T h e R o l e o f So c i o c u l t u r a l

F a c t o r s in th e E t i o l o g y o f
Eating D isorders
Ruth Striegel Weissman, PhD

KEYWORDS
 Eating disorder  Etiology  Sociocultural factors  Risk factors  Culture
 Epidemiology  Acculturation  Ethnicity

KEY POINTS
 In the eating disorder field, the risk factor literature reflects a lack of methodological con-
sistency, making it difficult to draw straightforward conclusions.
 The use of population-based registers and record linking is a major advance in the field
because it allows for accumulating very large case samples and accessing high-quality
risk factor data.
 Combining eating diagnoses into a transdiagnostic eating disorder category is premature.

INTRODUCTION

The individual does not exist apart from cultural influence, but is born into—and
can only develop within—particular worlds that come culturally configured.
—Adams and Markus, 2004, p. 346.1

Research into risk factors for the development of any disorder is undertaken in
hopes of improving understanding of and finding information needed for reducing
the burden of suffering related to the disorder. Eating disorders (EDs) are associated
with substantial personal and societal burdens, including alarmingly high mortality,
extensive medical and psychiatric comorbidity, and substantial direct and indirect
economic costs due to elevated health services utilization or adverse impacts on
educational attainment or employment.2–4 Hence, there is an urgency to answer the
questions of who develops an ED, why, and under what circumstances. Risk factor
studies may inform prevention interventions to eliminate or weaken the effect of modi-
fiable risk factors (eg, exposure to media images of extremely thin fashion models) or,
if the risk factor is not modifiable (eg, death of a parent), to provide support for at-risk

Disclosure Statement: Professor R.S. Weissman is Editor-in-Chief of the International Journal of

Eating Disorders and receives a stipend for her role.
Department of Psychology, Wesleyan University, 207 High Street, Middletown, CT 06359, USA
E-mail address: rweissman@wesleyan.edu

Psychiatr Clin N Am 42 (2019) 121–144

https://doi.org/10.1016/j.psc.2018.10.009 psych.theclinics.com
0193-953X/19/ª 2018 Elsevier Inc. All rights reserved.
122 Weissman

individuals. Risk factor research also informs treatment development, as recently

reviewed by Pennesi and Wade.5
The overrepresentation of girls and women among individuals with an ED,6 the
emergence of EDs at distinct historical periods and in geopolitically distinct regions
of the world,7,8 and the increased risk observed among individuals exposed through
migration or the diffusion of ideas, images and values associated with cultures of
modernity9,10 have led scholars to theorize that sociocultural factors play an integral
role in ED etiology.5 In the past century, anorexia nervosa (AN) and bulimia nervosa
(BN) were considered culture-bound syndromes and cross-cultural comparisons
were undertaken to test this assumption. Yet, although early studies found differences
in the incidence or prevalence when comparing samples representing Western culture
or cultures of modernity with samples representing other cultures, recent reviews
concluded that, increasingly, EDs are being identified in diverse countries and cultures
worldwide.8,11,12 A variant of testing the Western culture theory involves comparisons
of subcultures within a country (eg, racial or ethnic minorities vs majority populations,
or sexual minority vs sexual majority groups) with the guiding assumption that if these
subcultures espouse different beauty ideals or differ on other cultural variables from
the majority group, correspondingly, the minority groups should be less likely (or in
the case of homosexual men, more likely) to experience EDs. Cultural developments,
such as globalization and proliferation of social media, have contributed to broad
dissemination of Western culture and, increasingly, research has found that individ-
uals living in non-Western cultures or representing ethnic or sexual minority groups
also are at risk for developing an ED.8,13,14 Moreover, within a given culture, minority
populations are not insulated from exposure to the majority culture’s norms, institu-
tions, or policies; rather, they are exposed to multiple and often conflicting cultural
norms. Some studies suggest that such dual exposure may offset whatever protection
a person might experience as a minority group member or even increase risk due to
stresses arising from minority status (eg, discrimination).15
Early theoretic models focused on Western cultures’ adoption of an unrealistically
thin beauty ideal, norms of femininity that emphasized physical appearance as cen-
tral to self-worth and interpersonal success, and gender-role proscriptions that
limited girls’ and women’s agency and control over their lives.16 The list of sociocul-
tural risk factors has been greatly expanded, now encompassing culturally mediated
processes, such as urbanization, transnational migration or acculturation, social en-
vironments (eg, sororities and sports teams) that might amplify certain harmful
norms or values, and institutions or industries (media and agricultural or food con-
glomerates) responsible for the amplification or dissemination of risk factors.2,9,17,18
This article reviews the recent (2016–2018) empirical literature describing studies on
the contribution of sociocultural factors in ED etiology. Introduced first are chal-
lenges in studying ED etiology, followed by a description of how studies were
selected for inclusion. Results then are summarized and a final section draws
conclusions.

Key Challenges in Eating Disorders Risk Factor Research

Three challenges bear acknowledging:
1. Establishing precedence
2. What defines sociocultural risk factors is far from clear
3. Definitions of dependent variables vary greatly across studies
How these challenges informed the present article is discussed.
 Sociocultural Factors and Eating Disorders 123

The term risk factor implies precedence

As Kraemer and colleagues noted,18 a risk factor is “a measurable characteristic of a
subject in a specified population which precedes the outcome of interest and which
can be shown to divide the population into two groups: a high- and a low-risk
group.”18(p20) “Fixed markers” are risk variables that precede onset and cannot be
shown to be changed (eg, gender at birth). The requirement of temporal precedence
limits the designs suitable for testing risk factor hypotheses. Although fixed markers
can be established using cross-sectional designs, identifying modifiable risk factors
requires longitudinal or experimental designs.19 Increasingly, investigators have
used hybrid designs involving linking of population-based registers for extracting
risk factor data that had been collected in real time well prior to ED onset.
This review reports on results referring to variables where exposure occurred prior
to ED onset and excludes studies examining risk for symptom maintenance or wors-
ening.20 Although correlations with unknown temporal sequence are useful for gener-
ating risk hypotheses, such correlate findings were not considered.

Lack of a common list of sociocultural risk factors

Many ED theories acknowledge culture as an important risk domain, yet what social
variables beyond gender or race/ethnicity or what cultural factors beyond the ubiqui-
tously noted “thin female beauty ideal” should be studied is far from clear. The first
comprehensive review identified some 40 potential ED risk factors and classified 4
as “general and social factors”: gender, race/ethnicity, participation in weight-
related social or professional subculture, and sexual orientation.18 The universe of so-
ciocultural factors can be delineated by focusing on cultural theories of EDs; yet, a
systematic review of ED theories identified more than 50 theoretic models, each
enumerating multiple ED risk factors. Most theories implicate some cultural or social
factors. Some risk factors are model-specific, whereas many others (eg, female
gender, thin-ideal internalization, low self-esteem, and adverse interpersonal experi-
ences) are named in multiple theories or, although referenced as model-specific,
are conceptually similar to other variables (eg, appearance anxiety, body shame,
and fear of weight gain).5
The most widely studied cultural theory of ED is objectification theory.21–23 It posits
that cultures where the female body is objectified (seen as a tool for the sexual plea-
sure and gratification of men) put girls or women at increased ED risk through medi-
ating mechanisms, such as self-objectification, thin-ideal internalization, body
shame, or body surveillance. A meta-analysis of 53 cross-sectional studies found
moderately strong positive statistical associations between self-objectification and
disordered eating.21 Longitudinal studies are now needed for testing body objectifica-
tion (a cultural norm or gender-role proscription) or self-objectification (as the internal-
ization of the cultural norm) as an ED risk factor.
Another approach to testing cultural theories involves comparing individuals who
have undergone a cultural transition via temporary10 or permanent migration24 or
due to political or economic changes.25 Often the focus is on whether cultural transi-
tion is associated with increases in ED prevalence rather than on any sociocultural fac-
tor that might explain ED risk.
ED risk is multifactorial both within a domain (here sociocultural factors) and across
other major risk domains.2 Consistent with Adams and Markus,1 cultural influences
cannot be cleanly separated from other risk domains. Yet, few studies have investi-
gated multiple risk factors across major risk domains or used analytic strategies suit-
able for examining how risk factors might work together to result in distinct
vulnerability pathways and corresponding risk groups.26 Adding yet further
124 Weissman

complexity, ED risk factors cover a broad developmental spectrum (ranging from pre-
natal exposure to experiences into adolescence and possibly adulthood) and a given
risk factor may not apply uniformly across all ages but rather may be developmentally
specific.2,18,27,28
Given the lack of a commonly agreed-on set of sociocultural variables, this review
follows the recommendation18 to take an atheoretic approach and identified studies
for review using an inclusive search term (“risk factor”) and search terms associated
with major methodological approaches to identifying sociocultural risk factors (eg,
“acculturation”).

Risk for what outcome?

Risk factor studies have used various approaches to defining the dependent variable:
1. Using ED diagnoses, with AN and BN the most commonly studied28
2. Transdiagnostically combining diagnoses into one ED category, with the stated
rationale that transition from one to another ED diagnosis is common and that all
EDs share common risk factors
3. Targeting an ED symptom (eg, body image disturbance or binge eating)
4. Creating a disordered eating variable
The question of whether EDs should be combined into one transdiagnostic category
has not yet been settled. For example, both the likelihood of diagnostic transitions and
of transition patterns vary across ED diagnoses (eg, lifetime comorbidity among EDs is
lowest for AN and highest for binge-eating disorder [BED]; transition from AN to BED is
far less common than from AN to BN).29–33 Furthermore, diagnostic transition patterns
have been shown to predict clinical course, which, along with risk factor findings, has
been used as a criterion for diagnostic validity.34 Studies using a transdiagnostic case
definition may miss risk effects if a given variable increases risk for one but not another
ED. Symptom-based outcomes have the advantage over syndrome-based ap-
proaches of increased case samples; for example, far more individuals report binge
eating than meet diagnostic criteria for BN or BED.32,33,35 Symptom-based outcomes
have been widely used, including in prevention trials,15,28,36,37 yet few individuals with
an ED symptom ultimately develop an ED. The lack of a widely accepted composite
“disordered eating” category complicates study comparisons; moreover, composite
scores have the same limitations as transdiagnostic ED categories due to the hetero-
geneity of symptoms grouped into one score.38 This review describes studies
focusing on either ED diagnoses or a transdiagnostic ED category and excludes
studies or findings regarding symptom-based outcomes.

Method
A PubMed search was conducted for journal articles published in English between
January 1, 2016, and July 15, 2018, using the following terms: [sociocultural OR cul-
tural OR cross-cultural OR culture OR acculturation OR acculturative OR risk factor(s)
OR etiology OR etiologic OR epidemiology OR epidemiological] AND [anorexia nerv-
osa OR bulimia nervosa OR binge-eating disorder OR purging disorder OR ARFID OR
eating disorder(s)]. The search yielded 302 citations. Each article’s abstract was
reviewed considering the aforementioned inclusion/exclusion criteria. Additionally, ar-
ticles were discarded if they did not report data (eg, reviews and study protocols); dis-
cussed but did not study risk factors; focused solely on a risk domain other than
culture (eg, genetic studies); examined cultural differences in symptom expression,
treatment response, or health services use; or had fewer than 20 case samples.
Twenty-two risk factor studies were retained and are discussed.
 Sociocultural Factors and Eating Disorders 125

Update on Risk Factor Research in Eating Disorders

Overview
Most studies included individuals with AN (18 studies) or BN (18 studies); individuals
with BED or PD were included in 9 and 6 studies, respectively; 1 study used a code
capturing all EDs, and 1 study focused on feeding and eating disorders (FEDs) in in-
fants; 11 studies excluded boys and men. ED onset age was reported in 4 studies
and varied by diagnosis: earliest onsets were found for AN (median 1639; means:
17.40–19.3031,40,41), followed by BN (mean 20.2031; mean 21.2241) and highest onset
ages reported for BED (mean 23.1541; mean 24.50)31; 16 studies reported information
on incidence or prevalence; yet, given differences in sampling strategies, ED defini-
tions and assessment methods, not surprisingly, incidence and prevalence estimates
varied considerably. Novel contributions of these studies are the inclusion of purging
disorder (PD), which was found quite common,39,42,43 and of FED in very young chil-
dren. In several studies,32,39–41 lifetime or cumulative incidence estimates were higher
than previously reported (partly reflecting the change in diagnostic criteria) and, in 2
studies prevalence of AN even exceeded that of BN.31,32
Demographic characteristics indicating sociocultural risk
By providing information about a disorder’s distribution in the population, epidemio-
logic studies are a key resource for identifying fixed markers, which are commonly
used as proxies for sociocultural factors, including gender, ancestry, and parental ed-
ucation or income. Findings from 13 studies are summarized in Table 1. Shown first
are studies of nationally representative adult samples in Switzerland41 and the United
States,31 followed by a multiethnic study in the United Kingdom, with participants
ages 16 to 90.42 Shown next are studies of adolescent or young adults in Ger-
many,32,43 Finland,40 and Mexico,44 followed by 2 studies focused on middle-aged
women in England39 and in Puerto Rico.45 Shown last are 4 record linkage studies
based on Danish and/or Swedish patient registry data.24,46–48
Gender
Where both genders were included, girls or women outnumber boys or men in all ED
categories. Transdiagnostic groupings (any ED) obscure the fact that female-to-male
ratios vary across diagnosis. For example, gender differences in incidence or preva-
lence are smallest for BED, raising the question of what sociocultural factors might
explain differential risk relative to BN with which BED shares core symptoms with
the notable exception of inappropriate compensatory behaviors. A growing literature
supports adding weight/shape concerns (a defining symptom of BN) to the BED
criteria. Hence, clinically, the key distinction between BN and BED arises from dietary
restraint and inappropriate compensatory behaviors.49 In the only study of FED, haz-
ard ratios (HRs) for gender were highest for babies diagnosed in their first year and
nonsignificant when FED was first diagnosed in the second or third year of life.48
Ancestry
Prior reviews have noted the need to move beyond studying individuals representing
majority cultures in Europe, North America, and Australia or New Zealand. The studies
included still are heavily skewed toward such samples. Methodological differences,
including use of different indicators for prevalence or incidence across studies, pre-
clude confident interpretation of the results concerning race/ethnicity or nationality
as a risk factor. For example, a Mexican study44 identified 40 ED incident cases
over an 8-year period based on state-of-the-art methods, finding incidence estimates
comparable to studies of non-Hispanic samples. Yet, the sample comprised emerging
adults, the group at highest risk for ED onset. Typically, in adult samples, increasing
 126
Weissman
Table 1
Key methodological features and findings of studies of fixed markers for eating disorders

Eating Disorder
Study Location Identification, Incidence or Point
and Sample Diagnostic Criteria, Prevalence or Lifetime
Citation Characteristics and Study Instruments Prevalence (95% CI) Fixed Markers Comments
Mohler-Kuo Switzerland DMS-IV (and DSM-5 for AN) Weighted lifetime Gender Onset age: across EDs, a
et al,41 2016 Nationally representative Two-stage sampling prevalence Female > male across all majority of cases have
sample of 10,038 Stage 1 involved telephone Female EDs; statistical onsets between 10 y and
residents ages 15–60 y calls using questions Any ED: 5.3% (4.7–5.9) comparisons are not 20 y of age.
(52% female) about “household AN: 1.9% (1.6–2.3) reported Approximately 75% of
structure.” BN: 2.4% (2.0–2.8) Ancestry, parental AN, 13% and 20% for BN
Stage 2: CIDI 3.0, by phone BED: 2.4% (2.0–2.8) education, or income and BED, respectively,
Male Not reported reported onset before
Any ED: 1.5% (1.1–1.9) age 20.
AN: 0.2% (0.1–0.4)
BN: 0.9% (0.6–1.2)
BED: 0.7% (0.5–1.0)
Weighted 12-mo
prevalence
Female
Any ED: 1.5% (1.1–1.8)
AN: 0.07% (0.03–0.2)
BN: 0.6% (0.2–0.8)
BED: 0.9% (0.6–1.2)
Male
Any ED: 1.5% (1.1–1.9)
AN: 0.03% (0.004–0.02)
BN: 0.5% (0.3–0.7)
BED: 0.3% (0.5–1.0)
Udo & United States DSM-5 Weighted lifetime Shown are adjusted OR Age of onset: significantly
Grilo,31 2018 Nationally representative Alcohol Abuse and Alcohol prevalence (95% CI) later in BED vs AN or BN
sample of 36,309 Use Disorder and Female Gender Gender: for each ED,
noninstitutionalized Associated Disabilities AN: 1.42% (SE 0.12) AN: 12.00 (6.45–22.34) lifetime prevalence was
civilians ages 18 y or older Interview Schedule 5, in BN: 0.46% (SE 0.6) BN: 5.80 (2.82–11.92) significantly greater in
2012–2013 National person BED: 1.25% (SE 0.10) BED: 3.01 (2.17–4.16) women than men.
Epidemiologic Survey Male Ancestry for lifetime Ancestry: lifetime
Study on Alcohol and AN: 0.12% (SE 0.04) diagnosis (reference prevalence of AN was
Related Conditions BN: 0.08% (SE 0.03) group: non-Hispanic significantly more
Mean ages (SE) of ED BED: 0.42% (SE 0.06) white) common in non-Hispanic
sample Weighted 12-mo Hispanic white individuals vs non-
AN: 41.8 (0.96) y prevalence AN: 0.48 (0.33–0.72) Hispanic black or
BN: 39.1 (2.45) y Female BN: 0.65 (0.33–1.29) Hispanic individuals;
BED: 45.2 (1.21) y AN: 0.08% (SE 0.03) BED: 0.75 (0.38–0.92) lifetime prevalence of BN
BN: 0.22% (SE 0.05) Non-Hispanic black did not vary significantly
BED: 0.60% (SE 0.07) AN: 0.19 (0.11–0.33) by race/ethnicity; there
Male BN: 0.54 (0.25–1.19) were significantly fewer
AN: 0.01% (SE 0.01) BED: 0.60 (0.38–0.92) non-Hispanic black

Sociocultural Factors and Eating Disorders

BN: 0.05% (SE 0.02) Parental education or individuals than non-
BED: 0.26% (SE 0.05) income not reported Hispanic white
individuals with BED,
whereas prevalence of
BED did not differ
significantly between
Hispanic and non-
Hispanic white
individuals.

(continued on next page)

127
 128
Weissman
Table 1
(continued )
Eating Disorder
Study Location Identification, Incidence or Point
and Sample Diagnostic Criteria, Prevalence or Lifetime
Citation Characteristics and Study Instruments Prevalence (95% CI) Fixed Markers Comments
Solmi United Kingdom DSM-5 Weighted 12-mo Gender For establishing the
et al,42 2016 Cross-sectional prevalence Two-stage case and control prevalence No male cases of BN, PD, interview sample, ED
study selection Entire sample OSFED screen positive cases
South East London Stage 1: SCOFF Any ED: 7.4% (4.1%– BED: female 9/109, male were matched to ED
Community Health Study Stage 2: SCID in person, if 13.0%) 2/30, NS screen negative controls
1698 individuals ages 16– SCOFF score 2 or, for AN: 0 cases Ancestry on gender.
90 y (66% female- 25.4% controls, <2 and negative Female ED sample 54.9% of participants
minority) screen for psychiatric Any ED: 6.7% (3.1%– predominantly white, eligible for enrollment to
Study sample N 5 145 disorders 13.6%) no details provided be interviewed declined
Mean age of interview BN: 1.2% (0.5%–2.7%) Parental education or or were lost to follow-up.
sample 36.4 y BED: 4.7% (1.7%– income not reported ED diagnoses were made 2–
12.5%) 3 y after administration
PD: 0.8% (0.3%–2.2%) of the SCOFF and were
OSFED: 3.4% (1.2%– based retrospectively for
9.3%) the 12 mo preceding the
Male time when the SCOFF was
Any ED: 0.9% (0.2%– administered.
4.1%) “Any ED” included
BN: 0 “OSFED”
BED: 0.9% (0.2%– Sample size may have been
4.2%) too small to detect
PD: 0 differences in fixed
OSFED: 0 markers.
Nagl Germany DSM-IV Age-specific, weighted Gender (cumulative Peak threshold incidence
et al,32 2016 Representative community Munich CIDI 3.0 cumulative incidence by incident up to the age at periods for AN and BN:
sample of 30211 age 33: last assessment) 13–18 y
adolescents and young Female OR 22.5 (95% CI, 8.7– Gender: cumulative
adults from metropolitan AN: 3.24% (2.5%– 58.6) incidence up to the age
Munich, Germany, Early 4.3%) Ancestry, parental at last assessment of AN
Developmental Stages of BN: 2.2% (1.6%–3.1%) education, or income are or BN was significantly
Psychopathology study Male not reported higher in females than
Ages 14–24 at baseline and AN: 0.2% (0.07%– males.
ages 21–34 y at 10-y 0.6%)
follow-up BN: 0.08% (0.01%–
0.5%)
Weighted baseline 12-mo
prevalence
Female
AN: 0.4% (0.2%–0.9%)
BN: 0.3% (0.2%–0.7%)
Male

Sociocultural Factors and Eating Disorders

No AN or BN cases
Hammerle Germany DSM-5 Unweighted prevalence Gender Gender difference reached
et al,43 2016 1654 German 7th-grade Structured Interview for AN: 0.3% (0.1–0.7) Female/male ratio statistical significance for
and 8th-grade students Anorexia and Bulimia BN: 0.8% (0.4–1.4) AN: 5:0 atypical AN and PD but
from 9 schools in the Nervosa–Self Report, in BED: 0.5% (0.2–0.9) BN: 5:1 no other diagnosis
state of Rhineland- person OSFED: atypical AN: 3.6% BED: 5:3
Palatinate. Objective measures of (2.7–4.5)- PD: 1.9% Atypical AN: 45:13
Mean age 13.4 (SD 0.8) y height/weight (1.3–2.7) PD: 22:9
Parental education,
parental income,
ancestry not reported

(continued on next page)

129
 130
Table 1
(continued )

Weissman
Eating Disorder
Study Location Identification, Incidence or Point
and Sample Diagnostic Criteria, Prevalence or Lifetime
Citation Characteristics and Study Instruments Prevalence (95% CI) Fixed Markers Comments
Mustelin Finland DSM-5 15-y incidence (age interval None reported Gender: study sample
et al,40 2016 Wave 4 of the FinnTwin16 Two-stage case finding 10–24 y) of AN: included only women
birth cohorts study Stage 1: EDI subscales, self- 230 per 100,000 cases
Nationwide sample of reported ED, being (180–280) per person-
twins born in 1975–1979, suspected by others to year
restricted to those pairs have an ED, current and Lifetime prevalence
where at age 16, both past minimum weight (unweighted):
twins were alive resided Stage 2: DSM-IV SCID via AN: 3.6% (2.7%–4.2%)
in Finland; only women telephone, if screen
were included positive, if twin of a
Sample size N 5 2285 screen positive case, or if
Age range 22–27 y randomly selected from
Mean age 24.4 (SD 0.9) y among screen-negative
participants
Benjet Mexico DSM-IV 8-y incidence (covering Gender Published article uses
et al,44 2016 1074 young adults who had CIDI 3.0, adolescent version ages 12–26 y): ED relative risk ratio 1.33 terms, anorexia, bulimia,
participated 8 y before in at Wave 1, adult version Any ED: 3.7% female, NS and binge eating, but it
a study of adolescent at Wave 2, in person Female 4.3% Bulimia but not anorexia seems implied that these
residents of Mexico City Male 3.1% or binge eating terms refer to full-
metropolitan area. Female incidence significantly syndrome EDs.
Age range 12–26 y AN: 0.9% greater in females than Incidence case sample is
Bulimia: 2.7% males. very small (14 male and
Binge eating: 1.7% Ancestry 26 female ED cases).
Male Not reported Fixed markers as measured
AN: 1.7% Parental education, at Wave I
Bulimia: 0.8% parental income
Binge eating: 0.8% Not living with both
parents
Each tested for “any ED,”
NS
Micali United Kingdom DSM-IV/DSM-5 Weighted lifetime Gender Gender: all-female sample
et al,39 2017 5658 women who were the Two-phase assessment prevalence DNA
main carers of a child Phase 1: EDDS AN: 3.64 (2.81–4.72) Ancestry, parental
enrolled in the UK Avon questionnaire BN: 2.15 (1.7–2.74) education, or income not
Longitudinal Study of Phase 2: SCID for DSM-IV BED: 1.96 (1.52–2.51) reported
Parents and Children (Revised), in person, PD: 1.28 (0.85–1.92)
Phase 2 sample N 5 1036 including screen positive OSFED (including PD): 7.64
Mean age 47.78 (SD 4.5) y cases and a similar (6.32–9.24)
number of screen-
negative women
O’Brien United States DSM-5 (AN or BN; only cases Unweighted prevalence of Gender Gender: All female sample
et al,45 2017 Cross-sectional prevalence with onset between ages AN or BN with onset DNA Ancestry: Non-Hispanic
study 9 y and 22 y were between 9 and 22 y): 2% Results shown below are white > ED prevalence
Sister Study included) adjusted OR (95% CI). than Hispanic, non-
47,759 US or Puerto Rico ED diagnosis based on self- Ancestry (reference group Hispanic black, or Other,
women report (Have you ever non-Hispanic white) respectively
Age range 35–74 y had anorexia nervosa or Non-Hispanic black: Parental education: ED
Mean age 55.8 (9.0) no ED; bulimia?) 0.29 (0.19, 0.42) significantly more

Sociocultural Factors and Eating Disorders

49.8 (7.7) ED cases Fixed marker information: Hispanic: 0.47 (0.33, 0.74) prevalent among
retrospective Other: 0.78 (0.52, 1.19) offspring whose parents
questionnaire Parental education had some college, a
Markers based on was age (reference group high bachelor’s degree, or a
13 school or less) graduate degree,
Some college: 1.27 (1.06, respectively, compared
1.53) with high school or less
Bachelor’s degree: 1.67 Food insecurity: women
(1.41, 1.98) who reported that their
Graduate degree: 2.13 family did not have
(1.78, 2.56) enough to eat were
Income not reported significantly more likely
Food insecurity to self-report AN or BN
Ever vs never: 1.30 (1.03, compared with women
1.63) who had never
experienced food
insecurity before ED
onset.

(continued on next page)

131
 132
Table 1
(continued )

Weissman
Eating Disorder
Study Location Identification, Incidence or Point
and Sample Diagnostic Criteria, Prevalence or Lifetime
Citation Characteristics and Study Instruments Prevalence (95% CI) Fixed Markers Comments
Hvelplund Denmark ICD-10 codes for FEDs Cumulative incidence Gender Gender: FEC is more
et al,48 2016 901,227 children born from All data were extracted 1365 children, 1.6 per Female/male ratio: common in girls than
1997 to 2010, alive and from population and 1000 live births 1.1:0.9 boys
living in Denmark until patient registries HR 1.20 (95% CI, 1.08– Ancestry: FED more
age 48 mo 1.42) common in children with
Ancestry (immigration one or both parents born
status; reference group outside of Denmark
native born)
Both parents
HR 2.24 (1.92–2.61)
One immigrant parent
HR 1.30 (1.10–1.54)
Parental education or
income not reported
Razaz & Sweden ICD-9 code 307B and ICD-10 Overall incidence rate of Gender Gender: all female sample
Cnattingius,46 488,688 singleton girls born codes F500 or F501, based 8.54 per 10,000 person- DNA Ancestry: significantly
2
2018 in Sweden from 1992 to on nation-wide Patient years Adjusted HR (95% CI) lower risk for AN in
2002 as per Patient Registry and Cause of Ancestry females whose mothers
Registry and Cause of Death Register (Reference group: had been born in a non-
Death Register linked Demographic information mother born in a Nordic Nordic country
with Education Register (maternal education and country) Maternal education:
and the Total Population country of origin) Mother born in a non- incidence of AN in
Register extracted from the Nordic country: daughters increased
AN case sample (diagnosed Education and the Total HR 5 0.63 (95% CI significantly in a dose-
by 2012) Population Registers 0.52–0.76) response pattern with
N 5 2414 (0.5%) Maternal education increasing levels of
(Reference group < maternal education.
9 years of schooling)
10–11: HR 5 1.20 (95%
CI 0.97–1.48)
 12: HR 5 1.40 (95% CI
1.12–1.73)
13–14: HR 51.64 (95% CI
1.32–2.04)
15: HR = 1.90 (95% CI
1.54–2.35)
Sundquist Sweden ICD-10 code F50 (includes 0.2% of all individuals who Gender Gender: ED significantly
et al,47 2017 5,397,675 individuals ages AN, BN, BED, ARFID, were at least once Gender ratio in more common in women
18 y or older as per other specified EDs, and registered in the Primary prevalence female/ vs men
Primary Care Register for unspecified EDs) Care Register male: 10.33 (9.70; Mean parental education:
9 regions in Sweden 11.01) OR illustrates the
(1998 2016), linked with Ancestry decreased odds per 1 SD
Education Register Not reported increase in education.
AN case sample Mean parental education Risk for ED increased
N 5 12,633 (average across 5 levels) significantly with
OR: 1.11 (1.09; 1.13) increased level of
Income not reported parental education.
Mustelin Denmark ICD-10 EDs across immigration Analyses focused on Gender: females > males in

Sociocultural Factors and Eating Disorders

et al,24 2017 1,184,205 individuals born Patient registers (hospital status immigration status only; both countries and for
from 1984 to 2002, as per and outpatient) Denmark reference group: both AN and BN. Very
population registers Reported here: AN, BN AN: 4650 females, 359 offspring and parents are small case sample of male
Sweden Foreign migration codes1: males native born immigrants with BN in
1,222,593 individuals born first-generation BN: 2355 females, Reported are IRR, 95% CI, Denmark and N 5 0 in
from 1889 to 1999, as per immigrant (person and approximately 55 adjusted for calendar male immigrants in
population registers both parents born males (numbers too year and age Swedish sample
Inclusion criteria: child was abroad)2; second- low for exact Immigration status Immigration status: male
alive and the child and generation immigrant by reporting in some of Females, AN, Denmark and female first-
both parents were both foreign-born the immigration 1. IRR 0.40 (0.29–0.53) generation immigrants1
residing in Denmark or parents (person him/ status categories) 2. IRR 0.42 (0.34–0.50) are at significantly lower
Sweden, respectively, on herself born in Denmark/ Sweden 3. IRR 1.18 (1.02–1.62) risk for AN and BN, both
child’s 10th birthday. Sweden)3; second- Females, AN, Sweden in Denmark and Sweden
(continued on next page)

133
 134
Weissman
Table 1
(continued )
Eating Disorder
Study Location Identification, Incidence or Point
and Sample Diagnostic Criteria, Prevalence or Lifetime
Citation Characteristics and Study Instruments Prevalence (95% CI) Fixed Markers Comments
Individuals were followed generation immigrant by AN: 3424 females, 268 1. IRR 0.25 (0.19–0.32) Second-generation
from their 10th birthday foreign-born mother4; males 2. IRR 0.35 (0.45–0.62) immigrants with both
until ED onset, death, second-generation BN: 648 females, 9 Males, AN, Denmark foreign-born parents2
emigration from immigrant by foreign- males 1. IRR 0.25 (0.04–0.78) are at lower risk for AN in
Denmark/Sweden, or end born father; reference 2. IRR 0.45 (0.21–0.82) the Danish sample and in
of the follow-up period group: native Dane/ Males, AN, Sweden the female (but not male)
(end of 2013, Denmark, Swede (person and 1. IRR 0.52 (0.284–0.98) Swedish sample.
2009 Sweden), whichever parents born in Denmark Females, BN, Denmark In the female Danish
occurred first. or Sweden) 1. IRR 0.62 (0.43–0.85) sample, risk for AN was
2. IRR 0.57 (0.43–0.73) elevated in second-
3. IRR 1.97 (1.67–2.31) generation females with
Females, BN, Sweden a foreign-born father.4
1. IRR 0.30 (0.18–0.52) In regard to BN, findings
2. IRR 1.37 (1.01–1.87) varied by country,
3. IRR 1.48 (1.13–1.96) gender, and immigration
category. In females,
both in the Danish and
Swedish samples, lower
risk was observed in first-
generation Immigrants
and higher risks were
found in second-
generation immigrants
with foreign-born
fathers.

Only findings regarding “any eating disorder,” AN, BN, BED, PD, or Other Specified Eating Disorders (OSFED) are included.
Abbreviations: CIDI, Composite International Diagnostic Interview 3.0; DSM-5, Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition]; ICD-9, Interna-
tional Classification of Diseases, Ninth Revision; NS, not statistically significant; OR, odds ratio; SCID, Structured Clinical Interview for DSM.
 Sociocultural Factors and Eating Disorders 135

age and age cohort are associated with decreasing incidence and prevalence esti-
mates whereas in children, increasing age is associated with higher estimates.41,45,50
Two studies examined differences between non-Hispanic white (white), non-Hispanic
black (black) Hispanic individuals, yielding two overarching findings: (1) that EDs were
least common in black individuals and most common in white participants31,45 and (2)
in the only study to test for differences by ED diagnosis, that the degree of differences
by race/ethnicity varied across EDs (largest, and statistically significant, for AN and
smallest for BED).31 Udo and Grilo31 cautioned that the prevalence estimates for BN
and BED in the US study were unusually low in all ancestry groups; hence, the findings
of no differences across racial/ethnic groups in the prevalence of BN await replication.

Parental socioeconomic status

In 3 of 4 studies that reported on indicators of parental socioeconomic status, higher
levels of parental education were associated with increased risk for AN46 or risk for
ED.45,47 The exception was 1 study where neither parental educational nor family in-
come was associated significantly with ED risk, but the ED sample was small.44 ED
risk also was elevated in women who had experienced food insecurity (a proxy mea-
sure of family poverty),45 a result that is consistent with a recent study of elevated
prevalence of ED symptoms in patrons of soup kitchens or food pantries.51

Immigration
Results from three studies with data on immigration status suggest that risk varied by
ED. Specifically, FEC was more commonly diagnosed in children born to one or both
immigrant parents versus children of parents born in Denmark,48 but the early onset
and the extensive biological risk factors associated with FED in infants suggest that
FED is etiologically distinct from EDs with onsets in older children or adults.52
A second study reported significantly lower risk for AN in Swedish girls whose
mothers had been born in a non-Nordic country.46 A third study involving individuals
born in Denmark or Sweden specifically focused on immigrants versus the reference
group of native-born offspring of native-born parents.24 ED risk was significantly lower
in first-generation immigrants. Results concerning ED risk in second-generation immi-
grant children varied by country, diagnosis, and gender of the child.24 Among girls and
women in both countries, risk for AN was significantly lower in second-generation im-
migrants with two foreign-born parents; in sons, AN risk was lower only in the Danish.
Moreover, risk for AN was elevated in second-generation daughters with a foreign-
born father in the Danish sample only.
In regard to BN, male samples were too small for statistical tests. In Danish and
Swedish girls and women, lower risk was observed in first-generation immigrants
and higher risk in second-generation immigrants with foreign-born fathers. In the
Danish but not Swedish sample, second-generation daughters with two foreign-
born parents had lower risk for BN. Measured only in the Danish sample, foreign-
born adoptees and daughters born to expatriates had higher risk for BN.
Lower prevalence among first-generation immigrants has been well documented for
several psychiatric disorders, but as the investigators noted, the risk-lowering effect of
immigrant status was especially strong in AN, where risk was nearly halved among
first-generation immigrants or immigrants with two foreign-born parents. Because
ED diagnoses were extracted from patient registers, difference across immigrant
groups partly may reflect differences in treatment seeking or detection patterns. Prior
research found that failure to seek treatment or be properly diagnosed when present-
ing for treatment was more common among individuals with BED than AN or BN, mi-
norities, and boys and men.4 Future studies should explore why marriage involving a
136 Weissman

native-born and a foreign-born partner is associated with increased ED risk for

offspring. For example, perhaps these couples experience higher levels of unresolved
family disagreements, a variable reported to increase risk for AN or BN in two case-
control studies.53,54

Studies exploring variable risk factors

Variable risk factor studies examined data across 3 prevention trials designed to
reduce ED risk in at-risk women;26 tested risk using classification tree analysis
(CTA);55,56 used a case-control design and retrospectively measured exposure to
more than 120 risk factors;53,54 or used register-based record linkage designs to
test the role of early childhood adversities;57–59 immigration status;24 or school char-
acteristics60 in the etiology of EDs. Only findings pertaining to AN, BN, or BED are
discussed.

Multidomain risk factor studies

Using data from 3 targeted prevention trials, including 1271 women (mean age 18.5;
SD 4.2) with high levels of body image concerns (but not an ED diagnosis), Stice
and colleagues26 tested whether these baseline characteristics were predictive of
ED onset: thin-ideal internalization, thinness expectancy (expecting positive outcomes
from being thin), denial of thin-ideal costs, body dissatisfaction, dieting, negative
affect, overeating, fasting, excessive exercise, functional impairment, mental health
care, and body mass index (BMI). There were 9 women with AN, 77 with BN, 69
with BED, and 53 with PD. Multivariate models identified these significant predictors
of ED onset: elevated negative affect (95% CI, HR 2.23; 1.25–3.99) and lower BMI
(95% CI, HR 0.63; 0.44–0.91) for AN, body dissatisfaction (95% CI, HR 1.50; 1.08–
2.08), overeating (95% CI, HR 2.07; 1.65–2.61), and fasting (95% CI, HR 1.29; 1.08–
1.53) for BN, body dissatisfaction (95% CI, HR 1.49; 1.07–2.06), overeating (95%
CI, HR 1.99; 1.59–2.49), and functional impairment (95% CI, HR 1.40; 1.07–1.83) for
BED, and body dissatisfaction (95% CI, HR 1.88; 1.29–2.72) and dieting (95% CI,
HR 1.94; 1.35–2.79) for PD. None of the variables most commonly identified as tapping
sociocultural pressures related to the thin ideal (eg, thin ideal internalization or positive
expectancies about the benefits of thinness) were predictive of ED risk, possibly
because they were highly correlated with body dissatisfaction. Consistent with prior
studies, body dissatisfaction was a transdiagnostic ED risk factor.
Because many ED risk factors have been shown to be highly collinear,56 experts
have advocated using analytic methods that are suited for handling correlated data
and detecting interactions including higher-order interactions among variables, such
as classification tree analysis (CTA). Prior studies in ED using CTA included small
case samples,61 cross-sectional data,62 or symptom-based rather than syndrome-
based outcomes.63 Two studies were identified that used this method for testing for
ED. Specifically, in second publication based on the prevention trials data, Stice
and colleagues55 sought to address this gap and found the following. For women
with AN (defined in this second publication to include subthreshold cases, resulting
in a case sample increased from 9 to 26), a 2-way interaction of low BMI and body
dissatisfaction was found, with low BMI at baseline the most potent predictor, fol-
lowed by body dissatisfaction. CTA identified a complex 4-way interaction for BN.
The first partition occurred for high levels of overeating as the most potent predictor,
and amplification of this risk from overeating in those women with elevated positive
expectancies regarding thinness and body dissatisfaction. Paradoxically, among
those with this risk factor combination, lower levels of thin ideal internalization further
amplified risk for BN. Regarding, BED, a 4-way interaction was found with elevated
 Sociocultural Factors and Eating Disorders 137

body dissatisfaction as the strongest predictor. Risk was amplified by elevated over-
eating and low levels of dieting. High levels of thin ideal internalization further amplified
risk. For PD, a 3-way interaction was found involving dieting, the strongest predictor,
and negative affect and positive thinness internalization, each further amplifying PD
risk. Together, these findings suggest both transdiagnostic risk factors and
diagnosis-specific risk factors and that ED risk may involve different pathways, or
risk subgroups.
Utilizing data on 1297 adolescents (49% male) who participated in a longitudinal
cohort study in Australia,56 risk for ED onset by age 17 or age 20 was examined.
Excluding children who had developed an ED before age 14, there were 146 partici-
pants (26 boys) with ED onset (no AN). A transdiagnostic ED was created, representing
81 BN, 43 BED, and 22 PD cases. CTA found that the initial split was by gender with
ED incident of 18.1% for girls and 4% for boys. For boys, weight/shape concerns at
age 14 indicated differential risk as follows: boys with weight/shape concerns in the
bottom 12% of normalized scores or between the 12th and 59th percentiles, 1.2%
and 1.5%, respectively, developed an ED, compared with 17.1% of boys with
elevated weight/shape concerns (top 41% of the sample). For girls, weight/shape con-
cerns also were the most potent risk factor and, as for the boys, 3 nonlinear categories
were found, each with increasing degrees of risk: among girls in the bottom category,
1.8% developed an ED; among those in a middle category (ranging from the 12th to
the 77th percentiles), 8.2% had ED onset; in the upper category, 45.1% of girls devel-
oped an ED. Moreover, among girls in the midlevel weight/shape concern group, low
externalizing problems were identified to lower risk such that 2.8% had ED onset
compared with 12.3% onset among girls with average or above-average externalizing
problems. Because the Australian study used a different set of predictor variables
(except for BMI and dieting) and focused on a transdiagnosic rather than diagnosis-
specific outcome category, these findings cannot be compared directly to the US
sample. Weight/shape concern and body dissatisfaction, however, typically are highly
correlated. Taken together, consistent with sociocultural explanations, these two
studies support the etiologic role of body image concerns in EDs.
Machado and colleagues completed two related case-control studies involving
ED patients recruited at specialty treatment centers in Portugal, matching ED cases
(Diagnostic and Statistical Manual of Mental Disorders [Fourth Edition] [DSM-IV]) on
demographic variables to healthy controls (HCs) and psychiatric controls (PCs) and
examining a large (approximately 120) multifactorial set of risk variables.53,54 Expo-
sure was focused on the time before onset of the first ED symptom of AN or BN,
respectively. In both studies, most variables were found not to differentiate ED
from HC or PC cases (including several sociocultural factors, such as parental un-
employment, parents in risk occupations, and several childhood adversity
variables).
Specifically, in a sample of 60 women with BN, compared with 60 HCs,53 more
women with BN reported feeling self-conscious about their physical appearance,
negative family experiences related to eating or appearance (undue emphasis on
physical appearance, tensions during meals, and comments about weight),
appearance-related comments and teasing by friends or others, being overweight
as a child or adolescent, feeling fat in childhood, negative feelings about menarche,
negative attitudes about parents’ weight, childhood anxiety, deliberate self-harm,
parental depression, and family difficulties (unresolved disagreements, high parental
expectations and criticism, low parental involvement, and feeling inferior to siblings).
BN cases also differed significantly from PC (N 5 60) on most of these variables. In
logistic regressions, two variables were retained for differentiating BN cases versus
138 Weissman

HC cases, correctly classifying 93.3% of BN cases and 100% of HC cases: negative

attitude about parental weight and childhood overweight. Four variables differentiated
BN cases versus PC cases (correct classification: 86.7% BN and 88.3% PC): feeling
fat in childhood, deliberate self-harm, unresolved disagreements, and high maternal
expectations.
A second case-control study54 recruited 98 patients with AN, 79 patients with BN,
86 patients with HC, and 68 patients with PC (60 of the BN cases and all HC cases and
PC cases were included in the aforementioned study).53 Compared with HC, women
with AN were significantly more likely to report self-consciousness about physical
appearance, parental comments about eating, teasing (both unrelated and related
to weight/shape), feeling fat in childhood, negative attitudes about parents’ weight/
shape, a family history of AN or BN, perfectionism, and unresolved family disagree-
ments. Compared with the PC group, women with AN were significantly more likely
to report appearance self-consciousness, negative attitudes about parental weight/
shape and about menstruation, feeling fat in childhood, a family history of AN or
BN, being teased (unrelated to appearance), unresolved family disagreements, and
being perfectionistic. Compared with women with AN, women with BN were signifi-
cantly more likely to report exposure to high parental expectations, family emphasis
on fitness, feeling that adolescent overweight conveyed negative consequences,
and being overweight in adolescence.
A strength of these studies was the focus on a wide range of potential risk factors.
Yet, although the findings from these 2 case-control studies support cultural hypoth-
eses about the role of thinness expectations for girls and women and associated risk
factors, such as feeling fat or placing negative valence on overweight in self or others,
due to retrospective risk assessment, it cannot be ruled out that patients’ current
experience of an ED contributed to enhanced or selective recall of body image–
related, weight-related, and eating-related experiences. Case matching precluded
testing associations between fixed markers and AN or BN. In contrast to research
described later, where exposure was measured objectively or using much larger
case samples, in these case-control studies, measures of severe adverse childhood
events (including bereavement and physical or sexual abuse) were not found associ-
ated with AN or BN risk.

Childhood adversity
A Danish register–based study examined exposure (yes/no) before age 6 years to 9
adversity variables in a cohort of all girls born in Denmark to Danish-born parents be-
tween 1989 and 2007 (excluding children who died or emigrated before age 6 years)
who were followed up to age 25 years. Adversities included family disruption (child not
sharing same address with both parents for various reasons), parental somatic illness,
residential instability (more than 1 house move between 2 municipalities), parental
psychiatric illness, parental substance us disorder, severe parental criminality,
parental disability, familial death (loss of a parent or full or half sibling), and placement
in foster care. Due to human subjects’ concerns, the study did not examine childhood
physical abuse or sexual abuse. ED onset was defined as age of first hospitalization for
AN, BN, or EDNOS. Of 495,244 girls in the study sample, almost 35% had experienced
1 or more childhood adversities and 2892 (0.58%) were diagnosed with AN and 1027
(0.21%) with BN.
Cumulative incidence of AN was highest for women with no adversity and
decreased with increasing number of adversities; AN was associated inversely with
severe parental criminality or placement in foster care. In contrast, cumulative inci-
dence of BN was lowest for women with no adversity and increased with increasing
 Sociocultural Factors and Eating Disorders 139

number of adversities; family disruption and parental psychiatric illness were signifi-
cantly associated with BN. In regard to specificity of risk (ED rather than another psy-
chiatric disorder), the study found that exposure to most or all adverse events was
associated with increased risk for major mood disorder, anxiety disorder, and
obsessive-compulsive disorder. Because hospital-based care is the exception in
the treatment of ED and typically connotes greater illness severity,3,4 it is unclear
whether the findings generalize to individuals with no or only outpatient treatment.
Nonetheless, the large case samples and objective adversity measures are strengths
of this study.
A population-based cohort study using national registries from Denmark and
Sweden examined exposure to bereavement from 1 year preconception to 10 years
of age as a possible risk factor for broadly defined EDs.59 Results are described
concerning broadly defined AN and BN. Data were extracted for 934,610 girls
born in Denmark from 1970 to 2000 and 1,178,146 girls born in Sweden from
1973 to 1997 (excluding girls who had experienced bereavement between ages
10 and 26 years). Girls were followed for 10 years or until they were diagnosed
with an ED, emigrated, died, or reached age 26, whichever came first. In all,
64,453 (3.05%) of girls experienced exposure to bereavement, defined as loss of
a core relative (prenatal maternal loss of a spouse/child or postnatal loss of a parent
or a sibling) or loss of other relatives and by type of death (expected/unexpected).
Of the 5878 girls with AN, 166 (2.41%, adjusted incidence rate ratio [IRR] 0.96; 95%
CI, 0.82–1.13) were exposed, a nonsignificant association. IRRs also were nonsig-
nificant for timing of exposure, relationship to the deceased, or whether death was
expected. Of 1722 girls with BN, 63 were exposed to bereavement. Although the
overall association between bereavement and BN status was nonsignificant
(adjusted IRR 1.30; 0.97–1.72), the association of bereavement exposure to unex-
pected death prenatally or postnatally was associated significantly with risk for BN
(IRR 2.47; 1.67–3.65).
Using data from a nationally representative sample of US adults (N 5 36,309),
self-reported exposure before age 18 years to 7 types of childhood maltreatment
(harsh physical punishment, physical abuse, sexual abuse, emotional abuse,
emotional neglect, physical neglect, and exposure to intimate partner violence)
were examined as risk factors for AN, BN, or BED.57 Among women with AN
(N 5 240), except for emotional neglect and exposure to partner violence, maltreat-
ment was statistically significant, whereas for men with AN (N 5 36), only physical
abuse, sexual abuse, and exposure to intimate partner violence were significant pre-
dictors. When adjusting for all other forms of maltreatment, in both men and women
only the association of sexual abuse and AN diagnosis remained significant. In
women with BN (N 5 11), sexual abuse, emotional abuse, and emotional neglect
were significant; in men (N 5 11), associations with BN diagnosis were significant
for physical abuse, sexual abuse, emotional neglect, and exposure to intimate part-
ner violence. In fully adjusted models, emotional abuse remained a significant pre-
dictor in women with BN; the model could not be calculated for men due to
inadequate power and lack of variance. Among women with BED (N 5 203), asso-
ciations were significant for each of the maltreatment categories and, in fully
adjusted models, for sexual and emotional abuse. For men (N 5 82), sexual abuse
and physical neglect were significantly associated with BED; in fully adjusted
models, only physical neglect remained significant. Except for exposure to intimate
partner violence and physical neglect, which varied by diagnosis and gender, the
study suggests that childhood maltreatment is a transdiagnostic and gender-
neutral risk factor for AN, BN, and BED.
140 Weissman

High-risk environments
Cultural theories have proposed that certain environments confer increased risk for ED
development. For example, 1 study provided initial support for a social contagion ef-
fect among sororities,64 but few studies have tested environmental risk hypotheses
using systems, rather than individual-level variables. Bould and colleagues60
addressed this gap with a register-based record linkage study composed of data
on 55,059 girls born in Sweden from 1983 and finishing high school in 2002 to
2010, involving 409 schools. Individual-level and school-level variables were
measured and individual-level variables were used to estimate odds ratios for different
school environment factors and to calculate the contributions of each variable to
between-school variation in the incidence of ED at the school level. In all, 2.4% of girls
had an ED onset between 16 and 20 years of age; EDs were broadly defined (any In-
ternational Classification of Diseases, Tenth Revision [ICD-10] ED code) and extracted
from treatment registries. Of the total variation in the odds of ED, 4.4% (95% CI, 2.8–
7.1) was due to between-school differences. Individual-level variables with the great-
est impact on this variation included parental education, family disposable income,
maternal age, having a foreign-born parent, and parental psychiatric history. Models
adjusted for individual-level variables and testing the association of school-level vari-
ables on variance in the odds of ED found elevated ED risk in schools with a higher
proportion of girls or of children of highly educated parents.

SUMMARY

The myriad methodological variations across the 22 studies (including the wide range
of risk factors measured) defy reaching straightforward conclusions regarding the role
of sociocultural factors in ED risk, with 1 exception. Gender continues to be shown a
potent and consistent predictor, although the variations in risk ratios by diagnosis
point to the need to focus efforts to improve understanding of risk factors in male pop-
ulations. Reflecting a major gap, male samples were almost entirely limited to studies
reporting on fixed markers and missing from studies that assessed a broad range of
variable risk factors. Where included, with notable exceptions of population-based
register record linkage studies, boys and men with EDs represented small numbers,
underscoring the challenge of studying the etiology of male EDs.
A major novel development in the field is the use and linking of population-based
registers, which results in large case samples and provides reliable data on fixed
markers or exposure to certain risk factors that may be subject to biased recall.
Another novel contribution arises from the application of statistical methods that
help address the problem of collinearity of risk variables and elucidate risk subgroups.
Consistent with cultural theories of EDs, body dissatisfaction, or weight concerns, a
variable downstream from cultural pressures to achieve a thin beauty ideal was found
to increase risk transdiagnostically and across gender groups. In retrospective re-
ports, family or peer pressure about appearance also were significantly associated
with risk for AN or BN, although the motivation for such pressure may differ between
those 2 diagnostic groups. BMI has been found a bimodal risk factor, with elevated
BMI predicting BN or BED and low BMI predicting AN. The literature reviewed further
points to the need to continue to focus on diagnosis-specific outcome categories for
clarifying ED etiology. Studies of Scandinavian populations identified first-generation
immigration status as a protective factor yet also suggest that second-generation im-
migrants may experience elevated risk for reasons that await further exploration.
Across studies of fixed markers and variable risk factors, findings suggest that
beyond the transdiagnostic risk arising from body image concerns, there are
 Sociocultural Factors and Eating Disorders 141

disorder-specific risk factors, as exemplified by the role of BMI, immigration status,

and childhood aversities. Although most studies included individuals with AN or BN,
research of BED and PD continues to lag. Therefore, calls for a transdiagnostic
approach to ED classification seem premature.

REFERENCES

1. Adams G, Markus HR. Toward a conception of culture suitable for a social psy-
chology of culture. In: Schaller M, Crandall CS, Schaller M, editors. The psycho-
logical foundations of culture. Mahwah (NJ): Lawrence Erlbaum Associates
Publishers; 2004. p. 335–60.
2. Striegel-Moore RH, Bulik CM. Risk factors for eating disorders. Am Psychol 2007;
62:181–98.
3. Whiteford H, Weissman RS. Key factors that influence government policies and
decision making about healthcare priorities: lessons for the field of eating disor-
ders. Int J Eat Disord 2017;50:315–9.
4. Weissman RS, Rosselli F. Reducing the burden of suffering from eating disorders:
unmet treatment needs, cost of illness, and the quest for cost-effectiveness. Be-
hav Res Ther 2017;88:49–64.
5. Pennesi JL, Wade TD. A systematic review of the existing models of disordered
eating: do they inform the development of effective interventions? Clin Psychol
Rev 2016;43:175–92.
6. Rodin J, Silberstein L, Striegel-Moore R. Women and weight: a normative discon-
tent. Nebr Symp Motiv 1984;32:267–307.
7. Brumberg JJ. Fasting girls: the emergence of anorexia nervosa as a modern dis-
ease. Cambridge (MA): Harvard University Press; 1988.
8. Pike KM, Hoek HW, Dunne PE. Cultural trends and eating disorders. Curr Opin
Psychiatry 2014;27:436–42.
9. Becker AE. Sociocultural influences on body image and eating disturbance. In:
Brownell KD, Walsh BT, Brownell KD, editors. Eating disorders and obesity: a
comprehensive handbook. New York: Guilford Press; 2018. p. 127–33.
10. Swami V. Change in risk factors for eating disorder symptomatology in Malay stu-
dents sojourning in the United Kingdom. Int J Eat Disord 2016;49:695–700.
11. Kolar DR, Rodriguez DL, Chams MM, et al. Epidemiology of eating disorders in
Latin America: a systematic review and meta-analysis. Curr Opin Psychiatry
2016;29:363–71.
12. van Hoeken D, Burns JK, Hoek HW. Epidemiology of eating disorders in Africa.
Curr Opin Psychiatry 2016;29:372–7.
13. Becker AE, Fay KE, Agnew-Blais J, et al. Social network media exposure and
adolescent eating pathology in Fiji. Br J Psychiatry 2011;198:43–50.
14. Hoek HW. Review of the worldwide epidemiology of eating disorders. Curr Opin
Psychiatry 2016;29:336–9.
15. Calzo JP, Blashill AJ, Brown TA, et al. Eating disorders and disordered weight and
shape control behaviors in sexual minority populations. Curr Psychiatry Rep
2017;19:49.
16. Striegel-Moore RH, Silberstein LR, Rodin J. Toward an understanding of risk fac-
tors for bulimia. Am Psychol 1986;41:246–63.
17. Keel PK, Forney KJ. Psychosocial risk factors for eating disorders. Int J Eat Dis-
ord 2013;46:433–9.
142 Weissman

18. Jacobi C, Hayward C, de Zwaan M, et al. Coming to terms with risk factors for
eating disorders: application of risk terminology and suggestions for a general
taxonomy. Psychol Bull 2004;130:19–65.
19. Kraemer HC, Kazdin AE, Offord DR, et al. Coming to terms with the terms of risk.
Arch Gen Psychiatry 1997;54:337–43.
20. Davis HA, Smith GT. An integrative model of risk for high school disordered
eating. J Abnorm Psychol 2018;127:559–70.
21. Schaefer LM, Thompson JK. Self-objectification and disordered eating: a meta-
analysis. Int J Eat Disord 2018;51(6):483–502.
22. Velez BL, Campos ID, Moradi B. Relations of sexual objectification and racist
discrimination with Latina women’s body image and mental health. Couns Psy-
chol 2015;43:906–35.
23. Cheng H-L, Tran AGTT, Miyake ER, et al. Disordered eating among Asian Amer-
ican college women: a racially expanded model of objectification theory. J Couns
Psychol 2017;64:179–91.
24. Mustelin L, Hedman AM, Thornton LM, et al. Risk of eating disorders in immigrant
populations. Acta Psychiatr Scand 2017;136:156–65.
25. Nakai Y, Nin K, Noma S. Eating disorder symptoms among Japanese female stu-
dents in 1982, 1992 and 2002. Psychiatry Res 2014;219:151–6.
26. Stice E, Gau JM, Rohde P, et al. Risk factors that predict future onset of each
DSM-5 eating disorder: Predictive specificity in high-risk adolescent females.
J Abnorm Psychol 2017;126:38–51.
27. Bakalar JL, Shank LM, Vannucci A, et al. Recent advances in developmental and
risk factor research on eating disorders. Curr Psychiatry Rep 2015;17:42.
28. Culbert KM, Racine SE, Klump KL. Research review: what we have learned about
the causes of eating disorders - a synthesis of sociocultural, psychological, and
biological research. J Child Psychol Psychiatry 2015;56:1141–64.
29. Schaumberg K, Jangmo A, Thornton LM, et al. Patterns of diagnostic transition in
eating disorders: a longitudinal population study in Sweden. Psychol Med
2018;1–9 [Epub ahead of print].
30. Welch E, Jangmo A, Thornton LM, et al. Treatment-seeking patients with binge-
eating disorder in the Swedish national registers: clinical course and psychiatric
comorbidity. BMC Psychiatry 2016;16:163.
31. Udo T, Grilo CM. Prevalence and correlates of DSM-5-defined eating disorders in
a nationally representative sample of U.S. adults. Biol Psychiatry 2018;84:345–54.
32. Nagl M, Jacobi C, Paul M, et al. Prevalence, incidence, and natural course of
anorexia and bulimia nervosa among adolescents and young adults. Eur Child
Adolesc Psychiatry 2016;25:903–18.
33. Mustelin L, Bulik CM, Kaprio J, et al. Prevalence and correlates of binge eating
disorder related features in the community. Appetite 2017;109:165–71.
34. Striegel-Moore RH, Franko DL. Should binge eating disorder be included in the
DSM-V? A critical review of the state of the evidence. Annu Rev Clin Psychol
2008;4:305–24.
35. Schluter N, Schmidt R, Kittel R, et al. Loss of control eating in adolescents from
the community. Int J Eat Disord 2016;49:413–20.
36. Watson HJ, Joyce T, French E, et al. Prevention of eating disorders: A systematic
review of randomized, controlled trials. Int J Eat Disord 2016;49:833–62.
37. Le LK, Barendregt JJ, Hay P, et al. Prevention of eating disorders: a systematic
review and meta-analysis. Clin Psychol Rev 2017;53:46–58.
 Sociocultural Factors and Eating Disorders 143

38. Smink FRE, van Hoeken D, Dijkstra JK, et al. Self-esteem and peer-perceived so-
cial status in early adolescence and prediction of eating pathology in young
adulthood. Int J Eat Disord 2018;51(8):852–62.
39. Micali N, Martini MG, Thomas JJ, et al. Lifetime and 12-month prevalence of
eating disorders amongst women in mid-life: a population-based study of diagno-
ses and risk factors. BMC Med 2017;15:12.
40. Mustelin L, Silen Y, Raevuori A, et al. The DSM-5 diagnostic criteria for anorexia
nervosa may change its population prevalence and prognostic value. J Psychiatr
Res 2016;77:85–91.
41. Mohler-Kuo M, Schnyder U, Dermota P, et al. The prevalence, correlates, and
help-seeking of eating disorders in Switzerland. Psychol Med 2016;46:2749–58.
42. Solmi F, Hotopf M, Hatch SL, et al. Eating disorders in a multi-ethnic inner-city UK
sample: prevalence, comorbidity and service use. Soc Psychiatry Psychiatr Epi-
demiol 2016;51:369–81.
43. Hammerle F, Huss M, Ernst V, et al. Thinking dimensional: prevalence of DSM-5
early adolescent full syndrome, partial and subthreshold eating disorders in a
cross-sectional survey in German schools. BMJ Open 2016;6:e010843.
44. Benjet C, Borges G, Mendez E, et al. Eight-year incidence of psychiatric disor-
ders and service use from adolescence to early adulthood: longitudinal follow-
up of the Mexican Adolescent Mental Health Survey. Eur Child Adolesc Psychia-
try 2016;25:163–73.
45. O’Brien KM, Whelan DR, Sandler DP, et al. Predictors and long-term health out-
comes of eating disorders. PLoS One 2017;12:e0181104.
46. Razaz N, Cnattingius S. Association between maternal body mass index in early
pregnancy and anorexia nervosa in daughters. Int J Eat Disord 2018;51(8):
906–13.
47. Sundquist J, Ohlsson H, Sundquist K, et al. Common adult psychiatric disorders
in Swedish primary care where most mental health patients are treated. BMC Psy-
chiatry 2017;17:235.
48. Hvelplund C, Hansen BM, Koch SV, et al. Perinatal risk factors for feeding and
eating disorders in children aged 0 to 3 years. Pediatrics 2016;137:e20152575.
49. Lewer M, Bauer A, Hartmann AS, et al. Different facets of body image distur-
bance in binge eating disorder: a review. Nutrients 2017;9 [pii:E1294].
50. Volpe U, Tortorella A, Manchia M, et al. Eating disorders: what age at onset? Psy-
chiatry Res 2016;238:225–7.
51. Becker CB, Middlemass K, Taylor B, et al. Food insecurity and eating disorder
pathology. Int J Eat Disord 2017;50:1031–40.
52. Bryant-Waugh R, Markham L, Kreipe RE, et al. Feeding and eating disorders in
childhood. Int J Eat Disord 2010;43:98–111.
53. Goncalves S, Machado BC, Martins C, et al. Retrospective correlates for bulimia
nervosa: a matched case-control study. Eur Eat Disord Rev 2016;24:197–205.
54. Machado BC, Goncalves SF, Martins C, et al. Anorexia nervosa versus bulimia
nervosa: differences based on retrospective correlates in a case-control study.
Eat Weight Disord 2016;21:185–97.
55. Stice E, Desjardins CD. Interactions between risk factors in the prediction of
onset of eating disorders: exploratory hypothesis generating analyses. Behav
Res Ther 2018;105:52–62.
56. Allen KL, Byrne SM, Crosby RD, et al. Testing for interactive and non-linear effects
of risk factors for binge eating and purging eating disorders. Behav Res Ther
2016;87:40–7.
144 Weissman

57. Afifi TO, Sareen J, Fortier J, et al. Child maltreatment and eating disorders among
men and women in adulthood: results from a nationally representative united
states sample. Int J Eat Disord 2018;50(11):1281–96.
58. Larsen JT, Munk-Olsen T, Bulik CM, et al. Early childhood adversities and risk of
eating disorders in women: a Danish register-based cohort study. Int J Eat Disord
2017;50:1404–12.
59. Su X, Liang H, Yuan W, et al. Prenatal and early life stress and risk of eating dis-
orders in adolescent girls and young women. Eur Child Adolesc Psychiatry 2016;
25:1245–53.
60. Bould H, De Stavola B, Magnusson C, et al. The influence of school on whether
girls develop eating disorders. Int J Epidemiol 2016;45:480–8.
61. Striegel-Moore RH, Dohm FA, Kraemer HC, et al. Risk factors for binge-eating
disorders: an exploratory study. Int J Eat Disord 2007;40:481–7.
62. Racine SE, VanHuysse JL, Keel PK, et al. Eating disorder-specific risk factors
moderate the relationship between negative urgency and binge eating: a behav-
ioral genetic investigation. J Abnorm Psychol 2017;126:481–94.
63. Fairweather-Schmidt AK, Wade TD. Weight-related peer-teasing moderates ge-
netic and environmental risk and disordered eating: twin study. Br J Psychiatry
2017;210:350–5.
64. Crandall CS. Social contagion of binge eating. J Pers Soc Psychol 1988;55:
588–98.