Risk Factors for Eating Disorders Ruth H. Striegel-Moore Wesleyan University Cynthia M. Bulik University of North Carolina at Chapel Hill The authors review research on risk factors for eating disorders, restricting their focus to studies in which clear precedence of the hypothesized risk factor over onset of the disorder is established. They illustrate how studies of so- ciocultural risk factors and biological factors have pro- gressed on parallel tracks and propose that major ad- vances in understanding the etiology of eating disorders require a new generation of studies that integrate these domains. They discuss how more sophisticated and novel conceptualizations of risk and causal processes may inform both nosology and intervention efforts. Keywords: eating disorders, epidemiology, cultural risk factors, genetic risk factors EStevenson, ating disorders rank among the 10 leading causes of disability among young women (Mathers, Vos, & Begg, 2000), and anorexia nervosa has the highest mortality rate of all mental disorders (Millar et al., 2005 Sullivan, 1995 Zipfel, Lowe, Reas, Deter, & Herzog, 2000). Fueled by these ominous indicators of their clinical significance, efforts to identify risk or causal fac- tors for eating disorders are indicated for at least four critical reasons. First, identification of causal mechanisms satisfies the need to understand why certain people develop the problem in question and others do not. Indeed, some experts believe that such an understanding helps decrease the stigma associated with a mental disorder: If the cause is seen as out of the individual���s control, less blame is as- signed than if the disorder is seen as ���volitional��� (Crisp, Gelder, Rix, Meltzer, & Rowlands, 2000). A Newsweek magazine story titled ���Fighting Anorexia: No One to Blame��� stressed recent findings on genetic vulnerability to explain risk for the development of anorexia nervosa (Tyre, 2005). Second, ideally, nosology is based on etiology, yet the current classification schema for eating disorders, as artic- ulated in the Diagnostic and Statistical Manual of Mental Disorders (4th ed. [DSM���IV] American Psychiatric Asso- ciation, 1994), is based solely on the observed clustering of signs and symptoms. The eating disorder criteria remain the subject of considerable debate, in large part because they fail to result in clearly defined subgroups or to account for changing symptomatology over the course of the ill- ness. Most individuals who experience a clinically signif- icant eating disorder do not meet diagnostic criteria for anorexia nervosa or bulimia nervosa but, rather, meet cri- teria for eating disorder not otherwise specified (EDNOS), a diagnosis intended to capture a residual group (Hoek & van Hoeken, 2003 Hudson, Hiripi, Harrison, & Kessler, 2005 Striegel-Moore et al., 2005). Binge-eating disorder (BED) is the most widely studied specific example of an EDNOS (for review, see M. J. Devlin, Goldfein, & Do- brow, 2003). Also, diagnostic crossover is common, espe- cially from anorexia nervosa to bulimia nervosa (Tozzi et al., 2005) or from bulimia nervosa to BED (Fichter, Quad- flieg, & Hedlund, 2005). Risk-factor studies present an untapped source of information of potential value for re- vising the current classification system. Third, treatment is best accomplished when we know the causes of a disorder. The current evidence base for treatment of anorexia nervosa, in particular, is weak (Berk- man, Bulik, Brownley, & Lohr, in press). A classic exam- ple from the history of psychiatry is the treatment of neurosyphilis. As recently as the mid-19th century, ad- vanced neurosyphilis with psychiatric manifestations was a ticket to treatment in an asylum. With the discovery of the Spirochaeta pallida (Treponema pallidum Schaudinn & Hoffman, 1905), the search for biological cures ultimately led to the discovery of the efficacy of penicillin in the treatment of this sexually transmitted disease, leading to a rapid decline in incidence and changing treatment forever. Although the story with anorexia nervosa is unlikely to be so dramatic, this transformation of syphilis treatment from asylums to penicillin illustrates the power of understanding etiology in the search for effective interventions. Fourth, identification of risk factors is important for determining high-risk groups for targeted interventions, designing prevention program content, and informing public policy. For example, a growing literature on cross- generational transmission of eating disorders, termed cy- cles of risk (Bulik, Reba, Siega-Riz, & Reichborn-Kjenne- rud, 2005), suggests that targeted prevention strategies should be tested that focus on the offspring of women with eating disorders. Through the lens of sociocultural models, many prevention programs feature media literacy as a key component (Stice & Shaw, 2004). Finally, legislative ef- forts have included the introduction of laws prohibiting false advertising by the diet industry. Ruth H. Striegel-Moore, Department of Psychology, Wesleyan Univer- sity Cynthia M. Bulik, Department of Psychiatry, University of North Carolina at Chapel Hill. Correspondence concerning this article should be addressed to Ruth H. Striegel-Moore, Department of Psychology, Wesleyan University, 207 High Street, Middletown, CT 06459. E-mail: rstriegel@wesleyan.edu 181 April 2007 ��� American Psychologist Copyright 2007 by the American Psychological Association 0003-066X/07/$12.00 Vol. 62, No. 3, 181���198 DOI: 10.1037/0003-066X.62.3.181

The single best predictor of risk for developing an eating disorder is being female, prompting the question why women? And indeed, most studies have been restricted to girls or women yet, clearly, female sex is not a sufficient condition for explaining risk, prompting a second question: Which women in particular? (Striegel-Moore, Silberstein & Rodin, 1986). In the present review, we describe the state of knowledge of risk and causal factors for eating disorders as gleaned from studies on human populations. This illustrates that the ���prototypical��� eating disorder case is a young, White, middle- or upper-middle class woman residing in Europe or North America, reflecting in part the legacy of defining eating disorders on the basis of White girls or women and the common practice in the field of relying on patient samples (biasing samples toward White, educated, and more affluent individuals). Our understand- ing of risk factors will remain woefully incomplete until research definitions and sampling strategies accommodate diversity in the population within and across nations. We also address how risk-factor knowledge contributes to de- stigmatization of eating disorders, suggests revisions for the next DSM and International Classification of Diseases (World Health Organization, 1992), and informs treatment and prevention interventions. We conclude with recom- mendations for future studies and future thought about eating disorders. Changing Perspectives on Risk Factors for Eating Disorders Eating disorders (compared with many other psychiatric disorders) are newcomers to the official psychiatric nomen- clature. Since their earliest appearances in the literature, the field has been bifurcated in the search for etiological ex- planations. In this section, we review briefly the history of our understanding of anorexia nervosa and bulimia ner- vosa. Anorexia Nervosa Anorexia nervosa was introduced as a new illness in the late 19th century in separate yet almost simultaneous ac- counts by British psychiatrist William Gull (1874, 1888) and French physician Charles Lase `gue (1873). Although both Gull and Lase `gue characterized anorexia nervosa as a ���nervous��� disease, each man emphasized different aspects of his patients��� clinical presentation. Gull���s clinical de- scription elaborated on the physiological correlates that resulted from a ���perversion of the will��� and attributed them to ���simple starvation��� without detailed discussion of what might have caused this ���perversion.��� Lase `gue described anorexia nervosa as a ���hysteria of the gastric center��� and paid considerable attention to the psychological or social factors that he believed were involved in the development of this disorder (cited in Brumberg, 2000, pp. 118���119). At the risk of oversimplification, Gull focused more on bio- logical processes and disruptions, whereas Lase `gue fo- cused more on psychosocial and psychological roots. In both cases, however, the treatment was somatic, involving nutritional restoration. Nearly a century later, Hilde Bruch (1973, 1979) introduced a biopsychosocial conceptualiza- tion of anorexia nervosa that emphasized developmental factors and family dynamics. Theoretical models prolifer- ated in the 20th century ranging from the purely biological (Kaye, Frank, et al., 2005) to the biopsychosocial (Connan, Campbell, Katzman, Lightman, & Treasure, 2003) to the purely cultural (Bordo, 1993 Orbach, 1986), yet studies of risk factors typically have not tested comprehensive etio- logic models that incorporate biological, psychosocial, and environmental factors (for reviews, see Bulik et al., 2005 Jacobi, Hayward, de Zwaan, Kraemer, & Agras, 2004 Striegel-Moore & Cachelin, 2001). Bulimia Nervosa Russell (1979) described bulimia nervosa as ���an ominous variant of anorexia nervosa,��� and although he widely has been credited with introducing this disorder into the no- menclature, a few years prior, Boskind-Lodahl (1976) of- fered a feminist formulation of the binge���purge syndrome as a culture-bound syndrome that arose from Western cul- ture���s obsession with female thinness in particular and the restrictions of female gender role stereotypes in general. As with anorexia nervosa, numerous theoretical models pro- liferated, yet most risk-factor studies have tested variables representing various risk domains rather than testing a particular theoretical model (Jacobi et al., 2004 Stice, 2002 Striegel-Moore et al., 1986 for review, see Striegel- Moore & Cachelin, 2001). Nevertheless, too often, discus- sion of the etiology of eating disorders becomes polarized into ���cultural��� versus ���biological��� explanations that ignore the fact that biological and environmental variables are inextricably linked. Ruth H. Striegel-Moore 182 April 2007 ��� American Psychologist

Advances in Conceptualizing Risk Fundamentally, risk-factor research seeks to understand the factors that cause an outcome of interest (a full discussion of the epistemological and methodological challenges of such research is beyond the scope of this article see, e.g., Haynes, 1992). A risk factor is a characteristic (e.g., allele), event (e.g., teasing), or experience (e.g., growing up in a culture that values extreme thinness) that precedes the onset of the outcome of interest (e.g., an eating disorder) and that, ���if present, is associated with an increase in the probability (risk) of a particular outcome over the base rate of the outcome in the general (unexposed) population��� (Kazdin, Kraemer, Kessler, Kupfer, & Offord, 1997, p. 377). For causality to be inferred, it further needs to be shown that the association between the risk factor and the outcome is not due to confounding influences, that the results are replicable, and that there are plausible explana- tions for the processes mediating the relation between the hypothesized factor and the outcome (Kazdin et al., 1997). No one study can meet all of these requirements rather, risk-factor research involves piecing together the puzzle by drawing on multiple studies using a range of designs and methods. As discussed in previous comprehensive reviews (Ja- cobi et al., 2004 Stice & Shaw, 2002 Striegel-Moore & Cachelin, 2001), the state of knowledge concerning the risk and causal factors of eating disorders is frustratingly in- complete. Notwithstanding the fact that they are central to advancing etiological models, most epidemiological stud- ies have not included eating disorders among the psychi- atric disorders to be assessed in the population (Wittchen & Jacobi, 2005). In the United States, the first epidemiolog- ical study of a nationally representative sample was con- ducted only in 2001���2003 (Hudson, Hiripi, Pope, & Kessler, 2007). Few studies have moved beyond the first (important, yet preliminary) step of demonstrating statisti- cal association between the factor and the outcome to the second step of establishing the factor���s precedence, and fewer studies yet have used an experimental manipulation as the ultimate test of causal hypotheses (Jacobi et al., 2004). For example, anorexia nervosa has been shown to be associated with abnormalities in the serotonergic system during the acute illness stage as well as after recovery (Kaye, Bailer, Frank, Wagner, & Henry, 2005). This re- search suggests that the serotonin system is a potentially fruitful area of further study, yet before they can be con- sidered to be of etiological significance, it needs to be established whether these abnormalities simply represent concurrent symptoms or consequences (biological scars) of the eating disorder. Similarly, measuring personality char- acteristics or family functioning in acutely ill or recovered patients does not permit disaggregation of risk factors from clinical correlates or consequences of experiencing the disorder. We caution that in decades past, researchers pro- posed the now-debunked hypotheses that risk for schizo- phrenia and autism arose from particular mothering styles (Caplan & Hall-McCorquodale, 1985) on the basis of data gathered with designs ill-suited for testing risk-factor hy- potheses. Retrospective assessment of putative risk factors (e.g., asking about an individual���s history of weight-related teasing) is fraught with methodological problems such as difficulty with accurately timing the exposure relative to the age of onset of the disorder and inaccurate or selective recall of critical events. We focus our review on the second generation of research, which comprises genetic studies, studies with prospective and experimental designs, and evidence concerning fixed markers. We do not review the voluminous literature of cross-sectional or retrospective studies of personality characteristics or parenting styles and family functioning, a literature that has been critically discussed in numerous reviews (e.g., Jacobi et al., 2004 Ward, Ramsay, & Treasure, 2000 Wonderlich, Lilenfeld, Riso, Engel, & Mitchell, 2005). Research Approaches to Uncovering Risk Factors for Eating Disorders The core features of eating disorders include disturbance in body image (e.g., overvaluation of thinness, weight or shape concerns), over- or undercontrol of eating (e.g., severe dietary restriction, binge eating), and extreme be- haviors to control weight or shape (e.g., compulsive exer- cise, purging). This suggests a research focus on biological structures and processes involved in appetite, satiety, and weight regulation and on cultural factors that shape atti- tudes and behaviors related to body image and eating. For decades, the primary focus of risk-factor research has been sociocultural and family influences on etiology of eating disorders, yet very few specific, replicated candidate envi- ronmental risk factors have emerged (Jacobi et al., 2004). In contrast, studies over the past decade have consistently detected genetic effects. Furthermore, our ability to mea- Cynthia M. Bulik 183 April 2007 ��� American Psychologist