Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address Please contact the publisher to request reinstatement.
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
January 2008

Subjective Social Status in the School and Change in Adiposity in Female AdolescentsFindings From a Prospective Cohort Study

Author Affiliations

Author Affiliations: Departments of Society, Human Development, and Health (Ms Lemeshow and Dr Kawachi) and Epidemiology (Dr Colditz), Harvard School of Public Health; Channing Laboratory, Department of Medicine, Brigham and Women's Hospital, and Harvard Medical School (Ms Fisher and Drs Berkey and Colditz); and Department of pediatrics, Tufts-New England Medical Center and the Floating Hospital for Children (Dr Goodman), Boston, Massachusetts; New York City Department of Health and Mental Hygiene, Bureau of Tobacco Control, New York (Ms Lemeshow); and Alvin J. Siteman Cancer Center, Washington University School of Medicine, St Louis, Missouri (Dr Colditz).


Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008

Arch Pediatr Adolesc Med. 2008;162(1):23-28. doi:10.1001/archpediatrics.2007.11

Objective  To determine whether subjective social standing in school predicts a change in body mass index (BMI) in adolescent girls during a 2-year period.

Design  Prospective cohort study.

Setting  Self-report questionnaires from a community-based population of adolescent girls living across the United States from 1999 to 2001.

Participants  Of 5723 girls aged 12 to 18 years participating in the Growing Up Today Study (GUTS), adequate information was available for 4446 (78%), who provided the analytic sample.

Main Exposure  Low subjective social status in the school.

Main Outcome Measures  Change in BMI between 1999 and 2001 and multivariable odds ratio for a 2-U increase in BMI in girls with low subjective social status in the school compared with girls with higher subjective social status in the school.

Results  After adjusting for age, race/ethnicity, baseline BMI, diet, television viewing, depression, global and social self-esteem, menarche, height growth, mother's BMI, and pretax household income, adolescent girls who placed themselves on the low end of the school subjective social status scale had a 69% increased odds of having a 2-unit increase in BMI (odds ratio, 1.69; 95% confidence interval, 1.10-2.60) during the next 2 years compared with other girls.

Conclusion  Higher subjective social standing in school may protect against gains in adiposity in adolescent girls.

Between 1999 and 2004, the prevalence of overweight in girls in the United States increased significantly from 14% to 16%.1 The economic burden of childhood overweight is 3 times higher than it was 20 years ago, approximately $127 million per year.2 Children who are overweight experience many health complications but perceive the most immediate consequence of overweight to be social discrimination.3 To lessen this health and economic burden, it is important to identify factors that contribute to excess weight gain and the development of obesity.

There is evidence that social and emotional factors such as depression and low self-esteem and self-perception contribute to the obesity burden in adolescents.48 It has been argued that subjective perceptions of relative ranking may be more important determinants of health than objective indicators such as income, education, or occupation.5 One way subjective perception has been measured is with the Subjective Social Status (SSS) Scale, which was developed to address perceived placement within the social hierarchy among young adults using a visual scale. The scale uses a picture of a 10-rung ladder to represent both US society and the school community. For the school community, adolescents place themselves on the ladder according to where they believe they stand in relation to their classmates.

The objective of this analysis was to determine whether placement on the school SSS Scale predicted an increase in body mass index (BMI) (calculated as the weight in kilograms divided by the height in meters squared) in adolescent girls between 1999 and 2001. We anticipated that low placement on the school SSS Scale in 1999 would predict an increase in BMI between 1999 and 2001.


Established in 1996, the Growing Up Today Study (GUTS) cohort consists of 9039 girls and 7843 boys. All of these children have mothers who are participants in the established Nurses' Health Study II cohort. Details of the recruitment of the GUTS cohort are described elsewhere.9 At enrollment, the children in GUTS were 9 to 14 years of age and 93% were white; they resided across all 50 US states. GUTS collects information on a variety of adolescent behaviors but was originally designed to assess determinants of adolescent weight change, including dietary intake and physical activity and inactivity. Data were collected approximately annually via questionnaires mailed directly to the participants. This study was approved by the human subjects committees at the Harvard School of Public Health and Brigham and Women's Hospital.

Our analysis focused specifically on the association between SSS in the school and change in BMI in female adolescents. We limited our sample to female adolescents because we anticipated that the association between school SSS and change in BMI would be different between boys and girls and would be stronger in girls.4

Of 9039 girls enrolled in the GUTS cohort in 1996, 5723 (63%) returned both the 1999 and 2001 questionnaires. We excluded 383 girls for whom height or weight data were missing or who had an outlying BMI, as described elsewhere.10 We further excluded 461 girls with missing school SSS Scale data; 230 with missing data on their mother's BMI; 118 with missing diet data; 83 with missing data on other variables of interest including race/ethnicity, television viewing, depression, and self-esteem; and 2 who were outside the age range of 12 to 18 years. Our final sample for this analysis included 4446 adolescent girls aged 12 to 18 years in 1999, 78% of our baseline sample.


We assessed adiposity by computing BMI from self-reported heights and weights in 1999 and 2001. Body mass index is a useful tool for assessing weight in relation to stature in children because it provides a reference for adolescents that can be used beyond puberty, it compares well with laboratory measurements of body fat, and it can be used to track body size throughout life.11 Previous studies suggest high validity of self-reported heights and weights in boys and girls aged 10 to 17 years (0.62 < r < 0.98)12 and of BMIs computed from self-reported heights and weights (r = 0.92) in children in grades 7 through 12.13 Body mass index values higher than 3 SDs from the GUTS age-specific means (log scale) were excluded as outliers.10

Many studies examining adolescent BMI have used the Centers for Disease Control and Prevention growth z scores to measure change over time. However, recent research has found that analyzing change in BMI is preferable to using change in z scores.14 Before computing change in BMI from 1999 to 2001 (year 2001 minus year 1999), we subtracted from each girl's BMI the 50th percentile BMI for girls of the same age on the Centers for Disease Control and Prevention BMI charts. This eliminated typical BMI increases attributable to age during the 2-year period from the subsequently computed BMI changes. We then dichotomized BMI change into an increase of 2 U or more vs less than 2 U of gain (including BMI loss). On average, a 2-U increase in BMI corresponds to a 5-kg weight gain in girls in this age range, assuming no height growth. This BMI gain is larger than what would be considered normal according to the Centers for Disease Control and Prevention charts and, therefore, reflects substantial increases in body adiposity during 2 years in a group of adolescent girls. We dichotomized BMI to provide a simple and clinically meaningful indicator of excessively large increases in adiposity.


In 1999, Goodman et al5 modified an adult version of the SSS Scale to be applicable to adolescents and validated it in the GUTS cohort. The instrument had excellent 2-month test-retest reliability (r = 0.79), which suggests that adolescents are consistent in placing themselves within their immediate social environment.5 The question on the 1999 questionaire states: “At the top of the ladder are the people in your school with the most respect and the highest standing. At the bottom are the people who [sic] no one respects and no one wants to hang around with. Where would you place yourself on the ladder?” The school SSS Scale first appeared on the GUTS questionnaire in 1999. Because of the prospective nature of the analysis, we used school SSS placement in 1999 to predict adiposity change up to 2001.

We considered using school SSS as a continuous variable in the model, but it did not satisfy logistic model assumptions. Furthermore, a priori, we expected an effect in girls who placed themselves on the low end of the school SSS Scale. We, therefore, compared girls who placed themselves at or below 4 on the scale with girls who placed themselves at or above 5. This cutoff provided an adequate number of girls with very low school SSS.


Baseline BMI and race/ethnicity were kept, a priori, in the multivariable models regardless of statistical significance. We included baseline BMI in all analyses to account for regression to the mean: a girl with a very low BMI is less likely to subsequently have a large BMI decline, and, likewise, a girl with a very high BMI is more likely to have a subsequent BMI decline rather than a large BMI increase. Race/ethnicity (self-reported) was included because Goodman et al4 found that black girls had significantly lower school SSS placements and higher BMIs compared with white girls.4 We included race/ethnicity as a binary variable (white vs nonwhite).

Because diet was not assessed in 1999 or 2001, diet quality from 1998 was used in the analysis. Diet quality was assessed by adding individual scores for different food groups ranging from vegetables to fried foods. The scores ranged from 0 to 100, with 100 representing the highest quality diet.15 Diet was dichotomized into lower vs higher quality (≤50 vs >50).

To assess television viewing in 1999, girls selected their usual number of viewing hours separately for weekdays and weekends. Hours watched per week were calculated by adding up weekend and weekday viewing. Total hours of television viewing per week was dichotomized into 14 hours or less vs more than 14 hours. Hours per week of physical activity, also reported in 1999, were calculated by computing the total number of hours spent per week in 17 activities and team sports outside of gym class. Total physical activity hours per week ranged from 0 to 40.10

We used a 6-point Likert-type scale derived from the McKnight Risk Factor Survey to measure depression. The scale has good reliability (Cronbach α = 0.73); scores ranged from 6 to 30, with higher scores representing fewer depressive symptoms.5 Global (overall) and social self-esteem were assessed using a modified version of the Harter Self-Perception Profile for Children. Modifications were based on extensive pilot testing in students in public schools in Salem, Massachusetts.16 Self-esteem scores ranged from 6 to 18 for each domain, with higher scores indicating higher levels of self-esteem.5 Global self-esteem score was dichotomized into low vs high (≤13 vs >13).

Girls who responded yes to the question, “Have you started having menstrual periods?” in 1996, 1997, 1998, or 1999 were defined as having reached menarche. Height growth between 1999 and 2001 was calculated by subtracting height in 1999 from height in 2001. Girls with height more than 3 SDs from the mean were excluded as outliers, and girls whose height decreased by more than 2.54 cm between 1999 and 2001 were excluded on the basis of implausibility.10

Mother's BMI in 1999 was calculated the same way as for their daughters. We examined the mother's pretax household income and the educational achievement level of the mother's spouse to determine whether a relationship between school SSS and BMI change existed even after accounting for objective indicators of socioeconomic status. We included a missing category for the 791 girls who were missing income data. Mother's pretax household income was dichotomized into less than $50 000 vs $50 000 or more, and educational achievement level of the mother's spouse was categorized as high school or less, some college, and more than college.


Logistic regression was used to determine whether placement on the school SSS Scale was associated with BMI change in 4446 adolescent girls between 1999 and 2001. Statistical analyses were performed using commercially available software (SAS version 8.2; SAS Institute Inc, Cary, North Carolina).

Before modeling, variables were examined for linearity in the logit. Each continuous variable was first divided into quartiles, dummy coded, and placed in a logistic regression using BMI change (binary, as described in the “Covariates” section) as the main outcome of interest. These β coefficients were then plotted against the midpoints of each quartile. If the plot suggested that the variable increased or decreased linearly, the variable remained as a continuous variable in the model. Several of the variables were not linear in the logit and had to be transformed or recategorized. Both age and baseline BMI seemed to have parabolic relationships with change in BMI and, thus, were transformed to include squared terms in the analysis. We established 3 categories for educational achievement level of the mother's spouse and dichotomized the other variables that needed transformation to maintain a simple, easily interpretable model. Social self-esteem, depression, height growth, mother's BMI, and hours of physical activity per week remained as continuous variables.

We examined potential confounders in age-adjusted analyses to determine whether they were significantly associated with BMI change. Regardless of statistical significance, baseline BMI and race/ethnicity were kept, a priori, in the multivariable models (see “Covariates” section). Other covariates were included in the multivariable analysis if they were significantly associated with BMI change at the P < .10 level in the age-adjusted analysis.

Depression, global and social self-esteem, hours of television viewing, diet, height growth, menarche status, and mother's BMI and pretax household income were all significant at the 0.10 level in the age-adjusted analyses. Educational achievement level of the mother's spouse and hours of physical activity per week were not included in the multivariable analyses because they were not significantly associated with a BMI increase in the age-adjusted models.


The mean BMI in adolescent girls was 20.8 in 1999 and 22.1 in 2001. Between 1999 and 2001, 520 girls (11.7%) had at least a 2-U increase in BMI. In 1999, 182 girls (4.1%) placed themselves on the low end of the school SSS Scale and, among these girls, 35 (19.2%) had at least a 2-U increase in BMI. Among girls who placed themselves on the higher end of the scale, 485 (11.4%) had at least a 2-U increase in BMI (Table 1).

Table 1. 
Image not available
Baseline (1999) Characteristics According to School SSS in Girls in the Growing Up Today Study

Adolescent girls who placed themselves on the low end of the school SSS Scale had a significant 69% increased odds of having a 2-U increase in BMI (odds ratio, 1.69; 95% confidence interval, 1.10-2.60) after adjusting for age, age squared, baseline BMI, baseline BMI squared, depression, global and social self-esteem, television viewing, diet, height growth, menarche status, mother's BMI, and pretax household income (Table 2).

Table 2. 
Image not available
Association of 2-U Increase in BMI During 2 Years and School SSS in 4446 Girls in the Growing Up Today Studya

Based on these prospective data, our analysis shows that low SSS in the school community predicts a significant increase in the odds of gaining excess adiposity during 2 years in adolescent girls. These results are consistent with cross-sectional findings reported by Goodman and colleagues. One study5 using GUTS data reported a 9% decreased odds of overweight in girls with higher school SSS (odds ratio, 0.91; 95% confidence interval, 0.87-0.97),4 and another study using data from the Princeton School District in Greater Cincinnati reported a 22% increased odds of overweight in girls with lower school SSS (odds ratio, 1.22; 95% confidence interval, 1.13-1.32).4 The inference of cause and effect in these 2 studies was limited because the study designs were cross-sectional.

Our results are also consistent with previous research that found associations between social and emotional factors and overweight. One national longitudinal study reported a significant relationship between depression and BMI,6 a cross-sectional study found a significant inverse relationship between self-esteem and overweight,7 and another cross-sectional study found a significant relationship between overweight and depression, self-esteem, and school/social functioning in adolescents aged 12 to 14 years but not in older adolescents.8

Strengths of our analysis include a prospective study design, which enabled us to measure change in BMI over time while accounting for growth and maturation (height growth and menarche status). The unique nature of the GUTS data enabled us to link to the Nurses' Health Study II data set to combine data from mothers' reports of height, weight, household income, and educational achievement level of the mother's spouse. Although this observational study cannot prove causation, it provides evidence linking low SSS in the school community to a subsequent increase in BMI. Despite controlling for many covariates, residual and unmeasured confounding are still possible. The self-reported GUTS data are the greatest potential source of misclassification. However, self-reported data from other major studies of adolescent health, including the Youth Risk Behavior Surveillance System, the Longitudinal Survey of Adolescent Health, and the National Longitudinal Survey of Youth have reasonable validity and reliability.13,1720

The GUTS cohort includes participants from across the United States, but it does not represent a random sample of US adolescents. Participants are predominately white (93%) and their mothers hold nursing degrees. However, these potential limitations in generalizability do not negate the internal validity of the study.

Of the 5723 girls who returned questionnaires in 1999 and 2001, adequate information was available for 4446 (78%), who provided the analytic sample. Almost half of the excluded girls had missing information on the main outcome or predictor variables. We found no significant difference in either the difference in mean school SSS rankings between those included and excluded or the percentage difference between girls who had a 2-U BMI change vs those who did not. Despite the apparent lack of difference between those included and excluded, missing data are still a potential source of bias and may limit the power of the study to establish an association.

It is possible that girls who consistently gain too much body fat also experience decreased school SSS. To address this potential issue related to the direction of the association, we included baseline BMI in the model. By keeping baseline BMI in the model, the results of the analysis more likely reflect the prospective effect of school SSS on BMI change over time.

It is important that researchers consider physical, behavioral, environmental, and socioemotional factors that might contribute to the rising prevalence of overweight in adolescents. Previous research suggests that emotional factors such as depression and low self-esteem and self-perception contribute to the burden of overweight in adolescents.68 Our study contributes to this body of literature in that, to our knowledge, it is the first to prospectively evaluate the relationship between SSS in the school community and change in BMI, and our findings suggest that low school SSS may be an important contributor to increases in BMI in girls over time.

Back to top
Article Information

Correspondence: Adina R. Lemeshow, SM, New York City Department of Health and Mental Hygiene, Bureau of Tobacco Control, 2 Lafayette St, 21st Floor, Mailbox CN-18, New York, NY 10007 (

Accepted for Publication: June 15, 2007

Author Contributions: Ms Lemeshow and Dr Colditz had full access to all the study data and take responsibility for the integrity of the data and the accuracy of the data analyses. Study concept and design: Lemeshow and Fisher. Acquisition of data: Goodman, Kawachi, and Colditz. Analysis and interpretation of data: Lemeshow, Fisher, Goodman, Kawachi, and Berkey. Drafting of the manuscript: Lemeshow, Goodman, and Colditz. Critical revision of the manuscript for important intellectual content: Lemeshow, Fisher, Goodman, Kawachi, Berkey, and Colditz. Statistical analysis: Lemeshow, Fisher, Kawachi, Berkey, and Colditz. Obtained funding: Colditz. Study supervision: Kawachi and Colditz.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grant DK46834 from the National Institutes of Health (Dr Colditz).

Additional Contributions: Members of the Growing Up Today Study research group contributed their comments and expertise. We thank the children and their mothers for carefully completing the questionnaire.

Ogden  CLCarroll  MDCurtin  LRMcDowell  MATabak  CJFlega  KM Prevalence of overweight and obesity in the United States, 1999-2004. JAMA 2006;295 (13) 1549- 1555
Quattrin  TLiu  EShaw  NShine  BChiang  E Obese children who are referred to the pediatric endocrinologist: characteristics and outcome. pediatrics 2005;115 (2) 348- 351
 Overweight and obesity: health consequences. Accessed May 16, 2006
Goodman  EAdler  NEDaniels  SRMorrison  JASlap  GBDolan  LM Impact of objective and subjective social status on obesity in a biracial cohort of adolescents. Obes Res 2003;11 (8) 1018- 1026
Goodman  EAdler  NEKawachi  IFrazier  ALHuang  BColditz  GA Adolescents' perceptions of social status: development and evaluation of a new indicator. pediatrics 2001;108 (2) E31 Accessed July 17, 2006
Goodman  EWhitaker  RC A prospective study of the role of depression in the development and persistence of adolescent obesity. pediatrics 2002;110 (3) 497- 504
Mirza  NMDavis  DYanovski  JA Body dissatisfaction, self-esteem, and overweight among inner-city Hispanic children and adolescents. J Adolesc Health. 2005;36 ((3)) 267.e17-e22 Accessed July 17, 2006Article
Swallen  KCReither  ENHaas  SAMeier  AM Overweight, obesity, and health-related quality of life among adolescents: the National Longitudinal Study of Adolescent Health. pediatrics 2005;115 (2) 340- 347
Berkey  CSRockett  HField  A  et al.  Activity, dietary intake, and weight changes in a longitudinal study of preadolescent and adolescent boys and girls. pediatrics 2000;105 (4) E56. Accessed July 17, 2006
Berkey  CSRockett  HRGillman  MWColditz  GA One-year changes in activity and in inactivity among 10- to 15-year-old boys and girls: relationship to change in body mass index. pediatrics 2003;111 (4, pt 1) 836- 843
 BMI: body mass index.  Bethesda, MD Centers for Disease Control and Prevention Accessed March 22, 2006
Shannon  BSmiciklas-Wright  HWang  MQ Inaccuracies in self-reported weights and heights of a sample of sixth-grade children. J Am Diet Assoc 1991;91 (6) 675- 678
Goodman  EHinden  BRKhandelwal  S Accuracy of teen and parental reports of obesity and body mass index. pediatrics 2000;106 (1, pt 1) 52- 58
Berkey  CSColditz  GA Adiposity in adolescents: change in actual BMI works better than change in BMI z score for longitudinal studies. Ann Epidemiol 2007;17 (1) 44- 50
Feskanich  DRockett  HRColditz  GA Modifying the Healthy Eating Index to assess diet quality in children and adolescents. J Am Diet Assoc 2004;104 (9) 1375- 1383
Harter  S Manual of the Self-Perception Profile for Children.  Denver, CO University of Denver1985;
Brener  NDCollins  JLKann  LWarren  CWWilliams  BI Reliability of the Youth Risk Behavior Survey Questionnaire. Am J Epidemiol 1995;141 (6) 575- 580
Hornberger  LLRosenthal  SLBiro  FMStanberry  LR Sexual histories of adolescent girls: comparison between interview and chart. J Adolesc Health 1995;16 (3) 235- 239
Johnson  TPMott  JA The reliability of self-reported age of onset of tobacco, alcohol and illicit drug use. Addiction 2001;96 (8) 1187- 1198
Shew  MLRemafedi  GJBearinger  LH  et al.  The validity of self-reported condom use among adolescents. Sex Transm Dis 1997;24 (9) 503- 510