child AVOIDANT RESTRICTIVE FOOD INTAKE DISORDER

Eating Disorder Core Symptoms and Symptom Pathways Across Developmental Stages: A Network Analysis

Caroline Christian University of Louisville

Victoria L. Perko University of Kansas

Irina A. Vanzhula University of Louisville

Jenna P. Tregarthen Recovery Record, Inc., San Francisco, California

Kelsie T. Forbush University of Kansas

Cheri A. Levinson University of Louisville

Eating disorders (EDs) often develop during adolescence and early adulthood but may persist, arise, or reemerge across the life span. Research and treatment efforts primarily focus on adolescent and young adult populations, leaving large knowledge gaps regarding ED symptoms across the entire developmental spectrum. The current study uses network analysis to compare central symptoms (i.e., symptoms that are highly connected to other symptoms) and symptom pathways (i.e., relations among symptoms) across five developmental stages (early adolescence, late adolescence, young adulthood, early-middle adult- hood, middle-late adulthood) in a large sample of individuals with EDs (N � 29,902; N � 32,219) in two network models. Several symptoms related to overeating, food avoidance, feeling full, and overvaluation of weight and shape emerged as central in most or all developmental stages, suggesting that some core symptoms remain central across development. Despite similarities in central symptoms, significant differences in network structure (i.e., how symptom pathways are connected) emerged across age groups. These differences suggest that symptom interconnectivity (but not symptom severity) might increase across development. Future research should continue to investigate developmental symptom differences in order to inform treatment for individuals with EDs of all ages.

General Scientific Summary Connections between eating disorder symptoms vary across stages of development. Consistent with Habit Formation Theory, symptoms were more tightly connected in older individuals, who have on average a longer duration of illness. In contrast, eating disorder central symptoms (symptoms related to overeating, food avoidance, fullness, and overvaluation of weight and shape) were relatively consistent across age groups.

Keywords: eating disorder symptoms, development, age, network analysis, eating disorders

Supplemental materials: http://dx.doi.org/10.1037/abn0000477.supp

This article was published Online First November 11, 2019. X Caroline Christian, Department of Psychological and Brain Sciences,

University of Louisville; Victoria L. Perko, Department of Psychology, University of Kansas; Irina A. Vanzhula, Department of Psychological and Brain Sciences, University of Louisville; Jenna P. Tregarthen, Recovery Record, Inc., San Francisco, California; Kelsie T. Forbush, Department of Psychology, University of Kansas; Cheri A. Levinson, Department of Psychological and Brain Sciences, University of Louisville.

The present study is a new analysis of previously analyzed data. This study is the investigation of developmental differences in eating disorder symptoms using network analysis using this dataset. No other papers have addressed similar questions as those addressed in this article. All study procedures were approved by the University of Kansas Institutional Re-

view Board (Study IRB STUDY00003260). Authors complied with APA ethical standards in the treatment of their participants. The manuscript has not been and is not posted on a website. Jenna P. Tregarthen is a co-founder and shareholder of Recovery Record, Inc. Jenna P. Tregarthen made a substantial contribution as part of data collection and curation and ap- proved the final manuscript, but she did not participate in the analysis, interpretation, or drafting of the manuscript. Kelsie T. Forbush received an industry-sponsored grant from Recovery Record, Inc. No other authors have conflicts of interest to disclose.

Correspondence concerning this article should be addressed to Cheri A. Levinson, Department of Psychological and Brain Sciences, University of Louisville, Life Sciences Building 317, Louisville, KY 40292. E-mail: [email protected]

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

Journal of Abnormal Psychology © 2019 American Psychological Association 2020, Vol. 129, No. 2, 177–190 ISSN: 0021-843X http://dx.doi.org/10.1037/abn0000477

177

Eating disorders (EDs) are serious mental illnesses associated with negative health consequences, significant impairment, and high mortality (Crow et al., 2009; Rome & Ammerman, 2003; Stice, Marti, & Rohde, 2013). Peak age of ED onset is during adolescence, between 16 and 20 years of age (Stice et al., 2013). Although EDs most commonly develop during this period, evi- dence suggests that eating pathology may persist, return, or de- velop throughout an individual’s life (Fulton, 2016; Patrick & Stahl, 2009). Indeed, studies indicate that ED symptoms occur across all developmental stages, with approximately 11% of adults aged 42–55 and 4% of adults aged 60 –70 engaging in ED behav- iors, such as binge eating, laxative/diuretic misuse, or self-induced vomiting (Mangweth-Matzek et al., 2006; Marcus, Bromberger, Wei, Brown, & Kravitz, 2007). The presence of disordered eating among middle and older adults suggests that it is important to examine EDs across the full developmental spectrum; however, to date, research has primarily focused on EDs in adolescence and early adulthood.

Past research suggests that ED symptoms may change across development. However, the nature of these differences remains unclear. In terms of diagnoses, older individuals are more likely to be diagnosed with binge eating disorder, as compared to younger individuals with EDs (Jenkins & Price, 2018). Additionally, diag- nostic migration is extremely common in EDs, which suggests that symptomatology may shift as the person and illness develop (Cas- tellini et al., 2011; Fichter & Quadflieg, 2007). In terms of sever- ity, some research suggests that disordered eating behaviors, body dissatisfaction, and distorted cognitions surrounding food decline with age (Gadalla, 2008; Forman & Davis, 2005; Tiggemann & McCourt, 2013). Reduction of ED cognitions may be related to the changing social environment over the life span. In one study, the association between negative commentary about one’s weight and shape and bulimic symptoms diminished with older age (Tzoneva, Forney, & Keel, 2015).

However, other studies indicated body dissatisfaction and diet- ing behaviors remain prevalent and may strengthen with age (e.g., Fulton, 2016). Indeed, research supports that overvaluation of weight and shape is pervasive among middle age and older adults (Forman & Davis, 2005; Patrick & Stahl, 2009; Mangweth-Matzek et al., 2006). The Habit Formation Theory of EDs suggests that maladaptive eating behaviors may begin as goal-driven (e.g., di- eting to lose weight), but with repetition, these behaviors (e.g., restriction), coupled with the reward (e.g., praise from others on losing weight), develop into a deeply engrained habit (Walsh, 2013). Similarly, binge eating and purging behaviors may begin impulsively to cope with negative emotions but can develop into compulsive rituals with repetition (Pearson, Wonderlich, & Smith, 2015). Thus, Habit Formation Theory posits that maladaptive eating behaviors and cognitions will become more deeply en- grained and habitual in later developmental stages. Indeed, studies indicate that older age of onset and longer duration-of-illness are associated with poor treatment outcomes (Noordenbos, Olden- have, Muschter, & Terpstra, 2002; Norring & Sohlberg, 1993), highlighting the clinical importance of researching eating pathol- ogy across development.

Overall, the current state of eating disorder research provides an incomplete picture of cognitions and behaviors across the life span. Thus, additional research examining the differences in eating dis- order symptoms across developmental stages is urgently needed.

Specifically, it is unknown how specific symptoms and symptom relationships might change across developmental periods to main- tain ED psychopathology.

One novel way to conceptualize EDs is network theory. Net- work analysis (NA) is a statistical methodology based on network theory, which conceptualizes psychopathology as a web of inter- connecting nodes (symptoms) and edges (associations between symptoms) that are theorized to maintain a specific illness state (Borsboom, 2017). NA allows researchers to identify specific relationships among many symptoms at once and provides oppor- tunities to visualize illness pathways (relationships among individ- ual symptoms) and identify central symptoms (symptoms that are highly connected with other symptoms in the network). NA can also identify if two networks are significantly different from each other in structure (i.e., if two symptoms are similarly associated in both networks) and global strength (how strongly symptoms are associated with each other; van Borkulo et al., 2015). This tech- nique allows researchers to investigate if (and how) two popula- tions or subgroups of a population differ in symptom connected- ness.

Several studies have used NA to understand ED psychopathol- ogy. These studies found body checking (Forbush, Siew, & Vite- vitch, 2016), fear of weight gain (Elliott, Jones, & Schmidt, 2018; Forrest, Jones, Ortiz, & Smith, 2018; Levinson et al., 2017), and other symptoms related to overvaluation of weight and shape (DuBois, Rodgers, Franko, Eddy, & Thomas, 2017; Elliott et al., 2018; Forrest et al., 2018; Goldschmidt et al., 2018; Wang, Jones, Dreier, Elliott, & Grilo, 2018) to be central, maintaining symp- toms, consistent with the cognitive– behavioral theory of EDs (Cooper & Shafran, 2008; Fairburn, 2008). A few additional studies have identified additional important symptoms, such as dietary restraint (Goldschmidt et al., 2018; Solmi et al., 2018), interoceptive awareness, (Olatunji, Levinson, & Calebs, 2018; Solmi et al., 2018), and ineffectiveness (Olatunji et al., 2018; Solmi et al., 2018; Solmi, Collantoni, Meneguzzo, Tenconi, & Favaro, 2019), and the relationships among depression, anxiety, and ED symptoms (Solmi et al., 2018, 2019).

Although NA has been applied to increase the broad understand- ing of eating pathology, no research has examined differences in network models of ED symptoms across developmental stages. Past research suggests that there may be unique differences in ED presentations across the life span, including diagnostic differences, physical changes, and differences in treatment outcomes (Cas- tellini et al., 2011; Forman & Davis, 2005; Hudson & Pope, 2018; Jenkins & Price, 2018; Peat, Peyerl, & Muehlenkamp, 2008). Thus, it seems likely that ED symptom relationships may also differ across developmental stages. Better understanding of the differences in central ED symptoms across developmental stages could help determine if alternative treatments would be more beneficial for different age groups.

The current study utilizes NA to examine ED symptoms in five distinct developmental stages: early adolescence (11–14), late ad- olescence (15–18), young adulthood (19 –25), early-middle adult- hood (26 – 45), and middle-late adulthood (46�). These age ranges represent unique developmental stages in several aspects, includ- ing social environment, physiological and neurological develop- ment, maturity, and autonomy (Blonigen, Carlson, Hicks, Krueger, & Iacono, 2008; Steinberg, 2005; Williams & Currie, 2000). We examine symptom relationships across two widely used ED mea-

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

178 CHRISTIAN ET AL.

sures: the Eating Pathology Symptoms Inventory (EPSI; Forbush et al., 2013) and Eating Disorder Examination Questionnaire (EDE-Q; Fairburn & Beglin, 1994). Both questionnaires are con- sidered “gold-standard” measures of ED symptoms and are fre- quently used for network investigations (DuBois et al., 2017; Forbush et al., 2016; Forrest et al., 2018), yet they assess slightly different aspects of ED symptoms, such that the EDE-Q is based on the cognitive– behavioral model of EDs and the EPSI is de- signed to be a multidimensional assessment of ED symptoms. Thus, we include both measures to allow for a more comprehen- sive overview of ED symptoms and to gain insight into the replicability of networks.

We hypothesized that symptoms that were central in past studies using NA (e.g., overvaluation of weight and shape; Levinson, Vanzhula, Brosof, & Forbush, 2018) would remain central regard- less of age, as suggested by the literature (Forman & Davis, 2005; Patrick & Stahl, 2009; Mangweth-Matzek et al., 2006). Further, we hypothesized that there would be a significant difference in net- work structure across networks. Despite some common threads across EDs, specific connections between symptoms are likely to differ across developmental stages, given what the literature has described in terms of differences in symptom severity and treat- ment effectiveness (Hudson & Pope, 2018; Jenkins & Price, 2018; Peat et al., 2008; Forman & Davis, 2005). For example, although fear of weight gain may remain central across diverse ED presen- tations, the connection between fear of weight gain and binge eating may become stronger over time, consistent with the Habit Formation Theory (Walsh, 2013). This change would result in differences in network structure, which has implications for im- plementing effective treatments across age groups. Additionally,

we predict that the global strength would increase for networks with older participants compared to younger participants, reflec- tive of Habit Formation Theory, indicating increased severity across developmental stages.

Method

Participants

Participants were Recovery Record users (N � 29,902; N � 32,219), a smartphone application that is based on cognitive– behavioral treatment for EDs (Tregarthen, Lock, & Darcy, 2015). Participants provided consent for data to be used for research purposes when they agreed to the “Terms and Conditions” in the initial application setup. Participants who completed the EPSI (n � 29,902) were 11 to 85 years old (M � 26.23, SD � 10.46), and 94.0% identified as female. These participants reported their av- erage length of ED was 9.71 years (SD � 9.72, range � 0 – 65 years). Recovery Record allows users to connect their account with a clinician in order to share information and inform treatment planning. In our sample, 34.5% of participants had accounts con- nected with a treatment provider and had an official diagnosis of an ED based on clinician-report.

Participants who completed the EDE-Q (n � 32,219) were 11 to 79 years old (M � 23.43, SD � 8.89), and 96.5% identified as female. Average length of ED was 7.60 years (SD � 8.23, range � 0 – 60 years). In the present sample, 8.8% of participants had accounts connected with a treatment provider and had an official diagnosis of an ED based on clinician-report. See Table 1 for

Table 1 Demographic Breakdown

Demographic characteristic

Early adolescence n (%)

Late adolescence n (%)

Young adult n (%)

Early-middle adult n (%)

Middle-late adult n (%)

EDE-Q 1523 (100) 9838 (100) 11709 (100) 7955 (100) 1194 (100) Gender

Female 1468 (96.4) 9498 (96.5) 11310 (96.6) 7671 (96.4) 1131 (94.7) Male 42 (2.8) 248 (2.5) 288 (2.9) 228 (2.9) 56 (4.7) Missing 13 (.9) 92 (.9) 111 (.7) 56 (.7) 7 (.6)

Diagnosis AN 13 (.9) 85 (.9) 159 (1.4) 133 (1.7) 19 (1.6) BN 2 (.1) 37 (.4) 99 (.8) 104 (1.3) 12 (1.0) BED 4 (.3) 12 (.1) 49 (.4) 104 (1.3) 55 (4.6) Other 3 (.2) 42 (.4) 93 (.8) 102 (1.3) 18 (1.5) Missing 1501 (98.6) 9662 (98.2) 11309 (96.6) 7522 (94.6) 1090 (91.3)

Duration of illness (M[SD]) 1.71 (1.71) 2.92 (2.27) 5.82 (3.96) 13.86 (8.37) 29.00 (14.60) EPSI 1028 (100) 6171 (100) 10701 (100) 9929 (100) 2073 (100)

Gender Female 959 (93.3) 5786 (93.8) 10108 (94.5) 9412 (94.8) 1857 (89.6) Male 46 (4.5) 228 (3.7) 381 (3.6) 438 (4.4) 201 (9.7) Missing 23 (2.2) 157 (2.5) 212 (2.0) 79 (.8) 15 (.7)

Diagnosis AN 152 (14.8) 796 (12.9) 1456 (13.6) 995 (10.0) 172 (8.3) BN 30 (2.9) 307 (5.0) 870 (8.1) 825 (8.3) 81 (3.9) BED 31 (3.0) 165 (2.7) 514 (4.8) 1144 (11.5) 527 (25.4) Other 62 (6.0) 354 (5.7) 795 (7.4) 830 (8.4) 222 (10.7) Missing 753 (73.2) 4549 (73.7) 7066 (66.0) 6134 (61.8) 1071 (51.7)

Duration of illness (M[SD]) 2.06 (2.11) 3.19 (2.44) 6.05 (4.12) 14.54 (8.58) 29.12 (15.20)

Note. EDE-Q � Eating Disorder Examination Questionnaire; EPSI � Eating Pathology Symptoms Inventory; AN � anorexia nervosa; BN � bulimia nervosa; BED � binge eating disorder.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

179EATING DISORDER AGE NETWORKS

participants’ gender, ED diagnoses, and duration of illness across developmental categories.

Measures

EPSI. The EPSI is a 45-item multidimensional measure de- signed to assess ED symptoms. The EPSI has eight scales corre- sponding to unique facets of eating pathology: Body Dissatisfac- tion (i.e., satisfaction with body shape and body parts; e.g., hips, thighs), Binge Eating (i.e., tendency to overeat or eat mindlessly), Cognitive Restraint (i.e., attempting to restrict eating, whether successful or not), Excessive Exercise (i.e., intense or compulsive exercise), Restricting (i.e., efforts to avoid or reduce eating), Purging (i.e., self-induced vomiting and laxative/diuretic use), Muscle Building (i.e., cognitions and behaviors [supplement use] related to increasing muscularity), and Negative Attitudes Toward Obesity (i.e., negative judgment of individuals who are over- weight/obese). Between 32.6% and 73.6% of our sample scored above EPSI subscale means in an ED treatment sample (Forbush et al., 2013). Two scales of the EPSI, Negative Attitudes Toward Obesity and Muscle Building, were not included in the Recovery Record app; thus, these items were not included in the network. The EPSI has excellent convergent and discriminant validity, as well as excellent test-retest reliability (Forbush et al., 2013). The internal consistency of all items included in the EPSI network was adequate for the current sample (� � .73).

EDE-Q. The EDE-Q version 6.0 is a 28-item self-report ques- tionnaire designed to assess ED behaviors and thoughts. This version of the EDE-Q has four scales: Eating Concern (i.e., inter- fering thoughts about food, eating, or calories), Shape Concern (i.e., interfering thoughts about shape), Weight Concern (i.e., in- terfering thoughts about weight), and Restraint (i.e., attempts to reduce food intake; e.g., skipping meals, food rules). The mean EDEQ global score in our sample is 4.17 (SD � 1.10), and 63.3% (n � 20,390) of our sample scored above the recommended clinical cutoff (a score of 4.0 or higher) for EDs (Fairburn, Wilson, & Schleimer, 1993). One EDE-Q item (15) was excluded because it measures the same symptom (binge eating) as the previous question. Networks should not include two questions targeting the same symptom because it may artificially inflate centrality, poten- tially leading to false interpretation of that symptom as central (Fried & Cramer, 2017). The EDE-Q has demonstrated excellent test-retest reliability and internal consistency (Luce & Crowther, 1999) and good criterion and concurrent validity (Mond, Hay, Rodgers, Owen, & Beumont, 2004). The internal consistency of all items included in the EDE-Q network was good for the current sample (� � .86).

Procedure

Participants used the Recovery Record application to self- monitor ED cognitions and behaviors. The application encourages monthly completion of the EDE-Q and the EPSI. The present study used data from the initial completion of EDE-Q and EPSI by participants using the mobile application.

Participant data were categorized into five developmental stag- es: early adolescence (11–14), late adolescence (15–18), young adulthood (19 –25), early-middle adulthood (26 – 45), and middle- late adulthood (46�). Our ranges may not fully distinguish be-

tween all stages of development because we had few participants above the age of 45 (n � 1,194 for EPSI, n � 2,073 for EDE-Q) relative to the entire sample, so we used 45 as a cutoff for middle-late adulthood in order to ensure a large sample size for the networks. Using younger age ranges is not uncommon for clinical studies on EDs due to difficulty recruiting older adults with EDs (Forman & Davis, 2005; Jenkins & Price, 2018).

Glasso networks using the EDE-Q and EPSI were estimated at each developmental stage using the “estimateNetwork” function in the bootnet package in R (Epskamp, Maris, Waldorp, & Bors- boom, 2018). The Glasso function estimates partial correlations between nodes, meaning each correlation is unique, accounting for all other symptoms in the network while minimizing spurious relationships. We first created networks using the default setting (cor_auto), which uses polychoric correlations. However, because some networks did not have adequate stability, we estimated the networks again using Spearman correlations to obtain stable net- works, as suggested by Epskamp and Fried (2018). Stability esti- mates were calculated using the bootnet package in R (Epskamp et al., 2018).

Three indices of centrality were calculated using the “centrali- typlot” function in the qgraph package in R: strength (i.e., the sum of the absolute value of all of a node’s edges), closeness (i.e., degree of direct connections to other nodes), and betweenness (i.e., degree to which a node falls on the path between other nodes; Epskamp, Cramer, Waldorp, Schmittmann, & Borsboom, 2012). We interpret only strength centrality because it was the most stable, as has been done in prior NA investigations (e.g., DuBois et al., 2017; Epskamp et al., 2012). Centrality difference tests were conducted using the bootnet package in R (Epskamp et al., 2018) to determine if central symptoms were significantly more central than other symptoms. We included three to six central symptoms for each network based on the network centrality difference test. The number of symptoms included per network is based on sharp observable decreases in centrality differences among top symp- toms that were used as cutoffs for inclusion. We did not use a standard cutoff value across networks due to internetwork vari- ability.

Differences between networks across developmental stages were identified using the NetworkComparisonTest package in R (van Borkulo et al., 2015). Three metrics were utilized to analyze network differences: network invariance test (M; i.e., significant differences in the maximum edge strength in the networks), edge invariance test (E; i.e., significant differences between specific edges in the networks), and global strength invariance test (GSI; i.e., significant differences in the sum of the edge strengths; van Borkulo et al., 2015). Edge invariance was calculated for networks with significant network invariance in order to quantify the nature of these structural differences. Global strength is a particularly useful measure, as it may be related to symptom severity (van Borkulo et al., 2015).

A one-way ANOVA was conducted across developmental stages for both the EDE-Q and EPSI to investigate whether sig- nificant differences in symptom severity across groups were re- lated to global strength across networks, as has been theorized (van Borkulo et al., 2015). We conducted these analyses using the EDE-Q global score, as factor validity is strongest for the global index rather than the four subscales (Aardoom, Dingemans, Sloft Op’t Landt, & Van Furth, 2012) and six EPSI subscales, as the

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

180 CHRISTIAN ET AL.

EPSI was designed as a multidimensional measure of eating pa- thology, rather than a global subscale of severity (Forbush et al., 2013). A post hoc Bonferroni correction was used for multiple comparisons. The cutoff value after this correction is p � .007.

Results

Networks and Stability

See Figure 1 for EPSI networks and Figure 2 for EDE-Q networks. Table 2 includes descriptions of each of the EPSI and EDE-Q items. Stability for strength was excellent (strength � .75) for all the EPSI and EDE-Q networks (Epskamp, Borsboom, & Fried, 2018).

Central Symptoms

EPSI. See Figure 3 for the strength centrality of all symptoms in the EPSI networks. All central symptoms were significantly

more central than other symptoms in the network at p � .05. Overeating and feeling full after eating a small amount of food emerged as central symptoms across every developmental stage. Avoiding high calorie foods and planning days around exercise are central symptoms in late adolescence, young adulthood, early- middle adulthood, and middle-late adulthood. Fasting is a central symptom in early adolescence, late adolescence, young adulthood, and early-middle adulthood. Stuffing oneself to the point of feeling sick is a central symptom in young adulthood, early-middle adult- hood, and middle-late adulthood. The most central symptoms in the EPSI networks are described in Table 3.

EDE-Q. See Figure 4 for the strength centrality of all symp- toms in the EDE-Q networks. All central symptoms were signif- icantly more central than other symptoms in the network at p � .05. Desire for an empty stomach emerged as a central symptom across every developmental stage. Concentration prob- lems due to weight and shape is a central symptom in early adolescence, late adolescence, young adulthood, and early-middle

A. Early Adolescence B. Late Adolescence C. Young Adulthood

D. Early-middle Adulthood E. Middle-late Adulthood

clothesfit

unhealthyfood

nothungry

eatlittle

exercisedaily

supriseeat

exercisehard

snacking

fulleasy thinkdiuretics

outfits

thinklaxatives

dietteas

dietpills

dislikebody

full

countcals

planexercise

butt thighs

shapediff

vomit

noticeate

strenexercise

fullsmall

hips

eatmore

resist

stuffed

avoidhighcal

exerciseexhaust

diureticsuse

fast

autopilot

overeat

clothesfit

unhealthyfood

nothungry

eatlittle

exercisedaily

supriseeat

exercisehard

snacking

fulleasy

thinkdiuretics

outfits

thinklaxatives

dietteas

dietpills

dislikebody

full

countcals

planexercise

butt

thighs

shapediff

vomit

noticeate

strenexercise

fullsmall

hips

eatmore

resist stuffed

avoidhighcal

exerciseexhaust

diureticsuse

fast

autopilot

overeat

clothesfit

unhealthyfood

nothungry

eatlittle

exercisedaily

supriseeat

exercisehard

snacking

fulleasy

thinkdiuretics

outfits

thinklaxatives

dietteas

dietpills

dislikebody

full

countcals

planexercise

butt thighs

shapediff

vomit

noticeate

strenexercise

fullsmall

hips

eatmore

resist

stuffed

avoidhighcal

exerciseexhaust

diureticsuse

fast

autopilot

overeat

clothesfit

unhealthyfood

nothungry

eatlittle

exercisedaily

supriseeat

exercisehard

snacking

fulleasy

thinkdiuretics

outfits

thinklaxatives

dietteas dietpills

dislikebody

full countcals

planexercise

butt thighs

shapediff

vomit

noticeate

strenexercise

fullsmall

hips

eatmore

resist

stuffed

avoidhighcal

exerciseexhaust

diureticsuse

fast

autopilot

overeat

clothesfit

unhealthyfood

nothungry

eatlittle

exercisedaily

supriseeat

exercisehard

snacking fulleasy

thinkdiuretics

outfits

thinklaxatives

dietteas dietpills

dislikebody

full

countcals

planexercise

butt

thighs

shapediff

vomit

noticeate

strenexercise

fullsmall

hips eatmore

resist

stuffed

avoidhighcal

exerciseexhaust

diureticsuse

fast

autopilot

overeat

Figure 1. EPSI networks for (A) early adolescence (11–14), (B) late adolescence (15–18), (C) young adulthood (19 –25), (D) early-middle adulthood (26 – 45), and (E) middle-late adulthood (46�). Blue (solid) edges represent positive partial correlations. Red (dashed) lines represent negative partial correlations. Line thickness represents the strength of the partial correlation. See Table 2 for EPSI items corresponding to each node. See the online article for the color version of this figure.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

181EATING DISORDER AGE NETWORKS

adulthood. Feeling dissatisfied about one’s weight is a central symptom in early adolescence, young adulthood, early-middle adult- hood, and middle-late adulthood. Overeating is a central symptom in late adolescence, young adulthood, early-middle adulthood, and middle-late adulthood. Desire to lose weight is a central symptom in early and late adolescence. Judgment of self due to shape is a central symptom in early adolescence. Binge eating is a central symptom in young adulthood. Dissatisfaction about one’s shape is a central symp- tom in middle-late adulthood. The most central symptoms in the EDE-Q networks are described in Table 4.

EPSI networks. The network invariance test indicated that the early adolescence network was significantly different than late ado- lescence (M � 0.12, p � .05), young adulthood (M � 0.52, p � .05), early-middle adulthood (M � 0.23, p � .001), and middle-late adult- hood (M � 0.29, p � .001). The late adolescence network was significantly different from early-middle adulthood (M � 0.17, p � .001) and middle-late adulthood (M � 0.24, p � .001), but not young adulthood (p � .05). The young adulthood network was not signifi- cantly different from early-middle adulthood or middle-late adulthood

(p � .05). The early-middle adulthood network was significantly different than middle-late adulthood (M � 0.10, p � .02).

The edge invariance test indicated that two edges were significantly different (p � .05) between early adolescence and late adolescence, one edge significantly differed between early adolescence and young adulthood, 16 edges significantly differed between early adolescence and early-middle adulthood, 13 edges significantly differed between early adolescence and middle-late adulthood, 20 edges significantly differed between late adolescence and early-middle adulthood, 19 edges significantly differed between late adolescence and middle-late adulthood, and two edges significantly differed between early-middle adulthood and middle-late adulthood. See online supplemental mate- rials for all significantly different edges and corresponding E-values. The Global Strength Invariance test indicated that there were no significant differences in global strength among the EPSI networks of different developmental stages (p � .05).

EDE-Q. The structure of the early adolescence network was significantly different than young adulthood (M � 0.15, p � .001), early-middle adulthood (M � 0.17, p � .001), and middle-late adult-

A. Early Adolescence B. Late Adolescence C. Young Adulthood

D. Early-middle Adulthood E. Middle-late Adult Adulthood

restrict

fast

excludefood

foodrules

emptystomach

flatstomach

foodconc wsconc

losecontrolfeargain

feelfat

desirelose

overeat

binge

vomit

laxativescompex

eatsecret

guilty

otherseeeat

weightjudge

shapejudge

weighself

weighdiss

shapediss

seeself

otherseebody

restrict

fast

excludefood

foodrules

emptystomach

flatstomach

foodconc

wsconc

losecontrol

feargain

feelfat

desirelose

overeat

binge

vomit

laxatives

compex

eatsecret guilty

otherseeeat

weightjudge

shapejudge

weighself

weighdiss

shapediss

seeself

otherseebody

restrict

fast

excludefood

foodrules

emptystomach

flatstomach

foodconc

wsconc

losecontrol

feargain

feelfat

desirelose

overeat

binge

vomit

laxatives compex

eatsecret

guilty

otherseeeat

weightjudge shapejudge

weighself

weighdiss

shapediss seeself

otherseebody

restrict

fast

excludefood

foodrules

emptystomach

flatstomach

foodconcwsconc

losecontrol

feargain feelfat

desirelose

overeat

binge

vomit

laxatives

compex

eatsecret

guilty

otherseeeat

weightjudge

shapejudge

weighself

weighdiss

shapediss

seeself

otherseebody

restrict

fast

excludefood

foodrules emptystomach

flatstomach

foodconc wsconc

losecontrol

feargain

feelfat

desirelose

overeat

binge

vomit

laxatives

compex

eatsecret

guilty

otherseeeat

weightjudge

shapejudge

weighself

weighdiss

shapediss

seeself otherseebody

Figure 2. EDE-Q networks for (A) early adolescence (11–14), (B) late adolescence (15–18), (C) young adulthood (19 –25), (D) early-middle adulthood (26 – 45), and (E) middle-late adulthood (46�). Blue (solid) edges represent positive partial correlations. Red (dashed) lines represent negative partial correlations. Line thickness represents the strength of the partial correlation. See Table 2 for EDE-Q items corresponding to each node. See the online article for the color version of this figure.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

182 CHRISTIAN ET AL.

Table 2 Network Node (i.e., Symptom) Abbreviations

EPSI

clothesfit Dislike how clothes fit unhealthyfoods Attempt to exclude “unhealthy” foods nothungry Ate when not hungry eatlittle Told that I do not eat much exercisedaily Felt the need to exercise nearly daily supriseeat People would be surprised by how little I ate exercisehard Push myself hard when exercising snacking Snacked without realizing fulleasy Got full easily thinkdiuretics Considered taking diuretics outfits Tried on different outfits because of how I looked thinklaxatives Thought laxatives are good to lose weight dietteas Used diet teas or cleansing teas dietpills Used diet pills dislikebody Dislike how my body looked full Ate until uncomfortably full countcals Counted calories planexercise Planned days around exercising butt Thought butt was too big thighs Dislike size of thighs shapediff Wished shape of body was different vomit Vomited to lose weight noticeate Did not notice how much I ate until after strenexercise Engaged in strenuous exercise at least 5 days per week fullsmall Got full after eating a small amount of food hips Dissatisfied with the size of hips eatmore Others encouraged to eat more resist Felt I could not resist eating food offered stuffed Stuffed myself with food avoidhighcal Tried to avoid foods with high calories exerciseexhaust Exercised to exhaustion diureticsuse Used diuretics to lose weight fast Skipped two meals in a row autopilot Ate on autopilot overeat Ate a large amount of food in a short period of time

EDE-Q

restrict Tried to limit the amount of food eaten for shape or weight concerns fast Gone for long periods of time without eating for shape or weight concerns excludefood Tried to exclude foods that you like for shape or weight concerns foodrules Tried to follow food rules for shape or weight concerns emptystomac Definite desire to have an empty stomach flatstomach Definite desire to have a flat stomach foodconc Thinking about food, eating, or calories made it difficult to concentrate wsconc Thinking about shape or weight made it difficult to concentrate losecontrol Definite fear of losing control overeating feargain Fear that you might gain weight feelfat Felt fat desirelose Strong desire to lose weight overeat Ate an unusually large amount of food binge Had a sense of losing control over your eating and ate an unusually large amount of food vomit Made yourself sick (vomit) for shape or weight concerns laxatives Taken laxatives for shape or weight concerns compex Exercised in a “driven” or “compulsive” way for shape or weight concerns eatsecret Ate in secret guilty Felt guilty for eating due to shape or weight concerns otherseeeat Concerned about other people seeing you eat weightjudge Weight influenced self-judgment shapejudge Shape influenced self-judgment weighself Upset if had to weigh once a week weightdiss Dissatisfied with weight shapediss Dissatisfied with shape seeself Discomfort seeing your own body otherseebody Discomfort with others seeing your body

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

183EATING DISORDER AGE NETWORKS

hood (M � 0.17, p � .01), but not late adolescence. The late adolescence network was significantly different from young adult- hood (M � 0.07, p � .05), early-middle adulthood (M � 0.10, p � .001), and middle-late adulthood (M � 0.16, p � .001). The young adulthood network was significantly different than middle-late adult- hood (M � 0.14, p � .05), but not early-middle adulthood. The early-middle adulthood network was not significantly different than middle-late adulthood.

Eight edges were significantly different (p � .05) between early adolescence and young adulthood, 20 edges significantly differed

between early adolescence and early-middle adulthood, 23 edges significantly differed between early adolescence and middle-late adulthood, 12 edges significantly differed between late adoles- cence and young adulthood, 13 edges significantly differed be- tween late adolescence and early-middle adulthood, 19 edges significantly differed between late adolescence and middle-late adulthood, and seven edges significantly differed between young adulthood and middle-late adulthood. See online supplemental materials for all significantly different edges and corresponding E-values.

Figure 3. Centrality of EPSI symptoms for the (A) early adolescence, (B) late adolescence, (C) young adulthood, (D) early-middle adulthood, and (E) middle-late adulthood networks. Red (large) dots denote most central symptoms. See Table 2 for EPSI items corresponding to each node abbreviation. See the online article for the color version of this figure.

Table 3 EPSI Central Symptoms

Early adolescence Late adolescence Young adulthood Early-middle adulthood Middle-late adulthood

Overeat (1.88) Overeat (1.68) Overeat (1.72) Overeat (1.58) Overeat (1.98) Fullsmall (2.35) Fullsmall (1.45) Fullsmall (1.88) Fullsmall (1.92) Fullsmall (1.82)

Avoidhighcal (1.54) Avoidhighcal (1.44) Avoidhighcal (1.72) Avoidhighcal (1.47) Planexercise (1.29) Planexercise (1.57) Planexercise (1.69) Planexercise (1.30)

Fast (1.75) Fast (2.53) Fast (2.19) Fast (1.27) Stuffed (1.44) Stuffed (1.89) Stuffed (1.87)

Note. Standardized strength centrality coefficients included in parentheses. All symptoms in the table were significantly more central than over 75% of other symptoms in the network. See Table 2 for EPSI items corresponding to each node abbreviation.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

184 CHRISTIAN ET AL.

The early adolescence network (global strength � 11.82) had significantly lower global strength than middle-late adulthood (global strength � 12.56; GSI � 0.74, p � .05). Late adolescence (global strength � 12.73) had significantly lower strength than young adult- hood (global strength � 13.48; GSI � 0.75, p � .01) and early- middle adulthood (global strength � 13.61; GSI � 0.89, p � .05). There were no other significant differences in global strength among

the EDE-Q networks (p � .05). See Table 5 for an overview of network differences across developmental stages.

ANOVA Across Developmental Stages

The results of the one-way ANOVAs indicated a significant main effect of group for body dissatisfaction, F(4, 29,897) �

Figure 4. Centrality of EDE-Q symptoms for the (A) early adolescence, (B) late adolescence, (C) young adulthood, (D) early-middle adulthood, and (E) middle-late adulthood networks. Red (large) dots denote most central symptoms. See Table 2 for EDE-Q items corresponding to each abbreviation. See the online article for the color version of this figure.

Table 4 EDE-Q Central Symptoms

Early adolescence Late adolescence Young adulthood Early-middle adulthood Middle-late adulthood

Emptystomach� (.98) Emptystomach�� (1.11) Emptystomach�� (1.30) Emptystomach�� (1.72) Emptystomach�� (1.73) Wsconc�� (1.48) Wsconc�� (1.13) Wsconc� (1.07) Wsconc�� (1.13)

Overeat� (1.32) Overeat�� (1.43) Overeat�� (1.56) Overeat� (1.11) Weightdiss� (.98) Weightdiss�� (1.13) Weightdiss�� (1.41) Weightdiss� (1.09) Desirelose� (.95) Desirelose� (1.05) Shapejudge� (1.30)

Binge� (.99) Shapediss� (1.11)

Note. Standardized strength centrality coefficients included in parentheses. � Symptom is significantly more central than over 50% of other symptoms in the network. �� Symptom is significantly more central than over 75% of other symptoms in the network. See Table 2 for EDE-Q items corresponding to each node abbreviation.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

185EATING DISORDER AGE NETWORKS

10.03, cognitive restraint, F(4, 29,897) � 183.28, binge eating, F(4, 29,897) � 183.28, purging, F(4, 29,897) � 189.43, restric- tion, F(4, 29,897) � 700.49, excessive exercise, F(4, 29,897) � 215.31, and global ED symptoms, F(4, 32,214) � 107.64, p � .001. Post hoc pairwise comparisons indicated that body dissatis- faction was highest in late adolescence, young adulthood, and early-middle adulthood. Purging was highest in late adolescence and young adulthood. Restriction, excessive exercise, cognitive restraint, and global ED symptoms were highest in early adoles- cence and significantly declined across development. Binge eating was lowest in early adolescence and significantly increased across development. See Table 6 for means and standard deviations for these measures across each developmental stage.

Discussion

This study utilizes NA to explore ED symptoms across funda- mental developmental stages of adolescence and adulthood in a large sample of Recovery Record users. We hypothesized that central symptoms would be consistent across developmental stages but that the individual connections or pathways (edges) between symptoms may differ in strength. In support of our hypothesis, several symptoms emerged as central across all or most develop- mental stages. In partial support of our second hypothesis, there were significant differences in the network structure for all ED networks across both measures, but only significant differences in global strength among some of the EDE-Q networks. However, the results of the ANOVA contradicted these findings, as for most ED

symptoms, excluding binge eating, symptom severity was highest for adolescence and young adulthood and declined later in adult- hood, suggesting that the strength of the connections (but not the severity of symptoms) may increase across development. Overall, these network comparison results suggest that although many of the central symptoms remain consistent across developmental stages, the connections among symptoms significantly differ.

Central Symptoms

Several symptoms, including overeating and cognitions related to fullness, were central symptoms at every developmental stage. Several additional symptoms were central in four of the five networks, including symptoms related to food avoidance, overeat- ing, and overvaluation of weight and shape. The high proportion of symptoms that were central across most or all developmental stages suggests that these ED symptoms may be central regardless of developmental stage. Thus, these symptoms may represent important targets for intervention for individuals with EDs across all developmental stages. Some symptoms were unique to one or two developmental stages, including additional symptoms related to overvaluation of weight and shape (e.g., dissatisfaction about one’s shape; desire to lose weight). These symptoms may represent unique targets of intervention for the treatment of EDs in specific age populations.

Additionally, many of the central symptoms represent symp- toms related to overvaluation of weight and shape, including concentration problems due to weight and shape, dissatisfaction

Table 5 Network Comparison Tests

Early adolescence Late adolescence Young adulthood

Early-middle adulthood

Middle-late adulthood

Developmental stage M GSI M GSI M GSI M GSI M GSI

Early adolescence — — .11 .92 .15� 1.67 .17� 1.80 .18� .75�

Late adolescence .12� .81 — — .07� .75� .10� .88� .17� .17 Young adulthood .52� 2.88 .52 2.07 — — .06 .13 .14� .92 Early-middle adulthood .23� 1.60 .17� .78 .49 1.29 — — .12 1.05 Middle-late adulthood .29� .33 .24� .48 .51 2.55 .10� 1.26 — —

Note. Bottom left (not bold) values represent network comparisons among EPSI networks. Upper right (bold) values represent network comparisons among EDE-Q networks. M � network invariance test statistic; GSI � global strength invariance test statistic. � p � .05.

Table 6 Means and Standard Deviations of Study Measures Across Developmental Stages

Outcome Early

adolescence Late

adolescence Young

adulthood Early-middle

adulthood Middle-late adulthood

EDE-Q global 4.32 (1.07) 4.29 (1.07) 4.18 (1.11) 4.05 (1.10) 3.73 (1.14) EPSI body dissatisfaction 20.70 (6.29) 21.36 (5.67) 21.19 (5.88) 21.36 (6.05) 20.59 (6.40) EPSI cognitive restraint 8.26 (3.30) 8.11 (3.14) 7.84 (3.09) 7.20 (3.12) 6.44 (2.90) EPSI binge eating 12.39 (9.04) 15.62 (9.37) 16.71 (9.37) 17.95 (9.02) 17.58 (8.16) EPSI purging 5.63 (5.94) 6.58 (5.92) 6.10 (5.88) 5.08 (5.49) 3.08 (4.34) EPSI restriction 13.98 (6.25) 12.62 (6.37) 10.75 (6.49) 8.22 (6.49) 6.91 (5.69) EPSI excessive exercise 9.71 (5.81) 9.08 (5.65) 8.63 (5.82) 7.44 (5.67) 5.71 (4.82)

Note. Values reported as M (SD). Italicized values were not significantly different (p � .007) than at least one other developmental stage. Bolded values denote stages significantly different than three or more other developmental stages.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

186 CHRISTIAN ET AL.

about one’s weight, dissatisfaction about one’s shape, judgment about one’s shape, and desire to lose weight. This finding is consistent with past conceptualizations of eating pathology using NA (DuBois et al., 2017; Forrest et al., 2018; Levinson et al., 2017; Wang et al., 2018) and supports the theory that overvalu- ation of weight and shape are core ED symptoms (Fairburn, 2008). A few symptoms that were highly central, including overeating and food avoidance, had not previously emerged as central in past studies. Thus, more research should test if these results replicate in other samples.

Differences Across Development

Despite the number of central symptoms that remained similar across developmental stages, the network comparison tests re- vealed significant differences in how symptoms were related across networks. The adolescent networks for both the EDE-Q and EPSI were significantly different from all the adulthood networks, suggesting that symptom relationships during adolescence signif- icantly vary from adulthood. Additionally, the networks represent- ing stages of adulthood were significantly different from each other for both measures, indicating that symptom relationships also are highly variable across the developmental stages of adulthood. The edge invariance tests supported these findings, as there were many significantly different edges across networks. All signifi- cantly different edges are included in online supplemental mate- rials, as these edges represent pathways that may be differently important across developmental stages and provide insight into the clinical significance of network structure differences.

Overall, these findings suggest that important illness pathways may change across development, indicating that clinicians should expect fluctuations in the relationships among ED symptoms that occur with time and life experiences and that these changes may alter intervention targets. For example, fear of weight gain may be a common driving symptom across stages of development, but it may manifest differently over time (e.g., restriction may be more prevalent early on, but later shifts to judgment fears and isolation). Therefore, interventions may need to be tailored to address such changes.

In terms of global strength, only the EDE-Q networks exhibited significant differences, with trends indicating global strength in- creases for networks with older participants compared to younger participants. Higher global strength is theorized to be representa- tive of greater severity (Pe et al., 2015; van Borkulo et al., 2015). However, comparisons in EDE-Q global scores and EPSI subscale scores across stages of development indicated that symptoms (based on total symptom scores) were more severe (i.e., higher) in the adolescent and young adult groups for all symptoms except binge eating. Bos et al. (2018) also found increased network connectivity corresponding with decreased severity, contrary to findings by van Borkulo et al. and Pe et al. As such, global strength may not necessarily correspond to greater overall severity of symptoms, but instead tighter connections between symptoms. The high interconnectivity of symptoms in the later developmental stages may be attributed to the longer average duration of illness of older individuals with EDs in our sample, which would likely indicate stronger, more reinforced pathways among symptoms, as suggested by Habit Formation Theory (Walsh, 2013).

Contrary to this finding, no significant differences in the global strength emerged across EPSI networks. This result was surpris- ing, as the network comparison test detects even small differences. However, group comparisons indicated that some symptoms (e.g., binge eating) were stronger for older ages and other symptoms (e.g., restriction) were stronger in younger ages, so these opposing trends potentially “cancelled” each other out in the summation of strength across networks. It is also possible that this is an artifact of different measurement techniques that should be investigated in future research. Given the conflicting findings in the literature, future research should investigate how symptom interconnectivity (vs. symptom severity) may contribute to course of illness and outcomes.

Limitations

This study examines ED symptoms across developmental stages in the largest clinical ED sample used for NA to date, providing important insight into how ED symptomology may change across development. However, this study has limitations. One limitation is the missing diagnostic information in the data sets, which prevented us from using diagnosis-matched samples for each de- velopmental stage. Recovery record only provides participant di- agnostic information when the application is connected with a clinician, which was only applicable for 27.0 –48.7% of the EPSI participants and 1.4 –9.7% of the EDE-Q participants. Among the participants that did have ED diagnoses, there were significant differences in diagnoses across developmental stages. For exam- ple, in the EPSI network, the early adolescence group was primar- ily comprised of anorexia nervosa (55.3% of individuals with a clinician-provided diagnosis), and the middle-late adulthood group was primarily comprised of binge eating disorder (52.6% of indi- viduals with a clinician-provided diagnosis). Due to these differ- ences, it is possible that some of the network differences we found may be attributed to diagnostic differences as opposed to devel- opmental stage. Future research should use diagnosis-matched samples to test if our findings replicate. However, despite differ- ences in diagnoses across networks, many symptoms remained central across all networks. This finding supports the idea that despite differential diagnoses, EDs are transdiagnostic phenomena (Cooper & Dalle Grave, 2017; Lampard, Tasca, Balfour, & Bis- sada, 2013). Additionally, the ubiquity of core symptoms across diagnoses could contribute to the high diagnostic crossover in EDs (Castellini et al., 2011; Fichter & Quadflieg, 2007).

Further, because of the low prevalence of individuals above 46 that used the Recovery Record application, the middle-late adults network spans several decades (46 –79 years of age). Thus, this study is unable to contribute to parsing out ED symptom differ- ences across this large developmental category. Additional re- search should be conducted in middle and older adults, focused on identifying developmental differences in ED symptoms. Further, given that this is the first investigation of EDs across development from early adolescence to late adulthood, there are no established guidelines for distinct developmental periods in this population. Our categories are based on non-ED-specific developmental the- ories. Future research may refine these periods to ensure they reflect distinct stages of development for this population. Addi- tionally, data were self-reported from the Recovery Record app and limited by self-awareness and self-report biases. Two sub-

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

187EATING DISORDER AGE NETWORKS

scales of the EPSI, Muscle Building and Negative Attitudes To- ward Obesity, were not measured in the Recovery Record app, so it is unknown how these constructs might vary across develop- mental stages.

One primary concern with NA is that there is currently no empirical method for selecting items for inclusion. As depicted by differences in central symptoms and connections across the EPSI and EDE-Q, item inclusion can critically impact interpretation of the network. For example, the EPSI, comprised of more behavioral ED symptoms, had more behavioral symptoms emerge as central, as compared to the EDE-Q. Future research should develop and validate empirical methods of selecting items for a network and developing measures designed to perform well in NA. Researchers have also expressed concerns with sole reliance on centrality indices to determine central symptoms (see Bringmann & Eronen, 2018; Hallquist, Wright, & Molenaar, 2019). However, in general, many researchers have suggested that central symptoms may serve as useful targets for future interventions (McNally, 2016; Rode- baugh et al., 2018), and growing empirical data shows that central symptoms predict important outcomes, specifically in EDs (Elliott et al., 2018; Olatunji et al., 2018). Finally, these networks were conducted at the group level, so findings indicate trends across developmental stages and may not be representative of symptom relationships for an individual over time.

Implications and Future Research

This study examines ED symptoms across developmental stages in a large clinical ED sample, which has broad implications for future research and treatment development for individuals with EDs. Significant network differences across stages suggest that ED research should be inclusive of individuals from all ages, espe- cially older populations, who are typically left out of studies on treatment development (Forman & Davis, 2005). Additionally, differences across stages of development may impact treatment needs for subpopulations of EDs. For example, as symptom con- nections change in older populations, treatments may need to be adapted to focus on the strongest connections in order to disrupt the most salient illness pathways. Treatments for older individuals with EDs must also take into consideration the increased connec- tivity of symptoms, which may be contributing to the worse treatment outcomes for this population (Noordenbos et al., 2002; Norring & Sohlberg, 1993).

In addition, symptoms that are central to eating pathology across developmental stages, including items related to overeating, feel- ings of fullness, food avoidance, and overvaluation of weight and shape, are hypothesized to be good targets for intervention for individuals of all ages with EDs. Interventions that target these symptoms, including cognitive– behavioral and dialectical– behavior therapy interventions, such as thought challenging, ex- posure therapy, distress tolerance, and behavior chaining, are widely used and are among the most effective and empirically supported treatments for EDs (Fairburn, 2008; Linehan & Chen, 2005). Feelings of fullness can also be addressed using interocep- tive exposures, which little research has investigated in EDs (Boettcher, Brake, & Barlow, 2016). Central symptoms that are unique to specific developmental stages may also be suggested targets for treatment for individuals with EDs that fall within that stage. However, it should be noted that group-level trends across

development might not be reflective of the most important treat- ment targets for an individual. We hope that future research will explore similar questions within-persons. Overall, this study uti- lizes an emerging statistical approach to explore ED symptom differences across the life span, which future research will need to continue to address in order to develop more effective interven- tions for individuals of all ages who struggle with an ED.

References

Aardoom, J. J., Dingemans, A. E., Slof Op’t Landt, M. C., & Van Furth, E. F. (2012). Norms and discriminative validity of the Eating Disorder Examination Questionnaire (EDE-Q). Eating Behaviors, 13, 305–309. http://dx.doi.org/10.1016/j.eatbeh.2012.09.002

Blonigen, D. M., Carlson, M. D., Hicks, B. M., Krueger, R. F., & Iacono, W. G. (2008). Stability and change in personality traits from late adolescence to early adulthood: A longitudinal twin study. Journal of Personality, 76, 229 –266. http://dx.doi.org/10.1111/j.1467-6494.2007 .00485.x

Boettcher, H., Brake, C. A., & Barlow, D. H. (2016). Origins and outlook of interoceptive exposure. Journal of Behavior Therapy and Experimen- tal Psychiatry, 53, 41–51. http://dx.doi.org/10.1016/j.jbtep.2015.10.009

Borsboom, D. (2017). A network theory of mental disorders. World Psy- chiatry, 16, 5–13. http://dx.doi.org/10.1002/wps.20375

Bos, F. M., Fried, E. I., Hollon, S. D., Bringmann, L. F., Dimidjian, S., DeRubeis, R. J., & Bockting, C. L. H. (2018). Cross-sectional networks of depressive symptoms before and after antidepressant medication treatment. Social Psychiatry and Psychiatric Epidemiology, 53, 617– 627. http://dx.doi.org/10.1007/s00127-018-1506-1

Bringmann, L. F., & Eronen, M. I. (2018). Don’t blame the model: Reconsidering the network approach to psychopathology. Psychological Review, 125, 606 – 615. http://dx.doi.org/10.1037/rev0000108

Castellini, G., Lo Sauro, C., Mannucci, E., Ravaldi, C., Rotella, C. M., Faravelli, C., & Ricca, V. (2011). Diagnostic crossover and outcome predictors in eating disorders according to DSM–IV and DSM-V pro- posed criteria: A 6-year follow-up study. Psychosomatic Medicine, 73, 270 –279. http://dx.doi.org/10.1097/PSY.0b013e31820a1838

Cooper, Z., & Dalle Grave, R. (2017). Eating disorders: Transdiagnostic theory and treatment. In S. G. Hofmann & G. J. G. Asmundson (Eds.), The Science of Cognitive Behavioral Therapy (pp. 337–357). http://dx .doi.org/10.1016/B978-0-12-803457-6.00014-3

Cooper, Z., & Shafran, R. (2008). Cognitive behaviour therapy for eating disorders. Behavioural and Cognitive Psychotherapy, 36, 713–722. http://dx.doi.org/10.1017/S1352465808004736

Crow, S. J., Peterson, C. B., Swanson, S. A., Raymond, N. C., Specker, S., Eckert, E. D., & Mitchell, J. E. (2009). Increased mortality in bulimia nervosa and other eating disorders. The American Journal of Psychiatry, 166, 1342–1346. http://dx.doi.org/10.1176/appi.ajp.2009.09020247

DuBois, R. H., Rodgers, R. F., Franko, D. L., Eddy, K. T., & Thomas, J. J. (2017). A network analysis investigation of the cognitive-behavioral theory of eating disorders. Behaviour Research and Therapy, 97, 213– 221. http://dx.doi.org/10.1016/j.brat.2017.08.004

Elliott, H., Jones, P. J., & Schmidt, U. (2018). Central symptoms predict post-treatment outcomes and clinical impairment in anorexia nervosa: A network analysis in a randomized-controlled trial. http://dx.doi.org/10 .31234/OSF.IO/HW2DZ

Epskamp, S., Borsboom, D., & Fried, E. I. (2018). Estimating psycholog- ical networks and their accuracy: A tutorial paper. Behavior Research Methods, 50, 195–212. http://dx.doi.org/10.3758/s13428-017-0862-1

Epskamp, S., Cramer, A. O., Waldorp, L. J., Schmittmann, V. D., & Borsboom, D. (2012). qgraph: Network visualizations of relationships in psychometric data. Journal of Statistical Software, 48, 1–18.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

188 CHRISTIAN ET AL.

Epskamp, S., & Fried, E. I. (2018). A tutorial on regularized partial correlation networks. Psychological Methods, 23, 617– 634. http://dx.doi .org/10.1037/met0000167

Epskamp, S., Maris, G., Waldorp, L. J., & Borsboom, D. (2018). Network Psychometrics. In P. Irwing, T. Booth, & D. J. Hughes (Eds.), The Wiley handbook of psychometric testing (pp. 953–986). http://dx.doi.org/10 .1002/9781118489772.ch30

Fairburn, C. G. (2008). Cognitive behavior therapy and eating disorders. New York, NY: Guilford Press.

Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders: Interview or self-report questionnaire? International Journal of Eating Disorders, 16, 363–370.

Fairburn, C. G., Wilson, G. T., & Schleimer, K. (1993). Binge eating: Nature, assessment, and treatment. New York, NY: Guilford Press.

Fichter, M. M., & Quadflieg, N. (2007). Long-term stability of eating disorder diagnoses. International Journal of Eating Disorders, 40, S61– S66. http://dx.doi.org/10.1002/eat.20443

Forbush, K. T., Siew, C. S. Q., & Vitevitch, M. S. (2016). Application of network analysis to identify interactive systems of eating disorder psy- chopathology. Psychological Medicine, 46, 2667–2677. http://dx.doi .org/10.1017/S003329171600012X

Forbush, K. T., Wildes, J. E., Pollack, L. O., Dunbar, D., Luo, J., Patterson, K., . . . Watson, D. (2013). Development and validation of the Eating Pathology Symptoms Inventory (EPSI). Psychological Assessment, 25, 859 – 878. http://dx.doi.org/10.1037/a0032639

Forman, M., & Davis, W. N. (2005). Characteristics of middle-aged women in inpatient treatment for eating disorders. Eating Disorders: The Journal of Treatment & Prevention, 13, 231–243. http://dx.doi.org/10 .1080/10640260590932841

Forrest, L. N., Jones, P. J., Ortiz, S. N., & Smith, A. R. (2018). Core psychopathology in anorexia nervosa and bulimia nervosa: A network analysis. International Journal of Eating Disorders, 51, 668 – 679. http:// dx.doi.org/10.1002/eat.22871

Fried, E. I., & Cramer, A. O. J. (2017). Moving forward: Challenges and directions for psychopathological network theory and methodology. Perspectives on Psychological Science, 12, 999 –1020. http://dx.doi.org/ 10.1177/1745691617705892

Fulton, C. L. (2016). Disordered eating across the lifespan of women [PDF]. Retrieved from https://www.counseling.org/docs/default-source/ vistas/disordered-eating.pdf?sfvrsn�769e4a2c_4

Gadalla, T. M. (2008). Eating disorders and associated psychiatric comor- bidity in elderly Canadian women. Archives of Women’s Mental Health, 11, 357–362. http://dx.doi.org/10.1007/s00737-008-0031-8

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 sam- ple. Journal of Abnormal Psychology, 127, 251–264. http://dx.doi.org/ 10.1037/abn0000327

Hallquist, M., Wright, A. G. C., & Molenaar, P. C. M. (2019). Problems with centrality measures in psychopathology symptom networks: Why network psychometrics cannot escape psychometric theory. Multivariate Behavioral Research. Advance online publication. http://dx.doi.org/10 .1080/00273171.2019.1640103

Hudson, J. I., & Pope, H. G., Jr. (2018). Evolving perspectives on the public health burden of eating disorders. Biological Psychiatry, 84, 318 –319. http://dx.doi.org/10.1016/j.biopsych.2018.06.011

Jenkins, P. E., & Price, T. (2018). Eating pathology in midlife women: Similar or different to younger counterparts? International Journal of Eating Disorders, 51, 3–9. http://dx.doi.org/10.1002/eat.22810

Lampard, A. M., Tasca, G. A., Balfour, L., & Bissada, H. (2013). An evaluation of the transdiagnostic cognitive-behavioural model of eating disorders. European Eating Disorders Review, 21, 99 –107. http://dx.doi .org/10.1002/erv.2214

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, 67– 82. http://dx.doi.org/10.1007/s11920-018-0930-y

Levinson, C. A., Zerwas, S., Calebs, B., Forbush, K., Kordy, H., Watson, H., . . . Bulik, C. M. (2017). The core symptoms of bulimia nervosa, anxiety, and depression: A network analysis. Journal of Abnormal Psychology, 126, 340 –354. http://dx.doi.org/10.1037/abn0000254

Linehan, M. M., & Chen, E. Y. (2005). Dialectical behavior therapy for eating disorders. In A. Freeman, S. H. Felgoise, C. M. Nezu, A. M. Nezu, M. A. Reinecke (Eds.), Encyclopedia of Cognitive Behavior Therapy (pp. 168 –171). http://dx.doi.org/10.1007/0-306-48581-8_50

Luce, K. H., & Crowther, J. H. (1999). The reliability of the Eating Disorder Examination—Self-Report Questionnaire version (EDE-Q). International Journal of Eating Disorders, 25, 349 –351. http://dx.doi .org/10.1002/(SICI)1098-108X(199904)25:3�349::AID-EAT15�3.0 .CO;2-M

Mangweth-Matzek, B., Rupp, C. I., Hausmann, A., Assmayr, K., Mari- acher, E., Kemmler, G., . . . Biebl, W. (2006). Never too old for eating disorders or body dissatisfaction: A community study of elderly women. International Journal of Eating Disorders, 39, 583–586. http://dx.doi .org/10.1002/eat.20327

Marcus, M. D., Bromberger, J. T., Wei, H.-L., Brown, C., & Kravitz, H. M. (2007). Prevalence and selected correlates of eating disorder symptoms among a multiethnic community sample of midlife women. Annals of Behavioral Medicine: A Publication of the Society of Behavioral Med- icine, 33, 269 –277. http://dx.doi.org/10.1007/BF02879909

McNally, R. J. (2016). Can network analysis transform psychopathology? Behaviour Research and Therapy, 86, 95–104. http://dx.doi.org/10 .1016/j.brat.2016.06.006

Mond, J. M., Hay, P. J., Rodgers, B., Owen, C., & Beumont, P. J. (2004). Validity of the Eating Disorder Examination Questionnaire (EDE-Q) in screening for eating disorders in community samples. Behaviour Re- search and Therapy, 42, 551–567. http://dx.doi.org/10.1016/S0005- 7967(03)00161-X

Noordenbos, G., Oldenhave, A., Muschter, J., & Terpstra, N. (2002). Characteristics and treatment of patients with chronic eating disorders. Eating Disorders: The Journal of Treatment & Prevention, 10, 15–29. http://dx.doi.org/10.1080/106402602753573531

Norring, C. E. A., & Sohlberg, S. S. (1993). Outcome, recovery, relapse and mortality across six years in patients with clinical eating disorders. Acta Psychiatrica Scandinavica, 87, 437– 444. http://dx.doi.org/10 .1111/j.1600-0447.1993.tb03401.x

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. http://dx.doi.org/10.1016/j .psychres.2018.02.027

Patrick, J. H., & Stahl, S. T. (2009). Understanding disordered eating at midlife and late life. The Journal of General Psychology, 136, 5–20. http://dx.doi.org/10.3200/GENP.136.1.5-20

Pe, M. L., Kircanski, K., Thompson, R. J., Bringmann, L. F., Tuerlinckx, F., Mestdagh, M., . . . Gotlib, I. H. (2015). Emotion-network density in major depressive disorder. Clinical Psychological Science, 3, 292–300. http://dx.doi.org/10.1177/2167702614540645

Pearson, C. M., Wonderlich, S. A., & Smith, G. T. (2015). A risk and maintenance model for bulimia nervosa: From impulsive action to compulsive behavior. Psychological Review, 122, 516 –535. http://dx .doi.org/10.1037/a0039268

Peat, C. M., Peyerl, N. L., & Muehlenkamp, J. J. (2008). Body image and eating disorders in older adults: A review. The Journal of General Psychology, 135, 343–358. http://dx.doi.org/10.3200/GENP.135.4.343- 358

Rodebaugh, T. L., Tonge, N. A., Piccirillo, M. L., Fried, E., Horenstein, A., Morrison, A. S., . . . Heimberg, R. G. (2018). Does centrality in a

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

189EATING DISORDER AGE NETWORKS

cross-sectional network suggest intervention targets for social anxiety disorder? Journal of Consulting and Clinical Psychology, 86, 831– 844. http://dx.doi.org/10.1037/ccp0000336

Rome, E. S., & Ammerman, S. (2003). Medical complications of eating disorders: An update. Journal of Adolescent Health, 33, 418 – 426. http://dx.doi.org/10.1016/S1054-139X(03)00265-9

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, 680 – 692. http://dx.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, 24 –33. http://dx.doi.org/10.1002/erv.2633

Steinberg, L. (2005). Cognitive and affective development in adolescence. Trends in Cognitive Sciences, 9, 69 –74. http://dx.doi.org/10.1016/j.tics .2004.12.005

Stice, E., Marti, C. N., & Rohde, P. (2013). Prevalence, incidence, impair- ment, and course of the proposed DSM–5 eating disorder diagnoses in an 8-year prospective community study of young women. Journal of Ab- normal Psychology, 122, 445– 457. http://dx.doi.org/10.1037/a0030679

Tiggemann, M., & McCourt, A. (2013). Body appreciation in adult wom- en: Relationships with age and body satisfaction. Body Image, 10, 624 – 627. http://dx.doi.org/10.1016/j.bodyim.2013.07.003

Tregarthen, J. P., Lock, J., & Darcy, A. M. (2015). Development of a smartphone application for eating disorder self-monitoring. Interna-

tional Journal of Eating Disorders, 48, 972–982. http://dx.doi.org/10 .1002/eat.22386

Tzoneva, M., Forney, K. J., & Keel, P. K. (2015). The influence of gender and age on the association between “fat-talk” and disordered eating: An examination in men and women from their 20s to their 50s. Eating Disorders: The Journal of Treatment & Prevention, 23, 439 – 454. http://dx.doi.org/10.1080/10640266.2015.1013396

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 depression. Journal of the American Medical Association Psychiatry, 72, 1219 –1226. http://dx.doi.org/10 .1001/jamapsychiatry.2015.2079

Walsh, B. T. (2013). The enigmatic persistence of anorexia nervosa. The American Journal of Psychiatry, 170, 477– 484. http://dx.doi.org/10 .1176/appi.ajp.2012.12081074

Wang, S. B., Jones, P. J., Dreier, M., Elliott, H., & Grilo, C. M. (2018). Core psychopathology of treatment-seeking patients with binge-eating disorder: A network analysis investigation. Psychological Medicine, 49, 1923–1928. http://dx.doi.org/10.1017/S0033291718002702

Williams, J. M., & Currie, C. (2000). Self-Esteem and Physical Develop- ment in Early Adolescence. The Journal of Early Adolescence, 20, 129 –149. http://dx.doi.org/10.1177/0272431600020002002

Received February 11, 2019 Revision received August 16, 2019

Accepted August 19, 2019 �

E-Mail Notification of Your Latest Issue Online!

Would you like to know when the next issue of your favorite APA journal will be available online? This service is now available to you. Sign up at https://my.apa.org/portal/alerts/ and you will be notified by e-mail when issues of interest to you become available!

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

190 CHRISTIAN ET AL.