Part of this work was presented at the 15th Annual Meeting of the Eating Disorders Research Society in Brooklyn, NY.
Personality pathology in purging disorder and bulimia nervosa†
Version of Record online: 14 FEB 2011
Copyright © 2011 Wiley Periodicals, Inc.
International Journal of Eating Disorders
Volume 44, Issue 8, pages 735–740, December 2011
How to Cite
Brown, T. A., Haedt-Matt, A. A. and Keel, P. K. (2011), Personality pathology in purging disorder and bulimia nervosa. Int. J. Eat. Disord., 44: 735–740. doi: 10.1002/eat.20904
- Issue online: 9 NOV 2011
- Version of Record online: 14 FEB 2011
- Manuscript Accepted: 16 NOV 2011
- National Institute of Mental Health. Grant Numbers: R01 MH61836, R03 MH61320
- purging disorder;
- bulimia nervosa;
- personality pathology;
- eating disorders
To compare levels of personality pathology in women with purging disorder (PD), bulimia nervosa (BN), and controls and to compare women with PD and BN on associations between personality pathology and shared eating disorder features.
Women with BN (n = 73), PD (n = 48), and controls (n = 64) completed interviews and self-report questionnaires.
BN and PD were associated with significantly greater personality pathology compared to controls. Cluster C symptoms and trait anxiety were greater in BN compared to PD, but groups did not differ on Cluster B symptoms or impulsivity. Subjective binge episodes were associated with anxious and impulsive personality traits in PD but not BN. Purging in PD was associated with trait anxiety, while purging in BN was associated with impulsivity.
Although BN and PD share eating disorder features and personality disturbance, some of the underlying associations between these eating disorder and personality features differ between groups. © 2011 by Wiley Periodicals, Inc.
Purging Disorder (PD) is an eating disorder not otherwise specified1, 2 characterized by purging in the absence of Objective Binge Episodes (OBEs). Prior research supports the clinical significance and distinctiveness of PD, finding similar levels of overall eating pathology between women with PD and Bulimia Nervosa (BN)1, 2 and differences in appetite, disinhibition, and Cholecystokinin (CCK) response.3 In developing the nosological description of any psychiatric disorder, investigating personality pathology may provide important insights into the etiology, development, and maintenance of the syndrome. The purpose of the present study was to determine whether personality disorder features differentiate women with PD from women with BN and those with no eating pathology. Although PD and BN share several eating disorder features in common [e.g., recurrent purging, Subjective Binge Episodes (SBEs), and weight/shape concerns], differences in associations between these features and personality pathology may reveal further distinctions between syndromes with important treatment implications. Thus, a secondary aim was to explore whether PD and BN differ in associations between personality and eating pathology.
Substantial research has examined personality pathology in BN (for review, see Cassin and Von Ranson4). Generally, BN has been associated with a high degree of Cluster B personality disorders4, 5 and associated characteristics, such as high impulsivity, affective instability, and disinhibition.3, 6 In addition, numerous studies have found elevated rates of Cluster C disorders among BN groups compared with non-eating disorder controls.5, 6 Thus, BN may represent a manifestation of joint problems with affective instability, disinhibition, anxiety, and compulsive rigidity, which may contribute to the cycle of bingeing and purging.
The few studies that have examined personality pathology in PD have produced inconsistent findings. No differences in impulse control or personality disorder diagnoses were found between BN and PD in two previous studies,3, 7 and both groups demonstrated significantly more personality pathology than controls. Some studies comparing PD and BN on self-report measures of impulsivity have found greater impulsivity in BN,8, 9 while others have not.2, 7 A more consistent pattern has emerged when looking at anxious personality traits, with studies finding greater trait anxiety in women with BN compared to women with PD.7, 10 Importantly, the relatively small number of participants in studies comparing BN and PD may have limited statistical power for detecting differences, particularly for analyses based on personality disorder diagnoses. Indeed, studies utilizing continuous measures of personality pathology have been more likely to yield significant differences between PD and BN.7–10 Thus, looking at personality pathology symptoms as opposed to actual diagnoses in larger samples may provide a more powerful test of differences in personality pathology.
A premise for examining personality differences between BN and PD is that personality pathology may contribute to specific dysregulated eating behaviors that manifest themselves as different clinical presentations. However, no studies to date have examined how symptom configuration may influence associations between personality pathology and eating disorder behaviors common to BN and PD. The aims of the present study were (1) to compare the levels of personality pathology in women with PD, BN, and no eating pathology and (2) to compare the associations between personality pathology and shared eating disorder attitudes/behaviors in women with PD and BN. Given likely problems with statistical power in prior studies, the current study combined samples from two previously published studies3, 7 and examined personality disorder symptoms on self-report measures to provide more sensitive assessments. This approach reflects increasing evidence for the validity of dimensional over categorical approaches to personality pathology.11 Although the lack of consistent personality differences in the literature suggests that women with PD and BN may share similar anxious and impulsive personality traits, evidence for differential associations between these traits and eating disorder features may reflect distinctions between the disorders with important treatment implications.
Data were drawn from previously established community-based samples of women with BN (n = 73), PD (n = 48), and non-eating disorder controls (n = 64) who completed measures necessary for current analyses. Detailed information about the sampling procedures has been reported elsewhere.3, 7 Participants were between 18 and 45 years and had a body mass index (BMI) between 18.5 and 26.5 kg/m2, based on objective height and weight measurements. Eating disorder diagnoses were determined through the Structured Clinical Interview for DSM-IV axis I Disorders (SCID-I) and the Eating Disorder Examination (EDE). Specific inclusion criteria for BN were DSM-IV criteria for BN purging type. Specific inclusion criteria for PD were equivalent to BN purging subtype criteria with the exception that, in PD, purging occurs in the absence of OBEs. Exclusion criteria for PD participants were a lifetime history of DSM-IV BN or binge eating disorder. Exclusion criteria for controls were a lifetime history of eating disorder symptoms based on SCID-I and EDE assessments. Groups did not differ significantly in age [Mean (SD) = 23.15 (5.23); F(2,190) = 0.266, p = 0.77], BMI [Mean (SD) = 22.05 (1.73); F(2,187) = 0.619, p = 0.54], or ethnicity (83% Caucasian; χ2(8) = 9.965, p = 0.27).
Procedure and Measures
Participants completed in-person clinical interviews and questionnaires during a single assessment. Additionally, objective height and weight were obtained using a digital scale and wall-mounted ruler to determine BMI.
Eating Disorder Examination (EDE).12
The EDE is a semi-structured clinical interview that assesses eating disorder symptoms and features. For the current study, frequency of SBEs, purging frequency, and weight/shape concerns were taken from the EDE. Inter-rater reliability (Pearson's r) and internal consistency (Cronbach's alpha) were r = .995, alpha = 0.89 for Shape Concern, and r = .999, alpha = 0.93 for Weight Concern. Because body weight and shape concerns were highly correlated in the current study (r = .95, p < .001), a composite measure was created combining the weight and shape concern subscales (EDEwsc; Cronbach's alpha = 0.96). Inter-rater Reliability for an eating disorder diagnosis was K = 0.96.
SCID-II Personality Questionnaire (SCID-II PQ).13
The SCID-II PQ is a 119 item self-report questionnaire used to assess personality pathology. This study examined Cluster B and C symptoms. The SCID-II-PQ has been validated for use as a standalone instrument.14 In this study, we combined symptoms within clusters to capture two dimensions of personality dysfunction: impulsive/emotionally dysregulated (Cluster B) and anxious/rigid (Cluster C). Internal consistency was good for Cluster B (alpha = 0.89) and Cluster C (alpha = 0.78) symptom scores. Participants in the present study also completed SCID-II interviews, which have been reported elsewhere.3, 7 Supporting the concurrent validity of the SCID-II PQ, Cluster B symptom counts were significant higher in individuals with an interview-based Cluster B diagnosis [F(1,182) = 5.063, p < .001], and Cluster C symptom counts were significantly higher in those with an interview-based Cluster C diagnosis [F(1,184) = 5.693, p < .001].
Barratt Impulsiveness Scale-11 (BIS-11).15
The BIS-11 is a 30-item questionnaire that provides a comprehensive assessment of problems in impulse control. Previous research indicates that this measure has good internal reliability and good concurrent and criterion validity.15 Internal consistency in the present study was good (alpha = 0.85).
State-Trait Anxiety Inventory (STAI).16
The STAI is a 40-item questionnaire of current anxiety (state) and anxiety proneness (trait) with good concurrent validity and high test–retest reliability.16 As our aim was to examine personality characteristics, only the Trait subscale was examined. Previous studies have established the use of the Trait subscale as a measure of personality17 and this measure has demonstrated concurrent validity with personality measures of neuroticism.18 Internal consistency for the present study was good (alpha = 0.95).
Analyses of Variance (ANOVA) were conducted to examine group differences on personality measures and eating disorder features. To address our second aim, we examined correlations between personality and eating disorder features to explore whether these relationships differ as a function of eating disorder group. Correlation coefficients were compared using equations from Rosenthal and Rosnow.19 Given that BN and PD differ on presence of OBEs by definition, we only included direct comparisons on eating disorder features that can be shared by BN and PD (SBEs per week, purging episodes per week, and body image disturbance). To control for Type I error, a Bonferroni-adjusted p-value of .0167 was used to evaluate statistical significance of post-hoc comparisons.
Table 1 presents comparisons among BN, PD and controls on measures of personality pathology and disordered eating attitudes and behaviors. Significant group effects were found on all measures of personality and eating pathology. Both eating disorder groups had significantly greater pathology than controls on all personality and eating measures (all p values < 0.005). Women with BN and PD did not differ significantly on measures of impulsivity/emotional dysregulation (as measured by the BIS-11 and Cluster B symptomatology). In contrast, women with BN endorsed greater trait anxiety and greater Cluster C symptoms on the SCID-II PQ than did women with PD. Women with BN endorsed a lower number of SBEs and higher number of purging episodes per week than women with PD. No differences were found between BN and PD on EDE weight and shape concerns (EDEwsc).
|Measure||BN (n = 77)*||PD (n = 49)*||Control (n = 68)*||F(2,191)*||p|
Table 2 presents correlation analyses between personality and eating disorder features, separately, in women with BN and PD. Cluster B symptoms were not significantly correlated with purging frequency in either BN or PD, and associations did not differ between groups. Cluster B symptoms were significantly positively correlated with frequency of SBE episodes in PD, but not BN, and the magnitude of correlations was significantly greater in PD than BN. Additionally, Cluster B symptoms were not significantly correlated with EDEwsc in either BN or PD, and associations did not differ between groups. Finally, no significant association was found between Cluster B personality traits and weight/shape concerns in control women (r = .139, p = .28), and these associations did not differ from those found in BN or PD (all p values > .26).
|BN (n = 76)||PD (n = 49)|
|Cluster B with|
|Cluster C with|
Correlations between Cluster C symptoms and purging frequency were significant in PD but not BN; however, the magnitude of effect sizes did not differ significantly between groups. No significant associations were found between Cluster C symptoms and SBE frequency in BN or PD. Additionally, correlations between Cluster C personality traits and weight/shape concerns were not significant in BN or PD but were significant in control women (r = .305, p = .015).
Correlations between BIS-11 scores and purging frequency were significant in BN but not PD and the effect sizes did not significantly differ between BN and PD. No other significant associations were found between BIS-11 scores and disordered eating in BN or PD. Similarly, no significant association was found between impulsive traits and weight/shape concerns in control women (r = −.014, p = .91).
Correlations between STAI-Trait and purging frequency were significant in PD but not BN; however, the magnitude of effect sizes did not differ between groups. STAI-Trait scores were significantly correlated with SBE frequency in PD but not BN, and correlations differed significantly between PD and BN in magnitude. STAI-Trait was also significantly correlated with weight/shape concerns for BN, PD, and control women (control: r = .365, p = .003), with no differences in effect size magnitude among groups.
Although several significant associations were found between personality and SBEs in PD, no significant associations were found for SBEs in BN. This may be, in part, because individuals with BN also experience OBEs. Therefore, we examined associations of OBEs and measures of personality pathology in BN to explore whether associations with OBEs in BN were comparable with those found for SBEs in PD. OBEs were significantly associated with Cluster C symptoms (r = .34, p = .003), BIS-11 scores (r = .30, p = .008), and STAI-Trait scores (r = .29, p = .013) and approached significance for Cluster B symptoms (r = .22, p = .064) in women with BN.
Results show that both BN and PD are associated with significantly greater personality pathology than controls. While BN and PD did not differ on measures of impulsivity/emotional dysregulation or EDE weight and shape concerns, differences in measures of anxious/rigid personality traits and purging episodes per week signified greater pathology in BN, suggesting that increased number of eating disorder symptoms is associated with greater personality and eating pathology. This is consistent with Wade's2 suggestion that PD resides on a continuum with BN. However, results from correlations support a more nuanced interpretation.
In PD, anxiety seems to be the core underlying personality correlate of eating disorder behaviors. Purging in PD was associated with anxious/rigid personality, suggesting that increased anxiety proneness may contribute to exaggerated concerns about the effects of normal or small amounts of food on weight/shape and the propensity to purge in PD. SBEs in PD also were associated with anxiety (STAI-Trait) as well as impulsivity/emotional dysregulation (Cluster B symptoms). Thus, increased SBE frequency may be associated with greater personality disturbance and could be examined for their association with treatment outcomes.20 Although SBEs have not been reliably included in research definitions of PD,21 associations with these personality traits implies that SBEs may be clinically relevant.
In BN, we found a robust correlation between impulsivity (BIS-11) and purging frequency supporting previous research linking purging to impulsivity among those who binge.4, 22 Purging in BN may reflect an impulsive need to rid oneself of food following a binge episode or may be associated with impulsivity through a third underlying variable—frequency of OBEs. Although SBEs had no significant personality correlates in BN, both anxious/rigid personality measures (Cluster C and STAI-Trait) and impulsivity (BIS-11) were correlated with OBE frequency, suggesting that the combination of anxious and impulsive personality traits may underlie OBEs in BN just as these same features may underlie SBEs in PD.
Purging is a shared feature between PD and BN; however, the personality features underlying this behavior differ, supporting the idea that BN and PD may be distinct disorders. In PD, purging may emerge from a combination of feelings of excessive postprandial fullness3 and anxiety. In BN, purging may represent a more direct consequence of OBEs that emerge from a combination of blunted CCK response3 and problems with impulse regulation.
Although correlates of SBEs differed between PD and BN, we found similar underlying personality profiles for OBEs in BN and SBEs in PD as both were significantly associated with anxious and impulsive personality traits. For individuals with BN an SBE may reflect a less significant loss of control as the episode is smaller than an OBE. This may help explain why SBEs in BN are not associated with impulsivity/emotional dysregulation. In contrast, women with PD have shown greater satiety and postprandial fullness than BN.3 As such, an SBE for individuals with PD may represent a significant loss of control. The combination of anxiety and impulsivity may contribute to loss of control eating in both BN and PD with the amount of food consumed reflecting different physiological capacities. Thus, just as similar eating disorder behaviors in BN and PD can have different personality correlates, similar personality profiles may be expressed though different behaviors in BN and PD.
This study had several methodological strengths. Data were drawn from two previous community-based studies, resulting in sample sizes capable of detecting moderate effect sizes. Groups were similar on age, BMI, and ethnicity thus reducing the potential influence of confounds in group comparisons often seen in comparisons of other eating disorders. The study used multiple continuous measures of personality pathology that demonstrated good internal consistency, further increasing power for analyses and reducing the risk of type II error. Further, to our knowledge, this is the first study to compare associations between eating disorder features and personality traits between PD and BN.
However, the study had weaknesses as well. There were limitations in the measures used to assess personality. Due to the good internal consistency of the personality pathology measures in this study, the BIS-11, STAI-Trait, Cluster B and Cluster C were all treated as unidimensional measures. However, only the STAI-Trait accurately represents a unidimensional measure, with the BIS-11, Cluster B and Cluster C measuring broader ranges of personality disturbance. This may explain the superior performance of the STAI-Trait across analyses. Future studies should investigate the replicability of the current findings using unidimensional measures of personality pathology. While our sample sizes were adequate for detecting moderate effect sizes, there may be additional group differences of a smaller effect size that we could not detect. This may explain why some of the comparisons between BN and PD, such as the correlations between purging and Cluster C symptoms (BN r = .12, PD r = .37) and purging and BIS-11 scores (BN r = .33, PD r = .09) were not significant despite qualitative differences in effect size magnitude. In addition, analyses were based on cross-sectional data. Thus, it is not possible to determine whether these personality traits represent a cause or consequence of the disorder. Future prospective studies should examine whether these personality traits precede the onset of the eating disorder.
The current results may have implications for treatment by providing potential targets for intervention. The association between anxious personality features and frequency of SBEs and purging in PD suggests that these individuals may benefit from treatments that specifically target anxiety, such as exposure and response prevention therapies. In addition, the similar personality profile of anxious and impulsive traits underlying SBE frequency in PD and OBE frequency in BN implies that although SBEs are not required for a diagnosis of PD, they may be clinically relevant and might be included as a diagnostic specifier in research criteria. As high Cluster B symptoms have been shown to predict poorer outcome in BN,20 future studies examining course and outcome in PD should examine the association between SBEs and outcome.
In summary, this study suggests that although BN and PD share specific eating disorder features and levels of personality disturbance, the underlying associations between some personality and eating disorder features differ between groups. Findings imply that PD may not simply be BN without OBEs, providing further support that PD represents a clinically significant disorder of eating that appears to be distinct from BN. Finally, findings point to potential targets for intervention and the clinical significance of SBEs in PD.
Earn CE credit for this article!
- 12The eating disorder examination. In: Fairburn CG,Wilson GT. Binge Eating: Nature, Assessment, and Treatment, 12th ed. New York: Guilford Press, 1993, pp 317–331., .
- 13SCID-II Personality Questionnaire. Washington, DC: American Psychiatric Press, 1997., , , , .
- 16The State-Trait Anxiety Inventory. Palo Alto: Consulting Psychologists Press, 1983., , , , .
- 19Essentials of Behavioral Research: Methods and Data Analysis, 2nd ed. New York: McGraw-Hill, 1991., .