Eating Behaviors 16 (2015) 34–39

Contents lists available at ScienceDirect

Eating Behaviors

Problematic eating behaviors among bariatric surgical candidates: Apsychometric investigation and factor analytic approach

Bethany L. Gelinas ⁎, Chelsea A. Delparte, Kristi D. Wright, Regan HartUniversity of Regina

⁎ Corresponding author at: Department of PsychologySaskatchewan, CANADA, S4S 0A2. Tel.: +1 306 585 4221

E-mail address: gelinabe@uregina.ca (B.L. Gelinas).

http://dx.doi.org/10.1016/j.eatbeh.2014.10.0181471-0153/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 24 May 2014Received in revised form 23 September 2014Accepted 24 October 2014Available online 03 November 2014

Keywords:Bariatric surgeryPsychopathologyEating BehaviorPre-surgery Assessment

Psychological factors (e.g., anxiety, depression) are routinely assessed in bariatric pre-surgical programs, as highlevels of psychopathology are consistently related to poor program outcomes (e.g., failure to lose significantweight pre-surgery, weight regain post-surgery). Behavioral factors related to poor program outcomes andways in which behavioral and psychological factors interact, have received little attention in bariatric researchand practice. Potentially problematic behavioral factors are queried by Section H of the Weight and Lifestyle In-ventory (WALI-H), in which respondents indicate the relevance of certain eating behaviors to obesity. A factoranalytic investigation of the WALI-H serves to improve the way in which this assessment tool is interpretedand used among bariatric surgical candidates, and subsequent moderation analyses serve to demonstratepotential compounding influences of psychopathology on eating behavior factors. Bariatric surgical candidates(n =362) completed several measures of psychopathology and the WALI-H. Item responses from the WALI-Hwere subjected to principal axis factoring with oblique rotation. Results revealed a three-factor model including:(1) eating in response to negative affect, (2) overeating/desirability of food, and (3) eating in response to positiveaffect/social cues. All three behavioral factors of the WALI-H were significantly associated with measures of de-pression and anxiety. Moderation analyses revealed that depression did not moderate the relationship betweenanxiety and any eating behavior factor. Although single forms of psychopathology are related to eating behaviors,the combination of psychopathology does not appear to influence these problematic behaviors. Recommendationsfor pre-surgical assessment and treatment of bariatric surgical candidates are discussed.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The prevalence of psychopathology is often higher in bariatric surgi-cal candidate populations than the general population (Kalarchian et al.,2007). Specifically, bariatric surgical candidates experience high rates ofdepression and anxiety (e.g., Muhlhans, Horbach, & de Zwaan, 2009;Sarwer et al., 2004). A wealth of extant literature has contributed tothe evidence that psychological factors can influence bariatric surgicalcandidates’ pre-surgical and post-surgical success (Abilés et al., 2010;Kalarchian et al., 2007). In particular, high levels of psychopathologyare related to poor program outcomes, such as failure to lose significantweight pre-surgery or weight regain following surgery (de Zwaan et al.,2011; Kalarchian et al., 2007, 2008). Given the research indicating thatpsychopathology is a negative prognostic indicator for bariatricprograms, the practice of assessing such psychological factors has be-come common-place in bariatric surgery settings (Walfish, Vance, &Fabricatore, 2007). In fact, the detection of untreated or undertreateddepression and anxiety through psychological assessment are among

, University of Regina, Regina,; fax: +1 306 585 4772.

the top reasons that bariatric surgical candidates are denied or delayedsurgery (Walfish et al., 2007). The pre-surgical psychological assess-ment has proven useful in identifying psychological indicators of poorbariatric program success (Franks & Kaiser, 2008; Pawlow, O'Neil,White, & Byrne, 2005), and in doing so, identifying key targets forpsychological intervention.

Comparatively, a much smaller body of literature has examined be-havioral factors which may influence bariatric surgical candidates’pre- and post-surgical outcomes. Given the population, eating behaviorsin particularwarrant attention. Although examining eating behaviors ina bariatric population seems intuitive, a recent systematic literaturesearch (Carter & Jansen, 2012) failed to produce any studies that soughtto identify the target eating behaviors involved in obesity. The lack ofconsensus regarding the key behavioral factors involved in bariatricprogramsuccess is in stark contrast to thewell-established literature re-garding the emotional factors involved. Furthermore, there is little in-formation regarding potential interactions between the emotional andbehavioral components of bariatric surgical candidates’ cases.

Despite the dearth of research conducted on potentially problematiceating behaviors, a clinical tool exists which queries this information.The Weight and Lifestyle Inventory (WALI; Wadden & Foster, 2006) isa self-report questionnaire routinely used in pre-surgical bariatric

35B.L. Gelinas et al. / Eating Behaviors 16 (2015) 34–39

evaluations and includes a set of 24 items (Section H; from herein re-ferred to asWALI-H) that queries the relevance of certain eating behav-iors to obesity. Using a five-point Likert-style scale, respondents rate thedegree towhich they believe 24 different eating behaviors contribute totheir obesity. The WALI-H has been used in the past to make interven-tion recommendations based on bariatric surgical candidates’ endorse-ment of various eating behaviors (Walfish, 2004; Walfish & Brown,2009).

Fabricatore, Crerand, et al. (2006), Fabricatore,Wadden, et al. (2006)conducted a principal components analysis (with promax rotation) onthe WALI-H, which yielded five factors. The first factor was labellednegative affect as the seven pertinent items assess eating in responseto various negative emotions (i.e., stress, depression, anxiety, anger,boredom, loneliness, fatigue). The second factor, positive affect and socialcues, consisted of five items that assessed eating in response to positiveemotions or social events (e.g., eating when happy, eating whencelebrating). The third factor, overeating/impaired appetite, consisted ofseven items related to overeating, hunger, cravings, and not feelingfull. The fourth factor, overeating at early meals, only consisted of twoitems, and pertained to overeating at breakfast and lunch in particular.The fifth factor was titled snacking, and its two items related to snackingafter dinner and between meals. Fabricatore, Crerand, et al. (2006),Fabricatore,Wadden, et al. (2006) demonstrated adequate internal con-sistency (Cronbach’s alpha coefficients ranged from 0.65 to 0.88) andtest-retest reliability (reliability coefficients ranged from 0.61 to 0.81)for the five factors. Furthermore, all five factors demonstrated an associ-ation with depression and binge eating (with the exception of positiveaffect and social cues being unrelated to depression).

Although Fabricatore, Crerand, et al. (2006), Fabricatore, Wadden,et al. (2006)were able to demonstrate acceptable psychometric proper-ties for the five-factor model of the WALI-H; a few notable limitationswould suggest that further factor analytic studies are warranted. First,this factor structure has yet to be replicated in an independent sample.Second, in recent years the recommendations for factor analytic prac-tices have been updated. As such, principal axis factoring is now favoredabove principal components analysis (Costello & Osborne, 2005). Simi-larly, using eigenvalues ≥1 as the primary criteria for determining fac-tors has been amended. Currently, more stringent criteria, such ascomparing eigenvalues to parallel analyses is considered best practicefor determining factors (Costello & Osborne, 2005; Thompson, 2004).Moreover, it is generally accepted that factors should not have fewerthan three items, as these factors are unstable and should not be consid-ered robust (Costello & Osborne, 2005).

Given these recent advances in statistical practices specifically, andthe general need to replicate in independent samples, additional exam-ination of the factor structure of the WALI-H may prove beneficial. Assuch, the current investigation was designed to: 1) re-examine the fac-tor structure of the WALI-H; 2) investigate howmeasures of psychopa-thology typically encountered in a bariatric population are related to thefactors; and 3) investigate potential moderating or compounding influ-ences of psychopathology on eating behavior factors. The current inves-tigation will serve to improve the way in which a common pre-surgicalassessment tool is interpreted and used among bariatric surgicalcandidates, as well as improve the overall understanding of key eatingbehaviors and psychopathology in a bariatric population.

2. Material and Methods

2.1. Participants

Datawas collected from362 (274women and 88men) bariatric sur-gery candidates from theXXXXRegion (XXXX) bariatric surgical assess-ment clinic in XXXX, Canada. Candidates (M age =44.3; SD =10.4)were enrolled in a six-month pre-surgery program during which theyhad regular consultations with a multidisciplinary team consisting of asurgeon, clinical psychologist, dietician, nurse, and exercise therapist.

The majority of candidates were married (59%), Caucasian (84%) andmost had at least a high school education (84%). There were no signifi-cant demographic differences betweenmale and female bariatric surgi-cal candidates. Bariatric surgical candidates had an average weight of147.78 kg (SD =29.14), which corresponds to an average body massindex (BMI) of 52.16 (SD =8.46). This degree of obesity is typical ofother bariatric populations (e.g., Fabricatore, Crerand, et al., 2006;Fabricatore, Wadden, et al., 2006). Candidates approved for surgery re-ceived Roux en Y gastric bypass or gastric banding procedures.

2.2. Procedure

All bariatric surgical candidates completed a battery of question-naires as part of their involvement in the six-month pre-surgeryprogram. The data for the current studywas collected from the patients’initial pre-surgical psychological assessment. Furthermore, objectivemeasures of height and weight were obtained and converted to BMI.

2.3. Measures

2.3.1. Center for Epidemiological Studies Depression Scale (CES-D; Radloff,1977)

The CES-D is a 20-item self-report measure that assesses forthe presence of depressive feelings and behaviors over a one weekperiod. The CES-D queries symptoms associated with depression(e.g., depressed mood, feelings of guilt and worthlessness, feelings ofhelplessness and hopelessness, loss of appetite, sleep disturbance,psychomotor retardation) that have been used in previously validatedscales of depression (Radloff, 1977). Items are rated on a 4-point Likertscale based on the frequency of occurrence during the past week(i.e., responses range from rarely/none of the time – less than one day,to most/all of the time – 5 to 7 days). Total scores range from 0 to 60with higher scores indicating more depressive symptomatology. Cutoff scores are as follows: scores ranging from 15 to 21 indicate mild tomoderate depression, and scores greater than 21 suggest the possibilityof major depression. The CES-D has demonstrated high internal consis-tency across studies (α = .63 to .91; Devins, Orme, & Costello, 1988;Radloff, 1977) and moderate test-retest reliability (r = .61) in variousadult populations (Devins et al., 1988). Furthermore, the CES-D has suc-cessfully been utilized to assess depression among bariatric patients(Bond, Phelan, Leahey, Hill, & Wing, 2009). In the current study, theCES-D demonstrated excellent internal consistency (α = .92).

2.3.2. Self-reported Zung Anxiety Scale (ZAS; Zung, 1971)The ZAS is a 20-item measure developed to assess the frequency of

anxiety symptoms over a one-week period. The ZAS contains four factorsincluding: anxiety and panic, somatic control, vestibular sensations, andgastrointestinal/muscular sensations (Olatunji, Deacon, Abramowitz, &Tolin, 2006). Items are rated on a 4-point Likert scale based on the fre-quency of occurrence during the past week (i.e., responses range fromnone or a little of the time, to most/all of the time). Total scores rangefrom 20 to 80, with higher scores indicating more anxiety. A score of36 or higher indicates clinical anxiety (Zung, 1971). In past research,the ZAS has demonstrated good internal consistency (α = .81; Olatunjiet al., 2006). In the current study, the ZAS demonstrated comparablygood internal consistency (α = .83).

2.3.3. Weight and Lifestyle Inventory (WALI; Wadden & Foster, 2006)TheWALI is a self-report questionnaire that examines biological, en-

vironmental, and psychosocial factors related to weight difficulties. TheWALI is commonly used in bariatric pre-surgical settings. For the cur-rent study, Section A and Section H were utilized. Questions fromSection A of the WALI query demographic and weight information ofbariatric surgical candidates. Section H of theWALI lists 24 different eat-ing behaviors, and has respondents rate the degree to which they be-lieve each eating behavior contributes to their obesity. Items are rated

Table 1Principal axis factoring of the WALI-H items.

Item Factor Loading

1 2 3

17. Eating when anxious .90715. Eating when depressed .87316. Eating when angry .81514. Eating when stressed .79318. Eating when alone .59719. Eating when bored .53320. Eating when tired .4709. Eating too much food .8208. Overeating at dinner .8187. Eating because of inability to stop .75210. Continuing to eat due to not feeling full .7316. Eating because of good taste of the food .51421. Overeating at lunch .4495. Eating in response to sight or smell of food .41212. Eating because of feeling physically hungry .39911. Eating because of cravings .39923. Snacking after dinner .34524. Snacking between meals .3301. Eating with family/friends .8372. Eating when socializing/celebrating .8303. Eating at business functions .5974. Eating when happy .399

*Only loadings N0.30 are displayed. If an item loaded on more than one factor, it wasretained on the factor on which it had the highest loading. Two items are not included:the degree to which eating while cooking/preparing food contributes to weight gain;the degree to which overeating at breakfast contributes to weight gain.

36 B.L. Gelinas et al. / Eating Behaviors 16 (2015) 34–39

on a 5-point Likert scale based on each behavior’s contribution(i.e., responses range from does not contribute at all, to contributes thegreatest amount). Extant literature has indicated that the WALI-H hasacceptable test-retest reliability (r= .61) over a one to twoweek period(Crerand et al., 2006). No other sections of theWALIwere utilized in thecurrent study, as all other sections would have been extraneous to thecurrent goals of improving our understanding of eating behaviors andpsychopathology. Furthermore, separate factor analyses for each WALIsection are unwarranted and beyond the scope of the current study.

2.4. Statistical Analyses

Five separate sets of analyses were completed. First, descriptivestatistics were computed for demographic variables. Second, an explor-atory factor analysis using principal axis factoring (PAF)was conducted.An oblique rotation method (i.e., promax rotation) was chosen becausethe factors were not expected to be independent of each other. As rec-ommended in the current statistical literature (Costello & Osborne,2005), parallel analysis was used to determine factor retention. In paral-lel analysis, actual eigenvalues are compared with random ordereigenvalues. Factors are retained when actual eigenvalues surpass ran-dom ordered eigenvalues. In this way, factors are estimated using amathematical model, whereby the shared variance between linearitems is analyzed (Tabachnick& Fidell, 2007). Third, a series of indepen-dent sample t-tests were conducted to compare theWALI-H factors be-tween males and females. Fourth, a series of Pearson product-momentcorrelations were computed between the newly determined factorsand common measures of psychopathology. Fifth, a series of modera-tion analyses were conducted to determine the potential moderatingeffect of psychopathology scores on different problematic eatingbehavior factors. The analyses were conducted using the softwarepackage SPSS 21.0 (SPSS, Chicago IL) and PROCESS 2.11 (Hayes, 2013).

3. Results

3.1. Factor Structure of the WALI-H

The 24 items of the WALI-H were subjected to PAF using SPSSVersion 21.0. Prior to performing PAF, the suitability of data for factoranalysis was assessed. Inspection of the correlation matrix revealedthe presence of many coefficients of .30 and above. The Kaiser-Meyer-Oklin value was .899, exceeding the recommended value of .60(Kaiser, 1970, 1974) and Bartlett’s Test of Sphericity (Bartlett, 1954)reached statistical significance, supporting the factorability of the corre-lation matrix.

PAF initially revealed the presence of five factors with eigenvaluesexceeding 1.0, explaining 33.3%, 9.7%, 6.5%, 4.9%, and 4.6% of the vari-ance respectively; however, an inspection of the scree plot revealed amore distinct break after three factors. A parallel analysis using theMonte Carlo PCA, showedonly three factorswith eigenvalues exceedingthe corresponding criterion values for a randomly generated data ma-trix of the same size.

Based on themultiple tests that corroborated a three-factormodel, asecond PAF was run with three forced factors. One item (“degree towhich eating while cooking/preparing food contributes to weightgain”) was removed due to its extremely low communality (.061) andlow factor loading. The PAF was re-run a third time, and a second item(“degree to which overeating at breakfast contributes to weight gain”)was removed due to crossloading acrossmultiple factors. This problem-atic item loaded at .32 or higher on two factors, and as such, did notmeet the minimum factor loading criteria commonly recommendedby Tabachnick and Fidell (2001).

Ultimately, three factors were retained, including: Negative Affect(7 items), Overeating/Food Desirability (11 items), and Positive Affect/Social Cues (4 items). See Table 1 for the items’ factor loadings. Thefinal three factor solution explained a total of 52.7% of the variance,

with Negative Affect contributing 35.04%, Overeating/Food Desirabilitycontributing 10.65%, and Positive Affect/Social Cues contributing7.01%. The Negative Affect factor had a final eigenvalue of 7.71. TheOvereating/Food Desirability factor had a final eigenvalue of 2.34. ThePositive Affect/Social Cues factor had a final eigenvalue of 1.54. Toaid in the interpretation of the three factors, promax rotation wasperformed. The rotated solution revealed factors showing a numberof strong loadings, and all items loading substantially on only onecomponent. The inter-factor correlations ranged from low (r = .30) tomoderate (r = .60) positive correlations. See Table 2 for a summary ofthe descriptive characteristics and correlations among the three identi-fied factors.

3.2. Internal Consistency of the WALI-H

In the current sample, the entire WALI-H had excellent internalconsistency (α = .90). The three derived factors also proved to haveadequate to good internal consistency. Both the Negative Affect factor(α = .89) and the Overeating/Food Desirability factor (α = .86)had good internal consistency, whereas the Positive Affect/Social Cues(α= .77) factor had adequate internal consistency. Of note, the internalconsistency of the three-factor model favourably compares to the inter-nal consistency of the previous five-factor model (Fabricatore, Crerand,et al., 2006; Fabricatore, Wadden, et al., 2006).

3.3. Demographic Characteristics and the WALI-H

Bivariate correlations revealed that age, weight, and BMIwere unre-lated to the threeWALI-H factors. Independent sample t-testswere con-ducted to compare the WALI-H factors for males and females. Therewere no significant differences for the Overeating/Food Desirability fac-tor and Positive Affect/Social Cues factor. However, there was a signifi-cant difference in scores on the Negative Affect factor between malesand females, t (359)=2.20, p= .029, with females experiencing higherNegative Affect scores.

Table 2Descriptive characteristics and interfactor correlations among the three identified factors.

Factor Eigenvalue Variance Explained (%) α Mean S.D. 1 2 3

1. Negative affect 7.71 35.04 .89 18.81 7.50 1.00 .541 .6072. Overeating / Food desirability 2.34 10.65 .86 31.26 8.22 - 1.00 .6353. Positive affect / Social cues 1.54 7.01 .77 10.19 3.58 - - 1.00

p values for all interfactor correlations b0.01.

37B.L. Gelinas et al. / Eating Behaviors 16 (2015) 34–39

3.4. Correlational Analyses

Bivariate correlations were conducted between ZAS scores, CES-Dscores, and the three derived eating behavior factors. Table 3 presentsthe correlation coefficients, mean values and standard deviations forthese variables. ZAS scores and CES-D scores both showed positive,significant relationships with all three eating behavior factors.

3.5. Moderation Analyses

Moderation investigates the unique conditionsunderwhich twovar-iables are related (Hayes, 2013). Analyses to determinewhether depres-sive symptoms moderate relations between anxiety symptoms andeating behavior factors followed recommendations by Hayes (2013).Thismoderation analysis procedure yields the significance of the changein R2 produced by interactions between independent (i.e., anxiety) andmoderator variables (i.e., depression). This value serves as an index ofwhether the interaction significantly predicts relations between the in-dependent (i.e., anxiety) and dependent (i.e., eating behavior factor)variables. Moderation analysis allows examination of the link betweenindependent variable and dependent variable at low (i.e., -1 SD) andhigh levels (+1 SD) of the moderator, and in doing so, helps answerquestions such as when, or under which conditions anxiety is relatedto different types of eating behaviors. A series of moderation analyseswere conducted using PROCESS 2.11.

3.5.1. Moderation of the Negative Affect FactorThe analysis indicated that approximately 11% of the variation in the

dependent variable (i.e., negative affect eating behaviors) could be ex-plained by themain effects and the interaction effects (R2= .34, adjust-ed R2 = .11, F (3, 353) =15.01, p b .001). However, the Depression XAnxiety interaction did not significantly predict the Negative Affect fac-tor. Thus,moderating effects of depression on relations between anxietyand Negative Affect eating behaviors were not supported in the currentstudy. Although the interaction model was not significant, for the sakeof comprehensive reporting it is prudent to review the evidence of sig-nificant relationships between the dependent and independent vari-ables at certain levels of the moderating variable. These conditionaleffects can be interpreted at face value; however, the pattern of signifi-cance should not be mistaken as evidence of moderation (A. Hayes,personal communication, April 3, 2014). The moderation analysis pro-vided evidence that anxiety is significantly related to theNegative Affectfactor for low (B=0.21; t=2.84; p b .01) and average scores (B=0.16;t =2.66; p b .01) of depression, but not for high scores of depression.

Table 3Descriptive characteristics and interfactor correlations among the study variables.

Variable Mean S.D. 1 2 3 4 5

1. Negative affect 18.81 7.50 1.00 .541** .607** .300** .304**2. Overeating/Fooddesirability

31.26 8.22 - 1.00 .635** .192** .153**

3. Positive affect/Socialcues

10.19 3.58 - - 1.00 .170** .138**

4. Anxiety (ZAS) 35.22 8.48 - - 1.00 .676**5. Depression (CES-D) 13.68 10.81 - - 1.00

p values for all correlations b0.01. (2-tailed).

The absence of a significant relationship at high levels of depression,may indicate that for those individuals with high depression scores,their level of “negative affect” is substantial enough that the combina-tion of anxiety has no additive influence on negative affect related eat-ing behaviors. Whereas for individuals with low or average depressionscores, further negative affect (such as anxiety) may combine in suchaway that the relationship with negative affect related eating behaviorsbecomes significant.

3.5.2. Moderation of the Overeating/Food Desirability FactorThe analysis indicated that approximately 5% of the variation in the

dependent variable (i.e., overeating/food desirability related behaviors)could be explained by the main effects and the interaction effects(R2 = .21, adjusted R2 = .05, F (3, 351) =5.64, p b .001). However,the Depression X Anxiety interaction did not significantly predict theOvereating/Food Desirability factor. Thus, moderating effects of depres-sion on relations between anxiety and eating behaviors related to over-eating and the desirability of food were not supported in the currentstudy. Although the interaction model was not significant, evidence ofsignificant relationships between the dependent and independent var-iables at certain levels of the moderating variable warrant review. Themoderation analysis provided evidence that anxiety is significantly re-lated to the Overeating/Food Desirability factor for high values of de-pression (B =0.21; t =2.71; p b .01), but not for low or averagevalues of depression. In this case, it may be that a certain level ofpsychopathology is required before overeating behaviors are observed.As such, high levels of depression, in conjunction with anxiety, prove tohave a significant relationship with behaviors related to overeating andfood desirability.

3.5.3. Moderation of the Positive Affect/Social Cues FactorThe analysis indicated that about 4% of the variation in the depen-

dent variable (i.e., positive affect/social cues eating behaviors) could beexplained by the main effects and the interaction effects (R2 = .20,adjusted R2 = .04, F (3, 354) =4.64, p b .01). However, the DepressionX Anxiety interaction did not significantly predict the Positive Affect/Social Cues factor. Thus, moderating effects of depression on relationsbetween anxiety and eating behaviors related to positive affect and so-cial cues were not supported in the current study. For the sake of com-plete reporting, the evidence of significant relationships betweenanxiety and the Positive Affect/Social Cues factor at certain levels of de-pression should be mentioned. The moderation analysis indicated thatanxiety is significantly related to the Positive Affect/Social Cues factorfor high values of depression (B =0.08; t =2.38; p b .05), but not forlow or average values of depression. It may be that higher levels of de-pression and anxiety are indicative of general emotionality, and thisgeneral emotionality is related to eating in response to any emotionality,including positive affect.

4. Discussion

Although queried in clinical practice, problematic eating behaviorshave rarely been examined in bariatric research. The WALI-H, a self-report measure of eating behaviors perceived to be responsible forweight gain, has previously been demonstrated to comprise five factorsof problematic eating behaviors (Fabricatore, Crerand, et al., 2006;Fabricatore,Wadden, et al., 2006). Furthermore, the fiveWALI-H factors

38 B.L. Gelinas et al. / Eating Behaviors 16 (2015) 34–39

were significantly related to measures of depression and binge eating(Fabricatore, Crerand, et al., 2006; Fabricatore, Wadden, et al., 2006).However, previous research employed outdated statistical practices.The current investigation was designed to further examine the factorstructure of the WALI-H using newly recommended statistical tech-niques, its relationship to common measures of psychopathology, andwhether compounding influences of psychopathology may predictcertain problematic eating behaviors.

The current study found that three factors underlie the eating be-haviors queried in theWALI-H among a sample of bariatric surgery can-didates, in contrast to previous research that documented a five-factorstructure. The current factors include: eating in response to NegativeAffect, Overeating and Food Desirability, and eating in response toPositive Affect/ Social Cues. All three factor scores were found to havestrong reliability. Notably, these factors correspond to bariatric surgicalcandidates’ opinions and perceptions regarding eating behaviorsrelated to weight gain.

The first factor, eating in response to negative affect (e.g., stress,boredom, anger), has been documented in past literature as problematicfor bariatric surgical candidates. For example, in a sample of 122 femalebariatric surgical candidates, boredom, stress, and depression were re-ported by almost half of the sample as contributing a large amount totheir weight gain. Furthermore, only 38% of the female sample deniedthat their weight gain was attributable to eating in response to negativeemotions (Walfish, 2004). In a similar study (Walfish & Brown, 2009), asample of 100 male bariatric surgical candidates indicated that bore-dom, stress, and depression were the emotions that most often resultedin problematic eating and weight gain. In contrast to females, 59% ofthemale sample denied that their weight gainwas attributable to eatingin response to any negative emotion. In the current study, femalesreported eating in response to negative affect more than their malecounterparts, in line with past research.

Extant literature has also demonstrated that eating in response tonegative affect is differentially associated with body weight. Geliebterand Aversa (2003) demonstrated that overweight individuals reporteating more in response to negative emotions and negative situations(e.g., when alone, after an argument) than do individuals who are of av-erage weight or underweight. Given the high proportion of bariatricsurgical candidates who report eating in response to negative affect,and the demonstrated differences between overweight and average/underweight individuals in this regard, the current study providessupporting evidence of the importance of this factor.

The second factor, overeating and eating due to the desirability offood, appears to be assumed rather than actually researched, in bariatricsettings (Carter & Jansen, 2012). Furthermore, overeating is most oftenexamined in the context of binge eating, and thus focuses dispropor-tionately on this type of eating pathology (Carter & Jansen, 2012). Asurvey of pre-surgical bariatric evaluation practices indicated thatproblematic eating behaviors (such as overconsumption or eating inresponse to the desirability of food) are not formally assessed bymentalhealth professionals (Fabricatore, Crerand,Wadden, Sarwer, & Krasucki,2006). Results from the current study serve to indicate that overeatingand eating due to the desirability of food warrant attention, and shouldbe examined outside the context of binge eating.

In the current study, eating in response to positive affect (e.g., eatingwhen happy) and social cues (e.g., eatingwhen celebrating) emerged asa factor of problematic eating behaviors. Recent experimental studiesusing normal weight populations have asserted the importance of pos-itive emotions in overeating (Bongers, Jansen, Havermans, Roefs, &Nederkoorn, 2013; Evers, Adriaanse, de Ridder, & de Witt Huberts,2013). Positive emotions can act as salient triggers for unhealthy eating,but are often neglected when discussing emotional eating (Evers et al.,2013). Although very little is known about this type of problematiceating among bariatric surgical candidates, one study (Geliebter &Aversa, 2003) revealed that overweight individuals did not endorse eat-ing in response to positive emotionsmore than average or underweight

individuals. Future research that explores eating in response to positiveemotions/social cues is needed, and particularly research that clarifiesits role in bariatric populations.

Psychopathology appears to be related to the various eating behaviorsencountered as problematic by bariatric surgical candidates. The currentstudy demonstrated similar associations, as scores on depression andanxiety were invariably related to the three eating behavior factors.Over and above the identification of key eating behaviors among bariatricsurgical candidates, an understanding of the emotional context and emo-tional interaction is necessary (Carter & Jansen, 2012).

Bariatric surgical candidates are known to have comparatively higherrates of psychopathology than the general population (Kalarchian et al.,2007), and are also presupposed to have certain problematic eating be-haviors that have resulted in the need for bariatric surgery as a weightloss tool. The former and the latter are assumed to have a relationship,as one can inadvertently influence the other. For example, a recentstudy demonstrated that bariatric candidates with problematic bingeeating behaviors were more likely to have a current and lifetime mooddisorder and a current and lifetime anxiety disorder than bariatricpatients without binge eating disorder (Jones-Corneille et al., 2012).Furthermore, the amount of psychopathology present has been found toimpact bariatric patient outcomes. In a recent investigation of the courseand predictive significance of pre- and post-surgical anxiety and depres-sive disorders in 107 bariatric surgery patients de Zwaan et al. (2011)found that the more psychopathology present (i.e., comorbidities), theless weight loss observed post-surgery. Specifically, patients with bothdepressive and anxiety disorders at the initial assessment (current andlifetime) lost significantly less weight after surgery. The aforementionedresult suggests that the presence of comorbid psychopathology in a bar-iatric population appears associated with increased problematic eatingbehaviors as implied by less weight lost post-surgery. More specifically,the presence of multi-pathology, the interaction between anxiety anddepression, appears to be more associated with problematic eating be-haviors than just one type of psychopathology. This assumption was notsubstantiated in the current study. Although psychopathology is relatedto problematic eating (e.g., Jones-Corneille et al., 2012), depression didnot significantly moderate the relationship between anxiety and any ofthe three eating behavior factors. It may be that general negative affectiv-ity (of either anxiety or depression) is enough to influence problematiceating behaviors and multiple types of negative affectivity may not benecessary. Past research (de Zwaan et al., 2011) has examined actualweight outcomes influenced by problematic eating, whereas the currentstudy examined perceptions of reasons for weight outcomes. As such, acertain degree of general negative affect may color bariatric surgical can-didates’ perceptions of which eating behaviors are problematic. In whichcase, multiple types of negative affect may not further influence candi-dates’ perceptions.Recommendations for bariatric preoperative evalua-tions can be drawn from the current results. The three problematiceating behavior factors derived from the WALI-H could be used to moresystematically assess problem eating through the clinical interview. As-sessors can query the frequency, intensity, and results of the three typesof eating behaviors. In terms of assessing for psychopathology amongbar-iatric surgical candidates, assessors are cautioned to pay serious attentionto any indication of psychopathology. Given that depression and anxietyare related to problematic eating behaviors, but do not appear to have acompounding effect, it would appear that indications of single-type psy-chopathology are just as detrimental as comorbid or multiple types ofpsychopathology.

More targeted treatment interventions could be designed by examin-ing candidates’ self-reported problematic eating behaviors. Those candi-dates who perceive that eating in response to negative affect is mostproblematic could benefit from treatment techniques that target nega-tive affect. For example, relaxation techniquesmay be helpful in decreas-ing stress and anxiety (e.g., Barlow, 2007). Cognitive therapy techniquesmay be helpful in managing depression and anger (Walfish, 2004).Alternatively, mindfulness, or behavioral techniques could be helpful

39B.L. Gelinas et al. / Eating Behaviors 16 (2015) 34–39

for those candidates who perceive that overeating and the desirability offood is most problematic (Tapper et al., 2009). For example, mindfulnesstechniques (e.g., eating slowly, focusing on each bite) can improveawareness of consumption, and decrease feelings of uncontrollability(Tapper et al., 2009). If eating in response to positive affect or socialcues is considered most problematic, candidates could be tasked withfinding alternative non-food rewards or non-food related social activities,or find other ways in which to express and celebrate positive emotions.

5. Conclusion

The primary strength of the current study is the large clinical sample(i.e., 362), in that it provides sufficient power to conduct factor analyticand moderation analyses. Other strengths include the use of modernstatistical techniques and the most recently widely accepted recom-mendations for the factor analytic approach (Costello & Osborne,2005). The consideration of anxiety in relation to eating behaviors wasan additional strength of the current study.

It is important to note a number of limitations as well. The primarylimitation of the current study is the self-report nature of the data;particularly, the data obtained from theWALI-H. Rather than having ex-pert or objective ratings of problematic eating behaviors, the currentstudy utilized candidates’ perceptions of what was related to theirweight gain. As such, the accuracy of the present findings is dependenton the candidates’ insight and perception. The unequal gender distribu-tion (76% female) in the current sample is also a limitation; however, ison par with the fact that women seek bariatric surgery at much higherrates than men (Samuel et al., 2006).

Future researchmay endeavor to predict both pre-surgical and post-surgical weight loss and adjustment using the eating behavior factorsidentified in the current study. Such information would help in clarify-ing how much emphasis should be placed on the pre-surgical assess-ment and treatment of these behaviors. Although the presence ofdepression in combination with anxiety does not seem to influencethe expression of eating behaviors, future studies could examinewheth-er or not the presence of multi-pathology interferes with the treatmentof these problematic eating behaviors.

Role of Funding SourcesNo was funding was obtained for the current study.

ContributorsAuthor A designed the study and conducted the statistical analyses. Authors A and B

wrote the first draft of the manuscript. Authors A, B, C, and D contributed to and have ap-proved the final draft of the manuscript.

Conflict of InterestAll authors declare that they have no conflicts of interest.

References

Abilés, V., Rodriguez-Ruiz, S., Abilés, J., Mellado, C., Garcia, A., Pérez de la Cruz, A., et al.(2010). Psychological characteristics of morbidly obese candidates for bariatricsurgery. Obesity Surgery, 20(2), 161–167.

Barlow, D. H. (2007). Clinical handbook of psychological disorders. New York, NY: TheGuilford Press.

Bartlett, M. S. (1954). A note on the multiplying factors for various chi square approxima-tions. Journal of the Royal Statistical Society, 16(1), 296–298.

Bond, D. S., Phelan, S., Leahey, T. M., Hill, J. O., & Wing, R. R. (2009). Weight-loss mainte-nance in successful weight losers: Surgical vs non-surgical methods. InternationalJournal of Obesity, 33(1), 173–180.

Bongers, P., Jansen, A., Havermans, R., Roefs, A., & Nederkoorn, C. (2013). Happy eating.The underestimated role of overeating in a positive mood. Appetite, 67(1), 74–80.

Carter, F. A., & Jansen, A. (2012). Improving psychological treatment for obesity. Whicheating behaviors should we target? Appetite, 58(3), 1063–1069.

Costello, A. B., & Osborne, J. W. (2005). Best practices in exploratory factor analysis: Fourrecommendations for getting the most from your analysis. Practical AssessmentResearch & Evaluation, 10(7), 1–9.

Crerand, C. E., Wadden, T. A., Sarwer, D. B., Fabricatore, A. N., Kuehnel, R. H., Gibbons, L. M.,et al. (2006). A comparison of weight histories in women with class III vs. class I–IIobesity. Obesity, 14(S3), 63S–69S.

de Zwaan, M., Enderle, J., Wagner, S., Muhlhans, B., Ditzen, B., Gefeller, O., et al. (2011).Anxiety and depression in bariatric surgery patients: A prospective, follow-up studyusing structured clinical interviews. Journal of Affective Disorders, 133(1), 61–68.

Devins, G. M., Orme, C. M., & Costello, C. G. (1988). Measuring depressive symptoms in ill-ness populations: Psychometric properties of the center for epidemiologic studies de-pression (CES-D) scale. Psychology and Health, 2(2), 139–156.

Evers, C., Adriaanse, M., de Ridder, D. T., & de Witt Huberts, J. C. (2013). Good mood food.Positive emotion as a neglected trigger for food intake. Appetite, 68(1), 1–7.

Fabricatore, A. N., Crerand, C. E., Wadden, T. A., Sarwer, D. B., & Krasucki, J. L. (2006). Howdo mental health professionals evaluate candidates for bariatric surgery? Surveyresults. Obesity Surgery, 16(5), 567–573.

Fabricatore, A. N., Wadden, T. A., Sarwer, D. B., Crerand, C. E., Kuehnel, R. H., Lipschutz, P.E., et al. (2006). Self-reported eating behaviors of extremely obese persons seekingbariatric surgery: A factor analytic approach. Obesity, 14(S3), 83S–89S.

Franks, S. F., & Kaiser, K. A. (2008). Predictive factors in bariatric surgery outcomes: whatis the role of the preoperative psychological evaluation? Primary Psychiatry, 15(8),74–83.

Geliebter, A., & Aversa, A. (2003). Emotional eating in overweight, normal weight, andunderweight individuals. Eating Behaviors, 3(4), 341–347.

Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis:A regression-based approach. New York, NY: Guilford Press.

Jones-Corneille, L. R., Wadden, T. A., Sarwer, D. B., Faulconbridge, L. F., Fabricatore, A. N.,Stack, R. M., et al. (2012). Axis I psychopathology in bariatric surgery candidateswith and without binge eating disorder: Results of structured clinical interviews.Obesity Surgery, 22(3), 389–397.

Kaiser, H. (1970). A second generation Little Jiffy. Psychometrika, 35(4), 401–415.Kaiser, H. (1974). An index of factorial simplicity. Psychometrika, 39(1), 31–36.Kalarchian, M. A., Marcus, M. D., Levine, M. D., Courcoulas, A. P., Pilkonis, P. A., Ringham, R.

M., et al. (2007). Psychiatric disorders among bariatric surgery candidates: Relationshipto obesity and functional health status. American Journal of Psychiatry, 164(2), 328–334.

Kalarchian, M. A., Marcus, M. D., Levine, M. D., Soulakaova, J. N., Courcoulas, A. P., &Wisinski, M. S. C. (2008). Relationship of psychiatric disorders to 6-month outcomesafter gastric bypass. Surgery for Obesity and Related Diseases, 4(4), 544–549.

Muhlhans, B., Horbach, T., & de Zwaan, M. (2009). Psychiatric disorders in bariatric sur-gery candidates: A review of the literature and results of a German prebariatric sur-gery sample. General Hospital Psychiatry, 31(5), 414–421.

Olatunji, B. O., Deacon, B. J., Abramowitz, J. S., & Tolin, D. F. (2006). Dimensionality of so-matic complaints: factor structure and psychometric properties of the Self-RatingAnxiety Scale. Journal of Anxiety Disorders, 20(5), 543–561.

Pawlow, L. A., O'Neil, P. M., White, M. A., & Byrne, T. K. (2005). Findings and outcomes ofpsychological evaluations of gastric bypass applicants. Surgery for Obesity and RelatedDiseases, 1(1), 523–529.

Radloff, L. S. (1977). The CES-D scale: A self report for depression scale for research in thegeneral population. Applied Psychological Measuresment, 1(3), 385–401.

Samuel, I., Mason, E. E., Renquist, K. E., Huang, Y. H., Zimmerman, M. B., & Jamal, M.(2006). Bariatric surgery trends: an 18-year report from the International BariatricSurgery Registry. The American Journal of Surgery, 192(5), 657–662.

Sarwer, D. B., Cohn, N. I., Gibbons, L. M., Magee, L., Crerand, C. E., Raper, S. E., et al. (2004).Psychiatric diagnoses and psychiatric treatment among bariatric surgery candidates.Obesity Surgery, 14(9), 1148–1156.

Tapper, K., Shaw, C., Ilsley, J., Hill, A. J., Bond, F. W., & Moore, L. (2009). Exploratoryrandomised controlled trial of a mindfulness-based weight loss intervention forwomen. Appetite, 52(2), 396–404.

Thompson, B. (2004). Exploratory and confirmatory factor analysis: understanding conceptsand applications. Washington, DC: American Psychological Association.

Wadden, T. A., & Foster, G. D. (2006). Weight and lifestyle inventory (WALI). Surgery forObesity and Related Diseases, 2(2), 180–199.

Walfish, S. (2004). Self-assessed emotional factors contributing to increased weight gainin pre-surgical bariatric patients. Obesity Surgery, 14(10), 1402–1405.

Walfish, S., & Brown, T. A. (2009). Self-assessed emotional factors contributing to in-creased weight in presurgical male bariatric patients. Bariatric Nursing and SurgicalPatient Care, 4(1), 49–52.

Walfish, S., Vance, D., & Fabricatore, A. N. (2007). Psychological evaluation of bariatric sur-gery applicants: Procedures and reasons for delay or denial of surgery. ObesitySurgery, 17(12), 1578–1583.

Zung, W. W. (1971). A rating instrument for anxiety disorders. Psychosomatics, 12(6),371–379.