Received: 20 December 2019
Revised: 17 March 2020
Accepted: 17 March 2020
DOI: 10.1002/eat.23270
ORIGINAL ARTICLE
Eating disorder psychopathology in adults and adolescents
with anorexia nervosa: A network approach
Simona Calugi PhD
| Massimiliano Sartirana PsyD | Arianna Misconel PsyD
Camilla Boglioli PsyD | Riccardo Dalle Grave MD
Department of Eating and Weight Disorders,
Villa Garda Hospital, Garda, Italy
Correspondence
Simona Calugi, Department of Eating and
Weight Disorders, Villa Garda Hospital, via
Monte Baldo, Garda 89 I-37016, Italy.
Email: si.calugi@gmail.com
Action Editor: Ruth Weissman
|
Abstract
Objective: The aim of this study was to assess and compare eating disorder feature
networks in adult and adolescent patients with anorexia nervosa.
Methods: Patients seeking treatment for anorexia nervosa in inpatient and outpatient settings were consecutively recruited from January 2008 to September 2019.
Body mass index was measured, and each patient completed the Eating Disorder
Examination Questionnaire.
Results: The sample comprised 547 adolescent and 724 adult patients with anorexia
nervosa. Network analysis showed that in both adults and adolescents, the most
central and highly interconnected nodes in the network were related to shape overvaluation and desiring weight loss. The network comparison test identified similar
global strength and network invariance, confirming the similarity of the two network
structures.
Discussion: The network structures in adult and adolescent patients with anorexia
nervosa are similar, and lend weight to the cognitive behavioral theory that overvaluation of shape and weight is the core feature of anorexia nervosa psychopathology.
KEYWORDS
adolescent, adult, desiring weight loss, eating disorder psychopathology, feeling fat, shape
overvaluation
1
|
I N T RO DU CT I O N
Indeed, a study comparing the clinical presentation of adults and
adolescents with anorexia nervosa found that adolescents had a
Available data indicate that treatment outcomes in anorexia nervosa
shorter period of, but more rapid, weight loss, indicative of an illness
are generally better in adolescents than in adults (Ackard, Richter,
of shorter duration, but also less severe eating disorder psychopathol-
Egan, & Cronemeyer, 2014; Ambwani et al., in press; Calugi, Dalle
ogy severity; this led to the suggestion that these two characteristics
Grave, Sartirana, & Fairburn, 2015; Dalle Grave, Calugi, Doll, &
could explain why they are more amenable to treatment than adults
Fairburn, 2013; Fairburn, 2005). It has been hypothesized that these
(Fisher, Schneider, Burns, Symons, & Mandel, 2001). Nonetheless,
differences in treatment outcome could be due to the clinical impact
additional research examining the differences in eating disorder psy-
of neuroprogression (Treasure, Stein, & Maguire, 2015) and habit for-
chopathology between adults and adolescents with anorexia nervosa
mation (Walsh, 2013) or to the fact that adolescent patients have
is needed to clarify the relationship between specific eating disorder
more structured social networks and a briefer duration of illness than
psychopathology features in these two groups of patients.
adults (Calugi et al., 2015). However, it is also possible that some psy-
Eating disorder feature associations and relative contribution to
chopathological differences between adults and adolescents with
psychopathology have traditionally been modeled using empirical
anorexia nervosa could explain these findings.
approaches (e.g., latent class/profile analysis, factor analysis, structural
Int J Eat Disord. 2020;1–12.
wileyonlinelibrary.com/journal/eat
© 2020 Wiley Periodicals, Inc.
1
2
CALUGI ET AL.
equation modeling, and mixture models). However, these approaches
(Forrest, Jones, Ortiz, & Smith, 2018; Olatunji, Levinson, & Calebs, 2018;
assume the presence of latent categories and/or dimensions that give
Solmi et al., 2018; Solmi, Collantoni, Meneguzzo, Tenconi, &
rise to observable symptoms, which is largely reflective of a historical
Favaro, 2019) or adolescents alone (Goldschmidt et al., 2018;
tendency to apply a disease model to psychopathology (McNally, 2016).
Monteleone et al., 2019). Furthermore, different measures were used to
As Borsboom (2017) and McNally (2016) observed, latent variable
assess their eating disorder psychopathology, making it more difficult to
models of psychopathology are not without limitations. A new statistical
compare the results. That being said, summarizing the available findings,
approach to the study of psychopathology, network analysis, has been
studies combining both adults and adolescents with anorexia nervosa in
applied to address the limitations of prior latent variable model
a single sample found that desiring weight loss, dietary restraint, shape
approaches. Indeed, network theory suggests that psychiatric disorders,
and weight preoccupation, shape overvaluation, and fearing weight gain,
including eating disorders (Levinson, Vanzhula, Brosof, & Forbush, 2018;
as measured using the Eating Disorder Examination Questionnaire (EDE-
Smith et al., 2018), arise from a complex array of causal and reciprocal
Q), had the highest expected influence and strength (Forrest et al., 2018),
relationships among symptoms, rather than directly from latent diagno-
while symptoms such as interoceptive awareness, ineffectiveness,
ses (Borsboom, 2017; Costantini et al., 2015). Network analysis is a
depression, and anxiety were found to be central nodes in other net-
statistical method of enabling graphical and quantitative modeling
works (Olatunji et al., 2018; Solmi et al., 2018; Solmi et al., 2019). Studies
of associations between constructs to identify both specific relation-
including only adolescent patients with anorexia nervosa found feelings
ships between clinical features and central symptoms (symptoms
of fatness, dietary restraint, fear of losing control over eating, discomfort
that are highly connected with other symptoms in the network)
seeing one's body, and dissatisfaction with shape/weight, but also
(Borsboom, 2017). A recent procedural update also enables comparison
depression, personal alienation, asceticism, post-traumatic stress symp-
of network structures pertaining to different subpopulations; specific
toms, drive for thinness, low self-esteem, and physical symptoms as the
tests have now been developed to examine whether a network struc-
network nodes with the highest strength. Nevertheless, to our knowl-
ture is identical across subpopulations, whether specific correlations dif-
edge, no study has yet performed a network analysis to examine and
fer in strength between subpopulations, and whether the overall
compare eating disorder features in adults and adolescents with anorexia
connectivity is equal across subgroups (van Borkulo et al., 2015).
nervosa.
The studies using network analysis in patients with anorexia nervosa
In order to evaluate the differences in eating disorder psychopa-
published to date have included either adults and adolescents together
thology between adults and adolescents with anorexia nervosa, we
TABLE 1
Baseline characteristics of adolescent and adult patients with anorexia nervosa
Gender, female, n (%)
Adolescents (n = 547)
Adults (n = 724)
522 (96.3%)
686 (95.9%)
t-Test or chi-square test
p-Value
Effect sizea
0.11
.740
.009
34.48
<.001
.70
Other demographic and clinical characteristics, mean (SD) [range]
Age (years)
16.3 (1.9) [12–19]
29.7 (8.9) [20–61]
Body mass index (kg/m )
15.3 (1.8) [9.4–18.5]
15.2 (2.1) [8.1–18.5]
0.95
.340
.03
Age of onset, years
14.4 (1.6) [10–19]
18.2 (4.8) [12–46]
5.35
<.001
.15
2.2 (1.4) [0–5]
8.4 (6.2) [0–26]
4.59
<.001
.13
2
Duration of eating disorder (years)
Eating Disorder Examination Questionnaire, mean (SD)
Global score
3.4 (1.6)
3.4 (1.5)
0.07
.946
.002
Dietary restraint
3.5 (1.9)
3.4 (2.0)
1.31
.191
.04
Eating concern
2.8 (1.5)
3.1 (1.6)
2.59
.010
.07
Weight concern
3.4 (1.8)
3.4 (1.6)
0.24
.810
.01
Shape concern
3.9 (1.7)
3.8 (1.6)
0.92
.356
.03
Eating Disorder Examination Questionnaire, n (%) if presentb
Objective binge-eating episodes
209 (39.4%)
343 (48.7%)
10.55
<.001
.09
Self-induced vomiting
101 (18.6%)
246 (34.5%)
38.79
<.001
.17
Laxative misuse
62 (11.4%)
155 (21.9%)
23.27
<.001
.13
287 (53.2%)
309 (43.4%)
11.93
.001
.10
Inpatient
294 (53.7%)
528 (72.9%)
50.17
<.001
.20
Outpatient
253 (46.3%)
196 (27.1%)
Excessive exercise
Treatment setting, n (%)
Note: Data are presented as mean (SD) [range] or as n (%).
a
Phi or Cohen's d as appropriate (a value of .1 is considered as a small effect, .3 as a medium effect, and .5 as a large effect).
b
Number and percentage of patients who presented the eating disorder behavior.
3
CALUGI ET AL.
set out to apply network analysis to evaluate the complex interconnec-
diagnostic criteria for anorexia nervosa, as judged by both the refer-
tions in eating disorder features in both. Our hypothesis, in accordance
ring clinician and an eating disorder specialist (R. D. G.). Exclusion
with the cognitive behavioral theory of eating disorders (Dalle Grave &
criteria were the presence of an acute psychotic state (n = 3) and sig-
Calugi, 2020; Fairburn, Cooper, & Shafran, 2003), was that the core
nificant substance abuse (n = 7).
The Italian National Health System National ethics guidelines dic-
symptoms of eating disorder psychopathology in anorexia nervosa
would have similar network connections in adults and adolescents.
tate that this study should not be considered research per se, as all
treatment and assessment procedures were performed as part of existing clinical practice; being considered a routine service assessment;
2
METHODS
|
therefore, no ethical clearance was required. Nevertheless, informed
written consent for clinical data collection and its processing in an
2.1
|
Participants
anonymous, service-level research setting was obtained from each
patient and their parent(s) and/or legal guardian(s).
The study sample comprised 1,271 patients with anorexia nervosa
aged between 12 and 61 years. Patients were recruited between
January 2008 and September 2019 from the inpatient Eating Disorder
2.2
|
Measures
Unit at Villa Garda Hospital (northern Italy) (n = 822, 64.7%), and from
an outpatient Eating Disorder Service based in Verona (Italy) (n = 449,
The patients' eating disorder features were assessed using the EDE-Q,
35.3%). Patients were included in the study if they met the DSM-5
Italian version (Calugi et al., 2017). This questionnaire assesses the
F I G U R E 1 The network structure estimated from the graphical Least Absolute Shrinkage and Selection Operator in combination with
extended Bayesian information criterion model selection in adult patients with anorexia nervosa. Green lines represent positive correlations, and
red lines represent negative correlations. Thicker edges represent stronger correlations. Eating Disorder Examination Questionnaire:
restraint = dietary restraint; restrict = restrictive eating; exclude_food = excluding food; diet_rules = dietary rules; empty&flat_stomach = desire to
have an empty stomach and desire to have a flat stomach; preocc_food = food preoccupation; preocc_sw = shape or weight preoccupation;
fear_loc = fear of losing control over eating; fear_wtgain = fearing weight gain; desire_wtloss&feel_fat = desiring weight loss and feelings of
fatness; binge_eating = binge eating; vomit = vomiting to control shape or weight; laxatives = using laxatives to control shape or weight;
over_exercise = over-exercising to control shape or weight; eat_secret = eating in secret; guilt_eat = guilt about eating; social_eat = avoiding
social eating; weight_overval = weight overvaluation; shape_overval = shape overvaluation; rx_weighing = reaction to prescribed weighing;
diss_weight = dissatisfaction with weight; discomf_body = discomfort seeing body; avoid_exp&diss_shape = avoiding body exposure and
dissatisfaction with shape [Color figure can be viewed at wileyonlinelibrary.com]
4
CALUGI ET AL.
expression of eating disorder-related behaviors and psychopathology
participants (Calugi et al., 2017). The internal consistency reliability of
over the preceding 28 days; normative data are available for adults
the current sample, estimated using Cronbach's alpha for global
(Brewin, Baggott, Dugard, & Arcelus, 2014) and adolescents of 12 years
EDE-Q, was excellent (α = .94).
and over (Carter, Stewart, & Fairburn, 2001). Twenty-two out of the
In patients to be treated in an inpatient setting, the EDE-Q was
28 EDE-Q items are rated on a Likert scale from 0 to 6, on which higher
administered on the second day of hospitalization, while in outpatients,
scores reflect greater symptom severity. An additional six items assess
it was administered just before the first CBT-E session (Session #0).
eating disorder behaviors, in particular binge-eating episodes (items
13 and 14), binge-eating days (item 15), self-induced vomiting (item
16), laxative misuse (item 17), and excessive exercising (item 18). For
2.3
|
Statistical analysis
the purposes of this study, the 22 items and four eating disorder behaviors were included in the network analysis; EDE-Q item 15 was not
Data management and descriptive analyses were conducted using
included because it combines the information included in items 13 and
SPSS, version 26, and network analysis was performed using R, ver-
14. Items 13 and 14 formed a single measure of binge-eating episodes.
sion 3.5.2 (Team, 2018). Variables are presented as means and SDs, or
The Italian version of the EDE-Q has demonstrated very good
frequencies and percentages, as appropriate. As the Shapiro–Wilk test
internal consistency in terms of the global score (Cronbach's α = .94),
revealed that the study variables were not normally distributed, we
as well as high test–retest reliability (r = 0.80) and excellent criterion
calculated nonparametric correlations using nonparanormal transfor-
validity in eating disorder patients, as compared to healthy control
mation (Zhao, Liu, Roeder, Lafferty, & Wasserman, 2012).
F I G U R E 2 Centrality plots for adaptive Least Absolute Shrinkage and Selection Operator Network in combination with extended Bayesian
information criterion model selection in adult patients with anorexia nervosa. Strength and expected influence values of each node are presented.
Centrality indices are shown as standardized z-scores. Abbreviations for each node are provided in Figure 1
5
CALUGI ET AL.
Considering that, for both adults and adolescents, the frequency of
2.3.1
|
Network estimation
binge-eating episodes (EDE-Q items 13 + 14) had the highest percentage of missing data (3.0 and 2.7%, respectively), to manage missing data
Network analysis was carried out using qgraph Rpackage (Epskamp,
we adopted the assumption of “missing-at-random” (MAR) (e.g., cases
Cramer, Waldorp, Schmittmann, & Borsboom, 2012). Regularized par-
are missing as a function of measured characteristics). To assume MAR,
tial correlation networks (Epskamp, Borsboom, & Fried, 2018) were
we performed logistic regression analysis including “missingness” (0/1) as
estimated using the extended Bayesian information criterion (EBIC)
a dependent variable and baseline characteristics (age and BMI) as pre-
graphical LASSO (Epskamp & Fried, 2018), with polychoric correla-
dictors; we found that participants with one or more missing data points
tions as input when data points were ordinal. Regularization enables a
did not differ significantly from those for whom complete data were
statistical model to be estimated with an extra penalty for model com-
available (all ps > .005).
plexity. The resulting model is conservative, in that small or unstable
Missing data in each group were handled by means of a multiple
correlations are estimated to be zero; in this way, the network edges
imputation procedure using the fully conditional specification method;
(the connections between nodes) that are less likely to be genuine are
this is an iterative Markov chain Monte Carlo method that can be
removed. In addition, networks generated in this way are easier to
used when the pattern of missing data is arbitrary (Takahashi, 2017).
interpret (Epskamp, Kruis, & Marsman, 2017).
Thirty multiple imputed data sets were created; because pooled
The fit of the resulting networks were optimized by minimizing
results were unavailable, imputed data were aggregated to estimate
the EBIC (Chen & Chen, 2008). This has been shown to work particu-
networks. We checked the accuracy of the imputation procedures by
larly well in revealing the true network structure (Foygel &
randomly selecting 5 of the 30 imputed data sets and re-calculating
Drton, 2010; Foyger Barber & Drton, 2015), especially when the gen-
results for each (Azur, Stuart, Frangakis, & Leaf, 2011). Results were
erating network is sparse (i.e., contains few edges), and thereby
very similar across all individual imputation data sets.
enabled us to select the best network.
F I G U R E 3 Bootstrapped strength difference results in adult patients with anorexia nervosa. Symptoms are presented in descending order of
strength values. Grey boxes indicate nodes that do not differ significantly from one another, and black boxes represent nodes that do differ
significantly from one another. White boxes on the diagonal plot show the node strength
6
CALUGI ET AL.
The goldbricker function (Rpackage networktools) was used to
means of quantifying the importance of each node in the network,
identify nodes measuring the same underlying construct. We set the
strength is considered a more stable centrality index than between-
threshold to 0.25, and all pairs of nodes which fell below this thresh-
ness and closeness, (Epskamp & Fried, 2018). Since interpretation of
old were combined using the net_reduce function.
betweenness and closeness in networks is somewhat unclear (Forbes,
Wright, Markon, & Krueger, 2017), we calculated the node strength
and expected influence. The expected influence is similar to strength
2.3.2
|
Centrality indices
in that it identifies the strength of the relationships a given node has
with other nodes. However, unlike strength, expected influence
Computation of the centrality indices of a network structure is the
accounts for the direction of associations. If a node has only positive
typical method of assessing the importance of each of its nodes
associations with other nodes, the node's strength and expected influ-
(Costantini et al., 2015; Epskamp et al., 2018). Three such indices are
ence will be equal. If, on the other hand, a node has negative and posi-
node strength, closeness, and betweenness; node strength quantifies
tive associations with other nodes, the node's expected influence will
how strongly a node is directly connected to other nodes, closeness
be less than its strength. The core nodes in the network are consid-
quantifies how well a node is indirectly connected to other nodes, and
ered those with the highest influence or centrality. These indices are
betweenness quantifies how important a node is in the average path
normalized (mean = 0 and SD = 1), so that an index with a value >1
between two other nodes (Epskamp et al., 2018). However, as a
indicates that it is >1 SD from the mean.
F I G U R E 4 The network structure estimated from the graphical Least Absolute Shrinkage and Selection Operator in combination with
extended Bayesian information criterion model selection in adolescent patients with anorexia nervosa. Green lines represent positive correlations,
and red lines represent negative correlations. Thicker edges represent stronger correlations. Eating Disorder Examination Questionnaire:
exc_food&restraint = excluding food and dietary restraint; restrict = restrictive eating; diet_rules = dietary rules; empty&flat_stomach = desire to
have an empty stomach and desire to have a flat stomach; preocc_food = food preoccupation; fear_loc = fear of losing control over eating;
fear_wtgain = fearing weight gain; feel_fat = feelings of fatness; desire_wtloss = desiring weight loss; binge_eating = binge eating;
vomit = vomiting to control shape or weight; laxatives = using laxatives to control shape or weight; over_exercise = over-exercising to control
shape or weight; eat_secret = eating in secret; guilt_eat = guilt about eating; social_eat = avoiding social eating; wt_over&preocc_sw = weight
overvaluation and shape or weight preoccupation; shape_overval = shape overvaluation; rx_weighing = reaction to prescribed weighing;
diss_weight = dissatisfaction with weight; disc_body&diss_shape = discomfort seeing body and dissatisfaction with shape;
avoid_exposure = avoiding body exposure [Color figure can be viewed at wileyonlinelibrary.com]
7
CALUGI ET AL.
2.3.3
|
Differences in strength
proposed using subset bootstraps and calculating the correlation stability (CS) coefficient to quantify the stability of centrality indices; the
We used the Rpackage bootnet (Epskamp et al., 2018) to assess sig-
CS coefficient represents the maximum proportion of cases that can
nificant differences in symptom strength. Specifically, bootnet calcu-
be dropped so that the correlation between the original centrality
lated 1,000 nonparametric bootstraps of the differences between
indices and the centrality of networks based on subsets has a 95%
the strength values for all symptoms, generating a bootstrapped
probability of being 0.7 or higher. The CS coefficient needs to be
strength difference confidence interval. If the bootstrapped strength
above 0.25, preferably greater than 0.5, in order to enable differences
difference confidence interval does not span 0, this indicates a sig-
in centrality to be interpreted (Epskamp et al., 2018).
nificant difference in strength. Symptoms with significantly greater
strength have greater importance in the network structure
(Epskamp & Fried, 2018).
2.3.5
|
Differences between networks
Differences in network structure and global strength between and
2.3.4
|
Stability
within participants were tested for significance using the M-test and
the S-test included in the R-package network comparison test (NCT).
The Rpackage bootnet was also used to investigate the stability of
NCT uses permutation testing to compare network structures from
centrality indices of portions of the data. Epskamp et al. (2018)
two independent, cross-sectional data sets on invariance of
F I G U R E 5 Centrality plots for adaptive Least Absolute Shrinkage and Selection Operator Network in combination with extended Bayesian
information criterion model selection in adolescent patients with anorexia nervosa. Strength and expected influence values of each node are
presented. Centrality indices are shown as standardized z-scores. Abbreviations for each node are provided in Figure 4
8
CALUGI ET AL.
(a) network structure, (b) edge (connection) strength, and (c) global
global EDE-Q or EDE-Q dietary restraint, weight concern, or shape
strength (Van Borkulo et al., 2017).
concern subscale scores.
3
3.2 | Network structure in adult and adolescent
patients with anorexia nervosa
3.1
RESULTS
|
|
Patient characteristics
3.2.1
|
Adults
Among the 1,271 participants with anorexia nervosa, 547 (43%) were
adolescents (i.e., aged 12–19 years), as per the World Health Organi-
The goldbricker function found three pairs of variables to be combined
zation criteria (World Health Organization, 2015) and 724 (57%) were
in adult network: (a) “desire to have a flat stomach” and “desire to
adults (>19 years). Table 1 shows the demographic and clinical fea-
have an empty stomach”; (b) “desiring weight loss” and “feeling of fat-
tures of both groups. As compared to adults, adolescents displayed a
ness”; and (c) “avoiding body exposure” and “dissatisfaction with
significantly lower mean score for eating concern, and fewer episodes
shape.” Each pair was therefore combined into a single variable using
of binge eating, self-induced vomiting, and laxative misuse; they also
the net_reduce function. Thus, the final network for adult patients
showed a greater tendency to exercise excessively. A lower percent-
comprised 23 variables.
age of adolescents than adults were treated in an inpatient setting,
Figures 1 and 2 show the network structure, expected influence,
but no differences were found in terms of gender distribution, BMI,
and node strength values in adults with anorexia nervosa. Nodes
F I G U R E 6 Bootstrapped strength difference results in adolescent patients with anorexia nervosa. Symptoms are presented in descending
order of strength values. Grey boxes indicate nodes that do not differ significantly from one another, and black boxes represent nodes that do
differ significantly from one another. White boxes on the diagonal plot show the node strength
9
CALUGI ET AL.
presenting higher strength coefficients were: “dietary restraint”
3.4
|
Network comparisons
(S = 1.30), the combined variable “desiring weight loss”/“feeling fat”
(S = 1.10), and “shape overvaluation” (S = 1.39). Higher expected influ-
The NCT was used to identify differences between the overall struc-
ence coefficients were found for the following nodes: “dietary
tures of adult and adolescent networks, and to compare the cumula-
restraint” (EI = 1.00), the combined variable “desiring weight loss”/
tive strength of the connections within the networks.
“feeling fat” (EI = 1.67), “dietary rules” (EI = 1.25), and the combined
The NCT identified no significant differences in global strength
variable “avoid exposure”/“dissatisfaction with shape” (EI = 1.21). Lax-
(S = .18, p = .82) or network invariance (M = .30, p = .07), suggesting that
ative misuse and excessive exercising were the most peripheral nodes
the strengths of the connections within the networks were similar in
in the network. Using bootstrapped difference tests, we found no
adults and adolescents. Because no significant differences were found in
difference in node strength between “dietary restraint,” “shape
network invariance tests, specific edges were not tested because this
overvaluation,” and the combined variable “desiring weight loss”/
can increase the likelihood of Type I errors (Van Borkulo et al., 2017).
“feeling fat,” which were all significantly stronger than almost half of
the other symptoms (all ps < .05) (Figure 3).
4
3.2.2
Adolescents
|
|
DI SCU SSION
This study involved using a network approach to assess the relationships between psychopathological features in a large sample of adult
Four pairs of variables to be combined were found in the adoles-
and adolescent patients with anorexia nervosa recruited in inpatient
cent network using the goldbricker function: (a) “desire to have
and outpatient settings. There were three findings. Inspecting the
a flat stomach” and “desire to have an empty stomach”;
two networks separately, we found that some symptoms, in particu-
(b) “discomfort seeing body” and “dissatisfaction with shape”;
lar desiring weight loss and shape and weight preoccupation and
(c) “excluding food” and “dietary restraint”; and (d) “weight over-
overvaluation, were central nodes with strong connections to all the
valuation” and “shape or weight preoccupation.” These pairs were,
other eating disorder variables in both adult and adolescent net-
respectively, combined into single variables using the net_reduce
works. This finding is in line with the cognitive behavioral theory of
function. Thus, the final network of adolescent patients comprised
eating disorders, which postulates that overvaluation of shape and
22 variables.
weight is one of core psychopathological symptoms of eating disor-
The adolescent network, together with its node strengths and
ders, including anorexia nervosa, irrespective of patient age (Dalle
expected influence values, is presented in Figures 4 and 5. It indi-
Grave & Calugi, 2020). It is also supported by results from a previous
cates that the variables “discomfort seeing body”/“dissatisfaction
study on patients with anorexia nervosa which used the same statis-
with shape” (S = 2.05) and “weight overvaluation”/“shape or weight
tical approach and the same instrument to evaluate eating disorder
preoccupation” (S = 1.72), “desiring weight loss” (S = 1.03) and
psychopathology and behaviors (Forrest et al., 2018). Our second
“shape overvaluation” (S = 1.25) were nodes with the highest
finding was that, upon visual inspection, the two networks displayed
strengths. Nodes with higher expected influence coefficients
similar general structures, with nodes reflecting eating behaviors
also had higher strength, namely “discomfort seeing body”/
(restraint, restriction, excluded food, and dietary rules) and body
“dissatisfaction with shape” (EI = 1.94), “weight overvaluation”/
image concerns (body and weight dissatisfaction, discomfort seeing
“shape or weight preoccupation” (EI = 1.88), “desiring weight loss”
body, and shape and weight overvaluation) forming distinct, closed
(EI = 1.57), and “fear of gaining weight” (EI = 1.14). The most periph-
but connected clusters. This observation is confirmed by the results
eral and unconnected nodes in the network were laxative misuse
of the NCTs, which indicated no significant differences in network
and excessive exercising.
structure between the two groups. This finding suggests that
No differences were found between strength values for com-
anorexia nervosa psychopathology is similar in adults and adoles-
bined variables “discomfort seeing body”/“dissatisfaction with shape”
cents, and contrasts with recent data provided by Christian
and “weight overvaluation”/“shape or weight preoccupation” and
et al. (2020), who they found significant differences in eating disor-
nodes “desiring weight loss” and “shape overvaluation,” which were
der psychopathology networks across developmental stages. How-
significantly higher than almost half of the other symptoms in this
ever, their data pertain to a transdiagnostic sample, including
adolescents too (all ps < .05) (Figure 6).
patients with a diagnosis of binge-eating disorder, in which, as mentioned, overvaluation of shape and weight is only present in half of
cases (Coffino, Udo, & Grilo, 2019). Furthermore, the majority of
3.3
|
Stability of centrality indices
their participants did not have a clinician-informed diagnosis of an
eating disorder.
The strength CS coefficient was excellent (Epskamp & Fried, 2018) for
That being said, replicability remains a concern in network
both adult (strength [S] = 0.75) and adolescent (strength [S] = 0.75)
analysis.
samples.
Rhemtulla, and Cramer (2018) stated that if inconsistent findings arise,
Borsboom,
Robinaugh,
The
Psychosystems
Group,
10
CALUGI ET AL.
this may either be because the phenomenon is unstable or illusory
eating disorders, and this could explain the alignment of items with
(i.e., the finding is not replicable) or because of substantively meaning-
this model. Indeed, other studies using different instruments to eval-
ful differences between studies (i.e., the finding is not generalizable to
uate eating disorder psychopathology in patients with eating disor-
the context of another study). In our context, we suggest that the dif-
ders have found different central nodes, including interoceptive
ferences in findings between our study and the transdiagnostic study
awareness, ineffectiveness, depression, and anxiety (Monteleone
by Christian et al. (2020) could be explained by the fact that we
et al., 2019; Olatunji et al., 2018; Solmi et al., 2018; Solmi
included only patients with a clinical diagnosis of anorexia nervosa,
et al., 2019). Nevertheless, our results do lend further weight to the
thereby limiting the clinical differences across the adolescent and the
validity of cognitive behavioral theory, appearing to confirm that
adult samples.
overvaluation of shape and weight and additional indicators of shape
The third finding in our study was that some eating disorder
and weight concern are central features in the psychopathology of
behaviors, in particular laxative misuse and excessive exercising, had
both adults and adolescents with anorexia nervosa. Future studies
only a peripheral role in both adult and adolescent psychopathology.
should explore the network structure of anorexia nervosa psychopa-
This finding supports the current diagnostic definition of anorexia
thology in more depth, including the external maintenance mecha-
nervosa, which does not include laxative misuse and excessive
nisms (i.e., clinical perfectionism, core low self-esteem, interpersonal
exercising among its core symptoms.
difficulties, and mood intolerance) proposed by cognitive behavioral
The main strength of our investigation was the use of network
analysis to examine a large sample of patients with anorexia nervosa
theory, and evaluate treatment-related changes in adults and
adolescents.
to investigate the interconnections in both adult and adolescent eating disorder psychopathology. As in similar studies (Christian
DATA AVAILABILITY STAT EMEN T
et al., 2020; Forrest et al., 2018; Forrest, Sarfan, Ortiz, Brown, and
The data that support the findings of this study are available from the
Smith, 2019), the use of the same instrument to evaluate psycho-
corresponding author, S. C., upon reasonable request.
pathological features in the two groups enabled us to make an accurate comparison between their clinical characteristics. Nevertheless,
OR CID
to our knowledge, all previous studies using a network approach
Simona Calugi
https://orcid.org/0000-0002-4028-1877
included a transdiagnostic groups of patients with eating disorders
(Christian et al., 2020) or did not compare different age groups
(Forrest et al., 2018).
However, the study does present some limitations. The first is
the use of a self-report questionnaire, the EDE-Q, instead of a more
comprehensive semi-structured interview, to evaluate eating disorder psychopathology. That being said, the EDE-Q is a well validated
tool extensively used to evaluate patients with eating disorders in
clinical and research settings, and is strongly correlated with the
EDE interview (Berg, Peterson, Frazier, & Crow, 2012). The second
limitation of the study was the inclusion of items designed to assess
eating disorder psychopathology, but no items on general psychopathology (i.e., anxiety, depression, emotion-related variables, etc.);
this narrow range of variables included in the network analysis made
it impossible for us to evaluate the role of other features as maintenance mechanisms in the patients' psychopathology. The third limitation is the failure to evaluate gender differences. However, a
recent study found more similarities than differences in eating disorder networks in men and women (Perko, Forbush, Siew, &
Tregarthen, 2019). The fourth limitation is that it is impossible to
rule out the differences observed in our study being a function of
using different variables combining redundant nodes. Finally, the
cross-sectional nature of the study prevents us from making inferences about the directionality of the relationships we detected. This
means that we are not in a position to draw conclusions about clinical treatment, and we can merely point toward avenues for future
research.
Our results should be evaluated bearing in mind that the EDE-Q
was specifically developed from the cognitive behavioral theory of
RE FE RE NCE S
Ackard, D. M., Richter, S., Egan, A., & Cronemeyer, C. (2014). Poor outcome and death among youth, young adults, and midlife adults with
eating disorders: An investigation of risk factors by age at assessment.
International Journal of Eating Disorders, 47(7), 825–835. https://doi.
org/10.1002/eat.22346
Ambwani, S., Cardi, V., Albano, G., Cao, L., Crosby, R. D., Macdonald, P., …
Treasure, J. (in press). A multicenter audit of outpatient care for adult
anorexia nervosa: Symptom trajectory, service use, and evidence in support of "early stage" versus "severe and enduring" classification. International Journal of Eating Disorders. https://doi.org/10.1002/eat.23246
Azur, M. J., Stuart, E. A., Frangakis, C., & Leaf, P. J. (2011). Multiple imputation by chained equations: What is it and how does it work? International Journal of Methods in Psychiatric Research, 20(1), 40–49. https://
doi.org/10.1002/mpr.329
Berg, K. C., Peterson, C. B., Frazier, P., & Crow, S. J. (2012). Psychometric
evaluation of the eating disorder examination and eating disorder
examination-questionnaire: A systematic review of the literature. International Journal of Eating Disorders, 45(3), 428–438. https://doi.org/
10.1002/eat.20931
Borsboom, D. (2017). A network theory of mental disorders. World Psychiatry, 16(1), 5–13. https://doi.org/10.1002/wps.20375
Borsboom, D., Robinaugh, D. J., Psychosystems, G., Rhemtulla, M., &
Cramer, A. O. J. (2018). Robustness and replicability of psychopathology networks. World Psychiatry, 17(2), 143–144. https://doi.org/10.
1002/wps.20515
Brewin, N., Baggott, J., Dugard, P., & Arcelus, J. (2014). Clinical normative data
for eating disorder examination questionnaire and eating disorder inventory for DSM-5 feeding and eating disorder classifications: A retrospective study of patients formerly diagnosed via DSM-IV. European Eating
Disorders Review, 22(4), 299–305. https://doi.org/10.1002/erv.2301
Calugi, S., Dalle Grave, R., Sartirana, M., & Fairburn, C. G. (2015). Time to
restore body weight in adults and adolescents receiving cognitive
CALUGI ET AL.
behaviour therapy for anorexia nervosa. Journal of Eating Disorders, 3,
21. https://doi.org/10.1186/s40337-015-0057-z
Calugi, S., Milanese, C., Sartirana, M., El Ghoch, M., Sartori, F.,
Geccherle, E., … Dalle Grave, R. (2017). The Eating Disorder Examination Questionnaire: Reliability and validity of the Italian version. Eating
and Weight Disorders, 22(3), 509–514. https://doi.org/10.1007/
s40519-016-0276-6
Carter, J. C., Stewart, D. A., & Fairburn, C. G. (2001). Eating disorder examination questionnaire: Norms for young adolescent girls. Behaviour
Research and Therapy, 39(5), 625–632. https://doi.org/10.1016/
s0005-7967(00)00033-4
Chen, J., & Chen, Z. (2008). Extended Bayesian information criteria for
model selection with large model spaces. Biometrika, 95(3), 759–771.
https://doi.org/10.1093/biomet/asn034
Christian, C., Perko, V. L., Vanzhula, I. A., Tregarthen, J. P.,
Forbush, K. T., & Levinson, C. A. (2020). Eating disorder core symptoms and symptom pathways across developmental stages: A network
analysis. Journal of Abnormal Psychology, 129(2), 177–190. https://doi.
org/10.1037/abn0000477
Coffino, J. A., Udo, T., & Grilo, C. M. (2019). The significance of overvaluation of shape or weight in binge-eating disorder: Results from a
national sample of U.S. adults. Obesity, 27(8), 1367–1371. https://doi.
org/10.1002/oby.22539
Costantini, G., Epskamp, S., Borsboom, D., Perugini, M., Mõttusc, R.,
Waldorp, L. J., & Cramer, A. O. J. (2015). State of the aRt personality
research: A tutorial on network analysis of personality data in R. Journal of Research in Personality, 54, 13–29. https://doi.org/10.1016/j.jrp.
2014.07.003
Dalle Grave, R., & Calugi, S. (2020). Cognitive behavior therapy for adolescents with eating disorders. New York, NY: Guilford Press.
Dalle Grave, R., Calugi, S., Doll, H. A., & Fairburn, C. G. (2013). Enhanced
cognitive behaviour therapy for adolescents with anorexia nervosa: An
alternative to family therapy? Behaviour Research and Therapy, 51(1),
R9–r12. https://doi.org/10.1016/j.brat.2012.09.008
Epskamp, S., Borsboom, D., & Fried, E. I. (2018). Estimating psychological
networks and their accuracy: A tutorial paper. Behavior Research
Methods, 50(1), 195–212. https://doi.org/10.3758/s13428-017-0862-1
Epskamp, S., Cramer, A. O. J., Waldorp, L. J., Schmittmann, V. D., &
Borsboom, D. (2012). qgraph: Network visualizations of relationships
in psychometric data. Journal of Statistical Software, 48(4), 18. https://
doi.org/10.18637/jss.v048.i04
Epskamp, S., & Fried, E. I. (2018). A tutorial on regularized partial correlation networks. Psychological Methods, 23(4), 617–634. https://doi.org/
10.1037/met0000167
Epskamp, S., Kruis, J., & Marsman, M. (2017). Estimating psychopathological networks: Be careful what you wish for. PLoS One, 12(6),
e0179891. https://doi.org/10.1371/journal.pone.0179891
Fairburn, C. G. (2005). Evidence-based treatment of anorexia nervosa.
International Journal of Eating Disorders, 37(Suppl), S26–S42. https://
doi.org/10.1002/eat.20112
Fairburn, C. G., Cooper, Z., & Shafran, R. (2003). Cognitive behaviour therapy for eating disorders: A "transdiagnostic" theory and treatment.
Behaviour Research and Therapy, 41(5), 509–528. https://doi.org/10.
1016/s0005-7967(02)00088-8
Fisher, M., Schneider, M., Burns, J., Symons, H., & Mandel, F. S. (2001).
Differences between adolescents and young adults at presentation to
an eating disorders program. Journal of Adolescent Health, 28(3),
222–227. https://doi.org/10.1016/s1054-139x(00)00182-8
Forbes, M. K., Wright, A. G. C., Markon, K. E., & Krueger, R. F. (2017). Evidence that psychopathology symptom networks have limited replicability. Journal of Abnormal Psychology, 126(7), 969–988. https://doi.
org/10.1037/abn0000276
Forrest, L. N., Jones, P. J., Ortiz, S. N., & Smith, A. R. (2018). Core psychopathology in anorexia nervosa and bulimia nervosa: A network
11
analysis. International Journal of Eating Disorders, 51(7), 668–679.
https://doi.org/10.1002/eat.22871
Forrest, L. N., Sarfan, L. D., Ortiz, S. N., Brown, T. A., & Smith, A. R. (2019).
Bridging eating disorder symptoms and trait anxiety in patients with
eating disorders: A network approach. International Journal of Eating
Disorders, 52(6), 701–711. https://doi.org/10.1002/eat.23070
Foygel, R., & Drton, M. (2010). Extended Bayesian information criteria for
Gaussian graphical models. Advances in Neural Information Processing
Systems, 23, 2020–2028.
Foyger Barber, R., & Drton, M. (2015). High-dimensional Ising model selection with Bayesian information criteria. Electronic Journal of Statistics,
1, 567–607. https://doi.org/10.1214/15-EJS1012
Goldschmidt, A. B., Crosby, R. D., Cao, L., Moessner, M., Forbush, K. T.,
Accurso, E. C., & Le Grange, D. (2018). Network analysis of pediatric
eating disorder symptoms in a treatment-seeking, transdiagnostic sample. Journal of Abnormal Psychology, 127(2), 251–264. https://doi.org/
10.1037/abn0000327
Levinson, C. A., Vanzhula, I. A., Brosof, L. C., & Forbush, K. (2018).
Network analysis as an alternative approach to conceptualizing
eating disorders: Implications for research and treatment. Current
Psychiatry Reports, 20(9), 67. https://doi.org/10.1007/s11920018-0930-y
McNally, R. J. (2016). Can network analysis transform psychopathology?
Behaviour Research and Therapy, 86, 95–104. https://doi.org/10.1016/
j.brat.2016.06.006
Monteleone, A. M., Mereu, A., Cascino, G., Criscuolo, M.,
Castiglioni, M. C., Pellegrino, F., … Zanna, V. (2019). Reconceptualization of anorexia nervosa psychopathology: A network
analysis study in adolescents with short duration of the illness. International Journal of Eating Disorders, 52(11), 1263–1273. https://doi.
org/10.1002/eat.23137
Olatunji, B. O., Levinson, C., & Calebs, B. (2018). A network analysis of eating disorder symptoms and characteristics in an inpatient sample. Psychiatry Research, 262, 270–281. https://doi.org/10.1016/j.psychres.
2018.02.027
Perko, V. L., Forbush, K. T., Siew, C. S. Q., & Tregarthen, J. P. (2019). Application of network analysis to investigate sex differences in interactive
systems of eating-disorder psychopathology. International Journal of
Eating Disorders, 52(12), 1343–1352. https://doi.org/10.1002/eat.
23170
Smith, K. E., Crosby, R. D., Wonderlich, S. A., Forbush, K. T.,
Mason, T. B., & Moessner, M. (2018). Network analysis: An innovative
framework for understanding eating disorder psychopathology. International Journal of Eating Disorders, 51(3), 214–222. https://doi.org/
10.1002/eat.22836
Solmi, M., Collantoni, E., Meneguzzo, P., Degortes, D., Tenconi, E., &
Favaro, A. (2018). Network analysis of specific psychopathology and
psychiatric symptoms in patients with eating disorders. International
Journal of Eating Disorders, 51(7), 680–692. https://doi.org/10.1002/
eat.22884
Solmi, M., Collantoni, E., Meneguzzo, P., Tenconi, E., & Favaro, A. (2019).
Network analysis of specific psychopathology and psychiatric symptoms in patients with anorexia nervosa. European Eating Disorders
Review, 27(1), 24–33. https://doi.org/10.1002/erv.2633
Takahashi, M. (2017). Statistical inference in missing data by MCMC and
Non-MCMC multiple imputation algorithms: Assessing the effects of
between-imputation iterations. Data Science Journal, 16(37), 1–17.
https://doi.org/10.5334/dsj-2017-037
Team, R. C. (2018). R: A language and environment for statistical computing.
Vienna, Austria: R Foundation for Statistical Computing.
Treasure, J., Stein, D., & Maguire, S. (2015). Has the time come for a staging
model to map the course of eating disorders from high risk to severe
enduring illness? An examination of the evidence. Early Intervention in
Psychiatry, 9(3), 173–184. https://doi.org/10.1111/eip.12170
12
van Borkulo, C., Boschloo, L., Borsboom, D., Penninx, B. W. J. H.,
Waldorp, L. J., & Schoevers, R. A. (2015). Association of symptom network structure with the course of [corrected] depression. JAMA Psychiatry, 72(12), 1219–1226. https://doi.org/10.1001/jamapsychiatry.
2015.2079
van Borkulo, C., Boschloo, L., Kossakowski, J., Tio, P., Schoevers, R.,
Borsboom, D., & Waldorp, L. J. (2017). Comparing network structures
on three aspects: A permutation test. Retrieved from https://www.
researchgate.net/publication/314750838_Comparing_network_
structures_on_three_aspects_A_permutation_test
Walsh, B. T. (2013). The enigmatic persistence of anorexia nervosa. American Journal of Psychiatry, 170(5), 477–484. https://doi.org/10.1176/
appi.ajp.2012.12081074
World Health Organization. (2015). The global strategy for women's,
children's, and adolescents' health (2016–2030). Retrieved from
CALUGI ET AL.
http://www.who.int/life-course/partners/global-strategy/ewecglobalstrategyreport-200915.pdf?ua=1
Zhao, T., Liu, H., Roeder, K., Lafferty, J., & Wasserman, L. (2012). The huge
package for high-dimensional undirected graph estimation in R. Journal
of Machine Learning Research, 13, 1059–1062.
How to cite this article: Calugi S, Sartirana M, Misconel A,
Boglioli C, Dalle Grave R. Eating disorder psychopathology in
adults and adolescents with anorexia nervosa: A network
approach. Int J Eat Disord. 2020;1–12. https://doi.org/10.
1002/eat.23270