search 64 (2008) 87–95

Journal of Psychosomatic Re

Predicting changes in eating disorder symptoms among Chineseadolescents: A 9-month prospective study

Todd Jacksona,b, Hong Chena,c,⁎

aKey Laboratory of Cognition and Personality, Southwest University, Chongqing, ChinabSchool of Psychology, James Cook University, Townsville, Australia

cSchool of Psychology, Southwest University, Chongqing, China

Received 13 March 2007; received in revised form 16 July 2007; accepted 28 August 2007

Abstract

Objectives: Body image and eating disturbances have becomeglobal phenomena, yet prospective designs have rarely beenemployed in research on non-Western samples. This study testedthe extent to which select features of the dual-pathway accountof bulimic disturbances contributed to changes in eating disordersymptoms reported among adolescents from China. Methods: Asample of 593 Chinese middle school and high school students(217 boys, 376 girls) completed measures of eating disordersymptoms, body image concerns, internalized physical appear-ance ideals, negative affect, and appearance-based socialpressure, teasing, and comparison and returned 9 months later

⁎ Corresponding author. Key Laboratory of Cognition and Personality(Southwest University), Ministry of Education, China, Tiansheng RoadNo. 1, BeiBei, Chongqing, P.R. China 400715.

E-mail address: chenhg@swu.edu.cn (H. Chen).

0022-3999/08/$ – see front matter © 2008 Elsevier Inc. All rights reserved.doi:10.1016/j.jpsychores.2007.08.015

to complete the same measures. Results: For both girls andboys, increases in eating disorder symptoms between Times 1and 2 were predicted by higher baseline levels of fatness concernand perceived social pressure. Among the girls, negative affectalso contributed marginally to changes in eating disordersymptoms. Conclusions: Findings suggest specific risk factorsincluding personal concerns about being fat and negative socialfeedback about physical appearance, may help to explainchanges in eating disturbances of adolescents over time andacross specific cultures.© 2008 Elsevier Inc. All rights reserved.

Keywords: Disordered eating; Sociocultural; Body dissatisfaction; China; Adolescents; Gender

Introduction

Eating disorders, including anorexia nervosa, bulimianervosa (BN), and their subclinical variants, are prevalent,serious problems. Epidemiological studies from the UnitedStates indicate that weight dissatisfaction and eatingdisturbances are widespread [1,2], albeit these experiencesare no longer restricted to Western samples. Weightpreoccupation, disordered eating attitudes, and body dis-satisfaction have also been documented among adolescentsfrom developed regions of Asia such as Hong Kong [3–7]and rapidly developing countries such as China [8–10]. Forexample, Li et al. [8] observed that less than half of normal-

weight Chinese boys and girls were satisfied with theircurrent body weight. Chen and Jackson [9] found thatendorsements of compensatory behavior to control weightwere common for both males and females in a large sampleof Chinese adolescents and young adults; 15 girls (1.1%) inthis sample endorsed all or partial criteria for BN and ninegirls (0.7%) endorsed all criteria for binge eating disorder.

Because eating disorders pose significant risks formortality [11], physical health [12], and psychologicaladjustment [13–15], identifying factors that increase riskmay facilitate prevention and intervention. Unfortunately,the preponderance of past research has been cross-sectionaland cannot elucidate changes in disordered eating behaviorover time. Although well-designed prospective studies haveemerged in recent years [16–22], several issues have notbeen addressed in existing research.

Most notably, longitudinal research has been limited tosamples from Western countries, despite abundant evidence

88 T. Jackson, H. Chen / Journal of Psychosomatic Research 64 (2008) 87–95

that eating pathology is common across cultures. As such,little is known about factors that contribute to changes ineating disturbances in non-Western samples. Given theimportance of fitting in with peers during adolescence andthe comparatively strong orientation toward group member-ship as a source of identity in collectivistic cultures [23,24],it is reasonable to speculate that factors such as preferredphysical attractiveness ideals and appearance-related socialpressure and social comparison also contribute to eatingpathology in countries such as China.

Second, if body dissatisfaction is among the most reliablerisk factors for increases in disordered eating in Westernsamples [25–30], this contention has yet to be testedprospectively in non-Western groups. Furthermore, somecases of disordered eating in Chinese samples may arise fromappearance concerns that are not specific to “fat phobia” [31].For example, concerns about facial appearance have beenidentified as possible influences on eating pathology in selectstudies of Chinese adolescents [32,33]. If such findings areintriguing, it is not clear whether alternate appearanceconcerns contribute to changes in eating pathology over time.

Third, due to sex differences in prevalence [34],prospective research on young females was likely an earlypriority. More recently, muscularity has been identified as asource of body image concern common for young Westernmales [35,36], and males have been included in prospectiveresearch on eating disturbances [17–21]. Nonetheless, thereis a lag in knowledge about eating disorder risks amongmales, especially those from non-Western countries.

Finally, prospective tests of overlapping theoreticalaccounts have been largely piecemeal and have not assessedall salient theoretical constructs. For example, the dual-pathway model [37] posits, in part, that pressures to bethinner from family, peers, and media contribute to bodydissatisfaction and disordered eating because repeatedmessages that one is not thin enough promote disgruntlementwith one's appearance. Appearance-based comparisons withothers and internalizations of thin body ideals, at least amongwomen, also contribute to body dissatisfaction because suchideals are not attainable for most. In theory, bodydissatisfaction fosters dietary restraint and negative affect,which can subsequently increase risk for eating disordersymptoms. Changes in eating and body image disturbanceshave relatively consistent associations with features of themodel including body dissatisfaction [25–29], socialpressure [18,19,27,38], and internalized attractiveness ideals[18,21,27,38]. However, evidence is not as compelling forfactors such as dietary restraint [26,27] and has beeninconsistent for social pressure in the form of teasing aboutweight or appearance [16,20,22,29,30] and for negativeaffect [38–40]. Finally, dual-pathway features such associal comparison have been examined rarely in long-itudinal work [41].

In cross-sectional research on Chinese samples, disor-dered eating has been linked with fatness and facialappearance concerns, social pressure, social comparisons

and teasing based on physical appearance, and, to a lesserextent, negative affect [33]. That said, it is not clear whichfeatures of the dual-pathway model are relevant to explainingchanges in eating disturbances in a Chinese cultural context.

To address these issues, we examined, through this9-month prospective study, predictors of eating disturbancesamong adolescent girls and boys from China. Based onfairly consistent longitudinal evidence in Western samplesand cross-sectional research on Chinese samples, fatnessconcern and perceived social pressure about appearancewere expected to contribute to changes in eating pathologyin prediction models for each gender. We also explored thedegree to which preferred appearance ideals, socialcomparison, and negative affect were related to changes ineating disorder symptoms. Finally, in light of theory [31]and preliminary evidence [32,33] that appearance preoccu-pations not directly reflecting weight predict eatingdisturbances in Chinese adolescents, we assessed whetherdissatisfaction with facial appearance might contribute toeating disorder symptoms, independent of concerns withfatness/being overweight.

Method

Participants

From an initial sample of 855 adolescents who completedthe research measures at Time 1 (T1), 69.1% (217 boys,376 girls) also completed the Time 2 (T2) assessment.Respondents were recruited from middle schools and highschools from the Chinese cities of Chongqing, Guiyang, andNeijiang. Participants were between 12 and 19 years of age(mean=15.25 years, S.D.=1.63). Most participants were ofHan (82.8%), Buyi (8.2%), or Miao (5.7%) ethnicity, and theremaining participants were from eight other ethnic minorities.

In analyses comparing participants who finished bothphases of the study with those who completed only theinitial assessment, there were no differences for gender[χ2(1)=0.08, Pb.78], ethnicity [χ2(13)=16.03, Pb.25],parental education [χ2(3)=0.99, Pb.60], estimated house-hold income [χ2(3)=2.65, Pb.45], or rural versus urbanresidence [χ2(1)=0.09, Pb.77]. On interval-level measures(age, BMI, eating disorder symptoms, psychologicaland sociocultural measures), neither the multivariate effect[F(15, 827)=0.98, Pb.47] nor any of the univariate F values(all P valuesb .138) were significant.

Measures

Back translation and assessment of equivalenceAll scales had been used with Chinese samples in

previous published research [33,42]. English-languagemeasures had been translated into Chinese by the secondauthor and back-translated into English by a faculty memberin the English Department at Southwest University inChongqing, China. Because most scales were developed

89T. Jackson, H. Chen / Journal of Psychosomatic Research 64 (2008) 87–95

for use in Western female samples, we followed Jones' [41]procedure of subjecting items from each scale given at T1 toprincipal components analyses with varimax rotation toevaluate underlying factor structures. These analyses wereperformed separately for boys and girls.

Eating Disorder Diagnosis Scale (EDDS) [43]The EDDS is a 22-item self-report screening measure that

corresponds with DSM-IV eating disorder diagnostic criteriaand has demonstrated psychometric properties. Stice et al.[43,44] created an eating disorder symptom composite bysumming standardized scores of 20 EDDS items (excludingitems pertaining to height and use of birth control pills); thesymptom composite has high internal consistency, satisfac-tory test–retest reliability, and excellent concordance withvalidated measures of eating disturbances (i.e., dietaryrestraint, hunger, disinhibited eating, and weight/shapeconcerns and rituals), eating disorder risk factors, and socialimpairment [43,44]. In factor analyses of standardizedEDDS items for Chinese boys and girls, all items loadedon a single dimension except for actual body weight.Consequently, this item was omitted in calculating totalcomposite scores for each gender. In addition, the gender-specific item pertaining to missed menstrual cycles wasomitted from the symptom composite calculated for boys.For girls, EDDS composite alpha values of .78 and .81 wereobtained for T1 and T2, respectively. For boys, alphacoefficients for the composite were .72 for T1 and .86 for T2.

Negative Physical Self Scale (NPS) [45]The NPS is a 38-item measure of body image concerns

developed specifically for use in Chinese samples. Themeasure has subscales that assess specific physical appearanceconcerns related to Fatness, Stature, and Facial Appearance, aswell as a General Appearance Concern dimension. Each itemis rated on a 5-point scale of agreement. Chen et al. [45] foundthat each NPS dimension has a stable factor structure, isinternally consistent, and has satisfactory validity.

In this research, the 11-item Fatness Concerns (FC)subscale examined thoughts and feelings, projections, andbehaviors related to the self as fat. Sample items include “Ihave tried many ways to lose weight,” “I think I am fat inothers' eyes,” and “When I weigh myself, I feel depressed.”The 11-item Facial Appearance Concerns (FAC) subscale wasalso employed. FAC items reflect thoughts and feelings (e.g.,“I am depressed about how my face looks”), projections (e.g.,“People around me don't like the way my face looks”), andbehavior (e.g., “If it is possible, I will change the way my facelooks”). In this study, original factor structures for FC and FACwere replicated for both girls and boys. Alpha valueswere highfor girls (FC: T1=.88, T2=.87; FAC: T1=.89, T2=.89) and boys(FC: T1=.85, T2=.84; FAC: T1=.90, T2=.89).

Perceived Sociocultural Pressure Scale (PSPS) [37]The eight-item PSPS assesses pressure from friends,

media, dating partner(s), and family to change physical

appearance. Items were rated between 1 (none) and 5 (a lot).The PSPS has satisfactory psychometric properties. In thisresearch, all items loaded on one dimension for both girlsand boys. PSPB alpha values were high for girls (T1=.89,T2=.88) and boys (T1=.88, T2=.92).

Perception of Teasing Scale (POTS) [46]The six-item Weight Teasing frequency subscale of the

POTS assesses the occurrence of teasing about weight (e.g.,“How often have people made jokes about your weight?”) ona 5-point scale. Based on recent work [33], items weremodified slightly to reflect teasing about nonspecific aspectsof appearance (e.g., “How often have people made jokesabout some aspect of your physical appearance?”). In factoranalyses for each gender, all items loaded on a singledimension. Reliabilities were adequate for girls (T1=.83,T2=.83) and boys (T1=.79, T2=.84).

Physical Appearance Comparison Scale (PACS) [47]The PACS has five items that assess the tendency to

compare one's own appearance with that of others (e.g., “Atparties or other social events, I compare my physicalappearance to the physical appearance of others”). Responseoptions ranged from 1 (never) to 5 (always). Thompson et al.[46] found that the PACS had adequate psychometricproperties. Factor analyses on the PACS resulted in identicalsingle-factor structures for Chinese girls and boys. Con-sistent with recent American data [48], Item 4 failed to loadwith other PACS items and was dropped from analyses. Thefour-item PACS had satisfactory alpha values for girls(T1=.85, T2=.88) and boys (T1=.79, T2=.88).

Preferred appearance ideals [49]The Ideal Body Stereotype Scale-Revised assesses

preferences for a thinness ideal for females. The scaleincludes 10 items requiring participants to rate the degree towhich they agreed specific physical features (e.g., thin, longlegs, shapely) characterize attractive women. We created acomplementary 10-item scale to reflect stereotypicalmasculine physical appearance ideals (e.g., tall, muscular,broad shoulders, flat stomach). Participants rated their levelof agreement that such features characterized physicallyattractive men. For each scale, unidimensional factorstructures composed of all relevant items were obtainedfor each gender. Among girls, alpha values for the feminine(T1=.83, T2=.84) and masculine ideals (T1=.82, T2=.82)were adequate, as were those for feminine (T1=.85,T2=.81) and masculine ideals (T1=.85, T2=.86) amongboys. Although each appearance ideal was rated by bothgirls and boys at T1 and T2, only the gender-congruentdimension was included in the prediction model withineach gender.

Positive and Negative Affect Scale [50]This 20-item measure taps positive and negative affect

reported during a particular time interval. Participants

Table 1Descriptive statistics for girls and boys on research measures (n=593)

Measure

Girls Boys

Mean S.D. Mean S.D.

Time 1BMI 18.89 2.27 18.34 2.49ED symptom composite 1.35 9.12 −2.36 6.75Fatness concerns 12.21 8.51 5.33 6.36Facial appearance concerns 11.47 7.98 9.36 8.38Social comparison 8.87 3.52 8.53 3.37Social pressure 7.74 6.09 6.45 6.40Appearance teasing 8.99 3.59 8.09 2.76Internalized appearance ideal (female) 37.46 6.10 37.24 6.74Internalized appearance ideal (male) 33.25 6.75 37.87 6.79Negative affect 16.98 4.27 15.92 4.19Time 2BMI 19.51 2.36 18.89 2.52ED symptom composite 1.00 9.24 −1.56 10.11Fatness concerns 13.98 8.40 6.69 8.06Facial appearance concerns 12.67 7.69 11.50 8.89Social comparison 9.49 3.61 8.94 3.89Social pressure 8.35 6.11 7.18 6.92Appearance teasing 10.72 4.07 9.58 4.01Internalized appearance ideal (female) 38.41 6.05 38.63 6.44Internalized appearance ideal (male) 34.35 6.34 38.22 6.91Negative affect 18.42 4.24 17.42 5.18

90 T. Jackson, H. Chen / Journal of Psychosomatic Research 64 (2008) 87–95

rated how often they had 20 specific affective experiencesduring the last month. Items were rated on a 4-pointLikert scale from 1 (none or little of the time) to 4 (mostof the time). In this study, only negative affect was ofinterest. Despite the emergence of factors reflectingpositive and negative affect within each gender, therewas one deviation from the conceptualization of Watsonet al. [50]. “Alert” loaded with the 10 items reflectingnegative affect for boys and had a double loading forgirls, with a more pronounced association on negativeaffect. Consequently, the item was added with the 10negative affect items identified by Watson et al. incalculating negative affect for both genders. Alpha valuesfor negative affect were high for girls (T1=.83, T2=.84)and boys (T1=.82, T2=.82).

DemographicsAge, gender, ethnicity, and region of residence (rural vs.

urban) were solicited. As well, because self-reportedweight correlates very highly with confederate-measuredweight [51] and is less intrusive to assess, BMI wascalculated (kg/m2) from students' reports. Finally, partici-pants provided information about parents' educationallevels and estimated monthly household income. Specifi-cally, educational level of each parent was measured as“less than high school completion,” “high school gradu-ate,” or “more than high school graduate” (e.g., college oruniversity degree). Income was also based on threeoptions: “less than 1000 yuan per month,” “1000–3000yuan per month,” or “more than 3000 yuan per month”(US$1.00 is worth about 7.5 yuan).

Procedure

After receiving ethics approval from Southwest Uni-versity, the second author contacted colleagues from highschools and middle schools in Chongqing, Guiyang, andNeijiang, seeking permission to conduct a study on bodyimage issues. All contacted settings granted permission.Subsequently, one class from each setting was randomlyselected for inclusion. Aside from institutional consent, aparental assent form was sent to parents on eachassessment occasion to prevent their children fromparticipating if or when there were objections. Childrenwhose parents assented were given a survey packet bytheir teacher. The packet was composed of a statement ofthe research purpose (to foster knowledge about bodyimage among students from China), an informed consent,and the research measures. These were the same at T1 andT2, with the exception of a truncated demographicssection on the latter assessment. Participants wereencouraged to read items carefully before responding.Measures were completed in class and returned to teachersseparately from informed consent forms. Data werecollected in September/October 2005 and May/June 2006for T1 and T2, respectively.

Results

Stability and mean levels of predictor variables

With the exception of slight variations in the EDDSsymptom composite, factor structures underlying measureswere the same for boys and girls. Descriptive statistics foreach gender are presented in Table 1. Paired sample t testson changes in the EDDS composite indicated that eatingdisorder symptoms did not change for either boys [t(217)=−1.20, Pb.23] or girls [t(376)=0.92, Pb.38]. A series of2×2 (Gender×Time) repeated-measures analyses of co-variance was conducted on other measures. Age servedas a covariate because boys (mean=14.97, S.D.=1.49)were younger than were girls [mean=15.41, S.D.=1.69;t(591)=3.15, Pb.002]. To control for possible inflation oftype I error, we set the Bonferroni-adjusted significancelevel at Pb.005. Gender differences were obtained for FC[F(1, 588)=127.09, Pb.001], teasing [F(1, 588)=16.11,Pb.001], negative affect [F(1, 588)=8.01, Pb.005], andmale appearance ideals [F(1, 588)=85.77, Pb.001]. A mar-ginal difference was also observed for FAC [F(1, 588)=7.06,Pb.008]. Girls scored higher than boys did on eachscale with the exception of male appearance ideals. Thesample also reported increases in FC [F(1, 588)=10.46,Pb.001], FAC [F(1, 588)=8.70, Pb.003], social pressure[F(1, 588)=21.66, Pb.001], and endorsements of a femaleappearance ideal [F(1, 588)=9.84, Pb.002]. None of theGender×Time interactions was significant.

Table 2Correlations among Time 1 and Time 2 measures for girls (n=376)

Measure 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Time 11. ED symptom composite –2. Fatness concerns .54 –3. Facial appearance concerns .36 .36 –4. Social comparison .30 .29 .47 –5. Social pressure .38 .38 .55 .43 –6. Appearance teasing .32 .34 .26 .15 .30 –7. Internalized appearance ideals .13 .15 .18 .27 .17 −.01 –8. Negative affect .34 .18 .28 .12 .19 .15 .01 –Time 29. ED symptom composite .53 .40 .33 .24 .30 .18 .17 .27 –10. Fatness concerns .43 .70 .23 .23 .25 .20 .19 .07 .52 –11. Facial appearance concerns .26 .19 .62 .36 .35 .19 .20 .16 .38 .35 –12. Social comparison .28 .21 .31 .59 .29 .07 .25 .11 .40 .33 .47 –13. Social pressure .23 .25 .35 .30 .38 .16 .12 .17 .42 .36 .50 .44 –14. Appearance teasing .23 .30 .28 .10 .21 .50 .03 .13 .32 .31 .38 .17 .30 –15. Internalized appearance ideals .11 .04 .13 .21 .13 .00 .38 −.02 .21 .19 .19 .25 .16 −.01 –16. Negative affect .13 .08 .20 .14 .09 .09 −.05 .37 .28 .14 .28 .31 .32 .21 .02 –

Note: r=.11, Pb.05; r=.14, Pb.01; r=.17, Pb.001. Italicized correlations represent test–retest reliabilities of measures.

91T. Jackson, H. Chen / Journal of Psychosomatic Research 64 (2008) 87–95

Prospective prediction of eating disturbances

Tables 2 and 3 present zero-order correlations amongmeasures for girls and boys, respectively. Cross-timecorrelations were moderate to high and suggested consider-able stability in measurement between T1 and T2. Patterns ofcorrelation were typically in the expected direction, butbivariate relations between the EDDS composite andinternalized appearance ideals and social comparison wereattenuated, especially for boys.

To evaluate the extent to which changes in the EDDScomposite were predicted by other measures, we performedtwo sets of analyses. First, after controlling for baseline (T1)composite scores, associations between hypothesized T1

Table 3Correlations among Time 1 and Time 2 measures for boys (n=217)

Measure 1 2 3 4 5 6

Time 11. ED symptom composite –2. Fatness concerns .40 –3. Facial appearance concerns .26 .47 –4. Social comparison .08 .19 .34 –5. Social pressure .33 .26 .56 .43 –6. Appearance teasing .26 .33 .30 .05 .20 –7. Internalized appearance ideals .15 .12 .01 .06 .03 .18. Negative affect .33 .21 .28 .19 .23 .2Time 29. ED symptom composite .50 .35 .19 .03 .19 .110. Fatness concerns .29 .48 .22 .06 .17 .1111. Facial appearance concerns .26 .25 .55 .23 .40 .112. Social comparison .12 .28 .39 .52 .34 .013. Social pressure .23 .27 .48 .35 .45 .214. Appearance teasing .13 .26 .33 .20 .31 .215. Internalized appearance ideals .03 −.03 −.09 .03 −.07 −.016. Negative affect .09 .22 .40 .24 .31 .1

Note: r=.14, Pb.05; r=.18, Pb.01; r=.22, Pb.001. Italicized correlations represe

predictors and T2 EDDS composite scores were examined inunivariate regression models, a strategy that permitted anestimate of relations independent of potential multicollinear-ity [41]. Second, significant univariate predictors wereloaded simultaneously in the second step of a multivariateregression model on changes in eating disorder symptoms.

GirlsIn initial analyses, nonsignificant associations were

observed between changes in eating disorder symptomsand several demographic factors (age, parental education,region of residence, BMI). However, eating disturbances hada positive association with reported monthly householdincome at T1 (t=2.47, Pb.01). Hence, income was entered

7 8 9 10 11 12 13 14 15 16

6 –0 .12 –

8 .07 .23 –−.03 .07 .61 –

3 .00 .19 .50 .54 –4 .15 .12 .19 .23 .47 –1 .06 .19 .29 .31 .53 .44 –3 .08 .17 .28 .31 .39 .30 .48 –2 .31 .02 .05 .05 .03 .02 −.06 −.04 –0 .00 .30 .29 .23 .38 .32 .39 .31 −.02 –

nt test–retest reliabilities of measures.

Table 5Time 1 predictors to change in EDDS symptom composite for boys (n=217)

Univariate regression models

Time 1 predictor

Change in ED symptom composite

β t P R2

Fatness concerns .19 2.95 .004 .029Facial appearance concerns .07 1.13 .26 .004Social comparison .00 0.14 .89 .00Social pressure .15 2.43 .02 .02Teasing .05 0.81 .42 .002Internalized appearance ideals .00 0.06 .95 .00Negative affect .09 1.53 .13 .008

Multivariate multiple regression model

Step Time 1 predictor

Change inED symptomcomposite

β t P

1 ED symptom composite .50 8.46 .001R2=.25, ΔF(1, 214)=71.63, Pb.0012 Fatness concerns .16 2.37 .02

Social pressure .13 2.01 .05ΔR2=.043, ΔF(2, 213)=6.45, Pb.002

92 T. Jackson, H. Chen / Journal of Psychosomatic Research 64 (2008) 87–95

with T1 EDDS composite scores in Step 1 of the multivariatemultiple regression model.

Subsequent univariate multiple regression analysesrevealed that higher levels of T1 FC, FAC, social pressure,social comparison, preferred appearance ideals, and negativeaffect predicted increases in EDDS composite scores at T2(Table 4). These factors were entered into the multivariateregression equation, controlling for T1 composite scores andestimated household income. Overall, adjusted R2=.34 of thevariance in T2 EDDS scores was explained by the eightpredictors [F(8, 360)=24.21, Pb.0001]. More notably,psychological and sociocultural measures had a significantimpact on changes in EDDS scores [R2=.058, ΔF(6, 360)=5.37, Pb.001]. Girls reporting higher baseline levels ofsocial pressure and concerns about fatness were more likelyto report increases in eating disorder symptoms at T2.Negative affect also had a marginal impact of changes ineating disorder symptoms (Table 4).

BoysChanges in EDDS scores were not related to any of the

demographic measures. FC and social pressure were theonly T1 measures to predict changes in eating disordersymptoms in univariate multiple regression analyses (Table5). In the multivariate model, the three predictorsaccounted for adjusted R2=.28 of the variance in T2

Table 4Time 1 predictors to change in EDDS symptom composite for girls (n=376)

Univariate multiple regression models

Time 1 predictor

Change in ED symptom composite

β t P R2

Fatness concerns .16 3.10 .002 .018Facial appearance concerns .16 3.48 .001 .023Social comparison .09 2.03 .04 .008Social pressure .22 4.53 .001 .032Teasing .01 0.21 .83 .00Internalized appearance ideal .10 2.31 .02 .01Negative affect .10 2.15 .03 .009

Multivariate multiple regression model

Step Time 1 predictor

Change inED symptomcomposite

β t P

1 ED symptom composite .52 11.78 .001Monthly householdincome

.10 2.20 .03

R2=.29, ΔF(2, 366)=4.60, Pb.00012 Fatness concerns .11 2.02 .04

Facial appearanceconcerns

.05 0.94 .35

Social comparison .02 0.42 .67Social pressure .16 2.55 .01Internalized appearanceideals

.07 1.48 .14

Negative affect .08 1.75 .08ΔR2=.058, ΔF(6, 360)=5.37, Pb.001

EDDS scores [F(3, 212)=29.39, Pb.0001]. Both FC andsocial pressure added to the model [R2=.04, ΔF(2, 212)=6.45, Pb.002], independent of baseline EDDS compositescores (Table 5).

Discussion

To our knowledge, this is the first longitudinal study toassess the effects of psychological and sociocultural factorson changes in eating disturbances in Mainland Chineseadolescents. Findings suggest that select features of the dual-pathway model may have utility in explaining changes ineating disorder symptoms among adolescents from certainnon-Western cultures.

Among the girls, high initial levels of concerns aboutfatness and about facial appearance, appearance-based socialpressure and social comparison, internalization of femaleattractiveness ideals, and negative affect were univariatepredictors of increased eating disturbances 9 months later.When the measures were entered together in a multivariateregression model, only concerns with fatness and perceivedsocial pressure from others about one's physical appearancecontributed to changes in eating disorder symptoms. Thesefindings are consistent with results of Western studiesindicating that girls who are highly concerned with being fat[25–30] and who experience pressure from others to altertheir physical appearance [18,19,27,38] are particularlyprone to later increases in eating disturbances. Perhapsstronger collectivist value orientations, which are character-ized by goals of the group, social harmony, willingnessto adopt others' views, and maintenance of face in theeyes of others [52], contributed to the relative strength of

93T. Jackson, H. Chen / Journal of Psychosomatic Research 64 (2008) 87–95

appearance-based social pressure in univariate and multi-variate prediction models for girls.

Although facial appearance concerns, social comparison,thin attractiveness ideals, and negative affect wereunivariate predictors of changes in eating pathology, theireffects were attenuated in the multivariate model, perhaps,in part, due to intercorrelations among predictors [20]. Atthis juncture, it is premature and overly reductionistic toconclude that such experiences are unimportant in under-standing eating disturbances in Chinese samples. That said,cultural factors may attenuate effects of certain predictors.For example, the impact of internalized thinness ideals mayhave been diminished, at least over 9 months, becauseattractiveness ideals for Chinese females focus on physicalfeatures such as the face as much as or more than a thinbody type [53].

Less expectedly, girls from households with higherestimated incomes were also more likely to show subsequentincreases in eating pathology. Lee and Lee [54] havepreviously linked higher socioeconomic status with bodydissatisfaction and eating pathology in Chinese samples.These authors noted that economic prosperity is accompa-nied by numerous other changes including modernization,urbanization, westernization, increased exposure to massmedia, and competing pressures on adolescent girls to strivefor sometimes incompatible goals of achieving a successfulcareer and finding a desirable marital partner. Even morerecently, Lee and Lock [55] found that Asian Americaneating disorder patients more often came from high-incomefamilies than their non-Asian American patients. Asianpatients also reported comparatively higher levels ofachievement strivings and parental expectations in thefamily environment. To clarify why estimated income hasa positive association with eating disorder symptoms inChinese girls, features of social and family environments thatcorrespond with income must be considered in future work.

Among Chinese boys, negative self-evaluations aboutfatness and social pressure about physical appearance alsopredicted subsequent increases in eating disorder symptoms.Once again, relatively stronger collectivist orientations mayhave contributed to the salience of perceived social pressureon eating disturbances. Given the parallels observed betweenboys and girls in this research, prevention and interventionefforts targeting negative self-evaluations about appearanceand employing strategies to challenge or inoculate againstunrealistic appearance feedback from the mass media andclose interpersonal relationships may be of considerableutility for adolescents in general.

In contrast to Western research [19,20,35,36,41], negativeaffect and internalization of a masculine attractiveness idealwere not related to subsequent eating pathology in boys. Toan extent, this may have been because muscularity and bodydissatisfaction have been outcome measures in someWestern studies with boys. Furthermore, although muscu-larity is included among Chinese conceptions of masculineideals, nonphysical characteristics may be embraced more

strongly in defining masculinity among Chinese malescompared to their Western peers [56]. The identificationand assessment of physical attributes considered to be idealfor Chinese males may aid in understanding body imageconcerns of this group.

Consonant with past research on Chinese samples [45],girls expressed more concern about fatness as well asincreased appearance-based social comparison and socialpressure than boys did. These results could suggest thatadolescent girls in China are more susceptible to appearanceculture influences than Chinese boys are. It is also notablethat although girls and boys endorsed the female attractive-ness ideal to a similar extent, boys favored a masculineappearance ideal that emphasized a tall, lean, muscularappearance much more highly than the girls did. Thesefindings could reflect a stronger consensus among Chineseadolescents about attractiveness ideals for girls than for boys.

With respect to changes in measures over time, fatness andfacial appearance concerns rose over time as did socialpressure about appearance and endorsement of a femalephysical appearance ideal. These changes suggest thatnegative physical appearance evaluations increased substan-tially for both girls and boys during a 9-month interval. Otherrecent studies on Chinese children and adolescents havesuggested that adoption of thin appearance ideals [8] andpreoccupations with fatness [45] are particularly acute duringmiddle adolescence. Although the average age of our samplewas just over 15 years, there was also considerable variabilityin respondent ages. Hence, determining whether and whyparticular age groups are at increased risk for developingeating disturbances is an important issue for future research.In this vein, considerable variability was observed in a recentlongitudinal study on predictors of body dissatisfaction inlarge samples of American early and middle adolescent boysand girls [20]. Extending such designs to the prediction ofchanges in eating pathology within defined age samples (e.g.,13-year-olds vs. 16-year-olds vs. 19-year-olds) may help toenrich understanding of risk factors most relevant tooutcomes at particular points in development.

To our knowledge, this is the first longitudinal study toexamine factors that may predict changes in eating disordersymptoms of Chinese adolescents. Within each gender, highinitial levels of personal concern about fatness and socialpressure about appearance contributed to increases indisordered eating behavior 9 months later. These findingssuggest that at least select features of the dual-pathwaymodel extend to non-Western samples.

Despite these implications, the main limitations of thisstudy must be acknowledged. First and foremost, although alongitudinal design was employed, causal inferences must bemade with caution because random assignment and experi-mental manipulations were absent from the design. Conse-quently, extraneous variables cannot be ruled out asinfluences on the findings. Furthermore, the intervalexamined was relatively short; hence, extensions arewarranted to assess changes over longer intervals. Third,

94 T. Jackson, H. Chen / Journal of Psychosomatic Research 64 (2008) 87–95

incorporation of assessment methods such as behavioralobservation and experimentation in future work may informfindings based on self-reports. Finally, although we assessedconstructs derived from the dual-pathway model as well asselected factors that may be unique to China (i.e., facialappearance concerns), these factors accounted for a smallpercentage of the variance in eating pathology changes.Expansion of models to consider other personality, inter-personal, and cultural factors [57] may help to improveprediction models in future research.

Acknowledgments

This project was funded by a National Key SubjectFoundat ion of Fundamental Psychology grant(SNKSF06014) to Todd Jackson and a National KeySubject Foundation of Fundamental Psychology grant(SNKSF07001) and Philosophy and Social Science grant(05JZD0031) to Hong Chen from the China Ministry ofEducation.

References

[1] Minz LB, Betz NE. Prevalence and correlates of eating disorderedbehaviors among undergraduate women. J Couns Psychol 1988;35:463.

[2] Neumark-Sztainer D, Story M, Hannan PJ, Perry C, et al. Weight-related concerns and behaviors among overweight and non-overweightadolescents: implications for preventing weight-related disorders. ArchPediatr Adolesc Med 2002;156:171.

[3] Lee S. Fat phobic and non-fat phobic anorexia nervosa: a comparativestudy of 70 Chinese patients in Hong Kong. Psychol Med 1993;23:999.

[4] Lee AM, Lee S. Disorder eating and its psychosocial correlates amongChinese adolescent female in Hong Kong. Int J Eat Disord 1996;20:177.

[5] Lee S, Leung T, Lee AM, Yu H, et al. Body dissatisfaction amongChinese undergraduates and its implications for eating disorders inHong Kong. Int J Eat Disord 1996;20:77.

[6] Leung F, Lam S, Chan I. Disordered eating attitudes and behavioramong adolescent in Hong Kong. J Youth Stud 2001;4:36.

[7] Leung F, Lam S, Size S. Cultural expectations of thinness in Chinese.Eat Disord 2001;9:339.

[8] Li YP, Hu XQ, Ma WJ, Wu J, Ma G. Body image perception amongChinese children and adolescents. Body Image 2005;2:91.

[9] Chen H, Jackson T. Prevalence and sociodemographic correlates ofeating disorder endorsements among adolescents and young adultsfrom China. Eur Eat Disord Rev 2007 Oct 25;[Epub ahead of print].

[10] Luo Y, Parish WL, Lauman EO. A population-based study of bodyimage concerns among urban Chinese adults. Body Image 2005;2:333.

[11] Keel P, Dorer D, Eddy K, Franko D, et al. Predictors of mortality ineating disorders. Arch Gen Psychiatry 2004;60:179.

[12] Mickley DW. Medical aspects of anorexia and bulimia. In: Kinoy B,editor. Eating disorders. 2nd ed. New York: Columbia UniversityPress, 2001. pp. 7–16.

[13] Franko D, Dorer D, Keel P, Jackson S, et al. How do eating disordersand alcohol use disorder influence each other? Int J Eat Disord 2005;38:200.

[14] Franko D, Keel P, Dorer D, Blais M, et al. What predicts suicideattempts in women with eating disorders? Psychol Med 2004;34:843.

[15] Stice E, Hayward C, Cameron R, Killen JD, et al. Body image andeating related factors predict onset of depression in female adolescents.J Abnorm Psychol 2000;109:438.

[16] Barker ET, Galambos NL. Body dissatisfaction of adolescent girls andboys: risk and resources factors. J Early Adolesc 2003;23:141.

[17] Bearman S, Presnell K, Martinez E, Stice E. The skinny on bodydissatisfaction: a longitudinal study of adolescent girls and boy.J Youth Adolesc 2006;35:229.

[18] Gardner R, Stark K, Friedman B, Jackson N. Predictors of eatingdisorder scores in children ages 6 through 14: a longitudinal study.J Psychosom Res 2000;49:199.

[19] McCabe M, Ricciardelli LA. A longitudinal study of body image andstrategies to lose weight and increase muscles among children. J ApplDev Psychol 2005;26:559.

[20] Paxton S, Eisenberg M, Neumark-Sztainer D. Prospective predictors ofbody dissatisfaction in adolescent girls and boys: a five-yearlongitudinal study. Dev Psychol 2006;42:888.

[21] Presnell K, Bearman S, Stice E. Risk factors for body dissatisfaction inadolescent boys and girls: a prospective study. Int J Eat Disord 2004;36:389.

[22] Stice E, Whitenton K. Risk factors for body dissatisfaction in adolescentgirls: a longitudinal investigation. Dev Psychol 2002;38:669.

[23] Gabrenya W, Hwang K. Chinese social interaction. In: Bond MH,editor. The handbook of Chinese psychology. London: Oxford, 1996.pp. 309–21.

[24] Oetzel J, Ting-Toomey S. Face concerns in interpersonal conflict: across-cultural empirical test of the face negotiation theory. CommunRes 2003;30:599.

[25] Cooley E, Toray T. Body image and personality predictors of eatingdisorder symptoms during the college years. Int J Eat Disord 2001;30:28.

[26] Johnson F, Wardle J. Dietary restraint, body dissatisfaction, and psycholo-gical distress: a prospective analysis. J Abnorm Psychol 2005;114:119.

[27] Stice E. Risk and maintenance factors for eating pathology: a meta-analytic review. Psychol Bull 2002;128:825.

[28] Stice E, Shaw H. Role of body dissatisfaction in the onset andmaintenance of eating pathology: a synthesis of research findings.J Psychosom Res 2002;53:985.

[29] Thompson JK, Coovert M, Richards K, Johnson S, et al. Developmentof body image, eating disturbance, and general psychologicalfunctioning in female adolescents: covariance structure modeling andlongitudinal investigations. Int J Eat Disord 1995;18:221.

[30] Wertheim EH, Koerner J, Paxton SJ. Longitudinal predictors ofrestrictive eating and bulimic tendencies in three different age groupsof adolescent girls. J Youth Adolesc 2001;30:69.

[31] Lee S. Self-starvation in context: towards a culturally sensitiveunderstanding of anorexia nervosa. Soc Sci Med 1995;41:25.

[32] Jackson T, Chen H. Identifying the eating disorder symptomatic inChina: the role of sociocultural factors and culturally-definedappearance concerns. J Psychosom Res 2007;62:241.

[33] Lee S, Leung CM, Wing YK, Chiu HF. Acne as a risk factor foranorexia nervosa in Chinese. Aust N Z J Psychiatry 1991;25:134.

[34] Thompson JK, Heinberg LJ, Altabe M, Tantleff-Dunn S. Exactingbeauty: theory, assessment, and treatment of body image disturbances.Washington: APA, 1999.

[35] Fisher E, Dunn M, Thompson JK. Social comparison and body image.J Soc Clin Psych 2002;21:566.

[36] McCreary DR, Sasse DK. An exploration of the drive for muscularityin adolescent boys and girls. J Am Coll Health 2000;48:297.

[37] Stice E, Agras WS. Predicting onset and cessation of bulimic behaviorsduring adolescence: a longitudinal grouping analysis. Behav Ther1998;29:257.

[38] Stice E, Presnell K, Spangler D. Risk factors for binge eating onset: aprospective investigation. Health Psychol 2002;21:131.

[39] Vohs K, Heatherton T, Herrin M. Disordered eating and the transitionto college: a prospective study. Int J Eat Disord 2001;29:280.

[40] Stice E. A prospective test of the dual pathway model of bulimicpathology: mediating effects of dieting and negative affect. J AbnormPsychol 2001;110:124.

[41] Jones DC. Body image among adolescent girls and boys. Dev Psychol2004;40:823.

95T. Jackson, H. Chen / Journal of Psychosomatic Research 64 (2008) 87–95

[42] Chen H, Gao X, Jackson T. Predictive models for understanding bodydissatisfaction among young males and females in China. Behav ResTher 2007;45:1345.

[43] Stice E, Telch CF, Rizvi SL. Development and validation of the eatingdisorder diagnostic scale: a brief self-report measure of anorexia,bulimia, and binge-eating disorder. Psychol Assess 2000;12:123.

[44] SticeE, FisherM,MartinezE. Eating disorder diagnostic scale: additionalevidence of reliability and validity. Psychol Assess 2004;16:60.

[45] Chen H, Jackson T, Huang X. Initial development and validation of theNegative Physical Self Scale among Chinese adolescents and youngadults. Body Image 2006;3:401.

[46] Thompson JK, Cattarin J, Fowler H, Fisher E. The Perception of TeasingScale (POTS): a revision and extension of the Physical AppearanceRelated Teasing Scale (PARTS). J Pers Assess 1995;65:146.

[47] Thompson JK, Heinberg LJ, Tautleff S. The Physical AppearanceComparison Scale (PACS). Behav Ther 1991;14:174.

[48] Davison TE, McCabe MP. Relationships between men's and women'sbody image and their psychological, social and, sexual functioning.Sex Roles 2005;52:463.

[49] Stice E, Bearman SK. Body image and eating disturbancesprospectively predict increases in depressive symptoms in adolescentgirls: a growth curve analysis. Dev Psychol 2001;37:597.

[50] Watson D, Clark LA, Tellegen A. Development and validation of briefmeasures of positive and negative affect: the PANAS scales. J Pers SocPsychol 1988;54:1063–70.

[51] Attie I, Brooks-Gunn J. Development of eating problems in adolescentgirls: a longitudinal study. Dev Psychol 1998;25:70.

[52] Hui H, Triandis HC. Individualism–collectivism: a study of cross-cultural researchers. J Cross-Cult Psychol 1986;17:225.

[53] Frith K, Cheng H, Shaw P. Race and beauty: a comparison of Asianand Western models in women's magazine advertisements. Sex Roles2004;50:53.

[54] Lee S, Lee AM. Disordered eating in three communities of China. Int JEat Disord 2000;27:317.

[55] Lee HY, Lock J. Anorexia nervosa in Asian-American adolescents:do they differ from their non-Asian peers? Int J Eat Disord 2007;40:227.

[56] Yang C, Gray P, Pope HG. Male body image in Taiwan versus theWest: Yanggang Zhiqi meets the Adonis complex. Am J Psychiatry2005;162:263.

[57] Jackson T, Weiss KE, Lundquist JJ, Soderlind A. Sociotropy andperceptions of interpersonal relationships as predictors of eatingdisturbances among college women: two prospective studies. J GenetPsychol 2005;166:346.