Association Between Family Divorce and Children's BMI and Meal Patterns: The GENDAI Study
The aim of this work was to explore the associations between family factors, including divorce, and children's overweight as well as eating and physical activity patterns in a population-based sample of healthy school-aged children. In this cross-sectional study, 1,138 children (53% girls; age: 11.2 ± 0.7 years) from elementary schools in the Attica region participated. Their parents provided sociodemographic information, including their marital status. Overweight status classification was based on weight and height measurements and BMI evaluation. Children completed a physical activity checklist and a questionnaire on meal patterns and eating behaviors. The Eating Style score was calculated: the higher the score, the more frequent a child was engaged in less-structured feeding practices promoting food intake for reasons other than hunger. Analysis revealed significant association between family divorce and children's overweight: compared with children of married parents, those of divorced had significantly higher BMI levels (20.0 ± 3.6 kg/m2 vs. 21.3 ± 3.4 kg/m2, respectively, P = 0.007). Controlling for socioeconomic and physical activity factors, divorce remains a significant predictor of a higher BMI, along with older age, higher father's and mother's BMI, less children in the family, and more minutes of daily screen time. Children who had experienced a divorce in their family also reported higher Eating Style score, even after adjusting for potential confounders. In conclusion, in this sample of fifth and sixth graders, unfavorable family circumstances have been associated with children's overweight, as well as with aspects of their eating behavior, namely eating style in relation to conditions around food consumption and hunger, independent of other socioeconomic factors.
Discouraging reports all over the world describe the increase in the prevalence of childhood obesity and the need for research to explore contributing factors as well as effective interventions (1,2,3,4,5). Obesity phenotype lies on the interaction between gene-transmitted susceptibilities and environmental influences, such as abundance of energy-dense foods and changes in physical activity patterns (6). Apart from the genetic predisposition, family affects eating and physical activity habits through a number of factors, which were identified as potential predictors of childhood obesity in either cross-sectional or prospective studies. The most widely investigated factors are parental education, occupation, obesity and age, family income, number of siblings, home cognitive environment, and parental beliefs and practices, but the evidence so far has not been consistent with regard to the type or degree of the associations (7,8,9,10,11,12,13,14,15,16,17). The variability in the results may be related to the population under investigation, the definition of overweight, or the measures used; furthermore, mutual controlling for confounding variables (i.e., measures of socioeconomic status) may lessen or remove the significance of the findings.
One of the family features that has not been fully explored is the marital status of the parents. Some studies found that single maternal marital status unfavorably affects child's body weight (18,19,20,21,22), whereas some other did not (23,24,25). Strauss and Knight (10) reported that children who lived with single mothers were significantly more likely to develop obesity after a 6-year follow-up, compared to those with married parents; however, after adjusting for potential socioeconomic, cognitive, and emotional confounders, the risk of developing obesity became nonsignificant. Similarly, family structure in childhood (single vs. two biological parents, or biological plus step parent vs. two biological parents) had no significant effect on the likelihood of obesity in young adulthood, when confounders, such as BMI and age at examination in childhood, sex and social variables, were included in the model (14). Nevertheless, in all the above-mentioned studies, the role of family divorce has not been considered. Single-parent families were taken as an entity, without differentiating between unmarried and divorced single parents. Previous research has suggested that parental marital conflict and dissolution may increase the risk for adverse health consequences both in childhood and adulthood (26). Proposed biopsychosocial pathways accounting for this association include inadequate parental practices, child's emotional insecurity, and unfavorable changes in socioeconomic status.
In this work, we sought to explore the association between family factors, including divorce and other socioeconomic variables, and children's overweight as well as eating and physical activity practices in a population-based sample of healthy school-aged children.
Methods and Procedures
The Gene-Diet Attica Investigation (GENDAI) is a school-based cross-sectional study, designed to evaluate the pivotal interactions of genetic and environmental variables on children's overweight. The target population was children attending fifth and sixth grade, living in the Attica region of Greece. Several areas within the Attica region were selected in order to cover a wide range of neighborhoods with various socioeconomic levels. Before enrollment, registered fifth and sixth graders of participating schools and their respective parents or guardians were fully informed about objectives and methods of the study, understood that study participation was voluntary, and were assured about the data confidentiality. Parents or guardians of interested students signed an informed consent form, while participating children provided their verbal assent. The research study was approved by the Ethics Committee of Harokopio University and the Greek Ministry of Education. Of the 3,124 children invited to join the study, 1,207 agreed to participate but only 1,138 children participated finally (53% girls; mean age: 11.2 ± 0.7 years). Anthropometric, biological, and lifestyle characteristics were collected by trained personnel (paediatricians and dietitians), on the basis of a standard protocol. Details on the aims, methods, and design of the GENDAI Study have been published elsewhere (27).
Demographic information and clinical assessment
Parents or guardians of the participating children were asked to fill in a questionnaire on sociodemographic characteristics, namely marital status, annual average household income, age, origin, years of schooling and occupation of each parent, number of people permanently living at home, and family living space. With regard to marital status, parents were asked to classify themselves as “unmarried”, “married”, “divorced”, or “widowed.” Occupation was evaluated using a 4-point scale from mainly manual to mental work. For simplicity, we classified the educational level of the patients into three groups: group I: <9 years of school education, group II: up to high school or technical colleges (10–14 years), and group III: university degree. Self-reports of parental height and weight were also recorded to calculate their BMI values.
Assessment of meal patterns, dieting, and eating style
Children completed a questionnaire on their meal patterns and eating behaviors: The frequency of meal consumption was evaluated by the questions, “How often do you have breakfast?” “How often do you have full breakfast (more than a glass of milk or fruit juice)?”, “How often do you usually have lunch?”, and “How often do you usually have supper?”, with possible responses as “never”, “1–2 times/week”, “3–4 times/week”, “5–6 times/week”, or “everyday.” In addition to the main meals, there was a question for snacking, “How many times a day do you usually have a snack?”, with responses ranging from 0 to 6 or more times per day.
Eating style in relation to conditions around food consumption and hunger was also assessed. In particular, feeding practices related to each of the “five onlys”—eating only in the dining room, only while sitting, only from a proper plate, only when doing nothing else, and only when hungry—were assessed using 5-point scales, from always (=1) to never (=5), as described previously (28). Children's self-reports in each item were added in order to calculate a total Eating Style score: the higher the score, the more frequent a child was engaged in less-structured feeding practices promoting food intake for reasons other than hunger. Finally, they were asked about involvement in slimming diets, either on the time of study or in the past.
Anthropometry and physical activity
Physical measurements of body weight and height were obtained in light clothing without shoes. BMI, calculated as weight/height2 (kg/m2), was used for participants' classification as normal weight, overweight, or obese, according to the cutoff points adopted by the International Obesity Task Force (IOTF) (29). Students also completed a physical activity checklist recall for two nonconsecutive days (30). This instrument inquired about the time the children spent on mild, moderate, and strenuous exercise plus total screen time (viewing TV and playing computer/video games) during the previous 24 h.
Continuous variables are presented as mean values ± s.d., and categorical variables as absolute and relative frequencies. Associations between categorical variables were tested using the χ2 test. Multiple regression analysis was applied in order to evaluate the explanatory ability of various characteristics of the participants in relation to the investigated outcome, after adjusting for potential confounders. The results from the regression models are presented as standardized beta coefficients (β). All reported P values were based on two-sided tests. A P value <0.05 was considered statistically significant. Statistical calculations were performed using the Statistical Package for Social Sciences software 14.0 (SPSS, Chicago, IL).
Descriptive sociodemographic characteristics of the study participants according to their BMI status are presented in Table 1. No association was detected between parental education and children's overweight, either in the whole sample or in the analysis by sex. On the contrary, parental BMI, number of siblings, and family living space were significantly different between overweight and normal weight children. In particular, overweight children lived in 94 ± 27 m2 family living space, while normal weight ones in 98 ± 30 m2 (P = 0.010). Furthermore, 55.9% of overweight children reported to reside in <92 m2 living space (below the median), whereas the corresponding value for normal weight children was 46.1% (x 2 = 9.49, P = 0.002).
Table 1. Descriptive characteristics of the study participants by BMI status (mean (s.d.) or frequency)
A significant association was found between parents' divorce and obesity indexes of children. A higher frequency of family divorce was observed among overweight children compared with normal weight ones (10.8% vs. 7.3%, P = 0.045). ANOVA revealed that children of divorced parents had significantly higher BMI values (21.3 ± 3.4 kg/m2 vs. 20.0 ± 3.6 kg/m2, respectively, P = 0.007) than those of married parents; no similar differences were found in children of unmarried (19.6 ± 3.2 kg/m2, P = 0.393) or widowed parents (20.3 ± 3.3 kg/m2, P = 0.722), even in a multi-adjusted model, after controlling for child's age and sex, maternal BMI, parental education, and family living space. Therefore, in the subsequent analyses, children of divorced parents were compared with those of nondivorced parents, including those who were unmarried and widowed. Prevalence of child's overweight was 47% among divorced families and 37% among nondivorced families. Divorced families were also reported to have lower mean family income, lower family living space, less family cars, and a lower level of maternal education (Table 2).
Table 2. Sociodemographic, anthropometric, and biochemical features of children according to family divorce (mean (s.d.) or frequency)
Although an association was observed between family divorce and overweight, residual confounding might exist. For instance, differences in age, sex, socioeconomic factors, and physical activity status may explain the observed, unadjusted findings. Thus, in order to control for the aforementioned confounders, multiple regression analysis was performed with child's BMI as a dependent variable. It was revealed that a higher BMI was associated with older age (β = 0.394, P = 0.037), higher father's (β = 0.182, P < 0.001) and mother's BMI (β = 0.228, P < 0.001), less children in the family (β = −0.760, P < 0.001), more minutes of daily screen time (β = 0.003, P = 0.041), and having divorced parents (β = 1.280, P = 0.012). Variables, such as sex, maternal or paternal education or age, family living space, or income as well as minutes of everyday physical activities were not significantly associated with BMI. When the two sexes were analyzed separately, the significant associations remained the same.
In addition, we evaluated the effect of the divorce on eating and physical activity habits (Table 3). Children who had experienced a divorce in their family reported more frequently to engage on a slimming diet compared with those in the rest of sample (27.4% vs. 18.7%, P = 0.045); their total score in the Eating Style questionnaire, assessing the conditions around eating, was also significantly higher (9.7 ± 3.0 vs. 9.0 ± 2.5, P = 0.045). After controlling for potential confounding variables, namely parental BMI, family living space, and mean annual income, the difference in the Eating Style total score between children who experienced a divorce in their family and those who did not remained significant (respective adjusted means 9.8 ± 0.4 vs. 8.9 ± 0.1, P = 0.038). Analysis of the individual questionnaire items revealed that children from divorced families eat less frequently while sitting (P = 0.050) or in the dining room (P = 0.037); 56.8% of them responded that they usually eat only when hungry, whereas the corresponding frequency in the rest of the sample was 66.8% (P = 0.109). A trend has also been observed for children of divorced families to spend more time viewing TV and playing computer/video game, compared with their counterparts from nondivorced families (141 ± 96 and 126 ± 88 min/day, respectively, P = 0.136).
Table 3. Frequencies of eating behaviors of children according to family divorce (mean (s.d.) or frequency)
Our data indicate that family-related factors, namely divorce, parental BMI, number of siblings, and daily screen time, significantly predicted children's BMI at the age of 9–11 years. To the best of our knowledge, this is the first report that illustrates the potential role of divorce on weight status, but also on meal patterns and eating behavior of children. We found that such unfavorable family circumstances were associated with child's overweight; in particular, children of divorced parents had 6% higher BMI values compared with those from nondivorced parents, independent of other socioeconomic factors.
Many studies have investigated the effect of divorce on children's health, but it is not clear whether the effects are mainly attributed to the marital conflict or to the marital dissolution (26). For the association between divorce and overweight, a number of potential mediators may be suggested. Given the fact that behavioral problems sometimes precede the onset of overweight, one could hypothesize that children's adjustment problems following parent's divorce could mediate this association. Parental conflict and departure threaten children's sense of emotional security (31). Still, children from divorced families, compared with those from nondivorced families, are more likely to exhibit externalizing behaviors (noncompliance, conduct disorders, lack of self regulation) and internalizing disorders (depression, anxiety), to have lower self-esteem and more academic problems, and to be less socially responsive and competent (32,33,34). Lumeng et al. (23) have found that clinically meaningful behavioral problems, such as those described earlier, have been associated with an increased risk of becoming overweight in previously normal weight children 8–11-year-old, whereas psychosocial problems have been related to an increase in body weight of children between 7 and 10 years of age (35).
Diminished parental well-being and distress following marital dissolution could also be responsible for an increase in children's weight. Feelings, such as anger, anxiety, depression, loneliness, and impulsivity may emerge after a divorce (36,37,38). Parental mental health affects children's adjustment and health through diminished competence in parenting: after divorce, parenting is characterized by irritability and coercion, increased restrictiveness and diminished communication, affection, consistency, control, and monitoring (37,38,39,40). With regard to weight problems, parental neglect during childhood was found to predict a great risk of obesity in young adulthood (14) and, therefore, it may mediate a potential association between marital dissolution and children's weight changes. Ineffective parenting and dysfunctional family climate have been reported in families with obese children (41). Furthermore, mothers with an authoritarian, i.e., strict disciplinarian, parenting style or mothers with permissive and neglectful parenting styles are more likely to have overweight children compared with mothers with an authoritative parenting style, who respect children’ opinion but maintain clear boundaries (42).
A more permissive parenting style could also be related to our findings on the conditions around eating, such as less frequent eating while sitting and eating in the dining room as well as to the trend observed for the children of divorced families to spend more time on viewing TV and playing computer/video games. TV viewing, a risk factor for overweight, as revealed in this and other studies (43,44), has been previously reported to be at increased levels among children from divorced families (45).
Income decline after a divorce (46) could also contribute to weight changes. Gray et al. (22) have recently found that family income was a strong predictor of children's overweight, which mediated the effect of other demographic and family structure variables (i.e., married vs. nonmarried parents). Our results do not support this hypothesis; although several statistically significant differences were detected between divorced and nondivorced parents with regard to socioeconomic variables, family divorce remained a significant predictor of children's BMI in the multivariate model, adjusting for potential confounding factors.
Not surprisingly, parents' BMI was a significant predictor of child's BMI, after controlling for socioeconomic variables. The predictive power that parental obesity holds on the development and maintenance of childhood and adolescent obesity has been repeatedly reported (10,47,48,49,). Apart from the genetic inheritance, a number of behavioral patterns within the family have been recognized as linking parent and child weight status (51). In particular, parental weight status, related to parents' own eating patterns and their child-feeding practices, including food preferences and energy regulation, was found to influence children's eating behaviors and, consequently, overweight status.
Our findings on the association between number of siblings and overweight support similar limited evidence in children (7,9,52). Patterson et al. (9) reported that, in preadolescent girls, the odds ratio for obesity decreased by 14% for each additional sibling in the household. On the other hand, Wang et al. (52) found that the association of sib-size with the risk of overweight differs between boys and girls: in boys, a U-shaped relationship was found, with those from three-child families exhibiting the lowest risk, whereas in girls a strong negative dose-response relationship. Our analysis did not reveal any sex-specific difference on the effect of number of siblings on children's BMI.
Among the limitations of this work, we acknowledge that it constitutes a cross-sectional study. Even though we carefully adjusted for known and potential confounders, residual confounding may always exist in observational research. As this type of research is not appropriate for establishing cause-effect relationship or elucidating mechanisms, the hypotheses described here need to be assessed through future longitudinal or intervention studies. Furthermore, assessment of marital status and other socioeconomic variables was based on demographic information provided by one of the children's caretakers. In order to enhance sincerity of the reports, all questionnaires were precoded and no name of either child or parent appeared on it. Still, the information provided by the caretakers was not further explored, in order to clarify potential issues of cohabitation or separation without a divorce as well as type of the divorce (estranged or continued contact). Our results were solely based on their self-report of marital status. Similarly, parents' BMI values were calculated from one parent's report. Obesity status may have been misreported, as both men and women tend to underestimate weight and BMI and overestimate height (53,54); however, the variability in the estimates is high and no overall effect size could be estimated. Finally, physical activity and eating patterns were also self-reported as they were evaluated through the relevant questionnaires. Although the accuracy and validity of these self-reported methods in school-aged children has been questioned in previous studies (55,56,57), convenience reasons imposed by the population nature of our investigation did not allow us to use more accurate methods, such as the observation of meal patterns and eating styles.
In conclusion, in this sample of fifth- and sixth-graders from the Attica region, parental divorce was associated with children's BMI as well as with aspects of their eating behavior, namely eating style in relation to conditions around food consumption and hunger. Further research is needed to explore the underlying pathways. Psychometric data, as well as longitudinal information on child's body weight trajectory and timing of divorce, would provide better understanding on the type and direction of association between occurrence of divorce, psychological factors, body weight, and eating changes. Although preliminary, our results suggest that therapeutic measures for children's overweight should carefully recognize family-related factors, as well as the degree of their potential effect, and include as many as possible mediating behaviors in the intervention programs.
The study was supported by a research grant from Coca-Cola Hellas. We thank the following field investigators for their assistance in physical examination and nutritional and biochemical assessment: Nikoletta Vidra, Anastasia-Eleni Farmaki, Maria Tziirkalli, Evagelia Evagelidaki, Garifallia Kapravelou, Ioanna Kontele, Katerina Skenderi, Nektarios Lainakis, and Sofia Sakka.
The authors declared no conflict of interest.