 to $50,000/yr, or $50,000+/yr. For education, parents listed whether they had the following degrees/diplomas: high school, associates, technical/vocational school, bachelors, masters, PhD, MD, JD, or other.
Mothers’ and fathers’ height and weight were measured in the laboratory and used to calculate BMI [weight (kilograms)/height (meters)2]. Because we were specifically interested in examining differences between overweight and normal weight parents, we divided the sample into four groups using the widely accepted cut-off for defining overweight in mothers (BMI ≥ 25 kg/m2) (41). Because our purpose was to define groups in terms of parent overweight, we used a slightly more conservative BMI cut-off of ≥27 kg/m2 to define overweight in fathers to reduce the likelihood of misclassifying fathers whose slightly elevated BMIs reflect higher lean body mass rather than overfatness and to provide more equal group sizes. This resulted in four groups: neither parent overweight (n = 54), mother only overweight (n = 41), father only overweight (n = 40), and both parents overweight (n = 62).
Parents’ disinhibited eating behavior was assessed using the dietary disinhibition subscale from the Eating Inventory developed by Stunkard and Messick (42). The dietary disinhibition subscale consists of 16 items that measure parents’ reported disinhibition of cognitive control of eating or the tendency to eat in response to external factors such as the presence of food and emotional states. Respondents select among true/false and Likert-type response options. Total scores for each subscale are calculated by summing subscale items; higher scores reflect higher levels of reported disinhibited eating. Internal consistency scores for scale items on the disinhibition scale were 0.83 and 0.73 for mothers and fathers, respectively.
Girls’ BMI scores were generated using procedures described above for calculating BMI in parents. Height and weight were measured by a trained staff member following procedures described by Lohman et al. (43). Age- and sex-specific BMI percentiles were used to determine the prevalence of overweight in girls in this sample. Based on standardized reference criteria (44), a BMI score corresponding to the 85th percentile was used to classify girls as at-risk-for-overweight; BMI percentile scores at or above the 95th percentile were used to classify girls as overweight. Because our research focus is on predicting adiposity change across middle childhood, we used BMI to assess adiposity change, based on the evidence presented by Cole et al. (45) and Field et al. (46), indicating that, although BMI z-score is the preferred measure for assessing adiposity at a single occasion, absolute BMI is the best for assessing change over time because within-child variability over time is related to the child’s level of adiposity. Because all children in the study were girls and are in the same age cohort, standardized scores or percentile ranking were not needed, and BMI can be used as a relative measure of adiposity within our sample. For descriptive purposes, the BMI data for each group at each age are presented using the Centers for Disease Control and Prevention growth charts (47), which provide information on how the groups’ data relate to the BMI percentile data.
Girls’ disinhibited eating was assessed by measuring their intake of palatable foods in the absence of hunger. This procedure was developed in our laboratory and has been previously described elsewhere (38,48,49). Briefly, 20 minutes after eating and reporting that they were full, girls were left alone in a room and were asked to taste and rate a set of palatable snack foods. Each girl was told that she had a few minutes alone and that after rating the foods she could eat as much or as little of the foods as she desired. Large amounts of the foods were available so that energy intake was not limited by availability. Other available activities included listening to music, looking at books and magazines, and simple table games. Energy intake was calculated from gram weights obtained by pre -and post-weighing girls’ food intake; manufacturers’ information on energy content of foods was used to determine total energy intake, which was used as the indicator of EAH. The variable of interest in this study was the percentage of energy consumed from the total energy available during the period in which the girl had ad libitum access to the snack foods.
Data were analyzed using the SAS version 8.2 (50). Descriptive information was generated for all variables of interest. Each outcome variable was assessed for normality. ANOVA with Tukey honestly significant difference post hoc tests were used to assess differences among groups on parent weight status and disinhibited eating at study entry. Pearson’s correlation coefficients were used to assess linear associations between parent and daughter characteristics.
A mixed modeling approach (Proc MIXED; SAS Institute, Inc., Cary, NC) was used to assess the effects of parental overweight on the patterns of change in daughters’ BMI and disinhibited eating from 5 to 13 years of age. Mixed modeling is a useful tool for analyzing repeated measures over time, and a major advantage is its ability to retain cases with one or more missing data points (51). Determination of model fit was based on several criteria: 1) model convergence, 2) a positive definite G matrix, and 3) statistical fit comparison based on the Akaike Information Criteria (52). For the model predicting girls’ BMI change over time, an unstructured covariance matrix was selected, as determined by the aforementioned model fit criteria; main effects of time, parent overweight group, and a parent overweight group by time interaction were tested in this model. In the model predicting girls’ EAH over time, an auto-regressive covariance structure provided the best fit model; main effects of time, parent weight group, and a parent weight group by time interaction were tested in this model. In both models (BMI and EAH) inclusion of the interaction of parental weight status group by time provided a test of the major hypothesis, because a significant interaction effect provides evidence for a differential pattern of change (or rate of growth) over time for girls in different parental weight groups.
Logistic regression (Proc LOGISTIC; SAS Institute, Inc.) was used to model the outcome of girls’ overweight status at age 13 as a function of girl’s initial BMI and parent weight status at age 5. Parent overweight was identified as a classification variable in the model, and variables were arranged so that the three class contrasts made by SAS would be neither parent overweight vs. each of the other three parent weight status classes. Girl’s BMI at age 5 was included as a covariate in the logistic regression analysis.
Parental overweight groups did not differ in family income or parent education levels; all groups had the majority of members (75% of 85%) reporting an annual family income over $35,000, and the majority of parents within each group (62% to 75%) reported at least some college education. There were also no differences among groups in parent age at study entry; at the time of recruitment, parents’ mean age for the total sample was 35.4 ± 4.8 and 37.4 ± 5.4 years for mothers and fathers, respectively.
Table 1 presents data on parental weight status and dis-inhibited eating scores at study entry. Because groups were created based on parental weight status, maternal BMI at study entry was significantly higher in families in which mothers only or both parents were overweight compared with families in which fathers only or neither parent was overweight (p < 0.001). Similarly, fathers’ BMI at study entry was highest in families in which fathers only or both parents were overweight (p < 0.001). Additionally, these patterns held for mother and father disinhibition, as is discussed in more detail below.
Among the total sample, at 5, 7, 9, 11 and 13 years of age, 20%, 21%, 31%, 29%, and 26% of girls, respectively, were classified as at-risk-for-overweight (BMI ≥ 85th percentile) and 6%, 11%, 14%, 14%, and 11% were classified as overweight (BMI ≥ 95th percentile). Girls’ mean BMI increase corresponds to a mean weight gain of ~35 kg (MΔ5-13 = 35.4 ± 11.3 kg) over the 8-year period. Increase in BMI during the period from 5 to 13 years is normative, reflecting normal growth in height and weight, and does not necessarily reflect increasing adiposity.
Patterns of BMI change over time differed for girls from families with neither, one, or both parents overweight, as shown in Figure 1. Results of the mixed-model analyses revealed a significant parental weight group by time interaction (p < 0.01). Thus, the increase in BMI from age 5 to 13 was above and beyond what would be expected for normative growth among girls from families in which both parents were overweight (MΔ5-13 = 40.8 ± 13.3 kg) compared with girls with neither parent overweight [MΔ5-13 = 29.1 ± 6.2 kg; t(181) = 4.7, p < 0.001]. Girls with two overweight parents also showed greater increases in BMI than girls from families in which fathers only were overweight [t(177) = 2.7, p < 0.01] and mothers only were overweight [t(178) = 2.0, p < 0.05]. Girls from families in which mothers only were overweight showed more rapid increases in BMI over time compared with girls from families in which neither parent was overweight [MΔ5-13 = 34.9 ± 10.9 kg; t(182) = 2.3, p < 0.05]; there were no significant differences between the fathers only and neither parent overweight groups. Because the mother only and father only groups did not differ from each other [t(179) = 0.7, p = 0.50], these analyses were also conducted combining the mother only and father only groups, allowing comparisons among the families with neither, one, or both parents overweight. In this case, the group by time interaction was significant (p < 0.001), further showing that the neither, one, and two parent overweight groups differed significantly from each other in BMI increases over time. These analyses were also conducted using a BMI ≥25 kg/m2 to classify overweight in fathers, and the results were unchanged.
Results of the logistic regression analysis indicated that, compared with families in which neither parent was overweight, girls with both parents overweight at study entry were 8.1 times more likely to be overweight at age 13, when girls’ initial BMI at age 5 was included in the model as a covariate (p < 0.05). The contrasts of mother only overweight or father only overweight vs. neither parent overweight were not statistically significant. Girls’ initial BMI at age 5 was a significant covariate (p < 0.001).
Group means for parents’ disinhibited eating at study entry are shown in Table 1. As shown, mothers’ disinhibited eating was significantly higher in the two groups of families with overweight mothers (both parents and mothers only groups) in comparison with the two groups in which mothers were not overweight (neither parents nor fathers only groups). A similar pattern was noted for fathers; fathers’ disinhibited eating was significantly higher in the two family groups with overweight fathers (both parents and fathers only) compared with families in which fathers were not overweight (neither parent and mother only groups). Fathers’ disinhibition scores were consistently lower than mothers’ disinhibition scores. Table 2 presents correlations between parents’ disinhibited eating and girls’ EAH. Although mothers’ disinhibited eating was not significantly related to daughters’ disinhibited eating style (EAH) when daughters were age 5 or 7, mothers’ disinhibited eating was significantly and positively related to girls’ EAH at ages 9, 11, and 13. Fathers’ disinhibited eating was not related to girls’ EAH at any point in time.
The overall pattern of change in girls’ disinhibited eating from 5 to 13 years for the parent overweight groups is shown in Figure 2, which shows the percentage of total available energy consumed in the EAH procedure for each group across time. There was a significant interaction between parental weight status group and time (p < 0.001) in the mixed-model analysis, which indicates that there is evidence that the parental weight status groups differed significantly in daughters’ change in disinhibited eating over time. However, analysis of the data omitting age 5 reveals that the interaction term is no longer significant (p = 0.3470), indicating that the interaction in the complete analysis is caused solely by the relative shifting of means from age 5 to age 7. Examination of the relationships in Figure 2, although the lines do cross, most clearly shows a relatively similar pattern over time for the different parent overweight groups. Therefore, we examined the main effects, and there was a significant main effect for parental weight status group (p < 0.01), indicating that there were significant differences among groups in girls’ level of disinhibited eating. Post hoc analyses revealed that girls with both parents overweight consumed a higher percentage of calories during the EAH procedure across all ages than all other groups. The effect of time was significant (p < 0.001), showing that daughters’ disinhibited eating increased over time for all groups. Additionally, post hoc analyses revealed that girls with both parents overweight had larger increases in disinhibited eating over time compared with all other groups (p < 0.05). As was the case with the model predicting girls’ BMI over time, results were unchanged when we used BMI ≥25 kg/m2 to classify overweight in fathers.
Results from this study revealed that during childhood and into early adolescence, girls developing in families differing in parental overweight had divergent trajectories of weight gain and disinhibited overeating. The data provided support for the primary hypothesis that, relative to girls without overweight parents, girls living in families with overweight parents had greater BMI increases across this period. Additionally, daughters’ BMI increase from 5 to 13 years, as well as their risk for becoming overweight, was related to the number of overweight parents. Relative to girls with neither parent overweight, girls who had two overweight parents were 8.1 times more likely to be overweight at age 13. Overweight parents also reported higher levels of disinhibited eating than normal weight parents, and patterns of increase in girls’ disinhibited overeating during this same period of development also differed across parental weight status groups. At age 5, there were no systematic differences in BMI or disinhibited eating among girls, but diverging trajectories for BMI and disinhibited eating produced differences in both outcomes over time. Effects of parent overweight seem to be additive; girls with two overweight parents had the greatest gains in BMI and increases in disinhibited eating from 5 to 13 years of age.
The results of this study corroborate and extend previous studies reporting that having two overweight parents elevates children’s risk for rapid weight gain and becoming overweight during childhood and that having one parent overweight also elevates risk, relative to having neither parent overweight (9,11,16,21). In Figure 1, the BMI data for the four groups are imposed on the Centers for Disease Control and Prevention BMI percentiles growth charts (47) to show the differing patterns of change across groups against the backdrop of the reference data. The mean BMI of girls with neither parent overweight consistently tracks just above the 50th percentile from 5 to 13 years of age. In contrast, the BMI trajectory among girls with both parents overweight crosses percentiles, from 5 to 9 years of age, beginning at the 75th percentile, approaching the 90th percentile by age 9, and remaining at that level at 11 and 13 years of age. The trajectory of upward crossing of BMI percentiles, as shown by the both parent overweight group in this study, has been identified by the American Academy of Pediatrics as indicative of growth patterns placing children at higher risk for childhood overweight and obesity (53). These data, revealing that girls in this group were more than eight times more likely to become overweight, provide additional confirmation on this point.
Previous research has shown that disinhibited eating is consistently related to overweight among adults (29-33) and children (34-36), and these findings provide additional evidence for associations among disinhibited eating and elevated weight status among young girls. In addition, these findings revealed that, over time, disinhibited overeating increased among all groups, suggesting that such increases may be normative. These findings revealed group differences in patterns of change in BMI and differences in disinhibited eating; daughters of two overweight parents had significantly greater increases in BMI, significantly greater increases in disinhibited eating, and greater levels of disinhibited eating than girls with one or neither parent overweight. However, this pattern of association is not sufficient to causally link disinhibited eating to greater increases in BMI, given the lack of an experimental design. We have proposed elsewhere that restrictive parenting practices, which tend to be used by white, middle class parents with daughters who are more overweight, can promote disinhibited overeating. However, influence in parent-child dyads is bi-directional, suggesting that children’s weight status can be both a cause and a consequence of parental feeding practices (54); therefore, these findings raise the possibility of causal links, which must be addressed in future studies.
Mothers’ but not fathers’ disinhibited eating was associated with daughters’ disinhibited eating, but these links between mother’s and daughters’ disinhibited eating did not emerge until later childhood (ages 9, 11, and 13), perhaps because daughters’ disinhibited eating was lowest at age 5 and increased among all groups from age 5 to 13. Although associations between daughters’ and mothers’ disinhibited eating have been previously reported (55,56), relationships between fathers’ and daughters’ have not been evaluated. Both Jacobi et al. (56) and Cutting et al. (55) found a stronger association between mothers’ and daughters’ eating behavior than fathers’ and daughters’ eating behavior. These findings reveal that mothers and fathers are both influential in shaping daughters’ patterns of BMI change and that the development of disinhibited eating among girls was distinctly different only among families in which both parents were overweight. These findings are consistent with an additive model of parental influence, with both mothers’ and fathers’ genetic and environmental input contributing to daughters’ patterns of weight gain and the development of disinhibited eating, although perhaps in different ways. Our findings are consistent with relatively extensive findings showing maternal influence on daughters’ eating and weight status (31,38,55,57-62); there are very few data regarding fathers’ influence.
Research is needed to determine the mechanisms of intergenerational transmission of eating style from parent to child. This sample was selected to include only families in which daughters were living with both biological families. In such cases, parents both provide genes and shape the environments in which these genetic predispositions may be expressed. For example, there is experimental evidence that overeating can be fostered in children by experience with restrictive feeding environments (49) and that maternal modeling may be important (55). As indicated above, there is some evidence that disinhibited overeating has a genetic basis and can serve as a behavioral phenotype for obesity (37,63). These findings are consistent with this perspective and suggest how gene environment correlations shape the differential emergence of disinhibited eating among children growing up in families with and without overweight parents. For example, if overweight parents are genetically predisposed to disinhibited eating, and children share those genetic predispositions, parental displays of disinhibited eating can foster the acquisition of disinhibited eating by children.
Although there is evidence for the heritability of disinhibited eating behavior (63,64), de Castro et al. (64) found that 40% of the variance in disinhibited eating behavior was explained by a shared environmental component. Provencher et al. (63) found that only a small amount of the familial resemblance in eating behavior was caused by genetics; family environment was particularly important for dietary disinhibition. A recent study by Faith et al. (37) tested the hypothesis that our measure of disinhibited eating, EAH, was a genetic marker for childhood obesity. They found some support for this hypothesis among preschool-aged children, but only in boys and not girls. They concluded that eating in the absence of hunger might be more highly influenced by environmental as opposed to genetic factors. In this study, consistent with Faith et al., we failed to note initial associations between daughters’ disinhibited eating and weight status when girls were age 5, when levels of disinhibited eating were lowest among girls. However, our results reveal that parallels in patterns of change for girls’ BMI and disinhibited eating emerged as development proceeded, providing some evidence for disinhibited eating as a behavioral phenotype.
Study limitations include a sample of primarily middle-income, well-educated, and non-Hispanic white families with daughters, and only families in which daughters were living with both biological parents. This precludes generalizing the findings to boys, to single parent families, or to other ethnic and income groups, including those with overweight prevalence rates that are higher than this sample. To classify parents, we used parental BMI, based on measured heights and weights. We also used BMI as our primary measure of daughter overweight. BMI is not a direct measure of adiposity and may have led to misclassification of some parents; however, such misclassification would tend to reduce the likelihood of obtaining relationships among parental weight status and daughters’ weight gain. Data obtained from DXA on girls’ fat mass corroborated the BMI findings among daughters (data not shown). The use of longitudinal data is a positive feature, as was the inclusion of fathers in the research, because it allowed us to provide some initial information regarding paternal influence on daughters’ developing weight status and eating style during childhood and early adolescence.
In conclusion, although no epidemiologic data are available on the prevalence of overweight specifically among parents in the United States, among 20- to 54-year-old adults, an age range that would include most parents, about two thirds are overweight. These estimates suggest that the majority of children in the United States today are living in families with one or two overweight parents; only a small minority of children are growing up in families with neither parent overweight. These findings suggest that children in families with overweight parents are at elevated risk for obesity and that this demographic pattern may further accelerate the obesity epidemic. These findings indicate that, at least for girls, the effects of parental overweight on weight gain and disinhibited eating seem to be additive; girls growing up in families with at least one overweight parent showed accelerated patterns of weight gain from age 5 to 13 years relative to girls with neither parent overweight and that girls with two overweight parents show even more substantial weight gain over the same period.
The findings of this study corroborate previous research (10) indicating the elevated risk for childhood obesity conferred by having overweight parents and underscores the importance of developing prevention approaches that target overweight parents with young children. To inform such prevention efforts, research is needed to pinpoint the environmental and behavioral mediators of family resemblances in adiposity. Such efforts would include exploring the feasibility of approaches to prevent the transmission of disinhibited eating from parents to children. One approach could focus on reducing disinhibited overeating behaviors among parents to reduce children’s exposure to parental models of disinhibited overeating. Alternative approaches include providing anticipatory guidance for parents regarding approaches to parenting and child feeding that can promote children’s continued responsiveness to hunger and satiety cues in controlling energy intake and providing alternatives to parents’ use of feeding practices that can foster disinhibited overeating among children (38).
The services provided by the General Clinical Research Center of the Pennsylvania State University were appreciated. This research was supported by NIH Grants HD32973 and M01 RR10732.
The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.