Received: 20 February 2020

DOI: 10.1002/erv.2805

RESEARCH ARTICLE
- -
Revised: 2 October 2020 Accepted: 24 October 2020

Transdiagnostic vulnerability factors in eating disorders:

A network analysis

Myriam Vervaet1 | Louise Puttevils1,2 | Ria H. A. Hoekstra3 | Eiko Fried4 |

Marie‐Anne Vanderhasselt1,2,5

1
Department of Head and Skin, Ghent
University, Ghent, Belgium Abstract
2
Ghent Experimental Psychiatry (GHEP) Objective: Eating disorder (ED) symptoms and transdiagnostic vulnerability
Lab, Ghent, Belgium characteristics play a crucial role in the aetiology and maintenance of EDs.
3
Department of Psychological Methods, Over the last decade, researchers have started to model complex interrelations
University of Amsterdam, Amsterdam,
The Netherlands between symptoms using network models, but the literature is limited in that
4
Department of Clinical Psychology, it has focused solely on symptoms and investigated‐specific disorders while
University of Leiden, Leiden, The ignoring transdiagnostic aspects of mental health.
Netherlands
Method: This study tackles these challenges by investigating network re-
5
Department of Experimental‐Clinical and
Health Psychology, Ghent University,
lations among core ED symptoms, comorbid clinical symptoms (depression
Ghent, Belgium and anxiety) and empirically supported vulnerability and protective mecha-
nisms (personality traits, maladaptive cognitive schemata, perfectionism and
Correspodence
Myriam Vervaet, Department of Head and resilience) in a sample of 2302 treatment‐seeking ED patients. We estimated a
Skin, Faculty of Medicine and Health regularized partial correlation network to obtain conditional dependence re-
Sciences, Corneel Heymanslaan 10‐
lations among all variables. We estimated node centrality (interconnectivity)
1K12F (Entrance 17), B‐9000 GENT,
Ghent, Belgium. and node predictability (the overall magnitude of symptom inter‐
Email: [email protected] relationships).
Results: The findings indicate a central role of overvigilance, excessive focus on
inhibiting emotions and feelings, interoceptive awareness and perfectionism.
Conclusions: These results suggest that excessive control of bodily aspects by
dietary restraint (possibly through inhibition) and interoceptive awareness
may be important constructs that warrant future research in understanding

Abbreviations: AN, anorexia nervosa; Anx, anxiety; Aut, impaired autonomy and performance; Awa, interoceptive awareness; BDI‐II, Beck
Depression Inventory‐II; BED, binge‐eating disorder; BMI, Body Mass Index; BN, bulimia nervosa; Bod, body dissatisfaction; Bul, bulimia; CIs,
confidence intervals; Coa, cooperativeness; Cri, parental criticism; CS‐coefficient, centrality stability coefficient; Dft, drive for thinness; Dir,
disconnection and rejection; DSM, Diagnostic and Statistical Manual of Mental Disorders; ED, Eating Disorder; EDI, Eating Disorder Inventory;
EDNOS, eating disorder not otherwise specified; EDs, eating disorders; EI, expected influence; Exp, parental expectations; Fea, maturity fear;
FMPS, Frost Multidimensional Perfectionism Scale; Har, harm avoidance; Imp, impulse regulation; Ine, ineffectiveness; Inh, overvigilance and
inhibition; LASSO, least absolute shrinkage and selection operator; Lim, impaired limits; M, mean; Nov, novelty seeking; OSFED, other specified
feeding and eating disorders; Red, reward dependence; Res, resilience; RS‐NL, Resilience Scale (Dutch version); SCL‐90, Symptom Check‐List 90;

-
SD, standard deviation; Sed, self‐directedness; Set, self‐transcendence; Soc, social insecurity; Sta, personal standards; STAI, State‐Trait Anxiety
Inventory; TCI‐R, Temperament and Character Inventory Revised; TPQ, Tridimensional Personality Questionnaire; YSQ, Young Schema
Questionnaire.

Eur Eat Disorders Rev. 2020;1–15. wileyonlinelibrary.com/journal/erv © 2020 Eating Disorders Association and John Wiley & Sons Ltd. 1
2
- VERVAET ET AL.

vulnerability in EDs. We provide all code and data via the Open Science
Framework.

KEYWORDS
eating disorders, inhibition, interoceptive awareness, network analysis, perfectionism

1 | INTRODUCTION
Highlights
Eating disorders (EDs) are psychiatric illnesses with se-
� This network analysis revealed a relationship
vere disturbances in people's eating behaviours, emotions
between body dissatisfaction and trans-
and food/body‐related thoughts (i.e., preoccupation with
diagnostic vulnerability factors in ED patients
food, body weight and shape). Different diagnoses exist
� The results indicate a central role of personal
within EDs, such as anorexia nervosa (AN), bulimia
standards (perfectionism), overvigilance and
nervosa (BN), binge‐eating disorder (BED) and other
inhibition (maladaptive schemata), and the
specified feeding and eating disorders (OSFED; Diag-
specific ED symptoms, ineffectiveness and
nostic and Statistical Manual of Mental Disorders
interoceptive awareness
[DSM]‐V; American Psychiatric Association, 2013). ED
� Using motivational strategies for bridging ED
diagnoses as categories are faced with numerous chal-
symptoms, adequate methods of self‐improve-
lenges, such as high rates of comorbidity with depression
ment and adaptive emotion regulation were
and anxiety (Eddy et al., 2008), heterogeneity within
recommended in the treatment of ED patients
these diagnostic categories, poor discrimination, and
frequent migration between supposedly distinct di-
agnoses (Castellini et al., 2011; Eddy et al., 2008), high
rates of not otherwise specified diagnoses (Allen, Byrne, 2015). Klik of tik om tekst in te voeren. The network theory
Oddy, & Crosby, 2013) and common factors across the hypothesizes that symptoms influence each other irre-
diagnoses that influence the maintenance of ED behav- spective of traditional diagnosis DSM boundaries (Bors-
iours, emotions, and food‐related thoughts (Fairburn, boom, 2017). In network models, ED symptoms such as
Cooper, & Shafran, 2003). Hence, instead of tailoring body checking (Forbush, Siew, & Vitevitch, 2016), fear of
treatments only to the symptoms that fall within the weight gain and feeling fat (Christian et al., 2020; Forrest,
boundaries of a specific ED, it is becoming an increas- Jones, Ortiz, & Smith, 2018; Goldschmidt et al., 2018;
ingly common practice to take a transdiagnostic stance Levinson, Brosof, Ma, Fewell, & Lenze, 2017) and shape
(Harvey, Watkins, Mansell, & Shafran, 2004). Klik of tik and weight overvaluation (DuBois, Rodgers, Franko, Eddy,
om tekst in te voeren. The transdiagnostic approach in & Thomas, 2017; Forrest et al., 2018; Wang, Jones, Dreier,
EDs focuses on identifying the common and core un- Elliott, & Grilo, 2019; for a review see; Levinson et al., 2018)
derlying mechanisms that underpin a broad array of emerged as the core symptoms with the highest centrality.
diagnostic presentations. Although this approach can be The authors have interpreted such (statistically) central
applied to different categories of DSM diagnoses, for the symptoms as (theoretically) underlying clinical manifes-
scope of the current study, our transdiagnostic stance tation of ED psychopathology. Other researchers have re-
refers to a dimensional approach in EDs given that this ported that two ED symptoms, ineffectiveness and
patient population shares a distinctive core psychopa- interoceptive awareness, are central nodes in ED networks,
thology not seen in psychiatric disorders (Cooper & both at admission and discharge of a psychiatric treatment
Grave, 2017). (Cascino et al., 2019; Monteleone et al., 2019; Olatunji,
A new and fast‐moving development in the ED field, Levinson, & Calebs, 2018; Solmi et al., 2018). These
which aligns with the transdiagnostic framework, is the emerging results point towards the importance of specific
conceptualization of ED as networks of related features food‐related and body‐related thoughts as transdiagnostic
(Levinson, Vanzhula, Brosof, & Forbush, 2018). From a factors in patients diagnosed with ED, as these core ED
network perspective, disorders are conceptualized as symptoms have the strongest relations with other nodes
complex dynamic systems of interacting symptoms, for and are interpreted by some to potentially maintain ED
which some individual symptoms play a unique and central psychopathology. These cognitive preoccupations also
role in relation to other symptoms (Van Borkulo et al., seemed predictive of the post‐treatment outcome in
VERVAET ET AL.
- 3

patients with AN Klik of tik om tekst in te voeren (although features: interoceptive awareness and drive for thinness)
this might also be due to differences in node variance; and BMI, affective symptoms such as depression and
Elliott, Jones, & Schmidt, 2020). anxiety, interpersonal sensitivity and distrust, and inef-
What comes out of a statistical model depends on fectiveness were highly interconnected in the resulting
what you put in (Forbes, Wright, Markon, & network of all patients with a diagnosis of ED, and also
Krueger, 2019). Additionally, not only ED‐related symp- within each diagnostic subgroup (AN, BN and BED). In
toms but also symptoms that are linked to psychopa- line with this reasoning, a network analysis investigating
thology in general, such as anxiety and depression, play a personality characteristics in people seeking bariatric
key role in network models of EDs. For example, studies surgery revealed that low self‐directedness, a subscale of
examining the interplay between ED‐core symptoms and the Temperament and Character Inventory Revised–was
general psychiatric symptoms in patients with AN found an independent predictor of BMI at follow‐up (Mon-
that depression and anxiety symptoms were among the teleone et al., 2019).
most central nodes within the network (Monteleone, Additionally, as another well‐known transdiagnostic
Mereu, et al., 2019; Solmi, Collantoni, Meneguzzo, Ten- vulnerability mechanism in EDs, perfectionism has been
coni, & Favaro, 2019). Interestingly, network theory defined as striving for, and achievement of, personally
dictates that in addition to symptom variables, many demanding standards, despite adverse consequences
other variables could play important roles in the (Shafran, Cooper, & Fairburn, 2002). Studies show that
complexity of ED (e.g., Fried & Cramer, 2017). As stated patients diagnosed with an ED experience a sense of
by P. J. Jones, Heeren, and McNally (2017), many non‐ ineffectiveness in many areas of their lives (S. Wagner,
symptom variables play a causal role in the aetiology and Halmi, & Maguire, 1987), associated with perfectionistic
maintenance of mental disorders and might enrich a standards to control eating, shape and weight (Riley &
‘symptom network’. Indeed, non‐symptom variables such Shafran, 2005). In addition, Wade, Wilksch, Paxton,
as personality traits, maladaptive schemata, and child- Byrne, and Austin (2015) found that the relationship
hood maltreatment feature prominently in the aetiology, between perfectionism and increased risk for EDs was
symptomatic expression, and maintenance of EDs (Cas- mediated by levels of ineffectiveness (i.e., feelings of in-
sin & Von Ranson, 2005; Farstad, McGeown, & von adequacy, insecurity, worthlessness and having no con-
Ranson, 2016; Monteleone et al., 2019; Pauwels et al., trol over one's own life). Finally, Puttevils, Vanderhasselt,
2018; Pugh, 2015; Rodgers et al., 2019). As such, well‐ and Vervaet (2019) showed that perfectionism was a
known vulnerability and protective mechanisms in pa- significant predictor of ED symptoms in a large sample of
tients diagnosed with ED could be included in network patients diagnosed with ED (AN, BN, BED and OSFED/
modelling. These vulnerability or protective factors are EDNOS).
defined as mechanisms that, respectively, increase or Moreover, maladaptive schemata operate as an
reduce the risk of mental health problems and are rather important transdiagnostic vulnerability mechanism
interrelated (Fritz, Fried, Goodyer, Wilkinson, & van within the ED. Maladaptive schemata are defined as
Harmelen, 2018). unconditional, self‐defeating emotional and cognitive
As described in the Transtheoretical Model of EDs patterns that result from negative experiences and in-
(Brytek‐Matera & Czepczor, 2017), personality plays an teractions with significant others during childhood or
important role in the onset and/or maintenance of an ED adolescence and are stable constructs (Young, Klosko, &
and is considered an important transdiagnostic vulnera- Weishaar, 2003). A number of empirical studies have
bility mechanism. In line with these findings, a meta‐ indicated that maladaptive schema cognitions play an
analysis by Farstad et al. (2016) revealed that even though important role in the development and maintenance of
personality traits explain variance in ED symptomatic psychopathology, such as EDs (Gongora, Derksen, & Van
expression and prognosis, some personality traits such as Der Staak, 2004; Leung, Waller, & Thomas, 1999; Unoka,
high self‐directedness and avoidance motivation are Tölgyes, & Czobor, 2007; Waller, Ohanian, Meyer, &
frequent among all ED diagnoses relative to controls. As Osman, 2000) and might act as a vulnerability factor for
such, Solmi et al. (2018) included the Tridimensional ED relapse (C. Jones, Harris, & Leung, 2005). For
Personality Questionnaire for personality traits when example, Talbot, Smith, Tomkins, Brockman, and Simp-
estimating their network model in patients diagnosed son (2015) Klik of tik om tekst in te voeren. reported that
with EDs. Besides personality traits, their recent study AN, BN and OSFED groups each scored significantly
also used network psychometrics to model ED symptoms, higher than a community sample group for the majority
clinical variables (Body Mass Index [BMI] and duration of maladaptive schema modes In addition, Boone, Braet,
of illness) and the Symptom Check‐List 90. The results Vandereycken, and Claes (2013) found that maladaptive
showed that, in addition to ED‐core symptoms (cognitive schemata are positively related to body image concerns in
4
- VERVAET ET AL.

an ED sample. Furthermore, different aspects of clinical all the nodes included in the network, 276 individuals
perfectionism are linked to maladaptive schema domains. were excluded from the sample, leaving 1969 patients—
Finally, not only vulnerability factors but also pro- 1886 females (95.8%) and 82 males (4.2%)—in the final
tective mechanisms, such as resilience, play an important sample. Missing values were dealt with by using pairwise
role within ED. Resilience is frequently defined as the complete observations (i.e., participants were not deleted
ability to bounce back and adapt in the face of adverse listwise, but rather all available information was used to
conditions (Kalisch, Müller, & Tüscher, 2015). A study by estimate each correlation (cf. Santos, Fried, Asafu‐Adjei,
McGrath, Julie, and Caron (2012) revealed that increased & Jeanne Ruiz, 2017). The diagnostic rates are as follows:
levels of resilience are associated with an improved body 831 Anorexia Nervosa (42.2%), 617 Bulimia Nervosa
image. Research from de Vos et al. (2017) reported that (31.3%), 371 Eating Disorder Not Otherwise Specified
psychological well‐being and resilience were found to be (18.8%) and 150 Binge‐Eating Disorder (7.6%). The sam-
fundamental criteria for ED recovery and are associated ple members ranged in age from 13 to 67 years
with a reduction of ED symptoms over time (Calvete, las (M ¼ 23.93, SD ¼ 8.85), BMI ranged from 8.79 to 61.67
Hayas, & Gómez del Barrio, 2018). In addition, Ten Ham, (M ¼ 20.1, SD ¼ 6.53) and duration of illness (in years)
Hulsbergen, and Bohlmeyer (2016) indicated that resil- ranged from 0 to 46 (M ¼ 5.65, SD ¼ 6.72). Demographics
ience could be a protective factor across EDs. for each ED can be found in Table 1. The study was part
Overall, the present study aims to use network psy- of a larger project assessing different questionnaires in
chometrics to model the complex interrelations between patients with ED, and data were collected from 1998 to
a selection of constructs derived from a literature review: 2015. All participants provided informed consent for the
(1) core ED symptoms and common psychological/ use of data in an anonymous form.
behavioral features that are linked with EDs (measured
with validated questionnaires); (2) general psychiatric
symptoms that are well‐known comorbidity factors in 2.2 | Assessments
EDs (depression and anxiety) and (3) empirically sup-
ported vulnerability and protective mechanisms (beyond For all of the subscales described below, we calculated
symptoms) underlying the onset and maintenance of EDs the subscale scores by summing all the items belonging to
(personality traits, maladaptive cognitive schemata, a subscale.
perfectionism and resilience) in a large sample of treat-
ment‐seeking patients diagnosed with an ED (AN, BN,
BED and OSFED/EDNOS). We are especially interested 2.2.1 | Eating disorder symptoms
in the strongest edges between ED symptoms (most
central items) and non‐ED symptom variables (i.e., cross‐ The Eating Disorder Inventory (EDI)‐II (Garner, 1991) is
questionnaire edges). a self‐report measure, which consists of 91 items, with
responses on a 6‐point Likert scale (ranging from ‘never’
to ‘always’). The EDI‐II includes 11 subscales: Drive for
2 | METHOD Thinness (excessive concern with dieting, preoccupation
with weight and fear of weight gain; 7 items), Bulimia
The study was conducted according to the principles of (binge eating and purging; 7 items), Body Dissatisfaction
the ‘Declaration of Helsinki’ (as amended in Tokyo, (not being satisfied with one's physical appearance; 9
Venice, Hong Kong and Somerset West) and in accor- items), Ineffectiveness (feelings of inadequacy, insecu-
dance with the Guideline for Good Clinical Practice rity, worthlessness and having no control over one's
(CPMP/ICH/135/95—17th July 1996). own life; 10 items), Perfectionism (not being satisfied
with anything less than perfect; 6 items), Interpersonal
Distrust (reluctance to form close relationships; 7
2.1 | Participants items), Interoceptive Awareness (the ability of an indi-
vidual to discriminate between sensations and feelings,
A total of 2302 participants who registered at the Centre and between the sensations of hunger and satiety; 10
of Eating Disorders of the University Hospital of Ghent items), Maturity Fears (fear of facing the demands of
for ambulant or resident treatments, were asked to adult life; 8 items), Asceticism (avoidance of sexual re-
complete a series of questionnaires. A total of 2245 pa- lationships; 8 items), Impulse Regulation (ability to
tients meeting the criteria of an ED, as defined by the regulate impulsive behaviour such as binge behaviour;
DSM‐IV1 (American Psychiatric Association, 1994), were 11 items) and Social Insecurity (social fears and inse-
included in the current study. Due to incomplete data for curity; 8 items).
VERVAET ET AL.
- 5

T A B L E 1 Demographics for each

Mean (SD)
eating disorder separately
Number duration
female Mean Mean of illness
N (%) (SD) age (SD) BMI (in years)
Anorexia nervosa (n ¼ 831) 805 (96.9) 22.33 (8.08) 15.40 (1.60) 4.39 (5.79)
Bulimia nervosa (n ¼ 617) 590 (95.6) 23.65 (7.35) 21.84 (3.84) 5.55 (5.67)
Eating disorder not otherwise 349 (94.1) 24.57 (9.74) 20.88 (5.10) 5.76 (7.30)
specified (n ¼ 371)
Binge eating disorder (n ¼ 150) 142 (94.7) 32.38 (11.15) 34.79 (7.29) 11.24 (9.58)
Abbreviations: BMI, body mass index, SD, standard deviation.

2.2.2 | Personality traits (excessive focus on inhibiting one's emotions and feelings
in order to avoid making mistakes).
The Temperament and Character Inventory (TCI; Clo-
ninger, Svrakic, & Przybeck, 1993) comprises 240 items
with responses in a true/false answer format. The items 2.2.4 | Perfectionism
were divided into seven subscales. Four of them are
temperament subscales: novelty seeking (exploratory ac- The Frost Multidimensional Perfectionism Scale (FMPS;
tivity to obtain novel stimulation and impulsive decision Frost, Marten, Lahart, & Rosenblate, 1990) is a self‐report
making), harm avoidance (excessive worrying and being questionnaire that consists of 35 items (rated on a 5‐point
fearful, doubtful, shy and easily fatigued), reward depen- Likert scale, ranging from ‘Strongly Disagree’ to ‘Strongly
dence (depending on signs of reward such as social support Agree’). The FMPS covers six dimensions of perfectionism,
and approval) and persistence (perseverance in spite of namely concern over mistakes (reflecting negative re-
fatigue or frustration). The other three subscales examine actions to errors; 9 items), personal standards (setting high
character: self‐directedness (ability to regulate and adapt standards for evaluation; 7 items), parental expectations
behaviour in order to achieve personal goals and values), (the belief that one's parents set very high standards;
cooperativeness (being agreeable in relations with other 5 items), parental criticism (the belief that one's parents
people) and self‐transcendence (experiencing spiritual were overly critical; 4 items), doubting of actions (the
ideas). tendency to doubt about one's ability; 4 items) and orga-
nization (the importance placed on orderliness; 6 items).

2.2.3 | Maladaptive domains

2.2.5 | Depressive symptoms
The Young Schema Questionnaire (YSQ; Young et al.,
2003) consists of 205 items, which are divided into 16 The Beck Depression Inventory‐II (BDI‐II; Beck, Steer, &
subscales corresponding to the 16 early maladaptive Brown, 1996) is a 21‐item self‐report questionnaire that
schemata (EMS) scales. The items are graded on a 6‐point examines the severity of depressive symptoms (ranging
Likert scale ranging from 1 (‘Completely untrue of me’) to 6 from 0 to 3), with higher scores indicating greater levels
(‘Describes me perfectly’). According to Young's schema of depression.
therapy, the 16 EMS can be grouped into different schema
domains. Each of these domains represents one frustration
of different core emotional needs during childhood. There 2.2.6 | Anxiety symptoms
are five Schema Domains: Disconnection and Rejection
(unmet needs for personal safety and stability), Impaired Patients completed the Trait version of the State‐Trait
Autonomy & Performance (inability to function indepen- Anxiety Inventory (STAI; Spielberger, 1983; Van der
dently), Impaired Limits (inability to respect internal Ploeg, 1980), a 20‐item self‐report questionnaire to measure
limits and responsibility to others), Other‐directness trait anxiety. All items are rated on a 4‐point Likert
(excessive focus on meeting needs of others at the expense scale. Scores range from 20 to 80, with higher scores
of one's own needs) and Overvigilance & Inhibition indicating greater anxiety.
6
- VERVAET ET AL.

2.2.7 | Resilience network and usually leads to a sparse structure

(Epskamp & Fried, 2018).
To measure resilience, the Dutch version of the Resil-
ience Scale (RS‐NL) was used (Portzky, Wagnild, De
Bacquer, & Audenaert, 2010; Wagnild, 2009). The RS‐NL 3.2 | Centrality estimation
consists of 25 items rated on a 4‐point Likert subscale
(ranging from ‘strongly disagree’ to ‘strongly agree’). For the 32‐item ED network, we were interested in iden-
tifying which nodes were most influential. There are
numerous graph theoretical measures that can be used for
3 | DATA ANALYSIS this purpose; however, we opted for expected influence
(EI). EI reflects the sum of all edges of a node (Robinaugh,
To gain exploratory insights into the structure of mutual Millner, & McNally, 2016), instead of the absolute sum of all
dependence relations among the included constructs, we edges (node strength). Therewith, both negative and posi-
estimated a regularized partial correlation network. Next, tive edges were considered when estimating centrality.
we identified the most central nodes, that is, nodes with
the highest number of connections, that some interpreted
as potentially influencing many other nodes in the 3.3 | Accuracy and stability estimation
network (Bos et al., 2017). In addition to examining the
central nodes, we also investigated node predictability, In order to test the stability and accuracy of the estimated
which assesses the overall magnitude of symptom inter‐ network, we carried out the routines described in a
relationships as an absolute measure as well as network tutorial paper by Epskamp, Borsboom, and Fried (2018),
accuracy and stability (Haslbeck & Fried, 2017; Haslbeck using the R‐package bootnet. First, we used bootstrapping
& Waldorp, 2018). Data and analytic codes are available routines for the 95% confidence intervals of the edge
online on the Open Science Framework (https://osf.io/ weights to estimate the accuracy of edges in the network.
ks85g/). We hope this will enable future analysis of this Second, we examined the stability of the order of the
rich, transdiagnostic ED dataset. centrality measures by subsetting bootstrap: if a central
node remains central in the estimated network after
random participants are dropped, it is considered an in-
3.1 | Network estimation dicator of the stability of the order of centrality estimates.
Related to this, we estimated the centrality stability co-
Network models were estimated and analysed using the efficient (CS‐coefficient) as well as a ‘edge stability coef-
R‐package bootnet (Epskamp, Cramer, Waldorp, ficient’ (Epskamp & Fried, 2020), indicating what
Schmittmann, & Borsboom, 2012; Haslbeck & Waldorp, proportion of participants can be dropped to retain, with
2018). In general, a network consists of nodes (variables) 95% probability, a correlation of at least 0.7 between the
and edges (connections between nodes). A main centrality order in the bootstrapped datasets with
advantage of the network approach is that it estimates the original order of centrality in the full dataset, or
and visualizes the multivariate dependencies of the data for the edge stability coefficient, between the edge
that otherwise remain hidden. However, the interpreta- weights in the bootstrapped datasets and the original
tion of network models should be handled with caution edge weights. The CS‐coefficient should be at least 0.25,
since these models do not always match network theory, and preferably above 0.5. Currently, there are no clear
and the results are equivocal (Forbes et al., 2019; Fried, thresholds for the edge stability coefficient. Subsequently,
2020). We modelled edges akin to partial correlation we estimated mean explained variance as a measure of
coefficients, meaning that a connection between nodes A node predictability (the degree to which a given node can
and B is the connection after controlling for all other be predicted by all other nodes in the network). Node
edges in the network. In order to do so, we computed a predictability is important because while network cen-
Gaussian graphical model (Epskamp & Fried, 2018). trality is a relative measure, node predictability is an
With 32 nodes, 496 pairwise association parameters were absolute measure of potential controllability of a node,
estimated. Since the estimation of many parameters in- and provides the percentage of variance explained by the
creases the risk of false positive edges, the least absolute nodes' neighbours. Generally, predictability is highly
shrinkage and selection operator (LASSO; Tibshirani, related to expected influence, but the former is somewhat
1996) was used, which sets very small edges to zero. The more informative since it can give us potential informa-
LASSO procedure aims to keep only the relevant edges tion of the clinical relevance of connections (or edges)
in order to reveal the underlying structure of the between nodes (Haslbeck & Fried, 2017). Finally, we
VERVAET ET AL.
- 7

performed difference tests to compare individual edge pred ¼ 0.79), disconnection and rejection (Dir, YSQ
weights and to compare individual centrality estimates pred ¼ 0.77), impaired autonomy & performance (Aut,
(see Figure S1 in the supplementary materials). This non‐ YSQ pred ¼ 0.77), anxiety (Anx, STAI pred ¼ 0.71),
parametric bootstrapping method was used to determine personal standards (Sta, FMPS pred ¼ 0.69), novelty
if edges (or centrality values) significantly differed from seeking (Nov, TCI pred ¼ 0.41), reward dependence (Red,
one another; note that these difference tests are not cor- TCI pred ¼ 0.38), maturity fear (Fea, EDI‐II pred ¼ 0.36),
rect for multiple testing. bulimia (Bul, EDI‐II pred ¼ 0.34) and self‐transcendence
(Set, TCI, pred ¼ 0.152), respectively, from highest to
lowest.
3.4 | Visualization

Positive edges are represented by blue lines, while 5 | CENTRALITY ESTIMATION

negative edges are printed in red. Thicker and more
saturated edges visualize stronger connections. The Figure 2 shows the standardized EI estimates, that is
Fruchterman‐Reingold algorithm (Fruchterman & Rein- interconnectedness of all nodes. We list the five most
gold, 1991) was used, where stronger and/or more con- and least connected nodes below. Overvigilance and
nections are placed closer to one another. Inhibition (Inh, YSQ; EI ¼ 1.50) as maladaptive sche-
mata, interoceptive awareness, core ED symptoms
(Awa, EDI; EI ¼ 1.40) and personal standards, as
4 | RESULTS characteristic of perfectionism (Sta, FMPS; 1.40), serve
as core features in the transdiagnostic network. Inef-
4.1 | Estimation of transdiagnostic fectiveness (Ine, EDI; 1.30), Impaired Autonomy and
network model Performance (Aut, YSQ; both ¼ 1.10) also show
moderately high indices of centrality. The two least
The 32‐item network structure is presented in Figure 1. central nodes are Resilience (Res, RS; 0.42) and Harm
Overall, we found more positive (62%) than negative Avoidance (Har, TCI; 0.40). Other nodes with low
(38%) edges. interconnectedness are Novelty Seeking (Nov, TCI;
The strongest positive associations emerged between 0.32), self‐directedness (Sed, TCI; 0.25) and weak-
Parental Criticism (Cri [nodename; for an overview of ened limits (Lim, YSQ; 0.11).
full names see legend Figure 1], FMPS [questionnaire]) Given that the main analyses—consistent with prior
and Parental Expectations (Exp, FMPS; 0.62), two sub- studies—identified expected influence as the most sta-
scales of the perfectionism questionnaire. Furthermore, bly estimated centrality metric, we present the differ-
there are strong relationships between Drive for Thinness ence test for expected influence. Figure 3 represents
(Dft, EDI) and Body Dissatisfaction (Bod, EDI; 0.43), two significant differences between centrality estimates as
core ED symptoms as well as between depression (BDI) black boxes; of note, the test does not correct for
and anxiety (Anx, STAI; 0.39). multiple testing.
We also identified some strong negative relations, the
strongest being between Harm Avoidance (Har, TCI) and
Novelty Seeking (Nov, TCI; 0.28), two personality traits 5.1 | Network accuracy and stability
as well as Impaired Limits (Lim, YSQ) and Cooperative-
ness (Coa, TCI; 0.23), and social insecurity (Soc, EDI) We estimated the accuracy and stability of the trans-
and Resilience (Res, RS; 0.21). diagnostic ED network (available at https://osf.io/ks85g/).
Additionally, we examined which ED symptom vari- The results from the edge weight bootstrap show that the
ables (as measured by the EDI subscales) are most network model was fairly accurately estimated (however,
strongly related to other non‐symptom variables. The note that clear benchmarks to interpret the CIs are as yet
strongest negative edge is already stated above (social missing in the literature). The stability analysis for the
insecurity and resilience [ 0.21]), while the strongest centrality resulted in a CS‐coefficient of 0.75, which can be
positive edge occurs between impulse regulation (Imp, considered a highly stable order of centrality (Epskamp,
EDI) and novelty seeking (Nov, TCI; 0.13). Borsboom, & Fried, 2018). The stability analysis for the
Node predictability (degree to which a node can be edge weights resulted in an ES‐coefficient of 0.75. The
predicted by all the other nodes in the network). network has a mean explained variance (a measure of node
The five nodes with the highest and lowest predict- predictability) of 0.60, which is considered a high per-
ability were overvigilance and inhibition (Ihn, YSQ; centage of node predictability with regard to other network
8
- VERVAET ET AL.

F I G U R E 1 Transdiagnostic eating disorders network. Nodes represent eating disorders core features, whereas edges represent the
regularized partial correlation between any two nodes

papers investigating psychopathology (Haslbeck & symptoms (depression and anxiety) and (3) well‐known
Fried, 2017). Results from the centrality difference test and general/transdiagnostic vulnerability mechanisms
show that EI significantly differs for most nodes from each beyond symptoms (personality traits, maladaptive
other. The node with the largest expected influence over- cognitive schemata, perfectionism and resilience) in a
vigilance and inhibition (Inh, YSQ; EI ¼ 1.50) is signifi- large sample of treatment‐seeking patients with an ED
cantly larger than most of the other nodes, with the (AN, BN, BED and OSFED/EDNOS). The study results
exception of interoceptive awareness (Awa, EDI; identify maladaptive schemata with hypervigilance and
EI ¼ 1.40), Personal Standards (Sta, FMPS; 1.40), and excessive focus on inhibiting emotions and feelings in
Ineffectiveness (Ine, EDI; 1.30). Further accuracy and order to avoid mistakes (inhibition), interoceptive aware-
stability analyses for this network are available at https:// ness, regulation of feelings of ineffectiveness, and high
osf.io/ks85g/, including edge weight significance tests personal standards (perfectionism) as key characteristics
(testing for significant differences for all edges) and cen- in ED patients in the estimated network.
trality difference tests (testing for centrality differences for ‘Hypervigilance/Inhibition’ is defined as ‘excessive
all nodes). emphasis on suppressing one's spontaneous feelings,
impulses and choices or on meeting rigid, internalized
rules and expectations about performance and ethical
6 | DISCUSSION behaviour, often at the expense of happiness, self‐
expression, relaxation, close relationships or health’
We estimated a network model based on (1) core ED (Young et al., 2003), whereas ‘interoceptive awareness’
symptoms and common psychological/behavioural fea- measures the ability of an individual to discriminate be-
tures that are linked with EDs; (2) comorbid clinical tween sensations and feelings, and between the
VERVAET ET AL.
- 9

social outcomes, including reduced spontaneity, avoid-

ance, social withdrawal, aversion to novel situations and
lack of assertiveness (Cole, Michel, & Teti, 1994; Young
et al., 2003).
The high centrality of inhibiting emotions and feelings
in EDs is in line with recent studies in AN patients (Old-
ershaw, Lavender, & Schmidt, 2018; Oldershaw, Startup, &
Lavender, 2019). The highly interconnected ‘ineffective-
ness’ and ‘interoceptive awareness’, as found in their
analysis, may also explain experienced feelings of in-
adequacy, insecurity, worthlessness and a perceived lack of
control (McLaughlin, Karp, & Herzog, 1985). Probably
because of disturbances in ‘Interoceptive awareness’, ED
patients lose or miss the recognition and accurate identi-
fication of appetite signals and emotional cues (Garner
et al., 1983), which may not only contribute to dysregula-
tion of food intake, but also to the inhibition of emotion and
feelings as described above. Restrictive eating suppresses
emotions (primary avoidance behaviour), while binge
eating and purging are primarily aimed at weakening
emotions (secondary avoidance), and these eating behav-
iours are an expression of maladaptive emotion regulation
as a transdiagnostic factor across EDs. Given the major role
that emotions play in signalling, motivating, and commu-
nicating, ED patients may lose contact with themselves
and the outside world and try to steer their mood according
to certain (eating) rules. Because of this, they do not learn
to use social, cognitive, and physical strategies that help
them deal with negative events.
Furthermore, looking for the most important re-
lationships between ED symptoms and non‐symptom
variables in this network, the strongest negative associa-
tion was found between resilience, as a protective factor
in mental health, and the core EDI symptom, social
insecurity. Prior studies have shown that resilience fac-
F I G U R E 2 Expected Influence for the transdiagnostic eating tors predict improvements in psychological health and
disorders network. Larger numbers indicate that the item is more social relationship domains of quality of life and a
central to the network
reduction of ED symptoms over time (Calvete et al.,
2018). Additionally, the strongest positive connection
sensations of hunger and satiety (Garner, Olmstead, & occurs between the EDI symptom, Impulse Regulation,
Polivy, 1983). Personal standards, as a measure of and the temperament dimension, Novelty Seeking.
perfectionism, involve the setting of high standards and Several research results indicate that temperamental
goals for oneself (Frost et al., 1990). High standards features could be considered predictors of specific asso-
perfectionism can become maladaptive and evolve as an ciations between emotions and the tendency to eat (Atiye,
important risk and maintenance factor for EDs (Boone, Miettunen, & Raevuori‐Helkamaa, 2015; Rotella et al.,
Claes, & Luyten, 2014; Dimaggio et al., 2018; Duffy 2018; Vervaet, Van Heeringen, & Audenaert, 2004).
et al., 2019; Flett & Hewitt, 2002; Slof‐Op't Landt, Claes, Finally, another interesting result of this study was
& van Furth, 2016). In addition, this attitude serves as a based on measuring the node predictability. The five nodes
predictor of body dissatisfaction, an excessive fear of with the highest and lowest predictability were over-
gaining weight and a preoccupation to be thinner, even in vigilance and inhibition, disconnection and rejection,
healthy controls (Dickie, Wilson, McDowall, & Surgenor, impaired autonomy and performance, anxiety, personal
2012; Rivière & Douilliez, 2017). Hypervigilance or standards, novelty seeking, reward dependence, maturity
overcontrol as a schema mode is associated with negative fear, bulimia and self‐transcendence, respectively, from
10
- VERVAET ET AL.

F I G U R E 3 Degree centrality difference test for the 32‐item eating disorders network. Grey boxes indicate nodes or edges that do not
differ significantly from one‐another and black boxes represent nodes or edges that do differ significantly from one‐another and white
boxes in the centrality plot show the value of node strength

highest to lowest. As described earlier (Haslbeck & mechanisms of behavioural change. The central features of
Fried, 2017), node predictability can provide information the described network are in line with the transdiagnostic
on the clinical relevance of connections (or edges) between factors of Fairburn's model of ED (Fairburn et al., 2003),
nodes. This analysis shows again that overvigilance and that is, perfectionism, low self‐esteem, mood intolerance
inhibition, disconnection and rejection, impaired auton- and interpersonal difficulties. Moreover, the findings
omy, anxiety and perfectionism (personal standards) indicate that ED patients may benefit from treatments
might be clinically relevant constructs (as they are most designed to enhance cognitive flexibility, combined with
predicted by all other nodes in the network) that could be self‐compassion techniques (Pullmer, Coelho, & Zaitsoff,
taken into account when designing an intervention. Yet, 2019; A. F. Wagner & Vitousek, 2019). For example,
because node predictability informs about shared variance compassion focused therapy (Gilbert, 2009) learns patients
between nodes, and does not give any information on the to become milder to themselves instead of their extreme
direction of causation, future research should further perfectionism, which provokes constant negative emotions
investigate these predictions by conducting a study with and thus maintains the ED. Thus, the drive to excel in a
time‐series data to examine the direction of influence of the valued domain and their strength in top‐down regulation
highly central nodes from the current study. If it could be strategies can be a powerful attribute when redirected to
established that these nodes have temporal precedence, serve recovery. The harsh self‐criticism that accompanies
these results could be a first step towards identifying po- their perfectionistic striving can be diminished by
tential intervention targets. The current findings are rele- rewarding efforts by directing them to other more efficient
vant for the treatment of EDs by elucidating potential goals as an adaptive emotional regulation strategy.
VERVAET ET AL.
- 11

Without any doubt, motivational strategies need to precede subjects model, and how well the results will generalize
every intervention in bridging symptoms related to fear of to individual patients will have to be studied empirically,
forming close relationships with those related to feelings of for instance by collecting and analysing time‐series data
adequacy and perceived lack of control. (Epskamp, van Borkulo, et al., 2018).
Despite some important strengths of the study
(including the large, transdiagnostic study sample of
patients from all ED‐categories with various levels of 7 | CONCLUSION
symptom severity, and the measurement of both ED‐
specific symptoms and general psychiatric symptoms, This study aimed to explore the link between core ED
and vulnerability factors), a number of methodological symptoms and comorbid clinical symptoms and trans-
issues need to be addressed. Limitations include the diagnostic vulnerability mechanisms in a large sample of
cross‐sectional nature of the analysed data and the treatment‐seeking patients with an ED. The results
absence of data on potential medical comorbidities, indicate a central role of personal standards (a dimension
cognitive performance and social functioning that may of perfectionism), overvigilance and inhibition (a mal-
play important roles (Blinder, Cumella, & Sanathara, adaptive schemata with excessive focus on inhibiting
2006; Setia, 2016). Another limitation is that no addi- one's emotions and feelings in order to avoid making
tional network models were estimated for each ED mistakes), and ED symptoms (ineffectiveness and inter-
diagnosis separately, as has been done by previous oceptive awareness). These findings may suggest that
network studies (e.g., Solmi et al., 2018). Additionally, a over controlling bodily aspects by dietary restraint
statistical limitation is that the current study included a (through inhibition) and interoceptive awareness as a
patient sample in a hospital setting (in‐ and outpatients) maladaptive regulation of feelings of ineffectiveness are
that can lead to Berkson's bias (selecting a population key characteristics in ED patients. Probably, based on
based on scores of diagnostic criteria such as symptoms) their extreme personal standards with severe and rigid
when using statistical models such as structural equation rules, they continuously try to improve their control ef-
models that are estimated based on the correlation ma- forts instead of changing their method in order to develop
trix of items (de Ron, Fried, & Epskamp, 2019). More- a more adaptive and effective emotion regulation. These
over, current network models are based on pairwise, findings are relevant for the understanding and treatment
linear relationships, which should be seen as a lower of ED by underlining the importance of using motiva-
bound on the true complexity of the modelled system. tional strategies in bridging ED symptoms, adequate
Therefore, they will not successfully recover more methods of self‐improvement and adaptive emotion
complicated relationships between variables such as non‐ regulation.
linear effects, higher‐order interactions or threshold ef-
fects such that A only starts influencing B when A is at a
OR C I D
certain minimum or quadratic correlations among
Myriam Vervaet https://orcid.org/0000-0002-4831-5496
others. Furthermore, network models, such as any sta-
Louise Puttevils https://orcid.org/0000-0003-2162-4042
tistical model, can only explore variance between items
included in the model, and it is not a trivial question
E N DN OT E
which variables should be considered to be part of a 1
As data collection started 20 years ago, an older version of the
complex system. This means that future investigations DSM is used for the diagnosis of eating disorders. As the new
should examine how replicable the centrality results of DSM‐5 criteria are more sensitive to capture individuals within
the present study are to variations of included items. specified diagnoses, as compared to other or unspecified di-
Related to this issue, it is also very important to focus agnoses, we decided to focus our analyses on all patients diag-
more on item selection when creating a network model, nosed with EDs, without making a differentiation per category.
since centrality measures take items into account, but
this implies that constructs that hold more items (and REFERENCES
therefore are represented by more nodes in the network) Allen, K. L., Byrne, S. M., Oddy, W. H., & Crosby, R. D. (2013).
will benefit from intra‐measure correlations compared to DSM‐IV‐TR and DSM‐5 eating disorders in adolescents:
constructs with only a single node. Bridge symptoms Prevalence, stability, and psychosocial correlates in a popula-
tion‐based sample of male and female adolescents. Journal of
would help account for this issue because they only
Abnormal Psychology, 122(3), 720–732. https://doi.org/
include associations between clusters; however, this only 10.1037/a0034004
makes sense with clearly defined constructs and com- American Psychiatric Association. (2013). Diagnostic and statistical
munities in the network, which is not the case in our manual of mental disorders. Arlington, TX: American Psychiatric As-
current network model. Finally, the model is a between‐ sociation. https://doi.org/10.1176/appi.books.9780890425596.744053
12
- VERVAET ET AL.

Atiye, M., Miettunen, J., & Raevuori‐Helkamaa, A. (2015). A meta‐ Cole, P. M., Michel, M. K., & Teti, L. O. (1994). The development of
analysis of temperament in eating disorders. European Eating emotion regulation and dysregulation: A clinical perspective.
Disorders Review, 23(2), 89–99. https://doi.org/10.1002/erv.2342 Monographs of the Society for Research in Child Development,
Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Beck Depression 59, 73–102. https://doi.org/10.1111/j.1540-5834.1994.tb01278.x
Inventory‐Second Edition (BDI‐II). San Antonio, TX: Psycho- Cooper, Z., & Grave, R. D. (2017). Eating disorders: Transdiagnostic
logical Corporation. theory and treatment. In The Science of Cognitive Behavioral
Blinder, B. J., Cumella, E. J., & Sanathara, V. A. (2006). Psy- Therapy, (pp. 337–357). https://doi.org/10.1016/B978-0-12-
chiatric comorbidities of female inpatients with eating dis- 803457-6.00014-3
orders. Psychosomatic Medicine, 68(3), 454–462. https://doi. de Ron, J., Fried, E. I., & Epskamp, S. (2019). Psychological net-
org/10.1097/01.psy.0000221254.77675.f5 works in clinical populations: A tutorial on the consequences
Boone, L., Braet, C., Vandereycken, W., & Claes, L. (2013). Are of Berkson's bias. Psychological Medicine. 1–9. https://doi.org/
maladaptive schema domains and perfectionism related to 10.1017/S0033291719003209
body image concerns in eating disorder patients? European de Vos, J. A., Lamarre, A., Radstaak, M., Bijkerk, C., Bohlmeijer,
Eating Disorders Review, 21(1), 45–51. https://doi.org/10.1002/ E. T., & Westerhof, G. J. (2017). Identifying fundamental
erv.2175 criteria for eating disorder recovery: A systematic review and
Boone, L., Claes, L., & Luyten, P. (2014). Too strict or too loose? qualitative meta‐analysis (accepted proof). Journal of Eating
Perfectionism and impulsivity: The relation with eating disorder Disorders, 5(34), 1–14. https://doi.org/10.1186/s40337-017-
symptoms using a person‐centered approach. Eating Behaviors, 0164-0
15(1), 17–23. https://doi.org/10.1016/j.eatbeh.2013.10.013 Dickie, L., Wilson, M., McDowall, J., & Surgenor, L. J. (2012).
Borsboom, D. (2017). A network theory of mental disorders. World What components of perfectionism predict drive for thin-
Psychiatry, 16(1), 5–13. https://doi.org/10.1002/wps.20375 ness? Eating Disorders, 20(3), 232–247. https://doi.org/
Bos, F. M., Snippe, E., De Vos, S., Hartmann, J. A., Simons, C. J. P., 10.1080/10640266.2012.668484
Van Der Krieke, L., … Wichers, M. (2017). Can we jump from Dimaggio, G., MacBeth, A., Popolo, R., Salvatore, G., Perrini, F.,
cross‐sectional to dynamic interpretations of networks? Impli- Raouna, A., … Montano, A. (2018). The problem of over-
cations for the network perspective in psychiatry. Psychotherapy control: Perfectionism, emotional inhibition, and personality
and Psychosomatics, 86, 175–177. https://doi.org/10.1159/ disorders. Comprehensive Psychiatry, 83, 71–78. https://doi.
000453583 org/10.1016/j.comppsych.2018.03.005
Brytek‐Matera, A., & Czepczor, K. (2017). Models of eating disor- DuBois, R. H., Rodgers, R. F., Franko, D. L., Eddy, K. T., & Thomas,
ders: A theoretical investigation of abnormal eating patterns J. J. (2017). A network analysis investigation of the cognitive‐
and body image disturbance. Archives of Psychiatry and Psy- behavioral theory of eating disorders. Behaviour Research
chotherapy, 1, 16–26. https://doi.org/10.12740/APP/68422 and Therapy, 97, 213–221. https://doi.org/10.1016/j.brat.2017.
Calvete, E., las Hayas, C., & Gómez del Barrio, A. (2018). Longi- 08.004
tudinal associations between resilience and quality of life in Duffy, M. E., Rogers, M. L., Joiner, T. E., Bergen, A. W., Berrettini,
eating disorders. Psychiatry Research, 259(Jan), 470–475. W., Bulik, C. M., … Keel, P. K. (2019). An investigation of in-
https://doi.org/10.1016/j.psychres.2017.11.031 direct effects of personality features on anorexia nervosa
Cascino, G., Castellini, G., Stanghellini, G., Ricca, V., Cassioli, E., severity through interoceptive dysfunction in individuals with
Ruzzi, V., … Monteleone, A. M. (2019). The role of the lifetime anorexia nervosa diagnoses. International Journal of
embodiment disturbance in the anorexia nervosa psychopa- Eating Disorders, 52(2), 200–205. https://doi.org/10.1002/
thology: A network analysis study. Brain Sciences, 9(10), 1–11. eat.23008
https://doi.org/10.3390/brainsci9100276 Eddy, K. T., Dorer, D. J., Franko, D. L., Tahilani, K., Thompson‐
Cassin, S. E., & Von Ranson, K. M. (2005). Personality and eating Brenner, H., & Herzog, D. B. (2008). Diagnostic crossover in
disorders: A decade in review. Clinical Psychology Review, anorexia nervosa and bulimia nervosa: Implications for DSM‐
25(7), 895–916. https://doi.org/10.1016/j.cpr.2005.04.012 V. American Journal of Psychiatry, 165(2), 245–250. https://doi.
Castellini, G., Lo Sauro, C., Mannucci, E., Ravaldi, C., Rotella, org/10.1176/appi.ajp.2007.07060951
C. M., Faravelli, C., & Ricca, V. (2011). Diagnostic crossover Elliott, H., Jones, P. J., & Schmidt, U. (2020). Central symptoms predict
and outcome predictors in eating disorders according to DSM‐ posttreatment outcomes and clinical impairment in anorexia
IV and DSM‐V proposed criteria: A 6‐year follow‐up study. nervosa: A network analysis. Clinical Psychological Science, 8(1),
Psychosomatic Medicine, 73(3), 270–279. https://doi.org/ 139–154. https://doi.org/10.1177/2167702619865958
10.1097/PSY.0b013e31820a1838 Epskamp, S., Borsboom, D., & Fried, E. I. (2018). Estimating psy-
Christian, C., Williams, B. M., Hunt, R. A., Wong, V. Z., Ernst, S. E., chological networks and their accuracy: A tutorial paper.
Spoor, S. P., … Levinson, C. A. (2020). A network investigation Behavior Research Methods, 50(1), 195–212. https://doi.org/
of core symptoms and pathways across duration of illness 10.3758/s13428-017-0862-1
using a comprehensive cognitive‐behavioral model of eating‐ Epskamp, S., & Fried, E. I. (2020). Bootstrap Methods for Various
disorder symptoms. Psychological Medicine, 1–10. Advance Network Estimation Routines. Bootnet. Retrieved from https://
online publication. https://doi.org/10.1017/S cran.r-project.org/web/packages/bootnet/bootnet.pdf
Cloninger, C. R., Svrakic, D. M., & Przybeck, T. R. (1993). A psy- Epskamp, S., Cramer, A. O. J., Waldorp, L. J., Schmittmann, V. D.,
chobiological model of temperament and character. Archives of & Borsboom, D. (2012). Qgraph: Network visualizations of
General Psychiatry, 50(12), 975–990. https://doi.org/10.1001/ relationships in psychometric data. Journal of Statistical Soft-
archpsyc.1993.01820240059008 ware, 48(4), 1–18. https://doi.org/10.18637/jss.v048.i04
VERVAET ET AL.
- 13

Epskamp, S., & Fried, E. I. (2018). A tutorial on regularized partial Gilbert, P. (2009). Introducing compassion‐focused therapy. Ad-
correlation networks. Psychological Methods, 23(4), 617–634. vances in Psychiatric Treatment, 5(3), 199–208. https://doi.org/
https://doi.org/10.1037/met0000167 10.1192/apt.bp.107.005264
Epskamp, S., van Borkulo, C. D., van der Veen, D. C., Servaas, Goldschmidt, A. B., Crosby, R. D., Cao, L., Moessner, M., Forbush,
M. N., Isvoranu, A. M., Riese, H., & Cramer, A. O. J. (2018). K. T., Accurso, E. C., & Grange, D. Le (2018). Network analysis
Personalized network modeling in psychopathology: The of pediatric eating disorder symptoms in a treatment‐seeking,
importance of contemporaneous and temporal connections. transdiagnostic sample. Journal of Abnormal Psychology,
Clinical Psychological Science, 6(3), 416–427. https://doi.org/ 127(2), 251–264. https://doi.org/10.1037/abn0000327
10.1177/2167702617744325 Gongora, V. C., Derksen, J. J. L., & Van Der Staak, C. P. F.
Fairburn, C. G., Cooper, Z., & Shafran, R. (2003). Cognitive (2004). The role of core beliefs in the specific cognitions of
behaviour therapy for eating disorders: A “transdiagnostic” bulimic patients. Journal of Nervous and Mental Disease,
theory and treatment. Behaviour Research and Therapy, 41(5), 192(4), 297–303. https://doi.org/10.1097/01.nmd.0000120889.
509–528. https://doi.org/10.1016/S0005-7967(02)00088-8 01611.2f
Farstad, S. M., McGeown, L. M., & von Ranson, K. M. (2016). Harvey, A., Watkins, E., Mansell, W., & Shafran, R. (2004). Cognitive
Eating disorders and personality, 2004‐2016: A systematic re- behavioural processes across psychological disorders: A trans-
view and meta‐analysis. Clinical Psychology Review, 46, 91–105. diagnostic approach to research and treatment. New York, NY:
https://doi.org/10.1016/j.cpr.2016.04.005 Oxford University Press. Retrieved from https://www.
Flett, G. L., & Hewitt, P. L. (2002). Perfectionism: Theory, research, oxfordclinicalpsych.com/view/10.1093/med:psych/9780198528
and treatment. American Journal of Psychiatry, 161, 319–340. 883.001.0001/med-9780198528883
https://doi.org/10.1037/10458-000 Haslbeck, J. M. B., & Fried, E. I. (2017). How predictable are
Forbes, M. K., Wright, A. G. C., Markon, K. E., & Krueger, R. F. symptoms in psychopathological networks? A reanalysis of 18
(2019). The network approach to psychopathology: Promise published datasets. Psychological Medicine, 47(16), 2267–2276.
versus reality. World Psychiatry, 18(3), 272–273. https://doi. https://doi.org/10.1017/S0033291717001258
org/10.1002/wps.20659 Haslbeck, J. M. B., & Waldorp, L. J. (2018). How well do network
Forbush, K. T., Siew, C. S. Q., & Vitevitch, M. S. (2016). Application models predict observations? On the importance of predict-
of network analysis to identify interactive systems of eating ability in network models. Behavior Research Methods, 50,
disorder psychopathology. Psychological Medicine, 46(12), 853–861. https://doi.org/10.3758/s13428-017-0910-x
2667–2677. https://doi.org/10.1017/S003329171600012X Jones, C., Harris, G., & Leung, N. (2005). Core beliefs and eating
Forrest, L. N., Jones, P. J., Ortiz, S. N., & Smith, A. R. (2018). Core disorder recovery. European Eating Disorders Review, 13(4),
psychopathology in anorexia nervosa and bulimia nervosa: 237–244. https://doi.org/10.1002/erv.642
A network analysis. International Journal of Eating Disorders, Jones, P. J., Heeren, A., & McNally, R. J. (2017). Commentary:
51(7), 668–679. https://doi.org/10.1002/eat.22871 A network theory of mental disorders. Frontiers in Psychology,
Fried, E. I. (2020). Lack of theory building and testing impedes 16, 5–13. https://doi.org/10.3389/fpsyg.2017.01305
progress in the factor and network literature. PsyArXiv Pre- Kalisch, R., Müller, M. B., & Tüscher, O. (2015). A conceptual
prints, 1–50. https://doi.org/10.31234/osf.io/zg84s framework for the neurobiological study of resilience. The
Fried, E. I., & Cramer, A. O. J. (2017). Moving forward: Challenges Behavioral and Brain Sciences, 38(e92). https://doi.org/
and directions for psychopathological network theory and 10.1017/S0140525X1400082X
methodology. Perspectives on Psychological Science, 1–22. Leung, N., Waller, G., & Thomas, G. (1999). Core beliefs in anorexic
https://doi.org/10.1177/1745691617705892 and bulimic women. Journal of Nervous and Mental
Fritz, J., Fried, E. I., Goodyer, I. M., Wilkinson, P. O., & van Disease, 187(12), 736–741. https://doi.org/10.1097/00005053-
Harmelen, A. L. (2018). A network model of resilience factors 199912000-00005
for adolescents with and without exposure to childhood Levinson, C. A., Brosof, L. C., Ma, J., Fewell, L., & Lenze, E. J.
adversity. Scientific Reports, 8(15774), 1–13. https://doi.org/ (2017). Fear of food prospectively predicts drive for thinness in
10.1038/s41598-018-34130-2 an eating disorder sample recently discharged from intensive
Frost, R. O., Marten, P., Lahart, C., & Rosenblate, R. (1990). The treatment. Eating Behaviors, 27(November), 45–51. https://doi.
dimensions of perfectionism. Cognitive Therapy and Research. org/10.1016/j.eatbeh.2017.11.004
14(5), 449–468. https://doi.org/10.1007/BF01172967 Levinson, C. A., Vanzhula, I. A., Brosof, L. C., & Forbush, K. (2018).
Fruchterman, T. M. J., & Reingold, E. M. (1991). Graph drawing by Network analysis as an alternative approach to conceptual-
force‐directed placement. Software: Practice and Experience. izing eating disorders: Implications for research and treat-
https://doi.org/10.1002/spe.4380211102 ment. Current Psychiatry Reports, 20(67). https://doi.org/
Garner, D. M. (1991). Eating disorder inventory 2: Professional 10.1007/s11920-018-0930-y
manual. International Journal of Eating Disorders. Mcgrath, R. J., Julie, W., & Caron, R. M. (2012). The relationship
Retrieved from https://www.google.co.il/_/chrome/newtab? between resilience and body image in college women. The
espv¼2&ie¼UTF-8. Internet Journal of Health, 10(2). https://doi.org/10.5580/b81
Garner, D. M., Olmstead, M. P., & Polivy, J. (1983). Development McLaughlin, E. F., Karp, S. A., & Herzog, D. B. (1985). Sense of
and validation of a multidimensional eating disorder inventory ineffectiveness in women with eating disorders: A clinical
for anorexia nervosa and bulimia. International Journal of study of anorexia nervosa and bulimia. International Journal of
Eating Disorders, 2(2), 15–34. https://doi.org/10.1002/1098- Eating Disorders, 4(4), 511–523. https://doi.org/10.1002/1098-
108X(198321)2:2<15::AID-EAT2260020203>3.0.CO62 108X(198511)4:4<511::AID-EAT2260040411>3.0.CO;2-Z
14
- VERVAET ET AL.

Monteleone, A. M., Cascino, G., Pellegrino, F., Ruzzi, V., Patri- Psychotherapy, 33(3), 369–374. https://doi.org/10.1017/S135
ciello, G., Marone, L., … Maj, M. (2019). The association 2465805002122
between childhood maltreatment and eating disorder Rivière, J., & Douilliez, C. (2017). Perfectionism, rumination, and
psychopathology: A mixed‐model investigation. European gender are related to symptoms of eating disorders: A
Psychiatry, 61, 111–118. https://doi.org/10.1016/j.eurpsy.2019. moderated mediation model. Personality and Individual Dif-
08.002 ferences, 116, 63–68. https://doi.org/10.1016/j.paid.2017.04.041
Monteleone, A. M., Cascino, G., Solmi, M., Pirozzi, R., Tolone, S., Robinaugh, D. J., Millner, A. J., & McNally, R. J. (2016). Identifying
Terracciano, G., … Docimo, L. (2019). A network analysis of psy- highly influential nodes in the complicated grief network.
chological, personality and eating characteristics of people seeking Journal of Abnormal Psychology, 12(6), 747–757. https://doi.
bariatric surgery: Identification of key variables and their prog- org/10.1037/abn0000181
nostic value. Journal of Psychosomatic Research, 120(January), 81– Rodgers, R. F., DuBois, R., Thiebaut, S., Jaussent, I., Maimoun,
89. https://doi.org/10.1016/j.jpsychores.2019.03.010 L., Seneque, M., … Guillaume, S. (2019). Structural differ-
Monteleone, A. M., Mereu, A., Cascino, G., Criscuolo, M., Casti- ences in eating disorder psychopathology after history of
glioni, M. C., Pellegrino, F., … Zanna, V. (2019). Re‐concep- childhood abuse: Insights from a bayesian network analysis.
tualization of anorexia nervosa psychopathology: A network Journal of Abnormal Psychology, 128(8), 795–805. https://doi.
analysis study in adolescents with short duration of the illness. org/10.1037/abn0000470
International Journal of Eating Disorders, 52(11), 1263–1273. Rotella, F., Mannucci, E., Gemignani, S., Lazzeretti, L., Fioravanti,
https://doi.org/10.1002/eat.23137 G., & Ricca, V. (2018). Emotional eating and temperamental
Olatunji, B. O., Levinson, C., & Calebs, B. (2018). A network traits in eating disorders: A dimensional approach. Psychiatry
analysis of eating disorder symptoms and characteristics in an Research, 264(June), 1–8. https://doi.org/10.1016/j.psychres.
inpatient sample. Psychiatry Research, 262(July), 270–281. 2018.03.066
https://doi.org/10.1016/j.psychres.2018.02.027 Santos, H., Jr, Fried, E. I., Asafu‐Adjei, J., & Ruiz, R. J. (2017).
Oldershaw, A., Lavender, T., & Schmidt, U. (2018). Are socio‐ Network structure of perinatal depressive symptoms in latinas:
emotional and neurocognitive functioning predictors of ther- Relationship to stress and reproductive biomarkers. Research
apeutic outcomes for adults with anorexia nervosa? European in Nursing & Health, 40(3), 218–228. https://doi.org/10.1002/
Eating Disorders Review, 26(4), 346–359. https://doi.org/ nur.21784
10.1002/erv.2602 Setia, M. S. (2016). Methodology series module 3: Cross‐sectional
Oldershaw, A., Startup, H., & Lavender, T. (2019). Anorexia nervosa studies. Indian Journal of Dermatology, 61(3), 261–264. https://
and a lost emotional self: A psychological formulation of the doi.org/10.4103/0019-5154.182410
development, maintenance, and treatment of anorexia nerv- Shafran, R., Cooper, Z., & Fairburn, C. G. (2002). Clinical perfec-
osa. Frontiers in Psychology, 10(March), 1–22. https://doi.org/ tionism: A cognitive‐behavioural analysis. Behaviour Research
10.3389/fpsyg.2019.00219 and Therapy, 40(7), 773–791. https://doi.org/10.1016/S0005-
Pauwels, E., Dierckx, E., Schoevaerts, K., Santens, E., Peuskens, H., 7967(01)00059-6
& Claes, L. (2018). Early maladaptive schemas: Similarities Slof‐Op't Landt, M. C. T., Claes, L., & van Furth, E. F. (2016).
and differences between female patients with eating versus Classifying eating disorders based on “healthy” and “un-
substance use disorders. European Eating Disorders Review, healthy” perfectionism and impulsivity. International Journal
26(5), 422–430. https://doi.org/10.1002/erv.2610 of Eating Disorders, 49(7), 673–680. https://doi.org/10.1002/
Portzky, M., Wagnild, G., De Bacquer, D., & Audenaert, K. (2010). eat.22557
Psychometric evaluation of the Dutch Resilience Scale RS‐nl Solmi, M., Collantoni, E., Meneguzzo, P., Degortes, D., Tenconi,
on 3265 healthy participants: A confirmation of the association E., & Favaro, A. (2018). Network analysis of specific psy-
between age and resilience found with the Swedish version. chopathology and psychiatric symptoms in patients with
Scandinavian Journal of Caring Sciences, 24(Suppl. 1), 86–92. eating disorders. International Journal of Eating Disorders,
https://doi.org/10.1111/j.1471-6712.2010.00841.x 51(7), 680–692. https://doi.org/10.1002/eat.22884
Pugh, M. (2015). A narrative review of schemas and schema Solmi, M., Collantoni, E., Meneguzzo, P., Tenconi, E., & Favaro, A.
therapy outcomes in the eating disorders. Clinical Psychology (2019). Network analysis of specific psychopathology and psy-
Review, 39(July), 30–41. https://doi.org/10.1016/j.cpr.2015. chiatric symptoms in patients with anorexia nervosa. European
04.003 Eating Disorders Review, 27(1), 24–33. https://doi.org/10.1002/
Pullmer, R., Coelho, J. S., & Zaitsoff, S. L. (2019). Kindness begins erv.2633
with yourself: The role of self‐compassion in adolescent body Spielberger, C. D. (1983). State‐Trait Anxiety Inventory for Adults
satisfaction and eating pathology. International Journal of (STAI‐AD) [Database record]. APA Psyc Tests.
Eating Disorders, 52(7), 809–816. https://doi.org/10.1002/ Talbot, D., Smith, E., Tomkins, A., Brockman, R., & Simpson, S.
eat.23081 (2015). Schema modes in eating disorders compared to a
Puttevils, L., Vanderhasselt, M. A., & Vervaet, M. (2019). Investi- community sample. Journal of Eating Disorders, 3(41). https://
gating transdiagnostic factors in eating disorders: Does self‐ doi.org/10.1186/s40337-015-0082-y
esteem moderate the relationship between perfectionism and Ten Ham, B. V. H., Hulsbergen, M., & Bohlmeyer, E. (2016).
eating disorder symptoms? European Eating Disorders Review, Transdiagnostische factoren. Theorie & praktijk. Tijdschrift
27(4), 381–390. https://doi.org/10.1002/erv.2666 Voor Psychiatrie.
Riley, C., & Shafran, R. (2005). Clinical perfectionism: A pre- Tibshirani, R. (1996). Regression shrinkage and selection via the
liminary qualitative analysis. Behavioural and Cognitive lasso. Journal of the Royal Statistical Society: Series B
VERVAET ET AL.
- 15

(Methodological), 58(1), 267–288. https://doi.org/10.1111/ 6(4), 495–505. https://doi.org/10.1002/1098-108X(198707)

j.2517-6161.1996.tb02080.x 6:4<495::AID-EAT2260060406>3.0.CO;2-I.
Unoka, Z., Tölgyes, T., & Czobor, P. (2007). Early maladaptive Wagnild, G. (2009). A review of the resilience Scale. Journal of Nursing
schemas and body mass index in subgroups of eating disorders: Measurement, 17(2). https://doi.org/10.1891/1061-3749.17.2.105
A differential association. Comprehensive Psychiatry, 48(2), 199– Waller, G., Ohanian, V., Meyer, C., & Osman, S. (2000). Cognitive
204. https://doi.org/10.1016/j.comppsych.2006.09.002 content among bulimic women: The role of core beliefs. In-
Van Borkulo, C., Boschloo, L., Borsboom, D., Penninx, B. W. J. H., ternational Journal of Eating Disorders, 28(2), 235–241. https://
Lourens, J. W., & Schoevers, R. A. (2015). Association of doi.org/10.1002/1098-108X(200 009)28:2<235::AID-EAT15>
symptom network structure with the course of longitudinal 3.0.CO;2-1
depression. JAMA Psychiatry, 72(12), 1219–1226. https://doi. Wang, S. B., Jones, P. J., Dreier, M., Elliott, H., & Grilo, C. M. (2019).
org/10.1001/jamapsychiatry.2015.2079 Core psychopathology of treatment‐seeking patients with binge‐
Van der Ploeg, H. M. (1980). Validity of the Zelf‐Beoordelings‐ eating disorder: A network analysis investigation. Psychological
Vragenlijst (A Dutch version of the Spielberger State‐Trait Medicine, 49(11), 1923–1928. https://doi.org/10.1017/S003329
Anxiety Inventory). Nederlands Tijdschrift Voor de Psychologie 1718002702
En Haar Grensgebieden. Young, J. E., Klosko, J. S., & Weishaar, M. E. (2003). Schema therapy:
Vervaet, M., Van Heeringen, C., & Audenaert, K. (2004). Per- A practitioners's guide. New York, NY: The Guilford Press.
sonality‐related characteristics in restricting versus binging
and purging eating disordered patients. Comprehensive Psy-
S U P P O R T I NG I NF O R M A T I O N
chiatry, 45(1), 37–43. https://doi.org/10.1016/j.comppsych.
2003.09.008 Additional supporting information may be found online
Wade, T. D., Wilksch, S. M., Paxton, S. J., Byrne, S. M., & Austin, in the Supporting Information section at the end of this
S. B. (2015). How perfectionism and ineffectiveness influence article.
growth of eating disorder risk in young adolescent girls.
Behaviour Research and Therapy, 66(March), 56–63. https://
doi.org/10.1016/j.brat.2015.01.007 How to cite this article: Vervaet M, Puttevils L,
Wagner, A. F., & Vitousek, K. M. (2019). Personality variables and
Hoekstra RHA, Fried E, Vanderhasselt MA.
eating pathology. The Psychiatric Clinics of North America,
42(1), 105–119. https://doi.org/10.1016/j.psc.2018.10.012
Transdiagnostic vulnerability factors in eating
Wagner, S., Halmi, K. A., & Maguire, T. V. (1987). The sense of disorders: A network analysis. Eur Eat Disorders
personal ineffectiveness in patients with eating disorders: One Rev. 2020;1–15. https://doi.org/10.1002/erv.2805
construct or several? International Journal of Eating Disorders,