ORIGINAL ARTICLE
Personality Disorders in Eating Disorder Not Otherwise Specified and Binge Eating Disorder A Meta-analysis of Comorbidity Studies Oddgeir Friborg, PhD,*Þ Monica Martinussen, PhD,þ Sabine Kaiser, MSc,* Karl Tore Øverga˚rd, PsyD,* Egil W. Martinsen, MD, PhD,§ Pho¨be Schmierer, MSc,* and Jan Harald Rosenvinge, PhD* Abstract: A meta-analysis was conducted to identify the proportion of comorbid personality disorders (PDs) in patients with eating disorder not otherwise specified (EDNOS) and binge eating disorder (BED). A search identified 20 articles in the period of 1987 to 2010. For EDNOS and BED, the comorbid proportions for any PD were 0.38 and 0.29, respectively; for cluster C PDs, 0.38 and 0.30, respectively (avoidant PD, 0.18 and 0.12, and obsessive-compulsive PD, 0.11 and 0.10, respectively); and for cluster B PDs, 0.25 and 0.11, respectively (borderline, 0.12 and 0.10). This pattern converged with findings on anorexia nervosa and bulimia nervosa, except being lower. Because the comorbidity profiles for EDNOS and BED were highly similar, their underlying PD pathology seems similar. Few moderators were significant, except for interviews yielding lower estimates than that of questionnaires. The variance statistic for any PD comorbidity was wide for EDNOS and narrow for BED, thus partly supporting BED as a distinct eating disorder category and EDNOS as a potentially more severe condition than BED. Key Words: Personality disorders, eating disorder not otherwise specified, binge eating disorder, meta-analysis, comorbidity (J Nerv Ment Dis 2014;202: 119Y125)
E
ating disorder not otherwise specified (EDNOS) is a separate diagnostic category in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV; American Psychiatric Association [APA], 1994), comprising conditions in which the diagnostic criteria for anorexia nervosa (AN; i.e., degree of weight loss) or bulimia nervosa (BN; i.e., the duration or frequency of binging and vomiting) are not fully met. The eating disorder symptoms may be atypical, such as vomiting after a minor intake of food or repeatedly chewing but not swallowing large amounts of food. EDNOS is the most frequent eating disorder (Hay et al., 2008; Hudson et al., 2007), accounting for more than half of all current eating disorders in outpatient clinics (Fairburn and Cooper, 2007; Martin et al., 2000; Rockert et al., 2007) and up to 90% of psychiatric community-based treatment-seeking samples if binge eating disorder (BED) is included (Zimmerman et al., 2008). In a meta-analysis, the general level of psychopathology was comparable for EDNOS and AN, although patients with BN reported somewhat higher levels of pathology (mean d = 0.19) than patients with EDNOS (Thomas et al., 2009). Some subtypes of EDNOS, particularly the purging *Department of Psychology, Faculty of Health Sciences, University of Tromsø, Tromsø, Norway; †Department of Psychiatric Research, University Hospital of North Norway, Tromsø, Norway; ‡The Regional Centre for Child and Youth Mental Health and Child Welfare, Faculty of Health Sciences, University of Tromsø, Tromsø, Norway; and §Clinic for Mental Health, Oslo University Hospital, Oslo, Norway. Send reprint requests to Oddgeir Friborg, PhD, Department of Psychology, Faculty of Health Sciences, University of Tromsø, N-9037 Tromsø, Norway. E-mail: [email protected]. Copyright * 2014 by Lippincott Williams & Wilkins ISSN: 0022-3018/14/20202Y0119 DOI: 10.1097/NMD.0000000000000080
The Journal of Nervous and Mental Disease
one, may show a wider spectrum of clinical severity than that of AN and BN (Keel and Striegel-Moore, 2009). EDNOS can be a transitional stage to recovery (Agras et al., 2009), as well as to AN or BN (Milos et al., 2005). BED is listed as an appendix diagnosis in DSM-IV (APA, 1994). It is subsumed under EDNOS but is increasingly recognized as a distinct diagnostic disorder on the basis of taxometric analyses (Williamson, 2007), family aggregation studies (Javaras et al., 2008), treatment response research (Wilson et al., 2007), and studies of clinical course (Pope et al., 2006). Two extensive reviews (StriegelMoore and Franko, 2007; Wonderlich et al., 2009) also conclude positively with respect to the merit of BED as a distinct eating disorder diagnosis, and hence, it is now proposed as a separate ED diagnosis in the DSM-5 (APA, 2012). Although BED is currently part of the EDNOS category, our further use of the EDNOS acronym in the present article does not include studies on patients with BED. The provisional DSM-IV criteria for BED include recurrent binge-eating episodes defined as rapid eating in a discrete period characterized by poor self-regulation (e.g., not being able to stop binging). This is usually not preceded by feelings of hunger or satiety (Sysko et al., 2007). It departs from BN in terms of absence of extreme compensatory weight-control behaviors after a binge-eating episode and also has a less severe natural course than BN (Fairburn et al., 2000). Self-disgust and shame are typical concomitant feelings, and thus, patients often eat alone to reduce social embarrassment and distress. The transition from restrictive AN to BED is more common (Eddy et al., 2008) than the reverse process. Axis II personality disorder (PD) comorbidity has been found to worsen the recovery prognosis from AN, BN, and EDNOS (Rø et al., 2005); from BED (Masheb and Grilo, 2008); and in a large clinical study (N = 629) from any type of eating disorders (Helverskov et al., 2010). However, studies on the natural course of eating disorders (Grilo et al., 2007) do not indicate different rates of remission or relapse depending on a PD diagnosis, except for the avoidant PD among patients with EDNOS. Because avoidant and obsessivecompulsive PDs are reported as the most typical for patients with BED (Becker et al., 2010), cluster C comorbid PDs may be particularly relevant for the treatment prognosis of both EDNOS and BED. Eating disorders and PDs, furthermore, share important treatmentcomplicating clinical features, such as a profound self-devaluation; feelings of low mood and anxiety; as well as a poor self-insight into behavioral, cognitive, and interpersonal dysfunctional patterns that may reciprocally reinforce the severity of either condition (Chen et al., 2011; Kendall and Clarkin, 1992). Given the prevalence of BED and EDNOS in community and clinical settings, and considering how a PD may complicate treatment and worsen the prognosis, knowledge of comorbid patterns of PDs is highly warranted. The between-study variation in comorbid proportions (21% to 97%) of PDs in eating disorders is, however, large (Cassin and von Ranson, 2005; Sansone et al., 2005; Vitousek and Manke, 1994). This may also apply to BED and EDNOS. A metaanalysis may determine the mean comorbidity proportions more
& Volume 202, Number 2, February 2014
www.jonmd.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
119
The Journal of Nervous and Mental Disease
Friborg et al.
accurately and identify whether these estimates covary with methodological factors that need to be addressed in future studies. Moderators such as sample differences, variation in diagnostic criteria, and the methods of assessment may all account for the comorbidity variation (Sunday et al., 2001). For instance, self-report assessments yield consistently higher proportions of PDs compared with the use of structured clinical interviews (Echeburu´a and Maran˜on, 2001; Ramklint et al., 2010; Rosenvinge et al., 2000). Other sources may be the diagnostic system used and the type of patients recruited (e.g., inpatients versus outpatients). For example, in a meta-analysis on AN and BN (Rosenvinge et al., 2000), the PD comorbidity rate varied from 49% in outpatient to 75% in inpatient samples. Age of onset of an eating disorder is of particular interest as a possible moderator because an early onset (e.g., in teenage years) is related to more psychopathology in general (Wentz et al., 2009), which may impair personality development, whereas a late debut implies a fewer number of cluster C PDs (Cumella and Kally, 2008). One may therefore expect that an early age of onset would yield a higher proportion of any comorbid PD. Given the support for BED as a clinically relevant diagnosis (Javaras et al., 2008; Pope et al., 2006; Williamson, 2007; Wilson et al., 2007), one could expect a higher proportion of PDs in BED than in the clinically more heterogeneous EDNOS category (excluding BED), in which more patients are in a process of recovery compared with BED (Agras et al., 2009). However, because the levels of eating pathology and general psychopathology between EDNOS and BED have been reported comparable in a previous meta-analysis (Thomas et al., 2009), there is insufficient evidence to make any strong predictions. As the first study in the literature, we used a meta-analytic approach to assess the respective frequencies of comorbid axis II PD diagnoses between EDNOS and BED. In addition, the impact of sample characteristics, diagnostic criteria, type of assessment methods, body mass index (BMI), and age of onset for PD comorbidity was examined with regard to the clinical usefulness of the forthcoming DSM-5 diagnostic distinctions in terms of pathological personality functioning.
METHODS Literature Search The databases PsychINFO, EMBASE, and MEDLINE were searched for published articles in English or German language between 1987 (when EDNOS was introduced in the DSM-III-R) and June 2010. The search included the key words eating disorders or anorexia or bulimia and personality disorders and comorbidity. Initially, 459 articles were located, and these were supplemented by 37 articles from reference lists of previous meta-analyses and located articles (k = 33) and from a predecessor of OVID (i.e., SilverPlatter; k = 4). From the initial pool, 369 articles were removed because the patients either had been diagnosed with drug addiction or had recovered from their eating disorder or because the results were published as dissertation abstracts. From the remaining 127 studies, 42 were removed because these did not report proportions of PDs, reported the same results in multiple articles (the most recent were selected), or included subjects younger than 18 years. Studies that used PDs as an inclusion criterion were removed, as were those that failed to report current eating disorder diagnoses or diagnostic information. This yielded a pool of 88 articles, of which 19 articles had a suitable reporting of the comorbidity between PDs and the two forms of eating disorders in question. After this filtering process, an additional search was specifically done for the key word binge eating, which added one extra study not already in the study pool, thus yielding 11 articles addressing EDNOS and 9 addressing BED. 120
www.jonmd.com
& Volume 202, Number 2, February 2014
Coding Procedure and Interrater Reliability From each study, we coded publication year, comorbid events of PDs (converted to a proportion), sex (proportion of women participating), sample size, and mean age. The following variables were coded as moderators: sample characteristics (clinical versus community); diagnostic system (DSM-III-R versus DSM-IV); assessment methods for diagnosing BED, EDNOS, and PDs (structured interview, clinical assessment, and questionnaires); age at onset and duration of eating disorders; and BMI. In case of missing information, the corresponding authors were contacted, and in most cases, they provided the necessary data. The articles were coded by two psychology students trained by the same supervisor (J. H. R.). Disagreements in coding were discussed until consensus was reached. To assess interrater reliability, 17 (19%) of the 88 studies were randomly selected and coded by the first author (O. F.). Interrater agreement was estimated as kappa for categorical data and intraclass correlation coefficient (ICC) for continuous data. Kappa was 0.89 for type of sample (inpatient, outpatient, both, or community) and 1.0 for diagnostic method (interview versus questionnaire). ICC was perfect (1.00) for proportions of clusters A, B, and C PD and for the diagnostic categories of PD (except borderline disorder being lower at 0.97); for number of patients; for proportion of women; and for BMI. ICC was nearly perfect for mean age (0.98), age at onset (0.95), and duration (0.99) of eating disorders.
Procedure and Statistical Analysis The meta-analyses were conducted using the program Comprehensive Meta-Analysis version 2.2.057 (released December 2010; Borenstein et al., 2005). The mean weighted event rate (PD cases/ sample size) represented comorbid proportions and was used as effect sizes in the analyses. Calculations were based on a randomeffects model because the true proportion may vary across studies because of variations in factors such as severity or age of onset. Hence, variance in mean proportions is partitioned in two sources: a) variance within studies (?, sampling error) and b) variance between studies (5, variation in true proportions) (Borenstein et al., 2005; Hedges and Vevea, 1998). Each study was weighted by calculating the inverse of these variance components: 1/(? + 5). Small studies imply a larger within-study variance component (? = R2/n), thus lowering the weight given to data from small samples. A Q test statistic is used to test whether the variance between studies is larger than the variance within studies, thus indicating that variation in proportions truly exists from which predictors (or moderators) may be sought. The random-effects model reduces the type I error rate and enhances the generalizability of the findings (Field, 2003). An analogous Q test statistic for between-group differences may be conducted to examine whether the differences in proportions between two samples depend on a categorical covariate (moderator).
RESULTS Sample Characteristics The study pool consisted of nine BED studies (n = 838) and 11 EDNOS studies (n = 743; see Table 1 for descriptive characteristics). Among the EDNOS studies, eight studies were carried out on clinical samples (n = 373, 50%); two studies, on community samples (n = 337, 45%); and one study, on an unspecified sample (n = 34, 5%). Five studies based the diagnosis of PD on a structured interview (n = 136, 18%), five studies used clinical assessment (n = 294, 40%), and one study used questionnaires (n = 313, 42%). The DSM-IV system was used in four studies (n = 255, 34%); and the DSM-III/ DSM-III-R, in two studies (n = 337, 45%). It was unreported for five studies (n = 152, 21%). The mean age of the patients was 29.5 years (SD, 10.7). * 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
The Journal of Nervous and Mental Disease
& Volume 202, Number 2, February 2014
Personality Disorders in BED/EDNOS
TABLE 1. Overview of the Included Studies of EDNOS and BED Author, Year
N
Cooper et al., 1988 11 Yager et al., 1989 313 Schmidt and Telch, 23 1990 Hay and Hall, 1991 10 Herzog et al., 1995 32 Carlat et al., 1997 43 Striegel-Moore et al., 56 1999 Milos et al., 2003 34 Rø et al., 2005 22 Maran˜o´n et al., 2007 24 Godt, 2008 174 Yanovski et al., 1993 43 Specker et al., 1994 43 Telch and Stice, 1998 61 Wilfley et al., 2000 162 Stice et al., 2001 159 145/228 Grilo, 2002, 2004b Van Hanswijck et al., 15 2003 Picot and Lilenfeld, 50 2003 Masheb and Grilo, 75 2008
Sample
Mean Age
EDNOS EDNOS EDNOS
Out Comm Comm
25.7 24.7 19.0
100 100 100
19.3 V V
20.2 14.9 V
6.7 9.8 V
CA Q I
CA Q I
V III III-R
EDNOS EDNOS EDNOS EDNOS
In In and out In and out In
28.4 V V 56.7
72 V 0 0
V V V V
V V 19.1 V
V V V V
I Q+I CA CA
CA I CA CA
V V III, III-R, IV Othera
21.3 V V V 41.8 35.9 34.8 37.1 36.3 V 41.9
17.6 16.0 V V V V 18.0 V V V V
10.5 13.7 V V V V 25.5 V V V V
I I I Q+I I I I Q+I I I I
I I I
IV IV IV
I Q I I I I I
III-R III-R III-R III-R III-R IV IV
EDNOS EDNOS EDNOS EDNOS BED BED BED BED BED BED BED
% Onset Duration Assessment Assessment Women BMI of ED, yrs of ED, yrs of ED of PD
DSM Edition
Eating Disorder
In, out, and comm 28.1 100 In 29.7 100 Out V 100 In 23.8 100 Comm 36.1 77 Comm 39.2 100 Comm 43.5 100 Comm 45.2 83 Comm 40.0 100 Out 43.7/44.5 78/79 Out 36.9 100
BED
Comm
41.3
92
34.7
V
V
I
I
IV
BED
Out
46.0
81
35.3
28.1
V
I
I
IV
a
ICD-9-CM. Grilo has published two studies from the same sample. Data from the first study were included only if not reported in the second study. CA indicates clinical assessment; comm, community sample; in, inpatients; out, outpatients; Q, questionnaire; I, structured interview. b
For BED, 59% were community samples (n = 519, k = 6) and 41% were clinical samples (n = 319, k = 3). Only one and two studies reported duration of illness and age of onset, respectively. In contrast to the EDNOS studies, structured clinical interviews to diagnose a PD were dominant in eight studies (n = 795, 95%), whereas only one study used questionnaires (n = 43, 5%). Moreover, all subjects were assessed using either DSM-III-R (n = 469, k = 5, 56%) or DSM-IV (n = 369, k = 4, 44%) diagnostic criteria. The patients had a mean age of 41.6 years (SD, 3.3). BMI was reported in six BED studies (mean, 37.2; SD, 2.8) but only by two EDNOS studies (mean, 20.3; SD, 1.0).
Comorbidity Related to PD Clusters The mean proportions of comorbid PDs for EDNOS and BED as well as associated confidence intervals (CIs) and heterogeneity statistics are presented in Table 2. The mean proportion of patients having at least one PD (any PD) was higher for EDNOS (0.38) than for BED (0.29), but the difference was not statistically significant (Q = 2.23, p = 0.14). The variation in proportions for any PD was considerably larger among EDNOS (95% CI, 0.26Y0.51) than among BED studies (95% CI, 0.24Y0.33), also evidenced by a Qwithin test that was significant only for the EDNOS group. At PD cluster levels, no significant differences in mean proportions between EDNOS and BED emerged. The largest difference was seen in cluster B (0.25 and 0.11, respectively). Among the EDNOS studies, the comorbid proportions increased from cluster A, B, to C (0.09, 0.25, and 0.38, respectively), but it was not significant (Qdf = 2 = 4.93, p = 0.09). A similar trend was present for the * 2014 Lippincott Williams & Wilkins
BED studies (0.07, 0.11, and 0.30, respectively) but again statistically nonsignificant (p = 0.09). If BED and EDNOS were combined, the increase in proportions of clusters A, B, and C (0.08, 0.18, and 0.34, respectively) was significant (Qdf = 2 = 10.74, p G 0.01). The heterogeneity statistic (Qwithin) was significant for all PD clusters within EDNOS and BED, indicating that sampling error was not the sole reason for the variations in proportions.
Comorbidity Related to Specific PD Diagnoses No significant differences in mean proportions of PD were found between EDNOS and BED for any of the specific PDs. Avoidant PD was most frequent in EDNOS (0.18), followed by borderline (0.12) and obsessive-compulsive (0.11) PD. The lowest proportions were found for narcissistic PD (0.02) and the cluster AYspecific PDs (range, 0.01Y0.06). Avoidant PD was also most prevalent among the BED studies (0.12), followed by borderline and obsessive-compulsive PD (both 0.10). The lowest proportions were found for schizotypal (0.01), schizoid (0.02), and histrionic (0.02) disorders. The heterogeneity statistics (Qwithin) were significant for most specific PD proportions within both EDNOS and BED, except some of the rarer conditions (e.g., antisocial and schizoid PDs). The largest difference in PD proportions between EDNOS and BED was found for avoidant PD (0.18 and 0.12, respectively).
Moderator Analyses Type of sample influenced the comorbidity proportions minimally because significant differences between clinical and community www.jonmd.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
121
The Journal of Nervous and Mental Disease
Friborg et al.
TABLE 2. Proportions of PD Diagnoses for EDNOS and BED Based on a Random-Effects Model EDNOS Any PD Cluster A Paranoid Schizoid Schizotypal Antisocial Cluster B Borderline Histrionic Narcissistic Cluster C Avoidant Dependent Obsessive-compulsive BED Any PD Cluster A Paranoid Schizoid Schizotypal Antisocial Cluster B Borderline Histrionic Narcissistic Cluster C Avoidant Dependent Obsessive-compulsive
k
N
p (95% CI)
Qwithin
9 5 5 5 5 4 5 7 5 5 5 5 5 5
699 567 557 559 559 433 567 600 557 558 567 557 558 556
0.38 (0.26Y0.51) 0.09 (0.02Y0.31) 0.06 (0.02Y0.16) 0.03 (0.02Y0.04) 0.04 (0.00Y0.28) 0.01 (0.00Y0.03) 0.25 (0.10Y0.50) 0.12 (0.05Y0.25) 0.06 (0.02Y0.21) 0.02 (0.01Y0.03) 0.38 (0.25Y0.52) 0.18 (0.11Y0.30) 0.07 (0.02Y0.22) 0.11 (0.06Y0.20)
58.93*** 48.71*** 19.17*** 1.97 40.61*** 2.19 59.16*** 42. 23*** 28.36*** 1.45 28.06*** 17.95*** 35.20*** 15.19**
8 4 7 6 6 6 4 8 6 6 4 8 6 8
786 508 629 589 587 587 508 679 588 587 508 761 586 762
0.29 (0.24Y0.33) 0.07 (0.01Y0.34) 0.07 (0.03Y0.18) 0.02 (0.00Y0.07) 0.01 (0.01Y0.03) 0.03 (0.02Y0.05) 0.11 (0.02Y0.39) 0.10 (0.06Y0.16) 0.02 (0.00Y0.20) 0.03 (0.01Y0.08) 0.30 (0.21Y0.41) 0.12 (0.08Y0.17) 0.03 (0.01Y0.11) 0.10 (0.07Y0.16)
12.72 65.62*** 51.32*** 16.43** 5.22 3.31 69.15*** 24.95*** 63.89*** 15.20** 15.00** 22.24** 29.57*** 22.31**
**p G 0.01. ***p G 0.001. k indicates number of studies; p, mean proportion; Qwithin, heterogeneity statistic.
& Volume 202, Number 2, February 2014
samples emerged only for cluster A, which occurred more frequently in community samples. Similar tests for assessment method indicated that clinical interviews yielded consistently lower comorbid proportions than that of questionnaire-based methods (see Table 3). However, this comparison was based only on two studies in the questionnaire group. The use of different diagnostic systems (DSM-III or DSM-IV) was nonsignificant. A meta-regression analysis showed neither BMI (only estimable for BED) nor age of onset (only estimable for EDNOS) as significant moderators of the degree of PD comorbidity. Duration of eating disorder was not estimable (reported in only one BED and three EDNOS studies).
DISCUSSION Overall and Cluster Level Comparisons Having at least one PD of any kind (any PD) occurred more frequently in EDNOS (0.38) than in BED (0.29), but this could not be ascertained because of a statistical nonsignificant difference. None of the differences on the PD cluster and specific levels were significant either. The pathological personality functioning in BED and EDNOS thus stand out as more similar than different. The mean proportions of PDs showed an ascending increase across the clusters A, B, and C. Similar findings have been reported for AN (Herzog et al., 1992; Piran et al., 1988; Skodol et al., 1993; Wonderlich et al., 1990) and, in part, for BN (Herzog et al., 1992; Rosenvinge et al., 2000; Schmidt and Telch, 1990; Wonderlich and Mitchell, 1992; Yates et al., 1989). This may also indicate important similarities across the eating disorder diagnoses in terms of PD comorbidity. Although PD comorbidity occurs less frequently in BED and EDNOS than in AN and BN, the severity or impact on treatment may be of comparable concern when present. Moreover, the variance in the proportions of any PD was considerably wider for EDNOS than for BED, hence supporting the notion of EDNOS as a clinically more heterogeneous diagnosis than BED (Agras et al., 2009; Keel and Striegel-Moore, 2009; Milos et al., 2005). This implies that some patients with EDNOS may have more comorbidity conditions than patients with BED. Because cluster B PDs occurred more than twice as often in EDNOS than in BED, EDNOS stands out as a potentially more serious clinical condition than BED.
TABLE 3. Impact of Moderators on Proportions of PD Diagnoses in EDNOS and BED Moderators PDs a
ka
kb
na
nb
p (95% CI)a
p (95% CI)b
Qbetween
5
3c
448
518
0.05 (0.01Y0.14)
0.24 (0.08Y0.56)
4.21*
6 8 8 11 8 8
2c 2c 2c 2c 2c 2c
470 618 616 738 615 614
356 356 356 356 356 356
0.06 (0.03Y0.12) 0.01 (0.01Y0.03) 0.01 (0.00Y0.05) 0.09 (0.07Y0.12) 0.02 (0.01Y0.05) 0.02 (0.01Y0.04)
0.41 (0.20Y0.67) 0.05 (0.03Y0.09) 0.19 (0.07Y0.43) 0.30 (0.20Y0.41) 0.32 (0.14Y0.58) 0.26 (0.22Y0.31)
13.87*** 7.95** 15.24*** 20.48*** 19.71*** 72.10***
b
Clinical versus community Cluster A Interviewa versus questionnaireb Cluster A Schizoid Schizotypal Borderline Histrionic Dependent
All PD clusters and specific PDs were tested (only significant effects were reported). Diagnostic system (DSM-III-R versus DSM-IV) was tested but found nonsignificant. a Subgroup 1. b Subgroup 2. c The between-study variance (t2) was estimated as equal across both groups because of the low number of studies, especially in the questionnaire subgroup, consisting of one BED and one EDNOS study. *p G 0.05. **p G 0.01. ***p G 0.001. k indicates number of studies; p, mean proportion.
122
www.jonmd.com
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
The Journal of Nervous and Mental Disease
& Volume 202, Number 2, February 2014
Specific Level Comparisons The comorbidity profiles between BED and EDNOS were similar also on the specific PD level, with avoidant, borderline, and obsessive-compulsive disorders occurring in a descending order of frequency. Taken together, BED and EDNOS seem to share the PD comorbidity profile with AN and BN, and those differences mostly relate to the level of proportions. The most common comorbid PD in patients with BED were avoidant, obsessive-compulsive, and borderline PDs. These findings fit with previous studies on EDNOS and BED (Ramklint et al., 2010), as well as BED, AN, and BN (Cassin and von Ranson, 2005), and thus question the conclusions from a previous literature review (Sansone et al., 2005) claiming that obsessive-compulsive disorder is most typical with BED.
Moderator Analyses Contrary to a previous meta-analysis for AN and BN (Rosenvinge et al., 2000), the impact of clinical versus community samples was almost negligible. The only difference was cluster A PDs, which were more prevalent in community than in clinical samples. One explanation may be that treatment-seeking behaviors may be less likely to occur among patients with cluster A PDs, generally characterized as being distrusting and suspicious of others (APA, 1994). Still, the present results were more in line with the meta-analysis by Thomas et al. (2009), which reported no differences between EDNOS and BED in general psychopathology across patient and nonpatient samples. Similarly, Wilfley et al. (2000) found no differences in current and lifetime prevalence rates of psychiatric diagnoses or level of general psychiatric features between a clinical and a recruited treatment-seeking sample diagnosed with BED. However, community samples may be flawed with both overreporting and underreporting (Vitousek and Stumpf, 2006). When addressing the problems of treatment-seeking subjects by comparing frequencies of comorbid PDs among a BED-diagnosed clinical sample and a recruited community sample, Telch and Stice (1998) found a significantly higher relative risk for any axis I and axis II comorbidity in the clinical compared with the community sample. The lack of comparable differences in our meta-analysis may relate to the fact that 46% of the community samples consisted of treatmentseeking subjects, whereas the remaining subjects were recruited for research purposes. The use of self-report measures inflated the proportions. One caveat is the low number of studies using questionnaires, but the findings seem valid because these concur with numerous previous studies (e.g., Echeburu´a and Maran˜on, 2001; Godt, 2002; Ramklint et al., 2010; Rosenvinge et al., 2000). Indeed, the less use of self-report measures compared with, for instance, previous meta-analyses of PDs in AN and BN (Rosenvinge et al., 2000) is a positive trend in the literature in terms of improved scientific and clinical validity of findings. Unexpectedly, age of onset of an eating disorder did not moderate the proportions of any PD, although a declining trend was present (p = 0.13). Reasons for this finding may be the low number of studies (k = 4) reporting age of onset and the fact that these data are mostly available for younger age groups. However, available data across age groups (Cumella and Kally, 2008) also indicate a declining trend. Comparable studies have reported that earlier age of onset chronically impedes sexual functioning and libido (Pinheiro et al., 2010); reduces the likelihood of a more positive long-term outcome (Wentz et al., 2009); and, in general, implies more personality problems, a more profound body dissatisfaction, and a longer-lasting ascetic-driven pursuit of thinness (Abbate-Daga et al., 2007). An early debut may thus be a sizeable risk factor of developing an additional PD.
Strengths and Limitations A total of 30 studies were located in a meta-analysis (Thomas et al., 2009) comparing eating pathology and general psychopathology * 2014 Lippincott Williams & Wilkins
Personality Disorders in BED/EDNOS
between BED and EDNOS. Because PDs are less frequently studied than axis I symptoms, the identification of 20 studies in the current meta-analysis indicates a minor risk for having missed relevant studies. This point is supported by the fact that authors of publications were contacted and asked to provide unpublished data. Moreover, data comprehensibility was secured because authors were asked to provide data not appearing in their published articles. Finally, the findings were minimally affected by the coding procedure because the interrater reliability indicated high agreement. The moderator analyses were, however, hampered by the small number of total studies located, which prevented separate moderator analyses of EDNOS and BED. Moreover, an important sample bias has to be considered in the comparison of EDNOS and BED because 6 of 9 BED studies were based on community samples, whereas this was the case for only 2 of 11 EDNOS studies. Although the influence of sample type was modest, these sampling differences complicate interpretation of the differences. The higher heterogeneity (and slightly higher proportion of PDs) in EDNOS may therefore relate to these sample differences because levels of psychopathology are generally expected to be higher in clinical than in community samples. On the other hand, this concern is reduced considering that 46% of the community participants were also treatment-seeking subjects.
Implications for Further Research Future research on the comorbidity between eating disorders and PDs needs methodological improvements. This would include a systematic collection and reporting of demographic and clinical information, especially BMI, age of onset, and duration of both eating and PDs. This would certainly improve moderator analyses by allowing a simultaneous examination of multiple covariates. Second, future studies on EDNOS should provide more detailed information about the exact nature of the eating pathology, specifically about other eating disorder criteria that are not met. Third, to reduce error variance, we recommend using larger samples and similar recruitment and sampling procedures for both BED and EDNOS. Nevertheless, the present study indicates that structured clinical interviews are the preferred method, which certainly have improved the estimates of the true proportions of comorbid PDs. The previous (Cassin and von Ranson, 2005; Rosenvinge et al., 2000) and current meta-analyses suggest a remarkably consistent pattern of frequency across eating disorder diagnoses regarding distribution of proportions across PD clusters and specific PD diagnoses. This may support a transdiagnostic approach (Fairburn et al., 2003, 2008) to understanding the nature of eating disorders. This approach advocates the commonality of symptoms across eating disorder diagnoses and, hence, a more parsimonious approach than focusing on diagnostic differences of presumably remote clinical relevance. In addition, because the highest proportions were found for avoidant, obsessive-compulsive, and borderline PDs, this may support an inhibition-disinhibition dimension that is common across eating disorder diagnoses (Fassino et al., 2002; Rybakowsky et al., 2004), which is also relevant in a transdiagnostic understanding of eating disorders. The DSM-5 will probably be launched in 2014. At the DSM-5 Web site (http://www.dsm5.org/Pages/Default.aspx), several changes have been proposed for the disorders composing the present metaanalysis. In the DSM-5, the overall goal for the eating disorder classification is to restrict the spectrum of disorder clinical features now categorized as EDNOS. This is done by introducing more finegrained diagnostic labels and phenotypes under the heading Feeding or Eating Disorder Not Elsewhere Classified (FEDNEC; e.g., atypical AN, BN, or BED; a purging disorder; or a night eating syndrome). A recent study (Birgegard et al., 2012) indicates that this goal is realized using the proposed criteria. Another issue is whether www.jonmd.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
123
The Journal of Nervous and Mental Disease
Friborg et al.
more diagnostic labels are relevant in terms of specific psychopathology, clinical management, and outcome. Our findings that subjects with EDNOS do have comorbid PDs serve as an argument for FEDNEC as an overall diagnostic category. Future research, however, should study whether the clinical heterogeneity of PDs among subjects with the current EDNOS will persist or dissolve using the subgroup distinctions of FEDNEC. Another way of reducing heterogeneity of EDNOS is to include BED as an official DSM-5 diagnosis. The present results support this in the sense that comorbidity relates to well-known criteria for accepting a proposed clinical state as a diagnostic entity (Blashfield et al., 1990). Moreover, the variance in the comorbidity proportions is quite narrow compared with EDNOS, hence indicating a clinically more specific condition. The diagnostic criteria for BED will, however, also be more lenient, that is, by requiring binge eating once a week for 3 months compared with twice a week for 6 months in the current provisional DSM-IV criteria. However, preliminary studies indicate no substantial increase in the prevalence of BED (Hudson et al., 2012) and, hence, possibly in the proportion or heterogeneity of PD. Our findings are also relevant when considering the DSM-5 proposal for the organization of PD diagnoses. Although the DSM-5 will abandon the cluster model, the diagnostic-specific PD categories in the present study will be retained. New and more sophisticated models (Shedler et al., 2010; Tyrer et al., 2011) have been considered, which may offer new perspectives in studying comorbidity. DISCLOSURE The authors declare no conflict of interest.
REFERENCES Abbate-Daga G, Piero A, Rigardetto R, Gandione M, Gramaglia C, Fassino S (2007) Clinical, psychological and personality features related to age of onset of anorexia nervosa. Psychopathology. 40:261Y268. Agras WS, Crow S, Mitchell JE, Halmi KA, Bryson S (2009) A 4-year prospective study of eating disorder NOS compared with full eating disorder syndromes. Int J Eat Disord. 42:565Y570. American Psychiatric Association (1994) Diagnostic and statistical manual of mental disorders (DSM-IV) (4th ed). Washington, DC: American Psychiatric Association. American Psychiatric Association (2012) DSM-5 development. Arlington, TX: American Psychiatric Association. Retrieved from http://www.dsm5.org. Becker DF, Masheb RM, White MA, Grilo CM (2010) Psychiatric, behavioral, and attitudinal correlates of avoidant and obsessive-compulsive personality pathology in patients with binge-eating disorder. Compr Psychiatry. 51:531Y537. Birgegard A, Norring C, Clinton D (2012) DSM-IV versus DSM-5: Implementation of proposed DSM-5 criteria in a large naturalistic database. Int J Eat Disord. 45:353Y361. Blashfield RK, Sprock J, Fuller AK (1990) Suggested guidelines for including or excluding categories in the DSM-IV. Compr Psychiatry. 31:15Y19. Borenstein M, Hedges L, Higgins J, Rothstein H (2005) Comprehensive metaanalysis: Version 2.0. Englewood, NJ: Biostat. Carlat DJ, Camargo A, Herzog DB (1997) Eating disorders in males: A report on 135 patients. Am J Psychiatry. 154:1127Y1132.
& Volume 202, Number 2, February 2014
Echeburu´a E, Maran˜on I (2001) Comorbilidad de las alteraciones de la conducta alimentaria con los trastornos de personalidad. Psicologı´a Conductual. 9: 513Y525. Eddy KT, Dorer DJ, Franko DL, Tahilani K, Thompson-Brenner H, Herzog DB (2008) Diagnostic crossover in anorexia nervosa and bulimia nervosa: Implications for DSM-V. Am J Psychiatry. 165:245Y250. Fairburn CG, Cooper Z (2007) Thinking afresh about the classification of eating disorders. Int J Eat Disord. 40:S107YS110. Fairburn CG, Cooper Z, Doll HA, Norman P, O’Connor M (2000) The natural course of bulimia nervosa and binge eating disorder in young women. Arch Gen Psychiatry. 57:659Y665. Fairburn CG, Cooper Z, Shafran R (2003) Cognitive behaviour theory for eating disorders: A ‘‘transdiagnostic’’ theory and treatment. Behav Res Ther. 41: 509Y528. Fairburn CG, Cooper Z, Waller D (2008) ‘‘Complex cases’’ and comorbidity. In Fairburn CG (Ed), Cognitive behavior therapy and eating disorders (pp 245Y258). New York: Guilford Press. Fassino S, Abbate-Draga G, Amianto F, Leombruni P, Boggio S, Rovera G (2002) Temperament and character profile of eating disorders: A controlled study with the Temperament and Character Inventory. Int J Eat Disord. 32:412Y425. Field AP (2003) The problems in using fixed-effects models of meta-analysis on real-world data. Underst Stat. 2:77Y96. Godt K (2002) Personality disorders and eating disorders: The prevalence of personality disorders in 176 female patients with eating disorders. Eur Eat Disord Rev. 10:102Y109. Godt K (2008) Personality disorders in 545 patients with eating disorders. Eur Eat Disord Rev. 16:94Y99. Grilo CM (2002) Are there gender differences in DSM-IV personality disorders? Compr Psychiatry. 43:427Y430. Grilo CM (2004) Factorial structure and diagnostic efficiency of DSM-IV criteria for avoidant personality disorder in patients with binge eating disorder. Behav Res Ther. 42:1149Y1162. Grilo CM, Pagano ME, Skodol AE, Sanislow C, McGlashan TH, Gunderson JG, Stout RL (2007) Natural course of bulimia nervosa and of eating disorder not otherwise specified: 5-year prospective study of remissions, relapses, and the effects of personality disorder psychopathology. J Clin Psychiatry. 68:738Y746. Hay PI, Hall A (1991) The prevalence of eating disorders in recently admitted psychiatric in-patients. Br J Psychiatry. 159:562Y565. Hay PJ, Mond J, Nuttner P, Darby P (2008) Eating disorder behaviors are increasing. Findings from two sequential community surveys in South Australia. PloS One. 3:e1541. Hedges LV, Vevea JL (1998) Fixed- and random-effects models in meta-analysis. Psychol Methods. 3:486Y504. Helverskov JL, Clausen L, Mors O, Frydenberg M, Thomsen PH, Rokkedal K (2010) Trans-diagnostic outcome of eating disorders: A 30-month follow-up study of 629 patients. Eur Eat Disord Rev. 18:453Y463. Herzog DB, Keller MB, Lavori PW, Kenny GM, Sacks NR (1992) The prevalence of personality disorders in 210 women with eating disorders. J Clin Psychiatry. 53:147Y152. Herzog T, Stiewe M, Sandholz A, Hartmann A (1995) Borderline-Syndrom und Essto¨rungen. Literaturu¨bersicht und Interview-Studie an 172 konsekutiven InanspruchnahmepatientInnen der Freiburger Essto¨rungsambulanz. Psychother Psychosom Med Psychol. 45:97Y108. Hudson JI, Coit CE, LaLonde JK, Pope HG (2012) By how much will the proposed bew DSM-5 criteria increase the prevalence of binge eating disorder? Int J Eat Disord. 45:139Y141. Hudson JI, Hiripi E, Pope HG, Kessler RC (2007) The prevalence and correlates of eating disorders in the National Comorbidity Survey Replication. Biol Psychiatry. 61:348Y358.
Cassin SE, von Ranson KM (2005) Personality and eating disorders: A decade in review. Clin Psychol Rev. 25:895Y916.
Javaras KN, Laird NM, Reichborn-Kjennerud T, Bulik CM, Pope HG, Hudson JI (2008) Familiality and heritability of binge eating disorder: Results of a casecontrol family study and a twin study. Int J Eat Disord. 41:174Y179.
Chen EY, McCloskey MS, Michelson S, Gordon KH, Coccaro E (2011) Characterizing eating disorders in a personality disorders sample. Psychiatry Res. 185:427Y432.
Keel PA, Striegel-Moore RH (2009) The validity and clinical utility of purging disorder. Int J Eat Disord. 42:706Y719.
Cooper JL, Morrison TL, Bigman OL, Abrahamowitz SI, Blunden D, Nassi A, Krener P (1988) Bulimia and borderline personality disorder. Int J Eat Disord. 7:43Y49. Cumella EJ, Kally Z (2008) Profile of 50 women with midlife-onset eating disorders. Eat Disord. 16:193Y203.
124
www.jonmd.com
Kendall PC, Clarkin JF (1992) Introduction to special section: Comorbidity and treatment implications. J Consult Clin Psychol. 60:833Y834. Maran˜o´n I, Echeburu´a E, Grijalvo J (2007) Are there more personality disorders in treatment-seeking patients with eating disorders than in other kind of psychiatric patients? A two control groups comparative study using the IPDE. Int J Health Psychol. 7:283Y293.
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
The Journal of Nervous and Mental Disease
& Volume 202, Number 2, February 2014
Martin CK, Williamson DA, Thaw JM (2000) Criterion validity of the multiaxial assessment of eating disorders symptoms. Int J Eat Disord. 28:303Y310. Masheb RM, Grilo CM (2008) Examination of predictors and moderators for selfhelp treatments of binge-eating disorder. J Consult Clin Psychol. 76:900Y904. Milos G, Spindler A, Schnyder U, Fairburn CG (2005) Instability of eating disorder diagnoses: Prospective study. Br J Psychiatry. 187:573Y578. Milos GF, Spindler AM, Buddeberg C, Crameri A (2003) Axes I and II comorbidity and treatment experiences in eating disorder subjects. Psychother Psychosom. 72:276Y285. Picot AK, Lilenfeld LRR (2003) The relationship among binge severity, personality psychopathology, and body mass index. Int J Eat Disord. 34:98Y107. Pinheiro AP, Raney TJ, Thornton LM, Fichter MM, Berrettini WH, Goldman D, Halmi KA, Kaplan AS, Strober M, Treasure J, Woodside DB, Kaye WH, Bulik CM (2010) Sexual functioning in women with eating disorders. Int J Eat Disord. 43:123Y129. Piran N, Lerner P, Garfinkel PE, Kennedy SH, Brouillette C (1988) Personality disorders in anorexic patients. Int J Eat Disord. 7:589Y599. Pope HG, Lalonde JK, Pindyck LJ, Walsh T, Bulik CM, Crow SJ, McElroy SL, Rosenthal N, Hudson JI (2006) Binge eating disorder: A stable syndrome. Am J Psychiatry. 163:2181Y2183. Ramklint M, Jeansson M, Holmgren S, Ghaderi A (2010) Assessing personality disorders in eating disordered patients using the SCID-II: Influence of measures and timing on prevalence rate. Pers Individ Dif. 48:218Y223. Rockert W, Kaplan AS, Olmsted MP (2007) Eating disorder not otherwise specified: The view from a tertiary care treatment center. Int J Eat Disord. 40: S99YS103. Rosenvinge JH, Martinussen M, Østensen E (2000) The comorbidity of eating disorders and personality disorders: A meta-analytic review of studies published between 1983 and 1998. Eat Weight Disord. 5:52Y61. Rybakowsky F, Slopien A, Zakrzewska M, Hornowska E, Rajewski A (2004) Temperament and Character Inventory (TCI) in adolescents with anorexia nervosa. Acta Neuropsychiatr. 16:169Y174. Rø Ø, Martinsen EW, Hoffart A, Rosenvinge J (2005) Two-year prospective study of personality disorders in adults with longstanding eating disorders. Int J Eat Disord. 37:112Y118. Sansone RA, Levitt JL, Sansone LA (2005) The prevalence of personality disorders among those with eating disorders. Eat Disord. 13:7Y21. Schmidt NB, Telch MJ (1990) Prevalence of personality disorders among bulimics, nonbulimic binge eaters, and normal controls. J Psychopathol Behav Assess. 12:169Y185. Shedler J, Beck A, Fonagy P, Gabbard GO, Gunderson J, Kernberg O, Michels R, Westen D (2010) Personality disorders in DSM-5. Am J Psychiatry. 167: 1026Y1028. Skodol AE, Oldham JM, Hyler SE, Kellman HD, Doidge NN, Davies M (1993) Comorbidity of DSM-III-R eating disorders and personality disorders. Int J Eat Disord. 14:403Y416. Specker S, de Zwaan M, Raymond N, Mitchell J (1994) Psychopathology in subgroups of obese women with and without binge eating disorder. Compr Psychiatry. 35:185Y190. Stice E, Agras WS, Telch CF, Halmi KA, Mitchell JE, Wilson T (2001) Subtyping binge eating-disordered women along dieting and negative affect dimensions. Int J Eat Disord. 30:11Y27. Striegel-Moore RG, Franko DL (2007) Should binge eating disorder be included in the DSM-V? A critical review of the state of the evidence. Annu Rev Clin Psychol. 4:305Y324.
* 2014 Lippincott Williams & Wilkins
Personality Disorders in BED/EDNOS
Striegel-Moore RH, Garvin V, Dohm F-A, Rosenheck RA (1999) Psychiatric comorbidity of eating disorders in men: A national study of hospitalized veterans. Int J Eat Disord. 25:399Y404. Sunday SR, Peterson CB, Andreyka K, Crow SJ, Mitchell JE, Halmi KA (2001) Differences in DSM-III-R and DSM-IV diagnoses in eating disorder patients. Compr Psychiatry. 42:448Y455. Sysko R, Devlin MJ, Walsh BT, Zimmerli E, Kissileff HR (2007) Satiety and test meal intake among women with binge eating disorder. Int J Eat Disord. 40:554Y561. Telch C, Stice E (1998) Psychiatric comorbidity in women with binge eating disorder: Prevalence rates from a non-treatment seeking sample. J Consult Clin Psychol. 66:768Y776. Thomas JJ, Vartanian LR, Brownell KD (2009) The relationship between eating disorder not otherwise specified (EDNOS) and officially recognized eating disorders: Meta-analysis and implications for DSM. Psychol Bull. 135: 407Y433. Tyrer P, Crawford M, Mulder R (2011) Reclassifying personality disorders. Lancet. 377:1814Y1815. Van Hanswijck de Jonge P, Van Furth EF, Lacey JH, Waller G (2003) The prevalence of DSM-IV personality pathology among individuals with bulimia nervosa, binge eating disorder and obesity. Psychol Med. 33:1311Y1317. Vitousek K, Manke F (1994) Personality variables and disorders in anorexia nervosa and bulimia nervosa. J Abnorm Psychol. 103:137Y147. Vitousek KM, Stumpf R (2006) Difficulties in the assessment of personality traits and disorders in individuals with eating disorders. In Sansone RA, Levitt JL (Eds), Personality disorders and eating disorders: Exploring the frontier (pp 91Y117). New York: Routledge. Wentz E, Gillberg IC, Anckarsater H, Gillber C, Rastam M (2009) Adolescentonset anorexia nervosa: 18-year outcome. Br J Psychiatry. 194:168Y174. Wilfley DE, Friedman MA, Dounchis JZ, Stein RI, Welch RR, Ball SA (2000) Comorbid psychopathology in binge eating disorder: Relation to eating disorder severity at baseline and following treatment. J Consult Clin Psychol. 68:641Y649. Williamson DA (2007) Does the evidence point to a binge eating phenotype? Comment on Gordon et al. (2007) and Wonderlich et al. (2007). Int J Eat Disord. 40:S72YS75. Wilson GT, Grilo CM, Vitousek KM (2007) Psychological treatment of eating disorders. Am Psychol. 62:199Y216. Wonderlich SA, Gordon KH, Mitchell JA, Crosby RD, Engel SG (2009) The validity and clinical utility of binge eating disorder. Int J Eat Disord. 42: 687Y705. Wonderlich SA, Mitchell JE (1992) Eating disorders and personality disorders. In Yager J, Edelstein CK (Eds), Special problems in managing eating disorders (pp 51Y86). Washington, DC: American Psychiatric Press. Wonderlich SA, Swift WJ, Slotnick HB, Goodman S (1990) DSM-III-R personality disorders in eating-disorder subtypes. Int J Eat Disord. 9:607Y616. Yager J, Landsverk J, Edelstein CK, Hyler SE (1989) Screening for axis II personality disorders in women with bulimic eating disorders. Psychosomatics. 30:255Y262. Yanovski SZ, Nelson JE, Dubbert BK, Spitzer RL (1993) Association of binge eating disorder and psychiatric comorbidity in obese subjects. Am J Psychiatry. 150:1472Y1479. Yates W, Sieleni B, Reich J, Brass C (1989) Comorbidity of bulimia nervosa and personality disorder. J Clin Psychiatry. 50:57Y59. Zimmerman M, Francione-Witt C, Chelminski I, Young D, Tortolani C (2008) Problems applying the DSM-IV eating disorders diagnostic criteria in a general psychiatric outpatient practice. J Clin Psychiatry. 69:381Y384.
www.jonmd.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
125
Personality Disorders in Eating Disorder Not Otherwise Specified and Binge Eating Disorder A Meta-analysis of Comorbidity Studies Oddgeir Friborg, PhD,*Þ Monica Martinussen, PhD,þ Sabine Kaiser, MSc,* Karl Tore Øverga˚rd, PsyD,* Egil W. Martinsen, MD, PhD,§ Pho¨be Schmierer, MSc,* and Jan Harald Rosenvinge, PhD* Abstract: A meta-analysis was conducted to identify the proportion of comorbid personality disorders (PDs) in patients with eating disorder not otherwise specified (EDNOS) and binge eating disorder (BED). A search identified 20 articles in the period of 1987 to 2010. For EDNOS and BED, the comorbid proportions for any PD were 0.38 and 0.29, respectively; for cluster C PDs, 0.38 and 0.30, respectively (avoidant PD, 0.18 and 0.12, and obsessive-compulsive PD, 0.11 and 0.10, respectively); and for cluster B PDs, 0.25 and 0.11, respectively (borderline, 0.12 and 0.10). This pattern converged with findings on anorexia nervosa and bulimia nervosa, except being lower. Because the comorbidity profiles for EDNOS and BED were highly similar, their underlying PD pathology seems similar. Few moderators were significant, except for interviews yielding lower estimates than that of questionnaires. The variance statistic for any PD comorbidity was wide for EDNOS and narrow for BED, thus partly supporting BED as a distinct eating disorder category and EDNOS as a potentially more severe condition than BED. Key Words: Personality disorders, eating disorder not otherwise specified, binge eating disorder, meta-analysis, comorbidity (J Nerv Ment Dis 2014;202: 119Y125)
E
ating disorder not otherwise specified (EDNOS) is a separate diagnostic category in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV; American Psychiatric Association [APA], 1994), comprising conditions in which the diagnostic criteria for anorexia nervosa (AN; i.e., degree of weight loss) or bulimia nervosa (BN; i.e., the duration or frequency of binging and vomiting) are not fully met. The eating disorder symptoms may be atypical, such as vomiting after a minor intake of food or repeatedly chewing but not swallowing large amounts of food. EDNOS is the most frequent eating disorder (Hay et al., 2008; Hudson et al., 2007), accounting for more than half of all current eating disorders in outpatient clinics (Fairburn and Cooper, 2007; Martin et al., 2000; Rockert et al., 2007) and up to 90% of psychiatric community-based treatment-seeking samples if binge eating disorder (BED) is included (Zimmerman et al., 2008). In a meta-analysis, the general level of psychopathology was comparable for EDNOS and AN, although patients with BN reported somewhat higher levels of pathology (mean d = 0.19) than patients with EDNOS (Thomas et al., 2009). Some subtypes of EDNOS, particularly the purging *Department of Psychology, Faculty of Health Sciences, University of Tromsø, Tromsø, Norway; †Department of Psychiatric Research, University Hospital of North Norway, Tromsø, Norway; ‡The Regional Centre for Child and Youth Mental Health and Child Welfare, Faculty of Health Sciences, University of Tromsø, Tromsø, Norway; and §Clinic for Mental Health, Oslo University Hospital, Oslo, Norway. Send reprint requests to Oddgeir Friborg, PhD, Department of Psychology, Faculty of Health Sciences, University of Tromsø, N-9037 Tromsø, Norway. E-mail: [email protected]. Copyright * 2014 by Lippincott Williams & Wilkins ISSN: 0022-3018/14/20202Y0119 DOI: 10.1097/NMD.0000000000000080
The Journal of Nervous and Mental Disease
one, may show a wider spectrum of clinical severity than that of AN and BN (Keel and Striegel-Moore, 2009). EDNOS can be a transitional stage to recovery (Agras et al., 2009), as well as to AN or BN (Milos et al., 2005). BED is listed as an appendix diagnosis in DSM-IV (APA, 1994). It is subsumed under EDNOS but is increasingly recognized as a distinct diagnostic disorder on the basis of taxometric analyses (Williamson, 2007), family aggregation studies (Javaras et al., 2008), treatment response research (Wilson et al., 2007), and studies of clinical course (Pope et al., 2006). Two extensive reviews (StriegelMoore and Franko, 2007; Wonderlich et al., 2009) also conclude positively with respect to the merit of BED as a distinct eating disorder diagnosis, and hence, it is now proposed as a separate ED diagnosis in the DSM-5 (APA, 2012). Although BED is currently part of the EDNOS category, our further use of the EDNOS acronym in the present article does not include studies on patients with BED. The provisional DSM-IV criteria for BED include recurrent binge-eating episodes defined as rapid eating in a discrete period characterized by poor self-regulation (e.g., not being able to stop binging). This is usually not preceded by feelings of hunger or satiety (Sysko et al., 2007). It departs from BN in terms of absence of extreme compensatory weight-control behaviors after a binge-eating episode and also has a less severe natural course than BN (Fairburn et al., 2000). Self-disgust and shame are typical concomitant feelings, and thus, patients often eat alone to reduce social embarrassment and distress. The transition from restrictive AN to BED is more common (Eddy et al., 2008) than the reverse process. Axis II personality disorder (PD) comorbidity has been found to worsen the recovery prognosis from AN, BN, and EDNOS (Rø et al., 2005); from BED (Masheb and Grilo, 2008); and in a large clinical study (N = 629) from any type of eating disorders (Helverskov et al., 2010). However, studies on the natural course of eating disorders (Grilo et al., 2007) do not indicate different rates of remission or relapse depending on a PD diagnosis, except for the avoidant PD among patients with EDNOS. Because avoidant and obsessivecompulsive PDs are reported as the most typical for patients with BED (Becker et al., 2010), cluster C comorbid PDs may be particularly relevant for the treatment prognosis of both EDNOS and BED. Eating disorders and PDs, furthermore, share important treatmentcomplicating clinical features, such as a profound self-devaluation; feelings of low mood and anxiety; as well as a poor self-insight into behavioral, cognitive, and interpersonal dysfunctional patterns that may reciprocally reinforce the severity of either condition (Chen et al., 2011; Kendall and Clarkin, 1992). Given the prevalence of BED and EDNOS in community and clinical settings, and considering how a PD may complicate treatment and worsen the prognosis, knowledge of comorbid patterns of PDs is highly warranted. The between-study variation in comorbid proportions (21% to 97%) of PDs in eating disorders is, however, large (Cassin and von Ranson, 2005; Sansone et al., 2005; Vitousek and Manke, 1994). This may also apply to BED and EDNOS. A metaanalysis may determine the mean comorbidity proportions more
& Volume 202, Number 2, February 2014
www.jonmd.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
119
The Journal of Nervous and Mental Disease
Friborg et al.
accurately and identify whether these estimates covary with methodological factors that need to be addressed in future studies. Moderators such as sample differences, variation in diagnostic criteria, and the methods of assessment may all account for the comorbidity variation (Sunday et al., 2001). For instance, self-report assessments yield consistently higher proportions of PDs compared with the use of structured clinical interviews (Echeburu´a and Maran˜on, 2001; Ramklint et al., 2010; Rosenvinge et al., 2000). Other sources may be the diagnostic system used and the type of patients recruited (e.g., inpatients versus outpatients). For example, in a meta-analysis on AN and BN (Rosenvinge et al., 2000), the PD comorbidity rate varied from 49% in outpatient to 75% in inpatient samples. Age of onset of an eating disorder is of particular interest as a possible moderator because an early onset (e.g., in teenage years) is related to more psychopathology in general (Wentz et al., 2009), which may impair personality development, whereas a late debut implies a fewer number of cluster C PDs (Cumella and Kally, 2008). One may therefore expect that an early age of onset would yield a higher proportion of any comorbid PD. Given the support for BED as a clinically relevant diagnosis (Javaras et al., 2008; Pope et al., 2006; Williamson, 2007; Wilson et al., 2007), one could expect a higher proportion of PDs in BED than in the clinically more heterogeneous EDNOS category (excluding BED), in which more patients are in a process of recovery compared with BED (Agras et al., 2009). However, because the levels of eating pathology and general psychopathology between EDNOS and BED have been reported comparable in a previous meta-analysis (Thomas et al., 2009), there is insufficient evidence to make any strong predictions. As the first study in the literature, we used a meta-analytic approach to assess the respective frequencies of comorbid axis II PD diagnoses between EDNOS and BED. In addition, the impact of sample characteristics, diagnostic criteria, type of assessment methods, body mass index (BMI), and age of onset for PD comorbidity was examined with regard to the clinical usefulness of the forthcoming DSM-5 diagnostic distinctions in terms of pathological personality functioning.
METHODS Literature Search The databases PsychINFO, EMBASE, and MEDLINE were searched for published articles in English or German language between 1987 (when EDNOS was introduced in the DSM-III-R) and June 2010. The search included the key words eating disorders or anorexia or bulimia and personality disorders and comorbidity. Initially, 459 articles were located, and these were supplemented by 37 articles from reference lists of previous meta-analyses and located articles (k = 33) and from a predecessor of OVID (i.e., SilverPlatter; k = 4). From the initial pool, 369 articles were removed because the patients either had been diagnosed with drug addiction or had recovered from their eating disorder or because the results were published as dissertation abstracts. From the remaining 127 studies, 42 were removed because these did not report proportions of PDs, reported the same results in multiple articles (the most recent were selected), or included subjects younger than 18 years. Studies that used PDs as an inclusion criterion were removed, as were those that failed to report current eating disorder diagnoses or diagnostic information. This yielded a pool of 88 articles, of which 19 articles had a suitable reporting of the comorbidity between PDs and the two forms of eating disorders in question. After this filtering process, an additional search was specifically done for the key word binge eating, which added one extra study not already in the study pool, thus yielding 11 articles addressing EDNOS and 9 addressing BED. 120
www.jonmd.com
& Volume 202, Number 2, February 2014
Coding Procedure and Interrater Reliability From each study, we coded publication year, comorbid events of PDs (converted to a proportion), sex (proportion of women participating), sample size, and mean age. The following variables were coded as moderators: sample characteristics (clinical versus community); diagnostic system (DSM-III-R versus DSM-IV); assessment methods for diagnosing BED, EDNOS, and PDs (structured interview, clinical assessment, and questionnaires); age at onset and duration of eating disorders; and BMI. In case of missing information, the corresponding authors were contacted, and in most cases, they provided the necessary data. The articles were coded by two psychology students trained by the same supervisor (J. H. R.). Disagreements in coding were discussed until consensus was reached. To assess interrater reliability, 17 (19%) of the 88 studies were randomly selected and coded by the first author (O. F.). Interrater agreement was estimated as kappa for categorical data and intraclass correlation coefficient (ICC) for continuous data. Kappa was 0.89 for type of sample (inpatient, outpatient, both, or community) and 1.0 for diagnostic method (interview versus questionnaire). ICC was perfect (1.00) for proportions of clusters A, B, and C PD and for the diagnostic categories of PD (except borderline disorder being lower at 0.97); for number of patients; for proportion of women; and for BMI. ICC was nearly perfect for mean age (0.98), age at onset (0.95), and duration (0.99) of eating disorders.
Procedure and Statistical Analysis The meta-analyses were conducted using the program Comprehensive Meta-Analysis version 2.2.057 (released December 2010; Borenstein et al., 2005). The mean weighted event rate (PD cases/ sample size) represented comorbid proportions and was used as effect sizes in the analyses. Calculations were based on a randomeffects model because the true proportion may vary across studies because of variations in factors such as severity or age of onset. Hence, variance in mean proportions is partitioned in two sources: a) variance within studies (?, sampling error) and b) variance between studies (5, variation in true proportions) (Borenstein et al., 2005; Hedges and Vevea, 1998). Each study was weighted by calculating the inverse of these variance components: 1/(? + 5). Small studies imply a larger within-study variance component (? = R2/n), thus lowering the weight given to data from small samples. A Q test statistic is used to test whether the variance between studies is larger than the variance within studies, thus indicating that variation in proportions truly exists from which predictors (or moderators) may be sought. The random-effects model reduces the type I error rate and enhances the generalizability of the findings (Field, 2003). An analogous Q test statistic for between-group differences may be conducted to examine whether the differences in proportions between two samples depend on a categorical covariate (moderator).
RESULTS Sample Characteristics The study pool consisted of nine BED studies (n = 838) and 11 EDNOS studies (n = 743; see Table 1 for descriptive characteristics). Among the EDNOS studies, eight studies were carried out on clinical samples (n = 373, 50%); two studies, on community samples (n = 337, 45%); and one study, on an unspecified sample (n = 34, 5%). Five studies based the diagnosis of PD on a structured interview (n = 136, 18%), five studies used clinical assessment (n = 294, 40%), and one study used questionnaires (n = 313, 42%). The DSM-IV system was used in four studies (n = 255, 34%); and the DSM-III/ DSM-III-R, in two studies (n = 337, 45%). It was unreported for five studies (n = 152, 21%). The mean age of the patients was 29.5 years (SD, 10.7). * 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
The Journal of Nervous and Mental Disease
& Volume 202, Number 2, February 2014
Personality Disorders in BED/EDNOS
TABLE 1. Overview of the Included Studies of EDNOS and BED Author, Year
N
Cooper et al., 1988 11 Yager et al., 1989 313 Schmidt and Telch, 23 1990 Hay and Hall, 1991 10 Herzog et al., 1995 32 Carlat et al., 1997 43 Striegel-Moore et al., 56 1999 Milos et al., 2003 34 Rø et al., 2005 22 Maran˜o´n et al., 2007 24 Godt, 2008 174 Yanovski et al., 1993 43 Specker et al., 1994 43 Telch and Stice, 1998 61 Wilfley et al., 2000 162 Stice et al., 2001 159 145/228 Grilo, 2002, 2004b Van Hanswijck et al., 15 2003 Picot and Lilenfeld, 50 2003 Masheb and Grilo, 75 2008
Sample
Mean Age
EDNOS EDNOS EDNOS
Out Comm Comm
25.7 24.7 19.0
100 100 100
19.3 V V
20.2 14.9 V
6.7 9.8 V
CA Q I
CA Q I
V III III-R
EDNOS EDNOS EDNOS EDNOS
In In and out In and out In
28.4 V V 56.7
72 V 0 0
V V V V
V V 19.1 V
V V V V
I Q+I CA CA
CA I CA CA
V V III, III-R, IV Othera
21.3 V V V 41.8 35.9 34.8 37.1 36.3 V 41.9
17.6 16.0 V V V V 18.0 V V V V
10.5 13.7 V V V V 25.5 V V V V
I I I Q+I I I I Q+I I I I
I I I
IV IV IV
I Q I I I I I
III-R III-R III-R III-R III-R IV IV
EDNOS EDNOS EDNOS EDNOS BED BED BED BED BED BED BED
% Onset Duration Assessment Assessment Women BMI of ED, yrs of ED, yrs of ED of PD
DSM Edition
Eating Disorder
In, out, and comm 28.1 100 In 29.7 100 Out V 100 In 23.8 100 Comm 36.1 77 Comm 39.2 100 Comm 43.5 100 Comm 45.2 83 Comm 40.0 100 Out 43.7/44.5 78/79 Out 36.9 100
BED
Comm
41.3
92
34.7
V
V
I
I
IV
BED
Out
46.0
81
35.3
28.1
V
I
I
IV
a
ICD-9-CM. Grilo has published two studies from the same sample. Data from the first study were included only if not reported in the second study. CA indicates clinical assessment; comm, community sample; in, inpatients; out, outpatients; Q, questionnaire; I, structured interview. b
For BED, 59% were community samples (n = 519, k = 6) and 41% were clinical samples (n = 319, k = 3). Only one and two studies reported duration of illness and age of onset, respectively. In contrast to the EDNOS studies, structured clinical interviews to diagnose a PD were dominant in eight studies (n = 795, 95%), whereas only one study used questionnaires (n = 43, 5%). Moreover, all subjects were assessed using either DSM-III-R (n = 469, k = 5, 56%) or DSM-IV (n = 369, k = 4, 44%) diagnostic criteria. The patients had a mean age of 41.6 years (SD, 3.3). BMI was reported in six BED studies (mean, 37.2; SD, 2.8) but only by two EDNOS studies (mean, 20.3; SD, 1.0).
Comorbidity Related to PD Clusters The mean proportions of comorbid PDs for EDNOS and BED as well as associated confidence intervals (CIs) and heterogeneity statistics are presented in Table 2. The mean proportion of patients having at least one PD (any PD) was higher for EDNOS (0.38) than for BED (0.29), but the difference was not statistically significant (Q = 2.23, p = 0.14). The variation in proportions for any PD was considerably larger among EDNOS (95% CI, 0.26Y0.51) than among BED studies (95% CI, 0.24Y0.33), also evidenced by a Qwithin test that was significant only for the EDNOS group. At PD cluster levels, no significant differences in mean proportions between EDNOS and BED emerged. The largest difference was seen in cluster B (0.25 and 0.11, respectively). Among the EDNOS studies, the comorbid proportions increased from cluster A, B, to C (0.09, 0.25, and 0.38, respectively), but it was not significant (Qdf = 2 = 4.93, p = 0.09). A similar trend was present for the * 2014 Lippincott Williams & Wilkins
BED studies (0.07, 0.11, and 0.30, respectively) but again statistically nonsignificant (p = 0.09). If BED and EDNOS were combined, the increase in proportions of clusters A, B, and C (0.08, 0.18, and 0.34, respectively) was significant (Qdf = 2 = 10.74, p G 0.01). The heterogeneity statistic (Qwithin) was significant for all PD clusters within EDNOS and BED, indicating that sampling error was not the sole reason for the variations in proportions.
Comorbidity Related to Specific PD Diagnoses No significant differences in mean proportions of PD were found between EDNOS and BED for any of the specific PDs. Avoidant PD was most frequent in EDNOS (0.18), followed by borderline (0.12) and obsessive-compulsive (0.11) PD. The lowest proportions were found for narcissistic PD (0.02) and the cluster AYspecific PDs (range, 0.01Y0.06). Avoidant PD was also most prevalent among the BED studies (0.12), followed by borderline and obsessive-compulsive PD (both 0.10). The lowest proportions were found for schizotypal (0.01), schizoid (0.02), and histrionic (0.02) disorders. The heterogeneity statistics (Qwithin) were significant for most specific PD proportions within both EDNOS and BED, except some of the rarer conditions (e.g., antisocial and schizoid PDs). The largest difference in PD proportions between EDNOS and BED was found for avoidant PD (0.18 and 0.12, respectively).
Moderator Analyses Type of sample influenced the comorbidity proportions minimally because significant differences between clinical and community www.jonmd.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
121
The Journal of Nervous and Mental Disease
Friborg et al.
TABLE 2. Proportions of PD Diagnoses for EDNOS and BED Based on a Random-Effects Model EDNOS Any PD Cluster A Paranoid Schizoid Schizotypal Antisocial Cluster B Borderline Histrionic Narcissistic Cluster C Avoidant Dependent Obsessive-compulsive BED Any PD Cluster A Paranoid Schizoid Schizotypal Antisocial Cluster B Borderline Histrionic Narcissistic Cluster C Avoidant Dependent Obsessive-compulsive
k
N
p (95% CI)
Qwithin
9 5 5 5 5 4 5 7 5 5 5 5 5 5
699 567 557 559 559 433 567 600 557 558 567 557 558 556
0.38 (0.26Y0.51) 0.09 (0.02Y0.31) 0.06 (0.02Y0.16) 0.03 (0.02Y0.04) 0.04 (0.00Y0.28) 0.01 (0.00Y0.03) 0.25 (0.10Y0.50) 0.12 (0.05Y0.25) 0.06 (0.02Y0.21) 0.02 (0.01Y0.03) 0.38 (0.25Y0.52) 0.18 (0.11Y0.30) 0.07 (0.02Y0.22) 0.11 (0.06Y0.20)
58.93*** 48.71*** 19.17*** 1.97 40.61*** 2.19 59.16*** 42. 23*** 28.36*** 1.45 28.06*** 17.95*** 35.20*** 15.19**
8 4 7 6 6 6 4 8 6 6 4 8 6 8
786 508 629 589 587 587 508 679 588 587 508 761 586 762
0.29 (0.24Y0.33) 0.07 (0.01Y0.34) 0.07 (0.03Y0.18) 0.02 (0.00Y0.07) 0.01 (0.01Y0.03) 0.03 (0.02Y0.05) 0.11 (0.02Y0.39) 0.10 (0.06Y0.16) 0.02 (0.00Y0.20) 0.03 (0.01Y0.08) 0.30 (0.21Y0.41) 0.12 (0.08Y0.17) 0.03 (0.01Y0.11) 0.10 (0.07Y0.16)
12.72 65.62*** 51.32*** 16.43** 5.22 3.31 69.15*** 24.95*** 63.89*** 15.20** 15.00** 22.24** 29.57*** 22.31**
**p G 0.01. ***p G 0.001. k indicates number of studies; p, mean proportion; Qwithin, heterogeneity statistic.
& Volume 202, Number 2, February 2014
samples emerged only for cluster A, which occurred more frequently in community samples. Similar tests for assessment method indicated that clinical interviews yielded consistently lower comorbid proportions than that of questionnaire-based methods (see Table 3). However, this comparison was based only on two studies in the questionnaire group. The use of different diagnostic systems (DSM-III or DSM-IV) was nonsignificant. A meta-regression analysis showed neither BMI (only estimable for BED) nor age of onset (only estimable for EDNOS) as significant moderators of the degree of PD comorbidity. Duration of eating disorder was not estimable (reported in only one BED and three EDNOS studies).
DISCUSSION Overall and Cluster Level Comparisons Having at least one PD of any kind (any PD) occurred more frequently in EDNOS (0.38) than in BED (0.29), but this could not be ascertained because of a statistical nonsignificant difference. None of the differences on the PD cluster and specific levels were significant either. The pathological personality functioning in BED and EDNOS thus stand out as more similar than different. The mean proportions of PDs showed an ascending increase across the clusters A, B, and C. Similar findings have been reported for AN (Herzog et al., 1992; Piran et al., 1988; Skodol et al., 1993; Wonderlich et al., 1990) and, in part, for BN (Herzog et al., 1992; Rosenvinge et al., 2000; Schmidt and Telch, 1990; Wonderlich and Mitchell, 1992; Yates et al., 1989). This may also indicate important similarities across the eating disorder diagnoses in terms of PD comorbidity. Although PD comorbidity occurs less frequently in BED and EDNOS than in AN and BN, the severity or impact on treatment may be of comparable concern when present. Moreover, the variance in the proportions of any PD was considerably wider for EDNOS than for BED, hence supporting the notion of EDNOS as a clinically more heterogeneous diagnosis than BED (Agras et al., 2009; Keel and Striegel-Moore, 2009; Milos et al., 2005). This implies that some patients with EDNOS may have more comorbidity conditions than patients with BED. Because cluster B PDs occurred more than twice as often in EDNOS than in BED, EDNOS stands out as a potentially more serious clinical condition than BED.
TABLE 3. Impact of Moderators on Proportions of PD Diagnoses in EDNOS and BED Moderators PDs a
ka
kb
na
nb
p (95% CI)a
p (95% CI)b
Qbetween
5
3c
448
518
0.05 (0.01Y0.14)
0.24 (0.08Y0.56)
4.21*
6 8 8 11 8 8
2c 2c 2c 2c 2c 2c
470 618 616 738 615 614
356 356 356 356 356 356
0.06 (0.03Y0.12) 0.01 (0.01Y0.03) 0.01 (0.00Y0.05) 0.09 (0.07Y0.12) 0.02 (0.01Y0.05) 0.02 (0.01Y0.04)
0.41 (0.20Y0.67) 0.05 (0.03Y0.09) 0.19 (0.07Y0.43) 0.30 (0.20Y0.41) 0.32 (0.14Y0.58) 0.26 (0.22Y0.31)
13.87*** 7.95** 15.24*** 20.48*** 19.71*** 72.10***
b
Clinical versus community Cluster A Interviewa versus questionnaireb Cluster A Schizoid Schizotypal Borderline Histrionic Dependent
All PD clusters and specific PDs were tested (only significant effects were reported). Diagnostic system (DSM-III-R versus DSM-IV) was tested but found nonsignificant. a Subgroup 1. b Subgroup 2. c The between-study variance (t2) was estimated as equal across both groups because of the low number of studies, especially in the questionnaire subgroup, consisting of one BED and one EDNOS study. *p G 0.05. **p G 0.01. ***p G 0.001. k indicates number of studies; p, mean proportion.
122
www.jonmd.com
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
The Journal of Nervous and Mental Disease
& Volume 202, Number 2, February 2014
Specific Level Comparisons The comorbidity profiles between BED and EDNOS were similar also on the specific PD level, with avoidant, borderline, and obsessive-compulsive disorders occurring in a descending order of frequency. Taken together, BED and EDNOS seem to share the PD comorbidity profile with AN and BN, and those differences mostly relate to the level of proportions. The most common comorbid PD in patients with BED were avoidant, obsessive-compulsive, and borderline PDs. These findings fit with previous studies on EDNOS and BED (Ramklint et al., 2010), as well as BED, AN, and BN (Cassin and von Ranson, 2005), and thus question the conclusions from a previous literature review (Sansone et al., 2005) claiming that obsessive-compulsive disorder is most typical with BED.
Moderator Analyses Contrary to a previous meta-analysis for AN and BN (Rosenvinge et al., 2000), the impact of clinical versus community samples was almost negligible. The only difference was cluster A PDs, which were more prevalent in community than in clinical samples. One explanation may be that treatment-seeking behaviors may be less likely to occur among patients with cluster A PDs, generally characterized as being distrusting and suspicious of others (APA, 1994). Still, the present results were more in line with the meta-analysis by Thomas et al. (2009), which reported no differences between EDNOS and BED in general psychopathology across patient and nonpatient samples. Similarly, Wilfley et al. (2000) found no differences in current and lifetime prevalence rates of psychiatric diagnoses or level of general psychiatric features between a clinical and a recruited treatment-seeking sample diagnosed with BED. However, community samples may be flawed with both overreporting and underreporting (Vitousek and Stumpf, 2006). When addressing the problems of treatment-seeking subjects by comparing frequencies of comorbid PDs among a BED-diagnosed clinical sample and a recruited community sample, Telch and Stice (1998) found a significantly higher relative risk for any axis I and axis II comorbidity in the clinical compared with the community sample. The lack of comparable differences in our meta-analysis may relate to the fact that 46% of the community samples consisted of treatmentseeking subjects, whereas the remaining subjects were recruited for research purposes. The use of self-report measures inflated the proportions. One caveat is the low number of studies using questionnaires, but the findings seem valid because these concur with numerous previous studies (e.g., Echeburu´a and Maran˜on, 2001; Godt, 2002; Ramklint et al., 2010; Rosenvinge et al., 2000). Indeed, the less use of self-report measures compared with, for instance, previous meta-analyses of PDs in AN and BN (Rosenvinge et al., 2000) is a positive trend in the literature in terms of improved scientific and clinical validity of findings. Unexpectedly, age of onset of an eating disorder did not moderate the proportions of any PD, although a declining trend was present (p = 0.13). Reasons for this finding may be the low number of studies (k = 4) reporting age of onset and the fact that these data are mostly available for younger age groups. However, available data across age groups (Cumella and Kally, 2008) also indicate a declining trend. Comparable studies have reported that earlier age of onset chronically impedes sexual functioning and libido (Pinheiro et al., 2010); reduces the likelihood of a more positive long-term outcome (Wentz et al., 2009); and, in general, implies more personality problems, a more profound body dissatisfaction, and a longer-lasting ascetic-driven pursuit of thinness (Abbate-Daga et al., 2007). An early debut may thus be a sizeable risk factor of developing an additional PD.
Strengths and Limitations A total of 30 studies were located in a meta-analysis (Thomas et al., 2009) comparing eating pathology and general psychopathology * 2014 Lippincott Williams & Wilkins
Personality Disorders in BED/EDNOS
between BED and EDNOS. Because PDs are less frequently studied than axis I symptoms, the identification of 20 studies in the current meta-analysis indicates a minor risk for having missed relevant studies. This point is supported by the fact that authors of publications were contacted and asked to provide unpublished data. Moreover, data comprehensibility was secured because authors were asked to provide data not appearing in their published articles. Finally, the findings were minimally affected by the coding procedure because the interrater reliability indicated high agreement. The moderator analyses were, however, hampered by the small number of total studies located, which prevented separate moderator analyses of EDNOS and BED. Moreover, an important sample bias has to be considered in the comparison of EDNOS and BED because 6 of 9 BED studies were based on community samples, whereas this was the case for only 2 of 11 EDNOS studies. Although the influence of sample type was modest, these sampling differences complicate interpretation of the differences. The higher heterogeneity (and slightly higher proportion of PDs) in EDNOS may therefore relate to these sample differences because levels of psychopathology are generally expected to be higher in clinical than in community samples. On the other hand, this concern is reduced considering that 46% of the community participants were also treatment-seeking subjects.
Implications for Further Research Future research on the comorbidity between eating disorders and PDs needs methodological improvements. This would include a systematic collection and reporting of demographic and clinical information, especially BMI, age of onset, and duration of both eating and PDs. This would certainly improve moderator analyses by allowing a simultaneous examination of multiple covariates. Second, future studies on EDNOS should provide more detailed information about the exact nature of the eating pathology, specifically about other eating disorder criteria that are not met. Third, to reduce error variance, we recommend using larger samples and similar recruitment and sampling procedures for both BED and EDNOS. Nevertheless, the present study indicates that structured clinical interviews are the preferred method, which certainly have improved the estimates of the true proportions of comorbid PDs. The previous (Cassin and von Ranson, 2005; Rosenvinge et al., 2000) and current meta-analyses suggest a remarkably consistent pattern of frequency across eating disorder diagnoses regarding distribution of proportions across PD clusters and specific PD diagnoses. This may support a transdiagnostic approach (Fairburn et al., 2003, 2008) to understanding the nature of eating disorders. This approach advocates the commonality of symptoms across eating disorder diagnoses and, hence, a more parsimonious approach than focusing on diagnostic differences of presumably remote clinical relevance. In addition, because the highest proportions were found for avoidant, obsessive-compulsive, and borderline PDs, this may support an inhibition-disinhibition dimension that is common across eating disorder diagnoses (Fassino et al., 2002; Rybakowsky et al., 2004), which is also relevant in a transdiagnostic understanding of eating disorders. The DSM-5 will probably be launched in 2014. At the DSM-5 Web site (http://www.dsm5.org/Pages/Default.aspx), several changes have been proposed for the disorders composing the present metaanalysis. In the DSM-5, the overall goal for the eating disorder classification is to restrict the spectrum of disorder clinical features now categorized as EDNOS. This is done by introducing more finegrained diagnostic labels and phenotypes under the heading Feeding or Eating Disorder Not Elsewhere Classified (FEDNEC; e.g., atypical AN, BN, or BED; a purging disorder; or a night eating syndrome). A recent study (Birgegard et al., 2012) indicates that this goal is realized using the proposed criteria. Another issue is whether www.jonmd.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
123
The Journal of Nervous and Mental Disease
Friborg et al.
more diagnostic labels are relevant in terms of specific psychopathology, clinical management, and outcome. Our findings that subjects with EDNOS do have comorbid PDs serve as an argument for FEDNEC as an overall diagnostic category. Future research, however, should study whether the clinical heterogeneity of PDs among subjects with the current EDNOS will persist or dissolve using the subgroup distinctions of FEDNEC. Another way of reducing heterogeneity of EDNOS is to include BED as an official DSM-5 diagnosis. The present results support this in the sense that comorbidity relates to well-known criteria for accepting a proposed clinical state as a diagnostic entity (Blashfield et al., 1990). Moreover, the variance in the comorbidity proportions is quite narrow compared with EDNOS, hence indicating a clinically more specific condition. The diagnostic criteria for BED will, however, also be more lenient, that is, by requiring binge eating once a week for 3 months compared with twice a week for 6 months in the current provisional DSM-IV criteria. However, preliminary studies indicate no substantial increase in the prevalence of BED (Hudson et al., 2012) and, hence, possibly in the proportion or heterogeneity of PD. Our findings are also relevant when considering the DSM-5 proposal for the organization of PD diagnoses. Although the DSM-5 will abandon the cluster model, the diagnostic-specific PD categories in the present study will be retained. New and more sophisticated models (Shedler et al., 2010; Tyrer et al., 2011) have been considered, which may offer new perspectives in studying comorbidity. DISCLOSURE The authors declare no conflict of interest.
REFERENCES Abbate-Daga G, Piero A, Rigardetto R, Gandione M, Gramaglia C, Fassino S (2007) Clinical, psychological and personality features related to age of onset of anorexia nervosa. Psychopathology. 40:261Y268. Agras WS, Crow S, Mitchell JE, Halmi KA, Bryson S (2009) A 4-year prospective study of eating disorder NOS compared with full eating disorder syndromes. Int J Eat Disord. 42:565Y570. American Psychiatric Association (1994) Diagnostic and statistical manual of mental disorders (DSM-IV) (4th ed). Washington, DC: American Psychiatric Association. American Psychiatric Association (2012) DSM-5 development. Arlington, TX: American Psychiatric Association. Retrieved from http://www.dsm5.org. Becker DF, Masheb RM, White MA, Grilo CM (2010) Psychiatric, behavioral, and attitudinal correlates of avoidant and obsessive-compulsive personality pathology in patients with binge-eating disorder. Compr Psychiatry. 51:531Y537. Birgegard A, Norring C, Clinton D (2012) DSM-IV versus DSM-5: Implementation of proposed DSM-5 criteria in a large naturalistic database. Int J Eat Disord. 45:353Y361. Blashfield RK, Sprock J, Fuller AK (1990) Suggested guidelines for including or excluding categories in the DSM-IV. Compr Psychiatry. 31:15Y19. Borenstein M, Hedges L, Higgins J, Rothstein H (2005) Comprehensive metaanalysis: Version 2.0. Englewood, NJ: Biostat. Carlat DJ, Camargo A, Herzog DB (1997) Eating disorders in males: A report on 135 patients. Am J Psychiatry. 154:1127Y1132.
& Volume 202, Number 2, February 2014
Echeburu´a E, Maran˜on I (2001) Comorbilidad de las alteraciones de la conducta alimentaria con los trastornos de personalidad. Psicologı´a Conductual. 9: 513Y525. Eddy KT, Dorer DJ, Franko DL, Tahilani K, Thompson-Brenner H, Herzog DB (2008) Diagnostic crossover in anorexia nervosa and bulimia nervosa: Implications for DSM-V. Am J Psychiatry. 165:245Y250. Fairburn CG, Cooper Z (2007) Thinking afresh about the classification of eating disorders. Int J Eat Disord. 40:S107YS110. Fairburn CG, Cooper Z, Doll HA, Norman P, O’Connor M (2000) The natural course of bulimia nervosa and binge eating disorder in young women. Arch Gen Psychiatry. 57:659Y665. Fairburn CG, Cooper Z, Shafran R (2003) Cognitive behaviour theory for eating disorders: A ‘‘transdiagnostic’’ theory and treatment. Behav Res Ther. 41: 509Y528. Fairburn CG, Cooper Z, Waller D (2008) ‘‘Complex cases’’ and comorbidity. In Fairburn CG (Ed), Cognitive behavior therapy and eating disorders (pp 245Y258). New York: Guilford Press. Fassino S, Abbate-Draga G, Amianto F, Leombruni P, Boggio S, Rovera G (2002) Temperament and character profile of eating disorders: A controlled study with the Temperament and Character Inventory. Int J Eat Disord. 32:412Y425. Field AP (2003) The problems in using fixed-effects models of meta-analysis on real-world data. Underst Stat. 2:77Y96. Godt K (2002) Personality disorders and eating disorders: The prevalence of personality disorders in 176 female patients with eating disorders. Eur Eat Disord Rev. 10:102Y109. Godt K (2008) Personality disorders in 545 patients with eating disorders. Eur Eat Disord Rev. 16:94Y99. Grilo CM (2002) Are there gender differences in DSM-IV personality disorders? Compr Psychiatry. 43:427Y430. Grilo CM (2004) Factorial structure and diagnostic efficiency of DSM-IV criteria for avoidant personality disorder in patients with binge eating disorder. Behav Res Ther. 42:1149Y1162. Grilo CM, Pagano ME, Skodol AE, Sanislow C, McGlashan TH, Gunderson JG, Stout RL (2007) Natural course of bulimia nervosa and of eating disorder not otherwise specified: 5-year prospective study of remissions, relapses, and the effects of personality disorder psychopathology. J Clin Psychiatry. 68:738Y746. Hay PI, Hall A (1991) The prevalence of eating disorders in recently admitted psychiatric in-patients. Br J Psychiatry. 159:562Y565. Hay PJ, Mond J, Nuttner P, Darby P (2008) Eating disorder behaviors are increasing. Findings from two sequential community surveys in South Australia. PloS One. 3:e1541. Hedges LV, Vevea JL (1998) Fixed- and random-effects models in meta-analysis. Psychol Methods. 3:486Y504. Helverskov JL, Clausen L, Mors O, Frydenberg M, Thomsen PH, Rokkedal K (2010) Trans-diagnostic outcome of eating disorders: A 30-month follow-up study of 629 patients. Eur Eat Disord Rev. 18:453Y463. Herzog DB, Keller MB, Lavori PW, Kenny GM, Sacks NR (1992) The prevalence of personality disorders in 210 women with eating disorders. J Clin Psychiatry. 53:147Y152. Herzog T, Stiewe M, Sandholz A, Hartmann A (1995) Borderline-Syndrom und Essto¨rungen. Literaturu¨bersicht und Interview-Studie an 172 konsekutiven InanspruchnahmepatientInnen der Freiburger Essto¨rungsambulanz. Psychother Psychosom Med Psychol. 45:97Y108. Hudson JI, Coit CE, LaLonde JK, Pope HG (2012) By how much will the proposed bew DSM-5 criteria increase the prevalence of binge eating disorder? Int J Eat Disord. 45:139Y141. Hudson JI, Hiripi E, Pope HG, Kessler RC (2007) The prevalence and correlates of eating disorders in the National Comorbidity Survey Replication. Biol Psychiatry. 61:348Y358.
Cassin SE, von Ranson KM (2005) Personality and eating disorders: A decade in review. Clin Psychol Rev. 25:895Y916.
Javaras KN, Laird NM, Reichborn-Kjennerud T, Bulik CM, Pope HG, Hudson JI (2008) Familiality and heritability of binge eating disorder: Results of a casecontrol family study and a twin study. Int J Eat Disord. 41:174Y179.
Chen EY, McCloskey MS, Michelson S, Gordon KH, Coccaro E (2011) Characterizing eating disorders in a personality disorders sample. Psychiatry Res. 185:427Y432.
Keel PA, Striegel-Moore RH (2009) The validity and clinical utility of purging disorder. Int J Eat Disord. 42:706Y719.
Cooper JL, Morrison TL, Bigman OL, Abrahamowitz SI, Blunden D, Nassi A, Krener P (1988) Bulimia and borderline personality disorder. Int J Eat Disord. 7:43Y49. Cumella EJ, Kally Z (2008) Profile of 50 women with midlife-onset eating disorders. Eat Disord. 16:193Y203.
124
www.jonmd.com
Kendall PC, Clarkin JF (1992) Introduction to special section: Comorbidity and treatment implications. J Consult Clin Psychol. 60:833Y834. Maran˜o´n I, Echeburu´a E, Grijalvo J (2007) Are there more personality disorders in treatment-seeking patients with eating disorders than in other kind of psychiatric patients? A two control groups comparative study using the IPDE. Int J Health Psychol. 7:283Y293.
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
The Journal of Nervous and Mental Disease
& Volume 202, Number 2, February 2014
Martin CK, Williamson DA, Thaw JM (2000) Criterion validity of the multiaxial assessment of eating disorders symptoms. Int J Eat Disord. 28:303Y310. Masheb RM, Grilo CM (2008) Examination of predictors and moderators for selfhelp treatments of binge-eating disorder. J Consult Clin Psychol. 76:900Y904. Milos G, Spindler A, Schnyder U, Fairburn CG (2005) Instability of eating disorder diagnoses: Prospective study. Br J Psychiatry. 187:573Y578. Milos GF, Spindler AM, Buddeberg C, Crameri A (2003) Axes I and II comorbidity and treatment experiences in eating disorder subjects. Psychother Psychosom. 72:276Y285. Picot AK, Lilenfeld LRR (2003) The relationship among binge severity, personality psychopathology, and body mass index. Int J Eat Disord. 34:98Y107. Pinheiro AP, Raney TJ, Thornton LM, Fichter MM, Berrettini WH, Goldman D, Halmi KA, Kaplan AS, Strober M, Treasure J, Woodside DB, Kaye WH, Bulik CM (2010) Sexual functioning in women with eating disorders. Int J Eat Disord. 43:123Y129. Piran N, Lerner P, Garfinkel PE, Kennedy SH, Brouillette C (1988) Personality disorders in anorexic patients. Int J Eat Disord. 7:589Y599. Pope HG, Lalonde JK, Pindyck LJ, Walsh T, Bulik CM, Crow SJ, McElroy SL, Rosenthal N, Hudson JI (2006) Binge eating disorder: A stable syndrome. Am J Psychiatry. 163:2181Y2183. Ramklint M, Jeansson M, Holmgren S, Ghaderi A (2010) Assessing personality disorders in eating disordered patients using the SCID-II: Influence of measures and timing on prevalence rate. Pers Individ Dif. 48:218Y223. Rockert W, Kaplan AS, Olmsted MP (2007) Eating disorder not otherwise specified: The view from a tertiary care treatment center. Int J Eat Disord. 40: S99YS103. Rosenvinge JH, Martinussen M, Østensen E (2000) The comorbidity of eating disorders and personality disorders: A meta-analytic review of studies published between 1983 and 1998. Eat Weight Disord. 5:52Y61. Rybakowsky F, Slopien A, Zakrzewska M, Hornowska E, Rajewski A (2004) Temperament and Character Inventory (TCI) in adolescents with anorexia nervosa. Acta Neuropsychiatr. 16:169Y174. Rø Ø, Martinsen EW, Hoffart A, Rosenvinge J (2005) Two-year prospective study of personality disorders in adults with longstanding eating disorders. Int J Eat Disord. 37:112Y118. Sansone RA, Levitt JL, Sansone LA (2005) The prevalence of personality disorders among those with eating disorders. Eat Disord. 13:7Y21. Schmidt NB, Telch MJ (1990) Prevalence of personality disorders among bulimics, nonbulimic binge eaters, and normal controls. J Psychopathol Behav Assess. 12:169Y185. Shedler J, Beck A, Fonagy P, Gabbard GO, Gunderson J, Kernberg O, Michels R, Westen D (2010) Personality disorders in DSM-5. Am J Psychiatry. 167: 1026Y1028. Skodol AE, Oldham JM, Hyler SE, Kellman HD, Doidge NN, Davies M (1993) Comorbidity of DSM-III-R eating disorders and personality disorders. Int J Eat Disord. 14:403Y416. Specker S, de Zwaan M, Raymond N, Mitchell J (1994) Psychopathology in subgroups of obese women with and without binge eating disorder. Compr Psychiatry. 35:185Y190. Stice E, Agras WS, Telch CF, Halmi KA, Mitchell JE, Wilson T (2001) Subtyping binge eating-disordered women along dieting and negative affect dimensions. Int J Eat Disord. 30:11Y27. Striegel-Moore RG, Franko DL (2007) Should binge eating disorder be included in the DSM-V? A critical review of the state of the evidence. Annu Rev Clin Psychol. 4:305Y324.
* 2014 Lippincott Williams & Wilkins
Personality Disorders in BED/EDNOS
Striegel-Moore RH, Garvin V, Dohm F-A, Rosenheck RA (1999) Psychiatric comorbidity of eating disorders in men: A national study of hospitalized veterans. Int J Eat Disord. 25:399Y404. Sunday SR, Peterson CB, Andreyka K, Crow SJ, Mitchell JE, Halmi KA (2001) Differences in DSM-III-R and DSM-IV diagnoses in eating disorder patients. Compr Psychiatry. 42:448Y455. Sysko R, Devlin MJ, Walsh BT, Zimmerli E, Kissileff HR (2007) Satiety and test meal intake among women with binge eating disorder. Int J Eat Disord. 40:554Y561. Telch C, Stice E (1998) Psychiatric comorbidity in women with binge eating disorder: Prevalence rates from a non-treatment seeking sample. J Consult Clin Psychol. 66:768Y776. Thomas JJ, Vartanian LR, Brownell KD (2009) The relationship between eating disorder not otherwise specified (EDNOS) and officially recognized eating disorders: Meta-analysis and implications for DSM. Psychol Bull. 135: 407Y433. Tyrer P, Crawford M, Mulder R (2011) Reclassifying personality disorders. Lancet. 377:1814Y1815. Van Hanswijck de Jonge P, Van Furth EF, Lacey JH, Waller G (2003) The prevalence of DSM-IV personality pathology among individuals with bulimia nervosa, binge eating disorder and obesity. Psychol Med. 33:1311Y1317. Vitousek K, Manke F (1994) Personality variables and disorders in anorexia nervosa and bulimia nervosa. J Abnorm Psychol. 103:137Y147. Vitousek KM, Stumpf R (2006) Difficulties in the assessment of personality traits and disorders in individuals with eating disorders. In Sansone RA, Levitt JL (Eds), Personality disorders and eating disorders: Exploring the frontier (pp 91Y117). New York: Routledge. Wentz E, Gillberg IC, Anckarsater H, Gillber C, Rastam M (2009) Adolescentonset anorexia nervosa: 18-year outcome. Br J Psychiatry. 194:168Y174. Wilfley DE, Friedman MA, Dounchis JZ, Stein RI, Welch RR, Ball SA (2000) Comorbid psychopathology in binge eating disorder: Relation to eating disorder severity at baseline and following treatment. J Consult Clin Psychol. 68:641Y649. Williamson DA (2007) Does the evidence point to a binge eating phenotype? Comment on Gordon et al. (2007) and Wonderlich et al. (2007). Int J Eat Disord. 40:S72YS75. Wilson GT, Grilo CM, Vitousek KM (2007) Psychological treatment of eating disorders. Am Psychol. 62:199Y216. Wonderlich SA, Gordon KH, Mitchell JA, Crosby RD, Engel SG (2009) The validity and clinical utility of binge eating disorder. Int J Eat Disord. 42: 687Y705. Wonderlich SA, Mitchell JE (1992) Eating disorders and personality disorders. In Yager J, Edelstein CK (Eds), Special problems in managing eating disorders (pp 51Y86). Washington, DC: American Psychiatric Press. Wonderlich SA, Swift WJ, Slotnick HB, Goodman S (1990) DSM-III-R personality disorders in eating-disorder subtypes. Int J Eat Disord. 9:607Y616. Yager J, Landsverk J, Edelstein CK, Hyler SE (1989) Screening for axis II personality disorders in women with bulimic eating disorders. Psychosomatics. 30:255Y262. Yanovski SZ, Nelson JE, Dubbert BK, Spitzer RL (1993) Association of binge eating disorder and psychiatric comorbidity in obese subjects. Am J Psychiatry. 150:1472Y1479. Yates W, Sieleni B, Reich J, Brass C (1989) Comorbidity of bulimia nervosa and personality disorder. J Clin Psychiatry. 50:57Y59. Zimmerman M, Francione-Witt C, Chelminski I, Young D, Tortolani C (2008) Problems applying the DSM-IV eating disorders diagnostic criteria in a general psychiatric outpatient practice. J Clin Psychiatry. 69:381Y384.
www.jonmd.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
125
