EMPIRICAL ARTICLE

Does Childhood Bullying Predict Eating Disorder

Symptoms? A Prospective, Longitudinal Analysis

William E. Copeland, PhD1* ABSTRACT

Objective: Bullying is a common child-
of eating disorders, and bully-victims had
higher levels of anorexia symptoms. In
Cynthia M. Bulik, PhD2,3,4 hood experience with enduring psychoso- terms of individual items, victims were at
Nancy Zucker, PhD1 cial consequences. The aim of this study risk for binge eating, and bully-victims
was to test whether bullying increases had more binge eating and use of vomit-
Dieter Wolke, PhD5 risk for eating disorder symptoms. ing as a compensatory behavior. There
Suzet Tanya Lereya, PhD5 was little evidence in this sample that
Method: Ten waves of data on 1,420
Elizabeth Jane Costello, PhD1 participants between ages 9 and 25 were
these effects differed by sex. Childhood
bullying status was not associated with
used from the prospective population- increased risk for persistent eating disor-
based Great Smoky Mountains Study. der symptoms into adulthood (ages 19,
Structured interviews were used to assess 21, and 25).
bullying involvement and symptoms of
anorexia nervosa and bulimia nervosa as Discussion: Bullying predicts eating dis-
well as associated features. Bullying order symptoms for both bullies and vic-
involvement was categorized as not tims. Bullying involvement should be a
involved, bully only, victim only, or both part of risk assessment and treatment
bully and victim (bully-victims). planning for children with eating prob-
lems. V
C 2015 Wiley Periodicals, Inc.
Results: Within childhood/adolescence,
victims of bullying were at increased risk Keywords: eating behavior; body
for symptoms of anorexia nervosa and image; bullying; epidemiology; child-
bulimia nervosa as well as associated fea- hood; adolescence; psychosocial;
tures. These associations persisted after stress
accounting for prior eating disorder
symptom status as well as preexisting (Int J Eat Disord 2015; 48:1141–1149)
psychiatric status and family adversities.
Bullies were at increased risk of symp-
toms of bulimia and associated features

Introduction
Bullying involves targeting an individual perceived to
be vulnerable for repeated mistreatment.1 Bullying is
Accepted 14 August 2015
Additional Supporting Information may be found in the online
a common childhood experience with enduring
version of this article. social and psychological consequences (e.g., Refs.
Supported by grants MH63970, MH63671, MH48085, and 2–4). Victims of bullying are at increased risk of phys-
MH080230 from the National Institute of Mental Health; grant
DA/MH11301 from the National Institute on Drug Abuse; NARSAD ical health problems,5 behavior and emotional prob-
(Early Career Award to W.E.C.); the William T. Grant Foundation; lems,6 suicidality,7 psychotic symptoms,8 and poor
and grant ES/K003593/1 from the Economic and Social Research
school achievement.9 Victims who also bully others
Council (ESRC) in the United Kingdom.
*Correspondence to: William E. Copeland, Department of Psychi- (so-called bully-victims) have the worst outcomes.2,3
atry and Behavioral Sciences, Duke University Medical Center, Box Despite widespread evidence that bullying negatively
3454, Durham, NC 27710. E-mail: [email protected]
1
Department of Psychiatry and Behavioral Sciences, Duke
affects childhood functioning, few studies have
University Medical Center, Durham, North Carolina explored whether the negative effects of bullying
2
Department of Psychiatry, University of North Carolina at extend to eating problems.
Chapel Hill, Chapel Hill, North Carolina
3
Department of Nutrition, University of North Carolina at Eating behavior may be affected by bullying for
Chapel Hill, Chapel Hill, North Carolina
4
several reasons. First, bullying may increase nega-
Department of Medical Epidemiology and Biostatistics,
Karolinska Institutet, Stockholm, Sweden
tive perceptions of one’s body either directly via
5
Department of Psychology and Division of Mental Health and teasing about one’s weight/appearance10 or indi-
Well-Being, University of Warwick, Coventry, United Kingdom rectly through its effects on general self-esteem
Published online 4 September 2015 in Wiley Online Library
(wileyonlinelibrary.com). DOI: 10.1002/eat.22459 and emotional problems.11 Most studies of body
VC 2015 Wiley Periodicals, Inc. dissatisfaction and disordered eating have focused

International Journal of Eating Disorders 48:8 1141–1149 2015 1141

COPELAND ET AL.

on teasing that is appearance or weight-related.12 same time, the need to solidify one’s status by
These studies suggest moderate associations of harming or humiliating others may be indicative of
teasing with body dissatisfaction, dietary restraint, low self-esteem or poor self-image.27 Furthermore,
and bulimic behaviors.13 The associations, how- bullies commonly report regret following bullying
ever, are attenuated in longitudinal studies that incidents, even if this is not sufficient to change
account for prior levels of body dissatisfaction/dis- their future behavior.28 Given these findings, we
ordered eating, and almost no studies have predict bullies may also display increased disor-
accounted for other preexisting family and individ- dered eating.
ual factors that increase risk for being teased in the
first place.14 As such, questions remain about the
role of childhood teasing in eating disorder-related Method
problems.
Participants
Bullying involves a range of forms of peer victim-
ization beyond weight/appearance-related teasing The Great Smoky Mountains Study (GSMS) is a longi-
tudinal study of the development of psychiatric disorders
(e.g., overt aggression, social exclusion, and rumor
and the need for mental health services in rural and
mongering).15 Studies of teasing that is not appear-
urban youth.29 A representative sample of three cohorts
ance/weight related suggest negative effects on
of children, aged 9, 11, and 13 at intake, was recruited
body esteem both cross-sectionally12,16 and over
from 11 counties in western North Carolina. All children
time,17,18 with stronger effects in girls than in boys.
scoring above a predetermined cut point (the top 25% of
Studies that used broad measures of bullying also
the total scores) on a screener, plus a 1 - in - 10 random
suggested associations with disordered eating.19,20
samples of the remaining 75% of the total scores, were
Bullying often co-occurs with disordered eating,
recruited for detailed interviews. This oversampling
but it is not at all clear whether it is a risk factor or
approach allows us to estimate prevalence of common
predicts such problems.21
psychiatric disorders. By applying weights inversely pro-
The aim of this study was to test whether broadly portional to selection probability, results are unbiased
defined bullying—rather than specific appearance and representative of the population.30 About 8% of the
or weight-related teasing—is a risk factor for eating area residents and the sample are African American,
disorder symptoms. The study uses a representa- <1% are Hispanic, and 3% are American Indian. Of all
tive sample repeatedly assessed to test whether the participants recruited, 80% (N 5 1,420) agreed to par-
bullying increases risk for eating disorder symp- ticipate. Participants were assessed annually upto age 16
toms after accounting for preexisting levels of eat- and then again at ages 19, 21, and 25. Across all waves,
ing disorder symptoms. This design allows us to participation rates averaged 84% (range: 74%–94%).
test whether observed effects on eating disorder
symptoms are a direct effect of bullying or medi- Procedures
ated by the changes to emotional symptoms that The parent (biological mother for 83% of interviews)
are known to increase in response to bullying. We and participant were interviewed by trained interviewers
hypothesize that victims will be at increased risk separately until the participant was 16, and participants
for eating disorder symptoms and that a portion of only thereafter. Before the interviews began, parent and
this effect will be accounted for by increases in child signed informed consent approved by the Duke
negative effect. This hypothesis is based on the University Medical Center Institutional Review Board.
previous evidence on the effects of weight/body- Each parent and child received an honorarium for their
related teasing,13,22 associations between bullying participation.
and eating/weight outcomes,18,19 and the effects of
bullying on a range of related childhood/adoles- Assessment
cent outcomes.5,6,23 Finally, as the prevalence esti- Childhood Bullying Involvement. Bullying involvement
mates of both bullying involvement and eating required the child (or another child in the case of bullies)
disorder outcomes vary by sex,3,15,24,25 we will test to be a particular and preferred object of mockery, physi-
whether sex differences exist in any observed cal attacks, or threats. At each assessment between ages
associations. 9 and 16, the child and parent reported on whether the
Finally, few studies to date have considered the child had been bullied/teased multiple times or bullied
perpetrators of bullying—the bullies themselves. others in the 3 months immediately prior to the inter-
Bullies have been reported to enjoy high social view as part of the Child and Adolescent Psychiatric
standing and low levels of emotional distress while Assessment (CAPA).31 Participants were categorized as
having more conduct-related problems.26 At the victims, bullies, both (i.e., bully-victims) or neither.

1142 International Journal of Eating Disorders 48:8 1141–1149 2015

 EATING DISORDERS AND BULLYING

A bully-victim group was included rather than simply ior and eating disorders in the Child and Adolescent
modeling joint exposure with an interaction term as Psychiatric Assessment (ages 9–16) and its upward exten-
prior work suggests that this group has a different pat- sion: the Young Adult Psychiatric Assessment (ages 19, 21,
tern of correlates and outcomes than either bullies or and 25). The CAPA is a structured diagnostic interview
victims.2,3,32–34 Supplemental table 1 provides definitions that was used in this epidemiologic study to obtain prev-
for these categories as well as interview probes uses. alence estimates of various common childhood disor-
Being bullied or bullying others was counted if reported ders. The eating disorder module and associated glossary
by either the parent or the child. If the informant are included in the appendices. An item was counted as
reported that the participant had been bullied or bullied present if reported by either parent or child or both, as is
others, then the informant was asked separately how standard in child and adolescent epidemiological stud-
often the bullying occurred in the prior 3 months in the ies, approximating the process of combining information
following three settings: home, school, and the commu- from multiple informants in clinical practice. Two-week
nity. The focus in this article is on peer bullying in the test–retest reliability of CAPA diagnoses in children aged
school context only, as this is the most common setting 10 through 18 is comparable with that of other structured
for bullying.3 Weight/appearance-related teasing was not child psychiatric interviews.37,38 Construct validity as
assessed separately from bullying. Parent and child judged by 10 different criteria, including comparison to
agreement (j 5 0.24) was similar to that of other bullying other interviews and ability to predict mental health
measures.8 Parents were more likely to report their child service use, is good to excellent.31
was either a bully or victim than the child themselves The CAPA/YAPA eating disorder module assesses all
(62.2% and 55.2% of total cases, respectively). We have DSM-III-R and IV symptoms of anorexia nervosa and
found similar associations between parent- and child- bulimia nervosa. Associated eating disorder related fea-
reporting bullying involvement and long-term outcomes tures are also assessed. This included increased appetite
in prior work.2,3,35 or decreased appetite (defined as a definite change in
Childhood Bullying-Related Covariates. To clarify that food intake because of an appetite change that has per-
bullying involvement is an independent risk factor for sisted for at least a week) and a preoccupation with eat-
eating disorder symptoms, it is necessary to account for ing/food (defined as an unusual and excessive amount of
preexisting family and individual factors that might pre- time spent thinking or worrying about food and eating).
dict bullying involvement and eating disorder symptoms. In GSMS, it was rare for participants to meet full criteria
Childhood psychiatric and family hardships variables for an eating disorder.39 As such, the focus of this analy-
(except where indicated) were assessed by parent and sis is on individual symptoms of anorexia nervosa and
self-report using the Child and Adolescent Psychiatric bulimia nervosa and associated features. Two items were
Assessment (CAPA).31 Childhood psychiatric variables too rare to study individually: Amenorrhea (anorexia)
included any anxiety disorder, any depressive disorder, and use of medications to control weight (bulimia). Indi-
any behavioral disorder (conduct disorder, attention- vidual items were summed into scales for anorexia, buli-
deficit hyperactivity disorder, and oppositional defiant mia, and associated features. Finally, height and weight
disorder), and any substance abuse or dependence. See measures were collected at each observation, and over-
Ref. 36 for additional details. Four types of family hard- weight and obesity status was calculated using conven-
ships were assessed: low socioeconomic status (SES; tional BMI cutoffs.
including family poverty, low parental educational
attainment, and low parental occupational prestige), Analytic Framework
unstable family structure (indicators include single par- All models used SAS PROC GENMOD to run weighted
ent, divorce, parental separation, presence of step- regression models with robust variance (sandwich type)
parent, or change in parent structure), family dysfunc- estimates derived from generalized estimating equations
tion (including inadequate parental supervision, domes- to adjust the standard errors for the stratified design and
tic violence, parental over involvement, maternal repeated observations. Sampling weights were applied to
depression, marital relationship characterized by apathy, ensure that results are representative of the population
indifference, or high conflict, and high conflict between from which the sample was drawn. Odds/means ratios
parent and child), and maltreatment (including physical (OR/MR), 95% confidence intervals (CI), and p-values
abuse, sexual abuse, and parental neglect). Additional are provided for all analyses.
details are provided in Supplemental table 2. All bully- The primary analysis tested whether recent bullying
related covariates have been shown to be related to both involvement (the participant’s status at the most recent
bully and victim status in previous work.3 observation) predicted eating disorder symptom/features
Eating Disorder Symptoms and Associated Features. All within childhood/adolescence (ages 9 and 16). This simple
items were assessed as part of a module on eating behav- model included a dummy variable for comparing the

International Journal of Eating Disorders 48:8 1141–1149 2015 1143

COPELAND ET AL.

TABLE 1. Associations between childhood bully-victim groups and eating disorder related cognitions and behaviors
adjusting for prior status
Victims vs. Neither Bullies vs. Neither Bully-Victims vs. Neither
Neither Bully Only Victim Only Bully-Victim
(n 5 5,816) (n 5 204) (n 5 616) n 5 54) OR/MR OR/MR OR/MR
(%) (%) (%) (%) (95% CI) p-Values (95% CI) p-Values (95% CI) p-Values
Any anorexia symptom 5.6 8.6 11.2 22.8 1.8 (1.1–3.0) .02 1.1 (0.3–3.6) .87 4.6 (1.8–11.4) .001
Underweight for 2.2 1.1 4.0 0.0 2.2 (1.0–5.0) .05 0.8 (0.3–1.9) .62 — —
height/age
Fear of gaining weight 1.5 2.3 6.0 6.0 3.8 (2.0–7.1) <.001 0.9 (0.2–4.3) .88 4.3 (1.6–11.9) .005
Distorted body image 2.6 6.3 3.4 18.1 1.1 (0.5–2.5) .85 1.7 (0.4–7.8) .53 7.6 (2.5–22.4) <.001
Any bulimia symptom 17.6 30.8 27.9 28.4 1.5 (1.1–2.1) .007 1.7 (1.0–2.7) .04 1.7 (0.6–5.2) .34
Binge eating 0.4 0.9 1.6 4.8 4.6 (1.5–14.8) .01 2.3 (0.6–8.8) .22 14.2 (4.0–50.8) <.001
Attempts to cut weight 17.0 29.3 26.1 24.9 1.5 (1.1–2.0) .02 1.6 (1.0–2.6) .07 1.5 (0.4–4.9) .54
Diet 7.8 18.8 13.8 19.8 1.6 (1.0–2.4) .04 2.3 (1.1–4.8) .03 2.9 (0.9–9.3) .08
Exercise 14.1 24.6 22.4 9.9 1.5 (1.1–2.1) .01 1.6 (0.9–2.8) .12 0.5 (0.1–1.9) .33
Vomiting 0.3 0.3 0.2 2.4 0.9 (0.2–3.2) .83 0.5 (0.0–8.5) .63 9.3 (1.7–50.2) .01
Overconcern with body 1.9 5.3 8.0 9.6 3.6 (2.2–6.0) <.001 1.9 (0.3–11.8) .51 5.6 (2.6–12.1) <.001
Any associated features 28.2 43.3 42.5 33.6 1.8 (1.4–2.3) <.001 1.8 (1.1–3.0) .02 1.3 (0.5–2.9) .61
Preoccupied with eating 0.8 4.1 4.9 4.8 6.1 (3.0–12.5) <.001 5.6 (1.3–23.6) .02 6.0 (2.0–17.6) .001
Decreased appetite 7.8 12.9 13.6 15.5 1.9 (1.2–2.9) .003 1.6 (0.7–4.1) .28 2.4 (0.6–10.2) .24
Increased appetite 20.7 28.1 28.0 14.5 1.5 (1.1–2.0) .01 1.4 (0.8–2.3) .22 0.6 (0.2–1.6) .29

Bolded values are significant at p  0.05.

bullying group (bullies, victims, and bully-victims) to a likely to be female (44.2% vs. 50.1%, p 5 .74). Nei-
group not involved in bullying and status on the eating dis- ther victim, bully, nor bully-victim status was
order symptom/feature at the prior observation (typically related to being overweight or obese, but victims
1 year prior). Adjusted models also included covariates for were more likely to be underweight than those
sex, age, and psychiatric status and family adversities at uninvolved in bullying (46.7% vs. 36.1%, p < .001;
the prior observation (low SES, family instability, family results available upon request from first author).
dysfunction, maltreatment, depressive disorders, anxiety
disorders, disruptive behavior disorders, or substance dis- Short-term Associations
orders). A similar series of models tested whether having Table 1 provides prevalence estimates of child-
ever been involved in bullying in childhood or adolescence hood/adolescent (ages 9–16) eating disorder symp-
(ages 9–16) predicted eating disorder related outcomes in toms and associated features for different bullying
young adulthood (ages 19, 21, and 25). groups (Columns 2–5) and associations adjusted
for status on the eating disorder symptom/feature
at the prior observation (Columns 6–11). Not sur-
Results prising, the single best predictor of one’s current
level of eating disorder symptoms was the level of
Descriptive Information
symptoms at the most recent observation (e.g., dis-
A total of 6,674 assessments were completed on torted body image at one observation tended to
the 1,420 participants between ages 9 and 16 years predict a distorted body image at the next observa-
(median of 5 observations per subject). This was tion). By adjusting for prior status, our model is
84.4% of possible interviews during this period. No testing whether recent bullying involvement pre-
bullying involvement was reported at 5,800 person- dicts a change in eating disorder symptoms/fea-
observations, bullying only at 204, being bullied tures from the prior observation.
(victims) at 616, and 54 reported both bullying All bullying groups were at increased risk for
others and being bullied (bully-victim). Neither reporting eating disorder symptoms or associated
bullying role nor eating disorder symptoms were features as compared with those uninvolved in bul-
associated with missing prior or subsequent inter- lying. Victims of bullying were at increased risk for
views (bullies: p 5 .42; victims: p 5 .76; bully-vic- reporting a symptom of anorexia, bulimia, or an
tims: p 5 .10; anorexia symptoms: p 5 .56; bulimia associated feature. Specific symptoms that victims
symptoms: p 5 .38; and associated features symp- were at risk for included fear of gaining weight, fail-
toms: p 5 .79). ing to maintain weight, binge eating, a preoccupa-
Both bullies and bully-victims were less likely to tion with food/eating, and changes in appetite.
be female than those uninvolved in bullying (bul- Bully-victims were only at increased risk for report-
lies: 31.0% vs. 50.1%, p < .001; bully-victims: 22.4% ing an anorexic symptom, but they did have ele-
vs. 50.1%, p < .001). Victims were not more or less vated prevalence of specific bulimic symptoms,

1144 International Journal of Eating Disorders 48:8 1141–1149 2015

 EATING DISORDERS AND BULLYING

TABLE 2. Associations between childhood bully-victim groups and eating disorder symptoms/features adjusted for
sex, age, race, prior levels of eating symptoms/features, preexisting psychiatric status, and family adversitiesa
Victims vs. Neither Bullies vs. Neither Bully-Victims vs. Neither
OR/MR Sig. OR/MR Sig. OR/MR Sig.
(95% CI) p-Values Covariates (95% CI) p-Values Covariates (95% CI) p-Values Covariates
Any anorexia symptom 1.9 (1.2–3.2) .01 1,2,5,8,11 1.3 (0.4–3.2) .64 1,2,5,8,11 5.7 (2.0–16.7) .001 1,2,5,8,11
Underweight for height/age 2.1 (0.9–5.1) .09 1,2,6 0.8 (0.2–2.7) .73 1,2,6 – –
Fear of gaining weight 4.5 (2.5–8.0) <.001 1,8 1.2 (0.4–3.8) .74 1,8 4.9 (1.2–19.3) .02 1,8
Distorted body image 1.1 (0.5–2.3) .89 1,2,5,8 2.1 (0.4–10.0) .35 1,2,5,8 8.2 (2.4–28.5) <.001 1,2,5,8
Any bulimia symptom 1.5 (1.1–2.1) .02 1,2,11 1.8 (1.1–3.0) .03 1,2,11 1.6 (0.5–5.4) .44 1,2,11
Binge eating 3.7 (1.1–12.0) .03 3 2.1 (0.5–8.2) .31 3 11.1 (3.1–40.3) <0.001 3
Attempts to cut weight 1.4 (1.0–2.0) .04 1,2,11 1.7 (1.0–2.9) .05 1,2,11 1.4 (0.4–5.0) .64 1,2,11
Diet 1.5 (1.0–2.4) .07 1 2.5 (1.1–5.6) .02 1 2.9 (0.9–9.9) .08 1
Exercise 1.5 (1.1–2.2) .03 1,2,11 1.7 (0.9–3.1) .10 1,2,11 0.5 (0.1–1.9) .28 1,2,11
Vomiting 0.7 (0.1–4.4) .72 1,3,8,9 1.2 (0.2–8.6) .84 1,3,8,9 42.5 (6.6–271.5) <.001 1,3,8,9
Overconcern with body 3.8 (2.3–6.3) <.001 1,8 2.5 (0.6–10.6) .22 1,8 5.6 (2.1–15.1) <.001 1,8
Any associated features 1.7 (1.3–2.2) <.001 1,2,6 1.7 (1.1–2.8) .03 1,2,6 1.2 (0.6–2.8) .60 1,2,6
Preoccupied with eating 4.4 (2.3–8.4) <.001 1,3,6,9 6.0 (1.4–26.0) .02 1,3,6,9 4.4 (1.2–15.3) .02 1,3,6,9
Decreased appetite 2.2 (1.4–3.5) <.001 1,3 2.1 (0.8–5.2) .12 1,3 3.5 (0.8–14.5) .09 1,3
Increased appetite 1.4 (1.0–1.8) .05 1,2,3 1.2 (0.8–2.0) .39 1,2,3 0.5 (0.2–1.3) .16 1,2,3

Notes: OR 5 odds ratio; 95% CI 5 95% confidence interval.

Bolded ORs significant at p < 0.05.
a
Too few participants used medications to control weight or reported amenorrhea for these analyses. Childhood psychiatric and family hardships and
other covariates: 1 5 sex; 2 5 race; 3 5 age; 4 5 low SES; 5 5 family instability; 6 5 family dysfunction; 7 5 maltreatment; 8 5 depressive disorders;
9 5 anxiety disorders; 10 5 disruptive disorders; 11 5 substance disorder.

namely binge eating, use of vomiting to control tests whether bullying predicts new symptoms in
weight, or overconcern with body shape/weight. previously asymptomatic individuals. The pattern
Finally, bullies were at risk for both bulimic symp- of results is similar to that observed for the entire
toms and associated features. sample.
Including all individual items and bullying
groups, 17 of 34 independent tests of associations Are These Associations Sex-Specific?
were statistically significant. The likelihood of this The most common significant covariate in multi-
occurring by chance is 4.1 3 10214, suggesting this variable models was sex. It is not surprising that
is nonrandom. prevalence estimates of these outcomes vary by
This pattern of associations is suggestive. It is sex, but do the associations between bullying and
necessary, however, to account for preexisting indi- eating disorder related outcomes vary between
vidual and family factors that might increase risk boys and girls? To test for moderation by sex, all
for both bullying involvement and eating disorder
symptoms/features (possible confounders). Table FIGURE 1. Adjusted means scores for sum scales of eating disorder
2 presents results from models testing associations symptoms/features within childhood and adolescence by bully/victims
status. Means are adjusted for prior levels of eating symptoms/features
between bullying and these outcomes adjusted for as well as sex, race, age, and preexisting psychiatric status and family
sex, preexisting individual psychiatric problems adversities. Bars indicate the standard error of the mean. Starred bars
(depression, anxiety, behavior problems, and sub- indicate groups that are different from those uninvolved in bullying
(see Table 2 for means ratios, confidence intervals, and p-values).
stance use), preexisting family adversities (mal-
treatment, family instability, family dysfunction,
and low SES), in addition to the covariates from the
simple models. Associations between bullying
groups and the likelihood of having any symp-
toms/features were largely unchanged. A few indi-
vidual associations were no longer statistically
significant (victims and binge-eating). Figure 1
provides that adjusted mean scores for sum scales
of anorexic and bulimic symptoms and associated
features within childhood/adolescence by bullying
group. Supplemental figure 1 provides adjusted
mean scores for the sum scales when the sample is
restricted to those with no eating disorder symp-
toms at the prior wave. This analysis specifically

International Journal of Eating Disorders 48:8 1141–1149 2015 1145

COPELAND ET AL.

TABLE 3. Tests of depressive symptoms as a mediator of the association between bullying groups and eating disorder
symptoms/features
1. Predictors to 2. Mediator to 3. Predictor to 4. Sobel Test,
Predictor Outcome Mediator, b (SE) Outcome, b (SE) Outcome, b (SE) p-Value
Victim Anorexic symptoms 0.45 (0.09)* 0.35 (0.05)* 0.38 (0.23) <0.001
Victim Bulimic symptoms 0.45 (0.09)* 0.18 (0.03)* 0.35 (0.13)** <0.001
Victim Ass. features 0.45 (0.09)* 0.47 (0.03)* 0.11 (0.08) <0.001
Bullies Bulimic symptoms 0.61 (0.13)* 0.18 (0.03)* 0.36 (0.27) <0.001
Bullies Ass. features 0.61 (0.13)* 0.47 (0.03) 0.15 (0.16) <0.001
Bully-victims Anorexic symptoms 0.85 (0.18)* 0.33 (0.05)* 0.85 (0.43)*** <.001

Models were tested with Poisson regression. Columns numbered 2 and 3 provide results from models in which both predictor and potential mediator
predicted eating outcome status. Models adjusted for demographics, prior levels of eating symptoms/features, prior psychiatric status, and family adver-
sities. Sobel test assesses significance of indirect pathway.
*p  .0001; **p  .01; ***p  .05.

models in Table 2 were rerun including an interac- Long-Term Associations

tion term between sex and bully status. There was Finally, do these apparent effects of bullying on
no evidence to suggest widespread moderation by eating disorder symptoms persist long-term into
sex: the interaction terms were not significant for adulthood? Here, our goal was to test whether bul-
any of the summary scales (nine models) and only lying involvement was a risk factor for problems in
two of the individual indicators (34 models), con- young adulthood beyond one’s childhood status on
sistent with chance findings. eating disorder related measures. Table 4 presents
results predicting sum scales of young adult (ages
19, 21, and 25) eating disorder symptoms and asso-
Are These Associations Mediated by Changes in
ciated features from childhood bullying involve-
Emotional Distress?
ment (ages 9–16). The first set of models is only
We hypothesized that these associations could adjusted for childhood levels of disordered eating
be explained, in part, by the well-established emo- sum scales. The second set adjusts for prior status,
tional sequelae of bullying involvement such as and also childhood psychiatric status and family
anxiety or depressive symptoms. This is particu- adversities. Overall, there was little evidence that
larly the case with victims and bully-victims who bullying involvement is a risk factor beyond child-
are at risk for elevated depression and anxiety.3,6 hood status on eating disorder symptoms.
Five criteria had to be met to demonstrate medi-
ation by either anxiety or depressive symptoms: (1)
the bullying group was associated with the out-
Discussion
come; (2) the bullying group was associated with
mediator (Table 3, Column 1); (3) In models Eating disorders are relatively rare but even sub-
adjusting for the bullying group, the mediator was threshold presentations are associated with signifi-
associated with outcome (Column 2); (4) In models cant morbidity and impairment.24,25 This study
adjusting for mediator, the association between the used a community sample repeatedly assessed to
bully group and outcome was either no longer stat- test whether symptoms of eating disorders increase
istically significant or attenuated (Column 3); and following bullying involvement. All bullying
(5) A statistically significant indirect path existed groups—victims, bullies, and bully-victims—saw
between the bully group variable and the eating increases in at least one type of eating disorder
disorder outcome through the mediator, as meas- symptom even after accounting for prior eating
ured by the Sobel Test40 (Column 4). Table 3 and problems and preexisting psychiatric status and
supplemental table 3 tested mediation of the asso- family adversities. Victims displayed the most per-
ciations between bully group and different symp- vasive pattern with increases in anorexic and
tom scales by depressive and anxiety symptoms, bulimic symptoms as well as associated features.
respectively. In every case, the indirect path from Bully-victims had high prevalence of both binge
bullying status to the eating disorder symptom eating and vomiting, and victims were at increased
scale was statistically significant. This is consistent risk for binge eating. In all cases, there was evi-
with a common pathway by which bully status may dence that these associations might be mediated by
affect eating disorder symptoms through emo- increased depressive and anxiety symptoms. Child-
tional symptoms. hood/adolescent bullying involvement, however,

1146 International Journal of Eating Disorders 48:8 1141–1149 2015

 EATING DISORDERS AND BULLYING

TABLE 4. Associations between childhood bully-victim groups and young adult eating disorder related behaviors and
cognitions (ages 19, 21, and 25)a
Victims vs. Neither Bullies vs. Neither Bully-Victims vs. Neither
MR (95% CI) p-Values MR (95% CI) p-Values MR (95% CI) p-Values
Total anorexia symptoms
Adjusted for childhood status 0.9 (0.5–1.6) .68 1.4 (0.4–4.3) .58 1.0 (0.6–1.6) .88
Adjusted for psychiatric 0.9 (0.5–1.7) .71 1.7 (0.7–4.1) .29 1.1 (0.7–1.9) .69
status and adversities
Total bulimia symptoms
Adjusted for childhood status 0.8 (0.6–1.2) .38 1.2 (0.7–2.1 .59 0.8 (0.4–1.7) .60
Adjusted for psychiatric 1.0 (0.7–1.5) .90 1.6 (0.9–3.0) .12 1.3 (0.5–3.4) .57
status and adversities
Total associated features
Adjusted for childhood status 1.3 (1.0–1.8) .08 1.3 (0.7–2.5) .42 1.7 (1.2–2.6) .01
Adjusted for psychiatric 1.3 (1.0–1.8) .09 1.3 (0.7–2.6) .39 1.5 (0.9–2.6) .09
status and adversities

Notes: OR 5 means ratio; 95% CI 5 95% confidence interval.

Of the total young adult observations, 1,993 involved participants never involved in bullying, 243 involved participants that were bullies only, 774
involved victims only, and 205 involved bully-victims. Bolded MRs significant at p < 0.05.
a
Childhood psychiatric and family hardships and other covariates include sex, low SES, family instability, family dysfunction, maltreatment, depressive
disorders, suicidality, anxiety disorders, disruptive disorders, substance disorders, and age.

did not increase risk for eating problems in young surprising and novel. Bullies, like victims and
adulthood. bully-victims, had increased eating disorder out-
The findings in relation to victims and bully- comes, despite being perpetrators and presumably
victims should not be surprising given previous seeing bullying as way to access resources. More
evidence on the effects of weight/body-related surprisingly, observed associations were not atten-
teasing,13,22 associations between bullying and eat- uated in models accounting for preexisting status,
ing/weight outcomes,18,19 and the effects of bully- suggesting the experience of bullying itself may
ing on a range of related childhood/adolescent affect subsequent behaviors. This presents a few
outcomes.5,6,23 At the same time, the strength of nonexclusive explanations. First, weight loss efforts
our findings rests on the following features of this may be strategies to maintain the social influence/
study. First, the repeated assessments across child- dominance acquired through bullying. If bullying
hood and adolescence allowed us to predict is viewed as a tool of social control, then eating
changes in eating disorder outcomes associated behaviors and cognitions may be an extension of
with recent bullying involvement. As such, all anal- these efforts to master oneself and one’s environ-
yses accounted for the most potent predictor of ment. Second, the experience of criticizing/teasing
current behavior, namely, past behavior. Second, others (possible weight or body-related) may sensi-
the prospective design also allowed us to account tize bullies to their own physical attributes and
for preexisting individual and family factors that shortcomings. Third, bullies may experience regret
might increase risk for both bullying involvement or guilt following bullying incidents, and this con-
and eating disorder outcomes. This allowed us to tributes to impulsive eating behaviors and cogni-
make the strong inferences about bullying playing tions. Finally, bullying others may be triggered by
predicting subsequent eating disorder outcomes experiences (unmeasured here) that adversely
within the context of an observational study. Our affect one’s self-image. At the minimum, these
distinction between victims and bully-victims findings suggest that the experience of bullies may
allowed us to identify a small group at high risk of be more complex than previously suggested, and
problems that are of greatest clinical concern simple notions of “hale and hearty” perpetrators
(binge eating and vomiting). Finally, studying this are incomplete.
in the context of a broader study of emotional and Two other surprising findings were the lack of
behavioral functioning allowed us to test mood evidence of sex-specific associations and the lack
changes as a candidate mediator of the observed of long-term effects. In this study, bullies and
associations. Together, these considerations bully-victims were more likely to be boys, and it is
allowed us to build on our understanding of how well established that eating problems, weight loss
being bullied affects eating disorder outcomes. efforts, and negative body-image are more com-
The finding of increased risk of bulimic symp- mon in women.25 Despite the mismatch of sex dif-
toms and associated features for bullies is both ferences between the prevalence of the risk factor

International Journal of Eating Disorders 48:8 1141–1149 2015 1147

COPELAND ET AL.

and outcome, the associations were similar for Finally, the mediation analysis tested a plausible
men and women in this study. The absence of sex model of how bullying may affect eating disorder
differences serves as a useful corrective against symptoms. At the same time, these results may be
assessment or treatment planning based on bias due to potential confounding of the mediator-
gender-typical expectations. outcome association and should be interpreted
Finally, the bullying-related increased risk for cautiously.
eating problems in adolescence did not extend into Eating disorders are costly43 and incur consider-
young adulthood. This is surprising as we have able morbidity and mortality.24,44 Moreover, buli-
observed long-term effects of bullying on emo- mia and binge eating disorder lead to impairment
tional functioning in this sample.3 On the one and decrements in role attainment in adulthood.45
hand, it is relatively uncommon for childhood risk This study does not suggest that bullying preven-
factors to affect long-term functioning. It is possi- tion would eliminate eating problems, but it does
ble that adult eating disorder symptoms are identify a common, highly visible, childhood expe-
affected by a more proximal set of risk factors, than rience that may predict such problems for some
is the case for emotional problems. It is also still and exacerbate such problems for others. Bullying
possible (and perhaps likely) that bullying involve- can be assessed and monitored by parents, health
ment in adulthood may affect eating outcomes. professionals, and school personnel, and bullying
This should be a priority area of work for prospec- prevention programs that reduce victimization are
tive studies with adult samples. available.46 The most successful efforts to reduce
bullying typically involve improved supervision
Strengths and Limitations and surveillance from school personnel and
As reviewed earlier, the GSMS has the strengths parents, firm discipline for the perpetrators, and a
of prospective design, longitudinal assessment, collective understanding that bullying is not tolera-
and representative sampling. In addition, the study ble or a common rite of passage. Finally, bullying
involvement (including perpetration) should be
has maintained consistently high participation
part of the assessment armamentarium of clini-
rates across time to minimize selective attrition
cians working with patients struggling with eating
and used multiple informants across childhood
problems. Cognitive and emotional responses to
and adolescence. Limitations must also be consid-
such experiences provide a clear target for estab-
ered. The sample is not representative of the US
lished cognitive-behavioral treatment strategies.
population with Native Americans overrepresented
and African Americans and Latinos underrepre- Dr. Bulik is a consultant for Shire Pharmaceuticals. No
sented. The oversampling approach is useful for other authors report a biomedical financial interest or
ensuring an adequate number of cases for risk potential conflict of interest.
analyses, but does require use of weighting to
obtain population-based estimates. The time
between assessments was never less than a year, References
yet both bullying involvement and eating disorder
1. Olweus D. Bullying at School: What We Know and What We Can Do. Malden,
symptoms may vary over shorter periods. Because MA: Wiley-Blackwell, 1993.
this is a population-based study, the rates of DSM 2. Wolke D, Copeland WE, Angold A, Costello E. Impact of bullying in childhood
diagnosable eating disorders are very low. As such, on adult health, wealth, crime and social outcomes. Psychol Sci 2013;24:
we are unable to study whether the bullying 1958–1970.
3. Copeland W, Wolke D, Angold A, Costello E. Adult psychiatric outcomes of
related-changes in eating disorder symptoms
bullying and being bullied by peers in childhood and adolescence. JAMA
results in more children meeting criteria for eating Psychiatry 2013;70:419–426.
disorders. In looking at these associations, it would 4. Sourander AMD, Multimaki PMD, Nikolakaros GMD, Haavisto AMD, Ristkari
be helpful to know the precise nature of the bully- TMNS, Helenius HMS, et al. Childhood predictors of psychiatric disorders
ing behavior (e.g., weight/body related, physical vs. among boys: A prospective community-based follow-up study from age 8
years to early adulthood. J Am Acad Child Adolesc Psychiatry 2005;44:756–
relational). To the extent that such specific aspect
767.
of bullying would have a stronger effect on the out- 5. Gini G, Pozzoli T. Association between bullying and psychosomatic prob-
comes studied here, our findings may underesti- lems: A meta-analysis. Pediatrics 2009;123:1059–1065.
mate true associations. As an observation cohort 6. Reijntjes A, Kamphuis JH, Prinzie P, Telch MJ. Peer victimization and inter-
study, it is not possible to support causal inferences nalizing problems in children: A meta-analysis of longitudinal studies. Child
Abuse Neglect 2010;34:244–252.
in terms of risk factor-outcome associations.41 7. Brunstein Klomek A, Sourander A, Gould MS. The association of suicide and
Inferences in this study are limited by possible con- bullying in childhood to young adulthood: A review of cross-sectional and
founding between time-varying predictors.42 longitudinal research findings. Can J Psychiatry 2010;55:282–288.

1148 International Journal of Eating Disorders 48:8 1141–1149 2015

 EATING DISORDERS AND BULLYING

8. Schreier A, Wolke D, Thomas K, Horwood J, Hollis C, Gunnell D, et al. Pro- 27. Cook CR, Williams KR, Guerra NG, Kim TE, Sadek S. Predictors of bullying
spective study of peer victimization in childhood and psychotic symptoms and victimization in childhood and adolescence: A meta-analytic investiga-
in a nonclinical population at age 12 years. Arch Gen Psychiatry 2009;66: tion. School Psychol Quart 2010;25:65.
527–536. 28. Borg MG. The emotional reactions of school bullies and their victims. Educ
9. Nakamoto J, Schwartz D. Is peer victimization associated with academic Psychol 1998;18:433–444.
achievement? A meta-analytic review. Soc Dev 2010;19:221–242. 29. Costello EJ, Angold A, Burns B, Stangl D, Tweed D, Erkanli A, et al. The Great
10. Thompson JK, Coovert MD, Richards KJ, Johnson S, Cattarin J. Development Smoky Mountains Study of Youth: Goals, designs, methods, and the preva-
of body image, eating disturbance, and general psychological functioning in lence of DSM-III-R disorders. Archives Gen Psychiatry 1996;53:1129–1136.
female adolescents: Covariance structure modeling and longitudinal investi- 30. Pickles A, Dunn G, Vazquez-Barquero J. Screening for stratification in two-phase
gations. Int J Eat Disord 1995;18:221–236. (‘two-stage’) epidemiological surveys. Stat Methods Med Res 1995;4:73–89.
11. Perry DG, Hodges EV, Egan SK. Determinants of chronic victimization by 31. Angold A, Costello E. The Child and Adolescent Psychiatric Assessment
peers. In: Juvonen J, Graham S, editors. Peer Harassment in School: The (CAPA). J Am Acad Child Adolesc Psychiatry 2000;39:39–48.
Plight of the Vulnerable and Victimized. New York: The Guilford Press, 32. Veenstra R, Lindenberg S, Oldehinkel AJ, De Winter AF, Verhulst FC, Ormel J.
2001, pp. 73–104. Bullying and victimization in elementary schools: A comparison of bullies,
12. Lunde C, Frisen A, Hwang CP. Is peer victimization related to body esteem victims, bully/victims, and uninvolved preadolescents. Dev Psychol 2005;41:
in 10-year-old girls and boys? Body Image 2006;3:25–33. 672–681.
13. Menzel JE, Schaefer LM, Burke NL, Mayhew LL, Brannick MT, Thompson JK. 33. Wolke D, Woods S, Bloomfield L, Karstadt L. The association between direct
Appearance-related teasing, body dissatisfaction, and disordered eating: A and relational bullying and behaviour problems among primary school chil-
meta-analysis. Body Image 2010;7:261–270. dren. J Child Psychol Psychiatry 2000;41:989–1002.
14. Eisenberg M, Neumark-Sztainer D. Peer harassment and disordered eating. 34. Lereya ST, Copeland WE, Zammit S, Wolke D. Bully/victims: A longitudinal,
Int J Adolesc Med Health 2008;20:155–164. population-based cohort study of their mental health. Eur Child Adolesc
15. Wolke D, Stanford K. Bullying in school children. In: Messer D, Millar S, edi- Psychiatry 2015:1–11.
tors. Developmental Psychology. London: Arnold, 1999, pp. 341–360. 35. Copeland WE, Wolke D, Lereya ST, Shanahan L, Worthman C, Costello EJ.
16. Farrow CV, Fox CL. Gender differences in the relationships between bullying Childhood bullying involvement predicts low-grade systemic inflammation
at school and unhealthy eating and shape-related attitudes and behaviours. into adulthood. Proc Natl Acad Sci USA 2014;111:7570–7575.
Br J Educ Psychol 2011;81:409–420. 36. Copeland W, Shanahan L, Costello EJ, Angold A. Cumulative prevalence of
17. Lunde C, Frisen A, Hwang CP. Ten-year-old girls’ and boys’ body composition psychiatric disorders by young adulthood: A prospective cohort analysis
and peer victimization experiences: Prospective associations with body satis- from the Great Smoky Mountains Study. J Am Acad Child Adolesc Psychiatry
faction. Body Image 2007;4:11–28. 2011;50:252–261.
18. Lunde C, Frisen A. On being victimized by peers in the advent of adoles- 37. Angold A, Costello EJ. A test-retest reliability study of child-reported psychiat-
cence: Prospective relationships to objectified body consciousness. Body ric symptoms and diagnoses using the Child and Adolescent Psychiatric
Image 2011;8:309–314. Assessment (CAPA-C). Psychol Med 1995;25:755–762.
19. Kaltiala-Heino R, Rimpel€ A M, Rantanen P, Rimpel € AA. Bullying at school— 38. Angold A, Erkanli A, Copeland W, Goodman R, Fisher PW, Costello EJ. Psychi-
An indicator of adolescents at risk for mental disorders. J Adolesc 2000;23: atric diagnostic interviews for children and adolescents: A comparative
661–674. study. J Am Acad Child Adolesc Psychiatry 2012;51:506–517.
20. Striegel-Moore RH, Fairburn CG, Wilfley DE, Pike KM, Dohm F-A, Kraemer 39. Costello EJ, Mustillo S, Erkanli A, Keeler G, Angold A. Prevalence and devel-
HC. Toward an understanding of risk factors for binge-eating disorder in opment of psychiatric disorders in childhood and adolescence. Archives Gen
black and white women: A community-based case-control study. Psychol Psychiatry 2003;60:837–844.
Med 2005;35:907–917. 40. Sobel ME. Asymptotic confidence intervals for indirect effects in structural
21. Kraemer H, Stice E, Kazdin A, Offord D, Kupfer DJ. How do risk factors work equation models. In: Leinhardt S, editor. Sociological Methodology. Wash-
together? Mediators, moderators, and independent, overlapping, and proxy ington, DC: American Sociological Association 1982, pp. 290–312.
risk factors. Am J Psychiatry 2001;158:848–856. 41. De Stavola BL, Nitsch D, dos Santos Silva I, McCormack V, Hardy R, Mann V,
22. Thompson JK, Coovert MD. Body image, social comparison, and eating dis- et al. Statistical issues in life course epidemiology. Am J Epidemiol 2006;
turbance: A covariance structure modeling investigation. Int J Eat Disord 163:84–96.
1999;26:43–51. 42. Diggle P, Heagerty P, Liang K-Y, Zeger S. Analysis of Longitudinal Data.
23. Zwierzynska K, Wolke D, Lereya TS. Peer victimization in childhood and Oxford: Oxford University Press, 2002.
internalizing problems in adolescence: A prospective longitudinal study. 43. Agras WS. The consequences and costs of the eating disorders. Psychiatr Clin
J Abnormal Child Psychol 2013;41:309–323. North Am 2001;24:371–379.
24. Swanson SA, Crow SJ, Le Grange D, Swendsen J, Merikangas KR. Prevalence 44. Root TL, Pisetsky EM, Thornton L, Lichtenstein P, Pedersen NL, Bulik CM.
and correlates of eating disorders in adolescents: Results from the national Patterns of co-morbidity of eating disorders and substance use in Swedish
comorbidity survey replication adolescent supplement. Archives Gen Psychi- females. Psychol Med 2010;40:105.
atry 2011;68:714. 45. Kessler R, Shahly V, Hudson J, Supina D, Berglund P, Chiu W, et al. A com-
25. Hudson JI, Hiripi E, Pope HG Jr, Kessler RC. The prevalence and correlates of parative analysis of role attainment and impairment in binge-eating disor-
eating disorders in the National Comorbidity Survey Replication. Biol Psychi- der and bulimia nervosa: Results from the WHO World Mental Health
atry 2007;61:348–358. Surveys. Epidemiol Psychiatr Sci 2014;23:27–41.
26. Juvonen J, Graham S, Schuster MA. Bullying among young adolescents: The 46. Ttofi MM, Farrington DP. Effectiveness of school-based programs to reduce bul-
strong, the weak, and the troubled. Pediatrics 2003;112:1231–1237. lying: A systematic and meta-analytic review. J Exp Criminol 2011;7:27–56.

International Journal of Eating Disorders 48:8 1141–1149 2015 1149