Obesity and Psychiatric Disorder: Developmental Trajectories

Obesity and Psychiatric Disorder: Developmental Trajectories Sarah Mustillo, Carol Worthman, Alaattin Erkanli, Gordon Keeler, Adrian Angold and E. Jane Costello Pediatrics 2003;111;851-859 DOI: 10.1542/peds.111.4.851

The online version of this article, along with updated information and services, is located on the World Wide Web at: http://www.pediatrics.org/cgi/content/full/111/4/851

PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly publication, it has been published continuously since 1948. PEDIATRICS is owned, published, and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk Grove Village, Illinois, 60007. Copyright © 2003 by the American Academy of Pediatrics. All rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.

Downloaded from www.pediatrics.org at Duke University on September 25, 2007

Obesity and Psychiatric Disorder: Developmental Trajectories Sarah Mustillo, PhD*; Carol Worthman, PhD‡; Alaattin Erkanli, PhD*; Gordon Keeler, MS*; Adrian Angold, MRCPsych*; and E. Jane Costello, PhD* ABSTRACT. Objectives. To identify age-related trajectories of obesity from childhood into adolescence, and to test the association of these trajectories with the development of psychiatric disorders (conduct disorder, oppositional defiant disorder, attention-deficit/hyperactivity disorder, substance abuse, depression, and anxiety). Methods. White children (N ⴝ 991) 9 to 16 years old from the Great Smoky Mountains Study, a representative sample of rural youth, were evaluated annually over an 8-year period for height, weight, psychiatric disorder, and vulnerabilities for psychiatric disorder. Longitudinal analyses on the repeated measures data were conducted using developmental trajectory models and generalized estimating equation models. Results. Obesity was 3 to 4 times more common than expected from national rates using Centers for Disease Control and Prevention 2000 criteria. Four developmental trajectories of obesity were found: no obesity (73%), chronic obesity (15%), childhood obesity (5%), and adolescent obesity (7%). Only chronic obesity was associated with psychiatric disorder: oppositional defiant disorder in boys and girls and depressive disorders in boys. Conclusions. In a general population sample studied longitudinally, chronic obesity was associated with psychopathology. Pediatrics 2003;111:851– 859; obesity, psychopathology, developmental trajectories, depression, conduct disorder. ABBREVIATIONS. GSMS, Great Smoky Mountains Study; BMI, body mass index; NHANES, National Health Interview Survey; CAPA, Child and Adolescent Psychiatric Assessment; ADHD, attention-deficit/hyperactivity disorder; SPMM, semiparametric mixture model; GEE, Generalized Estimation Equation; CDC, Centers for Disease Control and Prevention.

O

besity among children and adolescents is a serious public health issue in the United States. The percentage of children defined as overweight has more than doubled since the 1970s and is still on the rise.1 Many studies have documented a secular trend in childhood obesity that is associated with increased health risks, both during childhood and later in adulthood.2– 4 During childhood, obesity is associated with neurologic, pulmonary, gastrointestinal, circulatory, and endocrine From the *Center for Developmental Epidemiology, Department of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Durham, North Carolina; and ‡Department of Anthropology, Emory University, Atlanta, Georgia. Received for publication May 22, 2002; accepted Oct 3, 2002. Address correspondence to Sarah Mustillo, PhD, Center for Developmental Epidemiology, Duke University Medical Center, Box 3454, Durham, NC 27710. E-mail: [email protected] PEDIATRICS (ISSN 0031 4005). Copyright © 2003 by the American Academy of Pediatrics.

conditions.5,6 In adulthood, a history of childhood obesity is associated with an increased risk for cardiovascular disease, diabetes, colon cancer, and allcause mortality.5,7 Overweight and obesity, or perceptions thereof, may also affect self-esteem, body image, and social mobility,5,8 and it is well-known that adults who are overweight or obese are at an increased risk for psychological disorders.9,10 Less is known about whether childhood obesity is associated with concurrent psychopathology, or with specific types of psychiatric disorder. Most of the published research on obesity and psychopathology in children has used clinical samples recruited because of obesity11–14 or depression.15 The risk of psychiatric disorder in obese children in these referred samples may not reflect what is seen in the general population. Among the structural, familial, and environmental vulnerabilities associated with both obesity and psychopathology in childhood are being from a poor16,17 or single-parent household,17,18 having harsh or abusive parents,17,19,20 and, for some psychiatric disorders, being female.21 However, there is little research that establishes the direction of effect, or the role of these factors as mediators or moderators of the link between obesity and psychopathology. Weight fluctuations are common in childhood, and even more so in adolescence.22 Because most studies of childhood obesity and psychopathology are crosssectional or retrospective, we do not know whether there are distinct patterns of obesity throughout childhood, nor do we know whether certain patterns are associated with particular outcomes. This study provides repeated measures of height and weight taken annually over 8 years, together with annual information about family characteristics and mental health. Longitudinal data permit identification of distinct developmental trajectories of obesity, separating the transiently obese from the chronically obese, and those whose obesity is confined to childhood or adolescence. The goals of this study are 1) to determine the number and type of distinct obesity trajectories in a general population sample of white children 9 to 16 years old; 2) to examine whether trajectory membership is associated with family vulnerabilities such as gender, poverty, single-parent family, parenting style, parental history of mental illness or drug abuse, and life events; and 3) to test the relationship between obesity trajectory membership and psychopathology, controlling for a range of potential mediators.

PEDIATRICS Vol. 111 No. 4 April 2003 Downloaded from www.pediatrics.org at Duke University on September 25, 2007

851

METHODS Study Design and Sample A full description of the setting, sample, and data collection methods can be found elsewhere.23 The Great Smoky Mountains Study (GSMS) is a longitudinal study of the development of psychiatric disorder and need for mental health services in rural youth. A representative sample of 3 cohorts of children, 9, 11, and 13 years old at intake was recruited from 11 counties in western North Carolina. Potential participants were selected from the population of some 20 000 children using a household equal probability, accelerated cohort design.24 The accelerated cohort design means that over several years of data collection each cohort reaches a given age in a different year, which controls for cohort effects.25 We used a 2-phase process to select the final sample for the longitudinal study. A screening questionnaire was administered to a parent (usually the mother) of the first stage random sample (N ⫽ 3896; 95% of those contacted). The questionnaire consisted mainly of the externalizing (behavioral) problems scale of the Child Behavior Checklist,26 and was administered by telephone or in person. All children scoring above a predetermined cut point (the top 25% of the total scores, in this case it was a score of 20) plus a 1-in-10 random sample of the remaining 75%, were recruited for detailed interviews. Eighty percent of those recruited agreed to participate. The contribution of each participant was weighted by the inverse of their selection probabilities, stratified by age and gender, to provide accurate prevalence estimates for the population of the study area.27 Families were reinterviewed annually until the child was 16, and every 2 to 3 years thereafter. The data presented here, based on the first 8 annual waves of the study (1993–2000), consist of 4600 interviews with 991 non-Hispanic white participants and their parents. Fewer than 10% of the area residents and the sample were African American. Because racial/ethnic differences in adiposity and its correlates are well-documented,28,29 non-white children were excluded from these analyses as power was not adequate for comparisons. The study protocol was approved by the Duke University School of Medicine Institutional Review Board.

Procedures Two interviewers visited the family each year, either at home or in a location convenient for them. Before the interviews began, parent and child signed informed consent forms. They were then interviewed in separate rooms. Each parent and child was paid $10 after the interview.

Measures Obesity Height and weight were both assessed at 2 different points during the interview, and the 2 averaged for these analyses. Participants wore normal clothing without shoes, socks, and belts. Height was measured to the nearest 0.1 cm using a stadiometer (CMS Weighing Equipment, London, United Kingdom) and standard techniques.30 The measurement was repeated if the first 2 measurements differed by ⬎0.5 cm, and the nearest 2 of the resultant 3 were averaged. Participants were weighed twice to the nearest 0.1 kg on a portable digital scale (Soehnle, Murrhardt, Germany); a repeat measure was taken if the first 2 differed by ⬎0.2 kg, and averaged as for height. Body mass index (BMI) was calculated using the standard wt/ht2 formula. For adults, obesity is generally defined as BMI ⬎30 kg/m2; but no equivalent standard exists for children.30,31 In this study we use age- and sex-specific 95th percentiles from the revised (2000) growth charts of the Centers for Disease Control and Prevention (CDC), which are based on data from the National Health Interview Survey (NHANES).32 The reference values are based on US national survey data and are intended for US children and adolescents.33 Because the GSMS children were weighed clothed, and the NHANES children were weighed wearing only a gown, undergarments, and foam slippers, we subtracted 1 kg from weight before calculating the BMI.

Psychopathology The psychopathology outcomes were assessed using the Child and Adolescent Psychiatric Assessment (CAPA), a psychiatric inter-

852

view for children 9 to 17 years old. The CAPA is an interviewerbased interview.34 The goal of interviews using this format is to combine the advantages of clinical interviews with those of highly structured epidemiologic interview methods.35 While using a highly structured format of questions and probes, the interviewerbased approach trains the interviewer to ensure that the parent or child being interviewed understands the construct under review, and provides enough detail and examples for a clear rating of the clinical severity of each symptom to be made. A detailed glossary provides the operational rules for identifying a clinically significant level of severity for each symptom. The CAPA interviews parent and child separately, using different interviewers. The presence of a symptom can be determined on the basis of information from a single respondent, or using the either/or rule common in clinical practice. For these analyses, we counted a symptom as present if reported by either parent or child or both. The time frame of the CAPA for determining the presence of most psychiatric symptoms is the past 3 months. Diagnoses were based on Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria.36 For these analyses we combined psychiatric diagnoses into 7 groups: conduct disorder, oppositional defiant disorder, depressive disorders (including major depression, dysthymia, and depression not otherwise specified), anxiety disorders (including separation anxiety, generalized anxiety disorder, simple phobia, social phobia, agoraphobia, and panic), bulimia, substance abuse, and attention-deficit/hyperactivity disorder (ADHD).

Background and Family Variables Potential mediators and moderators of the relationship between obesity and psychopathology included in these analyses were gender; family income (coded 0 ⫽ $0 –$25 000, 1 ⫽ $25 001– $45 000, and 2 ⫽ 45 001⫹); 1 or both parents with less than an 11th-grade education; single parent family; parental history of treated mental illness, drug abuse, or criminal conviction; harsh or overprotective parenting style; lax supervision; and traumatic life events.

Season We included in the trajectory models a variable to account for the season when the child was weighed and measured. All participants were interviewed each year on a day as close as possible to their birthday. Small seasonal variations in weight gain have been noted in 2 European samples,37,38 and studies relying on self-reports of weight have found that healthy adults believe their weight increases by ⬃5 pounds in the winter,39 – 41 although clinical research has failed to find support for seasonal weight gain.41 There is little literature on seasonal effects on children’s weight. Despite the lack of consensus on seasonal weight variation in the literature, we noted a substantial seasonal weight variation in our sample. On average, 29% of children measured in the winter months (January, February, and March) were obese, compared with 20% during the remaining months, with a mean BMI difference of 1. To correct for this, we included a 4-category variable for season (January to March, April to June, July to September, and October to December) in the trajectory model as a time-dependent covariate.

Interviewers and Interviewer Training Interviewers were residents of the area in which the study took place. All had at least bachelor’s level degrees. They received 1 month of training and constant quality control, maintained by postinterview reviews of each schedule, notes, and tape recordings by experienced interviewer supervisors and study faculty. Interviewers were trained by Department of Social Services staff in the States requirements for reporting abuse or neglect.

Data Management and Analysis Scoring programs for the CAPA, written in SAS software (SAS Institute, Inc, Cary, NC), combine information about the date of onset, duration, and intensity of each event and symptom to create scale scores and diagnoses. Prevalence rates were calculated using Stata 7.0 (Stata Corporation, College Station, TX), taking the survey design into account through weights and clustering. Significant differences were calculated using a design-based Pearson ␹2 with a second-order correction converted to an F statistic.

OBESITY AND PSYCHIATRIC DISORDER: DEVELOPMENTAL TRAJECTORIES Downloaded from www.pediatrics.org at Duke University on September 25, 2007

TABLE 1.

Mean BMI by Age for Boys and Girls in GSMS

Age

Girls

Boys

9 10 11 12 13 14 15 16

18.0 18.7 19.2 20.1 21.2 22.0 22.9 23.6

17.8 19.5 19.1 19.6 20.9 21.6 23.2 24.3

Obesity trajectories were determined by fitting a semiparametric mixture model (SPMM) to the data, using PROC TRAJ in SAS.42 SPMMs identify distinct groups of individual trajectories within the population. This approach to modeling growth curves is different from traditional latent growth curve modeling in that the latter assumes the random parameters to be bivariate normal distributed. In other words, all individuals belong to a single class of individuals who vary continuously on a latent trait. In contrast, the group-based method employed here assumes a number of discrete classes, each having a specific intercept and age slope and an estimated population prevalence.42 Because we suspected that individuals do not vary continuously on obesity, but rather that there are a distinct number of obesity trajectories, the group-based method was the most appropriate. The logit model was used to model the presence or absence of obesity predicted by age. By specifying the sampling weights, we invoked the robust variance estimator (ie, sandwich type estimator) to adjust the standard errors of the parameter estimates to account for the 2-phase sampling design. Thus, the classes that were identified, and their estimated prevalences and correlates, relate to the entire population of white children in the 11-county area from which participants were sampled. Children were classified into their most probable obesity trajectory class by the use of posterior probabilities. A posterior probability is the probability of each individual belonging to each

group.42 Thus, individuals are assigned to the group to which they have the highest probability of belonging. Once children were assigned to a trajectory class, we conducted analyses in 3 stages. First, we computed bivariate statistics to obtain prevalence rates of various background, familial, and psychopathological characteristics by obesity trajectory. An F test for each variable tested whether there was any difference among obesity groups. Second, we fit multinomial logit models to determine how the background and familial characteristics identified as significant in the first stage distinguished children on the different obesity trajectories from the comparison group of never-obese children. Third, we fit Generalized Estimation Equation (GEE) models to estimate the relationship between the different obesity trajectories and psychiatric disorders, controlling for the background factors identified in the second stage. The GEE models were run using binary regression with logit link function for dichotomous outcomes, such as the presence or absence of a psychiatric diagnosis, using the Stata program xtgee. The use of multiwave data with the appropriate sample weights capitalized on the availability of multiple observation points over time, while controlling for the effect on variance estimates of repeated measures on the same child, for overlapping cohorts, and for design effects.

RESULTS Obesity and Age

Table 1 shows the association between age and mean BMI for boys and girls. Gender differences in mean BMI were small and insignificant. BMI rose steadily through late childhood and adolescence, beginning at ⬃18 for 9 year olds, and rising to ⬃20 by age 12, 22 by age 14, and 24 by age 16. Fig 1 presents the overall prevalence of obesity, using the criterion of weight in the top 95th percentile for age and sex according to the 2000 CDC tables. Obesity rose between 9 and 10 years, dipped be-

Fig 1. Percent obese by age.


ARTICLES

853

tween 10 and 11 years, flattened out between 11 and 13 years, then rose again steadily to 16 years, at which point ⬃20% were obese. Generally, the prevalence of obesity increased with age. These results show that this rural, Southern sample was substantially heavier than the general population of children and adolescents in the United States, based on the CDC criteria, as 20% of the sample was above the 95th percentile by 16 years. We estimated SPMMs that allowed for 1, 2, 3, 4, and 5 classes of obesity over time. Improved fits were obtained from 1 to 2, from 2 to 3 classes, and from 3 to 4 classes. However, no improvement was obtained when 5 classes were allowed. The Bayesian Information Criterion supported the 4-class model as the best fitting. Thus, the analyses presented here are based on the 4-class model (Fig 2), with individuals being assigned to their most likely obesity class using posterior probabilities. Of the 4 trajectories of obesity, 1 consisted of children who were never or rarely obese. Children who fell into this category were not obese in any of the waves, or only in 1 wave. This group included 72.8% of the population. The second trajectory included the 7.5% of children who started off in the normal weight range during childhood but became obese over time (adolescent obesity). This curve was relatively flat until 12 years, at which point the prevalence of obesity rose sharply. The third trajectory consists of chronically obese children who were obese for most or all waves of observation. The chronically obese trajectory included 14.6% of children. The fourth trajectory, including 5.1% of the population, consisted of those who were obese during late childhood, and whose weight fell during adolescence (childhood obesity). Differences Among Obesity Trajectories

Table 2 shows the prevalence of a range of child and family characteristics in each obesity group. There were no overall differences among the groups in gender, family structure, parenting style, family history of mental illness, drug abuse, crime, or trau-

matic events. Childhood and chronic obesity, but not adolescent obesity, were associated with having uneducated parents and lower income. Bivariate analyses showed that childhood and chronically obese children were significantly more likely than nonobese children to have depression and oppositional defiant disorder. There was no significant difference among the trajectories in the proportion with anxiety disorders, conduct disorder, ADHD, or substance abuse disorders. Predictors of Obesity Trajectory Membership

We fit multinomial regression models to determine which factors predicted group membership (Table 3). Taking the never obese as the referent group, chronically obese children were significantly more likely to be in the lower and middle income groups, and to have uneducated parents. The childhood obesity group was also more likely to have uneducated parents. There were no other significant differences between the never-obese group and children on any of the other trajectories. Multinomial models run separately for boys and girls showed that adolescent-onset obese boys were less likely to be lower income, compared with the never obese. Chronically obese boys were more likely than never obese boys to have families with middle incomes, and to have uneducated parents. Compared with the never obese group, boys in the childhood-limited obesity group did not show any excess risk factors, nor did any of the factors examined predict chronic obesity among the girls. Uneducated parents were more common among girls with childhood-limited obesity than among the never obese. The pseudo R squared was small in all of the multinomial models, indicating that the familial and environmental variables tested did not explain a substantial portion of the variation in trajectory membership. Obesity Trajectories and Psychopathology

We tested for increased risk of any of 7 psychiatric disorders in the 3 obesity groups, relative to the

Figure 2. Four obesity trajectories.

854


TABLE 2.

Sociodemographic, Family Characteristics, and Psychiatric Disorders in 4 Obesity Trajectories Obesity Trajectories

N ⫽ 991 Male Income ⱕ$25,000 $25,001–$45,000 $45,001⫹ Uneducated parents Single parent Inadequate supervision Harsh discipline Overprotective Parent treated for MH problems Parent treated for drug problems Parent convicted of crime Violence between parents Loss events Violent events Adverse ongoing events Network events BMI Bulimia Any depression Any anxiety Oppositional defiant disorder Conduct disorder Substance use ADHD

F

Never Obese

Adolescent Obese

Obese

Childhood Obese

72.8% 47.7%

7.5% 60.5%

14.6% 55.3%

5.1% 64.7%

29.9% 32.1% 38.1% 13.1% 20.4% 6.7% 1.9% 7.0% 27.1% 6.4% 18.7% 1.8% 15.5% 15.6% 13.6% 12.0% 19.6 .0% 2.2% 2.4% 3.2% 2.1% 18.3% .9%

20.3% 42.2% 37.6% 16.4% 23.5% 9.0% 40.0% 4.0% 20.5% 2.5% 17.5% 60.0% 10.4% 15.2% 13.8% 1.6% 23.9 .0% 3.0% 1.1% 3.7% 1.4% 16.0% .9%

39.4% 41.6% 19.0% 29.7% 22.7% 9.2% 3.4% 1.1% 29.4% 8.8% 18.9% 1.5% 15.7% 19.3% 13.2% 13.1% 29.4 .4% 4.0% 2.8% 6.5% 3.4% 18.7% 1.3%

40.1% 27.3% 32.6% 32.0% 24.4% 10.9% 1.4% 2.1% 25.3% 8.5% 31.4% 2.4% 9.3% 19.7% 21.1% 16.0% 25.6 1.5% 2.0% 2.6% 7.5% 2.0% 21.3% 1.3%

P Value

1.14

.33

2.40

.03*

6.21 .18 1.14 2.98 1.48 .39 1.15 1.45 .41 1.72 .62 .78 .57

.00** .90 .33 .06 .22 .76 .32 .23 .68 .16 .60 .50 .63

2.56 3.40 .61 4.51 .94 .20 .38

.10 .04* .58 .01** .38 .89 .76

MH indicates mental health. * P ⱕ .05. ** P ⱕ .01.

nonobese group. Age, sex, and income were included in each model, as well as the other psychiatric disorders, to control for comorbidity. Oppositional disorder was more common in chronically obese boys and girls (odds ratio: 2.5; 95% CI: 1.36 – 4.61) and depression in chronically obese boys (odds ratio: 3.7; 95% CI: 1.27–10.2) but not girls. There were no significant associations between obesity trajectory membership and bulimia, ADHD, substance use, conduct disorder, or anxiety, controlling for comorbid psychiatric disorders. As shown in Table 3, the risk of both oppositional disorder (both sexes) and depression in boys were increased in youth with low incomes, and the risk of oppositional disorder was increased in youth with middle incomes as well. Testing for mediational effects, we found no mediational effects of income or any other family variables to explain the link between obesity trajectories and psychopathology. DISCUSSION

This study explores developmental aspects of obesity in a representative, predominantly rural, population of white non-Hispanic children between 9 and 16 years. The sample used for these analyses is quite small compared with the several national surveys on which data are available (reviewed in Ref. 31), and is restricted to white non-Hispanic children. However, it has 2 advantages: 1) repeated annual measurements on the same subjects, using careful measurement of height and weight (not self-report), and 2) measurement of a wide range of risk factors and

correlates, including state-of-the-art assessment for psychiatric disorder. Although 73% of the study participants were never obese across the 8 years of observation, ⬎22% of the young people were either chronically obese (15%) or had become obese by 16 years (7%). A smaller group (5%) started obese as children and then dropped below the obesity threshold by adolescence. Defining obesity, as we did, using the CDC 2000 formula which classifies as obese children in the top 5% of the age-matched NHANES population sample, between 16% and 24% of our sample were obese at any 1 of the 8 measurement points. Our data are consistent with both the widely noted secular trend toward an increasing prevalence of early obesity,3 and an increased risk of obesity with increased age.3 There is also evidence that children in the Southern states are heavier; for example, using the 1993 CDC growth curves, Strauss and Pollack1 found that in the Southern states the proportion of children who were “overweight” (above the 95th percentile) had increased to 17.1% by 1998, from 7.6% in 1968. In our sample, the prevalence of obesity was higher among the boys than the girls, although the difference was not significant. Several recent studies found obesity to be more prevalent among boys than girls,43– 45 with rates of obesity increasing faster in boys than in girls.1 However, gender patterns may vary by race; therefore, any gender differences may depend on the race of the sample. Gordon-Larsen et al44 examined prevalence of overweight and obesity by sex and race from the National Longitudinal Sur-


ARTICLES

855

TABLE 3.

Comparison with Never Obese Trajectory: Multinomial Regression

Independent Variables

Adolescent Obesity*

Chronic Obesity*

Childhood Obesity*

Relative Risk Ratios (Robust Standard Error) Male Low income† Middle income† Parental style Parental history Uneducated parents Life events Wald ␹2 Pseudo R2

1.72 (.85) .67 (.30) 1.27 (.49) 1.12 (.35) .74 (.13) 1.61 (.78) 1.37 (.28) 54.4§ .04

1.28 (.37) 2.16 (.74)‡ 2.31 (.72)§ 1.24 (.23) .92 (.09) 2.34 (.73)§ 1.01 (.15)

1.95 (1.13) 1.05 (.73) .80 (.50) 1.25 (.25) 1.09 (.15) 3.20 (1.68)‡ .80 (.20)

Girls only Low income† Middle income† Parental style Parental history Uneducated parents Life events Wald ␹2 Pseudo R2

1.11 (.81) .91 (.77) 1.18 (.57) .89 (.18) 2.37 (1.65) 1.44 (.50) 88.5§ .05

2.25 (1.14) 1.79 (.86) .80 (.23) .90 (.11) 2.08 (.97) 1.29 (.30)

1.39 (1.11) 1.73 (1.02) 1.12 (.24) 1.16 (.24) 7.44 (4.39)§ 1.39 (.29)

Boys only Low income† Middle income† Parental style Parental history Uneducated parents Life events Wald ␹2 Pseudo R2

.37 (.18)‡ 1.53 (.60) 1.08 (.44) .62 (.16) 1.10 (.69) 1.25 (.33) 47.5§ .05

2.08 (.93) 2.92 (1.10)§ 1.75 (.40) .95 (.16) 2.51 (1.06)‡ .80 (.15)

.93 (.83) .56 (.46) 1.27 (.41) 1.08 (.15) 1.70 (1.32) .50 (.16)‡

* Reference group is Group 1—Never/Rarely Obese. † Reference group is high income. ‡ P ⱕ .05. § P ⱕ .01.

vey of Adolescent Health and found that boys had a higher prevalence of overweight and obesity than girls among whites and Asians, but girls had a higher prevalence among Hispanics and blacks. Familial and environmental risk factors for chronic obesity were surprisingly few; of the ⬎20 examined, only income and lack of parental education remained significant in multivariable analyses. Children with chronic or childhood obesity were more than twice as likely as other children to come from families where 1 or both parents left school before the 11th grade. Lack of parental education remained significant after controlling for having a single, unemployed, or teenage parent, and other correlated risk factors (Table 3). An inverse relationship between income and BMI has been observed in several studies.46 – 48 The present study found a correlation between BMI and income of between ⫺.20 and ⫺.25 across each of the 8 annual assessments. However, this link between income and overweight may be specific to white children in the United States,1 and is clearly a phenomenon of developed economies. This study found an inverse relationship between chronic obesity and socioeconomic status in the combined sample, contrary to several previous studies, which find a stronger association among females than males.49 However, in our sample, different types of obesity had different associations with income so the apparent inconsistencies may be attributed to the difference between examining trajectories of obesity and obesity as a monolithic category. 856

Using developmental trajectory modeling, we were able to show that chronic and childhood-only obesity were more common in the lowest income group (annual family income below $25 000) than in the middle or upper income groups. Chronic obesity was also more common than expected in children from the middle income group ($25 000 –$45 000). This inverse relationship between weight and poverty, beginning in childhood and, in the case of the poorest children, persisting into adolescence, reflects earlier studies.16,50 It is particularly troubling in the light of evidence from the 1946 British birth cohort TABLE 4. GEE Models of Association Between Obesity Trajectories and Oppositional Defiance Disorder (Both Sexes) and Depression (Boys) (Semirobust Standard Errors in Parentheses) Independent Variables

Oppositional Defiance Disorder Total Sample

Sex Low income† Middle income† Age Adolescent obese* Chronic obese* Childhood obese* Wald ␹2 n (observations/subjects)

1.31 (.33) 2.85 (.85)§ 2.06 (.66)‡ 1.08 (.06) .94 (.41) 2.57 (.79)§ 2.00 (.91) 24.83§ 4288/983

Depression Boys Only

4.50 (2.23)§ 2.15 (1.08) 1.17 (.11) .24 (.26) 3.72 (1.94)§ 1.68 (1.15) 26.90§ 2374/555

* Reference group is Group 1—Never/Rarely Obese. † Reference group is high income. ‡ P ⬍ .05 § P ⬍ .01.


that the inverse association between socioeconomic status and BMI predicted higher BMI through to middle age48 and was impervious to the child’s later academic or economic achievements. The present study shows first and foremost that there are different types of childhood obesity. Additionally, this study shows that of the 4 basic weight trajectories—never obese, chronically obese, increasing risk, and decreasing risk— only chronic obesity was associated with a statistically significant increase in the risk of psychiatric disorder. The group with the next-highest prevalence of psychiatric disorder was in the small group who were obese in childhood but became less so as they moved through adolescence. Adolescent-onset obesity was not associated with increased risk of psychopathology. Chronically obese children had significantly higher rates of oppositional defiant disorder, and (for boys) depression. This is, to our knowledge, the first study to show an association between chronic obesity and behavioral problems in children and adolescents. Several studies have linked child and adolescent obesity to depression, but the majority of these are based on clinic samples, referred either because of obesity11–14 or depression.15 We cannot rule out the possibility that the co-occurrence of the 2 conditions increased the likelihood of a clinical referral.51 Among epidemiologic studies, the evidence linking obesity and depression is much less clear. The New York longitudinal study52 found an inverse relationship between adolescent depression and young adult obesity in males, and no relationship in females. One study53 argued that the association between overweight and depression may be caused by the effects of dieting, or of poor health. Another54 demonstrated that any association between depression and BMI in a sample of third graders was explained by concern about being overweight. Of course, this may be true across the whole population but not true for the smaller group of obese youth. Some studies have found that obesity is associated with lower levels of psychopathology, as opposed to higher levels.55 Friedman et al55 note that inconsistencies in prior studies of obesity and psychopathology may be attributed to methodological and sampling limitations. Additionally, we would add that inconsistencies in prior findings may be caused, in part, by the failure to consider developmental trajectories. In our sample, being obese at 1 point in time was not associated with an increased risk in psychopathology, rather being in the chronically obese trajectory was. One limitation of this study is that it does not address the issue of causality. Studies such as this beg the question, does obesity increase the risk of psychopathology or does psychopathology increase the risk of obesity? DiPietro et al56 rely on epidemiologic data to examine changes in weight as a function of changes in depressive symptoms and find that depression plays a role in weight change. Younger men (⬍55 years old) who were depressed at baseline gained more weight during the follow-up period than those not depressed, while young women who were depressed at baseline gained less

weight than the nondepressed. In a small prospective case-control study, Pine et al15 found that depression during childhood was positively associated with BMI during adulthood. Conversely, Pine et al52 found an association with current but not prior depression in a large, population-based sample. These inconsistencies between prior depression and current BMI may be due to differences in type of sample, measurement of psychopathology, and the use of self-report versus measured height and weight. In this study, we examine the association between obesity and psychopathology; however, we draw no conclusions about causal ordering. If psychopathology increases the risk of obesity, we would expect an association in the chronic and adolescent-onset groups. If obesity increases the risk of psychopathology, we would expect an association in the chronic and childhood-limited groups. Our findings show that psychopathology is most common in the chronically obese group first, and the childhood-limited group second. Therefore, these results could be suggestive of obesity increasing the risk of psychopathology; however, that conclusion is purely speculative. Future research investigating the sequence of events is necessary to draw any conclusions about causation. It is also possible that rather than 1 increasing the risk of the other, they are both associated with the same physiologic mechanisms. Evidence, largely from adults, links obesity and depression through underlying neuro-endocrine substrates. Specifically, activity of the hypothalamopituitary adrenal axis appears to be altered in both depressed57–59 and obese60,61 adults, although evidence from depressed obese adults is contradictory,62 and little is known about children. The trajectories of obesity identified here may manifest in part sequelae of suboptimal conditions for early development. For example, a growing body of research documents the role of maternofetal stress in later mental and physical health risk. Prenatal stress followed by postnatal overnutrition leads to childhood overweight63 and adult obesity64 as well as adult depression65 and increased hypothalamo-pituitary adrenal activity.66,67 Overall, relevant findings are more suggestive than definitive, leaving etiologic interrelationships between obesity and mood or behavior disorders and their possible common neuroendocrine mediators a subject for further investigation. Although chronic obesity and bulimia were linked in bivariate analyses, there were very few cases of bulimia in this sample, and all but 2 were comorbid for oppositional disorder and/or depression. Interestingly, bulimia was only associated with obesity when it was comorbid with a psychiatric disorder. Bulimia is clearly an important clinical problem, and the fact that in this population of 9 to 16 year olds it was seen almost exclusively in children with other psychiatric disorders argues for vigilance on the part of clinicians to identify these comorbid children early. There is debate about the accuracy of BMI as a measure of the prevalence of obesity.68,69 For example, Frankenfield et al69 compared BMI to percent body fat as measures of obesity, and found that BMI


ARTICLES

857

underidentified obesity. Defining obesity as a BMI ⱖ30 or at least 25% body fat for men and 30% for women, 30% of men and 40% of women who had BMI ⬍30 had obesity-level body fat. In adolescents, however, Malina68 found BMI to be a highly sensitive and specific marker of obesity, if not of overweight. A technical limitation of this study is that the children were weighed and measured clothed except for their shoes, socks, and belts. A constant 1 kg was subtracted from their weight before BMIs were calculated. The seasonal difference in weight observed in this study may have been caused in part by extra clothing worn in the winter, although American homes tend to be kept to a fairly even temperature throughout the year. Another limitation is that although the GEE models used repeated measures, they did not deal with the relative timing of obesity and psychiatric disorder. Thus, we could determine if an association existed between the obesity trajectories and Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition diagnoses, but not whether an association was more likely to occur at a particular age or if the association existed over time.

9.

10. 11.

12.

13.

14. 15.

16. 17.

18. 19.

CONCLUSIONS

Chronic obesity across 8 years from 9 to 16 years of age was observed in almost 15% of a general population sample of rural white youth, 3 times the rate predicted using the most recent CDC criteria. A further 7% became obese in adolescence, while 5% who had been obese in childhood moved out of obesity by adolescence. Chronic obesity was more common in children from poor and less-educated families, and it had implications for mental health. New methods of data analysis that permit the modeling of developmental trajectories offer the opportunity to deepen our understanding of pathways to health and illness across childhood and adolescence.

20.

21. 22.

23.

24. 25.

ACKNOWLEDGMENTS Work contributing to these analyses was supported by MH01167, MH57731, and MH48085 from the National Institute of Mental Health, and DA11301 from the National Institute on Drug Abuse. We are most grateful to the people of western North Carolina for their collaboration in the GSMS, to Dan Nagin and Mike Ezell for their methodological assistance, and to the reviewers for their helpful comments.

26.

27. 28.

29.

REFERENCES 1. Strauss RS, Pollack HA. Epidemic increase in childhood overweight, 1986 –1998. JAMA. 2001;286:2845–2848 2. Gidding SS, Bao W, Srinivasan SR, Berenson GS. Effects of secular trends in obesity on coronary risk factors in children: the Bogalusa Heart Study. J Pediatr. 1995;127:868 – 874 3. Tremblay MS, Willms JD. Secular trends in the body mass index of Canadian children. Can Med Assoc J. 2000;163:1429 –1433 4. Eisenmann JC, Katzmarzyk PT, Arnall DA, Kanuha V, Interpreter C, Malina RM. Growth and overweight of Navajo youth: secular changes from 1955 to 1997. Int J Obes Relat Metab Disord. 2000;24:211–218 5. Must S, Strauss RS. Risks and consequences of childhood and adolescent obesity. Int J Obes Relat Metab Disord. 1999;23:S2–11 6. Deckelbaum RJ, Williams CL. Childhood obesity: the health issue. Obes J. 2001;9:239S–243S 7. Dietz WH. Childhood weight affects adult mortality. J Nutr. 1998;128: 411S– 414S 8. Ge X, Elder GH, Regnerus M, Cox C. Pubertal transitions, perceptions

858

30. 31.

32.

33.

34.

35.

of being overweight, and adolescents’ psychological maladjustment: gender and ethnic differences. Soc Psychol Q. 2001;64:363–375 National Institutes of Health National Heart, Lung, and Blood Institute. Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults. 1998 Stunkard AJ, Wadden TA. Obesity: Theory and Therapy, 2nd ed. New York, NY: Raven Press; 1993 Britz B, Siegfried W, Ziegler A, et al. Rates of psychiatric disorders in a clinical study group of adolescents with extreme obesity and in obese adolescents ascertained via a population based study. Int J Obes. 2000; 24:1707–1714 Zipper E, Vila G, Dabbas M, et al. Obesity in children and adolescents, mental disorders and familial psychopathology. Presse Med. 2001;30: 1489 –1495 Sheslow D, Hassink S, Wallace W, DeLancey E. The relationship between self-esteem and depression in obese children. Ann N Y Acad Sci. 1993;699:289 –291 Csabi G, Tenyi T, Molnar D. Depressive symptoms among obese children. Eat Weight Disord. 2000;5:43– 45 Pine DS, Goldstein RB, Wolk S, Weissman MM. The association between childhood depression and adulthood body mass index. Pediatrics. 2001;107:1049 –1056 Stunkard AJ, Sorensen TIA. Obesity and socioeconomic status—a complex relation. N Engl J Med. 1993;329:1036 –1037 Costello EJ, Keeler GP, Angold A. Poverty, race/ethnicity and psychiatric disorder: a study of rural children. Am J Public Health. 2001;91: 1494 –1498 Racelli GP, Belmont L. Obesity in nineteen year old men: family size and birth order associations. Am J Epidemiol. 1975;109:66 –77 Felitti VJ, Anda RF, Nordenberg D, et al. Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults. Am J Prev Med. 1998;14:245–258 Jacoby A, Altman DG, Cook J, Holland WW, Elliot A. Influence of some social and environmental factors of the nutrient intake and nutritional status of school children. Am J Public Health. 1975;29:1494 –1498 Ambrosius WT, Newman SA, Pratt JH. Rates of change in measures of body size vary by ethnicity and gender. Ethn Dis. 2001;11:303–310 Guo S, Huang C, Maynard L, et al. Body mass index during childhood, adolescence, and young adulthood in relation to adult overweight and adiposity: the Fels Longitudinal Study. Int J Obes Relat Metab Disord. 2000;24:1628 –1635 Costello EJ, Angold A, Burns BJ, et al. The Great Smoky Mountains Study of Youth: goals, designs, methods, and the prevalence of DSMIII-R disorders. Arch Gen Psychiatry. 1996;53:1129 –1136 Schaie KW. A general model for the study of developmental problems. Psychol Bull. 1965;64:92–107 Willett JB, Singer JD, Martin NC. The design and analysis of longitudinal studies of development and psychopathology in context: statistical models and methodological recommendations. Dev Psychopathol. 1998; 10:395– 426 Achenbach TM, Edelbrock CS. Behavorial problems and competencies reported by parents of normal and disturbed children aged four through sixteen. Monogr Soc Res Child Dev. 1981;46:1– 82 Cochran WG. Sampling Techniques. New York, NY: Wiley; 1995 Kimm SY, Barton BA, Obarzanek E, et al. Racial divergence in adiposity during adolscence: the NHLBI growth and health study. Pediatrics. 2001;107(3). Available at: www.pediatrics.org/cgi/content/full/107/ 3/e34 Must A, Gortmaker SL, Dietz WH. Risk factors for obesity in young adults: Hispanics, African Americans and whites in the transition years, age 16 –28 years. Biomed Pharmacother. 1994;48(3– 4):143–56 Frisancho R. Anthropometric Standards for the Assessment of Growth and Nutritional Status. Ann Arbor, MI: University of Michigan Press; 1990 Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide international survey. BMJ. 2000;320:1– 6 Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, et al. CDC Growth Charts: United States advance data from vital and health statistics, no 314. National Center for Health Statistics. 2000 Flegal KM, Ogden CL, Wei R, Kuczmarski RL, Johnson CL. Prevalence of overweight in US children: comparison of US growth charts from the Centers of Disease Control and Prevention with other reference values for body mass index. Am J Clin Nutr. 2001;73:1086 –1093 Angold A, Fisher PW. Interviewer-based interviews. In: Shaffer D, Lucas C, Richters J, eds. Diagnostic Assessment in Child and Adolescent Psychopathology. New York, NY: Guilford Press; 1999:34 – 64 Wing JK, Cooper JE, Sartorius N. Present State Examination (PSE). In: Wing JK, Cooper JE, Sartorius N, eds. The Measurement and Classification


36.

37. 38.

39.

40.

41.

42. 43.

44.

45.

46. 47.

48.

49. 50. 51. 52.

of Psychiatric Symptoms. Cambridge, England: Cambridge University Press; 1974 American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition. Washington, DC: American Psychiatric Press, Inc; 1994 Andersson I, Rossner S. The Christmas factor in obesity therapy. Int J Obes Relat Metab Disord. 1992;16:1013–1015 Van Staveren WA, Seurenberg P, Burema J, DeGroot LC, Hautvast JG. Seasonal variation in food intake, pattern of physical activity and change in body weight in a group of young adult Dutch women consuming self-selected diets. Int J Obes. 1986;10:133–145 Rosenthal N, Genhart M, Jacobsen F, Skwerer R, Wehr T. Disturbances of appetite and weight regulation in seasonal affective disorder. Ann N Y Acad Sci. 1987;499:216 –230 Thompson C, Stinson D, Fernandez M, Fine J, Isaacs G. A comparison of normal, bipolar, and seasonal affective disorder subjects using the Seasonal Pattern Assessment Questionnaire. J Affect Disord. 1988;14: 257–264 Yanovski J, Yanovski S, Sovik K, Nguyen T, O’Neil P, Sebring N. A prospective study of holiday weight gain. N Engl J Med. 2000;342: 861– 867 Nagin DS. Analyzing developmental trajectories: a semiparametric, group-based approach. Psychol Methods. 1999;4:139 –157 Crespo CJ, Smit E, Troiano RP, Bartlett SJ, Macera CA, Andersen RE. Television watching, energy intake, and obesity in US children. Arch Pediatr Adolesc Med. 2001;155:360 –365 Gordon-Larsen P, McMurray RG, Popkin B. Adolescent physical activity and inactivity vary by ethnicity: The National Longitudinal Study of Adolescent Health. J Pediatr. 1999;135:301–306 Gauthier BM, Hickner JM, Ornstein S. High prevalence of overweight children and adolescents in the Practice Partner Research Network. Arch Pediatr Adolesc Med. 2000;154:625– 628 Seidell JC, Flegal KM. Assessing obesity: classification and epidemiology. Br Med Bull. 1997;53:238 –252 Goodman E. The role of socioeconomic status gradients in explaining differences in US adolescents’ health. Am J Public Health. 1999;89: 1522–1528 Hardy R, Wadworth M, Kuh D. The influence of childhood weight and socioeconomic status on change in adult body mass index in a British national birth cohort. Int J Obes. 2000;24:725–734 Sobal J, Stunkard AJ. Socioeconomic status and obesity: a review of the literature. Psychol Bull. 1989;105:260 –275 Braddon FE, Rodgers B, Wadsworth ME, Davies JM. Onset of obesity in a 36 year birth cohort study. Br Med J Clin Res Ed. 1986;293:299 –303 Berkson J. Limitations of the application of fourfold table analysis to hospital data. Biometrics Bull. 1946;2:47–52 Pine DS, Cohen P, Brook J, Coplan JD. Psychiatric symptoms in adolescence as predictors of obesity in early adulthood: a longitudinal study. Am J Public Health. 1997;87:1303–1310

53. Ross CE. Overweight and depression. J Health Soc Behav. 1994;35:63–79 54. Erickson SJ, Robinson TN, Haydel KF, Killen JD. Are overweight children unhappy? Body mass index, depressive symptoms, and overweight concerns in elementary school children. Arch Pediatr Adolesc Med. 2000;154:931–935 55. Friedman MA, Brownell KD. Psychological correlates of obesity: moving to the next research generation. Psychol Bull. 1995;117:3–20 56. Di Pietro L, Anda RF, Williamson DF, Stunkard AJ. Depressive symptoms and weight change in a national cohort of adults. Int J Obes. 1992;16:745–753 57. Rubin RT, Phillips JJ, McCracken JT, Sadow TF. Adrenal gland volume in major depression: relationship to basal and stimulated pituitaryadrenal cortical axis function. Biol Psychiatry. 1996;40:89 –97 58. Yehuda R, Teicher MH, Trestman RL, Levengood RA, Siever LJ. Cortisol regulation in posttraumatic stress disorder and major depression: a chronological analysis. Biol Psychiatry. 1996;40:79 – 88 59. Young EA, Lopez JF, Murphy-Weinberg V, Watson SJ, Akil H. Hormonal evidence for altered responsiveness to social stress in major depression. Neuropsychopharmacology. 2000;23:411– 418 60. Rosmond R, Bouchard C, Bjorntorp P. Tsp509I polymorphism in exon 2 of the glucocorticoid receptor gene in relation to obesity and cortisol secretion: a cohort study. Br Med J. 2001;322:652– 653 61. Bjorntorp R, Rosmond R. Obesity and cortisol. Nutrition. 2000;16: 924 –936 62. Katz JR, Taylor NF, Goodrick S, Perry L, Yudkin JS, Coppack SW. Central obesity, depression and hypothalamo-pituitary-adrenal axis in men and postmenopausal women. Int J Obes. 2000;24:246 –251 63. Okosum IS, Liao Y, Rotimi CNM, Dever GE, Cooper RS. Impact of birth weight on ethnic variations in subcutaneous and central adiposity in American children aged 5–11 years: a study from the Third National Health and Nutrition Examination Survey. Int J Obes. 2000;24:479 – 484 64. Eriksson J, Forsen T, Tuomilehto J, Osmond C, Barker D. Size at birth, childhood growth and obesity in adult life. Int J Obes. 2001;25:735–740 65. Thompson C, Syddall H, Rodin I, Osmond C, Barker DJ. Birth weight and the risk of depressive disorder in late life. Br J Psychiatry. 2001;179: 450 – 455 66. Phillips DI, Walker BR, Reynolds RM, et al. Low birth weight predicts elevated plasma cortisol concentrations in adults from 3 populations. Hypertension. 2000;35:1301–1306 67. Welberg LA, Seckl JR, Holmes MC. Inhibition of 11 ␤-hydroxysteroid dehydrogenase, the foeto-placental barrier to maternal glucocorticoids, permanently programs amygdala GR mRNA expression and anxietylike behavior in the offspring. Eur J Neurosci. 2000;12:1047–1054 68. Malina RM, Katzmarzyk PT. Validity of the body mass index as an indicator of the risk and presence of overweight adolescents. Am J Clin Nutr. 1999;70:131S–136S 69. Frankenfield DC, Rowe WA, Cooney RN, Smith JS, Becker D. Limits of body mass index to detect obesity and predict body composition. Nutrition. 2000;17:26 –30

NEONATOLOGY: PLEASE COPY

“People will keep claiming to have created cloned babies and eventually someone will succeed, but at what cost? A lot of damaged children and disappointed parents.”

The New York Times, January 19, 2003

—Thomas Murray, Bioethicist


ARTICLES

859

Obesity and Psychiatric Disorder: Developmental Trajectories Sarah Mustillo, Carol Worthman, Alaattin Erkanli, Gordon Keeler, Adrian Angold and E. Jane Costello Pediatrics 2003;111;851-859 DOI: 10.1542/peds.111.4.851 Updated Information & Services

including high-resolution figures, can be found at: http://www.pediatrics.org/cgi/content/full/111/4/851

References

This article cites 49 articles, 20 of which you can access for free at: http://www.pediatrics.org/cgi/content/full/111/4/851#BIBL

Citations

This article has been cited by 8 HighWire-hosted articles: http://www.pediatrics.org/cgi/content/full/111/4/851#otherarticle s

Post-Publication Peer Reviews (P3Rs)

One P3R has been posted to this article: http://www.pediatrics.org/cgi/eletters/111/4/851

Subspecialty Collections

This article, along with others on similar topics, appears in the following collection(s): Nutrition & Metabolism http://www.pediatrics.org/cgi/collection/nutrition_and_metabolis m

Permissions & Licensing

Information about reproducing this article in parts (figures, tables) or in its entirety can be found online at: http://www.pediatrics.org/misc/Permissions.shtml

Reprints

Information about ordering reprints can be found online: http://www.pediatrics.org/misc/reprints.shtml
