Risk Factors for Binge Eating Disorder:  A Community-Based, Case-Control Study FREE

BINGE EATING disorder is a new diagnostic concept, which is included in the appendix of DSM-IV, reserved for categories meriting further study.¹ The cardinal feature is binge eating, but, in contrast to bulimia nervosa, the binge eating is not accompanied by "the regular use of inappropriate compensatory behaviors" such as self-induced vomiting and misuse of laxatives.¹(p5501) The disorder commonly co-occurs with obesity.²

Research on binge eating disorder is still at an early stage,³ and its relationship with other eating disorders (particularly bulimia nervosa) continues to be debated.^4- 8 Its clinical features are being delineated,^3,9 and there have been a small number of controlled studies of its treatment.^10- 12 There has been little research on its causes or development. Several studies have focused on the timing of its onset,^13- 17 and reports exist of a raised rate of lifetime and family psychiatric disorder in the families of these patients.^7,18- 19 No comprehensive risk factor studies have been conducted.

We herein report the findings of a community-based, case-control study of risk factors for binge eating disorder. It complements a parallel and integrated study of risk factors for bulimia nervosa.²⁰ Our study had the following 3 aims: to identify risk factors for the development of binge eating disorder, to determine which of these risk factors are especially common among subjects with binge eating disorder compared with subjects with other psychiatric disorders, and to compare the risk factors for binge eating disorder with those for bulimia nervosa. On the basis of the clinical characteristics of the disorder, we made the following 2 main predictions: first, there would be 2 broad classes of risk factor for binge eating disorder, those that increase the risk for psychiatric disorder in general and those that increase the risk for obesity (parental and childhood obesity); and, second, compared with subjects with bulimia nervosa, those with binge eating disorder would have less exposure to risk factors for psychiatric disorder in general but more exposure to risk factors for obesity.

A case-control design was used with 3 related comparisons, corresponding to our 3 aims. Fifty-two women with binge eating disorder were compared with 104 without an eating disorder (healthy control subjects), 102 with other psychiatric disorders (general psychiatric control subjects), and 102 with bulimia nervosa. For the first comparison, both groups were individually matched within 1 year of age and within 3 parental social class bands²¹ (social classes I and II [high], III [middle], and IV and V [low]). Matching for current age reduced the risk for age-related recall bias, and matching for parental social class removed a potential confounding variable. To match the subjects for the time available for exposure, the healthy controls were questioned about their life until the age of onset of disturbed eating (index age) of their particular matched subject with binge eating disorder. The second and third comparisons were unmatched and used subjects who had been recruited for the study of bulimia nervosa.²⁰ These analyses were undertaken after adjusting for current age, parental social class, and index age.

Unless otherwise indicated, data are given as mean ± SD.

The recruitment and assessment strategy is described in detail in the article on risk factors for bulimia nervosa.²⁰ The sampling frame was the patient registers of 23 general practices situated in urban and rural parts of Oxfordshire, England. All women aged from 16 to 35 years listed on these registers were sent 2 self-report questionnaires: the Eating Disorder Examination-Questionnaire (EDE-Q), a questionnaire designed to detect eating disorders,^22- 25 and the 30-item General Health Questionnaire (GHQ),²⁶ an instrument designed to detect psychiatric disorder. Subjects who did not reply were sent a second set of questionnaires. Those whose responses on the EDE-Q suggested that they might have binge eating disorder were invited for an interview to establish their true case status. Subjects who met diagnostic criteria underwent a further interview to identify their age at onset of the eating disorder (see below) and to assess their exposure to putative risk factors for eating disorders. The subjects with bulimia nervosa were recruited in an equivalent manner.²⁰

Potential healthy controls were identified from their responses on the EDE-Q, and potential general psychiatric controls, from their scores on the GHQ, with true case status being established by interview. The subjects had their height and weight measured, and informed consent was obtained.

Subjects with eating disorders had to meet strict operational definitions of DSM-IV binge eating disorder or bulimia nervosa based on ratings on the EDE interview.^27- 29 The healthy controls were required to have no current or past eating disorder, established by interview. There were no other exclusion criteria. The general psychiatric controls had to have a current Axis I DSM-III-R psychiatric disorder, confirmed by interview using the Structured Clinical Interview for DSM-III-R.^30- 31 Those with a current or past eating disorder were excluded.

Exposure to putative risk factors for eating disorders was assessed by interviewing the subjects (mostly in their homes). The risk factor interview focused on the period before onset of the eating disorder, with onset (index age) being conservatively defined as the age at which the first significant and persistent behaviors characteristic of an eating disorder began (sustained dieting, regular episodes of overeating, self-induced vomiting, or laxative misuse) rather than the age at which the subject first met full diagnostic criteria for the disorder.²⁰ The presence of 5 risk factors was assessed before and after onset of the eating disorder, since they might have a hereditary influence (Table 1). The subjects also completed the Parental Bonding Instrument,³² a self-report measure of subjects' experience of both parents during the first 16 years of life.

The risk factor interview was investigator-based and used behavioral definitions of key concepts to minimize the problems associated with retrospective reporting.³³ Further details are provided in the article on bulimia nervosa²⁰ and a related article on sexual abuse.³⁴ (A copy of the interview schedule may be obtained by writing to C.G.F. A charge will be made to cover the cost of postage and photocopying.) A wide range of putative risk factors was assessed, and these were categorized a priori into domains and subdomains, each reflecting certain types of exposure (Table 1). Additional factors, which did not naturally belong to any of the domains, were age at onset of menstruation, a history of pregnancy or abortion, and parity.

The first comparison used conditional logistic regression analyses appropriate for a case-control design with individual matching.^35- 36 The other 2 comparisons used logistic regression analyses appropriate for an unmatched case-control design.^35- 36 The relationships between individual putative risk factors and case status were first assessed univariately. Each risk factor was considered as a single indicator variable and coded 0 for "no" or 1 for "yes." Statistical significance was assessed using the χ² likelihood ratio statistic and was set at the 1% level (P<.01) to achieve levels of statistical power comparable with those in the bulimia nervosa case-control study.²⁰ An overall measure of exposure to risk in each subdomain and domain was obtained by summing the numbers of component factors to which each subject had been exposed. The resulting index scores were grouped into categories of as equal size as possible, with varying numbers for the subdomains and 4 for the domains. The relationships between case status and exposure in each subdomain and domain were examined univariately and, to assess the relative importance of different types of exposure, in multiple stepwise logistic regression analyses. To test for the significance of any apparent linear trend, the categorized scores (each assigned a value of average exposure) were entered in factored and unfactored forms. Statistical significance for the subdomain and domain analyses was set at the 5% level (P<.05).

Subjects with binge eating disorder had a mean age of 24.5 ± 5.7 years and a parental social class distribution as follows: high (social classes I or II), 22 (43.1%); middle (III), 18 (35.3%); and low (IV or V), 11 (21.6%) (social class information was not available for 1 subject). Mean index age was 16.8 ± 4.6 years. Mean body mass index was 26.9 ± 5.5, and 22.2% (10 of the 45 subjects who provided this information) had a body mass index of more than 30. Few subjects had ever used extreme methods of weight control (4 subjects [7.7] had induced vomiting; 1 subject [1.9] had misused laxatives), and none had a history of bulimia nervosa. None were currently in treatment.

Healthy controls were individually matched to the subjects with binge eating disorder in terms of age and parental social class, with the result that the distribution of these characteristics was very similar.

Subjects with bulimia nervosa had a mean age of 23.7 ± 4.9 years and a parental social class distribution (which was not significantly different from that of the subjects with binge eating disorder) as follows: high (social classes I or II), 47 (46.1%); middle (III), 46 (45.1%); and low (IV or V), 9 (8.8%). Mean body mass index was 24.3 ± 4.6, and 12% (10 of the 84 subjects who provided this information) had a body mass index of more than 30. Mean index age was 15.5 ± 3.9 years, and 89% (85 of the 95 subjects who provided this information) were not in treatment.

General psychiatric controls were individually matched to the subjects with bulimia nervosa in terms of age and parental social class. Their diagnoses fell into the following principal categories: major depressive disorder, 83 (81.4%); bipolar disorder, 1 (1.0%); and an anxiety disorder, 18 (17.6%) . The overrepresentation of depressive disorder may have been the result of subjects with chronic anxiety disorder being missed by the GHQ.³⁷ Seventy (68.6%) were not receiving treatment.

Subjects with binge eating disorder reported greater levels of exposure to the following personal vulnerability factors: negative self-evaluation, parental depression (ever) (for both, P<.001), major depression (P=.003), marked conduct problems (P=.002), and deliberate self-harm (P=.004). With regard to the environmental factors, they reported greater levels of exposure to the following parental problems: parental criticism, high expectation, minimal affection (for all 3, P<.001); parental underinvolvement; and maternal low care and high overprotection (for both, P=.002). In addition, they were more likely to report sexual (P=.001) and repeated severe physical abuse (P=.01) and bullying (P=.005). They reported greater exposure to just the following 2 dieting vulnerability factors: critical comments by family about shape, weight, or eating (P<.001) and teasing about shape, weight, eating, or appearance (P=.01). They were also more likely to have been pregnant before index age (P<.001).

Subjects with binge eating disorder reported a greater level of exposure to the following 7 of the 12 subdomains: parental psychiatric disorder (ever), parental problems, dieting risk (for all 3, P<.001), childhood characteristics (P=.002), sexual and physical abuse (P=.003), obesity risk (P=.007), and disruptive events (P=.04). In each case, the more exposure, the greater the risk was for developing binge eating disorder. The following 3 subdomains entered the multiple regression model: parental problems (χ²₂=26.8; P<.001), parental psychiatric disorder (ever) (χ²₂=9.6; P=.008), and dieting risk (χ²₂=6.8; P=.03), 1 from each of the 3 domains.

On univariate analysis, subjects with binge eating disorder had a significantly greater level of exposure to all 3 domains (Table 2). The odds ratios increased linearly from those subjects exposed to the smallest number of factors to those exposed to the largest. Two domains entered the multiple regression model (Table 3). Exposure in the environmental domain was entered first (χ²₃=33.6; P<.001), followed by exposure in the dieting vulnerability domain (χ²₃=10.1; P=.02). There was no significant statistical interaction between the domains, suggesting that the effect of exposure in any 1 domain did not depend on the degree of exposure in the other and that the effect of combined exposures was additive.

Subjects with binge eating disorder had a greater level of exposure to 1 factor from the personal vulnerability and environmental domains, low parental contact (P<.001), and 2 of the dieting vulnerability factors: critical comments by family about shape, weight, or eating and childhood obesity (for both, P=.006).

Both groups did not differ significantly in their exposure to any of the personal vulnerability or environmental subdomains. Subjects with binge eating disorder, however, had been exposed to a significantly greater extent to all 3 dieting vulnerability subdomains. Only obesity risk, however, entered the multiple regression model (χ²₁=8.1; P=.004).

The two groups only differed significantly with respect to the dieting vulnerability domain (χ²₃=8.74; P=.03), with binge eating disorder 3 times as likely to develop in those subjects who had been exposed to 3 or more factors as those with 2 or fewer exposures (Table 2 and Table 3).

There were no statistically significant differences between the 2 groups with respect to individual risk factors.

Subjects with binge eating disorder reported less exposure to 2 of the subdomains: childhood characteristics (P=.001)and obesity risk (P=.02). Childhood characteristics entered the multiple regression model first (χ²₂=12.9; P=.002), followed by obesity risk (χ²₁=5.24; P=.02).

On univariate analysis, the two groups differed with respect to just 1 of the 3 domains, personal vulnerability, the level of exposure being lower in those with binge eating disorder (χ²₃=7.72; P=.05) (Table 2). The odds ratios decreased in a linear fashion, from those subjects exposed to the smallest number of factors to those exposed to the greatest. On multiple regression analysis, after adjusting for exposure in the personal vulnerability domain, no other domain had an independent relationship with case status (Table 3), although the subdomain of obesity risk (χ²₁=4.58; P=.03) remained significantly associated with case status.

To our knowledge, this is the first study of risk factors for the development of binge eating disorder. Certain aspects of our methods are of note. First, recruitment was direct from the community, thereby avoiding the sampling bias that is likely to arise from focusing on subjects who are seeking treatment. Such bias applies in bulimia nervosa³⁸ and is likely to apply in binge eating disorder. Second, we assessed a large number of putative risk factors, with the result that we were able to assess their relative contributions and possible interactions. Third, we had 3 comparison groups. The use of healthy controls made it possible to identify which of the putative risk factors were risk factors for binge eating disorder; the general psychiatric controls made it possible to determine which risk factors were particularly common among subjects with binge eating disorder compared with those with other psychiatric disorders; and the comparison with subjects with bulimia nervosa meant that we were able to compare the risk factors for these 2 eating disorders.

The comparison between subjects with binge eating disorder and healthy controls suggested that only certain of the risk factors implicated for eating disorders are risk factors for binge eating disorder. The factors with a major effect were negative self-evaluation; parental depression (ever); adverse childhood experiences (sexual and physical abuse and a range of parental problems); repeated exposure to negative comments from family members about shape, weight, or eating; and pregnancy before onset. One difference of note emerged from the comparison with the general psychiatric controls. This concerned factors belonging to the dieting vulnerability domain, most especially childhood obesity and negative comments from the family about shape, weight, and eating, with exposure being higher among subjects with binge eating disorder. Taken together, these findings support the prediction that binge eating disorder would be associated with exposure to 2 broad classes of risk factor: those that increase the risk for psychiatric disorder in general and those that increase the risk for obesity. Similarly, the findings of our bulimia nervosa risk factor study implicated exposure to the combination of general psychiatric risk factors and risk factors for dieting.²⁰

The comparison between subjects with binge eating disorder and those with bulimia nervosa revealed no differences with respect to individual risk factors, although 2 findings of note emerged from the subdomain and domain analyses. The main finding concerned the personal vulnerability domain, and most especially the childhood characteristics subdomain, where the exposure rates were higher among subjects with bulimia nervosa. Obesity risk (childhood and parental obesity) was also more common among subjects with bulimia nervosa, but the difference was less marked.

The findings from the first and third of the case-control comparisons, taken in conjunction with those from our study of risk factors for bulimia nervosa,²⁰ support the prediction that, compared with subjects with bulimia nervosa, those with binge eating disorder would report less exposure to risk factors for psychiatric disorder in general. When subjects with binge eating disorder were compared with their matched healthy controls, far fewer differences in exposure emerged than in the equivalent comparison involving subjects with bulimia nervosa,²⁰ and the strengths of the associations were in general weaker. This was true of risk factors for psychiatric disorder in general and dieting vulnerability factors. The findings of the direct (but statistically less powerful) comparison of both groups with eating disorders suggest that 2 classes of risk factor in particular differentiate the 2 disorders. The first consists of certain childhood traits, particularly negative self-evaluation, perfectionism, shyness, and extreme compliance, all of which appear to be especially common antecedents of bulimia nervosa. Whereas shyness and extreme compliance were also common antecedents of general psychiatric disorder, negative self-evaluation and perfectionism appeared to be especially characteristic of subjects in whom bulimia nervosa subsequently developed, which is in accord with cognitive accounts of the development of the disorder.^39- 40 The second distinctive risk factor is vulnerability to obesity, which, although increased in binge eating disorder, seems to be even more pronounced in bulimia nervosa. This finding is in contradistinction to our prediction, based on the observed clinical association between binge eating disorder and obesity,² that exposure to risk factors for obesity would be more marked in binge eating disorder. The strong association between bulimia nervosa and vulnerability to obesity may be disguised by the following 2 processes: the intensity of these subjects' efforts to avoid fatness, resulting in unremarkable body weight,⁴¹ and the influence of a sampling bias that results in subjects who are particularly vulnerable to obesity being underrepresented among those subjects who present for treatment.³⁸

Certain limitations of our study should be noted. First, we relied exclusively on information provided by the subjects. Other informants could not be used, since secrecy characterizes binge eating disorder and bulimia nervosa.^42- 43 Few of these community-based subjects would have given us permission to contact other informants. Second, it was not feasible for the risk factor interview to be conducted by assessors who were unaware of the subjects' case status, since this would have necessitated separate interviews with 2 different people, an arrangement that few subjects would have accepted. Third, the study was retrospective in nature and therefore vulnerable to the problems associated with retrospective reporting. In this context, it should be noted that the level of concern about the retrospective reports of subjects with psychiatric disorders is probably excessive; as Brewin et al³³(p82) note, "there is little reason to link psychiatric status with less reliable or less valid recall of early experiences." Furthermore, in our study, many such problems are likely to have been minimized by the use of an investigator-based assessment interview in which concepts were defined in behavioral terms. Nevertheless, a prospective study would have been preferable although extremely difficult, given the low incidence of eating disorders. The last point of note concerns the nature of the subjects with binge eating disorder who, although likely to be representative of many subjects in the community, differed in 2 main ways from those seen in clinics. First, they were younger, most patients with binge eating disorder being in their 40s,³ and second, fewer had comorbid obesity.³ Finally, like many other investigators, we focused on women, whereas it has been reported that a substantial minority of those with binge eating disorder are men.^44- 45 Future studies will need to determine whether our findings can be generalized to men with the disorder and to those who present for treatment.

Accepted for publication November 18, 1997.

Funded by program grant 18157 from the Wellcome Trust; Wellcome Principal Research Fellowship 046386 (Dr Fairburn); grants 030583 and 13123 from the Wellcome Trust (Drs Doll and Welch), London England; a Nuffield Medical Fellowship, University of Oxford, Oxford, England (Dr Hay); and a Girdlers' Junior Research Fellowship, University of Oxford (Dr Hay).

We thank George Brown, PhD, Bedford College, London, England, and Emily Simonoff, MD, Institute of Psychiatry, London, for their advice on the design of the risk factor interview; David Jones, MD, Park Hospital, Oxford, for advice on the assessment of sexual abuse; the participating general practitioners; and Faith Barbour, Jenny Burton, Valerie Dunn, Jan Francis, and Sue Shaw for their work as interviewers.

Corresponding author: Christopher G. Fairburn, DM, MPhil, FRCPsych, Department of Psychiatry, Warneford Hospital, Oxford University, Oxford OX3 7JX, England.