Association Between Eating Disorders and Low Self-esteem:

                                                                                              A Meta-analysis Study


                                                                      Lori L. Badura, PhD1,  and R. John Parker, CEO2
                                                                             Percepts, LLC, Canterbury, CT 06331

  Abstract

Objective: A meta-analysis was conducted to evaluate the association between eating disorders (ED) and low self-esteem (SE).  Method: Of 425 peer-reviewed articles identified through a series of literature searches, 10 met the overall criteria for inclusion in this analysis.  Two separate analyses were conducted, one on the complete set of 10 studies, and one on a subset of 8 articles that had more rigorously defined experimental and control groups.  Results: In both cases, the results indicated that ED is substantially associated with low self-esteem, mean effect sizes of -2.0 and -2.2 respectively.  Conclusion: These results are discussed with relevance to the potential role of SE in the etiology and treatment progression of ED.

 Introduction

Eating disorders (ED), while of relatively low incidence in the general population (1), represent a potentially life-threatening class of disorders that are historically difficult to treat, as well lacking in predictive markers of disease risk and progression.  ED typically fall into three general classes; anorexia nervosa (AN), characterized by excessive food restriction and pervasive body perception distortion,  bulemia (BU) involving binge eating typically with purging, and a third category of variable and diffuse traits classified as “eating disorder not otherwise specified” (EDNOS). 

While these forms of ED are symptomatically distinct, low self-esteem (SE) has been cited as a covariant in virtually every scientific report on ED where SE has been assessed concurrently.  Most clinicians and researchers readily agree that low SE is present in ED, yet it is surprising that the authors could locate only one review article that directly proposed a role for SE in the etiology of ED (2). Furthermore, no systematic analysis of the relationship between ED and SE appears to have been conducted to date.

However, there is some interest being generated amongst researchers as to whether low SE is a causative factor in ED, or simply exists as a comorbid trait. It has even been suggested that low SE can account for most, if not all, of the symptoms associated with ED. What is known is that it is extremely difficult to treat a condition if the cause is unknown. Therefore, evidence regarding a predictive association between SE and ED would be beneficial in helping to develop more efficacious therapies and better predictive screening modalities. SE has already shown potential as an indicator of treatment progression.  For example, Halvorson and Heyerdahl (3) reported data from AN patients who were actively exhibiting the disorder, those who were in the process of recovering but under strain, and those who were classified as recovered.  The SE levels for these groups tracked with the diagnostic classification (means = 8.9, 10.1,and 11.9 respectively). 

The current report seeks to review and assess the broad relationship in the literature between ED and SE by surveying those studies that compared SE in relevant ED and normal control groups.

Methods

Literature Search Parameters

            Multiple searches of Medline (1950-present) and Current Contents (1996-present) on keyword combinations of “self-esteem, anorexia” or “self-esteem, eating disorders” were conducted.  Once duplicate hits were removed, 425 articles remained for potential inclusion.  The abstracts for these articles were reviewed against the following criteria: experimental articles, published in English, contain a population with eating disorder, gender ,  age range > 16 years, and SE scores reported. This search gave a pool of 36 articles.  A thorough review was then conducted on these 36 articles, with further experimental criteria of comparison control group and single sample design, resulting in 10 articles for consideration in the analysis

Article Coding

            Articles were coded and data recorded on a number of demographic and experimental variables: country of origin, type of study, type of analysis (all were between group), experimental groups examined, sample size of selected comparison groups, % of sample that was female, age, body mass index, number of sample points, type of concurrent treatment, type of SE measure, means and standard deviations for SE, and F and p values where reported.  Articles were included in the analysis pool if they included either an AN or combined ED group for comparison. 

Statistical Analysis

            Effect sizes and pooled variance were calculated as standardized mean differences as per the random effects model method of Lipsey and Wilson (4).    P values for heterogeneity in this method are calculated by comparing the Q statistic with a Chi Square distribution based upon (k-1) degrees of freedom (k = number of studies).  One article (5) did not report the standard deviations, and thus they were estimated using the reported MS between and F ratios.  Of the 10 articles identified for inclusion, two articles did not have exact matches with the set criteria – Corning, et. al. (6) used an “ED symptomatic” experimental group of non-patients that scored high on ED characteristics using the Questionnaire for Eating Disorder Diagnoses scale and Halvorsen and Heyerdahl (3) used a control group of individuals who had recovered from AN and were currently asymptomatic.  In order to investigate the relationship between SE and ED as broadly as possible, two separate analyses were conducted – one on the total pool of 10 articles, and one with only the 8 articles that most closely aligned with the inclusion criteria

Results

            Table 1 summarizes the relevant demographic information for the 10 articles used in the subsequent analyses.  In those cases where more than one type of experimental group was included, only the most relevant ED group was included. For example, in Halvorson and Heyerdahl (3), data were reported for AN and for “AN unrecovered” and “AN strained” groups, but only the data for “AN unrecovered” was used in the analysis to reflect the ED experimental group.  In addition, some articles reported data from more than one type of control group, and the group most closely representing a “normal” population was chosen for the comparison.  For example, Geller, et. al, (7) reported data from a psychiatric control group that were not included in the analysis.  Similarly, Wilksch and Wade (8) reported data from two control groups: “normal restrained eaters” and “normal unrestrained eaters.”  Only the data from “normal unrestrained eaters” were included in the present analyses.  Table 1 identifies the comparison groups for each article that were selected from each study for the analyses.


Table1.  Demographic characteristics of articles included in meta-analyses.

 


 




 

 

 

The random effects model method of calculating the mean effect size for all 10 candidate studies (see Figure 1) revealed that individuals actively expressing ED, or ED symptomology, have substantially lower SE than normal controls, mean = -2.0, SE = 0.244, z = - 8.24, CL upper = -1.530, CL lower = -2.485.  A second analysis was conducted excluding two studies that had ED or control group criteria deviating slightly from the other investigations (i.e.,ED symptomology rather than formal diagnosis, and “AN recovered” rather than a normal control group never having expressed AN symptomology).  Re-analysis without these two studies yielded an even stronger effect, mean -2.21, SE = 0.191, z = -11.57, CL upper = -1.835, CL lower = -2.584, indicating an even stronger association between ED and low SE.

 

Figure 1.  Mean effect sizes (+ upper and lower confidence levels) for each study and overall effect size for entire cohort of 10 articles

Figure 2.  Mean effect sizes (+ upper and lower confidence levels) for each study and overall effect size for cohort of 8 articles with more rigorously defined experimental and control groups

Discussion

     The present results suggest a very strong relationship between low SE and the existence of an ED, even collapsing across various subtypes of ED.  According to conventional nomenclature, meta-analysis approaches use the following classification criteria for determining magnitude of the effect size: <0.20 = small, 0.50 = medium, > 0.80 = large (Lipsey & Wilson, 2001). The magnitude of the effect sizes obtained in the current analyses (-2.0 and -2.2) are therefore extremely large, and indicate virtually no overlap between control and ED groups. These results solidify the increasing belief within the clinical field that chronic low self-esteem is a critical component underlying the etiology of ED.

            Most therapeutic approaches for ED focus primarily upon restoration of nutritional intake and maintenance of stable weight, and rightfully so.  However, once these life-threatening factors are addressed, it may be most beneficial to center the continuing intervention upon techniques that improve SE.  The strength of the association between SE and ED reported here suggests that directive therapy toward SE would enhance the likelihood of recovery, shorten the time necessary for medical intervention, and serve to protect against relapse.  Indeed, one could argue that every therapeutic approach to ED should include a strong focus on SE.

Finally, the idea of a much greater causal role for SE in psychological disorders than previously recognized leads us to consider the possibility that we have grossly underestimated the impact of SE phenomenon on development and stability of ED.  One area of interest for researchers and clinicians is the ability to have early detection tools for those at risk for a disorder. Since not everyone with low SE develops an ED, such strongly entrenched behaviors as anorexia and bulimia do so because they fit with a particular configuration of SE (14). As such, there must be some potentially identifiable aspects of the global SE profile that closely correlate with the development of an ED.  Therefore, future research should survey a broader scope of relevant SE characteristics, including both explicit and implicit markers of SE, in terms of their relevance to both ED prevalence and potential for predicting treatment response. 

References

1. Heok, HW  The incidence and prevalence of anorexia nervosa and bulimia nervosa in primary care.  Psych.Medicine  1991; 21;  455-460.

2. Silverstone, PH   Is chronic low self-esteem the cause of eating disorders?  Med Hypotheses 2992; 39;  311-315

3. Halvorsen, I; Heyerdahl, S Girls with anorexia nervosa as young adults: Personality, self-esteem, and life satisfaction  Int J Eat Disorders 2006; 39:  285-293

4. Lipsey, MW,  Wilson, DB  Practical metaanalysis.  Applied Social research methods Series volume 49 London: Sage Publications; 2001.

5. Huon, G,  Brown, LB Psychological correlates of weight control among anorexia nervosa patients and normal girls. Brit J Med Psych 1984; 57; 61-66

6. Corning, AF, Krumm, AJ, Smitham, LA  Differential social comparison processes in women with and without eating disorder symptoms  J Coun Psych 2006; 53;  338-349

7. Geller, J, Johnston, C, Madsen, K, Goldner, EM, Remick, RA, Birmingham, CL.  Shape- and weight-based self-esteem and the eating disorders  Int J Eat Disorders 1998; 24;  285-298

8. Wilksch, S; Wade, TD  Differences between women with anorexia nervosa and restrained eaters on shape and weight concerns, self-esteem, and depression  Int J Eat Disorders 2004; 35;  571-578

9. McFarlane, T, McCabe, RE, Jarry, J, Olmstead, MP, Polivy, J  Weight-related and shape-related self-evaluation in eating-disordered and non-eating-disordered women  Int J Eat Disorders 2001; 29;  328-335

10. Jacobi, C, Paul, T, de Zwann, M, Nutzinger, DO, Dahme, B  Specificity of self-concept disturbances in eating disorders  Int J Eat Disorders 2004; 35;  204-210

11. Mendelson, BK, McLaren, L, Gauvin, L, Steiger, H  The relationship of self-esteem and body esteem in women with and without eating disorders Int j Eat Disorder 2002; 31;  318-323.

12. Williams, GJ, Power, KG, Millar, HR, Freeman, CP, Yellowlees, A, Dowds, T, Walker, M, Campsie, L, MacPherson, F., Jackson, MA  Comparison of eating disorders and other dietary/weight groups on measures of perceived control, assertiveness, self-esteem, and self-directed hostility  Int J Eat Disorders 1993; 14;  27-32

13. Cockell, SJ, Hewitt, PL, Seal, B, Sherry, S, Goldner, EM, Flett, GL, Remick, RA   Trait and self-presentational dimensions of perfectionism among women with anorexia nervosa  Cog Ther Res 2002; 26;  745-758 

14. Parker, RJ. An introduction to self esteem.  www.percepts.us  2007; 1-7.