Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Journal of Abnormal Psychology © 2016 American Psychological Association

2016, Vol. 125, No. 3, 369 –380 0021-843X/16/$12.00 http://dx.doi.org/10.1037/abn0000146

Characterizing and Predicting Trajectories of Disordered Eating


Over Adolescence
A. Kate Fairweather-Schmidt and Tracey D. Wade
Flinders University

This study sought to identify distinct trajectory classes of growth of disordered eating (DE), and their
respective correlates and risk factors. Females ages 12–19 years (Australian Twin Registry) were
interviewed by telephone with the Eating Disorder Examination on 3 occasions (Wave 1: N ⫽ 699; Wave
2: N ⫽ 669; Wave 3: N ⫽ 499). Each participant also completed self-report measures across all waves,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

and their parents at Waves 1 and 2. Growth mixture modeling determined the optimal classes represent-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

ing DE growth trajectories over 3 data-points. The best fit was a 3-class solution constituting: (a) no
growth in DE over time (90%; low-static) where 10% reported a lifetime eating disorder (ED); (b) a high
intercept but decreasing over time (7%; 53% had a lifetime ED; attenuating); (c) an increasing level of
DE (3%; 80% experiencing a lifetime ED; escalating). The majority (64%) of EDs in the attenuating
group were diagnosed at Wave 1, whereas 84% of EDs in the escalating group were diagnosed at
subsequent waves. The low-static group had fewest significant correlates/risk factors for DE. A multi-
variable GEE identified that over adolescence, internalization of the thin ideal was 2.5 times more likely
characterize an escalating rather than low-static DE symptom trajectory. These findings underscore the
fundamental importance of concentrating efforts on reducing the potency of the thin ideal message, and
strategies directed toward resolving life stressors by effective management of emotion and problem
resolution, in order to prevent the development of DE.

General Scientific Summary


This study was able to identify and characterize longitudinal changes in disordered eating symptom
trajectories of subpopulations of female adolescent twins. Internalization of the thin ideal was found
to distinguish a group with low-stable symptoms from a group whose symptoms escalated over
adolescence.

Keywords: disordered eating, adolescents, trajectory, longitudinal, internalization of the thin ideal

Supplemental materials: http://dx.doi.org/10.1037/abn0000146.supp

It is well known that risk for expression of disordered eating disorders by late adolescence is around 13% to 15% (Allen, Byrne,
increases over adolescence for girls, with new sources of genetic and Oddy, & Crosby, 2013; Stice, Marti, Shaw, & Jaconis, 2009; Wade &
environmental risk for disordered eating emerging at puberty (Klump, O’Shea, 2015), and by early adulthood almost one in four women
Burt, McGue, & Iacono, 2007) and during late adolescence report recent disordered eating, which is then associated with signif-
(Fairweather-Schmidt & Wade, 2015). Lifetime prevalence of eating icantly poorer physical and psychological quality of life over a 9-year
follow-up period compared with women without disordered eating
(Wade, Wilksch, & Lee, 2012).
However, a number of studies have identified that disordered
This article was published Online First February 4, 2016. eating does not increase uniformly for all adolescent girls, as
A. Kate Fairweather-Schmidt and Tracey D. Wade, School of Psychol-
shown in Table 1. This is consistent with the substantial influence
ogy, Flinders University.
The authors report no declarations of financial or conflicts of interest. of both genetic and environmental sources on disordered eating,
National Health and Medical Research Council (NHMRC) Grants 324715 with an estimated median genetic variance of 52% (Culbert,
and 480420 to Tracey D. Wade supported this work. Administrative Racine, & Klump, 2015), which suggests that trajectory of growth
support for data collection was received from the Australian Twin Registry, differs between groups who are differentially genetically suscep-
which is supported by an NHMRC Enabling Grant (ID 310667) adminis- tible to exposure to the different levels of various risk factors
tered by the University of Melbourne. We thank the twins and their present in the environment. It is also consistent with the findings
families for their participation in this research, and Ms. Judith Slater for
of studies using sample means of disordered eating, which show no
coordinating the data collection.
Correspondence concerning this article should be addressed to Tracey D.
coherent common trajectory across adolescence (Abebe, Lien, &
Wade, School of Psychology, Flinders University, GPO Box 2100, Ade- von Soest, 2012; Allen et al., 2013; Fairweather-Schmidt & Wade,
laide, South Australia 5001, Australia. E-mail: tracey.wade@flinders 2015; Klump et al., 2007; Kotler, Cohen, Davies, Pine, & Walsh,
.edu.au 2001; Neumark-Sztainer, Wall, Larson, Eisenberg, & Loth, 2011).
369
370 FAIRWEATHER-SCHMIDT AND WADE

Table 1
Studies Examining Trajectory of Disordered Eating (DE) Over Adolescence in Unselected Female Samples

N females Age at N waves of N disordered eating


Study baseline: M (SD) data collection DE measure trajectories: type (%)

Smith, Simmons, Flory, 394 3 Bulimia Test-Revised 4 binge eating: absence (77),
Annus, & Hill, 2007 (Thelen, Farmer, stable moderate (12),
Wonderlich, & Smith, increasing (7), decreasing
1991) from high to moderate (4)
12.84 (no SD 4 purging: absence (82),
reported) stable moderate (9),
increasing (8), decreasing
from high to moderate (1)
Aimé, Craig, Pepler, Jiang, & 426 4 Eating Attitudes Test 5: None (43), stable low
Connolly, 2008 (Garner & Garfinkel, level (29), high and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

1979) declining (16), chronic


This document is copyrighted by the American Psychological Association or one of its allied publishers.

high (8), increasing (4)


15.2 (.92)
Fay & Lerner, 2013 714 3 2 subscales of the Eating 6 drive for thinness: low
Disorder Inventory stable (32), decreasing
(Garner et al., 1983): (24), increasing/decreasing
drive for thinness and (17), increasing (12), high
body dissatisfaction stable (7.5), very low
stable (7.5)
14.90 (1.00) 5 body dissatisfaction: low
stable (34), very low
stable (32), moderate
stable (27), high stable
(6), none stable (1)
Pearson & Smith, 2015 938 8 3 items from the Eating 4 binge eating: none (55),
Disorder Examination- decreasing (23), increasing
Questionnaire (12), high steady (10)
(Fairburn & Beglin,
1994)
11 years (no M or SD 3 purging: none (74), dabble
reported) (19), increasing (7)

Across the four studies that have examined the longitudinal longer duration of sport decreased the impact of depressive symp-
trajectories of disordered eating phenotypes in adolescent girls, toms on drive for thinness). Smith et al. (2007) and Pearson and
three to six trajectories have been identified (Aimé, Craig, Pepler, Smith (2015) have examined predictors of increasing trajectories
Jiang, & Connolly, 2008; Fay & Lerner, 2013; Pearson & Smith, of binge eating and purging in adolescent girls. Expectations that
2015; Smith, Simmons, Flory, Annus, & Hill, 2007), which sug- eating and thinness would provide overgeneralized life improve-
gest that a “one size fits all” trajectory approach disregards differ- ment (reinforcement expectancies) and a tendency to act rashly
ent growth curves. Consistent across different age groups and when distressed (negative urgency) predicted increases in both
phenotypes are trajectories that include absent or low levels of binge eating and purging. Additionally, increased risk of binge
disordered eating at all time points, decreasing levels of disordered eating behavior was predicted by higher levels of subjective dis-
eating from a high level of symptoms, and an increasing level of tress. To date, however, theoretical justification of risk variables
disordered eating. Also present can be moderate to high levels of investigated in this context has been sparse with some exceptions
disordered eating across all time points. Across these studies the (Pearson & Smith, 2015; Smith et al., 2007).
largest group is represented by the “low stable” trajectory, ranging A number of theories suggest that many other variables may be
from 32% to 82% of the sample. The group experiencing decreas- implicated as risk factors for increasing trajectories of disordered
ing symptoms ranges from 1% to 23%, and the most consistently eating. However, although numerous models of the development
smallest group is represented by those experiencing an increase in of disordered eating have been posited in the literature, only 10
disordered eating over time (4%–12%). exist that have resulted in evaluated interventions, the efficacy of
These studies have shown a variety of variables to predict those which have provided validation of the informing model (Pennesi
girls who experience increasing levels of disordered eating, par- & Wade, 2016). Across these 10 models (Augustus-Horvath &
ticularly symptoms relating to binge eating and purging, which are Tylka, 2011; Cooper, Wells, & Todd, 2004; Fairburn, Cooper,
more common in older adolescence (Hudson, Hiripi, Pope, & & Shafran, 2003; Lyon et al., 1997; Neumark-Sztainer, Wall,
Kessler, 2007; Lewinsohn, Striegel-Moore, & Seeley, 2000). Aimé Story, & Perry, 2003; Stice, 2001; Treasure & Schmidt, 2013;
et al. (2008) found that externalizing and internalizing behavior Wildes, Ringham, & Marcus, 2010; Wilfley, Mackenzie, Welch,
problems predicted increased disordered eating. Fay and Lerner Ayres, & Weissman, 2000; Yamamiya, Shroff, & Thompson,
(2013) found depressive symptoms predicted increased drive for 2008) the types of variables most commonly included are preoc-
thinness, which was moderated by involvement in sports (i.e., a cupation with weight and shape, self-esteem deficits, emotional
GROWTH MIXTURE MODELING OF EDE GLOBAL 371

regulation difficulties, and problems with interpersonal function- least a low stable trajectory in addition to both an increasing and
ing. Preoccupation with weight and shape includes constructs such decreasing trajectory. We also hypothesized that of all our theo-
as weight and shape concern, which refers to both body dissatis- retically justified risk factors, baseline internalization of the thin
faction and importance of control over weight, shape and eating, ideal, external pressure to be thin, and perfectionism would predict
the latter somewhat similar to the idea of reinforcement expectan- membership of the increasing trajectory.
cies. Self-esteem deficits includes low self-esteem, ineffectiveness
and low self-efficacy. Emotion regulation difficulties include
Method
mood intolerance, emotional distress, emotional dysregulation,
emotional avoidance, and negative urgency, which can lead to
emotional eating. Problems with interpersonal functioning include Participants and Procedure
both peer and family relationships.
The current study came from three waves of data from adoles-
The two most robust models identified in the literature (Pennesi
cent female–female twin pairs depicted in Figure 1, described
& Wade, 2016), as determined by their role in the development of
previously (Wade, Byrne, & Bryant-Waugh, 2008; Wade et al.,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

both efficacious and effective interventions, are the Dual Pathway


2013; Wilksch & Wade, 2009, 2010). At Wave 1, twin pairs who
This document is copyrighted by the American Psychological Association or one of its allied publishers.

model (Stice, 2001) and the Transdiagnostic model (Fairburn et al.,


were between 12 and 15 years of age registered with the Australian
2003). These models further highlight the importance of internal-
Twin Registry (ATR) were approached, along with their parents,
ization of the media-presented body ideal as one’s own personal
to participate in the present study by the ATR. Seven hundred and
ideal (e.g., “I would like my body to look like people who are on
19 families were contacted and invited to participate, and of these
TV”), external pressure to be thin, and perfectionism. When inter-
411 (57.2%) agreed, 237 (32.9%) declined, and 71 (9.9%) did not
nalization is applied to females, the focus is on internalization of
reply. Researchers approached those agreeing to take part and sent
the thin ideal associated with Western culture, and the degree to
self-report questionnaires to both parents, including those families
which it is accepted and viewed as an appropriate standard for
where the parents did not live together. After parents returned the
one’s own appearance (Heinberg, Thompson, & Stormer, 1995),
questionnaires, the EDE (Fairburn & Cooper, 1993) was con-
despite the fact that for the vast majority of females this body
ducted by telephone, such that 699 twins were interviewed at
composition is very difficult to achieve (Thompson, van den Berg,
separate times with a different interviewer for each child in the
Roehrig, Guarda, & Heinberg, 2004). The Dual Pathway model
family. The sample was Caucasian and the socioeconomic indexes
informs a prevention approach named “The Body Project” for
for areas—a standardized measure of socioeconomic status with a
young women who experience body dissatisfaction, and change in
mean of 100 (SD ⫽ 15), using an amalgam of parental occupation,
internalization of the thin ideal has been shown to fully mediate the
education (years of school), and income from 2006 census data
effects of the program on change in body dissatisfaction, and to
related to the postcode of primary residence (Farish, 2004)—was
partially mediate changes in bulimic symptoms (Stice, Marti,
101.14 (SD ⫽ 11.36). The Flinders University Clinical Ethics
Rohde, & Shaw, 2011). The Transdiagnostic model, where per-
Committee approved the data collection process; parents gave
fectionism and low self-esteem are considered important pathways
written informed consent and twins gave written assent. All twins
to the development of disordered eating, has informed the devel-
were contacted again at Wave 2 and Wave 3 (including nonre-
opment of enhanced cognitive behavior therapy, considered the
sponders); the mean duration of time between Waves 1 and 2, and
treatment of choice for eating disorders where the body mass index
Waves 1 and 3, respectively, was 1.15 years (SD ⫽ 0.17), and 2.96
is greater than 17.5 (Fairburn et al., 2015; Poulsen et al., 2014).
years (SD ⫽ 0.27).
Therefore, the main purpose of the current study was to examine
the range of theoretically justified variables described above that
could be expected to predict growth of disordered eating in ado- Measures
lescent girls. Our disordered eating variable was derived from a
Disordered eating. The twin interview consisted of two parts
gold standard interview, namely the global score of the Eating
for Waves 1 and 2; the EDE (Fairburn & Cooper, 1993) and
Disorder Examination (EDE; Fairburn & Cooper, 1993). This
questions from self-report questionnaires assessing a range of
global score has been used as the primary indicator of outcome in
variables including weight-related peer teasing and life events
treatment trials of eating disorders, both anorexia nervosa and
(Wilksch & Wade, 2009, 2010). Parents also completed self-report
bulimia nervosa, in children and adults (Fairburn et al., 2009; le
questionnaires at Waves 1 and 2, including twin weight and height.
Grange, Crosby, Rathouz, & Leventhal, 2007; Lock et al., 2010;
The interview at Wave 3 comprised only the EDE interview.
Wade, Treasure, & Schmidt, 2011). It incorporates both cognitive
Sixteen postgraduate clinical psychology trainees competent in the
and behavioral indicators of disordered eating, but not the behav-
use of the EDE conducted all interviews.
iors included in the diagnostic criteria for eating disorders. As
The EDE interview generates a global measure of disordered
such, this represents the first study of trajectories of disordered
eating over the previous 28 days (22 items) that includes four
eating over adolescence that utilizes an interview rather than
subscales weight concern, shape concern, eating concern and di-
self-report. The EDE was used over three waves of data collection
etary restraint. The items related to importance of weight and
to assess disordered eating in adolescent female twins, and differ-
shape were adapted for the younger age of the sample simply by
ent trajectories over time were identified using growth mixture
encouraging participants to verbalize the list related to the follow-
modeling (GMM), which constitutes a “person-centered” approach
ing question:
while integrating variable-centered methodology (Muthén &
Muthén, 2000), known as cluster or latent class modeling. In line Imagine the things that influence how you feel about (judge, think,
with previous studies we hypothesized that we would identify at evaluate) yourself as a person—such as how you are doing at school,
372 FAIRWEATHER-SCHMIDT AND WADE
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 1. Flow diagram illustrating the data collection process over the three waves and their respective rates
of participation.

what sort of friend you are, how you get along with other people—and chiatric Association, 2013) diagnostic criteria, the relevant cogni-
put these things in order of importance, where does your weight/shape tive criteria at that time were also assessed. This information was
fit in?” used to obtain diagnoses for threshold eating disorders using a
The EDE is generally considered the “preeminent” eating dis- methodology that has been previously described, with interrater
order interview assessment tool with sound validity and reliability diagnostic kappas ranging from 0.67 for bulimia nervosa to 1.00
(Berg, Peterson, Frazier, & Crow, 2011, 2012; Berg, Peterson, et (Fairweather-Schmidt & Wade, 2014).
al., 2012). With the present data, the global EDE (Fairburn & Body mass index (BMI) centile. As data relating to parental
Cooper, 1993) has previously been found to possess construct and report of the twins’ weight and height were highly correlated, the
convergent validly, in addition to factorial invariance, medium- mothers’ report was used, aside from instances where these data were
large cross-wave correlations and high internal reliability missing, when the fathers’ report was used. At Wave 3, twins reported
(Fairweather-Schmidt & Wade, 2015). Additionally, the EDE also their own weight and height. As recommended by the Center for
contains behavioral diagnostic items, which were assessed over the Disease Control we use BMI centile for this adolescent sample.
lifetime. Over the time frame that behaviors met Diagnostic and Measures of theoretical correlates and risk. These measures
Statistical Manual of Mental Disorders (5th ed.; American Psy- were derived from Wave 1 self-report questionnaires completed in
GROWTH MIXTURE MODELING OF EDE GLOBAL 373

an interview format, as comprehensively described previously Statistical Analysis


(Wilksch & Wade, 2010). We offer a brief description below along
with internal reliability from the current sample. Patterns of missing data. An outcome variable was formulated
Preoccupation with weight and shape and self-esteem. Body for those who completed all three waves of data collection (71% of
dissatisfaction (␣ ⫽ .91) and Ineffectiveness (␣ ⫽ .88) were measured the sample) versus those who completed less than 3. Generalized
using the Eating Disorder Inventory (EDI; Garner, Olmstead, & estimating equation (GEE) models were undertaken specifying a
Polivy, 1983). The former subscale has been used reliability with binomial model with a logit link function in order to examine if any
populations as young as 11 years (Meltzer et al., 2001) and shows baseline variables differentiated between these two groups. To ac-
high test–retest reliability over a 3-week interval (r ⫽ .97; Wear & count for the nonindependence of these data, twins were clustered by
Pratz, 1987). The total EDI has been shown to significantly discrim- family (Hanley, Negassa, Edwardes, & Forrester, 2003).
inate individuals diagnosed with anorexia nervosa from a sample Derivation of latent classes. Initially, two linear mixed mod-
without anorexia nervosa (Garner et al., 1983). els (LMM; unconditional means model, and unconditional growth
Emotional regulation. Sensitivity to punishment (␣ ⫽ .74) model) were used to determine whether significant variance in the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

and sensitivity to reward (␣ ⫽ .64) subscales, which tap into EDE global score could be related time. LMM is able to accom-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

anxiety and impulsivity respectively, were drawn from the Sensi- modate correlated observations and statistical nonindependence
tivity to Punishment and Sensitivity to Reward Questionnaire (required due to the correlated nature of twin data) and is asymp-
(Torrubia, Avila, Molto, & Caseras, 2001), which has been found totically efficient with unbalanced data. Using means and standard
to have acceptable internal and test–retest reliability, as well as deviations of existing data points, LMM employs restricted max-
construct validity (Torrubia et al., 2001). The first subscale has imum likelihood approach (Nich & Carroll, 1997) to estimation of
been found to correlate positively with Eysenck Personality Ques- missing at random data (Han & Guo, 2014), and is considered a
tionnaire (EPQ: Eysenck & Eysenck, 1978) dimension of neurot- more accurate way of dealing with missing data such as
icism (r ⫽ .69 in males and r ⫽ .47 in females). The second expectation-maximization algorithm (von Hippel, 2007).
subscale is positively associated with the EPQ impulsiveness scale, The EDE global score over the three waves of data collection
with a correlation of 0.43. was investigated using GMM, which provides a structure for post
Interpersonal function. The conflict subscale of the Family hoc derivation and description of group differences over time
Environment Scale (Moos & Moos, 1986) was used to measure (Ram & Grimm, 2009), and constitutes an approach adopted by the
parental conflict (␣ ⫽ .81), and has shown good factorial validity majority of the studies pertinent to this investigation (e.g., Fay &
between two different samples of adolescents (Robertson & Hyde, Lerner, 2013; Pearson & Smith, 2015; Smith et al., 2007). Broadly,
1982). Parental care (␣ ⫽ .91) was assessed with the parental care theory, extant research findings, clinical knowledge and statistical
subscale of the Parental Bonding Inventory (Parker, Tupling, & fit indices are all necessary tenets in the decision making process
Brown, 1979), which has good test–retest reliability, and strong to determine number of classes.
factorial structure in adolescent populations (Parker, 1990). Given The determination of the optimal number of classes is informed
a high correlation between maternal and paternal scores (r ⫽ .77), by fit statistics, usually constituting Akaike information criteria
only the former were examined in the analyses. (AIC), Bayesian information criteria (BIC), and sample-size ad-
Internalization of the thin ideal. Internalization of the thin justed BIC (ABIC). Better fitting models generally produce lower
ideal (␣ ⫽ .89) was measured using a subscale from the Multidi- absolute values (Muthén, 2003; Nylund, Asparouhov, & Muthén,
mensional Media Influence Scale (Cusumano & Thompson, 2001), 2007). Subsequently, models can also be evaluated by classifica-
which has been demonstrated to correlate significantly and posi- tion precision or entropy, where higher values (where 1 denotes
tively with body dissatisfaction among girls ages 8 –11 years. perfect classification, high entropy values ⬎0.8) are indicative of
External pressure to be thin. The Perceived Sociocultural greater precision, and identify the preferred model in the event that
Pressure Scale (␣ ⫽ .80) assessed levels of perceived pressure to relative fit indices are essentially equivalent (Muthén, 2004). Fi-
be thin from friends, family, media and dating partners (Stice, nally, likelihood ratio tests, such as Vuong–Lo–Mendell–Rubin
Ziemba, Margolis, & Flick, 1996) and significantly correlates with likelihood ratio test (VLMR-LRT) and adjusted Lo–Mendell–
recollection of pressure to lose weight in childhood (r ⫽ .51; Stice Rubin likelihood ratio test (Adjusted LRT), compare a model with
et al., 1996). Weight-related peer teasing (␣ ⫽ .87) was measured C classes with an analogous model with one fewer classes (i.e.,
using the McKnight Risk Factor Survey (The McKnight Investi- C-1). All models adjusted for age at Wave 1, and accommodated
gators, 2003) and has shown good predictive ability in children for correlated observations and statistical nonindependence by clus-
detecting development of eating disorders (Field et al., 2008). tering within families. Analyses were undertaken with Mplus
Perfectionism. Concern over mistakes (␣ ⫽ .87), parental ex- version 7.11 (for Mac; Muthén & Muthén, 1999 –2010).
pectations (␣ ⫽ .72) and parental criticism (␣ ⫽ .78) were measured Characteristics of latent classes. First, across the identified
using the Multidimensional Perfectionism Scale (FMPS; Frost, Mar- latent classes; participants were described in relation to the (a) pres-
ten, Lahart, & Rosenblate, 1990). The FMPS possesses robust psy- ence of lifetime eating disorder, and (b) timing of first eating disorder
chometrics, and is used widely in eating disorder research. Conver- diagnosis. Comparisons were initially undertaken with the crosstab
gent validity has been demonstrated with the Perfectionism subscales function. To investigate the relative likelihoods of a lifetime eating
of the EDI within a sample with anorexia nervosa (r ⫽ .59; Garner et disorder across trajectory classes, GEE models were used. To account
al., 1983). The “just right” subscale (␣ ⫽ .85) from the Vancouver for the nonindependence of these data, twins were clustered by family
Obsessional Compulsive Inventory has good convergent and discrim- (Hanley et al., 2003). Second, using LMM, our theoretical correlates
inant validity (Thordarson et al., 2004). and risk factors were compared across the trajectory classes for
374 FAIRWEATHER-SCHMIDT AND WADE

variables available at Waves 1 and 2. Models incorporated Bonferroni model with an analogous model with one fewer class, and
adjustment for multiple comparisons. indicated that in all instances (models for C2–C5) the C-1
Risk factors. GEE modeling was used to identify Wave 1 model could not be rejected in favor of the C model.
variables distinguishing the different trajectory classes. Signifi- Current recommendations (e.g., Ram & Grimm, 2009) encour-
cance was adjusted using Bonferroni-corrected levels appropriate age model selection to be driven by all available fit indices in
for the multiple and post hoc testing required when undertaking a partnership with domain specific theory and existing research
number of comparisons. Initially, GEE were univariate, and vari- findings. Although the four-class model had slightly better fit in
ables with p values ⬍0.1 were subsequently simultaneously en- terms of AIC/BIC (but very similar entropy), it had very small
tered into a multivariate model to determine the predictors inde- numbers in one of the four classes (only 2% of the sample), and
pendently discriminating between a low-static and escalating class. split the increasing trajectory into two inclines. Further, the 3-class
Analyses incorporated Bonferroni adjustment for multiple com-
model was more consistent with previous pertinent trajectory
parisons and were undertaken with SPSS version 22.
studies (e.g., Pearson & Smith, 2015), in which the increasing
classes were amalgamated. Therefore, we decided that the three-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Results class model was most appropriate for these data based on extant
This document is copyrighted by the American Psychological Association or one of its allied publishers.

theory and empirical findings (e.g., Hudson Hiripi et al., 2007;


Patterns of Missing Data Pearson & Smith, 2015; Stice, Killen, Hayward, & Taylor, 1998).
Analysis of baseline predictors of participation across all three Figure 2 depicts these three classes as unique trajectories: Class 1
waves compared with participation in less than three waves indi- was low and horizontal, termed low-static, showing no change
cate no difference between key study variables (Table 2). There- over time; Class 2 was characterized by a decreasing EDE global
fore data was interpreted as being missing at random. score over time, termed attenuating; Class 3 increased, and was
termed escalating.
Identification of Trajectories of Growth of
Disordered Eating
Characteristics of the Latent Classes
Unconditional means model showed significant variance in
the EDE global, F(1, 353.93) ⫽ 283.28, p ⬍ .001, which the Table 4 presents the numbers and respective proportions of
unconditional growth model indicated was unrelated to time, eating disorders associated with each latent class. In low-static,
F(1, 1592.10) ⫽ 0.46, p ⫽ 0.50. Six models (specifying 1 about one third of eating disorders were diagnosed at each of the
through to 6 latent classes) were run with the EDE global score three waves. In contrast, the vast majority of eating disorders in
across the three waves of data. In order to identify the most attenuating were diagnosed at the first wave, whereas 84.2% of the
robust model for these observed data, three indicators of model eating disorders in escalating were diagnosed subsequent to Wave
fit were consulted. The first, the information criteria (AIC, BIC 1. Further GEE analyses identified that, relative to low-static, odds
and ABIC), are reported in Table 3. Each of AIC, BIC and of having any lifetime eating disorder were significantly elevated
ABIC criteria incrementally decreased with increasing numbers for attenuating (odds ratio [OR] ⫽ 10.36, 95% confidence interval
of latent classes specified. Second, increasing the number of [CI] [5.52, 19.45]), and escalating (OR ⫽ 34.63, 95% CI [12.49,
classes resulted in deteriorating accuracy of group membership 95.98]).
classification (entropy), where classification confidence was Table 5 presents comparisons of age-adjusted means (standard
lowest for the 6-class model. Third, likelihood ratio tests errors) and effect sizes for eating disorder risk factors for each of
(VLMR-LRT and adjusted LRT) were used to compare each
the three latent classes. At Wave 1, the majority of variables
showed attenuating to be significantly higher than low-static on the
Table 2 majority of risk factors, consistent with the finding that most of
Baseline Predictors of Participation at All Three Waves Versus those reporting a lifetime eating disorder in this group experienced
Less Than 3 Waves this earlier in adolescence. Despite the fact that most of the eating
disorders in the escalating trajectory group did not emerge until
Variable Wald ␹2(p) Exp B (95% CI) Waves 2 and 3, four key risk variables (ineffectiveness, internal-
Peer teasing 0.16 (.68) 0.98 [0.86, 1.10] ization of the thin ideal, body dissatisfaction, pressure to be thin)
Internalization of thin ideal 0.01 (.93) 1.00 [0.93, 1.08] were already significantly higher at Wave 1 compared with low-
Ineffectiveness 0.16 (.69) 0.98 [0.89, 1.08] static. However, mean scores on the aforementioned variables
Body dissatisfaction 1.43 (.23) 0.97 [0.93, 1.02]
were still notably lower in the escalating group compared with the
Concern over mistakes 1.43 (.23) 0.93 [0.83, 1.05]
Sensitivity to punishment 0.00 (.99) 1.00 [0.90, 1.11] attenuating group.
Sensitivity to reward 0.42 (.52) 0.97 [0.89, 1.06] Although the BMI centile were relatively comparable among the
Just right 0.09 (.76) 0.99 [0.95, 1.04] three classes at Wave 2, the EDE global score significantly dis-
Pressure to be thin 1.34 (.25) 0.92 [0.79, 1.06]
BMI centile 0.07 (.80) 1.00 [0.999, 1.002] criminated between low-static and the other two classes. However,
Global EDE 0.52 (.47) 0.96 [0.85, 1.08] there was little to distinguish attenuating and escalating trajectories
Age 0.15 (.70) 1.04 [0.85, 1.28] at this assessment point. Contrastingly, at Wave 3 the EDE global
Note. Exp ⫽ exponent; CI ⫽ confidence interval; BMI ⫽ body mass was now clearly highest in the escalating group, whereas low-static
index; EDE ⫽ Eating Disorder Examination. had the lowest (significantly) EDE global score.
GROWTH MIXTURE MODELING OF EDE GLOBAL 375

Table 3
Information Criteria for Growth Mixture Modeling With 1– 6 Classes

Fit statistics
No. of free
Classes parameters AIC BIC ABIC Entropy LRTb p

Nc ⫽ 1 10 3,432.19 3,477.67 3,445.92


Nc ⫽ 2 14 3,017.81 3,081.48 3,037.03 .972 .072/.077
Nc ⫽ 3 18 2,789.71 2,871.58 2,814.42 .967 .256/.266
Nc ⫽ 4 22 2,622.64 2,722.71 2,652.85 .949 .062/.066
Nc ⫽ 5 26 2,521.46 2,639.71 2,557.15 .947 .707/.697
Nc ⫽ 6 30 2,459.04a 2,595.48 2,500.23 .944 a

Note. AIC ⫽ Akaike information criteria; BIC ⫽ Bayesian information criteria; ABIC ⫽ sample size adjusted BIC; LRT ⫽ likelihood ratio tests; N ⫽
number; c ⫽ classes.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

a
Starting value run(s) did not converge. b LRT (e.g., Vuong–Lo–Mendell–Rubin likelihood ratio tests; Lo–Mendell–Rubin adjusted LRT test) quantify
This document is copyrighted by the American Psychological Association or one of its allied publishers.

specific comparisons between the model of interest and a model with one fewer class, C-1.

Baseline Factors Distinguishing Escalating From disordered eating, based on model fit and extant empirical findings
Flat-Static Groups (Aimé et al., 2008; Fay & Lerner, 2013; Pearson & Smith, 2015;
Smith et al., 2007), namely low-static, attenuating, and escalating
Correlations between the global EDE and Wave 1 risk factors is over time. In other words, this supports the suggestion that the
provided in the online supplement table, and ranged from 0.105 reason we do not observe sample-wide mean increases in EDE
(parental expectations) to 0.686 (body dissatisfaction). After ad- global with age is that different youth proceed along different
justing for age, BMI centile and zygosity, univariate GEE analyses trajectories, with some scores going down, others going up, and
identified 5 of 13 theoretically important variables significantly others staying the same.
differentiated escalating from those in low-static as shown in Table We were able to identify a number of correlates for each
6. These five variables (p ⬍ .05) were simultaneously entered into class—with the exception to the low-static trajectory that was
a multivariable model, which revealed that internalization of the characterized by a lack of EDE score growth over adolescence, and
thin ideal (OR ⫽ 2.47, 95% CI [1.36, 4.49]) was the only factor to little variation in the proportion of those first diagnosed with an
uniquely discriminate the escalating from low-static trajectory. eating disorder at each time point. Further, the low-static trajectory
had a relatively low prevalence (9.9%) of lifetime eating disorders,
Discussion and a reasonably stable BMI centile over time (around 50th per-
Consistent with some previous studies (Allen et al., 2013; centile). The second identifiable trajectory, attenuating, was pri-
Fairweather-Schmidt & Wade, 2015) our analyses showed no marily distinguishable by decreasing growth and a noticeably
evidence of mean levels of EDE global scores increasing among elevated baseline BMI centile. Accordingly, although approxi-
adolescent females over time, although significant variance in mately half of these adolescents had a lifetime eating disorder
these scores existed across adolescence. In line with our first diagnosis, two thirds of the eating disorder diagnoses were present
hypothesis, we were able to identify three different trajectories of at the first wave. In view of the early onset of the attenuating
trajectory, it is unsurprising 12 of the 13 theoretical risk factors
were elevated, and significantly distinguishable from the low-static
trajectory. The third trajectory was discernible from the other
classes by elevated growth over adolescence. Although the vast
majority of first time diagnoses in this escalating group were made
later in adolescence, significant differences from the low-static
trajectory were already observable at Wave 1— especially BMI
centile, ineffectiveness, internalization of the thin ideal, body
dissatisfaction, perceived pressure to be thin, and weight related
peer teasing.
The global EDE score was also found to robustly indicate
probability of future clinical diagnoses, as 84% of those in the
escalating trajectory group reported subsequent lifetime eating
disorders in mid to late adolescence. Thus, there appears to be
significant utility in use of EDE global information, the benefit of
which may reside in the detection of a preventive window inferred
Figure 2. Trajectories of three latent classes of mean Eating Disorder
Examination (EDE) global scores identified by growth mixture model-
by an early elevation in EDE global scores. Further, relative to the
ing analyses across three data points. These trajectories represent raw low-static trajectory, the escalating trajectory was much more
mean EDE global scores for Waves 1–3 (edew1– edew3). Class 1 ⫽ likely to have a lifetime eating disorder diagnosis, suggesting that
low-static, n ⫽ 627; Class 2 ⫽ attenuating, n ⫽ 47; and, Class 3 ⫽ the identification of key discriminating variables for these groups
escalating, n ⫽ 24. may provide opportunities to arrest DE development.
376 FAIRWEATHER-SCHMIDT AND WADE

Table 4
Proportions by Latent Class Detailing Time of First Eating Disorder Diagnosis and Presence of
Lifetime Eating Disorders

Class 1 low-static, Class 2 attenuating, Class 3 escalating,


Variable N ⫽ 627, N (%) N ⫽ 47, N (%) N ⫽ 24, N (%)

Lifetime eating disorder 62 (9.89) 25 (53.19) 19 (79.17)


First diagnosed with eating disorder
Wave 1 18 (29.03) 16 (64.00) 3 (15.78)
Wave 2 20 (32.26) 8 (32.00) 8 (42.11)
Wave 3 24 (38.71) 1 (4.00) 8 (42.11)

In addressing our main aim, we undertook univariate and mul- Braet, 2011; Cafri, Yamamiya, Brannick, & Thompson, 2005).
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

tivariate analyses to identify whether theoretically derived risk There is considerable evidence supporting this proposition, includ-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

factors were able to differentiate between the low-static and esca- ing studies of cultural change as their exposure to Western culture
lating trajectories. Univariate analyses identified a number of increases (Becker, Gilman, & Burwell, 2005; Hoek et al., 2005;
predictors, including anxiety, ineffectiveness, body dissatisfaction, Pavlova, Uher, Dragomirecka, & Papezova, 2010).
internalization of the thin ideal and pressure to be thin. Partially However, although psychosocial/cultural influences have gen-
consistent with our second hypothesis, only one variable, internal- erally been considered responsible for the promotion of thin ideal
ization of the thin ideal, significantly discriminated these two (Polivy & Herman, 2002), the broad exposure of society to ideal-
populations in the multivariate analysis, where a subpopulation ized media images of slimness begs the question as to why a
characterized by an escalating trajectory of disordered eating smaller proportion of women progress to developing eating disor-
across adolescence was associated with higher baseline internal- ders in Westernized societies, whereas the majority do not (Polivy
ization of the thin ideal. & Herman, 2004). This points to the second component of inter-
There are two components related to internalization of the thin nalization of the thin ideal, which is the adoption of this ideal for
ideal. The first is exposure to the unrealistic degree of thinness oneself, and an aspiration to work toward this ideal. Contemporary
valued by the Western beauty ideal, which does predict disordered work suggests that expectancies that thinness would provide over-
eating, including clinical eating disorders (Boone, Soenens, & generalized life improvement including intra- and interpersonal

Table 5
Age-Adjusted Characteristics of the Latent Classes

Class 1 low-static Class 2 attenuating Class 3 escalating


Variable (N ⫽ 627) M (SE) (N ⫽ 47) M (SE) (N ⫽ 24) M (SE) F(df), ES

BMI centile Wave 1 50.33 (1.20)a 68.98 (4.41)b 60.14 (4.16)b 9.22 (694)ⴱ, 0.23
EDE global Wave 1 0.26 (0.02)a 2.34 (0.06)b 0.91 (0.08)c 694.81 (694)ⴱ, 2.00
BMI centile Wave 2 54.67 (1.78) 61.57 (5.33) 55.53 (12.46) .75 (328), .10
EDE global Wave 2 0.25 (0.02)a 1.07 (0.09)b 1.12 (0.08)c 163.96 (663)ⴱ, .99
BMI centile Wave 3 50.68 (1.64) 64.10 (6.40) 61.36 (8.24) 2.73 (331), .18
EDE global Wave 3 0.28 (0.02)a 0.89 (0.07)b 2.81 (0.11)c 302.92 (494)ⴱ, 1.56
Preoccupation with weight and shape and self-esteem
Body dissatisfaction 2.26 (0.04)a 4.30 (0.14)b 3.00 (0.20)c 101.21 (693)ⴱ, .76
Ineffectiveness 1.78 (0.03)a 3.10 (0.10)b 2.31 (0.14)c 47.94 (693)ⴱ, .71
Emotional regulation
Sensitivity to reward 2.17 (0.01)a 2.36 (0.05)b 2.28 (0.07) 4.66 (694)ⴱ, .22
Sensitivity to punishment 2.24 (0.02)a 2.82 (0.08)b 2.48 (0.11)a 28.65 (694)ⴱ, .41
Interpersonal function
Parental conflict 1.87 (0.02)a 2.01 (0.06)b 2.02 (0.08) 4.67 (674)ⴱ, .17
Mother’s care 2.39 (0.02)a 2.06 (0.06)b 2.29 (0.08) 14.28 (692)ⴱ, .29
Internalization of the thin ideal
Internalization of the thin ideal 1.51 (0.03)a 2.74 (.10)b 2.15 (0.13)c 93.65 (694)ⴱ, .73
External pressure to be thin
Perceived pressure to be thin 1.28 (0.02)a 2.13 (0.06)b 1.52 (0.08)c 96.12 (692)ⴱ, .75
Peer teasing 1.35 (0.02)a 2.40 (0.07)b 1.52 (0.10)a 96.65 (694)ⴱ, .75
Perfectionism
Parental expectations 2.31 (0.02) 2.41 (0.07) 2.44 (0.10) 1.44 (693), .09
Parental criticism 1.74 (0.02)a 2.05 (0.07)b 1.78 (0.06) 10.55 (693)ⴱ, .25
Concern over mistakes 1.65 (0.02)a 2.03 (0.06)b 1.78 (0.09) 17.80 (694)ⴱ, .32
Just right 0.60 (0.02)a 1.13 (0.08)b 0.70 (0.11)a 20.49 (693)ⴱ, .34
Note. Post hoc comparisons Bonferroni-adjusted and use alphabetical superscripts to indicate where significant differences between groups exist. BMI ⫽
body mass index; EDE ⫽ Eating Disorder Examination.

p ⬍ .05.
GROWTH MIXTURE MODELING OF EDE GLOBAL 377

Table 6 predicated on the statistical and theoretical decisions during


Risk Factors Distinguishing Low-Static From Escalating Classes model specification. Although the number and various trajec-
tories of the classes identified in the present investigation make
Univariatea OR Multivariatea sense, further research with alternative data is needed to support
Wave 1 variable (95% CI) OR (95% CI)
these extant findings. Second, baseline response rate was 49%,
Concern about mistakes 2.08 [0.77, 5.61] but this equates with other large longitudinal epidemiological
Just right 1.37 [0.75, 2.49] studies of twins and singletons (Anstey et al., 2012; Wade,
Sensitivity to punishment 2.64 [1.21, 5.74] 1.33 [0.58, 3.05]
Bergin, Tiggemann, Bulik, & Fairburn, 2006). Third, it is
Sensitivity to reward 2.97 [0.60, 14.64]
Ineffectiveness 2.42 [1.52, 3.87] 1.53 [0.87, 2.69] possible that the use of the father’ report when the mother’s
Body dissatisfaction 1.99 [1.33, 2.98] 1.25 [0.76, 2.06] report was unavailable may have introduced greater variability,
Peer teasing 1.57 [0.82, 3.02] and potentially error, in the estimates of BMI centile. Addi-
Perceived pressure to be thin 2.84 [1.37, 5.89] 0.73 [0.25, 2.13] tional error may also have been introduced by transitioning
Internalization of the thin ideal 3.19 [1.89, 5.39] 2.47 [1.36, 4.49]
Mother’s care 0.57 [0.19, 1.73] from parental reporting of weight at Wave 1 and 2, to self-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Parental conflict 1.40 [0.64, 3.03] reported weight by twins at Wave 3. Finally, participants re-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Parental expectations 1.56 [0.70, 3.49] porting disordered eating at Wave 1 were offered referrals,
Parental criticism 2.05 [0.74, 5.69] which may have influenced subsequent levels of disordered
Note. Bolded univariate predictors met significance of p ⬍ .05, and were eating at Waves 2 and 3.
carried into the multivariate model. OR ⫽ odds ratio; CI ⫽ confidence By employing GMM methodology to describe latent subpopu-
interval. lations within EDE global data, this investigation has been able to
a
Model adjusted for Wave 1 age, Wave 1 body mass index centile, and
pursue an approach more sympathetic to dimensional psychiatric
zygosity.
nosology, recently receiving further attention and support (e.g.,
Wildes & Marcus, 2013). Using the EDE global as a screening tool
life functioning (Simmons, Smith, & Hill, 2002) increases risk for at early adolescence may offer vital intervention opportunities that
the development of bingeing and purging in girls ages 11–15 years could curtail development of eating disorders typically emerging
(Pearson & Smith, 2015). Similarly, importance of control over later in early adulthood. Future research could address which of the
weight and shape, which forms part of the diagnostic criteria for EDE global items are most discriminatory for the purpose of
both anorexia nervosa and bulimia nervosa, and has been described identifying future risk, in order to formulate a shorter measure that
as the “core psychopathology” of eating disorders (Fairburn & is more likely to be used in population settings.
Cooper, 1993), has been found to predict the growth of disordered Given the significant link between internalization and disordered
eating behaviors in young adolescent girls (Wilksch & Wade, eating, two important implications arise. The first is of the fundamen-
2010). This construct is defined in the EDE as the degree to which tal importance of efforts that concentrate on reducing the potency of
weight or shape influence how you feel about (judge, think, the thin ideal message. It is noteworthy that prevention approaches
evaluate) yourself as a person. Variation in expectancies for over- that focus on equipping young people with skills to stand up to this
generalized life improvement from thinness and in attaching im- ideal are among the most successful for decreasing risk for disorder
portance to control over weight and shape may lead to variation in eating, including the use of media literacy in universal prevention
internalization of the thin ideal. (Wilksch et al., 2015) and cognitive dissonance in selective preven-
The belief that control over eating and weight will make life better tion (Stice, Presnell, Gau, & Shaw, 2007). Second, strategies to
is also captured in qualitative research with anorexia nervosa (Serpell, strengthen emotional regulation and problem solving such that an
Treasure, Teasdale, & Sullivan, 1999; Sternheim et al., 2012), sug- escape to control over weight, shape and appearance is not substituted
gesting that the pursuit of low weight addresses a sense of ineffec- for facing, and dealing with, life stressors and challenges continues to
tiveness, makes the person feel safe, helps communicate distress be a priority. The connection between such strategies and the preven-
related to possible rejection and abandonment, and moderates the tion of disordered eating has shown some promise in older female
experience of negative emotions. Heritability plays a significant role adolescents (Atkinson & Wade, 2015).
in relation to internalization of the thin ideal (Suisman et al., 2012),
potentially explaining why some individuals more readily internalize
References
sociocultural messages compared with others. Similarly, importance
of weight and shape has also been found to be influenced significantly Abebe, D. S., Lien, L., & von Soest, T. (2012). The development of
by heritability, with a variance of 48% and 95% CIs ranging from bulimic symptoms from adolescence to young adulthood in females and
40 –56 (Wade, Zhu, & Martin, 2011). males: A population-based longitudinal cohort study. International
Journal of Eating Disorders, 45, 737–745. http://dx.doi.org/10.1002/eat
.20950
Strengths, Limitations and Implications Aimé, A., Craig, W. M., Pepler, D., Jiang, D., & Connolly, J. (2008).
This study was able to identify and characterize longitudinal Developmental pathways of eating problems in adolescents. Interna-
tional Journal of Eating Disorders, 41, 686 – 696. http://dx.doi.org/10
changes in EDE global trajectories of subpopulations of female
.1002/eat.20561
adolescent twins. However, there are important limitations that Allen, K. L., Byrne, S. M., Oddy, W. H., & Crosby, R. D. (2013).
inform interpretation of our results. First, GMM is best de- DSM–IV–TR and DSM–5 eating disorders in adolescents: Prevalence,
scribed as a post hoc “constrained exploratory technique” (au- stability, and psychosocial correlates in a population-based sample of
thors’ emphasis, Ram & Grimm, 2009, p. 575) that has the male and female adolescents. Journal of Abnormal Psychology, 122,
capacity to reveal latent patterns of change. However, these are 720 –732. http://dx.doi.org/10.1037/a0034004
378 FAIRWEATHER-SCHMIDT AND WADE

American Psychiatric Association. (2013). Diagnostic and statistical man- Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders:
ual of mental disorders (5th ed.). Washington, DC: Author. Interview or self-report questionnaire? International Journal of Eating
Anstey, K. J., Christensen, H., Butterworth, P., Easteal, S., Mackinnon, A., Disorders, 16, 363–370. http://dx.doi.org/10.1002/1098-108X(199412)
Jacomb, T., . . . Jorm, A. F. (2012). Cohort profile: The PATH through 16:4⬍363::AID-EAT2260160405⬎3.0.CO;2-#
life project. International Journal of Epidemiology, 41, 951–960. http:// Fairburn, C. G., & Cooper, Z. (1993). The eating disorder examination. In
dx.doi.org/10.1093/ije/dyr025 C. G. Fairburn & G. T. Wilson (Eds.), Binge eating: Nature, assessment,
Atkinson, M. J., & Wade, T. D. (2015). Mindfulness-based prevention for and treatment (pp. 317–360). New York, NY: Guilford Press.
eating disorders: A school-based cluster randomized controlled study. Fairburn, C. G., Cooper, Z., Doll, H. A., O’Connor, M. E., Bohn, K.,
International Journal of Eating Disorders, 48, 1024 –1037. http://dx.doi Hawker, D. M., . . . Palmer, R. L. (2009). Transdiagnostic cognitive-
.org/10.1002/eat.22416 behavioral therapy for patients with eating disorders: A two-site trial
Augustus-Horvath, C. L., & Tylka, T. L. (2011). The acceptance model of with 60-week follow-up. The American Journal of Psychiatry, 166,
intuitive eating: A comparison of women in emerging adulthood, early 311–319. http://dx.doi.org/10.1176/appi.ajp.2008.08040608
adulthood, and middle adulthood. Journal of Counseling Psychology, Fairburn, C. G., Cooper, Z., & Shafran, R. (2003). Cognitive behaviour
58, 110 –125. http://dx.doi.org/10.1037/a0022129 therapy for eating disorders: A “transdiagnostic” theory and treatment.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Becker, A. E., Gilman, S. E., & Burwell, R. A. (2005). Changes in Behaviour Research and Therapy, 41, 509 –528. http://dx.doi.org/10
This document is copyrighted by the American Psychological Association or one of its allied publishers.

prevalence of overweight and in body image among Fijian women .1016/S0005-7967(02)00088-8


between 1989 and 1998. Obesity Research, 13, 110 –117. http://dx.doi Fairweather-Schmidt, A. K., & Wade, T. D. (2014). DSM–5 eating disor-
.org/10.1038/oby.2005.14 ders and other specified eating and feeding disorders: Is there a mean-
Berg, K. C., Peterson, C. B., Frazier, P., & Crow, S. J. (2011). Convergence ingful differentiation? International Journal of Eating Disorders, 47,
of scores on the interview and questionnaire versions of the Eating 524 –533. http://dx.doi.org/10.1002/eat.22257
Disorder Examination: A meta-analytic review. Psychological Assess- Fairweather-Schmidt, A. K., & Wade, T. D. (2015). Changes in genetic and
ment, 23, 714 –724. http://dx.doi.org/10.1037/a0023246 environmental influences on disordered eating between early and late
Berg, K. C., Peterson, C. B., Frazier, P., & Crow, S. J. (2012). Psycho- adolescence: A longitudinal twin study. Psychological Medicine, 45,
metric evaluation of the eating disorder examination and eating disorder 3249 –3258. http://dx.doi.org/10.1017/S0033291715001257
Farish, S. (2004). Funding arrangements for non-government schools
examination-questionnaire: A systematic review of the literature. Inter-
2005–2008: Recalculation of the modified Socioeconomic Status (SES)
national Journal of Eating Disorders, 45, 428 – 438. http://dx.doi.org/10
Indicator using 2001 Australian Bureau of Statistics Census Data.
.1002/eat.20931
Canberra, Australia: Department of Education, Science and Training.
Berg, K. C., Stiles-Shields, E. C., Swanson, S. A., Peterson, C. B., Lebow,
Fay, K., & Lerner, R. M. (2013). Weighing in on the issue: A longitudinal
J., & Le Grange, D. (2012). Diagnostic concordance of the interview and
analysis of the influence of selected individual factors and the sports
questionnaire versions of the eating disorder examination. International
context on the developmental trajectories of eating pathology among
Journal of Eating Disorders, 45, 850 – 855. http://dx.doi.org/10.1002/eat
adolescents. Journal of Youth and Adolescence, 42, 33–51. http://dx.doi
.20948
.org/10.1007/s10964-012-9844-x
Boone, L., Soenens, B., & Braet, C. (2011). Perfectionism, body dissatis-
Field, A. E., Javaras, K. M., Aneja, P., Kitos, N., Camargo, C. A., Jr.,
faction, and bulimic symptoms: The intervening role of perceived pres-
Taylor, C. B., & Laird, N. M. (2008). Family, peer, and media predictors
sure to be thin and thin ideal internalization. Journal of Social and
of becoming eating disordered. Archives of Pediatrics & Adolescent
Clinical Psychology, 30, 1043–1068. http://dx.doi.org/10.1521/jscp
Medicine, 162, 574 –579. http://dx.doi.org/10.1001/archpedi.162.6.574
.2011.30.10.1043
Frost, R. O., Marten, P., Lahart, C., & Rosenblate, R. (1990). The dimen-
Cafri, G., Yamamiya, Y., Brannick, M., & Thompson, J. K. (2005). The
sions of perfectionism. Cognitive Therapy and Research, 14, 449 – 468.
influence of sociocultural factors on body image: A meta-analysis. http://dx.doi.org/10.1007/BF01172967
Clinical Psychology: Science and Practice, 12, 421– 433. http://dx.doi Garner, D. M., & Garfinkel, P. E. (1979). The Eating Attitudes Test: An
.org/10.1093/clipsy.bpi053 index of the symptoms of anorexia nervosa. Psychological Medicine, 9,
Cooper, M. J., Wells, A., & Todd, G. (2004). A cognitive model of bulimia 273–279. http://dx.doi.org/10.1017/S0033291700030762
nervosa. British Journal of Clinical Psychology, 43, 1–16. http://dx.doi Garner, D. M., Olmstead, M. P., & Polivy, J. (1983). Development and
.org/10.1348/014466504772812931 validation of a multidimensional eating disorder inventory for anorexia
Culbert, K. M., Racine, S. E., & Klump, K. L. (2015). Research review: nervosa and bulimia. International Journal of Eating Disorders, 2,
What we have learned about the causes of eating disorders—A synthesis 15–34. http://dx.doi.org/10.1002/1098-108X(198321)2:2⬍15::AID-
of sociocultural, psychological, and biological research. Journal of Child EAT2260020203⬎3.0.CO;2-6
Psychology and Psychiatry, and Allied Disciplines, 56, 1141–1164. Han, K. T., & Guo, F. (2014). Impact of violation of the missing-at-random
http://dx.doi.org/10.1111/jcpp.12441 assumption on full-information maximum likelihood method in multi-
Cusumano, D. L., & Thompson, J. K. (2001). Media influence and body dimensional adaptive testing. Practical Assessment, Research & Evalu-
image in 8 –11-year-old boys and girls: A preliminary report on the ation, 19, 1–11. Retrieved from http://pareonline.net/getvn
multidimensional media influence scale. International Journal of Eating .asp?v⫽19&n⫽2
Disorders, 29, 37– 44. http://dx.doi.org/10.1002/1098-108X(200101)29: Hanley, J. A., Negassa, A., Edwardes, M. D., & Forrester, J. E. (2003).
1⬍37::AID-EAT6⬎3.0.CO;2-G Statistical analysis of correlated data using generalized estimating equa-
Eysenck, S. B. G., & Eysenck, H. J. (1978). Impulsiveness and venture- tions: An orientation. American Journal of Epidemiology, 157, 364 –375.
someness: Their position in a dimensional system of personality descrip- http://dx.doi.org/10.1093/aje/kwf215
tion. Psychological Reports, 43, 1247–1255. http://dx.doi.org/10.2466/ Heinberg, L. J., Thompson, J. K., & Stormer, S. (1995). Development and
pr0.1978.43.3f.1247 validation of the Sociocultural Attitudes Towards Appearance Question-
Fairburn, C. G., Bailey-Straebler, S., Basden, S., Doll, H. A., Jones, R., naire. International Journal of Eating Disorders, 17, 81– 89. http://dx
Murphy, R., . . . Cooper, Z. (2015). A transdiagnostic comparison of .doi.org/10.1002/1098-108X(199501)17:1⬍81::AID-EAT226017
enhanced cognitive behaviour therapy (CBT-E) and interpersonal psy- 0111⬎3.0.CO;2-Y
chotherapy in the treatment of eating disorders. Behaviour Research and Hoek, H. W., van Harten, P. N., Hermans, K. M., Katzman, M. A.,
Therapy, 70, 64 –71. http://dx.doi.org/10.1016/j.brat.2015.04.010 Matroos, G. E., & Susser, E. S. (2005). The incidence of anorexia
GROWTH MIXTURE MODELING OF EDE GLOBAL 379

nervosa on Curaçao. The American Journal of Psychiatry, 162, 748 – Implications for prevention programs. Health Psychology, 22, 88 –98.
752. http://dx.doi.org/10.1176/appi.ajp.162.4.748 http://dx.doi.org/10.1037/0278-6133.22.1.88
Hudson, J. I., Hiripi, E., Pope, H. G. J., Jr., & Kessler, R. C. (2007). The Nich, C., & Carroll, K. (1997). Now you see it, now you don’t: A
prevalence and correlates of eating disorders in the National Comorbid- comparison of traditional versus random-effects regression models in the
ity Survey Replication. Biological Psychiatry, 61, 348 –358. http://dx analysis of longitudinal follow-up data from a clinical trial. Journal of
.doi.org/10.1016/j.biopsych.2006.03.040 Consulting and Clinical Psychology, 65, 252–261. http://dx.doi.org/10
Klump, K. L., Burt, S. A., McGue, M., & Iacono, W. G. (2007). Changes .1037/0022-006X.65.2.252
in genetic and environmental influences on disordered eating across Nylund, K. L., Asparouhov, T., & Muthén, B. O. (2007). Deciding on the
adolescence: A longitudinal twin study. Archives of General Psychiatry, number of classes in latent class analysis and growth mixture modeling:
64, 1409 –1415. http://dx.doi.org/10.1001/archpsyc.64.12.1409 A Monte Carlo simulation study. Structural Equation Modeling, 14,
Kotler, L. A., Cohen, P., Davies, M., Pine, D. S., & Walsh, B. T. (2001). 535–569. http://dx.doi.org/10.1080/10705510701575396
Longitudinal relationships between childhood, adolescent, and adult Parker, G. (1990). The Parental Bonding Instrument: A decade of research.
eating disorders. Journal of the American Academy of Child & Adoles- Social Psychiatry and Psychiatric Epidemiology, 25, 281–282. http://dx
cent Psychiatry, 40, 1434 –1440. http://dx.doi.org/10.1097/00004583- .doi.org/10.1007/BF00782881
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

200112000-00014 Parker, G., Tupling, H., & Brown, L. B. (1979). A Parental bonding
This document is copyrighted by the American Psychological Association or one of its allied publishers.

le Grange, D., Crosby, R. D., Rathouz, P. J., & Leventhal, B. L. (2007). instrument. The British Journal of Medical Psychology, 52, 1–10. http://
A randomized controlled comparison of family-based treatment and dx.doi.org/10.1111/j.2044-8341.1979.tb02487.x
supportive psychotherapy for adolescent bulimia nervosa. Archives of Pavlova, B., Uher, R., Dragomirecka, E., & Papezova, H. (2010). Trends in
General Psychiatry, 64, 1049 –1056. http://dx.doi.org/10.1001/ hospital admissions for eating disorders in a country undergoing a
archpsyc.64.9.1049 socio-cultural transition, the Czech Republic 1981–2005. Social Psychi-
Lewinsohn, P. M., Striegel-Moore, R. H., & Seeley, J. R. (2000). Epide- atry and Psychiatric Epidemiology, 45, 541–550. http://dx.doi.org/10
miology and natural course of eating disorders in young women from .1007/s00127-009-0092-7
adolescence to young adulthood. Journal of the American Academy of Pearson, C. M., & Smith, G. T. (2015). Bulimic symptom onset in young
Child & Adolescent Psychiatry, 39, 1284 –1292. http://dx.doi.org/10 girls: A longitudinal trajectory analysis. Journal of Abnormal Psychol-
.1097/00004583-200010000-00016 ogy, 124, 1003–1013. http://dx.doi.org/10.1037/abn0000111
Lock, J., Le Grange, D., Agras, W. S., Moye, A., Bryson, S. W., & Jo, B. Pennesi, J., & Wade, T. D. (2016). A systematic review of the existing
(2010). Randomized clinical trial comparing family-based treatment models of disordered eating: Do they inform the development of effec-
with adolescent-focused individual therapy for adolescents with an- tive interventions? Clinical Psychology Review, 43C, 175–192. http://
orexia nervosa. Archives of General Psychiatry, 67, 1025–1032. http:// dx.doi.org/10.1016/j.cpr.2015.12.004
dx.doi.org/10.1001/archgenpsychiatry.2010.128 Polivy, J., & Herman, C. P. (2002). Causes of eating disorders. Annual
Lyon, M., Chatoor, I., Atkins, D., Silber, T., Mosimann, J., & Gray, J. Review of Psychology, 53, 187–213. http://dx.doi.org/10.1146/annurev
(1997). Testing the hypothesis of the multidimensional model of an- .psych.53.100901.135103
orexia nervosa in adolescents. Adolescence, 32, 101–111. Polivy, J., & Herman, C. P. (2004). Sociocultural idealization of thin
The McKnight Investigators. (2003). Risk factors for the onset of eating female body shapes: An introduction to the special issue on body image
disorders in adolescent girls: Results of the McKnight Longitudinal Risk and eating disorders. Journal of Social and Clinical Psychology, 23,
Factor Study. The American Journal of Psychiatry, 160, 248 –254. 1– 6. http://dx.doi.org/10.1521/jscp.23.1.1.26986
http://dx.doi.org/10.1176/ajp.160.2.248 Poulsen, S., Lunn, S., Daniel, S. I. F., Folke, S., Mathiesen, B. B.,
Meltzer, L. J., Johnson, S. B., Prine, J. M., Banks, R. A., Desrosiers, P. M., Katznelson, H., & Fairburn, C. G. (2014). A randomized controlled trial
& Silverstein, J. H. (2001). Disordered eating, body mass, and glycemic of psychoanalytic psychotherapy or cognitive-behavioral therapy for
control in adolescents with Type 1 diabetes. Diabetes Care, 24, 678 – bulimia nervosa. The American Journal of Psychiatry, 171, 109 –116.
682. http://dx.doi.org/10.2337/diacare.24.4.678 http://dx.doi.org/10.1176/appi.ajp.2013.12121511
Moos, R. H., & Moos, B. S. (1986). Family Environment Scale (2nd ed.). Ram, N., & Grimm, K. J. (2009). Growth mixture modeling: A method for
Palo Alto, CA: Consulting Psychologists Press. identifying differences in longitudinal change among unobserved
Muthén, B. (2003). Statistical and substantive checking in growth mixture groups. International Journal of Behavioral Development, 33, 565–576.
modeling: Comment on Bauer and Curran (2003). Psychological Meth- http://dx.doi.org/10.1177/0165025409343765
ods, 8, 369 –377. http://dx.doi.org/10.1037/1082-989X.8.3.369 Robertson, D. U., & Hyde, J. S. (1982). The factorial validity of the Family
Muthén, B. (2004). Latent variable analysis. Growth mixture modeling and Environment Scale. Educational and Psychological Measurement, 42,
related techniques for longitudinal data. In D. Kaplan (Ed.), The Sage 1233–1241. http://dx.doi.org/10.1177/001316448204200433
handbook of quantitative methodology for the social sciences (pp. 345– Serpell, L., Treasure, J., Teasdale, J., & Sullivan, V. (1999). Anorexia
368). Thousand Oaks, CA: Sage. http://dx.doi.org/10.4135/97814 nervosa: Friend or foe? International Journal of Eating Disorders, 25,
12986311.n19 177–186. http://dx.doi.org/10.1002/(SICI)1098-108X(199903)25:
Muthén, B., & Muthén, L. K. (2000). Integrating person-centered and 2⬍177::AID-EAT7⬎3.0.CO;2-D
variable-centered analyses: Growth mixture modeling with latent trajec- Simmons, J. R., Smith, G. T., & Hill, K. K. (2002). Validation of eating
tory classes. Alcoholism: Clinical and Experimental Research, 24, 882– and dieting expectancy measures in two adolescent samples. Interna-
891. http://dx.doi.org/10.1111/j.1530-0277.2000.tb02070.x tional Journal of Eating Disorders, 31, 461– 473. http://dx.doi.org/10
Muthén, L. K., & Muthén, B. O. (1999 –2010). Mplus user’s guide (3rd .1002/eat.10034
ed.). Los Angeles, CA: Author. Smith, G. T., Simmons, J. R., Flory, K., Annus, A. M., & Hill, K. K.
Neumark-Sztainer, D., Wall, M., Larson, N. I., Eisenberg, M. E., & Loth, (2007). Thinness and eating expectancies predict subsequent binge-
K. (2011). Dieting and disordered eating behaviors from adolescence to eating and purging behavior among adolescent girls. Journal of Abnor-
young adulthood: Findings from a 10-year longitudinal study. Journal of mal Psychology, 116, 188 –197. http://dx.doi.org/10.1037/0021-843X
the American Dietetic Association, 111, 1004 –1011. http://dx.doi.org/ .116.1.188
10.1016/j.jada.2011.04.012 Sternheim, L., Startup, H., Saeidi, S., Morgan, J., Hugo, P., Russell, A., &
Neumark-Sztainer, D., Wall, M. M., Story, M., & Perry, C. L. (2003). Schmidt, U. (2012). Understanding catastrophic worry in eating disor-
Correlates of unhealthy weight-control behaviors among adolescents: ders: Process and content characteristics. Journal of Behavior Therapy
380 FAIRWEATHER-SCHMIDT AND WADE

and Experimental Psychiatry, 43, 1095–1103. http://dx.doi.org/10.1016/ Journal of Psychiatry, 40, 121–128. http://dx.doi.org/10.1080/j.1440-
j.jbtep.2012.05.006 1614.2006.01758.x
Stice, E. (2001). A prospective test of the dual-pathway model of bulimic Wade, T. D., Byrne, S., & Bryant-Waugh, R. (2008). The eating disorder
pathology: Mediating effects of dieting and negative affect. Journal of examination: Norms and construct validity with young and middle
Abnormal Psychology, 110, 124 –135. http://dx.doi.org/10.1037/0021- adolescent girls. International Journal of Eating Disorders, 41, 551–
843X.110.1.124 558. http://dx.doi.org/10.1002/eat.20526
Stice, E., Killen, J. D., Hayward, C., & Taylor, C. B. (1998). Support for Wade, T. D., Hansell, N. K., Crosby, R. D., Bryant-Waugh, R., Treasure,
the continuity hypothesis of bulimic pathology. Journal of Consulting J., Nixon, R., . . . Martin, N. G. (2013). A study of changes in genetic and
and Clinical Psychology, 66, 784 –790. http://dx.doi.org/10.1037/0022- environmental influences on weight and shape concern across adoles-
006X.66.5.784 cence. Journal of Abnormal Psychology, 122, 119 –130. http://dx.doi
Stice, E., Marti, C. N., Rohde, P., & Shaw, H. (2011). Testing mediators .org/10.1037/a0030290
hypothesized to account for the effects of a dissonance-based eating Wade, T. D., & O’Shea, A. (2015). DSM–5 unspecified feeding and eating
disorder prevention program over longer term follow-up. Journal of disorders in adolescents: What do they look like and are they clinically
Consulting and Clinical Psychology, 79, 398 – 405. http://dx.doi.org/10 significant? International Journal of Eating Disorders, 48, 367–374.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

.1037/a0023321 http://dx.doi.org/10.1002/eat.22303
Wade, T. D., Treasure, J., & Schmidt, U. (2011). A case series evaluation
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Stice, E., Marti, C. N., Shaw, H., & Jaconis, M. (2009). An 8-year
longitudinal study of the natural history of threshold, subthreshold, and of the Maudsley Model for treatment of adults with anorexia nervosa.
partial eating disorders from a community sample of adolescents. Jour- European Eating Disorders Review, 19, 382–389. http://dx.doi.org/10
nal of Abnormal Psychology, 118, 587–597. http://dx.doi.org/10.1037/ .1002/erv.1078
a0016481 Wade, T. D., Wilksch, S. M., & Lee, C. (2012). A longitudinal investiga-
Stice, E., Presnell, K., Gau, J., & Shaw, H. (2007). Testing mediators of tion of the impact of disordered eating on young women’s quality of life.
intervention effects in randomized controlled trials: An evaluation of Health Psychology, 31, 352–359. http://dx.doi.org/10.1037/a0025956
two eating disorder prevention programs. Journal of Consulting and Wade, T. D., Zhu, G., & Martin, N. G. (2011). Undue influence of weight
Clinical Psychology, 75, 20 –32. http://dx.doi.org/10.1037/0022-006X and shape: Is it distinct from body dissatisfaction and concern about
weight and shape? Psychological Medicine, 41, 819 – 828. http://dx.doi
.75.1.20
.org/10.1017/S0033291710001066
Stice, E., Ziemba, C., Margolis, J., & Flick, P. (1996). The dual pathway
Wear, R. W., & Pratz, O. (1987). Test–retest reliability for the eating
model differentiates bulimics, subclinical bulimics, and controls: Testing
disorder inventory. International Journal of Eating Disorders, 6, 767–
the continuity hypothesis. Behavior Therapy, 27, 531–549. http://dx.doi
769. http://dx.doi.org/10.1002/1098-108X(198711)6:6⬍767::AID-
.org/10.1016/S0005-7894(96)80042-6
EAT2260060611⬎3.0.CO;2-V
Suisman, J. L., O’Connor, S. M., Sperry, S., Thompson, J. K., Keel, P. K.,
Wildes, J. E., & Marcus, M. D. (2013). Incorporating dimensions into the
Burt, S. A., . . . Klump, K. L. (2012). Genetic and environmental
classification of eating disorders: Three models and their implications
influences on thin-ideal internalization. International Journal of Eating
for research and clinical practice. International Journal of Eating Dis-
Disorders, 45, 942–948.
orders, 46, 396 – 403. http://dx.doi.org/10.1002/eat.22091
Thelen, M. H., Farmer, J., Wonderlich, S., & Smith, M. (1991). A revision
Wildes, J. E., Ringham, R. M., & Marcus, M. D. (2010). Emotion avoid-
of the bulimia test: The BULIT-R. Psychological Assessment, 3, 119 –
ance in patients with anorexia nervosa: Initial test of a functional model.
124. http://dx.doi.org/10.1037/1040-3590.3.1.119 International Journal of Eating Disorders, 43, 398 – 404. http://dx.doi
Thompson, J. K., van den Berg, P., Roehrig, M., Guarda, A. S., & .org/10.1002/eat.20730
Heinberg, L. J. (2004). The Sociocultural Attitudes Towards Appearance Wilfley, D. E., Mackenzie, K. R., Welch, R. R., Ayres, V. E., & Weissman,
Scale–3 (SATAQ-3): Development and validation. International Jour- M. M. (2000). Interpersonal psychotherapy for group. New York, NY:
nal of Eating Disorders, 35, 293–304. http://dx.doi.org/10.1002/eat Basic Books.
.10257 Wilksch, S. M., Paxton, S. J., Byrne, S. M., Austin, S. B., McLean, S. A.,
Thordarson, D. S., Radomsky, A. S., Rachman, S., Shafran, R., Sawchuk, Thompson, K. M., . . . Wade, T. D. (2015). Prevention across the
C. N., & Ralph Hakstian, A. (2004). The Vancouver Obsessional Com- spectrum: A randomized controlled trial of three programs to reduce risk
pulsive Inventory (VOCI). Behaviour Research and Therapy, 42, 1289 – factors for both eating disorders and obesity. Psychological Medicine,
1314. http://dx.doi.org/10.1016/j.brat.2003.08.007 45, 1811–1823. http://dx.doi.org/10.1017/S003329171400289X
Torrubia, R., Avila, C., Molto, J., & Caseras, X. (2001). The Sensitivity to Wilksch, S. M., & Wade, T. D. (2009). An investigation of temperament
Punishment and Sensitivity to Reward Questionnaire (SPSRQ) as a endophenotype candidates for early emergence of the core cognitive
measure of Gray’s anxiety and impulsivity dimensions. Personality and component of eating disorders. Psychological Medicine, 39, 811– 821.
Individual Differences, 31, 837– 862. http://dx.doi.org/10.1016/S0191- http://dx.doi.org/10.1017/S0033291708004261
8869(00)00183-5 Wilksch, S. M., & Wade, T. D. (2010). Risk factors for clinically signif-
Treasure, J., & Schmidt, U. (2013). The cognitive–interpersonal mainte- icant importance of shape and weight in adolescent girls. Journal of
nance model of anorexia nervosa revisited: A summary of the evidence Abnormal Psychology, 119, 206 –215. http://dx.doi.org/10.1037/
for cognitive, socio-emotional and interpersonal predisposing and per- a0017779
petuating factors. Journal of Eating Disorders, 1, 13. http://dx.doi.org/ Yamamiya, Y., Shroff, H., & Thompson, J. K. (2008). The tripartite
10.1186/2050-2974-1-13 influence model of body image and eating disturbance: A replication
von Hippel, P. (2007). Regression with missing Ys: An improved method with a Japanese sample. International Journal of Eating Disorders, 41,
for analyzing multiply imputed data. Sociological Methodology, 37, 88 –91. http://dx.doi.org/10.1002/eat.20444
83–117. http://dx.doi.org/10.1111/j.1467-9531.2007.00180.x
Wade, T. D., Bergin, J. L., Tiggemann, M., Bulik, C. M., & Fairburn, C. G. Received September 22, 2015
(2006). Prevalence and long-term course of lifetime eating disorders in Revision received December 21, 2015
an adult Australian twin cohort. The Australian and New Zealand Accepted December 25, 2015 䡲

You might also like