Show Summary Details
Page of

Prevention of Eating Disorders 

Prevention of Eating Disorders
Prevention of Eating Disorders
Page of

PRINTED FROM OXFORD MEDICINE ONLINE ( © Oxford University Press, 2016. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Medicine Online for personal use (for details see Privacy Policy and Legal Notice).

date: 16 August 2018

For any disorder, understanding the risk factors, or variables that predict the development of the disorder, is vital to prevention efforts. In eating disorders, considerable progress has been made in identifying risk factors for the development of the syndrome of bulimia nervosa and for a number of the behavioral and psychological symptoms of eating disorders; however, much less is know about risk factors for anorexia nervosa. Through recent efforts we have begun to identify risk factors and to examine how they increase the probability that an eating disorder will develop. This information is being used to inform rational prevention efforts.


A risk factor is an agent or exposure that increases the probability of an adverse outcome, in this case, eating disorders. In order to be demonstrated conclusively as a risk factor, the agent in question should be assessed prospectively (prior to development of the eating disorder), show temporal precedence to the onset of the eating disorder, and show some degree of specificity with the eating disorder (i.e., not be merely a general risk factor for psychopathology).

One of the first factors mentioned in most discussions of risk factors for eating disorders is dieting (Schmidt, 2002). The term dieting is complex, laden with many meanings, and used to refer to a variety of attitudes and behaviors. The National Task Force on the Prevention and Treatment of Obesity (2000) defines dieting as “the intentional and sustained restriction of caloric intake for the purposes of reducing body weight or changing body shape, resulting in a significant negative energy balance.” This useful and relatively straightforward definition implies that dieting, because it results in negative energy balance, must be associated with weight loss. Therefore, attempts to restrict caloric intake that do not result in weight loss might properly be termed “unsuccessful dieting”; such attempts are frequently described by individuals with symptoms of eating disorders. The literature on eating disorders uses dieting to refer to both successful and unsuccessful attempts to restrict caloric intake, making it difficult to determine whether successful dieting and unsuccessful dieting play similar roles in the development of eating disorders.

The terms restrained eating and dietary restraint are theoretical constructs frequently employed in discussions of risk factors for the development and maintenance of eating disorders. Dietary restraint refers to a mental or cognitive set linked with the attempt to diet, and tends to be associated with unsuccessful dieting (Lowe, 1993). However, like the term dieting, dietary restraint and restrained eating are used to describe a range of attitudes and behaviors, including food avoidance.

Finally, the syndromes of anorexia nervosa and bulimia nervosa are characterized by what may be termed “unhealthy weight loss behaviors.” These include a wide range of activities associated with some risk of physical harm, such as self-induced vomiting, laxative and diet pill abuse, complete food avoidance for extended periods of time (fasting), and excessive exercise to lose weight. These types of behavior are relevant in discussions of risk factors for eating disorders, as a considerable number of young people engage in unhealthy weight loss behaviors, and it is possible that such individuals are at high risk for the development of eating disorders meeting full DSM-IV criteria.

In this chapter, we will attempt to use the terms dieting, unsuccessful dieting, dietary restraint, and unhealthy weight loss behaviors in the narrow sense just described. However, the range of ways in which these terms have been used in the literature and the failure of many studies to define them add to the difficulties of interpret-ing the literature on risk factors for eating disorders.


In order to explore the literature on risk factors for anorexia nervosa and bulimia nervosa, slightly different approaches are necessary. In part, this divergence stems from necessity. Anorexia nervosa is more rare than bulimia nervosa, and fewer prospective studies of anorexia ner vosa have been performed. Despite decades of study, no true risk factor for anorexia nervosa has been clearly demonstrated and confirmed. A greater number of prospective studies of bulimia nervosa exist. Nonetheless, the outcome is often varying degrees of bulimic pathology, rather than the occurrence of the syndrome of bulimia nervosa meeting full DSM-IV criteria. Unlike the study of risk factors for anorexia nervosa, there have been attempts to study bulimia nervosa risk factors experimentally, with some intriguing results. When prospective and experimental studies exist, they are highlighted, as these designs permit firmer inferences regarding the potential risk factors that promote psychiatric disturbances.

In the absence of prospective and experimental studies, the review of risk factors for anorexia nervosa necessitates a greater reliance on cross-sectional data. Although such data can be useful for generating hypotheses on potential risk factors for eating pathology, this type of design does not allow for differentiation of a precursor from a concomitant or consequence of a psychiatric condition. One exception is research on genetic factors, because one's genetic make-up is determined before birth, and therefore temporal precedence is ensured, although complicated by complex gene–environment interactions and gene–environment correlations. In addition, in this review, studies that used retrospective reports were de-emphasized or excluded, because temporal precedence could not be demonstrated and retrospective data are subject to a variety of biases.


Several factors have contributed to our relative lack of understanding of risk factors for anorexia nervosa. First, there is the pervasive sense that anorexia nervosa is a disorder “chosen” by women in pursuit of an unrealistic sociocultural body size ideal. Such beliefs can hamper attempts at identifying true risk factors. From a basic epidemiological perspective, identification of risk factors for anorexia nervosa has been difficult. The prevalence of the disorder in the population is relatively low, no more than 1% or 2% among females (see Chapter 13), and the prediction of a rare event is challenging, as evidenced by longitudinal studies in which no incident cases of anorexia nervosa emerged (McKnight Investigators, 2003; Stice, Presnell, & Bearman, 2001).

Unfortunately, few if any studies have satisfied the standards outlined above for identification of true risk factors. Although the available study designs have yielded suggestions about possible risk factors, confounds in their design have made it impossible to determine whether the traits examined are truly risk factors or phenomena associated with anorexia nervosa in some other way. Ultimately, novel designs will likely be required to identify factors that increase risk for anorexia nervosa.

In the absence of definitive studies, this discussion will include a range of potential risk factors that have been associated with anorexia nervosa via a variety of research designs. Three axes—temporal precedence, or, “Did the event occur prior to the onset of anorexia nervosa?”; time of reporting, or, “Was the temporal precedence determined prospectively or retrospectively?”; and the nature of the outcome event, or, whether the observed outcome is DSM-IV anorexia nervosa or a broadly defined anorexia nervosa syndrome or symptom—will be considered. Although longitudinal data suggest temporal stability of severe anorexia nervosa symptoms that emerge in childhood and persist throughout adolescence and adulthood (Kotler, Cohen, Davies, Pine, & Walsh, 2001), the relation between symptoms that do not meet criteria for DSM-IV anorexia nervosa and the subsequent development of DSM-IV anorexia nervosa is less certain. The correlation of anorexia nervosa symptoms ranges from 0.4 to 0.5 over time. Table 15.1 presents the putative risk factors and their ratings on each of the three axes.

Table 15.1 Risk Factors for Anorexia Nervosa






Persists after


Associated with

Full Syndrome of


Associated with














Societal emphasis on thinness




Perinatal events




Personality, temperament



Early feeding problems



Obesity (self)

Obesity (familial)





Body dissatisfaction














Family environment




Life events




Social support

High-risk activities


Family history, genetics





AN, anorexia nervosa; CSA, childhood sexual abuse; SES, socioeconomic status.

The mechanisms of action by which these putative risk factors may operate to increase risk of anorexia nervosa have not begun to be explored. Similarly, there are no data on the timing of exposure and whether there are critical windows for exposure that lead to differential risk. Nor have factors that may protect against the development of anorexia nervosa been examined in any detail.

General Risk Factors


Being female is perhaps the most reliable risk factor for anorexia nervosa. The female-to-male ratio for anorexia nervosa has been consistently estimated to be approximately 10:1 in both clinical and epidemiological samples (Garfinkel et al., 1995; Hoek, 1991; Jones, Fox, Babigan, & Hutton, 1980; Lucas, Beard, O'Fallon, & Kurland, 1988; Lucas, Beard, O'Fallon, & Kurland, 1991; Rand & Kuldau, 1992; Rastam, Gillberg, & Garton, 1989; Soundy, Lucas, Suman, & Melton, 1995; Wells, Bushnell, Hornblow, Joyce, & Oakley-Browne, 1989). The uneven gender distribution cannot yet be explained, and although theories ranging from sociocultural to biological (Garner & Garfinkel, 1980) factors exist, no definitive answer has been found as to why a differential risk for gender is so consistently observed across cultures.


The onset of anorexia nervosa typically occurs during the peripubertal or postpubertal period (Lucas et al., 1988, 1991). Prepubertal anorexia nervosa exists (Cooper, Watkins, Bryant-Waugh, & Lask, 2002), but is uncommon (Stein, Chalhoub, & Hodes, 1998). Disturbances in eating and weight-related behaviors are clearly present, however, in preadolescent girls (Graber, Brooks-Gunn, Paikoff, & Warren, 1994; Killen et al., 1994; Leon, Fulkerson, Perry, & Cudeck, 1993; Sands, Tricker, Sherman, Armatas, & Maschette, 1997). The extent to which these disturbances overlap with anorexia nervosa is not entirely known (Kotler et al., 2001). Onset of anorexia nervosa can occur throughout the life span (Beck, Casper, & Andersen, 1996; Inagaki et al., 2002), but the highest risk period for onset is around puberty (Lucas et al., 1988, 1991). As is true with gender, there is no definitive answer as to why puberty increases risk, and theories range from sociocultural (during puberty girls become more vulnerable to social pressures to be thin, especially in the context of their changing bodies; Gowers & Shore, 2001) to biological (hormonal changes during puberty trigger other relevant biological processes; see Muñoz & Argente, 2002, for a review). In contrast to binge eating disorder (Reichborn-Kjennerud, Bulik, Tambs, Harris, & Sullivan, submitted) early menarche has not been associated with risk for anorexia nervosa (Fairburn, Cooper, Doll, & Welch, 1999; Stice et al., 2001).

Societal Emphasis on Thinness

Weight concerns and dieting are normative in developed countries (Rodin, Silberstein, & Striegel-Moore, 1985; Striegel-Moore, Silberstein, & Rodin, 1986), a cultural phenomenon thought to be a necessary but not sufficient precondition for the development of eating disorders (Brownell, 1991; Garner & Garfinkel, 1980). The incidence of anorexia nervosa on Curaçao suggests, however, that anorexia nervosa is not confined to areas embracing the Western ideal of slimness (Hoek, van Harten, van Hoeken, & Susser, 1998). The precise role of sociocultural factors in increasing risk for anorexia nervosa is unclear, as exposure to thin ideals and dieting are nearly universal in industrialized countries, but only a very small number of young women actually develop clinically significant eating disorders. Individuals may be susceptible to the cultural pressures toward dieting and body dissatisfaction in proportion to their degree of genetic predisposition (Bulik, 2003). Thus, although all individuals are exposed to these forces, those at greater genetic risk may be more adversely affected. Immigration and the rapid introduction of Western body ideals have been shown to affect the incidence of eating disorder–related behaviors (Becker, Burwell, Gilman, Herzog, & Hamburg, 2002; Bulik, 1987; Fichter, Weyerer, Sourdi, & Sourdi, 1983; Katzman, Nasser, & Gordon, 2001). It is unclear whether exposure to Western ideals or major nonspecific life stress (i.e., immigration) is associated with the emergence of eating pathology.

Race and Socioeconomic Status

Traditionally, anorexia nervosa was considered to be a disorder confined to the white upper-middle class. This perception may result from the failure of many community-based studies to include non-white participants or from sampling highly homogeneous populations (Bushnell, Wells, Hornblow, Oakley-Browne, & Joyce, 1990; Götestam & Agras, 1995; Hoek et al., 1995; Kendler et al., 1991). Anorexia nervosa occurs across races and cultures (Hoek et al., 1998; Katzman et al., 2001). In terms of socioeconomic status, in a large population-based study of female twins, Walters and Kendler (1995) found that a greater number of years of parental education (a proxy variable for SES) was associated with anorexia nervosa. In addition, Striegel-Moore, Dohm, et al. (2003) found that 15 white (1.5%) and no black women in a geographically and economically diverse community sample of young women who had previously participated in the 10-year National Heart, Lung, and Blood Institute (NHLBI) Growth and Health Study met lifetime criteria for anorexia nervosa.

Perinatal Events

Cnattingius, Hultman, Dahl, and Sparen (1999), using the link between the Swedish birth registry and the Swedish psychiatric inpatient registry, found that girls born prematurely (especially if they were small for gestational age) and those born with cephalhematoma (a collection of blood under the scalp of a newborn) were at increased risk for developing anorexia nervosa. This study employed a design capable of identifying true prospective risk factors. The authors hypothesized that subtle brain damage at birth could result in early feeding difficulties and increase risk for anorexia nervosa. It is also possible that prematurity and small size for gestational age of infants were secondary effects of lingering eating disorders in the mother. Shoebridge and Gowers (2000) also noted higher rates of perinatal or infant loss in mothers of a clinical sample of individuals with anorexia nervosa, and greater maternal anxiety and concern during pregnancy. Emerging data in animals suggest that even moderate maternal undernutrition around the time of conception can produce a precocious fetal cortisol surge and preterm birth (Bloomfield et al., 2003), a phenomenon which could occur in women with anorexia nervosa symptoms around the time of conception. Additional data indicate that individuals with current or past anorexia nervosa have a higher risk of birth complications, cesarean deliveries, and postpartum depression (Bulik, Sullivan, Fear, Pickering, & Dawn, 1999; Franko & Walton, 1993; Franko et al., 2001). Perinatal adverse events may therefore increase the risk of developing anorexia nervosa, and a past history of anorexia nervosa may increase the risk of experiencing adverse perinatal events, thus perpetuating a cycle of risk.

Individual Risk Factors

Personality Characteristics

Psychometric studies have consistently linked anorexia nervosa to a cluster of personality and temperamental traits—specifically, negative self-evaluation, low self-esteem, extreme compliance, obsessionality, perfectionism, neuroticism, and harm avoidance (Anderluh, Tchanturia, Rabe-Hesketh, & Treasure, 2003; Brewerton, Dorn, & Bishop, 1992; Bulik, Sullivan, Weltzin, & Kaye, 1995; Bulik, Tozzi, et al., 2003; Fairburn, Cooper, et al., 1999; Gual et al., 2002; Karwautz et al., 2001; Kleifield, Sunday, Hurt, & Halmi, 1994; Srinivasagam et al., 1995; Strober, 1990; Tyrka, Waldron, Graber, & Brooks-Gunn, 2002; Vitousek & Manke, 1994; Waller et al., 1993; Walters & Kendler, 1995). These traits continue to characterize individuals with anorexia nervosa even after recovery (Bulik, Sullivan, Fear, & Pickering, 2000; Casper, 1990; Kaye, Weltzin, & Hsu, 1993; Srinivasagam et al., 1995). In the absence of premorbid personality data on large samples of individuals who later develop anorexia nervosa, some studies have used a “recovery” design, in which traits that persist after recovery are assumed to represent enduring traits that preceded the onset of the disorder and are thought to represent vulnerability factors. The potential weakness of this design is that these traits may not have existed premorbidly and instead could represent personality or trait “scars” from having had anorexia nervosa.

Specific Eating and Weight-Related Risk Factors

Obesity and Body Mass Index

Parental obesity is less frequently associated with risk for anorexia nervosa than for bulimia nervosa, and familial obesity appears to be more common in family members of women with bulimia nervosa than women with anorexia nervosa (Garfinkel, Moldofsky, & Garner, 1980; Grace, Jacobson, & Fullager, 1985; Vieselman & Roig, 1985). Fairburn, Welch, Doll, Davies, and O'Connor (1997) also found a differential obesity risk between individuals with anorexia nervosa and bulimia nervosa, with anorexia nervosa women having lower familial risk of obesity. Only one study explored the opposite phenomenon, whether parents of individuals with anorexia nervosa tend to be thinner than parents of healthy controls (Halmi, Struss, & Goldberg, 1978), and there was no evidence to suggest that either mothers or fathers weighed less than parents of controls.


The prevalence of dieting has been estimated to be between 14% and 77% and is highest in young women (French, Perry, Leon, & Fulkerson, 1994). In most of these studies dieting was simply defined as whether individuals had ever restricted their intake to lose weight, and therefore may have captured both successful and unsuccessful dieters. In a study of 36,320 public school students, dieting frequency was strongly related to poor body image, fears of being unable to control eating, and more prevalent history of binge eating. Dieting was also related in a “dose–re sponse” fashion to a range of psychosocial and health behavior variables (French, Story, Downes, Resnick, & Blum, 1995). Walters and Kendler (1995) found dieting status to be associated with anorexia nervosa in a population-based sample of twins; however, dieting was not measured prior to the onset of anorexia nervosa. In contrast, in a series of studies of anorexia nervosa, bulimia nervosa, and binge eating disorder in which subjects were asked to report on the premorbid presence of potential risk factors, Fairburn, Cooper, et al. (1999) found that although dieting dimensions were relevant to the emergence of both bulimia nervosa and binge eating disorder, they were not associated with anorexia nervosa. Dieting was not elevated in a study of affected sisters of discordant sister pairs (Karwautz et al., 2001). Thus, the data from clinical and epidemiological studies do not resolve whether dieting should be considered a risk factor for anorexia nervosa.

Body Dissatisfaction and Slim Body Ideal

Dissatisfaction with the size or shape of one's body is often thought to be the psychological motivator for dieting behavior (Stice, 1994) and a key contributor to the gender differential in prevalence of eating disorders. In puberty, body satisfaction begins to decrease in young girls, and this dissatisfaction may be secondary to the increase in body fat percentage associated with female pubertal development (Marino & King, 1980). Dissatisfaction with body shape or size is thought to be the driving force for the onset of dieting behavior (Cash & Henry, 1995; Hawkins & Clement, 1984; Rodin et al., 1985; Tiggemann, 1994; Van Strien, 1989).


As previously discussed in Chapter 13, other significant psychiatric disorders, especially major depression and anxiety disorders, commonly co-occur with anorexia nervosa (Braun, Sunday, & Halmi, 1994; Bulik, 2001; Bulik, Sullivan, Fear, & Joyce, 1997; Deep, Nagy, Weltzin, Rao, & Kaye, 1995; Halmi et al., 1991). Well over half of women with anorexia nervosa report a lifetime presence of an anxiety disorder—most commonly, overanxious disorder, obsessive-compulsive disorder, or social phobia. In many cases, onset of the anxiety disorder precedes the onset of anorexia nervosa (Bulik et al., 1997; Deep et al., 1995). Retrospective accounts of premorbid symptoms among children who later develop anorexia nervosa often emphasize the presence of pervasive anxiety (Bruch, 1973; Lask & Bryant-Waugh, 2000). This pattern of onset may reflect the natural course of the two disorders (i.e., the average age of onset of many anxiety disorders is younger than the average age of onset of anorexia nervosa), but it may also indicate that childhood anxiety is a significant risk factor for the development of anorexia nervosa. In a population-based sample of over 2,000 female twins, odds ratios for generalized anxiety disorder, phobias, and panic disorder were significantly elevated among women with varying definitions of anorexia nervosa (Walters & Kendler, 1995).

In addition, several outcome studies suggest that depression and anxiety commonly persist after recovery from anorexia nervosa (Löwe et al., 2001; Sullivan, Bulik, Fear, & Pickering, 1998). Significantly elevated relative risks for mood disorders have been reported among relatives of probands with anorexia nervosa (Gershon et al., 1984; Hudson, Pope, Jonas, & Yurgelun-Todd, 1983; Logue, Crowe, & Bean, 1989; Rivinus et al., 1984), although these rates may be highest among relatives of individuals with anorexia nervosa who are themselves depressed (Strober, Lampert, Morrell, Burroughs, & Jacobs, 1990). Twin studies have shown that the genetic risk factors for anorexia nervosa and depression are correlated (Wade, Bulik, Neale, & Kendler, 2000), and there appears to be a unique genetic factor that influences the emergence of both early eating and early anxiety disorder symptoms (Silberg & Bulik, submitted).

Other “Environmental” Risk Factors


Early family theories of the etiology of anorexia nervosa posited dysfunctional family patterns marked by enmeshment and rigidity (Minuchin, Rosman, & Baker, 1978). Studies have generally found that women with eating disorders display anxious attachment styles and separation distress (Armstrong & Roth, 1989), greater fears of abandonment, and lack of autonomy in relationships (Becker, Bell, & Billington, 1987; Kenny & Hart, 1992); the more insecure the attachment with parents, the greater the disordered eating pathology (Heesacker & Neimeyer, 1990). Shoebridge and Gowers (2000) noted higher rates of near exclusive maternal childcare, severe distress at first separation, high maternal trait anxiety, and later age for sleeping away from home in patients with anorexia nervosa in comparison to controls.

Family Environment

Research does not support the existence of a “typical” anorexia nervosa family; indeed, studies that have compared family environment across eating disorders tend to identify more family dysfunction in families with bulimia nervosa offspring than those with anorexia nervosa offspring. Schmidt, Tiller, and Treasure (1993) found that women with bulimia nervosa had experienced more family arrangements, parental indifference, excessive parental control, physical abuse, and violence among family members than members of other eating disorders subgroups. Murphy, Troop, and Treasure (2000) reported that sisters with anorexia nervosa had higher levels of maternal control and more antagonism toward and jealousy of their sisters than did their unaffected sisters. In a population-based sample of female twins, Walters and Kendler (1995) reported higher maternal overprotectiveness in individuals with anorexia nervosa.

Sexual Abuse

Childhood sexual abuse (CSA) has been reported in women with anorexia nervosa (Herzog, Staley, Carmody, Robbins, & van der Kolk, 1993; Horesh et al., 1995), and there is some evidence that CSA is more prevalent in women with anorexia nervosa who exhibit purging behavior (Waller, Halek, & Crisp, 1993). Given the relative rarity of anorexia nervosa, few population-based estimates of the prevalence of CSA in these women exist. Romans, Martin, and Mullen (1994) studied 3,000 women in New Zealand and found that there was a higher frequency of anorexia nervosa and bulimia nervosa in women who reported CSA. However, CSA did not appear to act independently, as poor parenting and growing up away from both parents also contributed independently to risk of eating disorders. Kinzl, Traweger, Guenther, and Biebl (1994) highlighted the importance of the familial environment in which CSA occurs and the difficulty in teasing out the long-term effects of CSA without considering the family context that mediates the experience of abuse (Alexander, 1992). The current data suggest overall that CSA and abuse of other types, although more common in women with anorexia nervosa, may be more general risk factors for psychopathology.

Life Events

Horesch et al. (1995) found that adolescents with anorexia nervosa reported more adverse life events than healthy controls, and more adverse life events relating to family than psychiatric controls. Fairburn, Cooper, et al. (1999) found that a history of events including abuse and death of close relatives increased the risk of anorexia nervosa. Schmidt, Tiller, Blanchard, Andrews, and Treasure (1997) found that individuals with anorexia nervosa experienced significantly more adverse events and difficulties with the potential to evoke sexual shame or disgust, in the year prior to onset of anorexia ner-vosa. An increase in such events was also found by Karwautz et al. (2001) in a study of discordant sisters: sisters with anorexia nervosa reported more teasing about breast development. Whether individuals with eating disorders actually experience more adverse life events, whether they remember adverse life events better than patients without such events, or whether they are more susceptible to the impact of life events remains unknown.

Social Support and Interpersonal Relationships

Women with eating disorders report less successful social adjustment and problematic inter personal relationships (Norman & Herzog, 1984; Pike & Rodin, 1991; Striegel-Moore, Silberstein, & Rodin, 1993). Problems with social adjustment often persist after recovery (Casper, 1990; Fava et al., 1990; Yager, Landsverk, & Edelstein, 1987; Yager, Landsverk, Edelstein, & Jarvik, 1988).

Participation in Sports and Activities, Professions with Focus on Body Shape and Weight

Activities that place substantial emphasis on weight and appearance (e.g., ballet, gymnastics) have been investigated as independent risk factors for the development of anorexia nervosa. Ballet participation has received substantial attention (Klump, Ringham, Marcus, & Kaye, 2001) because of the high levels of required exercise and pressures for thinness and athletic excellence (Garner & Garfinkel, 1980; Vaisman, Voet, Akivis, & Sive-Ner, 1996; Weeda-Mannak & Drop, 1985). An increased prevalence of both diagnosable eating disorders as well as disordered eating symptoms has been observed in ballet dancers, with a prevalence of DSM-IV anorexia nervosa that is 4 to 25 times higher in ballet dancers than that in the general population (4%–25%) (Garner & Garfinkel, 1980; Szmukler, Eisler, Gillies, & Hayward, 1985; Vaisman et al., 1996).

Unlike data from other athlete groups (Powers, Schocken, & Boyd, 1998), ballet dancers' scores on measures of eating pathology are similar to those of individuals with diagnosable eating disorders. In addition, disturbed eating attitudes and behaviors appear to persist after retirement in ballet dancers (Khan et al., 1996), which has not been observed in other athlete groups, including gymnasts (O'Connor, Lewis, Kirchner, & Cook, 1996). Moreover, athletic, artistic, and professional environments may attract individuals who are preoccupied with facets central to eating disorder pathology. This pairing can then contribute to the emergence of the syndromes.

Biological Risk Factors

Several lines of evidence suggest that women with anorexia nervosa have a disturbance of serotonin (5-hydroxytryptamine [5-HT]) neuronal transmission. Serotonin function in the brain is involved in the regulation of both mood and behavior, including eating behavior. Increased 5-HT activity may contribute to core emotional and attitudinal factors, such as anxiety, perfectionism, harm avoidance, and body image distortions, which, when coupled with psychosocial influences, may increase an individual's risk for developing anorexia nervosa (Kaye et al., submitted). Disturbances of 5-HT activity and anxious, obsessive symptoms persist after recovery from anorexia nervosa, and recovered women have increased levels of the major serotonin metabolite 5-hydroxyindoleacetic acid (5-HIAA) in cerebrospinal fluid (CSF) (Kaye, Gwirtsman, George, & Ebert, 1991). Moreover, the antidepressant medication fluoxetine, which affects the serotonergic system, may reduce the rate of relapse among anorexia nervosa patients (Kaye et al., 2001).

Genetic Risk Factors

A series of family studies have demonstrated a significantly greater lifetime prevalence of eating disorders among relatives of eating disordered individuals than among relatives of controls (Gershon et al., 1983; Hudson, Pope, Jonas, Yurgelun-Todd, & Frankenburg, 1987; Kassett et al., 1989; Lilenfeld et al., 1998; Strober et al., 1990; Strober, Freeman, Lampert, Diamond, & Kaye, 2000). Relatives of individuals with anorexia nervosa and bulimia nervosa also have significantly increased rates of eating disorders that do not meet full diagnostic criteria, compared to relatives of controls (Lilenfeld et al., 1998; Strober et al., 2000), which suggests a broad spectrum of eating-related pathology in families. Thus, anorexia nervosa appears to be familial (see Lilenfeld, Kaye, & Strober, 1997, for a review). In addition, twin studies (Klump, Miller, Keel, McGue, & Iacono, 2001; Kortegaard, Hoerder, Joergensen, Gillberg, & Kyvik, 2001; Wade et al., 2000) suggest that the observed familial links may be accounted for by genetic effects.

Linkage studies have begun to identify areas of the human genome that may harbor suscep tibility genes for anorexia nervosa, with the most promising areas being identified on chromosome 1 (Devlin et al., 2002; Grice et al., 2002). Several groups have investigated the relation between a variety of candidate genes and eating disorders (for a review, see Tozzi, Bergen, & Bulik, 2002). These studies have chosen candidate genes based on systems and functions associated with food intake, body weight, regulation of feeding behavior, motor activity, energy expenditure, metabolic adaptation to fasting, and other related characteristics and symptoms of eating disorders. Although a number of positive findings have been documented, no single gene or set of genes has consistently emerged as being strongly associated with either anorexia nervosa or bulimia nervosa. As anorexia nervosa is a complex disorder whose etiology is influenced by multiple genes and multiple environmental factors, ultimate understanding of risk will have to include the elucidation of Gene × Environment, Gene × Gene, and Environment × Environment interactions.


Numerous studies have been conducted to identify risk factors that predict the onset of anorexia nervosa, but enhanced efforts are needed to identify true risk and protective factors for anorexia nervosa. Studies of recovered individuals, while valuable, are not sufficient to discriminate between risk factors for anorexia nervosa and residual effects of anorexia nervosa. The identification of protective factors is equally critical, as the identification of factors that are associated with both the emergence of anorexia nervosa and the failure to develop the disorder has implications for both detection and prevention of the disorder.


As mentioned previously, this review of risk factors for bulimia nervosa will emphasize prospective and experimental studies. Such studies, properly conducted, are powerful in being able to establish with some clarity whether the introduction of a putative risk factor leads to an increase in symptoms. Cross-sectional data, although essential for hypothesis generation in studies of anorexia nervosa, are less valuable, as it is difficult or impossible to establish whether a variable is causally linked to the development of the disorder or only associated with it in some way. It should be noted, however, that the prospective and experimental studies focus on risk factors for the development of broadly defined bulimic symptoms rather than the full syndrome of bulimia nervosa. Crucial unanswered questions are the nature of the relation between bulimic symptoms and bulimia nervosa, and the degree of overlap between risk factors for bulimic symptoms and risk factors for bulimia nervosa.

Societal Emphasis on Thinness

As with anorexia nervosa, the cultural emphasis on thinness has been implicated as a risk factor for the development of bulimia nervosa. Pressure to be thin has been shown to predict increases in body dissatisfaction, dieting, and negative affect (Cattarin & Thompson, 1994; Field et al., 2001; Stice, 2001; Stice & Bearman, 2001; Stice & Whitenton, 2002; Wertheim, Koerner, & Paxton, 2001) and to contribute to bulimic symptom onset (Field, Camargo, Taylor, Berkey, & Colditz, 1999; Stice & Agras, 1998; Stice, Presnell, & Spangler, 2002). Experiments have found that acute and long-term exposure to ultraslender media images and peers produces increased body dissatisfaction, negative affect, dieting, and bulimic symptoms, but that these effects are stronger for girls with preexisting body image disturbances and social support deficits (Cattarin, Thompson, Thomas, & Williams, 2000; Groesz, Levine, & Murnen, 2002; Irving, 1990). Thus, perceived pressure to be thin appears to be a risk factor for body dissatisfaction, dieting, negative affect, and bulimic symptoms, and these effects may be amplified for at-risk individuals.

Early Feeding Problems

Several studies have examined the extent to which early problems with feeding constitute risk factors for the later development of broadly defined eating disorder symptoms. In data obtained from a long-term prospective study, Marchi and Cohen (1990) found that symptoms of bulimia nervosa in later adolescence were related to both digestive problems and pica in early childhood and to efforts at weight loss in early adolescence. Stice, Agras, and Hammer (1999) found that a higher infant body mass index (BMI) predicted the emergence of overeating and that longer duration of infant sucking predicted overeating-induced vomiting in children; however, the relationship of these phenomena to the development of bulimia nervosa is unknown.

Early Menarche

Early menarche is thought to produce increased adipose tissue, which may result in a perceived deviation from the thin ideal and decreased body dissatisfaction and could, in turn, lead to dieting, negative affect, and bulimic symptoms. Early sexual development can also lead to increased sexual attention and unwanted comments and teasing. However, early menarche did not predict future increases in body dissatisfaction, dieting, negative affect, or bulimic symptoms (Cooley & Toray, 2001a; Hayward et al., 1997; Stice & Whitenton, 2002). Therefore, the currently available research does not support early pubertal development as a risk factor for these symptoms.

Perfectionism and Impulsivity

Perfectionism has been evaluated as a risk factor for eating pathology because this personality trait may promote a relentless pursuit of the thin ideal. Three studies have found that perfectionism did not predict future bulimic symptomatology (Killen et al., 1996; Vohs, Bardone, Joiner, Abramson, & Heatherton, 1999; Vohs et al., 2001), although neither study used DSM-IV criteria for bulimia nervosa as an outcome. Another personality characteristic, a deficit in impulse control, has also been considered a risk factor for bulimic pathology, as deficits in impulse control might increase the propensity for episodes of uncontrollable binge eating. Data from three prospective studies indicate that impulsivity does not predict subsequent increases in bulimic symptoms (Wonderlich, Connolly, & Stice, in press). Thus, while there is currently no empirical evidence for a role of perfectionism or impulsivity as risk factors for bulimic pathology, additional studies are required to determine whether these traits influence risk for threshold bulimia nervosa.

Negative Affect

Negative affect has been shown to predict increases in bulimic symptoms and subthreshold/threshold bulimic pathology among adolescents (Cooley & Toray, 2001a; Killen et al., 1996; Stice, 2001; Stice & Agras, 1998), and experimentally induced negative affect produced increases in body dissatisfaction in one study (Taylor & Cooper, 1992). One prevention trial found that an intervention that reduced negative affect produced decreases in bulimic pathology (Burton, Stice, Bearman, & Rohde, submitted). These results are consistent in suggesting that negative affect is a risk factor for bulimic pathology.

Obesity and Body Mass Index

Elevated body mass, or being at a higher weight, has been shown to predict increases in perceived pressure to be thin, in body dissatisfaction, and in attempted dieting (Cattarin & Thompson, 1994; Field et al., 2001; Patton, Johnson-Sabine, Wood, Mann, & Wakeling, 1990; Stice & Whitenton, 2002; Vogeltanz-Holm et al., 2000), but not increases in depression (Lewinsohn et al., 1994; Stice & Bearman, 2001; Stice, Hayward Cameron, Killen, & Taylor, 2000). Therefore, individuals at a higher weight may experience pressure to be thin from family or peers, which could result in increased drive for thinness or body dissatisfaction, and increased risk for bulimic symptoms. Increased BMI predicted the onset of bulimic symptoms in several studies (Killen et al., 1994; Stice, Presnell, & Spangler, 2002; Vogeltanz-Holm et al., 2000), but other studies found no association (Cattarin & Thompson, 1994; Cooley & Toray, 2001b; Killen et al., 1996; Stice & Agras, 1998). Thus, while increased weight is a risk factor for pressure to be thin, body dissatisfaction, and attempted dieting, it may play a more important role in promoting other risk factors for bulimia nervosa than in directly fostering bulimic symptoms.


Self-report dietary restraint measures predicted increases in negative affect (Stice & Bearman, 2001; Stice et al., 2000) and onset of bulimic pathology (Field, Camargo, Taylor, Berkey, & Colditz, 1999; Killen et al., 1994, 1996; Stice, 2001; Stice & Agras, 1998). However, experiments manipulating dietary restriction have not supported dietary restriction as a risk factor for bulimic symptoms. Random assignment to a low-calorie diet (Goodrick, Poston, Kimball, Reeves & Foreyt, 1998; Presnell & Stice, 2003; Reeves et al., 2001) or a weight-maintenance diet (Klem, Wing, Simkin-Silverman, & Kuller, 1997; Stice, Presnell, Groesz, & Shaw, submitted), relative to a waiting-list condition, resulted in reduced negative affect and bulimic symptoms among obese and nonobese samples. Although these findings appear contradictory, it has been reported that self-report dietary restraint scales do not measure actual caloric restriction (Stice, Fisher, & Lowe, 2004), which indicates that these scales may not be valid measures of dietary restriction. Thus, although prospective studies suggest that dieting is a risk factor for negative affect and bulimic pathology, experimental studies suggest that prescribed dietary restriction is not a causal risk factor for bulimic pathology. It is important to note that caloric restriction as implemented in experimental situations may not mimic the types of dysfunctional dieting behavior that occur among individuals prone to eating disorders.

Body Dissatisfaction and Slim Body Ideal

Body dissatisfaction predicts increases in attempted dieting (Cooley & Toray, 2001a; Patton et al., 1990; Stice, 2001; Wertheim, Koerner, & Paxton 2001), negative affect (Rierdan, Koff, & Stubbs, 1989; Stice & Bearman, 2001; Stice et al., 2000), bulimic symptom onset (Field et al., 1999; Killen et al., 1994, 1996; Stice & Agras, 1998), and bulimic symptoms (Cooley & Toray, 2001a; Stice, 2001). An intervention that reduced body dissatisfaction produced decreases in negative affect and bulimic symptoms (Bearman, Stice, & Chase, 2003).

The societal ideal for thinness is very difficult to attain, and may promote dieting in the absence of body dissatisfaction. Body dissatisfaction and attempted dieting, in turn, are thought to increase the risk for negative affect and bulimic pathology. Thin-ideal internalization predicts increases in body dissatisfaction, attempted dieting, negative affect (Stice, 2001; Stice & Whitenton, 2002) and is associated with bulimic symptom onset (Field et al., 1999; Stice & Agras, 1998; Stice et al., 2002). The effects of thin-ideal internalization appear to be stronger for heavier girls (Stice et al., 2002; Stice & Whitenton, 2002). Interventions that reduce thin-ideal internalization result in decreased body dissatisfaction, dieting, negative affect, and bulimic symptoms (Stice & Shaw, 2004). Thus, the thin-ideal internalization may be a risk factor for body dissatisfaction, dieting, negative affect, and bulimic symptoms, and these effects are potentiated by elevated body weight.

Genetic Risk Factors

Twin studies have yielded heritability estimates ranging from 31% to 83% for bulimia nervosa (e.g., Bulik, Sullivan, & Kendler, 1998; Kendler et al., 1991; Wade et al., 1999). Significant linkage has been reported on chromosome 10p, which suggests that this is an area of the genome worthy of further investigation for containing genes that may influence risk for bulimia nervosa (Bulik, Devlin, et al., 2003). However, researchers have been unable to identify specific genes that may influence the risk of developing bulimia nervosa. It may be that bulimia nervosa is a complex syndrome with both genetic and environmental underpinnings.

Although it is likely that risk factors for bulimia are interactive and multiplicative, the prospective evidence examining this issue is mini mal. In a recent review, Stice (2002) concluded that there was some evidence that negative affect, perfectionism, and early menarche are potentiating factors that amplify the effects of other risk factors. He also noted that although early menarche did not emerge as a significant risk factor for eating pathology in univariate models, it did appear to interact with life stressors to predict emergence of negative affect and eating disturbances (Stice, 2002). These promising leads should be followed in future studies.

How Are Risk Factors Acquired?

As noted above, prospective studies have suggested that the perceived pressure to be thin, internalization of a thin ideal, body dissatisfaction, and negative affect are risk factors for the development of bulimic symptoms, and other studies have begun to examine the development of such factors (Stice, 2002). Their acquisition is undoubtedly complex, involving the adolescent's innate susceptibility, developmental stage, and exposure to sociocultural factors such as teasing from family members and peers, social modeling (e.g., Furman & Thompson, 2002; Jackson, Grilo, & Masheb, 2000; Stice, Maxfield, & Wells, 2003; van den Berg, Wertheim, Thompson, & Paxton, 2002), and media pressure (e.g., Field, Camargo, Taylor, Berkey, & Colditz, 1999; Field et al., 2001; Taylor et al., 1998). Although few longitudinal studies have demonstrated the developmental sequence of exposure to risk factors, interventions designed to reduce the impact of risk factors may need to address a range of sociocultural factors.


A number of factors, including perceived pressure to be thin, thin-ideal internalization, body dissatisfaction, and negative affect, have been identified as risk factors for the development of bulimic symptoms. Some studies suggest that attempted dieting is a risk factor for the development of bulimic symptoms, whereas other data indicate that actual experimentally prescribed dietary restriction is not a risk factor. It is not clear whether dysfunctional dieting, especially severe caloric restriction, is a risk factor for full-syndrome bulimia nervosa. Low self-esteem, perfectionism, early menarche, and impulsivity do not appear to be risk factors for bulimic symptoms. The conclusions from the studies of risk factors for bulimia nervosa indicate that prevention programs that target thin-ideal internalization, pressure to be thin, modeling of eating disturbances, body dissatisfaction, and negative affect may prove useful in preventing the development of bulimic symptoms. Although currently unproven, it seems logical to assume that a reduction in bulimic symptoms in a population would lead to a reduction in the incidence of full-syndrome bulimia nervosa as well.


As discussed previously, although the empirical evidence is mixed, dieting is frequently implicated in the pathogenesis of eating disorders (Schmidt, 2002). The presumed association between dieting and the development of symptoms of eating disorders has led school-based eating disorder prevention programs to warn students about the ill effects of dieting (e.g., Kater, Rohwer, & Levine, 2000). With obesity rapidly becoming a major public health problem for America's youth, it is important to understand whether treatments for obesity, specifically recommendations to restrict caloric intake, increase the risk for the development of eating disorders.

Recent data indicate that 15.5% of adolescents are overweight, a 3-fold increase since 1980 (National Center for Health Statistics, 2003; Troiano, Flegal, Kuczmarski, Campbell, & Johnson, 1995), and an additional 22% of adolescents are at risk of being overweight, compared to 15.7% in 1980 (Troiano et al., 1995). Approximately 80% of overweight teenagers will become obese adults and consequently will experience increased risks of cardiovascular disease, hyperlipidemia, hypertension, diabetes mellitus, gallbladder disease, several cancers, and psychosocial complications (Casey, Dwyer, Coleman, & Valadian, 1992; Garn, Sullivan, & Hawthorne, 1989). Adults suffer adverse health effects as a result of teenage obesity (DiPietro, Mossberg, & Stunkard, 1994; Must, Jacques, Dallal, Bajema, & Dietz, 1992), but obese teens may not be spared from health complications until they reach adulthood. Twenty years ago, Type 2 diabetes was rare in children and adolescents. However, recent reports demonstrated that one third of adolescents diagnosed with diabetes had the Type 2, or adult-onset, form of this disease, which represents a 10-fold increase from rates in 1982 (Pinhas-Hamiel et al., 1996; Glaser 1997; Phillips & Young, 2000). Investigators fear that diabetes progresses more quickly in youth than in adults (Styne, 2001).

A combination of decreased caloric intake (i.e., dieting) and increased physical activity is the cornerstone of weight management in overweight adolescents, as in obese adults. Some clinicians and researchers fear, however, that dieting may increase the risk of eating disorders, particularly in adolescent females, and that weight loss interventions may do more harm than good (Garner & Wooley, 1991; Hirschmann & Munter, 1988; Polivy & Herman, 1985). By contrast, obesity experts generally believe that early intervention is desirable. Family support for change is likely to be available, eating and activity habits may be more amenable to modification, and adipose tissue cell proliferation may be curtailed (Goldfield & Epstein, 2002). Early treatment may also be cost-effective. Preventing overweight children and adolescents from becoming obese adults could reduce the health-care costs of treating obesity-related complications.

The question of whether dieting increases the risk of eating disorders in overweight adolescents and older children who seek weight loss is particularly important given the increasing rates of obesity. However, the vast majority of adolescent dieters do not develop eating disorders, as 44% of teenage girls report trying to lose weight, but the prevalence of eating disorders is between 1% (anorexia nervosa) and 3% (bulimia nervosa) of women, and prevalence rates among men are approximately one tenth of those observed in women (Hoek, 2002). As discussed above, other factors, including a genetic predisposition and negative affect, appear to contribute to the development of eating disorders in the presence of attempts to diet (Schmidt, 2002).

Other issues must also be examined when considering whether dieting is a risk factor for the development of eating disorders in overweight adolescents. First, studies of average weight or lean youth may have limited relevance to overweight adolescents. Although average-weight individuals experience adverse behavioral and psychological effects from severe caloric restriction (Keys, Brozek, Henschel, Mickelson, & Taylor, 1950), obese adults who have lost 10% of their initial weight have shown improvements in mood and premorbid binge eating frequency (National Task Force on the Prevention and Treatment of Obesity, 2000). A second consideration already discussed is that dieting can take many forms, from unhealthy weight control practices such as fasting or starvation, to moderate energy restriction, to a preoccupation with purportedly “good” and “bad” foods. Some interventions would appear more likely than others to be associated with adverse effects, therefore, it is important to understand the types of treatments used with overweight or obese children and adolescents.

Weight Loss Interventions for Children and Adolescents

Effective management of overweight in children and adolescents consists of diet, physical activity, and behavior change, and often requires parental participation (Goldfield, Raynor, & Epstein, 2002). Dietary change may include reduction of calorie or fat intake, or improved adherence to dietary guidelines, such as the Food Guide Pyramid (Epstein, Myers, Raynor, & Saelens, 1998). A popular approach to diet modification in youth is provided by the Stoplight Diet for Children, which classifies foods into red (stop), yellow (caution), or green (go) categories based on caloric value and nutrient density (Epstein & Squires, 1988). Typically, the initial goal is to limit intake to 1,000–1,300 calories per day, adjusted to promote a weight loss of 0.25 kg/week.

To increase physical activity, programs typically encourage structured aerobic exercise, such as swimming, jogging, or basketball, and lifestyle activity, which increases physical activity throughout the day (e.g., using the stairs rather than escalators). Preliminary studies demonstrated that lifestyle activity was more effective than structured exercise in facilitating the maintenance of weight loss (Goldfield et al., 2002). Reducing sedentary behaviors, including watching television and playing video games, has also been shown to contribute significantly to weight management (Epstein et al., 1995).

Parental participation in treatment is critical for children and also benefits adolescents (Goldfield et al., 2002). Parents may reward changes in their child's diet or physical activity, or modify their own eating or activity habits to model healthy behaviors. Similarly, parents can limit high-fat and high-sugar foods available in the home, while increasing consumption of fruits, vegetables, and other healthy choices. One study found that using parents as the exclusive agents for their child's behavior change resulted in greater decreases in overweight than treating the child alone (Golan, Weizman, Apterm, & Fainaru, 1998).

Family-based behavioral programs reduce children's percentage of overweight by as much as 25% and produce successful weight maintenance for as long as 10 years (Goldfield & Epstein, 2002). More typical reductions in percentage overweight have ranged from 5% to 15% (Goldfield et al., 2002). Decreases in weight (or fat) have been associated with significant reductions in systolic and diastolic blood pressure, fasting serum cholesterol, triglycerides, and hyperinsulinemia, and significant increases in high-density lipoprotein serum cholesterol (Epstein et al., 1998).

Effects of Dieting and Weight Loss on Eating Behavior

A family-based behavior modification program for severe pediatric overweight (Levine, Ringham, Kalarchian, Wisniewski, & Marcus, 2001) produced an average decrease in overweight of 11% during treatment, but this reduction was not maintained approximately 8 months later. Symptoms of eating disorders were measured by the Children's Eating Attitudes Test (ChEAT), which is designed to assess attitudes toward eating and dieting behavior, perceived body image, obsessions and preoccupations with food, and dieting practices. At follow-up, preoccupation with dieting, unhealthy dieting behaviors, and concerns about being overweight decreased. Thus, neither significant weight loss nor weight regain caused an increase in eating disorder symptoms.

Epstein and colleagues evaluated an intervention in which all participants followed the standard Stoplight Diet for Children intervention (described above), and some also received problem-solving skills training (Epstein, Paluch, Saelens, Ernst, & Wilfley, 2001). Follow-up assessments, conducted 18 months after completion of treatment, indicated that percentage overweight decreased an average of 13% across conditions. Weight dissatisfaction, purging and restricting, and total symptoms of disordered eating, assessed by the Kids' Eating Disorder Survey (KEDS; Childress, Jarrell, & Brewerton, 1993), showed no significant changes over time.

A third study evaluated a cognitive-behavior modification (CBM) program that taught self-regulation and problem-solving skills and promoted lifestyle change (Braet & Van Winckel, 2000). Cognitive-behavior modification was delivered in a group, individual, or summer-camp format and was compared to a one-session advice condition. At the 4.6-year follow-up assessment, percentage overweight (which did not differ between groups) had decreased an average of 11% from baseline. None of the participants had anorexia nervosa at follow-up. Assessment with the Dutch Eating Behavior Questionnaire (DEBQ; Van Strien, Frijters, Bergers, & Defares, 1986) indicated that external eating decreased and restrained eating increased between baseline and follow-up, which indicated that the program helped children control external food stimuli and develop the restraint necessary for weight control. Emotional eating, also measured with the DEBQ, did not change. Five participants (9%) scored 1 standard deviation above the reference group mean for the bulimia subscale of the Eating Disorders Inventory, which indicates a greater risk for developing an eating disorder. However, the proportion of higher-risk participants did not appear elevated when compared with community samples of adolescents.

A 10-year follow-up study, the longest to date, evaluated outcomes for participants enrolled in one of four weight control programs during childhood (Epstein, Myers, Raynor, & Saelens, 1998). All of the interventions were family based and used the Stoplight Diet for Children. At follow-up, percentage overweight decreased an average of 10%–20% for participants in most treatment groups, whereas participants in control groups increased their percentage overweight by as much as 12%. Four percent of participants reported that they had been treated for bulimia nervosa over the course of the 10-year follow-up, and none reported treatment for anorexia nervosa. These rates are consistent with rates from community samples.

These four studies suggest that professionally administered weight loss interventions pose minimal risks of precipitating eating disorders in overweight children and adolescents. Cross-sectional studies examining the relationship between dieting and binge eating in clinical populations support this view, as past participation in diet programs does not appear to increase the occurrence of binge eating (Berkowitz, Stunkard, & Stallings, 1993). Similarly, about half of adults with binge eating disorder report that dieting did not precede the onset of their disorder (Yanovski, 2002).

Effects of Dieting on Psychological Status

Dieting and weight loss have also been identified as precipitants to adverse emotional reactions, including depression, anxiety, and irritability (Stunkard, 1957; Stunkard & Rush, 1974). Myers, Raynor, and Epstein (1998) evaluated children's psychological status, as determined by mothers' reports on the Child Behavior Checklist (CBCL; Achenbach, 1991), while they participated in a family-based behavioral program. From baseline to 1-year follow-up, participants' percentage overweight decreased an average of 20%, and during this time, global child psychopathology decreased significantly, while global competence increased. The proportion of children who met clinical criteria for at least one behavior problem decreased from 29% at baseline to 13% at follow-up. Improvements in some aspects of psychological status, including somatic complaints and social competence, were positively associated with weight loss.

Levine and colleagues (2001) found significant reductions in symptoms of depression and anxiety at the end of treatment that were maintained at 8-month follow-up. Epstein and colleagues (2001) observed that total behavior problems and internalizing behavior problems (as measured by the CBCL) decreased significantly at 18-month follow-up. Twelve percent of participants reported seeking treatment for depression during the decade of the Epstein, Valoski, Wing, & McCurley (1994) follow-up, a rate that does not appear high for children who have sought professional weight reduction services (Goldsmith et al., 1992). These findings, as a whole, do not indicate that dieting has a negative effect on mood (e.g., Polivy & Herman, 1985).

Evidence from Research on Obese Adults

Professionally administered weight control programs for overweight youth do not appear to precipitate disordered eating, a finding supported by research on the effects of dieting on binge eating behaviors in obese adults (National Task Force on the Prevention and Treatment of Obesity, 2000; Wilson, 2002). Behavioral programs that prescribed moderate caloric restriction were associated with significant decreases in binge eating episodes in individuals with preexisting binge eating pathology. Studies that used very low–calorie diets (VLCDs) also generally reported improvements in binge eating, although one investigation reported an increase in this behavior (Telch & Agras, 1993). In reviewing the literature, the National Task Force on the Prevention and Treatment of Obesity (2000) concluded that dieting and weight loss, in overweight or obese adults, were not associated with the development of eating disorders, and that weight loss was associated with improvements in depression, anxiety, and related complications. The inclusion, in most programs, of behavior therapy to promote weight loss may have contributed to the observed improvements in mood (Wadden, Stunkard, & Liebschutz, 1988; Wadden, Stunkard, & Smoller, 1986). Long-term studies also found that weight regain, while up setting to dieters, was not associated with significant increases in depression, anxiety, or binge eating (Foster, Wadden, Kendall, Stunkard and Vogt, 1996; National Task Force, 2000; Wilson, 2002).


Professionally administered weight loss programs for overweight children and adolescents generally do not increase symptoms of eating disorders and are associated with significant improvements in psychological status. Thus, concerns about the possible adverse effects of dieting should not deter our nation's growing number of overweight youth from pursuing sensible methods to lose weight or, at least, to prevent the progression of adiposity. A critical issue to address in future research is how to craft public health messages that are health promoting. Clearly approaches aimed at moderation (neither too much nor too little) could effectively target overweight children and youth who tend toward caloric restriction associated with eating disorders. Given the electricity and lack of clarity associated with the term dieting, other terminology should be sought that focuses on healthy portion sizes, moderation of intake, and healthy levels of physical activity.


The prevention of eating disorders remains an important but elusive goal. The literature on eating disorders prevention is relatively limited compared to the voluminous work on the prevention of other problems seen in adolescence. For instance, Durlak and Wells (1997) used meta-analysis to examine 177 primary prevention programs designed to prevent behavioral and social problems in young people under the age of 18. Tobler et al. (2000) found 207 universal prevention programs designed to reduce substance abuse. Nevertheless, the existing data can be useful in evaluating the current state of research on the prevention of eating disorders.

This section reviews empirical studies of prevention, with a focus on universal and targeted prevention activities. In universal prevention, attempts are made to reduce the incidence of a disease by eliminating or reducing risk factors in a population. Increasing exercise levels and reducing intake of high levels of saturated fat to reduce the prevalence of obesity would be categorized as a universal prevention intervention. Ideally, the reduction of risk factors would decrease the incidence of eating disorders such that the benefits of the change outweigh any attendant risks for the population as a whole. Targeted preventive interventions focus on reducing risk factors in individuals who are at high risk of developing subthreshold or threshold eating disorder syndromes. A more in-depth definition of universal and targeted prevention programs is given in the Introduction to this book.

For both universal and targeted interventions, risk factors must be identified and subsequently tested to determine whether a reduction in the risk factor decreases the incidence of the disorder. No risk factor for eating disorders has yet passed this test. Although some investigators have argued that preventive activities should also focus on “protective” factors, such as building higher levels of self-esteem to reduce the risk of developing an eating disorder, no such protective factors have been identified in prospective risk factor studies. Targeted preventive interventions must identify high-risk individuals accurately, which could involve the use of highly sensitive and specific screening tools to partition a population into no-risk, high-risk, or case (diagnosed with the disorder) groups. High-risk individuals could receive targeted preventive interventions, or monitoring, whereas cases could be referred for treatment. But there is currently no instrument that can satisfactorily partition individuals into these groups.

Universal Prevention

Most studies evaluating universal prevention efforts target older students and use curricula designed to change the knowledge, beliefs, attitudes, intentions, and behaviors of individual students. Most programs promote healthy weight regulation, discourage calorie-restrictive dieting, and address ways in which body image and eating are influenced by developmental, social, and cultural factors, or a healthy weight regulation (HWR) model. Other studies focus on broader issues such as increasing self-esteem, empowerment, confidence, and general skills, or a self-esteem/social competence (SESC) model. Many studies focus on reducing the onset of eating disorder symptoms, particularly bulimic pathology. Little is known about prevention of the onset of anorexia nervosa, but given the low prevalence of this disorder, it would be difficult for prevention studies to recruit sufficient sample sizes to detect significant effects.

Elementary School

Studies of universal prevention efforts in elementary school have produced some encouraging results in demonstrating increases in knowledge about eating disorders, but have been less successful in altering attitudes and behaviors. Smolak and Levine (2001) evaluated a 10-lesson HWR model curriculum for girls and boys ages 9 through 11 which also emphasized tolerance and appreciation for diversity in weight and shape. Students in the original sample were reassessed 2 years later. Compared to young adolescents from schools not included in the original study, participants were more knowledgeable, had higher body esteem, and used fewer unhealthy weight management techniques. Scores for the original control group were intermediate, suggesting “cross-contamination” or “spillover” between the original groups. Cross-talk among control and treatment groups at the same school creates a major confound for controlled school-based studies.

Middle School

At least 22 studies have evaluated universal prevention interventions with middle-school children. Killen et al. (1993) randomized 967 sixth-and seventh-grade girls to an 18-lesson program grounded in the HWR model or to standard curriculum control. The intervention produced only modest increases in knowledge and no short-or long-term changes in attitudes or behaviors. The authors also examined changes in the students at “risk,” on the basis of scores of a measure designed to predict the onset of eating disorders (Weight Concerns). At 2-year follow-up, the effect sizes of Weight Concerns for at-risk students in the preventive intervention and control classes were moderate. Thus, the intervention may have been effective for high-risk students.

McVey and Davis (2002) implemented a curriculum of six 1-hour lessons combining features of the HWR and SESC models for 11-to 12-year-old girls beginning the transition into adolescence. There were no significant differences in body satisfaction and eating attitudes between schools that received the intervention and schools that did not. In a controlled evaluation of another HWR intervention, Stewart, Carter, Drinkwater, Hainsworth, and Fairburn (2001) found significant decreases between the pre-and postintervention assessments, including shape concerns, Eating Disorders Examination Questionnaire (EDE-Q), and Eating Attitudes Test (EAT) scores, but scores on these variables reverted to baseline at 6-month follow-up.

A controlled evaluation of an intensive school-based obesity prevention program for youth ages 11 through 13 (Austin, Field, & Gortmaker, 2002) found more positive effects from curricula following the HWR model. Among the 188 girls who were not dieting or eating disordered at baseline, only 1 (0.5%) program participant reported purging or using diet pills 2 years later, as compared to 9 (5.5%) in the control condition. Promising results have also been reported in a series of studies using elements of the SESC model. For example, Steiner-Adair et al. (2002) developed a curriculum designed to help girls become more assertive and supportive of one another as they learned to critically evaluate cultural messages and recognize prejudices pertaining to gender, beauty, weight, and eating. This curriculum, called Full of Ourselves, consists of 70 activities, organized into eight units delivered across 2–4 months. Students learn assertion skills and learn how to be more supportive of one another. The curriculum also discusses issues related to prejudice about weight and teaches students to critically evaluate various cultural messages pertaining to gender, beauty, weight, and eating. Students are encouraged to take active leadership roles in social-justice issues concerning body image (Steiner-Adair, 1994). The girls are given the opportunity to work closely with trained adult mentors and to serve as mentors themselves for girls ages 9 to 11.

Among 500 seventh-grade girls in 24 schools, significant pre-to postintervention effects were found on measures of eating disorder knowledge and weight-related body esteem, which were maintained at 6-month follow-up. There were no apparent effects on weight management behavior.

High School

Universal prevention has been evaluated in at least 23 studies with high school students (ages 14 through 18.). Two studies (one in Israel and one in Italy) demonstrated sustained benefits for low-risk students receiving an HWR curriculum (Neumark-Sztainer, Butler, & Palti, 1995; Santonastaso et al., 1999). Four controlled studies of the HWR model with variations of the SESC model have produced positive pre-to postprogram changes, but the positive effects were limited to drive for thinness (Wiseman et al., 2002), attitudes about sociocultural factors (Kelton-Locke, 2001; Moriarty, Shore, & Maxim, 1990; Phelps, Sapia, Nathanson, & Nelson, 2000), or intentions to diet (Phelps et al., 2000). Two other well-designed studies of prevention programs with substantial elements of the HWR and SESC models (Buddeberg-Fischer, Klaghofer, Reed, & Buddeberg, 2000; Paxton, 1993) and a number of other small-scale or uncontrolled studies have found negative results.

Environmental Interventions

Eating-disordered attitudes and behaviors are difficult to alter because they are strongly reinforced by a variety of family, peer, medical, and other cultural factors. Consequently, some prevention researchers have argued for the need to change the environment of children and adolescents, specifically, the school environment (Neumark-Sztainer, 1996; Piran, 1999; see also Levine & Piran, 2001). Piran (1999), an advocate of this approach, has demonstrated that systemwide changes can reduce eating disorders in the high-risk setting of an elite ballet school. Neumark-Sztainer, Sherwood, Coller, and Hannan (2000) designed a community-based intervention to prevent disordered eating among preadolescent girls, and randomized 226 Girl Scout troop members into control and intervention groups. The intervention consisted of six 90-minute sessions focusing on media literacy and advocacy skills, with some training for troop leaders. At 3-month follow-up, the program demonstrated a positive influence on media-related attitudes and behaviors including internalization of sociocultural ideals, self-efficacy to impact weight-related social norms, and print media habits. Unfortunately, manipulation of system or setting variables to prevent the development of eating disorders has not been well tested in other settings or by other researchers.

Targeted Prevention

A number of studies, usually focused on older adolescents or college students, have shown that interventions targeted at high-risk students can be effective. Because most of these studies included self-selected and older samples, extrapolation of their success to adolescents should be made cautiously. Stice, Trost, and Chase (2003) randomly assigned high school and college students to a dissonance treatment, a healthy weight management condition, or a waiting-list control. With this approach, the participants were asked to help create a program to teach younger girls body acceptance and to avoid internalizing the thin-body ideal. The theory is that participating students will change their own attitudes and beliefs to better conform to the messages they are developing for the younger girls. At 6-month follow-up, the dissonance group had a sustained reduction in internalization of the thin ideal, but the effects for the other measures dissipated after the program (body dissatisfaction) or at 6-month follow-up (negative affect and bulimic behavior). If anything, the healthy weight intervention resulted in longer-term improvements in negative affect and bulimic symptoms.

Results from other studies reporting the effects of both brief and more intense psycho educational programs have also been mixed. Mann et al. (1997) evaluated the effects of recovered classmates describing their experience with having an eating disorder and providing information about eating disorders to fellow students. At the posttest, intervention participants had slightly more symptoms of eating disorders than did controls. Franko (1998) found little benefit from a more intensive eight-session psychosocial support group, whereas Stice and Ragan (2002) and Springer, Winzelberg, Perkins, and Taylor (1999) found some positive effects of college courses on body image and disordered eating.

In a series of studies using a computer-assisted HWR program, Taylor and colleagues found improvements in body image and eating behaviors of self-selected high-risk college students (Taylor, Winzelberg, & Celio, 2001). Studies of students with borderline symptoms of clinical disorders have also been promising. For instance, Kaminski and McNamara (1996) randomized 315 at-risk female college students to a no-treatment control or a cognitive-behavioral group, and at 1-month follow-up the intervention group demonstrated significant improvements in weight management behavior, body satisfaction, and self-esteem, and less fear of negative evaluation. These and more clinically focused studies show that intensive, targeted interventions can reduce risk factors, at least in the short term.

Combining Targeted and Universal Prevention

Luce et al. (submitted) demonstrated that students in a population can be screened for eating disorder risk and participate in interventions appropriate to their needs and interests. On the basis of answers to an on-line risk-factor screen, and self-reported height and weight, students in this study were offered various on-line options, including a general nutrition and healthy weight regulation program, and an intensive psychoeducational program focused on body image enhancement and/or weight maintenance. Students completed one of the programs, provided 1 hour/week for 4 weeks, and participated in a monitored discussion group germane to their group. Of the 11 students who reported vomiting and/or laxative abuse preintervention, 10 reported a decrease at postintervention, and the 11th entered therapy.

High-Risk Populations and Settings

Aside from a few studies with ballet dancers (Piran, 1999, Yannakoulia, Sitara, & Matalas, 2002), preventive interventions for particular at-risk populations or high-risk settings have received little attention. Olmsted, Daneman, Rydall, Lawson, and Rodin (2002) randomly assigned adolescent girls with insulin-dependent diabetes to a psychoeducational program. At 6-month follow-up, significant reductions in body dissatisfaction, drive for thinness, dietary restraint, and eating concerns were observed. Students participating in certain types of athletic activities can also be considered high risk, but there are no studies of adolescents with these characteristics.


The prevention of disordered eating is an important issue in public health. Many young girls and women, as well as boys, suffer from severe and potentially chronic problems with body image, eating, and various forms of unhealthy weight management. In addition, efforts to combine prevention efforts for eating disorders and obesity are important, as some of the fundamental factors that influence disordered eating may also contribute to obesity. The data on the prevention of both eating disorders and cigarette smoking and other drug use (Tobler et al., 2000) also suggest that curricula in the schools alone are not sufficient to produce sustained preventive effects.

Universal prevention efforts with elementary school children have produced positive changes in the relevant knowledge and attitudes of students. Programs that focus on changing factors with broad application, such as increasing self-esteem, creating a stronger sense of connection to peers and mentors, and transforming critical awareness into cultural change, have proved promising with middle school students, but it is unclear whether interventions using the HWR model are of benefit to children in middle school. High school students are a very difficult target audience for universal prevention, as evidenced by nine studies in five different countries producing no significant effects on attitudes and behaviors. Thus it remains unclear whether universal prevention interventions are effective for preventing eating disorders.

A recent meta-analysis of universal prevention studies concluded that the evidence does not “allow any firm conclusions to be made about the impact of prevention programs for eating disorders in children and adolescents” (Pratt and Woolfenden, 2002), and with similar data, others have concluded that universal prevention is ineffective and should be abandoned. However, a dearth of universal prevention studies is different from a proven lack of effectiveness, particularly when well-designed studies of multidimensional interventions are rare. Many questions remain about universal prevention programs, such as how these programs can have stronger and more long-lasting effects on risk factors, what the ideal age is for such interventions, what the advantages and disadvantages are of combined interventions, how to include environmental and family factors, and whether programs should be provided to both boys and girls in the same setting.

The general lack of effectiveness of programs aimed at preventing eating disorders in adolescents also needs to be put in the context of substantial research done on prevention of substance abuse, high-risk sexual behavior, and juvenile delinquency and violence (Nation et al., 2003). In an extensive review, the authors concluded that effective prevention programs need to be comprehensive, include varied teaching methods, provide sufficient dosage, be theory driven, provide opportunities for positive relationships, be developmentally appropriate and socioculturally relevant, include outcome evaluation, and involve well-trained staff. Few of the eating disorder prevention studies meet all these characteristics.

A number of studies have demonstrated that interventions targeting high-risk students can be effective, but because many of these studies focused on self-selected and older samples, caution is needed in generalizing the findings to adolescents. Analyses of universal prevention studies suggest that the HWR model might work for high-risk students, but surprisingly little research has focused on students in high-risk settings. Although targeted interventions have proven effective, their effects are generally short-lived and specific to a few dimensions. The challenges in delivering targeted interventions, particularly to populations, are substantial. For example, in school settings, the identification and motivation for high-risk individuals to participate in interventions may be difficult.