The prevalence of anorexia nervosa and bulimia nervosa in adolescent samples clearly is lower than rates reported for adult samples. The lack of studies with adequate sample sizes and the considerable variation in methodology make it difficult to answer confidently the question of how many adolescents experience an eating disorder. A major gap in studies from the United States is the lack of data adequately representing the ethnic diversity of its population.

Experts have expressed concern that the data on prevalence of eating disorders in adolescents are misleading because the strict diagnostic criteria do not permit diagnosis of anorexia nervosa or bulimia nervosa among adolescents who show evidence of the core features of these disorders yet have not yet developed the requisite severity or duration of symptoms (Golden et al., 2003). Examination of the prevalence of behavioral eating disorder symptoms, is therefore indicated. These symptoms may represent the first signs of development of a full-syndrome disorder, and data on their prevalence thus give an indication of the size of the “at-risk” group.

In 1990, the Centers for Disease Control and Prevention developed the Youth Risk Behavior Surveillance System to monitor health risk behaviors that contribute markedly to the leading causes of death, disability, and social problems among youth and adults in the United States. The system includes national, state, and local school-based Youth Risk Behavior Surveys (YRBS; 2003) conducted every 2 years, of representative samples of 9th through 12th grade students. Students complete the anonymous surveys during a class period at school. In the 2001 YRBS (Grunbaum et al., 2002), the overall participation rate was 63% but varied considerably across states. The YRBS includes questions about current attempts to lose weight and questions about weight loss or weight maintenance efforts, such as vomiting, diet pills, and “other methods.”

Consistently, the YRBS has found that about two of every three female students report trying to lose weight, compared to one in four male students. Trying to gain weight is quite common among boys (about 40%), whereas only a minority of girls (about 8%) engage in efforts to become heavier (Grunbaum et al., 2002; Lowry et al., 2002; Middleman, Vasquez, & Durant, 1998). The YRBS also includes questions about behav

iors that are considered in the DSM to be “inappropriate compensatory behaviors.” Data collected in 1999 and 2001 suggest that inappropriate efforts to lose weight or keep from gaining weight are disturbingly common, especially among girls. Specifically, in 1999, fasting, use of diet pills, and vomiting or laxative abuse were reported by 18.8%, 10.9%, and 7.5% of girls, respectively, compared to 6.4%, 4.4%, and 2.2% of boys (Lowry et al., 2002). Data for 2001 showed similar rates for these behaviors (Grunbaum et al., 2002).

The YRBS includes an ethnically diverse sample, permitting the examination of ethnic group differences in the prevalence of weight-related behaviors. In 2001, white and Hispanic girls were found to be significantly more likely than black girls to report inappropriate compensatory behaviors. Almost one in four (23.1%) female Hispanic students and one in five white students (19.7%), compared to 15% of black students, reported that she had gone without eating for 24 or more hours to control her weight. Vomiting or laxative use to control weight was reported by 10.8%, 8.2%, and 4.2% of Hispanic, white, and black girls, respectively (Grunbaum et al., 2002). These findings suggest that additional study of disordered eating among minority groups is needed. Whether girls in this sample would meet diagnostic criteria for an eating disorder is unclear, because the YRBS does not include questions covering the complete set of diagnostic criteria for anorexia nervosa and bulimia nervosa. Nevertheless, a considerable subset of female students practice potentially health-damaging behaviors, such as vomiting, which is cause for concern.

Although the YRBS does not assess binge eating, a recent study of boys and girls in public middle and high schools has provided information about binge eating in adolescents (Ackard, Neumark-Sztainer, Story, & Perry, 2003). Binge eating was considered present if the child answered “yes” to a question about overeating with loss of control, at least a few times a week, and feeling upset “some” or “a lot” by overeating. More girls (3.1%) than boys (0.9%) met criteria for binge eating, and the results suggested that binge eating was significantly correlated with body mass index (BMI) in girls and boys. Binge eating was only slightly more common in white girls (2.6%) than in black (1.6%) or Hispanic (1.7%) girls, and was reported by a surprisingly large number of Asian American girls (5.9%). Therefore, binge eating may occur among ethnically diverse groups.

Epidemiological research on eating disorders in adolescence is limited in several important ways. Nationally representative samples, including ethnic minority children, are needed to determine the prevalence of eating disorders among American youth. Studies of ethnic minorities in European countries (e.g., Bhugra & Kamaldeep, 2003) and among adolescents in non-Western countries (e.g., Huon, Mingyi, Oliver, & Xiao, 2002; Nobakht & Dezhkam, 2000) have reported significant rates of eating disorders among minorities. These results indicate that eating disorder research should be more inclusive in its sampling frames.

While the diagnostic criteria for anorexia nervosa and bulimia nervosa are well articulated, definitions of eating pathology that fail to meet diagnostic criteria are quite varied. Since eating pathology that fails to reach diagnostic significance may represent the early stages of a full-syndrome eating disorder for some adolescents, a more uniform definition of partial syndrome is needed. Studies should also include criteria for binge eating disorder to permit estimates of the prevalence of this syndrome.

The lack of a uniform instrument for measuring eating disorder symptoms also limits current epidemiological research. One-step assessment studies that use standardized psychiatric interviews, in which individuals “skip out” of the eating disorder module if a gated question is answered negatively, may have produced an underestimation of eating pathology, especially in young samples, in whom the clinical presentation of anorexia nervosa or bulimia nervosa could be atypical (Kreipe et al., 1995). Finally, parent reports might improve the detection of anorexia nervosa, a disorder in which denial is a hallmark.

Eating disorders may be transient (albeit in

many cases recurrent) and point prevalence rates may therefore not reflect fully the extent of eating pathology in adolescents (Patton, Coffey, & Sawyer, 2003). Even if an adolescent's eating disorder is time limited and nonrecurring, it may represent a marker for psychopathology that conveys important clinical information. For example, Johnson, Harris, Spitzer, & Williams (2000) demonstrated that an adolescent's history of an eating disorder is associated with elevated risk for other Axis I disorders in adulthood. Adolescent eating disorders, even when fully remitted, are associated with a broad range of indicators of impaired psychosocial functioning (Striegel-Moore, Seeley, & Lewinsohn, 2003).

Information about the incidence of eating disorders requires longitudinal data. Although a few studies have included longitudinal follow-ups, these studies have not focused specifically on eating disorders and did not provide incidence data. There is one exception, however. Lewinsohn and colleagues (2000) classified an eating disorder sample into onset prior to or after age 19 and reported that first incidence for anorexia nervosa was significantly more likely to occur prior to age 19. An incidence rate for anorexia nervosa of 0.1% was found for the age group of 19–23 years. For bulimia nervosa, the incidence rates for before age 19 and after were comparable: 1.5% and 1.3%, respectively. Clearly, more detailed data are needed on the incidence of eating disorders.

In conclusion, epidemiological research of eating disorders is quite limited. In light of the considerable public health significance of anorexia nervosa and bulimia nervosa and their spectrum variants, such research is urgently needed.

The lifetime rates of psychiatric comorbidity among patients with anorexia nervosa are approximately 80% (Halmi et al., 1991). Affective disorders, anxiety disorders, substance use disorders, and personality disorders are commonly associated with anorexia nervosa. The affective disorder that most commonly co-occurs with anorexia nervosa is major depressive disorder, with a lifetime comorbidity of 50%–68% (Herzog, Nussbaum, & Marmor, 1996). Lifetime rates of anxiety disorders are between 55% and 65% (Godart, Flament, Perdereau, & Jeammet, 2002; Halmi et al., 1991); the most common comorbid anxiety diagnoses are social phobia (55%) and obsessive-compulsive disorder (OCD) (25%–69%; Godart et al., 2002; Halmi et al., 1991). Lifetime prevalence of substance use disorders ranges between 12% and 21% (Herzog, Keller, Sacks, Yeh, & Lavori, 1992; Stock, Goldberg, Corbett, & Katzman, 2002), compared to 11% of women in the general population (Bulik, Sullivan, McKee, Weltzin, & Kaye, 1994). Patients with AN-B/P are more likely than those with AN-R to manifest substance use disorders (Herzog, Keller, Sacks, et al., 1992; Stock et al., 2002).

When patients with anorexia nervosa present for treatment, over 70% report an additional current psychiatric disorder. Approximately 66% have a co-occurring affective disorder, 49% are diagnosed with a personality disorder, 5% have a substance use disorder, and 56% report an anxiety disorder (Braun, Sunday, & Halmi, 1994; Herpertz-Dahlmann et al., 2001; Herzog, Keller, Sacks, et al., 1992; Wonderlich & Mitchell, 1997). Cluster C personality disorders, which include avoidant, dependent, obsessive-compulsive, and passive-aggressive personality disorder, are also common among patients with

anorexia nervosa (Herzog, Keller, Lavori, Kenny, & Sachs, 1992).

The developmental sequence of anorexia nervosa in relation to other comorbid conditions can vary significantly. Affective disorders may begin before or after the onset of anorexia nervosa, or the disorders can begin concurrently (Braun et al., 1994). Anxiety disorders, in particular social phobia and OCD, frequently predate the onset of anorexia nervosa (Anderluh, Tchanturia, Rabe-Hesketh, & Treasure, 2003; Braun et al., 1994; Bulik, Sullivan, & Joyce, 1997), whereas substance use disorders often develop after anorexia nervosa (Braun et al., 1994).

Nearly 83% of patients with bulimia nervosa report a lifetime history of an additional psychiatric disorder (Fichter & Quadfleig, 1997); affective disorders, anxiety disorders, substance use disorders, and personality disorders are commonly associated with bulimia nervosa. While there is a significant amount of variability in the rates of comorbidity, more than 50% of patients with bulimia nervosa have a lifetime history of a mood disorder. Major depressive disorder has been shown to be the most common mood disorder diagnosis among patients with bulimia nervosa. In community samples, approximately one third are depressed, a rate that increases to 65% in inpatient and outpatient samples. In clinical samples, the lifetime rates of comorbidity with at least one anxiety disorder ranges from 13% to 65% (Herzog, Keller, Sacks, et al., 1992). Social phobia (17%) and OCD (8%–33%) are the most frequently diagnosed anxiety disorders in bulimia nervosa, and panic disorder is also commonly observed (Brewerton et al., 1995; von Ranson, Kaye, Weltzin, Rao, & Matsunaga, 1999). The lifetime prevalence of substance use disorders is approximately 25% (Bulik et al., 1994), and patients with bulimia nervosa most frequently abuse alcohol, cocaine, and marijuana. Patients with bulimia nervosa and substance use disorders commonly exhibit impulsivity in multiple domains, including suicide attempts, self-injurious acts, and stealing.

When patients with bulimia nervosa present for treatment, approximately 75% meet criteria for an additional psychiatric disorder (Fichter & Quadfleig, 1997). Approximately 50% have a co-occurring affective disorder, 34% have a substance use disorder, and 56% have an anxiety disorder (Halmi et al., 2002; Herzog, Keller, Sacks, et al., 1992; Mitchell, Specker, & de Zwaan, 1991; Wonderlich & Mitchell, 1997). Cluster B personality disorders, such as antisocial, borderline, histrionic, and narcissistic personality disorder, are also common in bulimia nervosa (Herzog, Keller, Lavori, et al., 1992).

As in anorexia nervosa, the sequence development of bulimia nervosa and comorbid conditions varies, as onset of the comorbid disorder can occur prior to, at the same time as, or following the development of bulimia nervosa (Braun et al., 1994). As with anorexia nervosa, anxiety disorders commonly predate the onset of bulimia nervosa, whereas substance use disorders more often develop after the onset of bulimia nervosa (Braun et al., 1994; Bulik et al., 1997).

The available data suggest that approximately 50%–70% of adolescents with anorexia nervosa recover, 20% are improved but continue to have residual symptoms, and 10%–20% have chronic anorexia nervosa (Herpertz-Dahlmann et al., 2001; Morgan, Purgold, & Welbourne, 1983; Steinhausen, 2002; Steinhausen et al., 1997). Adolescents with anorexia nervosa continue to recover over time; for example, Strober, Freeman, and Morrell (1997) reported a 1% probability of adolescents reaching full recovery at 3 years, which increased to 72% after 10 years. Those anorexia nervosa patients experiencing persistent symptoms typically display abnormalities in weight, eating behaviors, menstrual function, comorbid psychopathology, and difficulties with psychosocial functioning (Herpertz-Dahlmann et al., 2001; Steinhausen, 1997; Strober et al., 1997; Wentz, Gillberg, Gillberg, & Rastam, 2001). Relapse is common after weight gain in hospitalized patients, with up to one third of adolescent anorexia nervosa patients relapsing soon after discharge (Herzog et al., 1999; Strober et al., 1997).

Anorexia nervosa has one of the highest mortality rates among psychiatric disorders. Approximately 5.6% of patients diagnosed with anorexia nervosa die per decade of illness (Sullivan, 1995), and anorexia nervosa patients are 12 times more likely to die than women of a similar age in the general population (Keel et al., 2003). Although the combined mortality rate for anorexia nervosa among adolescents and adults is over 5% (Steinhausen, 2002), the mortality rate during adolescence is low. The most common causes of death among patients with anorexia nervosa are suicide and the effects of starvation. The suicide rate among women with anorexia nervosa is 57 times higher than that for women of a similar age in the general population (Keel et al., 2003). Some studies have found lower weight at presentation, longer duration of illness, and severe alcohol use to be associated with higher risk of mortality (Keel, Mitchell, Miller, Davis, & Crow, 1999; Patton, 1988).

Few variables are consistently associated with outcome in anorexia nervosa, but the most positive outcomes are seen in patients between the ages of 12 and 18 with a short duration of illness. Poor outcome is associated with extremely low weight at presentation and, in some studies, by vomiting. The relationship of binge eating to outcome of anorexia nervosa is not clear, as patients with AN-R and those with AN-B/P have a similar time to recovery (Herzog et al., 1999).

Most adolescents and adults with bulimia nervosa improve over time, with recovery rates ranging from 35% to 75% at 5 or more years of follow-up (Fairburn et al., 2000; Fichter & Quadfleig, 1997; Herzog et al., 1999). Bulimia nervosa is a chronic relapsing condition, and approximately one third of individuals with bulimia nervosa relapse (Keel & Mitchell, 1997), often within 1 to 2 years of recovery (Herzog et al., 1996). Although approximately 50% of patients with bulimia nervosa recover, the remaining individuals continue to be symptomatic, often with substantial impact on physical and psychosocial functioning. Mortality is a rare outcome in bulimia nervosa, with rates as low as 0.5% (Keel et al., 1999). Few prognostic factors have been consistently reported across studies of bulimia nervosa, but low self-esteem, longer duration of illness prior to presentation, higher frequency or severity of binge eating, substance abuse history, and a history of obesity have been associated with poor outcome (Bulik, Sullivan, Joyce, Carter, & McIntosh, 1998; Fairburn, Stice, et al., 2003; Keel et al., 1999).

Few studies address diagnostic migration, or the movement from one eating disorder subtype, or eating disorder, to another, within the adolescent eating disorder population. While some patients migrate from bulimia nervosa to anorexia nervosa (Kassett, Gwirtsman, Kaye, Brandt, & Jimerson, 1988), the most frequent change among diagnostic categories is from the subtype AN-R to AN-B/P, reflecting the development of bulimic symptomatology. Some individuals gain weight in association with the binge eating, leading to a change in diagnostic status from the subtype AN-R or AN-B/P to bulimia nervosa. In one study, more than 50% of AN-R patients, both adolescents and adults, developed bulimic symptomatology (Eddy et al., 2002), and only a small fraction of patients with AN-R remained in that diagnostic subtype. The remaining patients with AN-R who did not develop binge eating or purging were partially or fully recovered. It is unknown what factors lead to the development of bulimic symptoms among patients with AN-R, and what the precise time course of this development is.

The occurrence of other psychiatric disorders is extremely common in association with both anorexia nervosa and bulimia nervosa, and this complicates treatment. Unfortunately, many treatment studies of eating disorders exclude patients with serious comorbid disorders, such as substance use disorders. Adolescents with anorexia nervosa have a better prognosis when they receive treatment early in the course of their