- Email: [email protected]
Relationship of Childhood Behavior Disorders to Weight Gain from Childhood into Adulthood
Relationship of Childhood Behavior Disorders to Weight Gain from Childhood into Adulthood Sarah E. Anderson, PhD; Patricia Cohen, PhD; Elena N. Naumova, PhD; Aviva Must, PhD Objective.—Obesity and behavior disorders are important conditions that affect the health of children and adolescents; some evidence suggests that they are associated. However, these relationships have not been studied longitudinally from childhood to adulthood. We investigated childhood to adulthood weight change associated with attention-deficit and disruptive behavior disorders. Methods.—We analyzed data from a prospective cohort study in which 655 individuals observed before age 16.6 years, and assessed in 1983 (when aged 9.1 to 16.6 years), 1985–1986 (when 11.1 to 20.8 years), 1991–1994 (when 16.6 to 26.9 years), and 2001–2003 (when 27.7 to 38.3 years). Attention-deficit/ hyperactivity disorder, oppositional defiant disorder, and conduct disorder (collectively, “disruptive disorders”) were assessed by trained interviewers who used the Diagnostic Interview Schedule for Children. We evaluated the association of disruptive disorders (assessed when youth were ⬍16.6 years) with longitudinal body mass index (BMI) z scores from childhood to adulthood using linear mixed-effects models.
Results.—Female subjects with disruptive disorders were estimated, at all ages, to have mean BMI z scores 0.23 (95% confidence interval, 0.03– 0.44) units higher than female subjects without disruptive disorders. Male subjects with disruptive disorders were estimated, at all ages, to have mean BMI z scores 0.20 (95% confidence interval, 0.00 – 0.39) units higher than male subjects without disruptive disorders. For male and female subjects, annual BMI z-score change was statistically unrelated to disruptive disorder history. Conclusions.—Disruptive disorders were associated with elevated weight status that was maintained from childhood into adulthood. These findings suggest that associations between behavior disorders and increased weight begin early in childhood and may have lifelong health effects.
O
likely to capitulate to children’s requests for high-calorie snack foods, or to try to control behavior by using television or food as rewards. Socioeconomic status and race/ ethnicity could influence these associations if resource availability, education, or cultural factors affected parenting decisions.19,20 Focus groups indicate that it is common for low-income mothers to calm their young children with food, or to offer food as a reward for good behavior.21 On the basis of these theories, we hypothesized that children with attention-deficit/hyperactivity disorder (ADHD), oppositional defiant disorder (ODD), or conduct disorder (CD) would have higher weights in childhood, and would gain more weight with increasing age compared with children without these disorders. Additionally, we hypothesized that the association of behavior disorders with relative weight might differ by race/ethnicity, and on the basis of previous analyses of weight trajectory in this cohort, by sex.7 To test these hypotheses, we analyzed data from a large prospective community-based cohort, the Children in the Community Study, with high-quality assessment of psychological disorders. In an earlier limited analysis of this cohort, Pine et al16 reported that adolescent CD symptoms predicted higher young adulthood body mass index (BMI). Differences between our current report and that of Pine et al16 include longitudinal analysis of 4 waves of data collected over 20 years versus 2 waves separated by10 years, analysis of relative weight trajectory
KEY WORDS: adolescents; attention-deficit/hyperactivity disorder; conduct disorder; longitudinal study; oppositional defiant disorder Ambulatory Pediatrics 2006;6:297–301
besity and behavior disorders, commonly encountered by pediatric clinicians, affect health and development of youth. Obesity prevalence in the United States has increased dramatically and continues to rise.1 Developing effective obesity prevention and treatment programs will require understanding childhood risk factors for obesity. Growing evidence indicates an association between childhood obesity and behavioral/psychosocial problems.2–7 Childhood obesity and behavior disorders share many risk factors, such as parental neglect, bullying, and social marginalization.8 –12 A number of studies have observed higher weight status among individuals with behavior disorders,2– 4,13–16 but other studies have observed no association or lower weight associated with behavior disorders.17,18 Behavior disorders in children could lead to weight gain if, for example, parents of children with behavior disorders were more
From the Gerald J. and Dorothy R. Friedman School of Nutrition Science and Policy, Tufts University, Boston, Mass (Drs Anderson and Must); Department of Psychiatry, College of Physicians and Surgeons, Columbia University, and Department of Epidemiology, New York State Psychiatric Institute, New York, NY (Dr Cohen); and Department of Public Health and Family Medicine, Tufts University School of Medicine, Boston, Mass (Drs Naumova and Must). Address correspondence to Sarah E. Anderson, PhD, Department of Public Health and Family Medicine, M&V 1, Tufts University, 136 Harrison Ave, Boston, MA 02111 (e-mail: [email protected]). Received for publication February 24, 2006; accepted June 13, 2006.
AMBULATORY PEDIATRICS Copyright © 2006 by Ambulatory Pediatric Association
297
Volume 6, Number 5 September–October 2006
298
Anderson et al
versus a single weight status assessment, and analysis of ADHD, ODD, and CD diagnoses versus symptoms of CD. METHODS Study Sample The Children in the Community Study is a prospective study of determinants and correlates of psychiatric disorders; design and operations have been described previously.22 Briefly, the cohort is based on a sample of Upstate New York families, with a child between the ages of 1 and 10 years in 1975, supplemented in 1983 with a new sample from urban poverty areas. In-home interviews were conducted by extensively trained and supervised lay interviewers with mothers and participants in 1983 (wave 1) and 1985–1986 (wave 2), and with participants in 1991– 1994 (wave 3) and 2001–2003 (wave 4). The sample was demographically representative of the area in 1983: 25% rural or small town, 21% poverty, and 25% upper middle class. Fifty percent of children were male. Race/ethnicity of the participant was reported by the mother and was categorized as white (91% of participants), African American (8%), or other (1%). Written informed consent (or assent for children) was obtained at each wave. The Institutional Review Boards of the New York State Psychiatric Institute and the Tufts–New England Medical Center approved this study. Our analysis is limited to participants younger than 16.6 years at wave 1 (n ⫽ 630) or wave 2 (n ⫽ 25). This age cutoff was chosen to maximize inclusion, and so that at wave 3, all participants were older than the age cutoff. Thus, 655 individuals contribute data to these analyses: 454 individuals (69%) with 4 waves, 161 (25%) with 3 waves, 29 (⬍5%) with 2 waves, and 11 (⬍2%) with 1 wave. Assessment of Disruptive Disorders Attention-deficit and disruptive behavior disorders, as classified in the Diagnostic and Statistical Manual of Mental Disorders (DSM), include ADHD, ODD, and CD.23 For brevity, we will refer collectively to ADHD, ODD, and CD as “disruptive disorders.” The onset of disruptive disorders typically occurs in childhood.23 Psychological disorders were assessed at each wave by use of a structured diagnostic interview consistent with DSM criteria. The Diagnostic Interview Schedule for Children (DISC-1R)24 was administered separately, by trained lay interviewers, to a parent and the participant at waves 1 and 2. The reliability and validity of the DISC has been extensively studied.24 –26 Parents and subjects provide unique, nonoverlapping information,25 and disorders were identified if either respondent reported a minimum number of items meeting DSM-III-R criteria, and the combined parent-child symptom severity score was at least 1 standard deviation above the mean. This approach aims to maximize sensitivity and specificity of diagnoses.27 We also evaluated a more conservative diagnostic definition in which the combined symptom severity score was at least 2 standard deviations above the mean27; we will refer to this more conservative diagnostic level as evidence of a “se-
AMBULATORY PEDIATRICS
vere” disruptive disorder. Participants were considered to have a history of a child/adolescent disruptive disorder if they met diagnostic criteria for ODD, CD, or ADHD (we did not assess whether individuals had combined-type or primarily inattentive ADHD) and were younger than 16.6 years at the wave when the disorder was first recognized. Assessment of Relative Weight BMI at each wave was calculated (weight [kg])/(height [m]2) from reported height and weight. At waves 1 and 2, a parent reported the participant’s height and weight. At waves 2, 3, and 4, participants self-reported height and weight. At wave 2, we used the participant’s report of weight and height if they were older than 13 years; otherwise we used the parent report. Participants were observed from childhood to adulthood, which necessitated a measure of relative weight comparable over a wide age range. BMI is commonly used in adults, but for youth, it is necessary to standardize BMI to age- and sex-specific external reference data.28,29 We used the Centers for Disease Control and Prevention BMIfor-age reference data30 to calculate BMI z scores. BMI z scores (BMIz) correspond with growth chart percentiles and enable tracking of relative weight from age 2 to 20 years. To provide continuity into adulthood, we used the age-20 reference when individuals were 20 or older. Covariates Socioeconomic status (SES) was defined on the basis of family income, parental education, work status, occupation, and receipt of public assistance.22 Medication use for emotional or behavioral problems before age 21 years was assessed at wave 4 and was considered a potential confounder because some psychoactive medications contribute to weight gain or loss.31 However, few individuals (n ⫽ 8) in this cohort reported early psychoactive medication use. Statistical Analysis We used linear mixed-effects models32 to estimate differences in BMIz level and annual change in BMIz associated with disruptive disorders for male and female subjects. We used age in years as the measure of time in these models. We analyzed whether history of a disruptive disorder observed before age 16.6 years was associated with BMIz trajectory; we have used the term “trajectory” to refer simultaneously to initial BMIz level and annual change in BMIz. To allow for nonlinearity, BMIz trajectory was estimated as a cubic function of age. We controlled for SES in all models. We explored whether associations of disruptive disorders with BMIz trajectory were different for whites and African Americans (individuals with other race/ethnicity [n ⫽ 5] were excluded from these subanalyses); we also tested for interactions between SES and disruptive disorders. We evaluated comorbidity of ADHD, ODD, and CD, and assessed associations by disruptive disorder type.
AMBULATORY PEDIATRICS
Relationship of Childhood Behavior Disorders to Weight Gain
299
Table 1. Analytic Sample Characteristics of Children in the Community Study Participants* Female (n ⫽ 323) Characteristic
Wave 1 (1983)
n 299 Age (y), mean 13.0 Age (y), range 9.1–16.6 BMI, mean (SD) 19.9 (3.2) BMI z score, mean (SD) 0.2 (0.9) ⱖ85th percentile BMI, n (%)† 59 (21%) ⱖ95th percentile BMI, n (%)‡ 10 (4%)
Wave 2 (1985–1986)
Wave 3 (1991–1994)
305 15.3 11.1–20.8 20.9 (3.2) 0.1 (0.9) 39 (14%) 7 (2%)
307 21.2 16.6–26.5 23.1 (4.4) 0.1 (0.9) 53 (18%) 15 (5%)
Male (n ⫽ 332) Wave 4 (2001–2003)
Wave 1 (1983)
284 296 32.2 12.8 27.8–38.2 9.2–16.6 26.3 (6.1) 20.5 (4.4) 0.7 (0.9) 0.4 (1.1) 108 (38%) 77 (28%) 50 (18%) 31 (11%)
Wave 2 (1985–1986)
Wave 3 (1991–1994)
Wave 4 (2001–2003)
320 15.2 11.2–19.8 21.6 (3.8) 0.3 (1.0) 63 (21%) 25 (8%)
313 21.2 16.7–26.9 24.4 (4.2) 0.3 (1.0) 59 (22%) 29 (11%)
244 32.2 27.8–38.3 27.3 (5.0) 0.9 (0.9) 107 (44%) 40 (17%)
*BMI indicates body mass index. BMI z-score from CDC BMI-for-age growth reference.30 †The 85th percentile BMI indicates BMI of 26.5 in female and 27.0 in male subjects at age 20 or older. At the 2001–2003 wave, 128 female subjects (46%) and 164 male subjects (68%) had BMI ⱖ25. ‡The 95th percentile BMI indicates BMI of 31.8 in female and 30.6 in male subjects at age 20 or older. At the 2001–2003 wave, 60 female subjects (21%) and 50 male subjects (21%) had BMI ⱖ30.
RESULTS Characteristics of Study Participants Average BMIz increased by over 0.5 units between waves 3 and 4 (Table 1). Twenty-five female (8%) and 51 male (15%) subjects were identified with a “severe” childhood disruptive disorder. By use of the less conservative diagnostic level, 74 female (23%) and 115 male subjects (35%) had a disruptive disorder; of these, 8 female (11%) and 15 male (13%) subjects were African American. In 80% of cases the disruptive disorder was observed at wave 1; however we designated an individual as having a disruptive disorder if they met diagnostic criteria at either wave 1 or wave 2, and thus estimates represent cumulative prevalence. Further, 2 informants (mother and child) contributed to identification of diagnoses which increases sensitivity, but may inflate prevalence.27 The prevalence of “severe” disruptive disorders in this cohort is similar to that seen in other community samples.33,34 ADHD, ODD, and CD were often comorbid; of 74 female subjects with a disruptive disorder, 10 (14%) concurrently met diagnostic criteria for ADHD, ODD, and CD, and 19 (26%) met criteria for 2 of these 3 disorders. For male subjects, 14 (12%) met criteria for 3 disorders, and 34 (30%) met criteria for 2 disorders. Thirty-two female and 62 male subjects met diagnostic criteria for ADHD, 53 female and 54 male subjects for ODD, and 28 female and 61 male subjects for CD. Only 11 female and
25 male subjects met diagnostic criteria for ADHD without comorbid ODD or CD. Association of Disruptive Disorders With BMIz Trajectory Individuals identified with a disruptive disorder before age 16.6 years had higher estimated mean BMIz at all ages compared with individuals who were not observed with a disruptive disorder. Female subjects with history of a disruptive disorder were estimated to have a mean BMIz trajectory level 0.23 (95% confidence interval [CI], 0.03– 0.44) units higher (Table 2; Figure) than female subjects without a disruptive disorder. Mean BMIz for female subjects with a “severe” disruptive disorder was 0.33 (95% CI, 0.02– 0.64) units higher than for female subjects without a “severe” disruptive disorder. Male subjects with history of a disruptive disorder were estimated to have a mean BMIz trajectory level 0.20 (95% CI, 0.00 – 0.39) units higher (Table 2; Figure) than male subjects without a disruptive disorder; for “severe” disruptive disorders, this estimate was also 0.20 (95% CI, ⫺0.07 to 0.46) units higher. For female and male subjects, no evidence of an interaction between disruptive disorders and age was observed. Thus, BMIz trajectories for individuals with and without a history of disruptive disorder were estimated to remain parallel, with a fixed offset, from childhood into adulthood (Figure). In post hoc analyses, we determined that results for male and female subjects could be pooled (P ⫽ .84, sex by disruptive-disorder interaction). In this com-
Table 2. BMI z Score Trajectory Level Associated With Disruptive Disorders† Subjects
Overall Disruptive Disorders
ADHD
ODD
CD
ODD or CD
Female (n ⫽ 323) Male (n ⫽ 332)
0.23 (0.03 to 0.44)** 0.20 (0.00 to 0.39)**
0.23 (⫺0.05 to 0.52) 0.21 (⫺0.03 to 0.44)*
0.34 (0.11 to 0.57)*** 0.21 (⫺0.02 to 0.45)*
0.09 (⫺0.20 to 0.37) 0.22 (⫺0.02 to 0.46)†
0.27 (0.05 to 0.48)** 0.24 (0.03 to 0.45)**
†Estimated mean difference and 95% confidence interval associated with disruptive disorders overall and by disruptive disorder type. Data were analyzed by linear mixed effects models (Proc Mixed, SAS version 8). Body mass index (BMI) z-score trajectory modeled as a cubic function of age to accommodate nonlinearity. All models adjusted for socioeconomic status. ADHD indicates attention-deficit/hyperactivity disorder; ODD, oppositional defiant disorder; and CD, conduct disorder. *P ⬍ .1 **P ⬍ .05 ***P ⬍ .01
300
Anderson et al
Figure. Predicted mean trajectories of BMI z score with age for male and female subjects with and without a disruptive disorder observed before age 16.6 years. Estimates from linear mixed effects models, at median SES.
bined analysis, individuals with a history of disruptive disorder were estimated to have a mean BMIz 0.21 (95% CI, 0.07– 0.35) units higher than individuals without a disruptive disorder. To assess sensitivity, we evaluated age cutpoints of ⬍14 years, ⬍15 years, or ⬍16 years for definition of child/ adolescent disruptive disorders. The estimates from these analyses were somewhat attenuated, although generally consistent, compared with the ⬍16.6 years definition. Three male and 5 female subjects reported use of medication for emotional or behavioral problems before age 21; results were not changed by exclusion of these individuals. When ADHD, ODD, and CD were modeled separately, the pattern of results did not differ substantially (Table 2); there is a suggestion that ODD was associated with the greatest BMIz difference in female subjects. We noted a marginally statistically significant interaction (P ⫽ .08) between disruptive disorder status and race/ethnicity in female subjects; African American female subjects with history of a disruptive disorder were estimated to have a mean BMIz trajectory level 0.58 units higher (95% CI, 0.07 units lower to 1.23 units higher) than African American female subjects of the same age and SES without history of a disruptive disorder. DISCUSSION Child/adolescent disruptive disorders were associated with persistently higher BMI z scores in female and male subjects. Compared with individuals without a history of disruptive disorder, a history of child/adolescent disruptive disorder predicted mean BMIz trajectory level 0.23 units higher for female subjects and 0.20 units higher for male subjects at all ages. These differences in relative weight are not great, but depending on height could equate to weight differences in adulthood of approximately 5 to 15 pounds. Our results indicate that participants with a disruptive disorder at wave 1 already had higher weights than those not identified with a disruptive disorder, and that this higher weight level was maintained into adulthood. The
AMBULATORY PEDIATRICS
tendency for weight status to track with age is well established.35 Therefore, we cannot discern whether disruptive disorder onset occurs before, after, or concurrently with increases in weight. Nevertheless, our analyses are unique in examining the influence of child/adolescent disruptive disorders on BMIz trajectory over a 20-year period, they support the hypothesis that an association between higher weight and disruptive disorders is established early in childhood, and they preclude a number of potential weight-related outcomes. For example, our data are not consistent with our initial hypothesis that individuals with disruptive disorders would experience larger annual weight gains compared with other individuals. Our results are generally in keeping with those of others who reported cross-sectional associations between higher weight and behavior problems in young children.3,4 Mustillo et al2 observed higher prevalence of ODD in chronically overweight children. Our results are also largely consistent with a previous more-limited analysis of Children in the Community Study data by Pine et al16; however, our conclusions differ substantially. Pine et al16 reported that wave 1 symptoms of CD predicted high BMI at wave 3; however, wave 1 BMI was not accounted for, and results were interpreted to support a directional association between adolescent CD and early adulthood obesity. In contrast, the results of our current analyses suggest that CD, ODD, and ADHD are associated with higher weight at an early age, and do not suggest that children with disruptive disorders will gain weight more rapidly than other children. We are not aware of studies that have examined associations of disruptive disorders and weight status among African Americans. Our observation that disruptive disorders in African American female subjects may be associated with BMIz ⬃0.6 units higher compared with other African American female subjects, was of marginal statistical significance (our power to detect interactions by race/ethnicity was limited). Nevertheless, further investigation could lead to greater understanding of the reasons for the greater obesity prevalence in the female African American population. Strengths of our analysis include longitudinal follow-up of a community cohort from childhood through adulthood. Four waves of data, combined with a wide age range, enabled us to investigate longitudinal associations for individuals aged 9 to 38 years. Another strength of our analysis is the assessment of disruptive disorders consistent with DSM diagnostic criteria. Several limitations are noteworthy. First, height and weight were self-reported. Although self-reported height and weight are considered accurate in adults and older teenagers,36,37 some error is likely, especially among younger adolescents. We would expect such measurement error to attenuate effects observed. Second, the youngest individuals in the cohort were 9 years old at wave 1, and assessments were widely spaced; if assessment were more frequent when these children were younger, we might have been better able to determine the directionality of the association between disruptive disorders and higher weight.
AMBULATORY PEDIATRICS
Relationship of Childhood Behavior Disorders to Weight Gain
Disruptive disorders were associated with higher BMI z scores in childhood that were maintained into adulthood. Further research is needed, and in order that the development of associations between weight and disruptive disorders can be most productively studied, this research should occur in a young, racially diverse cohort. Greater understanding of interrelationships between childhood weight and behavior problems will have implications for clinicians working with children and their families, and for broad-based obesity-prevention policy. ACKNOWLEDGMENTS This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (grants T32 DK62032-11, R21 DK64254), the National Institute of Child Health and Human Development (grant HD-40685), and the National Institute of Mental Health (grants MH-36971, MH-38916, MH-49191).
REFERENCES 1. Ogden CL, Carroll MD, Curtin LR, et al. Prevalence of overweight and obesity in the United States, 1999 –2004. JAMA. 2006;295: 1549 –1555. 2. Mustillo S, Worthman C, Erkanli A, et al. Obesity and psychiatric disorder: developmental trajectories. Pediatrics. 2003;111:851– 859. 3. Datar A, Sturm R. Childhood overweight and parent- and teacherreported behavior problems. Arch Pediatr Adolesc Med. 2004;158:804 – 810. 4. Lumeng JC, Gannon K, Cabral HJ, et al. Association between clinically meaningful behavior problems and overweight in children. Pediatrics. 2003;112:1138 –1145. 5. Goodman E, Whitaker RC. A prospective study of the role of depression in the development and persistence of adolescent obesity. Pediatrics. 2002;109:497–504. 6. Faith MS, Matz PE, Jorge MA. Obesity-depression associations in the population. J Psychosomat Res. 2002;53:935–942. 7. Anderson SE, Cohen P, Naumova EN, Must A. Association of depression and anxiety disorders with weight change in a prospective community-based study of children followed up into adulthood. Arch Pediatr Adolesc Med. 2006;160:285–291. 8. Lissau I, Sorensen TI. Parental neglect during childhood and increased risk of obesity in young adulthood. Lancet. 1994;343: 324 –327. 9. Janssen I. Associations between overweight and obesity with bullying behaviors in school-aged children. Pediatrics. 2004;113: 1187–1194. 10. Strauss RS, Pollack HA. Social marginalization of overweight adolescents. Arch Pediatr Adolesc Med. 2003;157:746 –752. 11. Burke JD, Loeber R, Birmaher MD. Oppositional defiant disorder and conduct disorder: a review of the past 10 years, part II. J Am Acad Child Adolesc Psychiatry. 2002;41:1275–1293. 12. Deater-Deckard K, Dodge KA, Bates JE, Pettit GS. Multiple risk factors in the development of externalizing behavior problems: group and individual differences. Dev Psychopathol. 1998;10: 469 – 493. 13. Altfas JR. Prevalence of attention deficit/hyperactivity disorder among adults in obesity treatment. BMC Psychiatry. 2002;2:1– 8. 14. Holtkamp K, Konrad K, Muller B, et al. Overweight and obesity in children with attention-deficit/hyperactivity disorder. Int J Obes. 2004;28:685– 689. 15. Lawlor DA, Mamun AA, O’Callaghan MJ, et al. Is being overweight associated with behavioural problems in childhood and adolescence? Findings from the Mater-University study of pregnancy and its outcomes. Arch Dis Child. 2005;90:692– 697. 16. 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.
301
17. Curtin C, Bandini LG, Perrin EC, et al. Prevalence of overweight in children and adolescents with attention deficit hyperactivity disorder and autism spectrum disorders: a chart review. BMC Pediatr. 2005;5:48. 18. Bardone AM, Moffitt TE, Caspi A, et al. Adult physical health outcomes of adolescent girls with conduct disorder, depression, and anxiety. J Am Acad Child Adolesc Psychiatry. 1998;37:594 – 601. 19. Kumanyika S. Ethnicity and obesity development in children. Ann N Y Acad Sci. 1993;699:81–92. 20. Sherry B, McDivitt J, Birch LL, et al. Attitudes, practices, and concerns about child feeding and child weight status among socioeconomically diverse white, Hispanic, and African-American mothers. J Am Diet Assoc. 2004;104:215–221. 21. Baughcum AE, Burklow KA, Deeks CM, et al. Maternal feeding practices and childhood obesity: a focus group study of low-income mothers. Arch Pediatr Adolesc Med. 1998;152:1010 –1014. 22. Cohen P, Cohen J. Life Values and Adolescent Mental Health. Mahwah, NJ: Lawrence Erlbaum Associates; 1996. 23. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th rev ed. Washington, DC: American Psychiatric Association; 2000. 24. Costello EJ, Edelbrock CS, Duncan MK, Kalas R. Testing of the NIMH Diagnostic Interview Schedule for Children (DISC) in a Clinical Population: Final Report to the Center for Epidemiological Studies, NIHM. Pittsburgh, Pa: University of Pittsburgh; 1984. 25. Costello EJ, Edelbrock CS, Costello AJ. Validity of the NIMH Diagnostic Interview Schedule for Children: a comparison between psychiatric and pediatric referrals. J Abnorm Child Psychol. 1985;13:579 –595. 26. Schwab-Stone ME, Shaffer D, Dulcan MK, et al. Criterion validity of the NIMH Diagnostic Interview Schedule for Children version 2.3 (DISC-2.3). J Am Acad Child Adolesc Psychiatry. 1996;35: 878 – 888. 27. Cohen P, Velez N, Kohn M, et al. Child psychiatric diagnosis by computer algorithm: theoretical issues and empirical tests. J Am Acad Child Adolesc Psychiatry. 1987;26:631– 638. 28. Barlow SE, Dietz WH. Obesity evaluation and treatment: expert committee recommendations. Pediatrics. 1998;102:e29. 29. Must A, Anderson SE. Body mass index in children and adolescents: considerations for population-based applications. Int J Obes. 2006;30:590 –594. 30. Ogden CL, Kuczmarski RL, Flegal KM, et al. Centers for Disease Control and Prevention 2000 growth charts for the United States: improvements to the 1977 National Center for Health Statistics version. Pediatrics. 2002;109:45– 60. 31. Wiener JMJ. Historical overview of childhood and adolescent psychopharmacology. In: Wiener JM, ed. Diagnosis and Psychopharmacology of Childhood and Adolescent Disorders. New York: John Wiley & Sons; 1985:3–50. 32. Singer JD, Willett JB. Modeling discontinuous and nonlinear change. In: Singer JD, Willett JB, eds. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press; 2003:189 –242. 33. Buka SL, Monuteaux M, Earls F. Epidemiology of child and adolescent mental disorders. In: Tsuang MT, Tochen M, eds. Textbook in Psychiatric Epidemiology. 2nd ed. New York: Wiley-Liss Inc; 2002:629 – 655. 34. Costello EJ, Mustillo S, Erkanli A, et al. Prevalence and development of psychiatric disorders in childhood and adolescence. Arch Gen Psychiatry. 2003;60:837– 844. 35. Guo SS, Chumlea WC. Tracking of body mass index in children in relation to overweight in adulthood. Am J Clin Nutr. 1999;70: 145S–148S. 36. Brener ND, McManus T, Galuska DA, et al. Reliability and validity of self-reported height and weight among high school students. J Adolesc Health. 2003;32:281–287. 37. Stewart AL. The reliability and validity of self-reported weight and height. J Chronic Dis. 1982;35:295–309.
Results.—Female subjects with disruptive disorders were estimated, at all ages, to have mean BMI z scores 0.23 (95% confidence interval, 0.03– 0.44) units higher than female subjects without disruptive disorders. Male subjects with disruptive disorders were estimated, at all ages, to have mean BMI z scores 0.20 (95% confidence interval, 0.00 – 0.39) units higher than male subjects without disruptive disorders. For male and female subjects, annual BMI z-score change was statistically unrelated to disruptive disorder history. Conclusions.—Disruptive disorders were associated with elevated weight status that was maintained from childhood into adulthood. These findings suggest that associations between behavior disorders and increased weight begin early in childhood and may have lifelong health effects.
O
likely to capitulate to children’s requests for high-calorie snack foods, or to try to control behavior by using television or food as rewards. Socioeconomic status and race/ ethnicity could influence these associations if resource availability, education, or cultural factors affected parenting decisions.19,20 Focus groups indicate that it is common for low-income mothers to calm their young children with food, or to offer food as a reward for good behavior.21 On the basis of these theories, we hypothesized that children with attention-deficit/hyperactivity disorder (ADHD), oppositional defiant disorder (ODD), or conduct disorder (CD) would have higher weights in childhood, and would gain more weight with increasing age compared with children without these disorders. Additionally, we hypothesized that the association of behavior disorders with relative weight might differ by race/ethnicity, and on the basis of previous analyses of weight trajectory in this cohort, by sex.7 To test these hypotheses, we analyzed data from a large prospective community-based cohort, the Children in the Community Study, with high-quality assessment of psychological disorders. In an earlier limited analysis of this cohort, Pine et al16 reported that adolescent CD symptoms predicted higher young adulthood body mass index (BMI). Differences between our current report and that of Pine et al16 include longitudinal analysis of 4 waves of data collected over 20 years versus 2 waves separated by10 years, analysis of relative weight trajectory
KEY WORDS: adolescents; attention-deficit/hyperactivity disorder; conduct disorder; longitudinal study; oppositional defiant disorder Ambulatory Pediatrics 2006;6:297–301
besity and behavior disorders, commonly encountered by pediatric clinicians, affect health and development of youth. Obesity prevalence in the United States has increased dramatically and continues to rise.1 Developing effective obesity prevention and treatment programs will require understanding childhood risk factors for obesity. Growing evidence indicates an association between childhood obesity and behavioral/psychosocial problems.2–7 Childhood obesity and behavior disorders share many risk factors, such as parental neglect, bullying, and social marginalization.8 –12 A number of studies have observed higher weight status among individuals with behavior disorders,2– 4,13–16 but other studies have observed no association or lower weight associated with behavior disorders.17,18 Behavior disorders in children could lead to weight gain if, for example, parents of children with behavior disorders were more
From the Gerald J. and Dorothy R. Friedman School of Nutrition Science and Policy, Tufts University, Boston, Mass (Drs Anderson and Must); Department of Psychiatry, College of Physicians and Surgeons, Columbia University, and Department of Epidemiology, New York State Psychiatric Institute, New York, NY (Dr Cohen); and Department of Public Health and Family Medicine, Tufts University School of Medicine, Boston, Mass (Drs Naumova and Must). Address correspondence to Sarah E. Anderson, PhD, Department of Public Health and Family Medicine, M&V 1, Tufts University, 136 Harrison Ave, Boston, MA 02111 (e-mail: [email protected]). Received for publication February 24, 2006; accepted June 13, 2006.
AMBULATORY PEDIATRICS Copyright © 2006 by Ambulatory Pediatric Association
297
Volume 6, Number 5 September–October 2006
298
Anderson et al
versus a single weight status assessment, and analysis of ADHD, ODD, and CD diagnoses versus symptoms of CD. METHODS Study Sample The Children in the Community Study is a prospective study of determinants and correlates of psychiatric disorders; design and operations have been described previously.22 Briefly, the cohort is based on a sample of Upstate New York families, with a child between the ages of 1 and 10 years in 1975, supplemented in 1983 with a new sample from urban poverty areas. In-home interviews were conducted by extensively trained and supervised lay interviewers with mothers and participants in 1983 (wave 1) and 1985–1986 (wave 2), and with participants in 1991– 1994 (wave 3) and 2001–2003 (wave 4). The sample was demographically representative of the area in 1983: 25% rural or small town, 21% poverty, and 25% upper middle class. Fifty percent of children were male. Race/ethnicity of the participant was reported by the mother and was categorized as white (91% of participants), African American (8%), or other (1%). Written informed consent (or assent for children) was obtained at each wave. The Institutional Review Boards of the New York State Psychiatric Institute and the Tufts–New England Medical Center approved this study. Our analysis is limited to participants younger than 16.6 years at wave 1 (n ⫽ 630) or wave 2 (n ⫽ 25). This age cutoff was chosen to maximize inclusion, and so that at wave 3, all participants were older than the age cutoff. Thus, 655 individuals contribute data to these analyses: 454 individuals (69%) with 4 waves, 161 (25%) with 3 waves, 29 (⬍5%) with 2 waves, and 11 (⬍2%) with 1 wave. Assessment of Disruptive Disorders Attention-deficit and disruptive behavior disorders, as classified in the Diagnostic and Statistical Manual of Mental Disorders (DSM), include ADHD, ODD, and CD.23 For brevity, we will refer collectively to ADHD, ODD, and CD as “disruptive disorders.” The onset of disruptive disorders typically occurs in childhood.23 Psychological disorders were assessed at each wave by use of a structured diagnostic interview consistent with DSM criteria. The Diagnostic Interview Schedule for Children (DISC-1R)24 was administered separately, by trained lay interviewers, to a parent and the participant at waves 1 and 2. The reliability and validity of the DISC has been extensively studied.24 –26 Parents and subjects provide unique, nonoverlapping information,25 and disorders were identified if either respondent reported a minimum number of items meeting DSM-III-R criteria, and the combined parent-child symptom severity score was at least 1 standard deviation above the mean. This approach aims to maximize sensitivity and specificity of diagnoses.27 We also evaluated a more conservative diagnostic definition in which the combined symptom severity score was at least 2 standard deviations above the mean27; we will refer to this more conservative diagnostic level as evidence of a “se-
AMBULATORY PEDIATRICS
vere” disruptive disorder. Participants were considered to have a history of a child/adolescent disruptive disorder if they met diagnostic criteria for ODD, CD, or ADHD (we did not assess whether individuals had combined-type or primarily inattentive ADHD) and were younger than 16.6 years at the wave when the disorder was first recognized. Assessment of Relative Weight BMI at each wave was calculated (weight [kg])/(height [m]2) from reported height and weight. At waves 1 and 2, a parent reported the participant’s height and weight. At waves 2, 3, and 4, participants self-reported height and weight. At wave 2, we used the participant’s report of weight and height if they were older than 13 years; otherwise we used the parent report. Participants were observed from childhood to adulthood, which necessitated a measure of relative weight comparable over a wide age range. BMI is commonly used in adults, but for youth, it is necessary to standardize BMI to age- and sex-specific external reference data.28,29 We used the Centers for Disease Control and Prevention BMIfor-age reference data30 to calculate BMI z scores. BMI z scores (BMIz) correspond with growth chart percentiles and enable tracking of relative weight from age 2 to 20 years. To provide continuity into adulthood, we used the age-20 reference when individuals were 20 or older. Covariates Socioeconomic status (SES) was defined on the basis of family income, parental education, work status, occupation, and receipt of public assistance.22 Medication use for emotional or behavioral problems before age 21 years was assessed at wave 4 and was considered a potential confounder because some psychoactive medications contribute to weight gain or loss.31 However, few individuals (n ⫽ 8) in this cohort reported early psychoactive medication use. Statistical Analysis We used linear mixed-effects models32 to estimate differences in BMIz level and annual change in BMIz associated with disruptive disorders for male and female subjects. We used age in years as the measure of time in these models. We analyzed whether history of a disruptive disorder observed before age 16.6 years was associated with BMIz trajectory; we have used the term “trajectory” to refer simultaneously to initial BMIz level and annual change in BMIz. To allow for nonlinearity, BMIz trajectory was estimated as a cubic function of age. We controlled for SES in all models. We explored whether associations of disruptive disorders with BMIz trajectory were different for whites and African Americans (individuals with other race/ethnicity [n ⫽ 5] were excluded from these subanalyses); we also tested for interactions between SES and disruptive disorders. We evaluated comorbidity of ADHD, ODD, and CD, and assessed associations by disruptive disorder type.
AMBULATORY PEDIATRICS
Relationship of Childhood Behavior Disorders to Weight Gain
299
Table 1. Analytic Sample Characteristics of Children in the Community Study Participants* Female (n ⫽ 323) Characteristic
Wave 1 (1983)
n 299 Age (y), mean 13.0 Age (y), range 9.1–16.6 BMI, mean (SD) 19.9 (3.2) BMI z score, mean (SD) 0.2 (0.9) ⱖ85th percentile BMI, n (%)† 59 (21%) ⱖ95th percentile BMI, n (%)‡ 10 (4%)
Wave 2 (1985–1986)
Wave 3 (1991–1994)
305 15.3 11.1–20.8 20.9 (3.2) 0.1 (0.9) 39 (14%) 7 (2%)
307 21.2 16.6–26.5 23.1 (4.4) 0.1 (0.9) 53 (18%) 15 (5%)
Male (n ⫽ 332) Wave 4 (2001–2003)
Wave 1 (1983)
284 296 32.2 12.8 27.8–38.2 9.2–16.6 26.3 (6.1) 20.5 (4.4) 0.7 (0.9) 0.4 (1.1) 108 (38%) 77 (28%) 50 (18%) 31 (11%)
Wave 2 (1985–1986)
Wave 3 (1991–1994)
Wave 4 (2001–2003)
320 15.2 11.2–19.8 21.6 (3.8) 0.3 (1.0) 63 (21%) 25 (8%)
313 21.2 16.7–26.9 24.4 (4.2) 0.3 (1.0) 59 (22%) 29 (11%)
244 32.2 27.8–38.3 27.3 (5.0) 0.9 (0.9) 107 (44%) 40 (17%)
*BMI indicates body mass index. BMI z-score from CDC BMI-for-age growth reference.30 †The 85th percentile BMI indicates BMI of 26.5 in female and 27.0 in male subjects at age 20 or older. At the 2001–2003 wave, 128 female subjects (46%) and 164 male subjects (68%) had BMI ⱖ25. ‡The 95th percentile BMI indicates BMI of 31.8 in female and 30.6 in male subjects at age 20 or older. At the 2001–2003 wave, 60 female subjects (21%) and 50 male subjects (21%) had BMI ⱖ30.
RESULTS Characteristics of Study Participants Average BMIz increased by over 0.5 units between waves 3 and 4 (Table 1). Twenty-five female (8%) and 51 male (15%) subjects were identified with a “severe” childhood disruptive disorder. By use of the less conservative diagnostic level, 74 female (23%) and 115 male subjects (35%) had a disruptive disorder; of these, 8 female (11%) and 15 male (13%) subjects were African American. In 80% of cases the disruptive disorder was observed at wave 1; however we designated an individual as having a disruptive disorder if they met diagnostic criteria at either wave 1 or wave 2, and thus estimates represent cumulative prevalence. Further, 2 informants (mother and child) contributed to identification of diagnoses which increases sensitivity, but may inflate prevalence.27 The prevalence of “severe” disruptive disorders in this cohort is similar to that seen in other community samples.33,34 ADHD, ODD, and CD were often comorbid; of 74 female subjects with a disruptive disorder, 10 (14%) concurrently met diagnostic criteria for ADHD, ODD, and CD, and 19 (26%) met criteria for 2 of these 3 disorders. For male subjects, 14 (12%) met criteria for 3 disorders, and 34 (30%) met criteria for 2 disorders. Thirty-two female and 62 male subjects met diagnostic criteria for ADHD, 53 female and 54 male subjects for ODD, and 28 female and 61 male subjects for CD. Only 11 female and
25 male subjects met diagnostic criteria for ADHD without comorbid ODD or CD. Association of Disruptive Disorders With BMIz Trajectory Individuals identified with a disruptive disorder before age 16.6 years had higher estimated mean BMIz at all ages compared with individuals who were not observed with a disruptive disorder. Female subjects with history of a disruptive disorder were estimated to have a mean BMIz trajectory level 0.23 (95% confidence interval [CI], 0.03– 0.44) units higher (Table 2; Figure) than female subjects without a disruptive disorder. Mean BMIz for female subjects with a “severe” disruptive disorder was 0.33 (95% CI, 0.02– 0.64) units higher than for female subjects without a “severe” disruptive disorder. Male subjects with history of a disruptive disorder were estimated to have a mean BMIz trajectory level 0.20 (95% CI, 0.00 – 0.39) units higher (Table 2; Figure) than male subjects without a disruptive disorder; for “severe” disruptive disorders, this estimate was also 0.20 (95% CI, ⫺0.07 to 0.46) units higher. For female and male subjects, no evidence of an interaction between disruptive disorders and age was observed. Thus, BMIz trajectories for individuals with and without a history of disruptive disorder were estimated to remain parallel, with a fixed offset, from childhood into adulthood (Figure). In post hoc analyses, we determined that results for male and female subjects could be pooled (P ⫽ .84, sex by disruptive-disorder interaction). In this com-
Table 2. BMI z Score Trajectory Level Associated With Disruptive Disorders† Subjects
Overall Disruptive Disorders
ADHD
ODD
CD
ODD or CD
Female (n ⫽ 323) Male (n ⫽ 332)
0.23 (0.03 to 0.44)** 0.20 (0.00 to 0.39)**
0.23 (⫺0.05 to 0.52) 0.21 (⫺0.03 to 0.44)*
0.34 (0.11 to 0.57)*** 0.21 (⫺0.02 to 0.45)*
0.09 (⫺0.20 to 0.37) 0.22 (⫺0.02 to 0.46)†
0.27 (0.05 to 0.48)** 0.24 (0.03 to 0.45)**
†Estimated mean difference and 95% confidence interval associated with disruptive disorders overall and by disruptive disorder type. Data were analyzed by linear mixed effects models (Proc Mixed, SAS version 8). Body mass index (BMI) z-score trajectory modeled as a cubic function of age to accommodate nonlinearity. All models adjusted for socioeconomic status. ADHD indicates attention-deficit/hyperactivity disorder; ODD, oppositional defiant disorder; and CD, conduct disorder. *P ⬍ .1 **P ⬍ .05 ***P ⬍ .01
300
Anderson et al
Figure. Predicted mean trajectories of BMI z score with age for male and female subjects with and without a disruptive disorder observed before age 16.6 years. Estimates from linear mixed effects models, at median SES.
bined analysis, individuals with a history of disruptive disorder were estimated to have a mean BMIz 0.21 (95% CI, 0.07– 0.35) units higher than individuals without a disruptive disorder. To assess sensitivity, we evaluated age cutpoints of ⬍14 years, ⬍15 years, or ⬍16 years for definition of child/ adolescent disruptive disorders. The estimates from these analyses were somewhat attenuated, although generally consistent, compared with the ⬍16.6 years definition. Three male and 5 female subjects reported use of medication for emotional or behavioral problems before age 21; results were not changed by exclusion of these individuals. When ADHD, ODD, and CD were modeled separately, the pattern of results did not differ substantially (Table 2); there is a suggestion that ODD was associated with the greatest BMIz difference in female subjects. We noted a marginally statistically significant interaction (P ⫽ .08) between disruptive disorder status and race/ethnicity in female subjects; African American female subjects with history of a disruptive disorder were estimated to have a mean BMIz trajectory level 0.58 units higher (95% CI, 0.07 units lower to 1.23 units higher) than African American female subjects of the same age and SES without history of a disruptive disorder. DISCUSSION Child/adolescent disruptive disorders were associated with persistently higher BMI z scores in female and male subjects. Compared with individuals without a history of disruptive disorder, a history of child/adolescent disruptive disorder predicted mean BMIz trajectory level 0.23 units higher for female subjects and 0.20 units higher for male subjects at all ages. These differences in relative weight are not great, but depending on height could equate to weight differences in adulthood of approximately 5 to 15 pounds. Our results indicate that participants with a disruptive disorder at wave 1 already had higher weights than those not identified with a disruptive disorder, and that this higher weight level was maintained into adulthood. The
AMBULATORY PEDIATRICS
tendency for weight status to track with age is well established.35 Therefore, we cannot discern whether disruptive disorder onset occurs before, after, or concurrently with increases in weight. Nevertheless, our analyses are unique in examining the influence of child/adolescent disruptive disorders on BMIz trajectory over a 20-year period, they support the hypothesis that an association between higher weight and disruptive disorders is established early in childhood, and they preclude a number of potential weight-related outcomes. For example, our data are not consistent with our initial hypothesis that individuals with disruptive disorders would experience larger annual weight gains compared with other individuals. Our results are generally in keeping with those of others who reported cross-sectional associations between higher weight and behavior problems in young children.3,4 Mustillo et al2 observed higher prevalence of ODD in chronically overweight children. Our results are also largely consistent with a previous more-limited analysis of Children in the Community Study data by Pine et al16; however, our conclusions differ substantially. Pine et al16 reported that wave 1 symptoms of CD predicted high BMI at wave 3; however, wave 1 BMI was not accounted for, and results were interpreted to support a directional association between adolescent CD and early adulthood obesity. In contrast, the results of our current analyses suggest that CD, ODD, and ADHD are associated with higher weight at an early age, and do not suggest that children with disruptive disorders will gain weight more rapidly than other children. We are not aware of studies that have examined associations of disruptive disorders and weight status among African Americans. Our observation that disruptive disorders in African American female subjects may be associated with BMIz ⬃0.6 units higher compared with other African American female subjects, was of marginal statistical significance (our power to detect interactions by race/ethnicity was limited). Nevertheless, further investigation could lead to greater understanding of the reasons for the greater obesity prevalence in the female African American population. Strengths of our analysis include longitudinal follow-up of a community cohort from childhood through adulthood. Four waves of data, combined with a wide age range, enabled us to investigate longitudinal associations for individuals aged 9 to 38 years. Another strength of our analysis is the assessment of disruptive disorders consistent with DSM diagnostic criteria. Several limitations are noteworthy. First, height and weight were self-reported. Although self-reported height and weight are considered accurate in adults and older teenagers,36,37 some error is likely, especially among younger adolescents. We would expect such measurement error to attenuate effects observed. Second, the youngest individuals in the cohort were 9 years old at wave 1, and assessments were widely spaced; if assessment were more frequent when these children were younger, we might have been better able to determine the directionality of the association between disruptive disorders and higher weight.
AMBULATORY PEDIATRICS
Relationship of Childhood Behavior Disorders to Weight Gain
Disruptive disorders were associated with higher BMI z scores in childhood that were maintained into adulthood. Further research is needed, and in order that the development of associations between weight and disruptive disorders can be most productively studied, this research should occur in a young, racially diverse cohort. Greater understanding of interrelationships between childhood weight and behavior problems will have implications for clinicians working with children and their families, and for broad-based obesity-prevention policy. ACKNOWLEDGMENTS This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (grants T32 DK62032-11, R21 DK64254), the National Institute of Child Health and Human Development (grant HD-40685), and the National Institute of Mental Health (grants MH-36971, MH-38916, MH-49191).
REFERENCES 1. Ogden CL, Carroll MD, Curtin LR, et al. Prevalence of overweight and obesity in the United States, 1999 –2004. JAMA. 2006;295: 1549 –1555. 2. Mustillo S, Worthman C, Erkanli A, et al. Obesity and psychiatric disorder: developmental trajectories. Pediatrics. 2003;111:851– 859. 3. Datar A, Sturm R. Childhood overweight and parent- and teacherreported behavior problems. Arch Pediatr Adolesc Med. 2004;158:804 – 810. 4. Lumeng JC, Gannon K, Cabral HJ, et al. Association between clinically meaningful behavior problems and overweight in children. Pediatrics. 2003;112:1138 –1145. 5. Goodman E, Whitaker RC. A prospective study of the role of depression in the development and persistence of adolescent obesity. Pediatrics. 2002;109:497–504. 6. Faith MS, Matz PE, Jorge MA. Obesity-depression associations in the population. J Psychosomat Res. 2002;53:935–942. 7. Anderson SE, Cohen P, Naumova EN, Must A. Association of depression and anxiety disorders with weight change in a prospective community-based study of children followed up into adulthood. Arch Pediatr Adolesc Med. 2006;160:285–291. 8. Lissau I, Sorensen TI. Parental neglect during childhood and increased risk of obesity in young adulthood. Lancet. 1994;343: 324 –327. 9. Janssen I. Associations between overweight and obesity with bullying behaviors in school-aged children. Pediatrics. 2004;113: 1187–1194. 10. Strauss RS, Pollack HA. Social marginalization of overweight adolescents. Arch Pediatr Adolesc Med. 2003;157:746 –752. 11. Burke JD, Loeber R, Birmaher MD. Oppositional defiant disorder and conduct disorder: a review of the past 10 years, part II. J Am Acad Child Adolesc Psychiatry. 2002;41:1275–1293. 12. Deater-Deckard K, Dodge KA, Bates JE, Pettit GS. Multiple risk factors in the development of externalizing behavior problems: group and individual differences. Dev Psychopathol. 1998;10: 469 – 493. 13. Altfas JR. Prevalence of attention deficit/hyperactivity disorder among adults in obesity treatment. BMC Psychiatry. 2002;2:1– 8. 14. Holtkamp K, Konrad K, Muller B, et al. Overweight and obesity in children with attention-deficit/hyperactivity disorder. Int J Obes. 2004;28:685– 689. 15. Lawlor DA, Mamun AA, O’Callaghan MJ, et al. Is being overweight associated with behavioural problems in childhood and adolescence? Findings from the Mater-University study of pregnancy and its outcomes. Arch Dis Child. 2005;90:692– 697. 16. 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.
301
17. Curtin C, Bandini LG, Perrin EC, et al. Prevalence of overweight in children and adolescents with attention deficit hyperactivity disorder and autism spectrum disorders: a chart review. BMC Pediatr. 2005;5:48. 18. Bardone AM, Moffitt TE, Caspi A, et al. Adult physical health outcomes of adolescent girls with conduct disorder, depression, and anxiety. J Am Acad Child Adolesc Psychiatry. 1998;37:594 – 601. 19. Kumanyika S. Ethnicity and obesity development in children. Ann N Y Acad Sci. 1993;699:81–92. 20. Sherry B, McDivitt J, Birch LL, et al. Attitudes, practices, and concerns about child feeding and child weight status among socioeconomically diverse white, Hispanic, and African-American mothers. J Am Diet Assoc. 2004;104:215–221. 21. Baughcum AE, Burklow KA, Deeks CM, et al. Maternal feeding practices and childhood obesity: a focus group study of low-income mothers. Arch Pediatr Adolesc Med. 1998;152:1010 –1014. 22. Cohen P, Cohen J. Life Values and Adolescent Mental Health. Mahwah, NJ: Lawrence Erlbaum Associates; 1996. 23. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th rev ed. Washington, DC: American Psychiatric Association; 2000. 24. Costello EJ, Edelbrock CS, Duncan MK, Kalas R. Testing of the NIMH Diagnostic Interview Schedule for Children (DISC) in a Clinical Population: Final Report to the Center for Epidemiological Studies, NIHM. Pittsburgh, Pa: University of Pittsburgh; 1984. 25. Costello EJ, Edelbrock CS, Costello AJ. Validity of the NIMH Diagnostic Interview Schedule for Children: a comparison between psychiatric and pediatric referrals. J Abnorm Child Psychol. 1985;13:579 –595. 26. Schwab-Stone ME, Shaffer D, Dulcan MK, et al. Criterion validity of the NIMH Diagnostic Interview Schedule for Children version 2.3 (DISC-2.3). J Am Acad Child Adolesc Psychiatry. 1996;35: 878 – 888. 27. Cohen P, Velez N, Kohn M, et al. Child psychiatric diagnosis by computer algorithm: theoretical issues and empirical tests. J Am Acad Child Adolesc Psychiatry. 1987;26:631– 638. 28. Barlow SE, Dietz WH. Obesity evaluation and treatment: expert committee recommendations. Pediatrics. 1998;102:e29. 29. Must A, Anderson SE. Body mass index in children and adolescents: considerations for population-based applications. Int J Obes. 2006;30:590 –594. 30. Ogden CL, Kuczmarski RL, Flegal KM, et al. Centers for Disease Control and Prevention 2000 growth charts for the United States: improvements to the 1977 National Center for Health Statistics version. Pediatrics. 2002;109:45– 60. 31. Wiener JMJ. Historical overview of childhood and adolescent psychopharmacology. In: Wiener JM, ed. Diagnosis and Psychopharmacology of Childhood and Adolescent Disorders. New York: John Wiley & Sons; 1985:3–50. 32. Singer JD, Willett JB. Modeling discontinuous and nonlinear change. In: Singer JD, Willett JB, eds. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press; 2003:189 –242. 33. Buka SL, Monuteaux M, Earls F. Epidemiology of child and adolescent mental disorders. In: Tsuang MT, Tochen M, eds. Textbook in Psychiatric Epidemiology. 2nd ed. New York: Wiley-Liss Inc; 2002:629 – 655. 34. Costello EJ, Mustillo S, Erkanli A, et al. Prevalence and development of psychiatric disorders in childhood and adolescence. Arch Gen Psychiatry. 2003;60:837– 844. 35. Guo SS, Chumlea WC. Tracking of body mass index in children in relation to overweight in adulthood. Am J Clin Nutr. 1999;70: 145S–148S. 36. Brener ND, McManus T, Galuska DA, et al. Reliability and validity of self-reported height and weight among high school students. J Adolesc Health. 2003;32:281–287. 37. Stewart AL. The reliability and validity of self-reported weight and height. J Chronic Dis. 1982;35:295–309.