Objective. To determine whether there is a relationship between clinically meaningful behavior problems and concurrent and future overweight in 8- to 11-year-old children.
Methods. 1998 National Longitudinal Survey of Youth interview data for 8- to 11-year-old children and their mothers were analyzed. A Behavior Problems Index score >90th percentile was considered clinically meaningful. Child overweight was defined as a body mass index (BMI) ≥95th percentile for age and sex. Multiple logistic regression was used to control for potential confounders (selected a priori): child’s sex, race, use of behavior-modifying medication, history of academic retention, and hours of television per day; maternal obesity, smoking status, marital status, education, and depressive symptoms; family poverty status; and Home Observation for Measurement of the Environment-Short Form (HOME-SF) cognitive stimulation score. In an attempt to elucidate temporal sequence, a second analysis was conducted with a subsample of normal-weight children who became overweight between 1996 and 1998 while controlling for BMI z score in 1996.
Results. The sample included 755 mother-child pairs. Of the potential confounding variables, race, maternal obesity, academic grade retention, maternal education, poverty status, and HOME-SF cognitive stimulation score acted as joint confounders, altering the relationship between behavior problems and overweight in the multiple logistic regression model. With these covariates in the final model, behavior problems were independently associated with concurrent child overweight (adjusted odds ratio: 2.95; 95% confidence interval: 1.34–6.49). The relationship was strengthened in the subsample of previously normal-weight children, with race, maternal obesity, HOME-SF cognitive stimulation score, and 1996 BMI z score acting as confounders (adjusted odds ratio: 5.23; 95% confidence interval: 1.37–19.9).
Conclusions. Clinically meaningful behavior problems in 8- to 11-year-old children were independently associated with an increased risk of concurrent overweight and becoming overweight in previously normal-weight children.
Childhood overweight is a growing public health problem. More than 20% of children between the ages of 6 and 11 years are overweight,1 and the severity of overweight has increased in the past 10 years.2 Well-described risk factors for childhood overweight include race,2 elevated maternal body mass index (BMI),3,4 and amount of television viewing.5 The literature diverges on whether low socioeconomic status is a risk factor.4,6,7
Several risk factors for childhood overweight, including television viewing8 and low socioeconomic status,9–12 are also associated with increased rates of behavior problems. However, the relationship of behavior problems and childhood overweight has not been explored thoroughly. Previous studies have shown an association between behavior problems and childhood overweight.13–20 Interpretation of these data has been limited, however, by methodologic constraints, including small and homogeneous samples, limited use of standardized behavior rating scales, and bias as a result of samples’ being drawn only from weight-loss clinics. Individuals who attend weight-loss clinics have higher rates of psychological problems than overweight individuals who do not attend such a clinic.21,22 Furthermore, the previous studies have focused on statistical differences in behavior rating scales, as opposed to clinically meaningful behavioral differences. No study has controlled adequately for potentially confounding variables such as television viewing, socioeconomic status, or quality of the home environment. Finally, recent data are lacking; the most pertinent studies were conducted at least 10 years ago.13–16
To address issues raised by previous research, the current analysis investigates the relationship between clinically meaningful behavior problems in 8- to 11-year-old children and childhood overweight in a nationally representative, nonreferred population. In addition, in an attempt to explore the temporal sequence of this association, we conducted a second analysis investigating the relationship between clinically meaningful behavior problems in normal-weight children and becoming overweight 2 years later.
The National Longitudinal Survey of Youth (NLSY) is a multipurpose survey sponsored by the US Department of Labor. It originally included >12 600 individuals, ages 14 to 21 years, who have been interviewed annually since 1979. Since 1986, biennial assessments have been administered to the children of women in the original cohort. We included children who were between 8 years, 0 months and 11 years, 0 months at the 1998 survey. This age range was selected because overweight prevalence significantly increases between ages 6 and 11 years compared with ages 5 years and younger.7,23 Given the underlying hypothesis that behavior problems precede the onset of overweight, we elected to assess the association beginning at 8 years of age so that the potential effect of behavior problems having their onset during the grade-school years could be measured. Of the original NLSY cohort, 4944 women were eligible to be surveyed in 1998. Of these, 87% (4301) were interviewed. Approximately 82% (3533) of the women interviewed were mothers. Response rates differed by <5% among major racial groups for both the mother and the child assessments.24 The eligible age range included 1416 children, and 90% (1268) of this group had weight and height data. We limited the sample to children with complete data for all covariates. The final sample, therefore, consisted of 755 children. This study was approved by the Institutional Review Board at Boston University School of Medicine.
Demographic data consisted of child’s sex, race, mother’s marital status, mother’s education, and family poverty status (Table 1). Additional characteristics that may modify an association between behavior problems and overweight were identified from the literature and included as potential covariates (Table 1). These were mother’s obesity, mother’s depressive symptoms, Home Observation for Measurement of the Environment-Short Form (HOME-SF) cognitive stimulation score, mother’s smoking status, use of behavior-modifying medication, hours of television per day, and history of academic grade retention.
Height and weight were measured by an in-home interviewer in 81% and 78% of subjects, respectively. For the remainder of subjects, parental report was used. There was no statistically significant difference between measured and reported height and weight data (data available on request). Child overweight was defined as a BMI ≥ 95th percentile for age and sex based on the newest norms from the National Center for Health Statistics.25
Mothers answered questions regarding their child’s behaviors in the previous 3 months using the Behavior Problems Index (BPI). The BPI was created by Zill and Peterson26 to measure the frequency, range, and type of childhood behavior problems for children ages 4 years and older. Many items were derived from the Achenbach Child Behavior Checklist.27 Mothers could answer “often,” “sometimes true,” or “not true” to each item. Responses were dichotomized into “often or sometimes true” or “not true” by the NLSY. These dichotomized scores are the basis for the age- and sex-based normed total standard scores provided by the NLSY. Externalizing and internalizing standard scores were created by the NLSY on the basis of the trichotomous “often,” “sometimes true,” or “not true” rating. Norms for externalizing and internalizing standard scores are age based but not sex based. Some questions in the BPI contribute to both the internalizing subscore and the externalizing subscore. The BPI is also divided into 6 subscales: antisocial, anxious/depressed, dependent, headstrong, hyperactive, and peer conflicts/withdrawn. The specific questions included in the BPI are provided in Table 2. Higher scores represent more behavior problems. The overall completion rate of the BPI was approximately 93%, with Hispanic children having slightly lower levels of completion.24 We categorized total BPI standard scores as clinically meaningful when they were >90th percentile for the entire cohort of 8- to 11-year-old children in the NLSY in 1996 and in 1998. We selected this cutoff on the basis of previous data indicating that BPI scores at this level are associated with higher rates of referrals for mental health services.9 The >90th percentile cutoff was also used to categorize internalizing and externalizing standard scores as clinically meaningful.
All data analysis was done using SAS (SAS Institute Inc, Cary, NC). The survey oversampled children who were born to poor and minority women. Therefore, we weighted the data by using the sampling weights given for 1998 so that the sample is nationally representative of children among the same age cohort. Descriptive statistics are provided as unweighted frequencies to demonstrate the actual composition of the sample. Clustering caused by the presence of siblings was accounted for by using generalized linear models fit via generalized estimating equations.28
We performed multiple logistic regression accounting for clustering with children’s overweight status as the dependent variable. We computed both unadjusted and adjusted odds ratios (ORs) and their corresponding 95% confidence intervals (CIs) from the logistic regression models. All P values presented are 2-tailed. An α level of 0.05 was used to determine statistical significance. The covariates described in Table 1 were included in a model. Each of these candidate covariates was tested in a model with behavior problems and overweight by excluding them one by one while keeping all others in the model. Covariates that altered the relationship of clinically meaningful behavior problems to child overweight in these models by >10% were retained in the final model.29 To determine whether internalizing or externalizing behaviors were specifically accounting for the relationship of behavior problems to childhood overweight, internalizing and externalizing BPI scores >90th percentile were each tested individually in the model. In an attempt to elucidate temporal sequence, the analysis was repeated using a subsample of children who were not overweight in 1996, using 1996 BPI scores, 1996 BMI z scores, and 1998 data for the remaining covariates.
To account for missing data, we created a model to determine whether selection bias altered the relationship between significant behavior problems and being overweight. We examined the influence of potential selection bias by adding a propensity score to model the probability of inclusion in the sample with complete data.30 Our estimation of the association between behavior problems and child overweight was not altered by the inclusion of this variable.
The 1998 sample included 755 mother-child pairs. The ages of the children included in the analysis were 8 years, 0 months to 11 years, 0 months. Approximately half of the sample was male, 12% were overweight, and 10% had a BPI score >90th percentile (Table 3). The average unweighted standard scores for the BPI subscales in overweight versus not overweight children were similar although consistently slightly greater in the overweight children (Table 4).
Behavior Problems and Concurrent Overweight
A relationship between clinically meaningful behavior problems and overweight was present in the bivariate analysis (OR: 2.36; 95% CI: 1.13–4.91). Other variables found to be associated with overweight in the bivariate analysis were race, maternal obesity, and maternal education (Table 5).
A relationship between clinically meaningful behavior problems and overweight persisted in the adjusted analysis (OR: 3.08; 95% CI: 1.30–7.29). Although maternal education was no longer significant, race and maternal obesity continued to be independently associated with an increased risk of overweight in children (Table 5). In the analysis restricted to include only those covariates that altered the relationship between behavior problems and overweight by >10%, joint confounding was evident. Joint confounding occurs when the simultaneous control of 2 or more variables gives different results from those obtained by controlling for each variable separately. Therefore, race, maternal obesity, academic grade retention, maternal education, poverty status, and HOME-SF cognitive stimulation score all were retained in the model. In this final model, clinically meaningful behavior problems were independently associated with concurrent child overweight (adjusted OR: 2.95; 95% CI: 1.34–6.49; Table 6).
We tested for the presence of interactions between the covariates by adding terms to the multiple logistic regression model. We examined interactions of race by significant behavior problems, by maternal obesity, by academic grade retention, by maternal education, by poverty status, by maternal depressive symptoms, and by HOME-SF cognitive stimulation score; maternal obesity by significant behavior problems, by academic grade retention, by maternal education, by poverty status, by maternal depressive symptoms, and by HOME-SF cognitive stimulation score; academic grade retention by significant behavior problems, by maternal education, by poverty status, by maternal depressive symptoms, and by HOME-SF cognitive stimulation score; maternal education by significant behavior problems, by poverty status, by maternal depressive symptoms, and by HOME-SF cognitive stimulation score; family poverty status by significant behavior problems, by maternal depressive symptoms, and by HOME-SF cognitive stimulation score; and maternal depressive symptoms by significant behavior problems and HOME-SF cognitive stimulation score. We could not evaluate HOME-SF cognitive stimulation score by significant behavior problems as we did not have the power to do so. None of the interactions evaluated was statistically significant. Although a total BPI score >90th percentile was related to overweight, neither an externalizing nor an internalizing subscale score >90th percentile was significantly related to overweight.
Unlike previous research with the NLSY,4,31 there was no independent relationship between the number of hours of television watched per day or cognitive stimulation in the home environment and childhood overweight. To determine whether the absence of a relationship was attributable to controlling for behavior problems, we removed the clinically meaningful behavior problems variable from the model adjusted for all covariates that is listed in Table 5. There was no significant change in the relationship of either number of hours of television per day (OR: 1.00; 95% CI: 0.93–1.07) or low cognitive stimulation in the home environment (OR: 1.13; 95% CI: 0.28–4.60) with childhood overweight.
Behavior Problems in Normal-Weight Children and Future Overweight
Of the 755 children in the 1998 sample, 746 were interviewed in 1996 and 700 had height and weight data. Of these 700 children, 90% were not overweight in 1996, resulting in a final sample size of 629. Six percent of the 629 children who were not overweight in 1996 became overweight by 1998. This subsample of 629 children was similar to the original sample: 9% had clinically meaningful behavior problems, 7% of the mothers were obese, 10% of the families were living in poverty, 4% of the children were taking medication to modify behavior, and 9% had repeated a grade (additional data available on request).
When this subsample was entered into the model, a relationship of significant behavior problems in normal-weight children in 1996 to becoming overweight 2 years later was present in the unadjusted analysis (OR: 2.87; 95% CI: 1.06–7.80). We again entered the same candidate covariates (1998 data) and the BMI z score from 1996 into the model one by one and retained those that altered the relationship between behavior problems and future overweight by >10%. The only covariates that altered the relationship in this manner were race, maternal obesity, HOME-SF cognitive stimulation score, and BMI z score in 1996. After adjustment for these covariates, clinically significant behavior problems in normal-weight children in 1996 were independently associated with becoming overweight 2 years later (adjusted OR: 5.23; 95% CI: 1.37–19.93; Table 7).
We found in a nationally representative sample that clinically meaningful behavior problems in 8- to 11-year-old children were independently associated with concurrent child overweight and, in a separate analysis, with an increased risk of becoming overweight 2 years later in normal-weight children. The magnitude of the relationship was similar to that of several previously recognized risk factors, including maternal obesity and race.
To our knowledge, our study is the first to investigate systematically the relationship of a clinically meaningful level of behavior problems and childhood overweight in a nationally representative sample while analytically controlling for a large number of potentially confounding factors. Three previous studies demonstrated marginally increased scores on the Child Behavior Checklist in overweight children in non-clinic-referred populations.15,18,19 The number of overweight children in these studies, however, varied from 25 to 30, and the samples were overwhelmingly minority (83%–100%) and of low socioeconomic status. The only study that used a large sample that we are aware of demonstrated an increased teacher report of “behavioral problems” in association with an increased rate of weight gain in 5399 Swedish schoolchildren in the 1970s.14 Demographic characteristics of this sample were neither described nor controlled for in the analysis.
Maternal obesity and race have strong independent associations with child overweight, both in our analyses and in others.3,4,7 Controlling for these factors did not explain or diminish the independent relationship between behavior problems and overweight. We hypothesized that poverty would be a significant confounding factor, given the relationship of socioeconomic status to behavior problems.9 Although poverty did conjointly modify the relationship between behavior problems and overweight in combination with race, maternal obesity, maternal education, academic grade retention, and HOME-SF cognitive stimulation score, the effect was relatively small. Unlike previous work with the NLSY data set,31 the amount of television viewing did not independently predict child overweight in our analysis or act as a confounder. Controlling for behavior problems also did not explain the absence of an association. This may reflect that the children in our sample are slightly younger than those in the previous study, and the association between television viewing and overweight is less clear in younger children.5,32 The discrepancy may also reflect different definitions of outcome. We used the definition for “overweight” (BMI ≥95th percentile) compared with the previous study, which used the less stringent definition of “at risk for overweight” (BMI ≥85th percentile).
We also did not replicate the findings of a previous study demonstrating that low cognitive stimulation in the home is independently associated with childhood overweight,4 and this was also not explained by controlling for behavior problems. The absence of an independent association may be because the previous study included children initially ages birth to 8 years and evaluated the cumulative incidence of overweight 6 years later. The age range in our study was older, and the follow-up was not as long. Low cognitive stimulation in the home did, however, act as a joint confounder in our analysis, slightly diminishing the relationship between behavior problems and overweight. This observation supports the previous data indicating a significant role of the home environment in mediating the child’s weight status and speaks to the complexity of the association. The differences in findings in both the case of television viewing and cognitive stimulation in the home may also reflect different confounders controlled for in each analysis.
Although a relationship between a BPI score >90th percentile and overweight was present, when we attempted to disaggregate internalizing and externalizing behavior problems, neither was related to overweight. These results suggest that of the children with clinically significant behavior problems, there was not a large proportion with either an internalizing standard score >90th percentile or a large proportion with an externalizing score >90th percentile. Rather, it seems that the majority of children with BPI scores >90th percentile had subthreshold internalizing and externalizing BPI scores. This is consistent with our clinical experience and with the data presented in Table 4. The overweight children had slightly higher scores on all subscales; there were not particular subscales that seemed to be accounting for the relationship.
The results of our analysis of the temporal relationship between child behavior problems and overweight suggest that behavior problems may precede the onset of overweight at least in some children. Controlling for the child’s previous weight status only strengthened the association between behavior problems and subsequent overweight, which supports our underlying hypothesis. Given that it seems that behavior problems sometimes precede becoming overweight, the higher prevalence of behavior problems in overweight children may not then simply be a response to stigmatization.
Previous research has demonstrated a relationship between childhood depression and adult overweight33 in addition to a relationship between adolescent depression and an increased risk of overweight 1 year later.34 Although the BPI cannot make mental health diagnoses, behavior problems are sometimes indicative of underlying mental health diagnoses.9 A significant change in weight is a diagnostic criterion for depression,35 and food binging behavior has been reported as comorbid with psychiatric diagnoses in adults.36–40 Thus, just as a mental health diagnosis in childhood may be a risk factor for adult obesity, extreme behavior problems in childhood may be a more proximal indicator of overweight risk in childhood.
There are potential limitations to the study reflecting residual confounding and study design. Our measure of maternal mental health was available only in 1992. There are no data, to our knowledge, regarding the stability of maternal depressive symptoms over a period of 6 years. It is possible that a more proximal measure of maternal depressive symptoms would have slightly altered our results. We also were unable to control for paternal obesity, which is a risk factor for child overweight of similar magnitude to the risk of maternal obesity.3 There may also be other confounding variables for which we were unable to control. Finally, our sample size of overweight children was limited. Unlike other studies with limited sample size, however, our sample was nationally representative.
These data strongly suggest a significant relationship of behavior problems to childhood overweight. On the basis of our second analysis, it seems that children with behavior problems may be at increased risk of becoming overweight, suggesting that the clinical evaluation of children with significant behavioral problems should include an assessment of overweight risk based on physical activity and nutrition. It should be emphasized that most overweight children in our study did not have behavior problems. However, a better understanding of a child’s behavioral profile may be helpful in addressing the motivational issues that pediatricians perceive to be a major barrier to treatment41 of childhood overweight. Additional research is needed to understand how a child’s behavioral profile may be associated with physical activity and eating patterns leading to overweight.
This study was supported by HRSA MCHB 5T77 MC00015 09, the American Heart Association Fellow-to-Faculty Transition Award, and the Joel and Barbara Alpert Endowment for Children of the City.
We thank Howard Bauchner, MD, for thoughtful review of the manuscript.
- ↵Strauss R, Knight J. Influence of the home environment on the development of obesity in children. Pediatrics.1999;103(6) . Available at: http://www.pediatrics.org/cgi/content/full/103/6/e85
- ↵Dietz W, Gortmaker S. Do we fatten our children at the television set? Obesity and television viewing in children and adolescents. Pediatrics.1985;75 :807– 812
- ↵Troiano R, Flegal K. Overweight children and adolescents: description, epidemiology, and demographics. Pediatrics.1998;101(suppl 3) :497– 504
- ↵Gortmaker S, Walker D, Weitzman M, Sobol A. Chronic conditions, socioeconomic risks, and behavioral problems in children and adolescents. Pediatrics.1990;85 :267– 276
- Fergusson D, Horwood L, Shannon F. Relationship of family life events, maternal depression, and child-rearing problems. Pediatrics.1984;73 :773– 776
- ↵Clarizo H, McCoy G. Behavior Disorders in School-Aged Children. Scranton, PA: Chandler; 1970
- ↵Mellbin T, Vuille J. Further evidence of an association between psychosocial problems and increase in relative weight between 7 and 10 years of age. Acta Paediatr Scand.1989;78 :574– 580
- ↵Israel A, Shapiro L. Behavior problems of obese children enrolling in a weight reduction program. J Pediatr Psychol.1985;10 :449– 460
- ↵Pine D, Wasserman G, Coplan J, Staghezza-Jaramillo B. Cardiac profile and disruptive behavior in boys at risk for delinquency. Psychosom Med.1996;58 :342– 353
- ↵Strauss C, Smith K, Frame C, Forehand R. Personal and interpersonal characteristics associated with childhood obesity. J Pediatr Psychol.1985;10 :337– 343
- ↵Ogden C, Troiano R, Briefel R, Kuczmarski R, Flegal K, Johnson C. Prevalence of overweight among preschool children in the United States, 1971 through 1994. Pediatrics.1997;99(4) . Available at: http://www.pediatrics.org/cgi/content/full/99/4/e1
- ↵Center for Human Resource Research. NLSY79: 1998 Child and Young Adult Data Users Guide, Columbus, OH: The Ohio State University; 2002
- ↵Zill N, Peterson J. Behavior Problems Index. Washington, DC: Child Trends, Inc; 1986
- ↵Achenbach T, Edelbrock C. Manual for the Child Behavior Checklist and Revised Child Behavior Profile. Burlington, VT: University of Vermont, Department of Psychology; 1983
- ↵Diggle P, Liang K, Zeger S. Analysis of Longitudinal Data. Oxford, UK: Clarendon Press; 1994
- ↵Rothman K, Greenland S. Modern Epidemiology. Philadelphia, PA: Lippincott-Raven; 1998
- ↵DuRant R, Baranowski T, Johnson M, Thompson W. The relationship among television watching, physical activity, and body composition of young children. Pediatrics.1994;94 :449– 454
- ↵Pine D, Goldstein B, Wolk S, Weissman M. The association between childhood depression and adulthood body mass index. Pediatrics.2001;107 :1049– 1056
- ↵Goodman E, Whitaker R. A prospective study of the role of depression in the development and persistence of adolescent obesity. Pediatrics.2002;110 :497– 504
- ↵First M. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. Washington, DC: American Psychiatric Association; 1994
- Zwaan MD. Binge eating disorder and obesity. Int J Obes.2001;1 :S51– S55
- ↵Fergusson D, Horwood L, Gretton M, Shannon F. Family life events, maternal depression, and maternal and teacher descriptions of child behavior. Pediatrics.1985;75 :30– 35
- ↵Radloff L. The CES-D Scale: a self-report depression scale for research in the general population. J Appl Psychol Measurement.1977;1 :385– 401
- ↵Stunkard A, Albaum J. The accuracy of self-reported weights. Am J Clin Nutr.1981;34 :1593– 1599
- ↵World Health Organization. Measuring Obesity: Classification and Description of Anthropometric Data. Report on a WHO consultation on the epidemiology of obesity. Copenhagen, Denmark: WHO Regional Office for Europe, Nutrition Unit; 1988
- ↵Caldwell B, Bradley R. Home Observation for Measurement of the Environment. Little Rock, AR: University of Arkansas at Little Rock, Center for Child Development and Education; 1984
- ↵Weitzman M, Gortmaker S. Maternal smoking and behavior problems of children. Pediatrics.1992;90 :342– 349
- ↵Logan S, Spencer N. Smoking and other health related behaviour in the social and environmental context. Arch Dis Child.1996;74 :176– 179
- ↵Barkley R, McMurray M, Edelbrock C, Robbins K. Side effects of methylphenidate in children with attention deficit hyperactivity disorder: a systemic, placebo-controlled evaluation. Pediatrics.1990;86 :184– 192
- Copyright © 2003 by the American Academy of Pediatrics